article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the star experiment at the relativistic heavy ion collider ( rhic ) at the brookhaven national laboratory has a unique capability of precise measurement of charged and neutral hadrons and photon multiplicity at forward rapidity . through this capability
we can carry out a systematic study of various aspects of particle production in relativistic heavy ion collisions .
the study of nuclear modification factor ( @xmath3 ) as a function of transverse momentum ( @xmath4 ) and pseudorapidity ( @xmath5 ) can reveal interesting information on the mechanism of particle production . a @xmath3(@xmath4 ) of unity
will indicate the absence of nuclear effects such as shadowing , multiple scattering ( cronin effect ) and gluon saturation . in a color glass condensate picture ( cgc ) of particle production
it is expected that the nuclear modification factor will decrease with the increase in rapidity @xcite .
measurement of baryons at forward rapidity will help in understanding the baryon transport and quantifying the nuclear stopping in high energy collisions .
forward rapidity is also the region where we observe the particle production per participating nucleon pair as a function of @xmath5 - y@xmath6 , where y@xmath6 is the beam rapidity , to be independent of beam energy .
this phenomenon is known as limiting fragmentation ( lf ) @xcite .
a detailed study of energy , centrality and species dependence of lf behaviour will provided useful insight to particle production mechanism in forward rapidity .
comparative study of lf for photons ( primarily from decay of @xmath7 mesons ) and charged particles will help in understanding the centrality dependence lf behaviour .
the detectors at the forward rapidity @xcite in star experiment are forward time projection chamber ( ftpc ) , photon multiplicity detector ( pmd ) and forward @xmath7 detector ( fpd ) .
the ftpcs detect charged particles in the pseudorapidity region 2.5 @xmath8 @xmath9 @xmath8 4.0 .
the pmd consists of two planes of an array of cellular gas proportional counters separated by a lead convertor .
it detects photons within 2.3 @xmath8 @xmath5 @xmath8 3.7 .
the fpd consists of lead - glass cells 21 radiation lengths deep .
it detects high energy @xmath10 mesons within @xmath11 .
= 200 gev .
the results are compared to brahms and phenix results .
the charged hadron measurements are also compared to hijing .
, title="fig : " ] = 200 gev .
the results are compared to brahms and phenix results .
the charged hadron measurements are also compared to hijing .
, title="fig : " ] -11.0 mm in fig . [ nmf ] ( left ) we show the nuclear modification factor ( @xmath12 ) of charged hadrons for two centrality classes at @xmath13 @xmath14 3.1 and up to @xmath4 = 3 gev/@xmath15 .
@xmath16 is larger on the gold side than on the deuteron side . on the deuteron side
we observe a stronger centrality dependence of @xmath16 compared to the gold side .
our measurement compares well with the results from brahms and phenix experiments @xcite .
the lines are calculations from hijing with ( dashed ) and without ( solid ) shadowing .
the nuclear effects on particle production have been also quantified by studying the @xmath17 ( fig .
[ nmf ] ( right ) ) , the ratio of the inclusive yield of @xmath7 in d+au to p+p collisions normalized by the number of nucleon - nucleon collisions .
the ratio @xmath18 at @xmath19 is significantly smaller than @xmath20 at smaller @xmath5 @xcite , consistent with the trend expected from cgc - based models @xcite .
yields in central and peripheral collisions compared to hijing .
right : net baryon density for three centrality classes on the deuteron side of the collision .
, title="fig : " ] yields in central and peripheral collisions compared to hijing .
right : net baryon density for three centrality classes on the deuteron side of the collision .
, title="fig : " ] -13.0 mm in the ftpc s , @xmath1 and @xmath2 are reconstructed by obtaining the invariant mass of their dominant decay modes @xmath21 and @xmath22 ( branching ratio of 64% ) . in fig .
[ lambda ] ( left ) we show the ratio of the @xmath1 rapidity density in central collisions to that in peripheral collisions as a function of rapidity .
the data show significant deviations from expectations from wounded nucleon scaling ( arrows ) .
the results when compared to hijing calculations show a good agreement on the deuteron side of the collisions . on the gold side
the data is above hijing indicating possible presence of substantial nuclear effects . in fig .
[ lambda ] ( right ) we show the measured net baryon density for three centrality classes .
the midrapidity results are from measured p and @xmath23 in the time - of - flight detector and at y = 2.75 on the deuteron side from the @xmath1 measurements in ftpc .
lines show fits to the net baryon distributions with global constraints on the total number of baryons .
within such a model , we observe significant stopping in central collisions and a large degree of transparency in peripheral collisions . for charged particles and photons @xcite for various collision systems : ( left ) at different @xmath0 and ( right ) for various event centrality classes at @xmath0 = 62.4 gev .
, title="fig : " ] for charged particles and photons @xcite for various collision systems : ( left ) at different @xmath0 and ( right ) for various event centrality classes at @xmath0 = 62.4 gev .
, title="fig : " ] -14.0 mm
in fig . [ lim ] we present the energy dependence of lf for inclusive photons and charged particles . our data from central collisions has been compared to corresponding measurements from nucleus - nucleus collisions at different @xmath0 in phobos , brahms , wa98 , wa93 experiments and p+p(@xmath23 ) collisions in isr and ua5 experiments @xcite .
the @xmath24 per participant pair as a function of @xmath5 - y@xmath6 is observed to be independent of beam energy . in fig .
[ lim ] we show the centrality dependence of lf for charged particles and photons . at forward rapidity in au+au collisions at @xmath0 = 62.4 gev the charged particle yield normalized to number of participating nucleons ( @xmath25 ) as a function of @xmath5 - y@xmath6 is higher for peripheral collisions compared to central collisions , whereas within the measured @xmath5 range of 2.3 to 3.7 , the photon yield normalized to @xmath25 as a function of @xmath5 - y@xmath6 is independent of centrality @xcite .
we also observe that the photon results in the forward rapidity region from @xmath26 collisions at @xmath0 = 540 gev are in close agreement with the measured photon yield in au+au collisions at @xmath0 = 62.4 gev .
however , the @xmath27 and @xmath26 inclusive charged particle results are very different from those for au+au collisions at @xmath0 = 62.4 gev .
the photon results and comparison with p+p(@xmath23 ) collisions indicate that , in the @xmath5 region studied , there is apparently a significant charged baryon contribution in au+au collisions and such a contribution may be responsible for the observed centrality dependent lf behaviour in charged particles .
we have presented the results from the star experiment on charged hadrons , neutral hadrons and photons at forward rapidity in d+au collisions at @xmath0 = 200 gev and au+au collisions at @xmath0 = 62.4 gev .
the nuclear modification factor in d+au collisions is observed to decrease with increasing rapidity , which is consistent with predictions from cgc based models .
@xmath16 is larger on the gold side than on the deuteron side in d+au collisions . from the net baryon density measurements we conclude that there is significant stopping in central d+au collisions and large transparency in peripheral collisions .
our measurements of photons and charged particle multiplicity show that photons follow centrality independent limiting fragmentation behaviour whereas charged particles follow a centrality dependent behaviour .
k. h. ackermann , _ et al .
_ , nucl .
meth . a 499 , 713 ( 2003 ) ; m.m .
et al . _ ,
meth . a 499 , 751 ( 2003 ) ; c. allgower _ et al .
_ , nucl .
meth . a 499 , 740 ( 2003 ) .
i. arsene , _ et al . _
( brahms collaboration ) , phys .
93 , 242303 ( 2004 ) ; s. s. adler , _ et al . _
( phenix collaboration ) , phys .
94 , 082302 ( 2005 ) .
|
we present the recent results from the star experiment on charged and neutral particle measurements at the forward rapidity in d+au collisions at @xmath0 = 200 gev and au+au collisions at @xmath0 = 62.4 gev .
the nuclear modification factor for charged and neutral hadrons in d+au collisions are presented .
measurements of @xmath1 and @xmath2 production at forward rapidity and the variation of net baryon density as a function of collision centrality are discussed .
we have also studied the limiting fragmentation of photons and charged particles in au+au collisions .
the photons and charged particles separately follow the energy independent limiting fragmentation behaviour .
however they have been observed to follow a different centrality dependence of limiting fragmentation .
|
when electrons in a crystal band are driven by an external time - independent electric field , they move periodically across the brillouin zone , creating characteristic bloch oscillations . the frequency of the oscillations , @xmath0 , where @xmath1 is the unit cell size , coincides with the energy separation between neighboring states localized on a wannier - stark ladder @xcite .
the effect has been observed for electrons / holes in semiconducting superlattices@xcite , for atoms trapped in a periodic optical potential@xcite , and for light propagating in a periodic array of waveguides , with gradient of the temperature or of the refraction index working as an effective electric field@xcite .
combining the effects of a strong , time - periodic driving field , with the nonlinearity of the bloch oscillations leads to higher harmonic generation of the driving frequency@xcite .
this effect has recently been observed in bulk zno crystals strongly driven by a few - cycle pulsed infrared laser@xcite .
the application of the infrared field in short , 100-femtosecond , pulses was necessary to ensure that the absorbed energy could be transferred to the lattice and dissipated . in this work , we suggest that frequency multiplication due to periodically - driven bloch oscillations could also be observed in a steady - state setting , e.g. , a periodically modulated nanowire ( or an array of such nanowires ) continuously driven by high - amplitude terahertz radiation ( see fig .
[ fig : nanowire ] ) . in the weak - scattering limit , the quantum efficiency of frequency tripling for such a system can be as high as 93% . for a nanowire in mechanical contact with an insulating , optically transparent substrate
, a quasi - equilibrium electron distribution will be reached as the power absorbed from the driving field will be continuously dissipated into phonon modes .
this distribution can be quite different from the initial , equilibrium fermi distribution .
in particular , at the driving field amplitude which is optimal for third harmonic generation , the distribution can be both broadened and inverted .
the inversion of the distribution occurs once the driving field amplitude exceeds the dynamical localization threshold@xcite . in our analytic derivation , we combine the floquet quasi - energy description with the keldysh green s function technique to obtain the semiclassical master equation for a one - dimensional band of strongly and rapidly driven electrons in the presence of weak scattering by phonons .
we solve these equations numerically to find the electron distribution function for a cosine energy band at a given driving field frequency ( fixed at @xmath2thz ) and the field amplitude chosen to suppress the generation of the principal harmonic .
this electron distribution is used as an input for calculating the time - dependent current and the intensity radiated at different harmonics of the driving field frequency .
we use these results to find the optimal dimensions of a periodically modulated inas / inp nanowire , which would yield the most efficient frequency tripling of @xmath3thz radiation .
we consider a single - band one - dimensional metallic wire driven by a harmonic electric field with the amplitude @xmath4 and frequency @xmath5 , and coupled to substrate phonons , @xmath6 where the electron , electron - phonon , and phonon hamiltonians are , respectively @xmath7 here @xmath8 @xmath9 is the annihilation ( creation ) operator for an electron with one - dimensional momentum @xmath10 and energy @xmath11 . to apply our results to a periodically modulated nanowire , we assume a tight - binding model with the electronic spectrum , @xmath12 where @xmath13 is the hopping matrix element and @xmath1 is the period of the potential along the chain .
the electric field is incorporated into the hamiltonian through the vector potential @xmath14 with @xmath15 representing the vector potential of the driving field .
phonon annihilation ( creation ) operators @xmath16 and @xmath17 are labeled with the three - dimensional wavevector @xmath18 and @xmath19 is the phonon frequency ( electron spin and phonon branch indices are suppressed ) .
the factors @xmath20 are the matrix elements for electron - phonon scattering .
we ignore the effects of disorder or electron - electron interactions , and consider lattice phonons in thermal equilibrium at temperature @xmath21 .
we do not include directly the scattering by phonon modes of the nanowire , assuming that they are strongly hybridized with those of the substrate , with the corresponding effects incorporated in the matrix elements @xmath22 .
the electron - phonon coupling is considered to be weak , meaning that the phonon scattering time is long compared to the period @xmath23 of the driving field .
the dynamics of the strongly - driven electrons with the hamiltonian ( [ eq : ham - electron ] ) is characterized by non - monotonous phases @xmath24 the phase accumulated over a period , @xmath25 , can be expressed in terms of the average particle energy with the momentum @xmath10 , @xmath26 clearly , this energy can be also identified as the floquet energy of a single - electron state . while eq .
( [ eq : average - energy ] ) does not include the usual additive uncertainty @xmath27 , this particular choice has the advantage that in the weak - field limit , @xmath28 , @xmath29 recovers the zero - field spectrum @xmath11 .
the average energy ( [ eq : average - energy ] ) also coincides with that introduced in the theory of dynamical localization@xcite .
dynamical localization occurs when the effective band becomes flat , i.e. , @xmath30 .
the corresponding condition is most easily obtained in the special case of tight - binding model with the spectrum ( [ eq : tight - binding ] ) , @xmath31 where @xmath32 is the zeroth order bessel function . with the driving field amplitude increasing from zero the bandwidth
is gradually reduced ; it switches sign at the roots of the bessel function , @xmath33 .
the first time this happens corresponds to the electric field @xmath34 , where @xmath35 .
we obtain the instantaneous current by averaging the canonical velocity operator @xmath36 over the electron distribution function @xmath37 , @xmath38 where we assumed the tight - binding spectrum ( [ eq : tight - binding ] ) and used the definitions @xmath39 in the limit of weak scattering , the distribution function @xmath40 is time- independent and always symmetric , @xmath41 .
thus , @xmath42 while @xmath43 is a time - independent pre - factor .
the fourier components of the current are obtained directly , @xmath44 where the summation is over the odd harmonics @xmath45 . by choosing @xmath46 , the first harmonic
can be fully suppressed , which leaves the third harmonic dominant .
the maximal value for the fraction of the energy emitted into the third harmonic ( @xmath47 ) is found in close vicinity of this amplitude , see fig .
[ fig : analyticharmonics ] . , for @xmath48 ( red dashed ) , @xmath49 ( black , solid ) , and @xmath50 ( blue , dotted ) plotted as a function of the dimensionless amplitude of the vector potential of the driving field , see eq .
( [ eq : ham - electron ] ) .
the intensities @xmath51 correspond to the power emitted in the corresponding harmonics when multiple nanowires are used in a planar geometry , see fig .
[ fig : nanowire ] ( b),(c).,width=288 ] we use the keldysh non - equilibrium green s function ( gf ) formalism along with a perturbation theory expansion with respect to the entire time - dependent electron hamiltonian ( [ eq : ham - electron ] ) ; the corresponding evolution is solved exactly in terms of the phases ( [ eq : phases ] ) .
previously , related approaches have been used , e.g. , for describing ionization of atoms@xcite and the high - order harmonic generation@xcite in the field of ultrashort laser pulses . here , instead of solving the corresponding equations numerically , we take the limit of weak electron - phonon coupling and
analytically derive the semiclassical master equation for electron distribution function averaged over the period of the driving field , see eqs .
( [ eq : qke ] ) and ( [ eq : rates ] ) .
the same master equation can also be derived from the formalism by konstantinov and perel@xcite with the help of an appropriate resummation of the perturbation series@xcite . in the interaction representation with respect to the time - dependent hamiltonian ( [ eq : ham - electron ] )
, the electron operators acquire time - dependence @xmath52 with quasiperiodic phases ( [ eq : phases ] ) .
we separate these phases by defining the `` lower - case '' gfs @xmath53 where the `` upper - case '' @xmath54 is any of the conventional gfs introduced in the keldysh formalism .
these phases introduce rapid oscillations in the self - energy , making the direct wigner transformation difficult .
we notice , however , that in the limit of weak electron - phonon coupling , the gfs ( [ eq : gf ] ) are expected to change only weakly when both time arguments are incremented by the driving period @xmath55 .
this implies that in the following decomposition , @xmath56 @xmath57 is the `` fast '' time , while @xmath58 is the `` slow '' time when it appears as an argument of thus defined floquet components @xmath59 of the gf . the dyson equations for thus defined keldysh @xmath60 and retarded @xmath61 gfs@xcite have the form @xmath62 where @xmath63 and @xmath64 are the collision integrals originating from the corresponding self - energy functions .
the collision integrals being relatively small , both @xmath60 and @xmath61 are dominated by the @xmath65 components@xcite . to derive the semiclassical master equation , we write the equations for the @xmath65 components of the `` lesser '' @xmath66 and `` greater '' @xmath67 gfs@xcite , perform the wigner transformation replacing the fast time variable @xmath68 by the frequency @xmath69 , and use a version of the kadanoff - baym approximation@xcite @xmath70 for the corresponding spectral function , @xmath71=2\ , \im g^r_{k,0}(\omega , t)$ ] , where @xmath72 is the non - equilibrium electron distribution function averaged over the period .
this requires that the electron - phonon coupling be weak , and assumes that the electron spectrum renormalization has been included in the hamiltonian ( [ eq : ham - electron ] ) .
the resulting master equation for weak electron - phonon interactions has the following standard form@xcite @xmath73 f_k(t ) \\ & & \qquad\;\ ; -\gamma_{k',k}\,f_{k'}(t ) [ 1 - f_{k}(t ) ] \bigr\ } , \label{eq : qke}\end{aligned}\ ] ] where the transition rates are @xmath74.\;\;\strut \label{eq : rates}\end{aligned}\ ] ] here @xmath75 is the phonon spectral function ( density of states weighted by the square of the coupling ) for a given momentum @xmath76 along the wire , see eq .
( [ eq : ham - e - ph ] ) , @xmath77^{-1}$ ] is the phonon distribution function , and the energy increment @xmath78 is the energy carried in or out by phonons , depending on its sign .
note that this energy includes @xmath45 quanta of the driving field , emitted or absorbed , depending on the sign of @xmath79 .
the matrix elements @xmath80 are the fourier expansion coefficients of the product of the two phase factors , @xmath81 , where @xmath82 is the periodic part of the phase .
they satisfy the sum rule @xmath83 clearly , the equilibrium fermi distribution for @xmath40 is only obtained in the limit of small electric field amplitudes , such that @xmath84 with @xmath65 gives the dominant contribution .
the results presented in this section have been obtained by numerically finding the stationary solution of the discretized version of the master equation ( [ eq : qke ] ) with transition rates ( [ eq : rates ] ) .
a simple model for the phonon spectral function , @xmath85 , was used , with the sound speed @xmath86 m/s as appropriate for typical 3d acoustical phonons .
since we assume no other scattering mechanisms , the quasi - equilibrium distribution functions @xmath40 and other results do not depend on the magnitude of the electron - phonon coupling @xmath87 .
we fix the phonon temperature at @xmath88k , the lattice period @xmath89 nm , the average electron filling at @xmath90 and choose the driving field frequency @xmath91 hz ( energy @xmath92mev ) .
also , the amplitude @xmath93 is fixed , which corresponds to the point where the first harmonic generation is fully suppressed [ see fig .
[ fig : analyticharmonics ] ] . at this point
the effective coupling is @xmath94 , which creates an inverted and somewhat narrowed band .
the effective bandwidth is smaller than @xmath95 for @xmath96mev . in fig .
[ fig : i3vj ] , we show the intensity @xmath97 of the radiated third harmonic ( in arbitrary units ) as a function of the tight - binding hopping parameter @xmath13 .
the overall upward trend reflects the linear scaling of the current with @xmath13 .
the plot has a series of pronounced maxima and minima related to the structure of the distribution function @xmath40 , see fig .
[ fig:3distributions ] .
indeed , at the first maximum of the radiated intensity @xmath97 , @xmath98mev , the distribution function has a well - defined minimum at @xmath99 and symmetric maxima at @xmath100 [ fig .
[ fig:3distributions](b ) ] ; notice the population inversion consistent with negative @xmath101 . on the other hand , the distribution in fig .
[ fig:3distributions](c ) corresponding to the first minimum of radiated intensity , @xmath102mev , is much flatter .
this flattening can be traced to a sharp increase of the transition rates connecting the regions of momentum space near @xmath99 and @xmath103 .
this is illustrated in fig .
[ fig : gammarateplot1 ] , where transition rates between @xmath99 and @xmath104 are shown . the corresponding phases @xmath105 have only even harmonics @xmath27 , @xmath106 , and the threshold values of @xmath13 for different @xmath45 correspond to sharp maxima of @xmath107 . , computed with @xmath108 discrete momentum points as indicated in the caption .
see text for other simulation parameters .
the pronounced minima are caused by the flattening of the distribution function near thresholds of @xmath45-photon - assisted scattering between the vicinities of @xmath99 and @xmath100 , with @xmath45 even.,width=288 ] in fig .
[ fig : powervsj ] , we show how the average power @xmath109 radiated into the phonon modes scales with the tight - binding parameter @xmath13 . while general dependence on @xmath13 is monotonic , at @xmath102mev , where the third harmonic has a minimum , @xmath109 changes slope . ) , ( [ eq : rates ] ) with @xmath110 momentum points and the tight - binding parameters @xmath13 as indicated .
see text for other simulation parameters .
dashed lines : equilibrium fermi distribution functions .
, width=259 ] ) ( arbitrary units ) for scattering between the sites at @xmath111 and @xmath112 ( black , solid ) and the individual contributions from @xmath45-photon assisted processes as indicated .
the vertical dashed line at @xmath113mev indicates the threshold for the @xmath114 transition , @xmath115 ; the peaks to the left and to the right of this point correspond to phonon emission and absorption , respectively.,width=288 ] .
the four curves on top of each other correspond to the same numbers of discrete momentum points @xmath116 as in fig .
[ fig : i3vj].,width=278 ]
the simulation results in the previous section suggest that the optimal system for third harmonic generation would be a one - dimensional metallic conductor with an unrenormalized bandwidth close to @xmath117 times the energy @xmath95 of the driving field quanta ( bandwidth of about @xmath118mev for @xmath2thz is needed ) , and a wide gap to reduce the absorption of the generated harmonics .
one option to satisfy these requirements is to use modulated semiconductor nanowires .
here we estimate the growth parameters of an inas / inp nanowire@xcite , which would have a near optimal band structure for generating the third harmonic of a 1thz driving field .
we calculate the band structure of the modulated nanowire modeling it as a stack of cylinders with isotropic ( bulk ) electron effective masses @xmath119 and @xmath120 for the inas and inp carriers respectively , as appropriate for the nanowire diameter we used@xcite .
we used the barrier height of @xmath121ev , found from the four - band model simulations , which is close to experimentally observed@xcite @xmath122ev . to ensure a relatively large gap ,
we chose the nanowire diameter @xmath123 nm , and inas well width @xmath124 nm . separating the radial and angular parts of the corresponding wave functions , we obtained a version of the kronig - penney model with effective mass modulation , and effective barrier dependent on the transverse momentum @xmath125 .
we plot the first few allowed energy bads as a function of inp barrier width @xmath126 in fig .
[ fig : kplevj0r1 ] .
nm , inas well width @xmath124 nm , plotted as a function of inp barrier width , @xmath127 .
the labels indicate the radial @xmath128 and angular @xmath129 quantum numbers of dimensional quantization .
the dashed line at @xmath130 nm gives a bandwidth @xmath131mev , or tight - binding parameter @xmath98 , corresponding to the first maximum of the third harmonic in fig .
[ fig : i3vj].,width=288 ] in particular , we conclude that an inas / inp nanowire of diameter @xmath132 nm , well width of @xmath124 nm , and barrier width of @xmath130 nm [ fig .
[ fig : nanowire ] ( a ) ] would have the lowest band with a width of approximately @xmath131mev .
the next band would be separated by a gap of @xmath133mev [ fig .
[ fig : kplevj0r1 ] ] .
these parameters are near optimal for third harmonic generation at @xmath2thz
. one possible device design could involve depositing of a number of parallel modulated nanowires on a substrate , with an @xmath134-polarized driving field incident on the surface at angle @xmath135 so that the electric field of the wave be directed along the nanowires [ fig .
[ fig : nanowire ] ( b),(c ) ] . then both the reflected signal and the first harmonic are going to be propagating at the same reflection angle @xmath135 , while the propagation direction of the third harmonic can be found from the snell s law , @xmath136 , which accounts for the wavelengths ratio .
in this work we suggest a possibility that frequency multiplication due to periodically - driven bloch oscillation may be possible in a quasistationary setting , with the help of a narrow - band one - dimensional conductor .
a quasi - equilibrium electron distribution is possible because the energy absorbed from the driving field is continuously dissipated by the bulk phonons . for
a periodically modulated inas / inp nanowire with the period @xmath89 nm , and the driving field frequency @xmath2thz , the emission of the first harmonic is suppressed with the dimensionless vector potential amplitude @xmath137 , which gives the electric field amplitude @xmath138v / m , corresponding to the energy flux of about @xmath139mwt/@xmath140 . at this kind of power , many effects could lead to eventual run - away overheating of the system , e.g. , direct absorption by the substrate , or even a relatively weak disorder scattering in the nanowire .
we hope that a quasi - continuous operation would still be possible , with the driving field pulse duration of a few microseconds , as opposed to few picoseconds in the experiment@xcite .
the authors are grateful to goutam chattopadhyay , ken cooper , and robert a. suris for multiple helpful discussions , and to craig pryor for letting us use his dot code for nanowire calculations .
this work was supported in part by the u.s .
army research office grant no .
w911nf-11 - 1 - 0027 , and by the nsf grant no
. 1018935 .
bloch , f. _ z. phys .
* 1928 * , _ 52 _ , 555600 yakovlev , v. a. _ sov .
phys .- solid state _ * 1961 * , _ 3 _ , 1442 , [ fiz .
tela * iii * 1983 ( 1961 ) ] yakovlev , v. a. _ sov .
jetp - ussr _
* 1961 * , _ 13 _ , 1194 , [ jetp * 40 * , 1695 ( 1961 ) ] keldysh , l. v. _ sov .
jetp - ussr _
* 1963 * , _ 16 _ , 471 , [ jetp * 43 * , 661 ( 1962 ) ] wannier , g. h. _ rev .
phys . _ * 1962 * , _ 34 _ , 645655 fukuyama , h. ; bari , r. a. ; fogedby , h. c. _ phys .
b _ * 1973 * , _ 8 _ , 55795586 waschke , c. ; roskos , h. g. ; schwedler , r. ; leo , k. ; kurz , h. ; khler , k. _ phys . rev .
_ * 1993 * , _ 70 _ , 33193322 mendez , e. e. ; bastard , g. _ physics today _ * 1993 * , _ 46 _ , 34 ben dahan , m. ; peik , e. ; reichel , j. ; castin , y. ; salomon , c. _ phys .
* 1996 * , _ 76 _ , 45084511 pertsch , t. ; dannberg , p. ; elflein , w. ; bruer , a. ; lederer , f. _ phys .
lett . _ * 1999 * , _ 83 _ , 47524755 morandotti , r. ; peschel , u. ; aitchison , j. s. ; eisenberg , h. s. ; silberberg , y. _ phys .
lett . _ * 1999 * , _ 83 _ , 47569 christodoulides , d. n. ; lederer , f. ; silberberg , y. _ nature _ * 2003 * , _ 424 _ , 81723 faisal , f. h. m. ; kamiski , j. z. _ phys . rev .
* 1997 * , _ 56 _ , 748762 gupta , a. k. ; alon , o. e. ; moiseyev , n. _ phys . rev .
b _ * 2003 * , _ 68 _ , 205101 golde , d. ; meier , t. ; koch , s. w. _ phys . rev .
b _ * 2008 * , _ 77 _ , 075330 ghimire , s. ; dichiara , a. d. ; sistrunk , e. ; agostini , p. ; dimauro , l. f. ; reis , d. a. _ nature physics _ * 2011 * , _ 7 _ , 138141 dunlap , d. h. ; kenkre , v. m. _ phys . rev .
b _ * 1986 * , _ 34 _ , 36253633 grossmann , f. ; dittrich , t. ; jung , p. ; hnggi , p. _ phys . rev
* 1991 * , _ 67 _ , 516519 keldysh , l. v. _ zh . eksp . teor .
fiz . _ * 1964 * , _ 47 _ , 1515 , [ sov . phys .
jetp * 20 * , 1018 ( 1965 ) ] rammer , j. ; smith , h. _ rev . mod .
phys . _ * 1986 * , _ 58 _ , 32359 kamenev , a. lectures notes for 2004 les houches summer school on `` nanoscopic quantum transport '' haug , h. ; jauho , a .- p .
_ quantum kinetics in transport and optics of semiconductors _ , 2nd ed . ;
springer : new york , 2008 devries , p. l. _ j. opt .
b _ * 1990 * , _ 7 _ , 517520 joachain , c. j. ; kylstra , n. j. ; potvliege , r. m. _ atoms in intense laser fields _ ; cambridge university press : cambridge , uk , 2011 kemper , a. f. ; moritz , b. ; freericks , j. k. ; devereaux , t. p. _ new journal of physics _ * 2013 * , _ 15 _ , 023003 konstantinov , o. v. ; perel , v. i. _ zh .
_ * 1960 * , _ 39 _ , 197208 pryadko , l. p. ; sengupta , p. _ phys . rev .
* 2006 * , _ 73 _ , 085321 hamilton , k. e. ; kovalev , a. a. ; pryadko , l. p. unpublished kadanoff , l. p. ; baym , g. _ quantum statistical mechanics _ ; benjamin : new york , 1962 bjork , m. t. ; ohlsson , b. j. ; sass , t. ; persson , a. i. ; thelander , c. ; magnusson , m. h. ; deppert , k. ; wallenberg , l. r. ; samuelson , l. _ applied physics letters _ * 2002 * , _ 80 _ , 10581060 moreira , m. d. ; venezuela , p. ; miwa , r. h. _ nanotechnology _ * 2010 * , _ 21 _ , 285204 thelander , c. ; bjrk , m. t. ; larsson , m. w. ; hansen , a. e. ; wallenberg , l. r. ; samuelson , l. _ solid state communications _ * 2004 * , _ 131 _ , 573 579
|
we consider the possibility of observing continuous third - harmonic generation using a strongly driven , single - band one - dimensional metal . in the absence of scattering ,
the quantum efficiency of frequency tripling for such a system can be as high as 93% . combining the floquet quasi - energy spectrum with the keldysh green s function technique
, we derive the semiclassical master equation for a one - dimensional band of strongly and rapidly driven electrons in the presence of weak scattering by phonons . the power absorbed from the driving field is continuously dissipated by phonon modes , leading to a quasi - equilibrium in the electron distribution .
we use the kronig - penney model with varying effective mass to establish growth parameters of an inas / inp nanowire near optimal for third harmonic generation at terahertz frequency range .
|
agglomerative hierarchical clustering has been the dominant approach to constructing embedded classification schemes .
it is our aim to direct the reader s attention to practical algorithms and methods both efficient ( from the computational and storage points of view ) and effective ( from the application point of view ) .
it is often helpful to distinguish between _ method _ , involving a compactness criterion and the target structure of a 2-way tree representing the partial order on subsets of the power set ; as opposed to an _ implementation _ , which relates to the detail of the algorithm used .
as with many other multivariate techniques , the objects to be classified have numerical measurements on a set of variables or attributes .
hence , the analysis is carried out on the rows of an array or matrix .
if we do not have a matrix of numerical values to begin with , then it may be necessary to skilfully construct such a matrix .
the objects , or rows of the matrix , can be viewed as vectors in a multidimensional space ( the dimensionality of this space being the number of variables or columns ) .
a geometric framework of this type is not the only one which can be used to formulate clustering algorithms .
suitable alternative forms of storage of a rectangular array of values are not inconsistent with viewing the problem in geometric terms ( and in matrix terms for example , expressing the adjacency relations in a graph ) .
motivation for clustering in general , covering hierarchical clustering and applications , includes the following : analysis of data ; interactive user interfaces ; storage and retrieval ; and pattern recognition .
surveys of clustering with coverage also of hierarchical clustering include gordon ( 1981 ) , march ( 1983 ) , jain and dubes ( 1988 ) , gordon ( 1987 ) , mirkin ( 1996 ) , jain , murty and flynn ( 1999 ) , and xu and wunsch ( 2005 ) .
lerman ( 1981 ) and janowitz ( 2010 ) present overarching reviews of clustering including through use of lattices that generalize trees .
the case for the central role of hierarchical clustering in information retrieval was made by van rijsbergen ( 1979 ) and continued in the work of willett ( e.g. griffiths et al .
, 1984 ) and others .
various mathematical views of hierarchy , all expressing symmetry in one way or another , are explored in murtagh ( 2009 ) .
this article is organized as follows . in section [ sect5 ]
we look at the issue of normalization of data , prior to inducing a hierarchy on the data . in section [ sect1 ]
some historical remarks and motivation are provided for hierarchical agglomerative clustering . in section [ sect3 ]
, we discuss the lance - williams formulation of a wide range of algorithms , and how these algorithms can be expressed in graph theoretic terms and in geometric terms . in section [ sect2 ] , we describe the principles of the reciprocal nearest neighbor and nearest neighbor chain algorithm , to support building a hierarchical clustering in a more efficient manner compared to the lance - williams or general geometric approaches .
in section [ sect6 ] we overview the hierarchical kohonen self - organizing feature map , and also hierarchical model - based clustering .
we conclude this section with some reflections on divisive hierarchical clustering , in general .
section [ sect7 ] surveys developments in grid- and density - based clustering .
the following section , section [ sect8 ] , presents a recent algorithm of this type , which is particularly suitable for the hierarchical clustering of massive data sets .
before clustering comes the phase of data measurement , or measurement of the observables . let us look at some important considerations to be taken into account .
these considerations relate to the metric or other spatial embedding , comprising the first phase of the data analysis _
stricto sensu_. to group data we need a way to measure the elements and their distances relative to each other in order to decide which elements belong to a group .
this can be a similarity , although on many occasions a dissimilarity measurement , or a `` stronger '' distance , is used . a distance between any pair of vectors or points @xmath0 satisfies the properties of : symmetry , @xmath1 ; positive definiteness , @xmath2 and @xmath3 iff @xmath4 ; and the triangular inequality , @xmath5 .
if the triangular inequality is not taken into account , we have a dissimilarity .
finally a similarity is given by @xmath6 max@xmath7 .
when working in a vector space a traditional way to measure distances is a minkowski distance , which is a family of metrics defined as follows : @xmath8 where @xmath9 is the set of positive integers .
the manhattan , euclidean and chebyshev distances ( the latter is also called maximum distance ) are special cases of the minkowski distance when @xmath10 and @xmath11 . as an example of similarity we have the _ cosine _
similarity , which gives the angle between two vectors .
this is widely used in text retrieval to match vector queries to the dataset . the smaller the angle between a query vector and a document vector ,
the closer a query is to a document .
the normalized cosine similarity is defined as follows : @xmath12 where @xmath13 is the dot product and @xmath14 the norm .
other relevant distances are the hellinger , variational , mahalanobis and hamming distances .
anderberg ( 1973 ) gives a good review of measurement and metrics , where their interrelationships are also discussed .
also deza and deza ( 2009 ) have produced a comprehensive list of distances in their _ encyclopedia of distances_. by mapping our input data into a euclidean space , where each object is equiweighted , we can use a euclidean distance for the clustering that follows .
correspondence analysis is very versatile in determining a euclidean , factor space from a wide range of input data types , including frequency counts , mixed qualitative and quantitative data values , ranks or scores , and others .
further reading on this is to be found in benzcri ( 1979 ) , le roux and rouanet ( 2004 ) and murtagh ( 2005 ) .
agglomerative hierarchical clustering algorithms can be characterized as _
greedy _ , in the algorithmic sense . a sequence of irreversible algorithm steps is used to construct the desired data structure .
assume that a pair of clusters , including possibly singletons , is merged or agglomerated at each step of the algorithm .
then the following are equivalent views of the same output structure constructed on @xmath15 objects : a set of @xmath16 partitions , starting with the fine partition consisting of @xmath15 classes and ending with the trivial partition consisting of just one class , the entire object set ; a binary tree ( one or two child nodes at each non - terminal node ) commonly referred to as a dendrogram ; a partially ordered set ( poset ) which is a subset of the power set of the @xmath15 objects ; and an ultrametric topology on the @xmath15 objects .
an ultrametric , or tree metric , defines a stronger topology compared to , for example , a euclidean metric geometry . for three points , @xmath0 , metric and ultrametric respect the properties of symmetry ( @xmath17 , @xmath18 ) and positive definiteness ( @xmath19 and if @xmath20 then @xmath21 ) . a metric
though ( as noted in section [ sect5 ] ) satisfies the triangular inequality , @xmath5 while an ultrametric satisfies the strong triangular or ultrametric ( or non - archimedean ) , inequality , @xmath22 . in section [ sect5 ] , above ,
there was further discussion on metrics .
the single linkage hierarchical clustering approach outputs a set of clusters ( to use graph theoretic terminology , a set of maximal connected subgraphs ) at each level or for each threshold value which produces a new partition .
the single linkage method with which we begin is one of the oldest methods , its origins being traced to polish researchers in the 1950s ( graham and hell , 1985 ) .
single linkage _ arises since the interconnecting dissimilarity between two clusters or components is defined as the least interconnecting dissimilarity between a member of one and a member of the other .
other hierarchical clustering methods are characterized by other functions of the interconnecting linkage dissimilarities . as early as the 1970s
, it was held that about @xmath23 of all published work on clustering employed hierarchical algorithms ( blashfield and aldenderfer , 1978 ) .
interpretation of the information contained in a dendrogram is often of one or more of the following kinds : set inclusion relationships , partition of the object - sets , and significant clusters .
much early work on hierarchical clustering was in the field of biological taxonomy , from the 1950s and more so from the 1960s onwards .
the central reference in this area , the first edition of which dates from the early 1960s , is sneath and sokal ( 1973 ) .
one major interpretation of hierarchies has been the evolution relationships between the organisms under study .
it is hoped , in this context , that a dendrogram provides a sufficiently accurate model of underlying evolutionary progression . a common interpretation made of hierarchical clustering is to derive a partition .
a further type of interpretation is instead to detect maximal ( i.e. disjoint ) clusters of interest at varying levels of the hierarchy .
such an approach is used by rapoport and fillenbaum ( 1972 ) in a clustering of colors based on semantic attributes .
lerman ( 1981 ) developed an approach for finding significant clusters at varying levels of a hierarchy , which has been widely applied . by developing a wavelet transform _ on _ a dendrogram ( murtagh , 2007 ) , which amounts to a wavelet trasform in the associated ultrametric topological space , the most important in the sense of best approximating
clusters can be determined .
such an approach is a topological one ( i.e. , based on sets and their properties ) as contrasted with more widely used optimization or statistical approaches . in summary
, a dendrogram collects together many of the proximity and classificatory relationships in a body of data .
it is a convenient representation which answers such questions as : `` how many useful groups are in this data ? '' , `` what are the salient interrelationships present ? '' .
but it can be noted that differing answers can feasibly be provided by a dendrogram for most of these questions , depending on the application .
a wide range of agglomerative hierarchical clustering algorithms have been proposed at one time or another .
such hierarchical algorithms may be conveniently broken down into two groups of methods .
the first group is that of linkage methods the single , complete , weighted and unweighted average linkage methods .
these are methods for which a graph representation can be used .
sneath and sokal ( 1973 ) may be consulted for many other graph representations of the stages in the construction of hierarchical clusterings .
the second group of hierarchical clustering methods are methods which allow the cluster centers to be specified ( as an average or a weighted average of the member vectors of the cluster ) .
these methods include the centroid , median and minimum variance methods .
the latter may be specified either in terms of dissimilarities , alone , or alternatively in terms of cluster center coordinates and dissimilarities .
a very convenient formulation , in dissimilarity terms , which embraces all the hierarchical methods mentioned so far , is the _ lance - williams dissimilarity update formula_. if points ( objects ) @xmath24 and @xmath25 are agglomerated into cluster @xmath26 , then we must simply specify the new dissimilarity between the cluster and all other points ( objects or clusters ) .
the formula is : @xmath27 where @xmath28 , @xmath29 , @xmath30 , and @xmath31 define the agglomerative criterion .
values of these are listed in the second column of table [ tabhier ] .
.specifications of seven hierarchical clustering methods . [ cols= " < , < , < , < " , ] notes : @xmath32 is the number of objects in cluster @xmath24 .
@xmath33 is a vector in @xmath34-space ( @xmath34 is the set of attributes ) , either an intial point or a cluster center .
@xmath35 is the norm in the euclidean metric .
the names upgma , etc .
are due to sneath and sokal ( 1973 ) .
coefficient @xmath29 , with index @xmath25 , is defined identically to coefficient @xmath28 with index @xmath24 .
finally , the lance and williams recurrence formula is ( with @xmath36 expressing absolute value ) : @xmath37 in the case of the single link method , using @xmath38 , @xmath39 , and @xmath40 gives us @xmath41 which , it may be verified , can be rewritten as @xmath42 using other update formulas , as given in column 2 of table [ tabhier ] , allows the other agglomerative methods to be implemented in a very similar way to the implementation of the single link method . in the case of the methods which use cluster centers , we have the center coordinates ( in column 3 of table [ tabhier ] ) and dissimilarities as defined between cluster centers ( column 4 of table [ tabhier ] ) .
the euclidean distance must be used for equivalence between the two approaches . in the case of the _ median method _ , for instance , we have the following ( cf.table [ tabhier ] ) .
let @xmath43 and @xmath44 be two points ( i.e. @xmath34-dimensional vectors : these are objects or cluster centers ) which have been agglomerated , and let @xmath45 be another point
. from the lance - williams dissimilarity update formula , using squared euclidean distances , we have : @xmath46 & { { \| { \bf a}-{\bf c } \|^2 } \over { 2 } } + { { \|{\bf b}-{\bf c}\|^2 } \over { 2 } } - { { \|{\bf a}-{\bf b}\|^2 } \over { 4 } } .
\end{array}\ ] ] the new cluster center is @xmath47 , so that its distance to point @xmath48 is @xmath49 that these two expressions are identical is readily verified . the correspondence between these two perspectives on the one agglomerative criterion is similarly proved for the centroid and minimum variance methods . for cluster center methods , and with suitable alterations for graph methods ,
the following algorithm is an alternative to the general dissimilarity based algorithm .
the latter may be described as a `` stored dissimilarities approach '' ( anderberg , 1973 ) .
* stored data approach * step 1 : : examine all interpoint dissimilarities , and form cluster from two closest points .
step 2 : : replace two points clustered by representative point ( center of gravity ) or by cluster fragment .
step 3 : : return to step 1 , treating clusters as well as remaining objects , until all objects are in one cluster . in steps 1 and 2 , ``
point '' refers either to objects or clusters , both of which are defined as vectors in the case of cluster center methods .
this algorithm is justified by storage considerations , since we have @xmath50 storage required for @xmath15 initial objects and @xmath50 storage for the @xmath16 ( at most ) clusters . in the case of linkage methods ,
the term `` fragment '' in step 2 refers ( in the terminology of graph theory ) to a connected component in the case of the single link method and to a clique or complete subgraph in the case of the complete link method . without consideration of any special algorithmic `` speed - ups '' , the overall complexity of the above algorithm is @xmath51 due to the repeated calculation of dissimilarities in step 1 , coupled with @xmath50 iterations through steps 1 , 2 and 3 . while the stored data algorithm is instructive , it does not lend itself to efficient implementations . in the section to follow ,
we look at the reciprocal nearest neighbor and mutual nearest neighbor algorithms which can be used in practice for implementing agglomerative hierarchical clustering algorithms . before concluding this overview of agglomerative hierarchical clustering algorithms
, we will describe briefly the minimum variance method .
the variance or spread of a set of points ( i.e. the average of the sum of squared distances from the center ) has been a point of departure for specifying clustering algorithms .
many of these algorithms , iterative , optimization algorithms as well as the hierarchical , agglomerative algorithms are described and appraised in wishart ( 1969 ) .
the use of variance in a clustering criterion links the resulting clustering to other data - analytic techniques which involve a decomposition of variance , and make the minimum variance agglomerative strategy particularly suitable for synoptic clustering .
hierarchies are also more balanced with this agglomerative criterion , which is often of practical advantage
. the minimum variance method produces clusters which satisfy compactness and isolation criteria .
these criteria are incorporated into the dissimilarity .
we seek to agglomerate two clusters , @xmath52 and @xmath53 , into cluster @xmath54 such that the within - class variance of the partition thereby obtained is minimum .
alternatively , the between - class variance of the partition obtained is to be maximized .
let @xmath55 and @xmath56 be the partitions prior to , and subsequent to , the agglomeration ; let @xmath57 , @xmath58 , be classes of the partitions : @xmath59 letting @xmath60 denote _ variance _ , then in agglomerating two classes of @xmath55 , the variance of the resulting partition ( i.e. @xmath61 ) will necessarily decrease : therefore in seeking to minimize this decrease , we simultaneously achieve a partition with maximum between - class variance .
the criterion to be optimized can then be shown to be : @xmath62 & { { \mid c_1 \mid \
\mid c_2 \mid } \over { \mid c_1 \mid + \mid c_2 \mid } } \vert { \bf c_1 } - { \bf c_2 } \vert ^2 \ , \end{array}\ ] ] which is the dissimilarity given in table [ tabhier ] .
this is a dissimilarity which may be determined for any pair of classes of partition @xmath55 ; and the agglomerands are those classes , @xmath52 and @xmath53 , for which it is minimum .
it may be noted that if @xmath52 and @xmath53 are singleton classes , then @xmath63 , i.e. the variance of a pair of objects is equal to half their euclidean distance .
early , efficient algorithms for hierarchical clustering are due to sibson ( 1973 ) , rohlf ( 1973 ) and defays ( 1977 ) .
their @xmath64 implementations of the single link method and of a ( non - unique ) complete link method , respectively , have been widely cited . in the early 1980s a range of significant improvements ( de rham , 1980 ; juan , 1982 ) were made to the lance - williams , or related , dissimilarity update schema , which had been in wide use since the mid-1960s .
murtagh ( 1983 , 1985 ) presents a survey of these algorithmic improvements .
we will briefly describe them here .
the new algorithms , which have the potential for _ exactly _ replicating results found in the classical but more computationally expensive way , are based on the construction of _ nearest neighbor chains _ and _ reciprocal _ or mutual nns ( nn - chains and rnns ) .
a nn - chain consists of an arbitrary point ( @xmath65 in fig .
[ figrnns ] ) ; followed by its nn ( @xmath66 in fig .
[ figrnns ] ) ; followed by the nn from among the remaining points ( @xmath54 , @xmath17 , and @xmath67 in fig .
[ figrnns ] ) of this second point ; and so on until we necessarily have some pair of points which can be termed reciprocal or mutual nns .
( such a pair of rnns may be the first two points in the chain ; and we have assumed that no two dissimilarities are equal . ) ( 0,3)(6,2.5 ) ( 7,4 ) ( 5,4 ) ( 4,4 ) ( 2,4 ) ( -1,4 ) ( 7,3)e ( 5,3)d ( 4,3)c ( 2,3)b ( -1,3)a ( -1,4)(1,0)3 ( 2,4)(1,0)2 ( 4,4)(1,0)1 ( 5,4)(-1,0)1 in constructing a nn - chain , irrespective of the starting point , we may agglomerate a pair of rnns as soon as they are found .
what guarantees that we can arrive at the same hierarchy as if we used traditional `` stored dissimilarities '' or `` stored data '' algorithms ?
essentially this is the same condition as that under which no inversions or reversals are produced by the clustering method .
[ figinv ] gives an example of this , where @xmath68 is agglomerated at a lower criterion value ( i.e. dissimilarity ) than was the case at the previous agglomeration between @xmath69 and @xmath70 .
our ambient space has thus contracted because of the agglomeration .
this is due to the algorithm used in particular the agglomeration criterion and it is something we would normally wish to avoid .
this is formulated as : @xmath71 @xmath72 this is one form of bruynooghe s _ reducibility property _ ( bruynooghe , 1977 ; see also murtagh , 1984 ) . using the lance - williams dissimilarity update formula , it can be shown that the minimum variance method does not give rise to inversions ; neither do the linkage methods ; but the median and centroid methods can not be guaranteed _ not _ to have inversions . to return to fig
. [ figrnns ] , if we are dealing with a clustering criterion which precludes inversions , then @xmath54 and @xmath17 can justifiably be agglomerated , since no other point ( for example , @xmath66 or @xmath67 ) could have been agglomerated to either of these .
the processing required , following an agglomeration , is to update the nns of points such as @xmath66 in fig .
[ figrnns ] ( and on account of such points , this algorithm was dubbed _ algorithme des clibataires _ , or bachelors algorithm , in de rham , 1980 ) .
the following is a summary of the algorithm : * nn - chain algorithm * step 1 : : select a point arbitrarily .
step 2 : : grow the nn - chain from this point until a pair of rnns is obtained .
step 3 : : agglomerate these points ( replacing with a cluster point , or updating the dissimilarity matrix ) .
step 4 : : from the point which preceded the rnns ( or from any other arbitrary point if the first two points chosen in steps 1 and 2 constituted a pair of rnns ) , return to step 2 until only one point remains . in murtagh ( 1983 , 1984 , 1985 ) and day and edelsbrunner ( 1984 )
, one finds discussions of @xmath64 time and @xmath50 space implementations of ward s minimum variance ( or error sum of squares ) method and of the centroid and median methods .
the latter two methods are termed the upgmc and wpgmc criteria by sneath and sokal ( 1973 ) .
now , a problem with the cluster criteria used by these latter two methods is that the reducibility property is not satisfied by them .
this means that the hierarchy constructed may not be unique as a result of inversions or reversals ( non - monotonic variation ) in the clustering criterion value determined in the sequence of agglomerations .
murtagh ( 1983 , 1985 ) describes @xmath64 time and @xmath64 space implementations for the single link method , the complete link method and for the weighted and unweighted group average methods ( wpgma and upgma ) .
this approach is quite general vis vis the dissimilarity used and can also be used for hierarchical clustering methods other than those mentioned .
day and edelsbrunner ( 1984 ) prove the exact @xmath64 time complexity of the centroid and median methods using an argument related to the combinatorial problem of optimally packing hyperspheres into an @xmath34-dimensional volume .
they also address the question of metrics : results are valid in a wide class of distances including those associated with the minkowski metrics .
the construction and maintenance of the nearest neighbor chain as well as the carrying out of agglomerations whenever reciprocal nearest neighbors meet , both offer possibilities for distributed implementation .
implementations on a parallel machine architecture were described by willett ( 1989 ) .
evidently ( from table [ tabhier ] ) both coordinate data and graph ( e.g. , dissimilarity ) data can be input to these agglomerative methods .
gillet et al .
( 1998 ) in the context of clustering chemical structure databases refer to the common use of the ward method , based on the reciprocal nearest neighbors algorithm , on data sets of a few hundred thousand molecules . applications of hierarchical clustering to bibliographic information retrieval are assessed in griffiths et al .
ward s minimum variance criterion is favored . from details in white and mccain ( 1997 ) , the institute of scientific information ( isi ) clusters citations ( science , and social science ) by first clustering highly cited documents based on a single linkage criterion , and then four more passes are made through the data to create a subset of a single linkage hierarchical clustering .
in the clustan and r statistical data analysis packages ( in addition to hclust in r , see flashclust due to p. langfelder and available on cran , `` comprehensive r archive network '' , cran.r-project.org ) there are implementations of the nn - chain algorithm for the minimum variance agglomerative criterion .
a property of the minimum variance agglomerative hierarchical clustering method is that we can use weights on the objects on which we will induce a hierarchy . by default , these weights are identical and equal to 1 . such weighting of observations to be clustered is an important and practical aspect of these software packages .
it is quite impressive how 2d ( 2-dimensional or , for that matter , 3d ) image signals can handle with ease the scalability limitations of clustering and many other data processing operations .
the contiguity imposed on adjacent pixels or grid cells bypasses the need for nearest neighbor finding .
it is very interesting therefore to consider the feasibility of taking problems of clustering massive data sets into the 2d image domain .
the kohonen self - organizing feature map exemplifes this well . in its basic variant ( kohonen , 1984 , 2001 )
is can be formulated in terms of k - means clustering subject to a set of interrelationships between the cluster centers ( murtagh and fernndez - pajares , 1995 ) .
kohonen maps lend themselves well for hierarchical representation .
lampinen and oja ( 1992 ) , dittenbach et al . (
2002 ) and endo et al .
( 2002 ) elaborate on the kohonen map in this way .
an example application in character recognition is miikkulanien ( 1990 ) .
a short , informative review of hierarchical self - organizing maps is provided by vicente and vellido ( 2004 ) .
these authors also review what they term as probabilistic hierarchical models .
this includes putting into a hierarchical framework the following : gaussian mixture models , and a probabilistic
bayesian alternative to the kohonen self - organizing map termed generative topographic mapping ( gtm ) .
gtm can be traced to the kohonen self - organizing map in the following way .
firstly , we consider the hierarchical map as brought about through a growing process , i.e. the target map is allowed to grow in terms of layers , and of grid points within those layers . secondly , we impose an explicit probability density model on the data .
tino and nabney ( 2002 ) discuss how the local hierarchical models are organized in a hierarchical way . in wang
et al . ( 2000 ) an alternating gaussian mixture modeling , and principal component analysis , is described , in this way furnishing a hierarchy of model - based clusters .
aic , the akaike information criterion , is used for selection of the best cluster model overall .
murtagh et al .
( 2005 ) use a top level gaussian mixture modeling with the ( spatially aware ) plic , pseudo - likelihood information criterion , used for cluster selection and identifiability . then at the next level and potentially also for further divisive , hierarchical levels the gaussian mixture modeling is continued but now using the marginal distributions within each cluster , and using the analogous bayesian clustering identifiability criterion which is the bayesian information criterion , bic . the resulting output is referred to as a model - based cluster tree . the model - based cluster tree algorithm of murtagh et al .
( 2005 ) is a divisive hierarchical algorithm .
earlier in this article , we considered agglomerative algorithms .
however it is often feasible to implement a divisive algorithm instead , especially when a graph cut ( for example ) is important for the application concerned .
mirkin ( 1996 , chapter 7 ) describes divisive ward , minimum variance hierarchical clustering , which is closely related to a bisecting k - means also .
a class of methods under the name of spectral clustering uses eigenvalue / eigenvector reduction on the ( graph ) adjacency matrix . as von luxburg ( 2007 ) points out in reviewing this field of spectral clustering , such methods have `` been discovered , re - discovered , and extended many times in different communities '' .
far from seeing this great deal of work on clustering in any sense in a pessimistic way , we see the perennial and pervasive interest in clustering as testifying to the continual renewal and innovation in algorithm developments , faced with application needs .
it is indeed interesting to note how the clusters in a hierarchical clustering may be _ defined _ by the eigenvectors of a dissimilarity matrix , but subject to carrying out the eigenvector reduction in a particular algebraic structure , a semi - ring with additive and multiplicative operations given by `` min '' and `` max '' , respectively ( gondran , 1976 ) . in the next section , section [ sect7 ] ,
the themes of mapping , and of divisive algorithm , are frequently taken in a somewhat different direction .
as always , the application at issue is highly relevant for the choice of the hierarchical clustering algorithm .
many modern clustering techniques focus on large data sets . in xu and wunsch ( 2008 , p. 215 )
these are classified as follows : * random sampling * data condensation * density - based approaches * grid - based approaches * divide and conquer * incremental learning from the point of view of this article , we select density and grid based approaches , i.e. , methods that either look for data densities or split the data space into cells when looking for groups . in this section
we take a look at these two families of methods .
the main idea is to use a grid - like structure to split the information space , separating the dense grid regions from the less dense ones to form groups .
in general , a typical approach within this category will consist of the following steps as presented by grabusts and borisov ( 2002 ) : 1 . creating a grid structure , i.e. partitioning the data space into a finite number of non - overlapping cells .
2 . calculating the cell density for each cell .
sorting of the cells according to their densities .
4 . identifying cluster centers .
traversal of neighbor cells .
some of the more important algorithms within this category are the following : * * sting : * statistical information grid - based clustering was proposed by wang et al .
( 1997 ) who divide the spatial area into rectangular cells represented by a hierarchical structure .
the root is at hierarchical level 1 , its children at level 2 , and so on .
this algorithm has a computational complexity of @xmath73 , where @xmath74 is the number of cells in the bottom layer .
this implies that scaling this method to higher dimensional spaces is difficult ( hinneburg and keim , 1999 ) .
for example , if in high dimensional data space each cell has four children , then the number of cells in the second level will be @xmath75 , where @xmath34 is the dimensionality of the database .
* * optigrid : * optimal grid - clustering was introduced by hinneburg and keim ( 1999 ) as an efficient algorithm to cluster high - dimensional databases with noise .
it uses data partitioning based on divisive recursion by multidimensional grids , focusing on separation of clusters by hyperplanes .
a cutting plane is chosen which goes through the point of minimal density , therefore splitting two dense half - spaces .
this process is applied recursively with each subset of data .
this algorithm is hierarchical , with time complexity of @xmath76 ( gan et al .
, 2007 , pp .
210212 ) .
* * gridclus : * proposed by schikute ( 1996 ) is a hierarchical algorithm for clustering very large datasets .
it uses a multidimensional data grid to organize the space surrounding the data values rather than organize the data themselves .
thereafter patterns are organized into blocks , which in turn are clustered by a topological neighbor search algorithm .
five main steps are involved in the gridclus method : ( a ) insertion of points into the grid structure , ( b ) calculation of density indices , ( c ) sorting the blocks with respect to their density indices , ( d ) identification of cluster centers , and ( e ) traversal of neighbor blocks . * * wavecluster : * this clustering technique proposed by sheikholeslami et al .
( 2000 ) defines a uniform two dimensional grid on the data and represents the data points in each cell by the number of points .
thus the data points become a set of grey - scale points , which is treated as an image .
then the problem of looking for clusters is transformed into an image segmentation problem , where wavelets are used to take advantage of their multi - scaling and noise reduction properties .
the basic algorithm is as follows : ( a ) create a data grid and assign each data object to a cell in the grid , ( b ) apply the wavelet transform to the data , ( c ) use the average sub - image to find connected clusters ( i.e. connected pixels ) , and ( d ) map the resulting clusters back to the points in the original space .
there is a great deal of other work also that is based on using the wavelet and other multiresolution transforms for segmentation .
further grid - based clustering algorithms can be found in the following : chang and jin ( 2002 ) , park and lee ( 2004 ) , gan et al .
( 2007 ) , and xu and wunsch ( 2008 ) .
density - based clustering algorithms are defined as dense regions of points , which are separated by low - density regions .
therefore , clusters can have an arbitrary shape and the points in the clusters may be arbitrarily distributed .
an important advantage of this methodology is that only one scan of the dataset is needed and it can handle noise effectively . furthermore the number of clusters to initialize the algorithm is not required .
some of the more important algorithms in this category include the following : * * dbscan : * density - based spatial clustering of applications with noise was proposed by ester et al .
( 1996 ) to discover arbitrarily shaped clusters .
since it finds clusters based on density it does not need to know the number of clusters at initialization time .
this algorithm has been widely used and has many variations ( e.g. , see gdbscan by sander et al .
( 1998 ) , pdbscan by xu et al .
( 1999 ) , and dbcluc by zaane and lee ( 2002 ) . *
* bridge : * proposed by dash et al . ( 2001 ) uses a hybrid approach integrating @xmath77-means to partition the dataset into @xmath77 clusters , and then density - based algorithm dbscan is applied to each partition to find dense clusters . *
* dbclasd : * distribution - based clustering of large spatial databases ( see xu et al . , 1998 ) assumes that data points within a cluster are uniformly distributed .
the cluster produced is defined in terms of the nearest neighbor distance .
* * denclue : * density based clustering aims to cluster large multimedia data .
it can find arbitrarily shaped clusters and at the same time deals with noise in the data .
this algorithm has two steps .
first a pre - cluster map is generated , and the data is divided in hypercubes where only the populated are considered .
the second step takes the highly populated cubes and cubes that are connected to a highly populated cube to produce the clusters . for a detailed presentation of these steps
see hinneburg and keim ( 1998 ) . * * cubn : * this has three steps .
first an erosion operation is carried out to find border points .
second , the nearest neighbor method is used to cluster the border points .
finally , the nearest neighbor method is used to cluster the inner points .
this algorithm is capable of finding non - spherical shapes and wide variations in size .
its computational complexity is @xmath50 with @xmath15 being the size of the dataset . for a detailed presentation of this algorithm
see wang and wang ( 2003 ) .
in the last section , section [ sect7 ] , we have seen a number of clustering methods that split the data space into cells , cubes , or dense regions to locate high density areas that can be further studied to find clusters . for large data sets clustering via an @xmath34-adic ( @xmath34 integer ,
which if a prime is usually denoted as @xmath78 ) expansion is possible , with the advantage of doing so in linear time for the clustering algorithm based on this expansion .
the usual base 10 system for numbers is none other than the case of @xmath79 and the base 2 or binary system can be referred to as 2-adic where @xmath80 .
let us consider the following distance relating to the case of vectors @xmath81 and @xmath82 with 1 attribute , hence unidimensional : @xmath83 this distance defines the longest common prefix of strings .
a space of strings , with this distance , is a baire space .
thus we call this the baire distance : here the longer the common prefix , the closer a pair of sequences .
what is of interest to us here is this longest common prefix metric , which is an ultrametric ( murtagh et al . , 2008 ) .
for example , let us consider two such values , @xmath81 and @xmath82 .
we take @xmath81 and @xmath82 to be bounded by 0 and 1 .
each are of some precision , and we take the integer @xmath84 to be the maximum precision . thus we consider ordered sets @xmath85 and @xmath86 for @xmath87 .
so , @xmath88 is the index of the first decimal place of precision ; @xmath89 is the index of the second decimal place ; . .
; @xmath90 is the index of the @xmath84th decimal place .
the cardinality of the set k is the precision with which a number , @xmath81 , is measured .
consider as examples @xmath91 ; and @xmath92 . in these cases ,
start from the first decimal position . for @xmath88
, we have @xmath94 . for @xmath89 , @xmath95 .
but for @xmath96 , @xmath97 . hence their baire distance is @xmath98 for base @xmath79 .
it is seen that this distance splits a unidimensional string of decimal values into a 10-way hierarchy , in which each leaf can be seen as a grid cell . from equation ( [ eqnbaire ] ) we can read off the distance between points assigned to the same grid cell .
all pairwise distances of points assigned to the same cell are the same .
clustering using this baire distance has been successfully applied to areas such as chemoinformatics ( murtagh et al . ,
2008 ) , astronomy ( contreras and murtagh , 2009 ) and text retrieval ( contreras , 2010 ) .
hierarchical clustering methods , with roots going back to the 1960s and 1970s , are continually replenished with new challenges . as a family of algorithms
they are central to the addressing of many important problems .
their deployment in many application domains testifies to how hierarchical clustering methods will remain crucial for a long time to come .
we have looked at both traditional agglomerative hierarchical clustering , and more recent developments in grid or cell based approaches .
we have discussed various algorithmic aspects , including well - definedness ( e.g. inversions ) and computational properties .
we have also touched on a number of application domains , again in areas that reach back over some decades ( chemoinformatics ) or many decades ( information retrieval , which motivated much early work in clustering , including hierarchical clustering ) , and more recent application domains ( such as hierarchical model - based clustering approaches ) .
1 . anderberg mr _ cluster analysis for applications_. academic press , new york , 1973 .
2 . benzcri et coll .
, jp _ lanalyse des donnes
. i. la taxinomie _ , dunod , paris , 1979 ( 3rd ed . ) .
3 . blashfield rk and aldenderfer ms the literature on cluster analysis _ multivariate behavioral research _ 1978 , 13 : 271295 .
bruynooghe m mthodes nouvelles en classification automatique des donnes taxinomiques nombreuses _ statistique et analyse des donnes _ 1977 , no . 3 , 2442 . 5 .
chang j - w and jin d - s , a new cell - based clustering method for large , high - dimensional data in data mining applications , in sac 02 : proceedings of the 2002 acm symposium on applied computing .
new york : acm , 2002 , pp .
contreras p search and retrieval in massive data collections phd thesis .
royal holloway , university of london , 2010 .
contreras p and murtagh f fast hierarchical clustering from the baire distance . in classification as a tool for research , eds . h. hocarek - junge and c. weihs , springer , berlin , 235243 , 2010 .
dash m , liu h , and xu x , @xmath99 : merging distance and density based clustering , in dasfaa 01 : proceedings of the 7th international conference on database systems for advanced applications .
washington , dc : ieee computer society , 2001 , pp .
3239 . 9 .
day whe and edelsbrunner h efficient algorithms for agglomerative hierarchical clustering methods _ journal of classification _ 1984 , 1 : 724 .
defays d an efficient algorithm for a complete link method _ computer journal _ 1977 , 20 : 364366
de rham c la classification hirarchique ascendante selon la mthode des voisins rciproques _
les cahiers de lanalyse des donnes _ 1980 , v : 135144 .
deza mm and deza e _ encyclopedia of distances_. springer , berlin , 2009 . 13 .
dittenbach m , rauber a and merkl d uncovering the hierarchical structure in data using the growing hierarchical self - organizing map _ neurocomputing _ , 2002 , 48(14):199216 .
endo m , ueno m and and tanabe t a clustering method using hierarchical self - organizing maps _ journal of vlsi signal processing _ 32:105118 , 2002
ester m , kriegel h - p , sander j , and xu x , a density - based algorithm for discovering clusters in large spatial databases with noise , in 2nd international conference on knowledge discovery and data mining .
aaai press , 1996 , pp .
226231 . 16 .
gan g , ma c and wu j _ data clustering theory , algorithms , and applications _ society for industrial and applied mathematics .
siam , 2007 .
gillet vj , wild dj , willett p and bradshaw j similarity and dissimilarity methods for processing chemical structure databases _ computer journal _ 1998 , 41 : 547558 . 18 .
gondran m valeurs propres et vecteurs propres en classification hirarchique _ rairo informatique thorique _ 1976 , 10(3 ) : 3946 . 19 .
gordon ad _ classification _ , chapman and hall , london , 1981 .
gordon ad a review of hierarchical classification _ journal of the royal statistical society a _ 1987 , 150 : 119137 .
grabusts p and borisov a using grid - clustering methods in data classification , in parelec 02 : proceedings of the international conference on parallel computing in electrical engineering.washington , dc : ieee computer society , 2002 . 22 .
graham rh and hell p on the history of the minimum spanning tree problem _ annals of the history of computing _ 1985 7 : 4357 .
griffiths a , robinson la and willett p hierarchic agglomerative clustering methods for automatic document classification _ journal of documentation _ 1984 , 40 : 175205 .
hinneburg a and keim da , a density - based algorithm for discovering clusters in large spatial databases with noise , in proceeding of the 4th international conference on knowledge discovery and data mining .
new york : aaai press , 1998 , pp .
5868 . 25 .
hinneburg a and keim d optimal grid - clustering : towards breaking the curse of dimensionality in high - dimensional clustering , in vldb 99 : proceedings of the 25th international conference on very large data bases .
san francisco , ca : morgan kaufmann publishers inc . , 1999 , pp .
506517 . 26
jain ak and dubes rc _ algorithms for clustering data _ prentice - hall , englwood cliffs , 1988 .
jain ak , murty , mn and flynn pj data clustering : a review _ acm computing surveys _ 1999 , 31 : 264323 . 28 .
janowitz , mf _ ordinal and relational clustering _ , world scientific , singapore , 2010 .
juan j programme de classification hirarchique par lalgorithme de la recherche en chane des voisins rciproques _ les cahiers de lanalyse des donnes _ 1982 , vii : 219225 .
kohonen t _ self - organization and associative memory _ springer , berlin , 1984 .
kohonen t _ self - organizing maps _ , 3rd edn . , springer , berlin , 2001 .
lampinen j and oja e clustering properties of hierarchical self - organizing maps _ journal of mathematical imaging and vision _ 2 : 261272 , 1992 . 33 .
lerman , ic _ classification et analyse ordinale des donnes _ , dunod , paris , 1981 .
le roux b and rouanet h _
geometric data analysis : from correspondence analysis to structured data analysis _ , kluwer , dordrecht , 2004 .
von luxburg u a tutorial on spectral clustering _ statistics and computing _ 1997 , 17(4 ) : 395416 .
march st techniques for structuring database records _ acm computing surveys _ 1983 , 15 : 4579 . 37 .
miikkulainien r script recognition with hierarchical feature maps _ connection science _ 1990 , 2 : 83101 .
mirkin b _ mathematical classification and clustering _ kluwer , dordrecht , 1996 .
murtagh f a survey of recent advances in hierarchical clustering algorithms _ computer journal _ 1983 , 26 , 354359 .
murtagh f complexities of hierarchic clustering algorithms : state of the art _ computational statistics quarterly _ 1984 , 1 : 101113 .
murtagh f _ multidimensional clustering algorithms _ physica - verlag , wrzburg , 1985 . 42 . murtagh f and hernndez - pajares m the kohonen self - organizing map method : an assessment , _ journal of classification _ 1995 12 , 165 - 190 .
murtagh f , raftery ae and starck jl bayesian inference for multiband image segmentation via model - based clustering trees , _ image and vision computing _ 2005 , 23 : 587596 . 44 .
murtagh f _ correspondence analysis and data coding with java and r _ , chapman and hall , boca raton , 2005 .
murtagh f the haar wavelet transform of a dendrogram _ journal of classification _ 2007 , 24 : 332 .
murtagh f symmetry in data mining and analysis : a unifying view based on hierarchy _ proceedings of steklov institute of mathematics _ 2009 , 265 : 177198 .
murtagh f and downs g and contreras p hierarchical clustering of massive , high dimensional data sets by exploiting ultrametric embedding _ siam journal on scientific computing _ 2008 , 30(2 ) : 707730 .
park nh and lee ws , statistical grid - based clustering over data streams , _
sigmod record _ 2004 , 33(1 ) : 3237 .
rapoport a and fillenbaum s an experimental study of semantic structures , in eds .
romney , r.n .
shepard and s.b .
nerlove , _ multidimensional scaling ; theory and applications in the behavioral sciences .
2 , applications _ , seminar press , new york , 1972 , 93131 . 50 .
rohlf fj algorithm 76 : hierarchical clustering using the minimum spanning tree _
computer journal _ 1973 , 16 : 9395 . 51 .
sander j , ester m , kriegel h .-
p , and xu x , density - based clustering in spatial databases : the algorithm gdbscan and its applications _ data mining knowledge discovery _ 1998 , 2(2 ) : 169194 .
schikuta e grid - clustering : an efficient hierarchical clustering method for very large data sets , in icpr 96 : proceedings of the 13th international conference on pattern recognition .
washington , dc : ieee computer society , 1996 , pp .
101105 . 53 .
sheikholeslami g , chatterjee s and zhang a , wavecluster : a wavelet based clustering approach for spatial data in very large databases , _ the vldb journal _
, 2000 , 8(34 ) : 289304 .
sibson r slink : an optimally efficient algorithm for the single link cluster method _ computer journal _ 1973 , 16 : 3034 .
sneath pha and sokal rr _ numerical taxonomy _ , freeman , san francisco , 1973 .
tino p and nabney i hierarchical gtm : constructing localized non - linear projection manifolds in a principled way , _ ieee transactions on pattern analysis and machine intelligence _
, 2002 , 24(5 ) : 639656 .
van rijsbergen cj _ information retrieval _ butterworths , london , 1979 ( 2nd ed . ) .
wang l and wang z - o , cubn : a clustering algorithm based on density and distance , in proceeding of the 2003 international conference on machine learning and cybernetics .
ieee press , 2003 , pp . 108112 .
wang w , yang j and muntz r sting : a statistical information grid approach to spatial data mining , in vldb 97 : proceedings of the 23rd international conference on very large data bases.san francisco , ca : morgan kaufmann publishers inc . , 1997 , pp
wang y , freedman m.i . and kung s .- y .
probabilistic principal component subspaces : a hierarchical finite mixture model for data visualization , _ ieee transactions on neural networks _ 2000 , 11(3 ) : 625636 .
white hd and mccain kw visualization of literatures , in m.e .
williams , ed . , _ annual review of information science and technology ( arist ) _ 1997 , 32 : 99168
vicente d and vellido a review of hierarchical models for data clustering and visualization . in r. girldez , j.c .
riquelme and j.s .
aguilar - ruiz , eds .
, tendencias de la minera de datos en espaa .
red espaola de minera de datos , 2004 .
willett p efficiency of hierarchic agglomerative clustering using the icl distributed array processor _ journal of documentation _ 1989 , 45 : 145 . 64 .
wishart d mode analysis : a generalization of nearest neighbour which reduces chaining effects , in ed .
cole , _ numerical taxonomy _
, academic press , new york , 282311 , 1969 .
xu r and wunsch d survey of clustering algorithms _ ieee transactions on neural networks _ 2005 , 16 : 645678 .
xu r and wunsch dc _ clustering _
ieee computer society press , 2008 .
xu x , ester m , kriegel h - p and sander j a distribution - based clustering algorithm for mining in large spatial databases , in icde 98 : proceedings of the fourteenth international conference on data engineering .
washington , dc : ieee computer society , 1998 , pp .
324331 . 68 .
xu x , jger j , and kriegel , h - p a fast parallel clustering algorithm for large spatial databases , _ data mining knowledge discovery _ 1999 , 3(3 ) : 263290 . 69
zaane or and lee c - h , clustering spatial data in the presence of obstacles : a density - based approach , in ideas 02 : proceedings of the 2002 international symposium on database engineering and applications.washington , dc : ieee computer society , 2002 , pp .
|
we survey agglomerative hierarchical clustering algorithms and discuss efficient implementations that are available in r and other software environments .
we look at hierarchical self - organizing maps , and mixture models .
we review grid - based clustering , focusing on hierarchical density - based approaches .
finally we describe a recently developed very efficient ( linear time ) hierarchical clustering algorithm , which can also be viewed as a hierarchical grid - based algorithm .
|
recent progress in experimental implementation of quantum algorithms has demonstrated in principle that quantum computers could solve specific problems in fewer steps than any classical machine @xcite .
these first generation quantum computers were 2-spin molecules in solution .
they were initialized , manipulated and measured at room - temperature using bulk nuclear magnetic resonance ( nmr ) spectroscopy techniques @xcite .
classical redundancy in the ensemble and the discrete nature of the answers ensured that the correct answers were obtained despite gate imperfections and moderate rates of decoherence .
however , the accumulation of errors would be detrimental in larger quantum computers and for longer computations in the future , and methods to protect quantum information will be needed .
classically , errors ( bit flips ) are detected and fixed by error correcting codes .
information is encoded redundantly , and the output contains information on both the encoded input and the errors that have occurred , so that the errors can be reversed .
generalization to quantum information is non - trivial since it is impossible to clone an arbitrary quantum bit ( qubit ) and to measure quantum states without disturbing the system .
furthermore , there is a continuum of possible errors , and finally , entanglement can cause errors to propagate rapidly throughout the system . despite the apparent difficulties , quantum error correction was shown to be possible theoretically , and can be useful for reliable computation even when coding operations are imperfect .
shor and steane @xcite realized a major breakthrough by constructing the first quantum error correcting codes .
prudent use of quantum entanglement enables the information on the errors to be obtained by non - demolition measurements without disturbing the encoded inputs ; it also enables digitization of errors .
these schemes correct for storage errors but not for the extra errors introduced by the coding operations .
the extension to handle gate errors and to achieve reliable computation with a certain accuracy threshold were subsequently developed by many others @xcite . in this paper , we report experimental progress toward this elusive goal of continued quantum computation .
we implement a simple phase error detection scheme @xcite that _ encodes one qubit in two _ and detects a single phase error in either one of the two qubits .
the output state is rejected if an error is detected so that the probability to accept an erroneous state is reduced to the smaller probability of having multiple errors .
our aim is to study the effectiveness of quantum error correction in a real experimental system , focusing on effects arising from imperfections of the logic gates .
therefore , the experiment is designed to eliminate potential artificial origins of bias in the following ways .
first , we compare output states stored with and without coding ( the latter is unprotected but less affected by gate imperfections ) .
second , by ensuring that all qubits used in the code decohere at nearly the same rate , we eliminate apparent improvements brought by having an ancilla with lifetime much longer than the original unencoded qubit .
third , our experiment utilizes only _ naturally occuring _ error processes .
finally , the main error processes and their relative importance to the experiment are thoroughly studied and simulated to substantiate any conclusions . in these aspects ,
our study differs significantly from previous work @xcite demonstrating quantum error correction working only in principle . using nuclear spins as qubits , we performed two sets of experiments : ( 1 )
the `` coding experiments '' in which input states were encoded , stored and decoded , and ( 2 ) the `` control experiments '' in which encoding and decoding were omitted . comparing the output states obtained from the coding and the control experiments , both error correction by coding and extra errors caused by the imperfect coding operations
were taken into account when evaluating the actual advantage of coding . in our experiments ,
coding reduced the net error probabilities to second order as predicted , but at the cost of small additional errors which decreased with the original error probabilities .
we identified the major imperfection in the logic gates to be the inhomogeneity in the radio frequency ( rf ) field used for single spin rotations .
simulation results including both phase damping and rf field inhomogeneity confirmed that the additional errors were mostly caused by rf field inhomogeneity .
the causes and effects of other deviations from theory were also studied . _ tomography experiments _ giving the full density matrices at major stages of the experiments further confirmed the agreement between theoretical expectations and the actual results .
the rest of the paper is structured into five sections : section [ sec : theory ] consists of a comprehensive review of the background material for the subsequent sections of the paper .
this background material includes the phase damping model , the two bit coding scheme , and the theory of bulk nmr quantum computation .
these are reviewed in sections [ sec : pd ] , [ sec:2bitcode ] and [ sec : nmr ] .
readers who are familiar with these subjects can skip the appropriate parts of the review .
section [ sec:2bitnmr ] describes the methods to implement the two bit coding scheme in nmr , and the fidelity measures to evaluate the scheme .
section [ sec : experiment ] presents the experimental details .
section [ sec : result ] consists of the experimental results together with a thorough analysis .
the effects of coding , gate imperfections and the causes and effects of other discrepancies are studied in detail . in section [ sec : conclusion ] , we conclude with a summary of our results . we also discuss syndrome measurement in bulk nmr , the equivalence between logical labeling and coding , the applicability of the two bit detection code as a correction code exploiting classical redundancy in the bulk sample and the signal strength issue in error correction in bulk nmr .
sections [ sec:2bitnmr ] and [ sec : result ] contain the main results of the paper ; the remainder is included for the sake of completeness and to develop notation and terminology we believe will be helpful to the general reader .
phase damping is a decoherence process that results in the loss of coherence between different basis states .
it can be caused by random phase shifts of the system due to its interaction with the environment .
for example , let @xmath0 be an arbitrary pure initial state .
a phase shift , @xmath1 , can be represented as a rotation about the @xmath2 axis by some angle @xmath3 , @xmath4 % = \left [ \begin{array}{cc } % { e^{-i \theta /2}}&{0}\\{0}&{e^{i
\theta /2 } } % \end{array } \right ] % \label{eq : phasekick } \,,\ ] ] where @xmath5 is a pauli matrix .
the resulting state is given by @xmath6 .
let @xmath7 be the density matrix of the initial qubit , @xmath8 % \label{eq : denmat } \,.\ ] ] after the phase shift given by eq.([eq : phasekick ] ) , the density matrix becomes @xmath9 % \,.\ ] ] here , we model phase randomization as a stochastic markov process with @xmath3 drawn from a normal distribution .
the density matrix resulting from averaging over @xmath3 is @xmath10 , % \label{eq : phasedamp}\end{aligned}\ ] ] where @xmath11 is the variance of the distribution of @xmath3 . by markovity ,
the total phase shift during a time period @xmath12 is a random walk process with variance proportional to @xmath12 .
therefore , we replace @xmath13 by @xmath14 in eq.([eq : phasedamp ] ) when the time elapsed is @xmath12 .
since the diagonal and the off - diagonal elements represent the populations of the basis states and the quantum coherence between them , the exponential decay of the off - diagonal elements caused by phase damping signifies the loss of coherence without net change of quanta .
one can represent an arbitrary density matrix for one qubit as a bloch vector @xmath16 , defined to be the real coefficients in the _ pauli matrix decomposition _
@xmath17 where the pauli matrices are given by @xmath18,~ % \sigma_y = \left [ \begin{array}{cc } { 0}&{-i}\\{i}&{0 } \end{array } \right],~ % \sigma_z = \left [ \begin{array}{cc } { 1}&{0}\\{0}&{-1 } \end{array } \right ] \ , .
\label{eq : pauli}\ ] ] the space of all possible bloch vectors is the unit sphere known as the bloch sphere .
phase damping describes the axisymmetric exponential decay of the @xmath19 and @xmath20 components of any bloch vector , as depicted in fig .
[ fig : phasedamping ] .
in contrast to the above picture of phase damping as a continuous process , we now describe an alternative model of phase damping as a discrete process .
this will facilitate understanding of quantum error correction .
the essence is that any quantum process @xmath21 can be written in the _ operator sum representation _ as @xcite @xmath22 where the sum is over a _
finite _ number of _ discrete _ events , @xmath23 , which are analogous to quantum jumps , and @xmath24 is positive . for instance , eq.([eq : phasedampmix ] ) describing phase damping can be rewritten as @xmath25 where @xmath26 . in eq.([eq : opsumrep ] ) , the output @xmath27 can be considered as a ( 1-@xmath28):@xmath28 mixture of @xmath7 and @xmath29 ; in other words , @xmath27 is a mixture of the states after the event `` no jump '' ( @xmath30 ) or `` a phase error '' ( @xmath5 ) has occurred .
the weights @xmath31 and @xmath28 are the probabilities of the two possible events . in general ,
each term @xmath32 in eq.([eq : osr ] ) represents the resulting state after the event @xmath23 has occurred , with probability @xmath33 .
this important interpretation is used throughout the paper . note that the decomposition of @xmath27 is a mathematical interpretation rather than a physical process .
the component states @xmath32 of @xmath27 are not generally obtainable because they are not necessarily orthogonal to each other .
we emphasize that eqs.([eq : phasedampmix ] ) and ( [ eq : opsumrep ] ) describe the same physical process .
eq.([eq : opsumrep ] ) provides a discrete interpretation of phase damping , with the continuously changing parameter @xmath34 embedded in the probabilities of the possible events . for a system of multiple qubits , we _ assume _ independent decoherence on each qubit .
for example , for two qubits @xmath35 and @xmath36 with error probabilities @xmath37 and @xmath38 , the joint process is given by @xmath39 where @xmath7 now denotes the @xmath40 @xmath41 @xmath40 density matrix for the two qubits .
the events @xmath42 and @xmath43 are first order errors , while @xmath44 is second order .
first and second order events occur with probabilities linear and quadratic in the small error probabilities .
having described the noise process , we now proceed to describe a coding scheme that will correct for it .
quantum error correction is similar to its classical analogue in many aspects .
first , the input is encoded in a larger system which goes through the decoherence process , such as transmission through a noisy channel or storage in a noisy environment .
the encoded states ( codewords ) are chosen such that information on the undesired changes ( error syndromes ) can be obtained in the extra degrees of freedom in the system upon decoding .
then corrections can be made accordingly .
however , in contrast to the classical case , quantum errors occur in many different forms such as phase flips and not just as bit flips .
furthermore , the quantum information must be preserved without ever measuring it , because measurement which obtains information about a quantum state inevitably disturbs it .
there are excellent references on the theory of quantum error correction @xcite .
we limit the present discussion to detection codes only .
for a code to detect errors , it suffices to choose the codeword space @xmath45 such that all errors to be detected map @xmath46 to its orthogonal complement . in this way
, detection can be done unambiguously by a projection onto @xmath46 _ without _ distinguishing individual codewords ; hence without disturbing the encoded information . to make this concrete ,
consider the code @xcite @xmath47 where the subscript @xmath48 denotes logical states .
an arbitrary encoded qubit is given by @xmath49 \label{eq : psi } \,.\end{aligned}\ ] ] after the four possible errors in eq.([eq : opsumrep2 ] ) , the possible outcomes are @xmath50 with the _ first order _ erroneous states @xmath51 and @xmath52 orthogonal to the correct state @xmath53 .
therefore , it is possible to distinguish ( [ eq : iipsi ] ) from ( [ eq : zipsi ] ) and ( [ eq : izpsi ] ) by a projective measurement during decoding , which is described next . the encoding and decoding can be performed as follows .
we start with an arbitrary input state and a ground state ancilla , represented as qubits @xmath35 and @xmath36 in the circuit in fig .
[ fig : codingcircuit ] . to encode the input qubit , a hadamard transformation , @xmath54 ,
is applied to the ancilla , followed by a controlled - not from the ancilla to qubit @xmath35 ( written as @xmath55 ) .
let @xmath35 and @xmath36 be the first and second label .
then , the two operations have matrix representations @xmath56~~ % c\!\!n_{ba } = \left [ \begin{array}{cccc } { 1}&{0}&{0}&{0}\\{0}&{0}&{0}&{1}\\ { 0}&{0}&{1}&{0}\\{0}&{1}&{0}&{0 } \end{array } \right ] \,,\ ] ] and the qubits transform as ( fig .
[ fig : codingcircuit ] ) @xmath57 where eq.([eq : psi33 ] ) is the desired encoded state .
the decoding operation is the inverse of the encoding operation ( see fig .
[ fig : codingcircuit ] ) so as to recover the input @xmath58 note that the ancilla becomes @xmath59 upon decoding if and only if a _
single _ phase error has occurred .
moreover , qubits a and b are in product states but they are classically correlated . therefore , syndrome can be read out by a projective measurement on @xmath36 without measuring the encoded state .
the decoding operation transforms the codeword space and its orthogonal complement to the subspaces spanned by @xmath60 and @xmath61 in qubit @xmath36 , while all the encoded information , either with or without error , goes to qubit @xmath35 .
we illustrate the role of entanglement in the digitization and detection of errors as follows .
suppose the error is a phase shift on qubit @xmath35 : @xmath62 , @xmath63 .
then , the encoded state becomes @xmath64 \\ & = & \frac{1}{\sqrt{2}}\frac{1+e^{i \theta}}{2 } \left[\rule{0pt}{2.4ex } a ( |00 \rangle + |11 \rangle ) + b ( |10 \rangle + |01 \rangle ) \right ] \nonumber \\ & + & \frac{1}{\sqrt{2}}\frac{1-e^{i \theta}}{2 } \left[\rule{0pt}{2.4ex } a ( |00 \rangle -|11 \rangle ) + b ( -|10 \rangle + |01 \rangle ) \right ] \,.\end{aligned}\ ] ] the decoded state is now a superposition of the states given by eqs.([eq : phi ] ) and ( [ eq : ziphi ] ) : @xmath65 measurement of qubit @xmath36 projects it to either @xmath66 or @xmath67 . _ because of entanglement _ , qubit @xmath35 is projected to having no phase error , or a complete phase flip .
we quantify the error correcting effect of coding using the discrete interpretation of the noise process , leaving a full discussion of the fidelity to section [ sec:2bitnmr ] .
recall from eq.([eq : opsumrep2 ] ) that the errors @xmath68 , @xmath69 , @xmath42 , and @xmath44 occur with probabilities @xmath70 , @xmath71 , @xmath72 and @xmath73 respectively , and only in the first and the last cases will the output state be accepted .
the probability of accepting the output state is @xmath74 whereas the probability of accepting the correct state is @xmath70 .
the conditional probability of a correct , accepted state is therefore @xmath75 for small @xmath37 , @xmath38 .
the code improves the conditional error probability to second order , as a result of screening out the first order erroneous states .
we conclude this section with a discussion of some properties of the two bit code .
first , the code also applies to mixed input states since the code preserves all constituent pure states in the mixed input .
second , we show here that two qubits are the minimum required to encode one qubit and to detect any phase errors .
let @xmath46 be the 2-dimensional codeword space and @xmath76 be a non - trivial error to be detected .
for phase damping , @xmath76 is unitary and therefore @xmath77 is also 2-dimensional .
moreover , @xmath46 and @xmath77 must be orthogonal if @xmath76 is to be detected .
therefore the minimum dimension of the system is @xmath40 , which requires two qubits .
however , using only two qubits implies other intrinsic limitations .
first , the code can detect but can not distinguish errors .
therefore , it can not correct errors .
moreover , @xmath52 decodes to a correct state in spin @xmath35 which is rejected .
these affect the absolute fidelity ( the overall probability of successful recovery ) but not the conditional fidelity ( the probability of successful recovery if the state is accepted ) .
second , the error @xmath44 can not be detected .
this affects both fidelities but only in second order . to understand why these limitations are intrinsic ,
let @xmath78 be the set of non - trivial errors to be detected .
since @xmath79 has to be orthogonal to @xmath45 for all @xmath80 , and since @xmath46 has a unique orthogonal complement of dimension 2 in a two bit code , it follows that all @xmath79 are equal , and it is impossible to distinguish ( and correct ) the different errors . by the same token , for any distinct errors @xmath81 and @xmath82 , @xmath83 because they are both orthogonal to @xmath84 , which has a unique 2-dimensional orthogonal complement .
therefore , a two bit code that detects single phase errors can never detect double errors .
finally , since a detection code can not correct errors , it can only improve the conditional fidelity of the _ accepted _ states but not the absolute fidelity . here
, we only remark that the conditional fidelity is a better measure in our experiments due to the bulk system used to implement the two bit code .
a discussion of fidelity measures in our experiments and quantum error correction in bulk systems will be postponed until sections [ sec:2bitnmr ] and [ sec : conclusion ] .
the system in which the two bit code is implemented will be described next .
nuclear spin systems are good candidates for quantum computers for many reasons .
nuclear spins can have long coherence times .
coupled operations involving multiple qubits are built in as coupling of spins within molecules .
complex sequences of operations can be programmed and carried out easily using modern spectrometers .
however , the signal from a single spin is so weak that detection is not feasible with current technology unless a bulk sample of identical spin systems is measured .
these identical systems run the same quantum computation in classical parallelism .
computation can be performed at room temperature starting with mixed initial states by distilling the signal of the small excess population in the desired ground state .
how nmr quantum computation can be done is described in detail in the following . + * the quantum system ( hardware ) * in our two bit nmr system , @xmath60 and @xmath59 describe the ground and excited states of the nuclear spin ( the states aligned with and against an externally applied static magnetic field @xmath85 in the @xmath86 direction ) .
as in the previous section , we call the spins denoted by the first and second registers @xmath35 and @xmath36 .
the _ reduced _ hamiltonian for our system is well approximated by ( @xmath87 ) @xcite ( see also fig . [ fig : energydiagram ] ) @xmath88 the first two terms on the right hand side of eq.([eq : nmrhamiltonian ] ) are zeeman splitting terms describing the free precession of spins @xmath35 and @xmath36 about the @xmath89 direction with frequencies @xmath90 and @xmath91 .
the third term describes a spin - spin coupling of @xmath92 hz which is electron mediated .
it is known as the @xmath92 coupling .
@xmath93 represents coupling to the reservoir , such as interactions with other nuclei , and higher order terms in the spin - spin coupling .
+ * universal set of quantum logic gates * a set of logic gates is _ universal _ if any operation can be approximated by some suitable sequence of gates chosen from the set .
depending on whether computation is fault - tolerant or not , the minimun requirements for universality are different . in the latter case , any coupled two - qubit operation together with the set of all single qubit transformations form a _ universal set _ of quantum gates @xcite .
both requirements are satisfied in nmr as follows .
+ * single qubit rotations * spin flip transitions between the two energy eigenstates can be induced by pulsed radio frequency ( rf ) magnetic fields .
these fields , oriented in the @xmath94-plane perpendicular to @xmath85 , selectively address either @xmath35 or @xmath36 by oscillating at angular frequencies @xmath95 or @xmath96 . in the classical picture ,
an rf pulse along the axis @xmath97 rotates a spin about @xmath97 by an angle @xmath3 proportional to the product of the pulse duration and amplitude of the oscillating magnetic field . in the quantum picture
, the rotation operator @xmath98 _ rotates _ the bloch vector ( eq.([eq : blochvec ] ) ) likewise . throughout the paper ,
we denote rotations of @xmath99 along the @xmath100 and @xmath101 axes for spins @xmath35 and @xmath36 by @xmath102 , @xmath103 , @xmath104 , and @xmath105 with respective matrix representations @xmath106 , @xmath107 , @xmath108 , and @xmath109 .
the rotations in the reverse directions are denoted by an additional `` bar '' above the symbols of the original rotations , such as @xmath110 .
the angle of rotation is given explicitly when it differs from @xmath99 .
this set of rotations generates the lie group of all single qubit operations , @xmath111 . for example
, the hadamard transformation can be written as @xmath112 which can be implemented in two pulses .
+ * coupled operations * quantum entanglement , essential to quantum information processing , can be naturally created by the time evolution of the system . in the respective rotating frames of the spins
( tracing the free precession of the uncoupled spins ) , only the @xmath92-coupling term , @xmath113 , is relevant in the time evolution .
entanglement is created because the evolution depends on the state of _ both _ spins .
a frequently used coupled `` operation '' is a time delay of @xmath114 , denoted by @xmath115 , which corresponds to the evolution @xmath116 .
for instance , appending @xmath115 with the _ single qubit rotations _
@xmath117 and @xmath118 about the @xmath119 axes of @xmath35 and @xmath36 , we implement the unitary operation @xmath120 \,,\ ] ] which is a cross - phase modulation between the two qubits .
together with the set of all single qubit transformations , @xmath121 completes the requirement for universality .
for instance , the controlled - not , @xmath55 , mentioned in the previous section , can be written as @xmath122 , and can be implemented by concatenating the sequences for each constituent operation .
it is also crucial that the free evolution which leads to creation of entanglement can be reversed by applying _ refocusing _
@xmath123-pulses , such that creation of entanglement between qubits can be stopped . this technique will be described in detail later .
+ * measurement * the measured quantity in nmr experiments is the time varying voltage induced in a pick - up coil in the @xmath94-plane : @xmath124 .
\label{eq : fid } \end{aligned}\ ] ] the signal @xmath125 , known as the _ free induction decay _ ( fid ) , is recorded with a phase - sensitive detector . in eq.([eq : fid ] ) , the onset of acquisition of the fid is taken to be @xmath126 . if the density matrix @xmath127 has pauli matrix decomposition @xmath128 where @xmath129 are the identity matrix and @xmath130 respectively , then the spectrum of @xmath125 has four lines at frequencies @xmath131 , @xmath132 , @xmath133 , @xmath134 , with corresponding integrated areas ( `` peak integrals '' ) @xmath135 \label{eq : ahigh } \\
i_{a_{low } } & = & -\left[\rule{0pt}{2.4ex } i ( c_{10 } + c_{13 } ) + c_{20 } + c_{23 } \right ]
\label{eq : alow } \\
i_{b_{high } } & = & -\left[\rule{0pt}{2.4ex } i ( c_{01 } - c_{31 } ) + c_{02 } - c_{32 } \right ] \label{eq : bhigh } \\
i_{b_{low } } & = & -\left[\rule{0pt}{2.4ex } i ( c_{01 } + c_{31 } ) + c_{02 } + c_{32 } \right ] \label{eq : blow } \,.\end{aligned}\ ] ] note that the expression @xmath136 ( @xmath137 ) occuring in the _ high _ ( _ low _ ) frequency line of spin @xmath35 is the coefficient of @xmath138 in @xmath127 . similarly , @xmath139 ( @xmath140 ) is the coefficient of @xmath141 .
they signify transitions between the states @xmath142 for spin @xmath35 when spin @xmath36 is in @xmath59 ( @xmath60 ) .
similar observations hold for the high and low lines of spin @xmath36 .
+ * thermal states * in bulk nmr quantum computation at room temperature , a pure initial state is not available due to large thermal fluctuations ( @xmath143 ) .
instead , a convenient class of initial states arises from the thermal equilibrium states ( thermal states ) . in the energy eigenbasis ,
the density matrix is diagonal with diagonal entries proportional to the boltzmann factors , in other words , @xmath144 , where @xmath145 is the thermal energy and @xmath146 is the partition function normalization factor . at room temperature , @xmath147 , @xmath148 and @xmath149 to first order .
for most of the time , the identity term in the above expansion is omitted in the analysis for two reasons .
first , it does not contribute to any signal in eqs.([eq : ahigh])-([eq : blow ] ) .
physically , it represents a completely random mixture which is _ isotropic _ and , by symmetry , has no _ net _ magnetization at any time .
second , the identity is invariant under a wide range of processes .
processes which satisfy @xmath150 are called _
unital_. these include all unitary transformations and phase damping . under unital processes ,
the evolution of the state is completely determined by the evolution of the deviation from identity @xcite , in which case the identity can be neglected .
the unitality assumption is a good approximation in our system
. the main cause of non - unitality is amplitude damping .
in general , for a non - unital but trace preserving process @xmath151 , the observable evolution of the deviation can be understood as follows @xcite . rewriting @xmath152 where @xmath153 is the traceless deviation from the identity and @xmath154 .
then , @xmath155 the observed evolution of the deviation is @xmath156 .
the second term in eq.([eq : deltaevolve ] ) comes from the evolution of @xmath157 when the identity is neglected , and the first term is the correction due to non - unitality . in our experiment , amplitude damping is slow compared to all other time scales , therefore @xmath158 is small and can be treated as a small extra distortion of the state when making the unitality assumption .
+ * temporal labelling * one convenient method to create arbitrary _ initial deviations _ from the thermal mixture is temporal labelling @xcite .
the idea is to add up the results of a series of experiments that begin with different preparation pulses before the intended experiment , so as to cancel out the signals from the undesired components in the initial thermal mixture .
mathematically , let @xmath159 be the set of initial pulses and @xmath160 be the intended computation process .
by linearity , @xmath161 .
summing over the experimental results ( on the left side ) is equivalent to performing the experiment with initial state @xmath162 ( on the right side ) .
temporal labelling assumes the repeatability of the experiments , which is true up to small fluctuations .
+ * example * as an example of the above theories , consider applying the pulse sequence in fig .
[ fig : cnshort ] to the thermal state .
[ fig : cnshort ] the pulses are short compared to other relevant time scales .
therefore , other changes of the system during the pulses are ignored .
the unitary operation implemented by the above sequence is given by @xmath163 \,,\end{aligned}\ ] ] similar to @xmath55 described in section [ sec : nmr ] .
the deviation density matrix of the thermal state is proportional to @xmath164 . neglecting the @xmath92-coupling term , which is much smaller than the zeeman terms , @xmath165 where `` diag '' indicates a diagonal matrix with the given elements . for simplicity ,
we omit the proportionality constant in eq.([eq : rhoth ] ) , and rename the right hand side as @xmath166 .
the sequence transforms @xmath166 to @xmath167 in which the populations of @xmath168 and @xmath169 are interchanged .
therefore , @xmath170 and @xmath55 act in the same way _ on the thermal state_. by inspection of eqs.([eq : rhoth ] ) and ( [ eq : rhocn ] ) , it can be seen that @xmath166 and @xmath171 have zero peak integrals given by eqs.([eq : ahigh])-([eq : blow ] ) . to obtain more information about the states
, a readout pulse @xmath102 can be applied to transform the two states to @xmath172 and @xmath173 now , in @xmath174 is a term @xmath175 with coefficient @xmath176 which contributes to two spectral lines at @xmath177 with equal and positive , real peak integrals .
the readout pulse transforms the unobservable coefficient @xmath178 in @xmath166 to the observable @xmath179 in @xmath174 , yielding information on _ the state before the readout pulse_. similarly , @xmath180 has a @xmath181 term with coefficient @xmath182 which gives rise to two spectral lines with real and opposite peak integrals ( fig . [
fig : cnspect ] ) .
all outputs in our experiments are peak integrals of this type carrying information on the decoded states .
we now describe how the two bit code experiment can be implemented in an ensemble of two - spin systems .
modifications of the standard theories in section [ sec:2bitcode ] are needed .
these include methods for state preparation , designing encoding and decoding pulse sequences , methods to store the qubit with controllable phase damping , and finally methods to read out the decoded qubit .
fidelity measures for deviation density matrices are also defined .
spins @xmath35 and @xmath36 are designated to be the input and the ancilla qubits respectively .
the output states of spin @xmath35 are reconstructed from the peak integrals at frequencies @xmath183 .
[ fig : scheme ] schematically summarizes the major steps in the experiments , with details given in the text .
some notation is defined as follows .
initial states and input states refer to @xmath184 and @xmath185 in fig .
[ fig : scheme ] .
the phrase `` ideal case '' refers to the scenario of having perfect logical operations throughout the experiments and pure phase damping during storage .
+ * initial state preparation * it is necessary to initialize spin @xmath36 to @xmath60 before the experiment .
this can be done with temporal labeling using two experiments : the first experiment starts with no additional pulses ; the second experiment starts with @xmath170 ( fig .
[ fig : cnshort ] ) .
therefore , the equivalent initial state is @xmath186 ( all symbols are as defined in the example in section [ sec : nmr ] ) : @xmath187 \\ & = & \omega_a ~\sigma_z \otimes |0 \rangle \langle 0| % + \omega_b ~i \otimes \sigma_z \label{eq : pureancilla } \,.\end{aligned}\ ] ] the first term in eq.([eq : pureancilla ] ) is the desired initial state
. the second term can not affect the observable of interest , the spectrum at @xmath90 , because of the following .
the identity in @xmath35 is invariant under the preparation pulse @xmath188 .
the input state is thus the identity , which has no coherence to start with .
therefore , the output state after phase damping in both the control and the coding experiment is still the identity .
this is non - trivial in the coding experiment .
however , inspection of eqs.([eq : phi])-([eq : zzphi ] ) shows that spin @xmath35 is changed at most by a phase in the coding experiment .
while eqs.([eq : phi])-([eq : zzphi ] ) applies only to the case when @xmath36 starts in @xmath60 , the result can be generalized to any arbitrary diagonal density matrix in @xmath36 ( proof omitted ) .
it follows from eqs.([eq : ahigh])-([eq : blow ] ) that the second term is not observable in the output spectrum of @xmath35 .
in contrast , the input state in spin @xmath35 can be a mixed state as given by the first term in eq.([eq : pureancilla ] ) , since the phase damping code is still applicable .
different input states can be prepared by rotations about the @xmath101-axis of different angles @xmath189 $ ] to span a semi - circle in the bloch sphere in the @xmath190-plane . due to axisymmetry of phase damping ( eq.([eq : phasedamp ] ) ) , these states suffice to represent all states to test the code .
we conclude with an alternative interpretation of the initial state preparation .
let the fractional populations of @xmath191 , @xmath192 , @xmath193 and @xmath169 be @xmath194 , @xmath195 , @xmath196 and @xmath197 in the thermal state .
then , the initial state after temporal labelling is @xmath198 where the identity term is not omitted , unlike eq.([eq : pureancilla ] ) .
temporal labelling serves to randomize spin @xmath35 in the first term in eq.([eq : initstate ] ) when @xmath36 is @xmath59 .
we have shown previously that the identity input state of @xmath35 is preserved throughout both the coding and the control experiments in the ideal case .
consequently , only the last term in eq.([eq : initstate ] ) contributes to any detectable signals in all the experiments , and we can consider the last term as the initial state .
having justified both pictures using the first term in eq.([eq : pureancilla ] ) and the last term in eq.([eq : initstate ] ) as initial state , both will be used throughout rest of the paper . + * encoding and decoding *
the original encoding and decoding operations are composed of the hadamard transformation and @xmath55 , as defined in section [ sec : nmr ] .
the actual sequences can be simplified and are shown in fig .
[ fig : seqs ] . the operator @xmath199 can be found by multiplying the component operators in fig .
[ fig : seqs ] , giving @xmath200 \label{eq : uenc } \,.\ ] ] the encoded states are slightly different from those in section [ sec:2bitcode ] : @xmath201 but the scheme is nonetheless equivalent to the original one .
the decoding operation @xmath202 is given by @xmath203 = u_{enc}^{\dagger } \label{eq : udec } \,.\ ] ] the possible decoded outputs are the same as in section [ sec:2bitcode ] except for an overall sign in the single error cases .
+ * storage * the time delay between encoding and decoding corresponds to storage time of the quantum state . during this delay time , phase damping , amplitude damping and @xmath92-coupling evolution occur simultaneously . how to make phase damping the dominant process during storage
is explained as follows .
first of all , the time constants of amplitude damping , @xmath204 s , are much longer than those of phase damping , @xmath205 s .
storage times @xmath206 are chosen to satisfy @xmath207 .
this ensures that the effects of amplitude damping are small .
the remaining two processes , phase damping and @xmath92-coupling , can be considered as independent and commuting processes in between any two pulses since all the phase damping operators commute with the @xmath92-coupling evolution @xmath208 .
we choose @xmath209 to be even integers to approximate the identity evolution . as @xmath92
is known with limited accuracy , we add refocusing @xmath123-pulses @xcite to spin @xmath36 ( about the @xmath20-axis ) in the middle and at the end of the phase damping period to ensure trivial evolution under @xmath92-coupling .
these pulses flip the @xmath2 axis for @xmath36 during the second half of the storage time so that evolution in the first half is always reversed by that in the second half . in this way , controllable amount of phase damping is achieved to good approximation .
+ * control experiment * for each storage time @xmath206 , input state and temporal labelling experiment , a control experiment is performed with the coding and decoding operations omitted . since phase damping and @xmath92-coupling
can be considered as independent processes , and @xmath92-coupling is arranged to act trivially , the resulting states illustrate phase damping of spin @xmath35 without coding .
+ * output and readout * for an input state prepared with @xmath188 , the state after encoding , dephasing and decoding ( @xmath210 in fig .
[ fig : scheme ] ) is derived in appendix [ sec : mixstate ] and is given by eq.([eq : codedout ] ) ( from here onwards , @xmath95 is omitted ) @xmath211 ~\otimes~ ( i+\sigma_z)/2 \nonumber \\ + ~ & \left[\rule{0pt}{2.1ex}\right . & % \cos\theta ~ ( p_a + p_b-2 p_a p_b)~ \sigma_z \nonumber \\ & & + \sin\theta ~(- p_a + p_b)~ \sigma_x % \left.\rule{0pt}{2.1ex}\right ] ~\otimes~ ( i-\sigma_z)/2 % \label{eq : textcodedout } \ , . \end{aligned}\ ] ] in the control experiment , the corresponding output state is given by ( eq.([eq : controlout ] ) ) @xmath212 \otimes ( i+\sigma_z)/2 \label{eq : textcontrolout } \,.\ ] ] the initial state used in the derivation of eq.([eq : textcontrolout ] ) is the first term in eq.([eq : pureancilla ] ) , and the encoding and decoding operations are as given by eqs.([eq : uenc ] ) and ( [ eq : udec ] ) . in the ideal case
, the output state can be read out in a single spectrum .
recall that the coefficients of @xmath213 and @xmath214 are the real and imaginary parts of the low and the high frequency lines of @xmath35 .
therefore , the coefficients of @xmath215 and @xmath214 in @xmath216 and @xmath217 can be read out as the real and imaginary parts of the low and the high frequency lines of @xmath35 , if @xmath102 is applied before acquisition .
this pulse transforms the @xmath218-component of @xmath35 to the @xmath219-component leaving the @xmath220-component unchanged , as described in section [ sec : nmr ] .
note that only states with spin @xmath36 being @xmath221 ( @xmath59 ) contribute to the low ( high ) frequency line .
therefore , in the coding experiments , the accepted ( rejected ) states of @xmath35 can be read out separately in the low ( high ) frequency line .
there are no rejected states in the control experiments .
the rest of the paper makes use of the following notation .
`` output states '' or `` accepted states '' refer to the reduced density matrices of @xmath35 _ before _ the readout pulse , and are denoted by @xmath222 and @xmath223 .
rejected states refer to @xmath224 from the coding experiments .
the accepted and rejected states for a given input as calculated from eq.([eq : textcodedout ] ) and eq.([eq : textcontrolout ] ) are summarized in table [ table : outcome ] .
@xmath225 the output states @xmath226 and @xmath227 , as predicted by table [ table : outcome ] , are plotted in fig .
[ fig : prediction ] in the @xmath228-plane of the bloch sphere of @xmath35 .
the north and south poles represent the bloch vectors @xmath229 ( @xmath60 and @xmath59 ) .
the time trajectories of various initial states are indicated by the arrows .
the bloch sphere is distorted to an ellipsoid after each storage time .
we concentrate on the cross - section in one half of the @xmath228-plane , and call the curve an `` ellipse '' for convenience .
the storage times plotted have equal spacing and correspond to @xmath230 .
for each ellipse , @xmath38 is chosen to be the same as @xmath37 .
the main experimental results will comprise of information of this type .
+ * fidelity * one can quantify how well the input states are preserved using various fidelity measures .
in classical communication , the fidelity can be defined as the probability of successful recovery of the input bit string in the worse case . in quantum information processing , when the input is a _ pure _ state , the above definition generalizes to the _ minimal overlap fidelity _
, @xmath231 we emphasize that eq.([eq : fidelity1 ] ) applies to _ pure _ input states only . for simplicity
, fidelities for mixed input states will not be given here .
the reason why eq.([eq : fidelity1 ] ) is sufficient for our purpose will become clear later .
when @xmath232 and @xmath233 are qubit states of unit trace with respective bloch vectors @xmath234 and @xmath235 , eq.([eq : fidelity1 ] ) can be rewritten as @xmath236 recall from eq.([eq : phasedampmix ] ) that , for phase damping , when @xmath237 , @xmath238
. therefore , @xmath239 where we have used the fact @xmath240 for pure states and @xmath241 .
the minimum in eq.([eq : fidelity2 ] ) is attained for input states on the equatorial plane with @xmath242 .
therefore @xmath243 with coding , the accepted state is ( see eqs.([eq : phi])-([eq : zzphi ] ) ) @xmath244 if one considers the conditional fidelity in the accepted state , @xmath233 in eq.([eq : fidelity1 ] ) should be taken as the post measurement density matrix , @xmath245 note that the above expression is identical to the expression for single qubit phase damping but with error probability @xmath246 .
therefore , coding changes the conditional error probability to second order , and the conditional fidelity is improved to @xmath247 .
the amount of distortion can also be summarized by the ellipticities of the `` ellipses '' that result from phase damping .
the ellipticity @xmath248 is defined to be the ratio of the major axis to the minor axis . without coding ,
the major axis remains unchanged under phase damping , and the minor axis shrinks by a factor of @xmath34 , therefore @xmath249 .
using eq.([eq : rate ] ) , @xmath250 with coding , @xmath251 is given by the same expression on the right hand side of eq.([eq : fande ] ) .
in the ideal case , the overlap fidelity and the ellipticity have a one - to - one correspondence . in the presence of imperfections , the overlap fidelity and the ellipticity , one
being the minimum of the input - output overlap and the other being an average parameter of distortion , are more effective in reflecting different types of distortion .
we now generalize to new definitions of fidelity for deviation density matrices for the two bit code . in nmr
, quantum information is encoded in the small deviation of the state from a completely random mixture .
the problem with the usual definitions of fidelity is that they do not change significantly even when the small deviation changes completely .
this is true whether the fidelities are defined for pure or mixed input states . to overcome this problem ,
we introduce the strategy of identifying the initial excess population in @xmath191 as the pure initial state so that usual definitions of fidelity for pure input states are applicable .
this improves the sensitivity of the fidelity measures and provides a closer connection to the pure state picture .
the initial state in eq.([eq : initstate ] ) can be rewritten as @xmath252 where @xmath253 , and @xmath254 \,.\end{aligned}\ ] ] it has already been shown that @xmath255 is irrelevant to the evolution and the measurement of @xmath256 when all processes are unital .
therefore @xmath255 is neglected and the small signal resulting from the slow non - unital processes will be treated as extra distortion to the observable component . the input state prepared by @xmath188
can be written as @xmath257 for the state change @xmath258 , we consider the overlap between @xmath259 and @xmath260 in place of the overlap between @xmath232 and @xmath261 .
this defines a new overlap fidelity @xmath262 @xmath263 @xmath264 similar to the pure state case .
@xmath265 can be calculated from the experimental results in the following manner .
the measured bloch vector of @xmath35 , @xmath266 , is proportional to that defined by @xmath267 . due to limitations in the measurement process
, this proportionality constant @xmath268 is not known a priori . however ,
when @xmath269 in the control experiment , @xmath270 and @xmath271 .
therefore , @xmath272 can be determined . in other words , @xmath273 is used to normalize all other measured output states before using the expression for @xmath265 .
the expression for @xmath265 can also be used for the conditional fidelity in the coding experiment if the post - measurement accepted output state is known .
this requires @xmath274 @xmath263 ( @xmath275 ) to be determined for each storage time . the correct normalization is again given by the output at @xmath276 , which equals @xmath277 .
in summary , each ellipse obtained in the coding and the control experiment is normalized by the amplitude at @xmath269 : @xmath278 \ , .
\label{eq : expfid}\ ] ] it is interesting to note that in contrast to the fidelity measure , the ellipticity measure naturally performs an equivalent normalization , and thus can be used for deviations without modifications .
we now turn to the experimental results , beginning with a description of our apparatus .
we performed our experiments on carbon-13 labeled sodium formate ( choo@xmath279na@xmath280 ) ( fig .
[ fig : formate ] ) at 15@xmath281c .
the nuclear spins of proton and carbon were used as input and ancilla respectively . note that the system was heteronuclear .
the sodium formate sample was a 0.6 milliliter 1.26 molar solution ( 8:1 molar ratio with anhydrous calcium chloride ) in deuterated water @xcite .
the sample was degassed and flame sealed in a thin walled , 5 mm nmr sample tube .
the time constants of phase damping and amplitude damping are shown in table [ table : t12 ] .
the fact @xmath282 ensures that the effect of amplitude damping is small compared to phase damping .
the experimental conditions are chosen so that proton and carbon have almost equal @xmath205 s : the most theoretically interesting scenario in which identical quantum systems are available for coding . coding with qubits having very different @xmath205 s is described in appendix [ sec : difft2 ] .
the time constants of phase damping and amplitude damping are shown in table [ table : t12 ] .
the fact @xmath282 ensures that the effect of amplitude damping is small compared to phase damping .
the experimental conditions are chosen so that proton and carbon have almost equal @xmath205 s .
this eliminates potential bias caused by having a long - lived ancilla when evaluating the effectiveness of coding .
this also realizes a common assumption in coding theory that identical quantum systems are available for coding .
subsidiary experiments with qubits having very different @xmath205 s are described in appendix [ sec : difft2 ] .
@xmath283 phase damping arises from constant or low - frequency non - uniformities of the `` static '' magnetic field which randomize the phase evolution of the spins in the ensemble .
several processes contribute to this inhomogeneity on microscopic or macroscopic scales .
which process dominates phase damping varies from system to system @xcite . for instance
, intermolecular magnetic dipole - dipole interaction dominates phase damping in a solution of small molecules , whereas the modulation of direct electron - nuclear dipole - dipole interactions becomes more important if paramagnetic impurities are present in the solution . for molecules with quadrupolar nuclei ( spin @xmath284 1/2 ) , modulation of the quadrupolar coupling dominates phase damping .
other mechanisms such as chemical shift anisotropy can also dominate phase damping in other circumstances .
these microscopic field inhomogeneities have no net effects on the static field when averaged over time , but they result in irreversible phase randomization with parameters intrinsic to the sample .
another origin of inhomogeneity comes from the macroscopic applied static magnetic field .
in contrast to the intrinsic processes , phase randomization due to this inhomogeneity can be reversed by applying refocusing pulses as long as diffusion of molecules is insignificant .
phase damping caused by the intrinsic irreversible processes alone has a time constant denoted by @xmath205 , while the combined process has a shorter time constant denoted by @xmath285 .
@xmath205 is measured by the carr - purcell - meiboom - gill @xcite experiment using multiple refocusing pulses .
@xmath285 can be estimated from the line - width of the nmr spectral lines : during acquisition , the signal decays exponentially due to phase damping , resulting in lorentzian spectral lines with line - width @xmath286 . in our experiment , @xmath285 s for proton and carbon
were estimated to within 0.05 s to be @xmath287 0.35 and 0.50 s. the storage times @xmath206 were approximately 0 , 62 , 123 , 185 , 246 , 308 ms ( @xmath288 for @xmath289 ) .
the maximum storage time was 120 @xmath115 , long compared to the clock cycle and was comparable to @xmath285 .
the decay constant @xmath290 , defined in section [ sec : pd ] , was given by @xmath291 .
the error probabilities after a storage time of @xmath206 were @xmath292 for spins @xmath293 . to reconstruct the ellipse for each storage time ,
11 experiments were run with input states spanning a semi - circle in the @xmath294-plane .
each input state was prepared by a @xmath188 pulse with @xmath295 for @xmath296 .
all experiments were performed on an oxford instruments superconducting magnet of 11.7 tesla , giving precession frequencies of @xmath297 @xmath287 500 mhz for proton and @xmath298 @xmath287 125 mhz for carbon .
a varian @xmath299_inova _ spectrometer with a triple - resonance probe was used to send the pulsed rf fields to the sample and to measure the fid s .
the rf pulses selectively rotated a particular spin by oscillating on resonance with it .
the @xmath123/2 pulse durations were calibrated , and they were typically 8 to 14 @xmath300s . to perform logical operations in the respective rotating frames of the spins , reference oscillators were used to keep track of the free precession of both spins , leaving only the @xmath92-coupling term of 195.0 hz in the time evolution .
each fid was recorded for @xmath287 6.8 s ( until the signal had faded completely ) .
the thermal state was obtained after a relaxation time of 80 s ( @xmath301 s ) before each pulse sequence . using the above apparatus and procedures
, we performed the experiments as outlined in section [ sec:2bitnmr ] .
the experimental results are described in the next section .
tomography results following the evolution of the state are presented in appendix [ sec : tomo ] .
the output states , @xmath302 and @xmath303 , obtained as described in sections [ sec:2bitnmr ] and [ sec : experiment ] , and the analysis that confirms the correction effects of coding , are presented in this section .
[ fig : sodformellips ] shows the accepted states in the @xmath304-plane of the bloch sphere of @xmath35 .
@xmath216 and @xmath217 are plotted in figs .
[ fig : sodformellips ] a and b. the ellipse for each storage time is obtained by a least - squares fit described later . the most important feature in fig .
[ fig : sodformellips ] is the reduction of the ellipticities of the ellipses due to coding , which represents partial removal of the distortion caused by phase damping - the signature of error correction .
coding is effective throughout the range of storage times tested .
we quantify the correction effects due to coding using the ellipticities .
when deviations from the ideal case such as offsets of the angular positions of the points along the ellipses and attenuation of signal strength with increasing @xmath3 exist , the minimum overlap fidelities and the ellipticities are no longer related by eq.([eq : fande ] ) .
since the ellipticity is an average measure of distortion which is less susceptible to scattering of individual data points , we first study the ellipticities . moreover , since the deviations from the ideal case are small , we can still _ infer _ the fidelities from the ellipticities using eq.([eq : fande ] ) .
a discussion of the discrepancies and the exact overlap fidelities will given later .
+ * ellipticities * in the ideal case , the ellipticity for each ellipse can be obtained experimentally as @xmath305 where @xmath30 denotes the intensity ( amplitude square ) of the peak integral .
@xmath30 is given by the @xmath19 and @xmath306-components of the output states as @xmath307 in the ideal case , @xmath308 can be found from table [ table : outcome ] : @xmath309 and both are of the functional form @xmath310 experimentally , the output bloch vectors do not form perfect ellipses .
we modify eq.([eq : intensity ] ) to include signal strength attenuation with increasing @xmath3 and constant offsets in the angular positions : @xmath311 and perform non - linear least - squares fits of the experimental data to determine @xmath312 and @xmath313 .
the fitted ellipses plotted in fig .
[ fig : sodformellips ] follow from eq.([eq : intensity_mod ] ) and the fitted parameters .
the ellipticities @xmath248 are found using eq.([eq : epsilon ] ) by interpolating the intensities at @xmath276 and @xmath314 .
the ellipticities are plotted in fig .
[ fig : case1 ] a. the uncertainties of the fitted parameters originate from the uncertainties in the data , which are estimated to be @xmath287 1% for the amplitude and 1.5 degrees for the phase in the measured peak integrals .
these uncertainties are propagated numerically to the ellipticities as plotted in fig .
[ fig : case1 ] a. ideal case predictions and simulation results are also plotted in fig .
[ fig : case1 ] a. the simulation takes into account the major imperfection in the pulses and will be described later .
+ * error correction * the effectiveness of coding to correct errors is evident when comparing the ellipticities from the coding and the control experiments ( fig .
[ fig : case1 ] a ) . without coding ,
the ellipticity grows exponentially as @xmath315 ( for @xmath316 fitted to be @xmath317 ) . with coding ,
the growth is slowed down , with almost zero growth for small @xmath206 .
the suppression of linear growth of the ellipticity can be further quantified by weighted quadratic fits @xmath318 to the ellipticities . for the control experiments , @xmath319 , @xmath320 and @xmath321
whereas for the the coding experiments , @xmath322 , @xmath323 and @xmath324 ( @xmath206 in seconds ) .
therefore , the linear term `` vanishes '' due to coding .
the small negative coefficient for the linear term originates from the scattering of the data point at zero storage time . to quantify the `` cost of the noisy gates '' caused by the imperfect pulses
, we compare the ellipticities from the coding experiments and from the ideal case , the quadratic fits of which are respectively @xmath325 and @xmath326 .
the imperfections cause the ellipticity to increase by 0.1 at @xmath327 and this extra distortion _ decreases _ with @xmath206 .
we take advantage of the fact that the simulation results are close to the data points but are not as scattered to have a better estimate of this `` cost of the noisy gates '' .
the simulation data can be fitted by @xmath328 .
compared to the ideal case , the coding operations increase the ellipticity by @xmath287 0.06 at @xmath329 , and this extra distortion remains almost constant for all @xmath206 .
the error probabilities as inferred from the ellipticities @xmath330 are plotted in fig .
[ fig : case1 ] b as a function of storage time .
error correction is also manifested by expressing @xmath331 in the coding experiment as a function of the original @xmath331 in the control experiment , as plotted in fig .
[ fig : eorder ] . the quadratic fit to the experimental results gives @xmath332 where @xmath28 stands for @xmath331 in the control experiment , @xmath333 , @xmath334 and @xmath335 .
therefore , the expected improvement @xmath336 is confirmed .
experimentally , the error probabilities are larger than in the ideal case by at most 4.7% and these extra errors decrease with @xmath28 .
the quadratic fit to the simulation results ( which is a good approximation of the experimental data ) gives @xmath337 and differs from the ideal case by a constant amount of @xmath338 for all @xmath28 , which represents the cost of the noisy gates . in conclusion , coding with noisy gates is still effective in our experiments , even though the noisy gates add a constant amount of distortion .
while the data exhibit a clear correction effect , there are notable deviations from the ideal case .
first , the ellipses with coding are smaller than their counterparts without coding .
this is most obvious when the storage time is zero , in which case the coding and the control experiments should produce equal outputs .
second , the signal strength is attenuated with increasing @xmath3 relative to ideal ellipses .
third , although the data points are well fitted by ellipses , their angular positions are not exactly as expected ( `` @xmath3-offsets '' ) .
finally , the spacings between the ellipses deviate from expectation .
the causes of these discrepancies and their implications on error correction are discussed next .
+ * gate imperfections : rf field inhomogeneity * the major cause of experimental errors is rf field inhomogeneity , which causes gate imperfections .
this was determined by a series of experiments ( details of which are not given here ) , and a thorough numerical simulation , as described below .
the physical origin of the problem is as follows .
the coil windings produce inhomogeneous rf fields that randomize the angles of rotation among molecules . for a single rotation ,
the signal averaged over the ensemble decreases exponentially with the pulse duration to good approximation .
a measure of the rf field inhomogeneity is the signal strength after a @xmath99 pulse .
they are measured to be @xmath287 0.96 and 0.92 for proton and carbon respectively .
in other words , a single @xmath99 pulse has an error of @xmath287 4 - 8% .
rf field inhomogeneity affects our experiments in many ways .
first , it attenuates the signal in both the coding and the control experiments , but the effects are much more severe in the coding experiments which have eight extra pulses .
for instance , when the storage time is zero , the two experiments should have identical outputs , but the ellipse in the coding experiment is actually 5 - 15 % smaller .
second , for each ellipse , attenuation increases with @xmath3 as the preparation pulse @xmath188 becomes longer .
the effects of the rf field inhomogeneity are complicated , because the errors from different rf pulses are correlated , and the correlation depends on the temporal separation between the pulses and the diffusion rate of the molecules .
the correlation time of the rf field inhomogeneity is comparable to the experimental time scales .
for this reason , predictions of the effects of rf field inhomogeneity are analytically intractable .
numerical simulations were performed to model the dominant effects of rf field inhomogeneity .
we followed the evolution of the states assuming random rf field strengths drawn from lorentzian distributions ( also known as cauchy distributions ) with means and standard deviations matching pulse calibration and attenuation for the @xmath99 pulses .
all parameters in the simulation , including @xmath285 s , were determined experimentally without introducing any free parameters .
as the exact time correlation function for the errors was unknown , except for numerical evidence of a long correlation time , we _ assumed _ perfect correlation in the errors .
the simulated ensemble output signal was obtained by gaussian integration with numerical errors bounded to below 1.5% .
the results were shown in fig .
[ fig : simellips ] . besides phase damping and error correction effects in the data ,
the simulations also reproduce extra signal attenuation in the coding experiments .
the ellipticities obtained from simulations ( see fig . [
fig : case1 ] ) approximate the experimental values very well .
such agreement to experimental results is surprising in the absence of free parameters in the model .
simulation results allow the discrepancy between the observed and the ideal ellipticities to be explained in terms of the rf field inhomogeneity and allow the `` cost of coding '' to be better estimated to be the constant 6 % increase in ellipticity or the @xmath287 3 % increase in error probabilities .
the simulation results also predict increasing attenuation with @xmath3 . from the fitted parameter @xmath339 ( see eq.([eq : intensity_mod ] ) ) , the amplitudes at @xmath340 are @xmath287 4% _ weaker _ than the corresponding values at @xmath341 in the simulations .
experimentally , this attenuation increases from @xmath287 8 to 15% ( as storage time increases from 0 to 308 ms ) in the control experiments , and remains @xmath287 8% in the coding experiments .
therefore , rf field inhomogeneity contributes to the attenuation but only partially .
we conclude that rf field inhomogeneity as we have modeled explains the diminished signal strength in the coding experiments .
the simulation quantifies the `` cost of the noisy gates '' .
rf field inhomogeneity also explains part of the attenuation with increasing @xmath3 .
we can also conclude that other discrepancies not predicted by the simulations are _ not _ caused by rf field inhomogeneity and these discrepancies are described next .
+ * other discrepancies * the simulation results show that rf field inhomogeneity does _ not _ explain why the attenuation at large @xmath3 increases with storage time without coding , and it does not explain the @xmath3-offsets along the ellipses and the unexpected spacings between them .
the increased attenuation with storage time at large @xmath3 can be caused by amplitude damping .
a precise description @xcite of amplitude damping during storage is out of scope of this paper and we consider only the dominant effect predicted by a simple picture . _ phenomenologically _ , the loss of energy to the lattice is described in the nmr literature by @xmath342 where @xmath343 is the thermal equilibrium magnetization .
as @xmath344 at @xmath276 and @xmath123 , we expect no changes at @xmath276 , but expect @xmath345 at @xmath340 to decrease by @xmath346 % for @xmath347 ms and @xmath348 s for proton .
note that refocusing does not affect spin @xmath35 in the control experiments @xcite but it swaps @xmath349 and @xmath350 halfway during storage , symmetrizing the amplitude damping effects in the coding experiments .
therefore , we expect increased attenuation with storage time in the control experiments only .
this matches our observations that @xmath351 decreases from 8 to 16 % in the control experiments , and remains 8% in the coding experiments
. moreover , earlier data taken without refocusing ( not presented ) have the same trend of increased attenuation with storage time in both the coding and the control experiments .
these are all in accord with the hypothesis that amplitude damping is causing the observed effect .
the second unexplained discrepancy is that the output states span more than a semi - ellipse in the coding experiment but slightly less than a semi - ellipse in the control experiment .
we are not aware of any quantum processes that can be a cause of it .
it is notable that the output states and the fitted ellipticities can be used to infer the initial values of @xmath3 , and they are roughly proportional to the expected values for each ellipse .
the proportionality constants are 5 - 8 % higher than unity in the coding experiment , and 0 - 1.6% lower in the control experiments .
moreover , similar effects are observed in many other experimental runs .
therefore , this is likely to be a systematic error .
we have no convincing explanation for the anomalous spacings between the ellipses in the experiment .
however , from the fact that all the data points belonging to the same ellipse are well fitted by it , the anomalous spacings are unlikely to be caused by random fluctuations on the time scales of each ellipse - experiment .
the effect of the anomalous spacings is reflected in the scattering of the ellipticities of the data , and the large uncertainties in the quadratic fits . while it is impossible to eliminate or to fully explain these imperfections , it is possible to show that the deviations can not affect the conclusion that error correction is effective .
+ * effects of the discrepancies * we now consider the effects of the discrepancies on the ellipticities and the inferred fidelities in the experiments .
first of all , radial attenuation of the signal due to rf field inhomogeneity does not affect the ellipticities nor the inferred fidelities ( taken as conditional fidelities ) .
second , different expressions for the `` ellipticity '' are not equivalent when the output states do not form perfect ellipses .
however , they differ by no more than 7 and 3 % in the control and the coding experiments .
@xmath3-offsets along the ellipses are not reflected in the ellipticities , resulting in overestimated inferred fidelities .
this is bounded by 3% .
the scattering of the ellipticities due to anomalous spacings between the ellipses is averaged out with curve - fits to the data .
the most crucial point is , none of these effects have a dependence on the storage time that can be mistaken as error correction .
therefore , the effects of error correction can still be confirmed in the presence of all these small discrepancies .
the two previous subsections dealing with the ellipticities provide an analysis of the global performance of the code .
a stricter analysis is provided in this section using the overlap fidelities given by eq.([eq : expfid ] ) in section [ sec:2bitnmr ] .
the minimal overlap reflects all defects and deviations hidden in the ellipticities as well as other distortions such as that caused by amplitude damping .
all measurements are normalized using the amplitudes at @xmath269 as described in section [ sec:2bitnmr ] . in the control experiments ,
the output states at @xmath269 are least affected by amplitude damping and rf field inhomogeneity .
therefore , the normalization can be done accurately . in the coding experiment
, the signal attenuation at @xmath276 due to rf field inhomogeneity can lead to overestimated fidelities .
we determine the uncertainties due to rf field inhomogeneity by the following method . for each storage time ,
the amplitudes of the accepted and the rejected states at @xmath276 are summed .
the sum is compared with the corresponding amplitude at @xmath276 in the control experiment to estimate the attenuation due to rf field inhomogeneity .
the effects on the overlap fidelities are bounded to below 2% .
the errors in the measured peak integrals are propagated to the fidelities which result in standard deviations no more than 0.7% .
we apply similar procedures to the simulation results .
the net error probabilities , given by @xmath352 for the control and @xmath353 for the coding experiments , are plotted in fig .
[ fig : rawfidelity ] .
the large difference in the rates of growth of error probabilities confirms the effectiveness of coding even when a stricter measure of fidelity is used .
we have demonstrated experimentally , in a bulk nmr system , that using a two bit phase damping detection code , the distortion of the accepted output states can be largely removed .
these experimental results also provided quantitative measures of the major imperfections in the system .
the principle source of errors , rf field inhomogeneity , was studied and a numerical simulation was developed to model our data . despite the imperfections , a net amount of error correction was observed , when comparing cases with and without coding , and including gate errors in both cases .
our analysis also addresses several theoretical questions in quantum error correction in bulk samples such as the fidelity measures of deviation density matrices . in the following ,
we conclude with some observations regarding quantum error correction in bulk systems , including syndrome measurements , the equivalence between error correction and logical labelling @xcite , the applicability and advantages of detection codes and some issues in signal strength .
projective syndrome measurements traditionally employed in the standard theory of quantum error correction are impossible in ensemble quantum computation .
measurements via the acquisition of the fid do not reduce individual quantum states and provide only `` average syndromes '' .
moreover , the quantum states are destroyed after acquisition . however , the important point is that syndrome measurement is _ not necessary _ in error correction @xcite . in each molecule ,
the syndrome bits carry the error syndrome for that particular molecule after decoding .
these bits can either be used in a controlled - operation to correct the error @xcite , or in the case of a detection code , to `` logically label '' the correct and erroneous states .
conversely , logical labelling to obtain effective pure states can be considered as error detection : unsuitable initial states are `` detected '' and are labelled as `` bad '' to start with .
both processes involve ejecting the entropy of the system to the ancilla bits .
error detection and logical labelling are therefore similar concepts .
the distinction between correction and detection codes is blurred when using bulk systems .
the objective of error correction is to achieve reliable data transmission or information processing with high probability of success .
when information is encoded in single systems , encoding with an error detection scheme will fail to provide reliable output for two different reasons : accepting an erroneous state or losing the state upon the detection of an error .
therefore , coding schemes capable of _ distinguishing and correcting _ errors are necessary to improve the probability of successful data processing .
in contrast , in a bulk sample , a large initial redundancy exists upon preparation , and the combined signal of all the accepted cases forms the output .
therefore , rejection of the erroneous cases results in a reduction of the signal strength in the improved accepted cases without necessarily causing a failure .
detection codes thus provide a tradeoff between probability of error - free computation and signal strength .
as suggested by this analysis , and in concert with our experimental results , it thus makes sense to use detection codes instead of correction codes in bulk quantum computing systems under certain circumstances .
fundamentally , it is valuable to be able to interchange resources depending on their relative costs , this is illustrated by the following simple example .
suppose a total pool of @xmath354 qubits is available for transmission , and one just wants to correct for single phase flip errors of probability @xmath28 . using a three bit code , one would obtain an aggregate signal strength of @xmath355 , with fidelity @xmath356 , whereas with a two bit code , the accepted signal strength would be @xmath357 , with fidelity @xmath358 . therefore , when @xmath359 , the two - qubit code performs better in this model due to its higher rate .
another example relevant to bulk computation arises when the encoding and decoding circuits fail with probability proportional to the number of elementary gates used .
although errors in consecutive gates can be made to cancel sometimes , this basic scenario is substantiated by our experiment , in which imperfect pulses contribute significantly to the net error .
assume now that we have @xmath360 molecules , which are either two or three qubit systems .
let us compare the performance of the two and three bit codes , based on the strength of the correct output signal . because the correction code requires at least three times as many operations as the detection code @xcite , the figures of merit obtained for the two schemes are @xmath361 and @xmath362 respectively , where @xmath363 is the gate failure probability . in this model , the detection code performs better for @xmath364 due to the simplicity of the coding operations .
a third example is the case of current state nmr quantum computation at room temperature , in which the intrinsic signal strength decreases exponentially with the number of qubits@xcite . in this model ,
the initial signal strength of an effective pure state of @xmath354 qubits is approximately of order @xmath365 , and thus , for an ensemble of @xmath360 molecules , the signal strengths of the outputs from the correction and detection codes are about @xmath366 and @xmath367 respectively .
according to this measure of performance , the detection code outperforms the correction code for @xmath368 ( @xmath369 in our experiments ) .
if signal strength indeed decreases exponentially with @xmath354 , then some interesting generalizations can be made .
for arbitrary qubit errors , a @xmath12-error detection code has distance @xmath370 , while a @xmath12-error correction code has distance @xmath371 @xcite .
if one encodes @xmath80 bits in @xmath372 , the extra number of qubits used , @xmath373 , satisfies the singleton bound @xcite , @xmath374 .
therefore , the output signal strengths for detection and correction codes would be approximately proportional to @xmath375 and @xmath376 , where @xmath377 is a polynomial .
the detection code is thus always better asymptotically in this model@xcite .
this work illustrates how a careful study of dynamics in bulk quantum systems can provide a valuable opportunity to demonstrate and test theories of quantum information and computation .
the development of temporal , spatial , logical , and related labeling techniques opens a window allowing information about the dynamics of single quantum systems to be extracted from bulk systems .
furthermore , by systematically developing an experimental toolbox of quantum circuits and quantum error correction and detection codes , experiments which test multiple particle quantum behavior become increasingly accessible . with improvements in the initial polarization in the systems and new labeling algorithms which do not incur exponential signal loss @xcite , and with better methods to control the major source of error , the rf field inhomogeneity
, we believe that further study of bulk quantum systems will complement the study of single quantum systems , provide new insights into the dynamical behavior of open quantum systems , and further the potential for quantum information processing .
this work was supported by the darpa ultrascale program under contract daag55 - 97 - 1 - 0341 .
d.l . acknowledges support of an ibm graduate fellowship . l.v .
acknowledges a yansouni family fellowship under the stanford graduate fellowship program .
l.v . , x.z . and d.l .
are indebted to prof .
james harris and prof .
yoshihisa yamamoto for their patience and support .
we thank thorsten hesjedal , lois durham and jody puglisi for experimental assistance , nabil amer , charles bennett , hoi fung chau , hoi - kwong lo and shigeki takeuchi for enjoyable and helpful discussions , and rolf landauer for his guiding wisdom .
10 i. chuang , n. gershenfeld , m. kubinec , phys .
lett . * 80 * , 3408 ( 1998 ) .
i. chuang , l. vandersypen , x. zhou , d. leung , s. lloyd , nature * 393 * , 143 ( 1998 ) .
j. jones , m. mosca , r. hansen , nature * 393 * , 344 ( 1998 ) .
j. jones , m. mosca , j. chem .
* 109 * , 1648 ( 1998 ) .
j. jones , m. mosca , e - print quant - ph/9808056 , ( 1998 ) .
n. gershenfeld and i. chuang , science * 275 * , 350 ( 1997 ) .
d. cory , a. fahmy and t. havel , proc .
usa * 94 * , 1634 - 1639 ( 1997 ) .
d. cory , m. price , t. havel , physica d , * ( 120 ) 1 - 2 * p. 82 - 101
( 1998 ) .
i. chuang , n. gershenfeld , m. kubinec , and d. leung , proc .
a ( 1998 ) 454 , 447 - 467 .
p. shor , phys .
rev a. * 52 * , 2493 ( 1995 ) .
a. steane , phys . rev
* 77 * ( 1996 ) .
p. shor , proc .
37th annual symposium foundations of computer science p. 56 - 65
d. aharonov and m. ben - or , proc .
29th annual acm symposium on theory of computing , p. 176 - 188
j. preskill , proc .
a ( 1998 ) 454 , 385 - 410 .
a. kitaev , proc .
3rd international conference on quantum communication and measurement ( 1996 ) .
e. knill , r. laflamme and w. zurek , proc .
a ( 1998 ) 454 , 365 - 84 .
d. gottesman , phys .
a. * 57 * , 127 ( 1998 ) . i. l. chuang and r. laflamme , e - print quant - ph/9511030 , ( 1995 ) .
d. cory , m. price , w. mass , e. knill , r. laflamme , w. zurek , t. havel and s. somaroo , phys .
lett . * 81 * 2152 ( 1998 ) .
k. kraus , annals of physics , * 64 * , 311 , ( 1970 ) .
b. schumacher , phys .
a. * 54 * , 2614 ( 1996 ) .
a. ekert and c. macchiavello , phys .
lett . * 77 * , 2585 ( 1996 ) .
e. knill and r. laflamme , phys .
a. * 55 * , 900 ( 1997 ) . c. h. bennett , d. p. divincenzo , j. a. smolin , and w. k. wootters , phys
a. * 54 * , 3824 ( 1996 ) .
m. a. nielsen , h. barnum , c. m. caves and b. w. schumacher , proc .
a ( 1998 ) 454 , 277 - 304 .
d. gottesman , ph.d thesis , california institute of technology ( 1997 ) c. slichter , _ principles of magnetic resonance _
( springer , berlin , 1990 ) .
a. abragam , _ the principles of nuclear magnetism _
( oxford university press , oxford , 1961 ) .
d. divincenzo , phys .
a. * 51 * , 1015 , ( 1995 ) .
a. barenco , proc .
a ( 1995 ) 449 , 679 - 83 .
d. deutsch , a. barenco and a. ekert , proc .
a ( 1995 ) 449 , 669 - 77 .
a. barenco , c. bennett , r. cleve , d. divincenzo , n. margolus , p. shor , t. sleater , j. smolin and h. weinfurter , phys .
a. * 52 * , 3457 , ( 1995 ) .
m. a. nielsen , private communication .
e. knill , i. chuang and r. laflamme , phys .
a. * 57 * , 3348 ( 1998 ) .
r. r. ernst , g. bodenhausen , and a. wokaun , _ principles of nuclear magnetic resonance in one and two dimensions _ ( oxford university press , oxford , 1994 ) . formic-13c acid and calcium chloride were supplied by aldrich ( respective catolog nos .
27,941 - 2 and 23,922 - 4 ) .
chemicals were used without further purification .
we note that @xmath92-coupling , phase damping and amplitude damping are non - commuting and simultaneous processes . moreover , the effects of generalized amplitude damping at finite temperature on deviation density matrices are complicated .
the refocusing pulses were applied to spin @xmath36 , therefore spin @xmath35 was not affected directly .
refocusing affected the output indirectly by flipping spin @xmath36 for half of the storage time during which it was susceptible to amplitude damping .
a. peres , proc .
4th workshop on physics and computation , p. 275 - 77 ( 1996 ) .
two hadamard transformations and two controlled - nots are needed in the detection code , whereas six hadamard transformations , four controlled - nots and one toffoli gate are needed in the correction code . simulating the toffoli gate by three controlled - nots , the correction code still requires at least three times as many operations as required by the detection code . despite these observations ,
it must be kept in mind that the proper figure of merit depends on the application , and our conclusions do not imply advantages for detection codes , or even coding schemes in general .
for example , in the exponential signal loss model , each extra bit used in coding reduces the absolute signal strength by half . a similar observation
was made by cory _
. n. cerf and r. cleve , phys .
a , * 56 * , 1721 ( 1997 ) . a. calderbank , e. rains , p. shor , and n. sloane , ieee trans .
theory , vol.44 , no.4 , p. 1369 - 87 ( 1998 ) .
l. schulman and u. vazirani , proc .
31st annual acm symposium on theory of computing ( 1999 ) .
the chloroform sample was a 0.5 milliliter , 0.2 molar solution in d6-acetone supplied by cambridge isotope labs , inc .
( catolog no .
clm-262 ) .
experiments were performed at room temperature .
@xmath285 s for the input and the ancilla , as measured from the line widths , were 0.13 and 0.53 s when carbon was the input , and 0.92s and 0.16 s when proton was the input .
the storage times were 0 , 37.2 , 74.4 , 111.6 , 148.8 ms when carbon was the input , and 0 , 223 , 446 ms when proton was the input .
recall that the initial state after ancilla preparation is given by @xmath378 ( see eq.([eq : pureancilla ] ) , with @xmath95 omitted ) .
after @xmath188 , the new density matrix is given by @xmath379 without coding , phase damping changes the density matrix to @xmath380 \otimes ( i+\sigma_z)/2 \label{eq : controlout } \,.\ ] ] with coding , the encoding , phase damping , and decoding change the density matrix to @xmath381 , @xmath382 , and @xmath210 : @xmath383/2 \nonumber \\ & + & \sin\theta \sigma_x \otimes \left[\rule{0pt}{2.4ex } % ( 1-p_a - p_b ) ( i+\sigma_z ) + ( -p_a+p_b ) ( i-\sigma_z ) \right]/2 \\ & = & \left[\rule{0pt}{2.4ex } \cos\theta ( 1-p_a - p_b+2 p_a p_b ) \sigma_z % + \sin\theta ( 1-p_a - p_b ) \sigma_x \right ] \otimes ( i+\sigma_z)/2 \nonumber \\ & + & \left[\rule{0pt}{2.4ex } \cos\theta ( p_a+p_b-2 p_ap_b ) \sigma_z % + \sin\theta ( -p_a+p_b ) \sigma_x \right ] \otimes ( i-\sigma_z)/2 % \label{eq : codedout } \,.\end{aligned}\ ] ] * alternative understanding of the code * the operation of the code can be further understood using the above mixed state description . without coding , the input qubit is stored in the terms @xmath42 and @xmath384 which decay at different rates and distort the state asymmetrically . with coding , the input qubit is stored in the terms @xmath385 and @xmath386 which decay at the same overall rate to first order .
the resulting decoded state therefore shrinks radially up to first order . when subject to phase damping , the points on the bloch sphere move inwards transversely without coding , distorting the shape axisymmetrically ( fig .
[ fig : flows ] ) .
in contrast , the points move somewhat radially with coding , preserving the spherical shape better .
while the case of equal @xmath205 s is interesting from a theoretical standpoint , different spins in a molecule typically have quite different @xmath205 s . to study the two bit code in this regime ,
we performed experiments with carbon-13 labelled chloroform dissolved in acetone @xcite .
all parameters were similar to the sodium formate sample , except for the relaxation time constants . in the chloroform experiment ,
@xmath204 s were 16 s and 18.5 s and @xmath205 s were 7.5 s and 0.35 s for proton and carbon respectively .
separate experiments with the ancilla dephasing much slower or faster than the input were performed by interchanging the roles of proton and carbon .
@xmath285 s and @xmath206 s were as listed in @xcite .
the ellipticities are shown in fig .
[ fig : case23 ] . from fig .
[ fig : case23 ] a , it is apparent that coding almost removes the distortion entirely when a much better ancilla is available . the question is , is coding advantageous over storing in the good ancilla alone ? theoretically , coding is always advantageous because the error probability is always reduced from @xmath388 ( @xmath389 being the input spin ) to @xmath73 .
[ fig : case23 ] b shows that experimentally , such improvement is marginal , because the advantage of coding is offset by the noise introduced .
therefore , when the @xmath205 s are very different , the bottle neck is the dephasing of the bad qubit .
quantum state tomography @xcite is a procedure to reconstruct the density matrix given a certain set of measurements . in 2-spin nmr systems
, eight out of the fifteen coefficients , @xmath390 , in the pauli decomposition are obtainable from the peak integrals , eq.([eq : ahigh])-([eq : blow ] ) .
the remaining 7 parameters can be obtained by repeating the measurement process with additional readout pulses before acquisition .
these pulses permute the coefficients @xmath390 .
a series of 9 experiments with different readout pulses is sufficient to reconstruct the complete deviation density matrix .
we reconstructed the deviation density matrices in the coding experiments .
the results for @xmath391 and @xmath392 ms are shown in fig .
[ fig : timeseq ] .
the ideal matrices were calculated using equations derived in appendix [ sec : mixstate ] .
|
using nuclear magnetic resonance techniques , we experimentally investigated the effects of applying a two bit phase error detection code to preserve quantum information in nuclear spin systems .
input states were stored with and without coding , and the resulting output states were compared with the originals and with each other . the theoretically expected result , net reduction of distortion and conditional error probabilities to second order , was indeed observed , despite imperfect coding operations which increased the error probabilities by approximately 5% .
systematic study of the deviations from the ideal behavior provided quantitative measures of different sources of error , and good agreement was found with a numerical model .
theoretical questions in quantum error correction in bulk nuclear spin systems including fidelity measures , signal strength and syndrome measurements are discussed .
|
heavy quarkonia , both @xmath16 and @xmath17 bound states , have provided good laboratories for the study of the strong interaction @xcite . for the hadronic transitions between the heavy quarkonium states , yan @xcite characterized it as the emission of two soft gluons from the heavy quarks and the conversion of gluons into light hadrons .
based on this scheme , a series of decay rates , such as the e1-e1 hadronic transition , e1-m1 hadronic transition , m1-m1 hadronic transition have been calculated @xcite . it has been shown that the multipole expansion can make quite successful predictions for many hadronic transitions between the heavy quarkonia @xcite
. however , most of these studies are for the transitions among the @xmath18 states ; the hadronic transitions of @xmath19 states are seldom explored . using a sample of @xmath20 , cleo measured for the first time the transition rate of @xmath21-wave bottomonium @xmath22 @xcite , and the results are consistent with the theoretical predictions @xcite . for the hadronic transition of the @xmath21-wave charmonium states ,
there is only an upper limit of 2.2% at the 90% confidence level ( c.l . ) on the @xmath23 transition rate recently reported by the babar experiment @xcite .
the most promising process @xmath24 , which is dominated by an e1-m1 transition , is calculated in the multipole expansion formalism , and a transition rate @xmath25% is predicted @xcite . in this article , we search for @xmath0 with @xmath4 decays into @xmath5 and @xmath6 , where the @xmath26 is reconstructed in @xmath27 and @xmath28 in @xmath29 final states .
we also report the first measurement of the branching fractions of @xmath30 , @xmath13 , @xmath14 , and @xmath15 .
the data sample for this analysis consists of @xmath2 events produced at the peak of the @xmath3 resonance @xcite ; an additional 42 pb@xmath31 of data were collected at a center - of - mass energy of @xmath32=3.65 gev to determine non - resonant continuum background contributions .
the data are accumulated with the besiii detector operated at the bepcii @xmath33 collider . the besiii detector , described in detail in ref .
@xcite , has an effective geometrical acceptance of 93% of 4@xmath34 .
it contains a small cell helium - based main drift chamber ( mdc ) which provides momentum measurements of charged particles ; a time - of - flight system ( tof ) based on plastic scintillator which helps to identify charged particles ; an electromagnetic calorimeter ( emc ) made of csi ( tl ) crystals which is used to measure the energies of photons and provide trigger signals ; and a muon system ( muc ) made of resistive plate chambers ( rpc ) .
the resolution of the charged particles is @xmath35% at @xmath36 in a 1 tesla magnetic field . the energy loss ( @xmath37 ) measurement provided by the mdc has a resolution better than 6% for electrons from bhabha scattering
. the photon energy resolution can reach @xmath38% ( @xmath39% ) at @xmath40 in the barrel ( endcaps ) of the emc .
and the time resolution of tof is @xmath41 ps in the barrel and @xmath42 ps in the endcaps .
monte carlo ( mc ) simulated events are used to determine the detection efficiency , optimize the selection criteria , and study the possible backgrounds .
the simulation of the besiii detector is geant4 @xcite based , where the interactions of the particles with the detector material are simulated .
the @xmath3 resonance is produced with kkmc @xcite , while the subsequent decays are generated with evtgen @xcite .
the study of the background is based on a sample of @xmath43 @xmath3 inclusive decays which are generated with known branching fractions taken from the particle data group ( pdg ) @xcite , or with lundcharm @xcite for the unmeasured decays .
a charged track should have good quality in the track fitting and be within the angle coverage of the mdc , @xmath44 . a good charged track ( excludes those from @xmath26 decays )
is required to be within 1 cm of the @xmath33 annihilation interaction point ( ip ) transverse to the beam line and within 10 cm of the ip along the beam axis .
charged - particle identification ( pid ) is based on combining the @xmath37 and tof information in the variable @xmath45 .
the values @xmath46 and the corresponding confidence levels @xmath47 are calculated for each charged track for each particle hypothesis @xmath48 ( pion , kaon , or proton ) .
photons are reconstructed from isolated showers in the emc which are at least 20 degrees away from any of the charged tracks . in order to improve the reconstruction efficiency and the energy resolution ,
the energy deposited in the nearby tof counter is included .
photon candidates are required to have the energy greater than 25 @xmath49 in the emc barrel region ( @xmath50 ) , while in the emc endcap region ( @xmath51 ) , the energy threshold requirement is increased to 50 @xmath49 .
emc timing requirements are used to suppress noise and energy deposits unrelated to the event .
@xmath26 candidates are reconstructed from secondary vertex fits to all the charged - track pairs in an event ( assume the tracks to be @xmath34 ) .
the combination with the best fit quality is kept , and the @xmath26 candidate must have an invariant mass within @xmath52 of the @xmath26 nominal mass and the secondary vertex be at least 0.5 cm away from the ip .
the reconstructed @xmath26 information is used as input for the subsequent kinematic fit .
the @xmath28 candidates are reconstructed from pairs of photons with an invariant mass in the range @xmath53 . in selecting @xmath54 , a candidate event should have at least six charged tracks and at least one good photon .
after @xmath26 selection , the event should have exactly four additional good charged tracks with zero net charge . while in selecting @xmath55 , a candidate event should have four good charged tracks with zero net charge and at least three good photons .
the @xmath56 ( @xmath57 ) candidate is then subjected to a four - constraint ( 4c ) kinematic fit to reduce background and improve the mass resolution .
determination of the species of the final state particles and selection of the best photons when additional photons ( and @xmath28 candidates ) are found in an event are achieved by selecting the combination with the minimum value of @xmath58 , where @xmath59 is the chi - square from the 4c kinematic fit .
events with @xmath60 are kept as @xmath56 ( @xmath57 ) candidates .
there is substantial background from @xmath61 decays .
the @xmath62 events are removed by requiring the recoil mass of any @xmath27 pair to be outside of a @xmath63 window around the @xmath64 nominal mass , where @xmath65 is the resolution of the @xmath27 recoil mass .
the @xmath66 , @xmath67 events are rejected if @xmath68 . to reject background with one more photon than the signal events ( @xmath69 or @xmath70 )
, @xmath71 is required when a 4c kinematic fit to @xmath69 or @xmath70 is applied to the event ; and to suppress @xmath72 background , @xmath73 or @xmath74 is required .
after the above selection , the invariant mass distributions of the hadron system are shown in fig . [
fig : invariant mass kkpipipi ] , and clear @xmath7 signals are observed with very low background level in the two modes . using the inclusive mc events sample ,
the potential backgrounds from the @xmath3 decays which may contaminate @xmath75 ( @xmath76 ) are estimated .
events from @xmath77 , @xmath78 and @xmath79 ( @xmath77 , @xmath80 ) may peak at the signal region , and the contributions from these peaking backgrounds are estimated using the detection efficiencies determined from the mc simulations and the corresponding branching fractions from previous measurements @xcite , and then subtracted .
these background events after final event selection are listed in table [ table : peaking bkg ksk3pi ] and [ table : peaking bkg 2k2pipi0 ] ; the errors in the numbers of events are from the uncertainties of the detection efficiency and branching fractions .
the other backgrounds are composed of dozens of decay modes and smoothly distributed in the full mass region ( @xmath81 ) ; this kind of background is described by a second - order chebyshev function . the besiii collaboration thanks the staff of bepcii and the computing center for their hard efforts .
this work is supported in part by the ministry of science and technology of china under contract no .
2009cb825200 ; national natural science foundation of china ( nsfc ) under contracts nos .
10625524 , 10821063 , 10825524 , 10835001 , 10935007 , 11125525 ; joint funds of the national natural science foundation of china under contracts nos . 11079008 , 11179007 ; the chinese academy of sciences ( cas ) large - scale scientific facility program ; cas under contracts nos .
kjcx2-yw - n29 , kjcx2-yw - n45 ; 100 talents program of cas ; istituto nazionale di fisica nucleare , italy ; ministry of development of turkey under contract no .
dpt2006k-120470 ; u. s. department of energy under contracts nos .
de - fg02 - 04er41291 , de - fg02 - 91er40682 , de - fg02 - 94er40823 ; u.s .
national science foundation ; university of groningen ( rug ) ; the helmholtzzentrum fuer schwerionenforschung gmbh ( gsi ) , darmstadt ; and wcu program of national research foundation of korea under contract no .
r32 - 2008 - 000 - 10155 - 0 .
q. liu and y .- p .
kuang , phys .
d * 75 * , 054019 ( 2007 ) .
the predicted branching fraction is recalculated with the parameters in the model determined using the updated expermential data : @xmath82 kev , @xmath83% , and @xmath84 .
the total width of @xmath3 and the branching fraction of @xmath85 are taken from pdg , while the @xmath86 is the combined result from the measurement in cleo and besiii collaboration .
|
hadronic transitions of @xmath0 ( @xmath1 , 1 , 2 ) are searched for using a sample of @xmath2 @xmath3 events collected with the besiii detector at the bepcii storage ring .
the @xmath4 is reconstructed with @xmath5 and @xmath6 final states .
no signals are observed in any of the three @xmath7 states in either @xmath4 decay mode . at the 90% confidence level ,
the upper limits are determined to be @xmath8 , @xmath9 , and @xmath10 . the upper limit of @xmath11 is lower than the existing theoretical prediction by almost an order of magnitude . the branching fractions of @xmath12 , @xmath13 , @xmath14 and @xmath15 ( @xmath1 , 1 , 2 ) are measured for the first time .
|
the humankind studies space weather for more than 4000 years starting from the first mentions of auroras in ancient chinese literature . the term `` space weather '' itself exists for almost a century .
the official definition adopted by cospar states that `` space weather describes the physical processes induced by solar activity that have impact on our terrestrial and space environment , on ground based and space technological systems , and on human activities and health . ''
the first part of this definition actually covers two spatial scales of space weather , because when we speak about space weather in space , e.g. in connection with spacecraft failures , we usually mean some local parameters of the environment , and when we speak about space weather on the earth , e.g. in connection with human health , we usually mean some integral characteristics like the geomagnetic indices
. since this article centers on the variations of the geomagnetic field , the latter meaning will be used .
the second part of this definition indicates practical manifestations of space weather .
the impact of the space weather on technological systems is generally accepted ( see @xcite ) due to a number of spectacular events like the superstorm of 1989 when canada s power grid was disabled for 9 hours and numerous spacecraft failures due to `` killer electrons '' causing arcing in electronic components , see @xcite .
the impact on human health , however , is disputed by most specialists .
nevertheless , the latest reports ( e.g. @xcite , @xcite ) indicate that there is indeed a strong correlation between the rate of sudden cardiac deaths and the space weather .
the space weather problem is twofold .
the first aspect is purely practical and aims for prediction and , eventually , mitigation of adverse effects of space weather .
ideally , this task should be accomplished by launching a vast number of spacecraft which will monitor the sun - earth region for large - scale structures like cmes .
unfortunately , the resources of the humankind are insufficient to produce and maintain such a large space fleet as well as to process all the data delivered by these spacecraft .
so , today we should use the resources at hand , which include a few solar wind spacecraft ( ace , wind , soho , and stereo ) , magnetospheric spacecraft ( cluster , themis ) , and ground - based stations ( intermagnet , magdas , etc . ) , to develop forecast techniques that will be used in future .
thus , we should try to predict space weather with what data we have , and we should aim for possibly longer prediction times to allow for some kind of countermeasures .
the second aspect is mostly academic and involves study of the processes in the near - earth space and , specifically , understanding of interaction between the solar wind and the magnetosphere .
naturally , improving our knowledge of the underlying physics significantly improves predictive capabilities , so fulfilling the second task will significantly help with the first one .
modern conceptions of solar wind - magnetosphere interaction are mostly based on phenomenological models constructed in 1960 s .
however , there are numerous problems these models can not answer .
this is largely due to the fact that these models were developed at the very beginning of the space era when data quality and quantity were immeasurably worse than today . for more than 40 years we collected astonishing amounts of data about solar wind parameters and geomagnetic activity and
now it is time to put them to good use .
space weather prediction is a challenging and nontrivial activity , see @xcite .
the most straightforward approach to space weather prediction is studying the whole complex chain of physical processes involved in magnetospheric dynamics and conjugating them in a global model of the evolution of the magnetosphere under the influence of the solar wind .
unfortunately , this is not yet possible due to our poor understanding of the physics of the interaction between the solar wind and the magnetosphere .
for this reason , different approaches should be tried . according to @xcite ,
today there are several established methods of space weather prediction , listed below .
morphological analysis of solar images .
this method provides the longest prediction time ( up to a week ) .
its accuracy is unknown since it is used for the academic purposes only .
today it is purely manual and thus almost useless for practical implications .
detection of large - scale perturbations in the solar wind , see e.g. @xcite , @xcite .
this method provides a very good prediction time ( up to several days ) , but is capable of predicting less than 10% of the most intense storms .
while it is very inaccurate when used alone , it can prove to be useful in combination with one of the following short - term methods .
construction of empirical models , see e.g. @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
this method provides the shortest prediction time ( up to 1 hour ) with moderate accuracy ( @xmath1 ) .
potentially this method could demonstrate far better results if the physics behind the magnetic storms was less of a mystery .
\4 . numerical modeling , see e.g. @xcite , @xcite .
this method provides a good prediction time ( up to several days ) but its accuracy varies in huge limits .
the accuracy of these methods is limited by their inability to correctly describe plasma instabilities .
besides , the ring current can not be described in the framework of ideal mhd , which forms the basis of most numerical models .
however , they can adequately describe the motion of e.g. magnetic clouds in the interplanetary environment , but rely on different methods to detect them . of the corresponding term in equation ( [ eqn:1 ] ) on the time offset for the @xmath2 autocorrelation model ] \5
. multidimensional time series analysis .
this method provides a moderate prediction time ( up to several hours ) with the highest accuracy ( @xmath3 ) .
they are very effective and easy to use but strongly depend on satellite data availability .
these are `` black box '' or `` input - output '' models , which seek only to reproduce the system s output in response to changes of its inputs .
the model terms are usually physically interpretable and thus useful for construction of new phenomenological models .
for this reason , this method can not only provide space weather forecast per se , but also can improve our knowledge of the physics involved and thus increase the efficiency of other methods .
further we shall speak about the last method , keeping in mind that its results can be used later to assist other methods .
first of all , let us discuss its existing implementations .
multidimensional time series analysis can be performed using the methods of statistics , signal processing , informatics , fuzzy logic etc .
the most widely used variations are artificial neural networks ( e.g. @xcite , @xcite , @xcite , @xcite ) , optimization ( e.g. @xcite , @xcite , @xcite ) , and correlation analysis ( e.g. @xcite , @xcite , @xcite , @xcite , @xcite ) .
neural network approach provides short - term predictions up to 4 hours with the correlation coefficient of 0.79 in the paper by @xcite .
earlier implementations of this approach experienced significant difficulties predicting strong geomagnetic storms with @xmath4 , but this approach remains one of the most popular alongside the empirical methods .
optimization approach seems to be more successful being able to provide 8-hour predictions in the paper by @xcite .
however , in the papers based upon the optimization methods the volume of the dataset used is insufficient to correctly describe secular variations of geomagnetic activity .
correlation analysis gives interesting results , but it was used solely for developing and constraining empirical models ( see @xcite ) .
however , most of these methods have a common feature : they lead to a regression relationship at some point , so it seems natural to skip all the preliminary steps and instantly use the regression analysis without unnecessary multiplication of entities .
regression analysis itself was attempted earlier by @xcite , but it was used to estimate the probability of intense / super - intense storm occurence depending on the solar and interplanetary parameters .
@xcite was able to predict 2 of 4 super - intense and 5 of 5 intense cme driven storms during the 1996 - 2002 period using another 46 cme driven storms to train his model . hereafter we propose a new approach , named `` regression modeling '' , which already allows achieving accurate ( @xmath5 ) 6 hours ahead forecast of the dst index , which we will use as a quantitative characteristics of space weather .
this method can be easily extended to predict other geomagnetic indices like kp or ap .
the proposed method is statistical , but has some features of empirical models .
it is based upon the regression analysis and the mathematical statistics . in its framework
the predicted dst value is sought in the form @xmath6 where @xmath7 is the number of current step ( number of hours since jan 1 , 1963 ) , @xmath8 is the prediction length , @xmath9 are the regression coefficients , and @xmath10 are the regressors , which are functions and combinations of input quantities , which are already measured at the time when prediction is made .
values of @xmath9 are determined by least square method ( lsm ) over a large sample of solar wind and geomagnetic data ( see next chapter ) , with equal statistical weights of all points .
the statistical significance of the regressors was determined by fisher test ( f - test ) ( see @xcite , @xcite ) .
this test allows separating significant and insignificant regressors .
the insignificant parameters are then rejected and the routine is repeated until the regression contains only significant regressors . of course , this method does not guarantee that all the significant regressors will enter the regression , but physical considerations and brute force in the form of trial and error provide us with requested reliability .
the regressors @xmath10 are generally nonlinear , so from the control theory s point of view , this method is able to describe discrete dynamical systems with strong nonlinearity .
this is an essential feature of the regression modeling method .
there is only one manual operation in this method selection of regressors to be considered . for this purpose
all known models , basic physical considerations , and random choice are used .
naturally , common sense also counts : for example it would be silly to add imf components in gse and gsm coordinates at the same time .
if some regressors @xmath10 appeared to be statistically significant , we also checked the significance of products of their powers @xmath11 , where @xmath12 can be any real number , including zero , but for practical purposes we used integer values of @xmath12 in the range from 0 to 6 .
this yields a very important feature of the regression analysis : it allows checking the statistical significance of any regressor , which can be useful for verifying different physical hypotheses . in this sense
we will call a parameter `` geoeffective '' if it appears at least in one statistically significant regressor . more details of this method can be found in the article @xcite .
the @xcite database was used .
it contains imf , solar wind and geomagnetic data , averaged over 1-hour intervals ( 49 parameters in total , starting from jan 1 , 1963 ) .
it was supplemented with provisional dst data , taken from wdc for geomagnetism ( kyoto ) .
thus a continuous 44-year dst time series was obtained .
we estimated the geoeffectiveness of a parameter by coefficients and statistical significances of all regressors , which contain this parameter .
this was done in the following way .
after processing the data with the least square method , fisher significance parameter @xmath13 was determined for each regressor .
all the @xmath13 values were compared to the values 2.7055 , 3.84 , 5.02 , 6.635 , 7.879 , 10.83 and 12.1 , which correspond to statistical significance of 90 , 95 , 97.5 , 99 , 99.5 , 99.9 and 99.95% respectively .
then , insignificant regressors were rejected and the routine was repeated until all the regressors were significant .
the number of significant regressors depends on the selected significance threshold .
all results given herein correspond to the significance threshold of 90% .
in contrast to empirical models we do not add fitting parameters and all the regressors have physical meaning . the described routine was applied to the sample , obtained from the initial dataset after rejecting filled values .
this sample can be divided into two subsamples , corresponding to quiet ( @xmath14 ) and perturbed ( @xmath15 ) conditions .
first , we determined which previous dst values are statistically significant .
for this purpose , we constructed an autoregression ( see details in @xcite ) @xmath16 where @xmath17 is the `` age '' of the oldest dst value .
this model alone is not sufficient to correctly predict dst , but it sets a basis for the construction of models that are able to do so .
let us determine the maximum reasonable value of @xmath17 .
for this purpose , we plot the autocorrelation function ( acf ) of the dst index for @xmath18 ( see fig .
[ fig:1 ] ) .
one can see that acf tends to a sinusoid with a period close to half a year .
this is caused by seasonal variations .
this yields a question : if there were no temporal variations , what would acf tend to at large offsets ?
if the distribution of dst was normal , the answer would be zero .
however , the distribution is not normal , so acf can tend to some non - zero quantity . to determine this quantity we need to remove temporal variations .
for this purpose we need to calculate the acf of a random sample with the same statistical characteristics as the dst sample .
the easiest way to get such a sample is to process the dst sample with a permutation method , which is widely used for determination of correlation functions , e.g. in astronomy .
this method is based on random shuffle of the sample .
using this method many times ( 10000 times in our case ) and calculating the correlation coefficient each time , we get the distribution of the correlation coefficient by monte - carlo method .
the distribution of the correlation coefficient for this sample ( fig . [ fig:2 ] ) appeared to be very close to a normal distribution with mean 0.008 and variance @xmath19 .
the maximum recorded value in 10000 trials was equal to 0.015 .
the top and the mean values are depicted on fig .
[ fig:1 ] by horizontal lines . as one can see on fig .
[ fig:1 ] , in reality the correlation coefficient exceeds this value at most times due to temporal variations .
the acf crosses the top line for the first time at @xmath20 hours , though the difference between the acf and the sine with a half - year period crosses it at @xmath21 hours , which is about 3.5 27-day periods , so we will assume the latter value as a rough estimation of @xmath17 .
this hints that rather old dst values can be quite significant . besides the half - a - year periodicity one can also notice the 27-day periodicity , caused by carrington rotation of the sun , which can be taken into account by adding the sunspot number @xmath22 to the regression .
let us return to equation ( [ eqn:2 ] ) . applying the f - test we can determine which previous dst values are statistically significant ( see fig .
[ fig:3 ] ) .
we did not search statistically significant dst values for @xmath23 , but it is possible that there are even older statistically significant values .
a similar situation was reported by @xcite regarding kp : `` the significance is often quite large for extended periods of time ( 10 - 20 days ) '' .
as our analysis shows , dst index can `` remember '' its previous values for significantly longer periods of time .
in fact , after adding regressors , corresponding to satellite data , some of the previous dst values become insignificant . we found that after the addition of these regressors there are still statistically significant values as far as 801 hours ago ( 33 days and 9 hours ) for @xmath18 . the statistical significance of this oldest value
is over 99.9% . at this point
we already have a large number of regressors , describing just the previous dst values ( autoregression ) , without satellite data and nonlinear terms .
if we add those , the number of regressors will only increase . after determining which previous dst values are statistically significant , we added all the solar wind parameters available in the omni 2 database .
then , we added nonlinear terms as discussed in section [ s : description ] . after adding a new regressor ,
all the significances are recalculated , and some of the old regressors can become insignificant . the total number of regressors is about 150 - 200 . since it is very large , we will not give here any lists of regressors or coefficients even for the simplest case @xmath18 , but the preliminary list is given in the paper @xcite .
in this section we will demonstrate how this method can be used for identification of geoeffective parameters . we will use four parameters as an example : doy ( day of the year ) , ut ( universal time ) and latitudinal and longitudinal flow angles of the solar wind . on fig .
[ fig:1 ] one could see a clear seasonal dependence of the dst index .
this dependence was described in many articles , for example by @xcite , @xcite , @xcite , and @xcite , but the reason behind it is still disputed .
most authors believe these asymmetries are caused by either of two cusps turning to the sunlit side due to annual rotation of the earth with respect to the sun .
however , @xcite state that this mechanism would give only 17% of observed asymmetry .
@xcite connected the diurnal variations of dst with an uneven distribution of dst network stations .
let us use this known effect to validate our method .
if we select two subsamples , corresponding to summer and winter in northern hemisphere , bounded by vernal and autumnal equinoctia , and verify the hypothesis that the difference between the corresponding average dst values is statistically significant using a one - sided student test , we obtain @xmath24 , which is well over 99.95% significant .
values of @xmath25 corresponding to 99 and 99.95% significance levels are equal to 2.334 and 3.31 respectively
. for diurnal asymmetry student test gives @xmath26 , which corresponds to significance level of more than 99.95% .
note that formally student test is applicable when dst is normally distributed .
in fact , the distribution of dst is slightly asymmetric , but taking into account the obtained values of @xmath25 , we can be sure in qualitative conclusions made .
[ fig:4 ] and [ fig:5 ] show the histograms of seasonal and diurnal variations of dst index .
taking this known geoeffective factor as an example we demonstrate how easily one can take it into account using regression approach .
to do so one should simply add terms @xmath27 and @xmath28 into the regression . here
@xmath7 is once again the number of hours since jan 1 , 1963 , 1920 is the number of hours between the beginning of the year and the vernal equinox , and 4383 is the number of hours in half a year .
the first of these terms is significant and describes summer / winter asymmetry , and the second one ( which appears statistically insignificant ) describes an absent spring / autumn asymmetry .
likewise , for diurnal asymmetry the corresponding terms will be @xmath29 and @xmath30 . here 2 is the time difference between ut and the northern geomagnetic pole , and 12 is the number of hours in half a day .
both these terms are significant .
the coefficient of the @xmath31 term is 30 times less than the observed difference between mean dst values of summer and winter subsamples .
this can be explained in the following way : there are other regressors , which depend on parameters with statistically significant summer / winter asymmetry , e.g. previous dst values .
they provide the lion share of summer / winter asymmetry of dst .
a good example of such a regressor is the sunspot number @xmath22 , which describes the 27-day periodicity .
nevertheless , there is a small difference which can not be expressed with these terms . including it into regression , we obtain these statistically significant regressors . to further illustrate this point , let us consider as an example a value @xmath32 . in the regression it will look like @xmath33 = x_n+a\left [ ( \cos\omega\delta t - 1 ) \sin\omega t + \cos\omega t\sin\omega\delta t\right]$ ] .
the first term in brackets is of order @xmath34 , and the second @xmath35 in the natural assumption that @xmath36 .
so , it will seem that the coefficient is @xmath37 rather than @xmath38 .
note that this is just an example and has nothing to do with actual regressors . however if we look on the distribution of mean dst values vs. time of the year ( fig .
[ fig:4 ] ) , we see a much more complicated pattern of seasonal variations of geomagnetic activity . among other features there is a strong asymmetry between summer and winter on one side and spring and autumn on the other . to take it into account we introduced additional terms into our regression , which are powers of @xmath31 and their products with powers of @xmath39 .
the sum of regressors with the corresponding coefficients , depicted on fig .
[ fig:6 ] , is very similar to fig .
[ fig:4 ] .
note that fig .
[ fig:6 ] was obtained independently from fig .
[ fig:4 ] .
we did the same thing with the diurnal asymmetry .
the distribution is plotted on fig .
[ fig:5 ] , and the sum of regressors with the corresponding coefficients on fig .
[ fig:7 ] .
the term @xmath40 is also significant and should be included in the regression .
after this we obtained a joint distribution of semiannual and diurnal variations of dst index , plotted on fig .
[ fig:8 ] .
it contains 18 regressors . increasing the number of regressors describing temporal variations of geomagnetic activity
we can improve the accuracy of this distribution .
in particular , one could add 11-year and 22-year solar cycles , higher powers of @xmath41 and @xmath42 etc .
thus , we demonstrated how easily one can take into account new geoeffective parameters in this method s framework .
now let us discuss parameters , whose geoeffectiveness was determined by this method , and demonstrate that they are indeed geoeffective .
latitudinal flow angle @xmath43 was mostly associated with the southern component of imf .
i plotted the distribution of its value and the corresponding mean dst value on fig .
[ fig:9 ] .
the distribution looks similar to a normal distribution , but it significantly differs from the normal one according to @xmath44 test .
this manifests in much larger number of points with deviations more than @xmath45 than follows from the normal distribution .
this is mostly caused by the number of points in the wing bins @xmath46 being equal to 196 points versus 11 points in the case of normal distribution .
however , most of these points were obtained in the 1960s , when quality of measurements was much worse then today .
this period includes the maximum and minimum values of @xmath43 , equal to @xmath47 and @xmath48 .
nevertheless , these points constitute only a minor fraction of all points and did nt affect the linear regression routine . assuming normal distribution we obtain @xmath49 and @xmath50 .
thus , the distribution is insignificantly shifted towards positive values .
if we ignore the wing bins in the distribution of mean dst values against @xmath43 , which are somewhat random due to small amount of points in them , we will notice a slight almost linear trend .
if we plot the sum of terms containing @xmath43 ( fig .
[ fig:10 ] ) , we will notice a similar trend . if we select two subsamples , one @xmath51 and other @xmath52 , and verify the hypothesis that the difference between the corresponding average dst values is statistically significant using a one - sided student test , we obtain @xmath53 , which is over 99.95% significant .
if we divide the sample in two subsamples : one for northern summer and one for northern winter , we obtain such a picture ( fig .
[ fig:11 ] ) . once again
, let us not look at the wing bins . what do we see ?
summer distribution has an obvious linear trend , but the winter one has not .
if we apply the student test to the same intervals now , we obtain that @xmath25 is 5.44 in the summer and only 0.059 in the winter .
the prior corresponds to more than 99.95% significance , while the latter to less than 10% .
this could mean that there are two factors connected with the latitudinal flow angle , which work together in the summer and against each other in the winter .
the physical explanation of this phenomenon , however , lies beyond the scope of this paper .
longitudinal flow angle @xmath54 was only occasionally used in models .
however , it appeared to be even more significant than the latitudinal flow angle .
its distribution together with corresponding mean dst values is plotted on fig .
[ fig:12 ] , where white bars show the complete dataset sans rejects , and the grayed bars show the quiet - time sample with @xmath14 . like the latitudinal flow angle ,
the distribution of the longitudinal flow angle resembles normal distribution . however , @xmath44 test disproves the relevant null - hypothesis . once again
, this is mostly due to wing bins which are mostly formed of data points , corresponding to measurements in 1960s , including maximum and minimum values equal to @xmath55 and @xmath56 . assuming normal distribution we obtain @xmath57 and @xmath58 .
a significant trend is the most prominent feature of this figure .
if we plot a sum of regressors , which contain @xmath54 ( fig . [ fig:13 ] ) , we see a very similar trend .
like before , we plot the distribution for summer and winter subsamples separately ( fig .
[ fig:14 ] ) .
we see that the trend is identical on both plots , so the corresponding effect is season - independent .
the list of regressors for @xmath18 , containing @xmath43 and @xmath54 , is given in table [ tbl:1 ] .
it contains the regressors themself , their coefficients and @xmath13 values .
thus , we demonstrated that our method is truly capable of pointing out new geoeffective parameters and verified the geoffectiveness of two such quantities .
rc@ cc @xmath59&@xmath10&@xmath9&@xmath60 + @xmath61 & @xmath62 & @xmath63 & @xmath64 + @xmath65 & @xmath66 & @xmath67 & @xmath68 + @xmath69 & @xmath70&@xmath71 & @xmath72 + @xmath73 & @xmath74 & @xmath75&@xmath76 + @xmath77 & @xmath78 & @xmath79 & @xmath80 + @xmath81 & @xmath82 & @xmath83&@xmath84 + @xmath85 & @xmath86 & @xmath87&@xmath88 + @xmath89 & @xmath90 & @xmath91 & @xmath92 + @xmath93 & @xmath94 & @xmath95&@xmath96 + @xmath97&@xmath98 & @xmath99&@xmath100 + @xmath101&@xmath102 & @xmath103&@xmath104 + rrccc @xmath8 , h&rms ,
nt&lc&pe¬e + @xmath61 & @xmath105 & @xmath106&@xmath107 & + @xmath61 & @xmath108 & @xmath109 & @xmath110&autoregression + @xmath61 & @xmath111 & @xmath112 & @xmath113&@xmath14 + @xmath61 & @xmath114 & @xmath115 & @xmath116&@xmath15 + @xmath69 & @xmath117 & @xmath118 & @xmath119 & + @xmath81 & @xmath120&@xmath121 & @xmath122 & + @xmath93 & @xmath123&@xmath124 & @xmath125 & + @xmath126&@xmath127&@xmath128 & @xmath129&for reference + @xmath130&@xmath131&@xmath132 & @xmath133&for reference + @xmath134&@xmath135&@xmath136 & @xmath137&for reference + taking into account the considered parameters together with parameters , whose geoeffectiveness was beyond doubts , like previous values of dst , dawn - dusk electric field , ram pressure of the solar wind and most of other parameters from the omni2 database , we constructed models for predicting dst 1 , 3 , 6 , 9 , 12 , 18 , and 24 hours ahead , and 3 more models for predicting dst 1 hour ahead for quiet and perturbed conditions and for the case when satellite data are unavailable ( autoregression , see more in @xcite ) .
the statistical characteristics of these models are summarized in table [ tbl:2 ] .
they include residual mean square ( rms ) , linear correlation coefficient ( lc ) , and prediction efficiency ( @xmath138 , where sd is the sample s standard deviation ) . in divided cells
the top number corresponds to the actual model , and the bottom one to the simplest possible model @xmath139 .
it is noteworthy that despites good correlation for all the models , in reality only the 1-hour and 3-hour models are ready for practical use , and the 6-hour model can potentially reach this state .
this is due to a significant time shift being present in further predicting models .
note that since the proposed method is statistical , there is little difference whether the `` training '' sample contains the period when prediction is made ( `` test '' sample ) or not . to further illustrate this point
let us consider an example : i predicted dst 3 hours ahead using the `` test '' sample from 2007 to 2008 and 3 `` training '' samples : the sample from 1976 to 2003 gives lc=0.851 , the sample from 1976 to 2006 gives lc=0.854 , and the sample from 1976 to 2008 gives lc=0.854 as well .
the first two `` training '' samples do not contain the `` test '' sample , but the results are slightly different .
the third `` training '' sample contains the `` test''sample , yet the result is the same as for the second sample .
this yields a conclusion that the volume and statistical properties of the `` training '' sample affect the prediction results stronger than the inclusion of the `` test '' sample .
so , the inclusion of the `` test '' sample to the `` training '' sample has little or no effect on the lc .
pe is calculated independently from the `` test '' sample and is not affected by its selection in any way .
we also present graphical representations of prediction in comparison with results of other authors : @xcite ( fig .
[ fig:15 ] , [ fig:16 ] ) , @xcite ( fig .
[ fig:17 ] ) , and @xcite ( fig . [ fig:18 ] ) .
note , that the intervals for prediction were selected by authors of the original papers , and the coefficients in eq .
( [ eqn:1 ] ) were the same all the time and for all the figures .
it is clearly visible that our method provides much more precise forecast than most empirical models and typical neural network models .
ridiculously , even our @xmath140 model is more precise than most empirical @xmath2 models .
the autoregression model , described by eq .
( [ eqn:2 ] ) , though , lags in the left part of the plot due to a rapid positive change of dst at 1500 ut .
the lag persists through the growth phase and the main phase , and vanishes only in the recovery phase .
for this reason , the autocorrelation model holds little practical value and should be considered as a transitional result , required to construct the full model .
it is , however , possible to improve it by adding terms describing temporal variations , and , for example , the number of sunspots , but then the term `` autoregression '' will no longer be applicable . on fig .
[ fig:19 ] we present the results of prediction 3 , 6 and 9 hours ahead for a number of events , kindly selected for us by v.g .
fainshtein , which are particularly hard to predict by medium - term methods , such as @xcite , to verify the efficiency of our method .
we can see that this method s accuracy is higher for stronger storms , which are of greater interest .
a huge advantage of this method is that the most resource - demanding operation the calculation of the regression coefficients should be performed only once for each model .
the prediction itself is just a summation of a polynom , which usually takes no more than 4 - 6 seconds on an average pc ( including disk i / o ) , which allows for creation of fully automated operational on - line space weather forecast services .
the proposed regression approach appeared to be more than adequate for space weather forecasting . for the forecasting per se , its main advantages are quite good correlation ( about 90% for 6 hours forecast ) , adaptability to any samples , and very fast forecasting code ( typically about 5 seconds on an average pc ) . for the identification of geoeffective parameters
it is extremely convenient and easy to use . in particular
, it allowed to uncover 2 new geoeffective parameters the latitudinal and the longitudinal flow angles of the solar wind .
this is just a short summary of the regression modeling method , since its full description would take much more space .
of course , this method can be used in conjunction with other methods , first of all , with physical methods of detection of large - scale perturbations in the solar wind and with empirical models .
the author would like to thank prof .
cheremnykh , prof .
yatsenko , and academician v.m .
kuntsevich for fruitful discussion , prof .
fainshtein for useful remarks and for providing a list of geomagnetic events for validation of the model , and reviewer # 1 for valuable comments which greatly improved this article .
the author is grateful to the space physics data facility ( spdf ) and the national space science data center ( nssdc ) for the free online omni2 catalogue and to the world data center for geomagnetism ( wdc - b ) at kyoto university for the free online catalogue of geomagnetic indices .
balikhin , m.a .
2001 , geophys .
, 28 , 1123 ballatore .
p. & gonzalez , w.d .
2003 , earth planets space , 55 , 427 burton , r.k .
1975 , j. geophys .
res . , 80 , 4204 campbell , w.h .
1996 , j. atm . & terr .
phys . , 58 , 1171 cerrato , y. et al .
, 2004 , in `` lecture notes and essays in astrophysics '' , pp.131 - 142 cid , c. et al . , 2005 , in solar wind 11 soho 16 workshop , `` connecting sun and heliosphere '' , pp.116 - 119 cliver , e.w .
2000 , j. geophys .
res . , 105 , 2413 eselevich , v.g . & fainshtein , v.g .
1993 , annales geophysicae , 11(8 ) , 678 eselevich , v.g .
et al . , 2009 , cosmic research , 47(2 ) , 95 fenrich , r.r .
& luhmann , j.g .
1998 , geophys .
lett . , 25 , 2999 fisher , r.a . , 1954 , `` statistical methods for research workers '' , oliver and boyd : london harrison , r.f . & drezet , p.m. , 2001 , in les woolliscroft memorial conf .
, `` multipoint measurements versus theory '' , pp.141 - 146 hudson , d.j . , 1964 , `` statistics lectures on elementary statistics and probability '' , cern : geneva johnson , j.r . & wing , s. 2004 , report pppl-3919rev .
+ http://www.pppl.gov/pub_report/2004/pppl-3919rev.pdf johnson , j.r .
& wing , s. 2005 , j. geophys .
res . , + doi:10.1029/2004ja010638 joselyn , j.a .
1995 , rev .
geophys . , 33 , 383 khabarova , o.v .
2007 , sun and geosphere , 2(1 ) , 32 khabarova , o.v . &
dimitrova , s. , 2008 , in international conference , `` fundamental space research '' , pp.306 - 309 kugblenu , s. et al . 1999 , earth planets space , 51 , 307 li , x. et al .
2003 , eos , 84 , 37 liu , y. et al .
2008 , preprint , arxiv:0802.2423v3 [ astro - ph ] lyatsky , w. et al .
2001 , geophys .
lett . , 28 , 2353 marubashi , k. 1989 , space sci .
, 51 , 197 mckenna - lawlor , s.m.p .
et al . 2008 , j. geopys .
, + doi:10.1029/2007ja012577 obrien , t.p . & mcpherron , r.l .
2000a , j. atm .
& sol .- terr .
phys . , 62 , 1295 obrien , t.p . & mcpherron , r.l .
2000b , j. geophys .
res . , 105 , 7707 obrien , t.p . & mcpherron , r.l .
2002 , j. geophys .
res . , + doi:10.1029/2002ja009435 oh , s.y .
& yi , y. 2004 , j. korean astron . soc . , 37 , 151 omni2 database .
national space science data center / space physics data facility .
+ http://nssdc.gsfc.nasa.gov/omniweb/ cited 04 jan 2009 pallocchia , g. et al .
2006 , mem .
, 9 , 120 parnowski , a.s .
2008 , kosmichna nauka i technologiya ( space science & technology ) , 14(3 ) , 48 parnowski , a.s . 2009a , j. autom .
, 41(3 ) , 128 parnowski , a.s . 2009b , earth planets space , 61(5 ) , 621 rangarajan , g.k .
& barreto , l.m .
1999 , earth planets space , 51 , 363 romanova .
2005 , kosmicheskie issledovaniya ( cosmic research ) , 43(3 ) , 186 siscoe , g. et al .
2005 , j. geophys .
res . , + doi:10.1029/2004ja010465 srivastava , m. 2005 , annales geophysicae , 23 , 2969 stoupel , e. et al .
2006 , sun and geosphere , 1(2 ) , 13 temerin , m. & li , x. 2002 , j. geophys .
res . , 107 , 1472 temerin , m. & li , x. 2006 , j. geophys . res .
, + doi:10.1029/2005ja011257 takalo , j. & mursula , k. 2001 , j. geophys .
res . , 106 , 10905 valdivia , j.a .
1996 , geophys .
lett . , 23 , 2899 watanabe , s. et al .
2002 , earth planets space , 54 , 1263 wei , h.l .
2004 , nonlinear processes in geophysics , 11 , 303 wing , s. et al .
2005 , j. geophys .
res . , + doi:10.1029/2004ja010500 zhou , x .- y .
& wei , f .- s .
1998 , earth planets space , 50 , 839
|
a regression modeling method of space weather prediction is proposed .
it allows forecasting dst index up to 6 hours ahead with about @xmath0 correlation .
it can also be used for constructing phenomenological models of interaction between the solar wind and the magnetosphere . with its help
two new geoeffective parameters were found : latitudinal and longitudinal flow angles of the solar wind .
it was shown that dst index remembers its previous values for 2000 hours .
|
cet article sinscrit dans le cadre de la correspondance de langlands @xmath0-adique .
dans ses articles @xcite , breuil a dfini une correspondance qui une reprsentation @xmath0-adique @xmath1 de dimension @xmath6 qui est cristalline ou semi - stable , associe une reprsentation @xmath2 de @xmath7 .
le fait que @xmath2 est non - nul , irrductible et admissible a t dmontr par colmez pour les reprsentations semi - stables dans @xcite puis par breuil et lauteur pour les reprsentations cristallines dans @xcite .
la correspondance a ensuite t tendue aux reprsentations triangulines par colmez dans @xcite .
dans @xcite , breuil a par ailleurs dfini une correspondance en caractristique @xmath0 et conjectur quelle tait compatible avec la premire . lobjet
de cet article est de dmontrer cette conjecture pour les reprsentations triangulines cristalline non @xmath8-semi - simple ou bien @xmath1 non - cristalline mais le devenant sur une extension ablienne , la dmonstration repose sur des rsultats non - rdigs . ] .
afin dnoncer ce thorme , nous devons introduire quelques notations qui nous serviront par la suite .
dans tout cet article , @xmath0 est un nombre premier et le corps @xmath9 ( le corps des coefficients ) est une extension finie de @xmath10 do nt on note @xmath11 lanneau des entiers , @xmath12 lidal maximal et @xmath13 le corps rsiduel . on note @xmath14 le groupe de galois @xmath15 , @xmath16 la rduction modulo @xmath0 du caracre cyclotomique @xmath17 ,
et pour @xmath18 ou @xmath19 on note @xmath20 le caractre non - ramifi de @xmath14 qui envoie @xmath21 ( cest - - dire linverse du frobenius arithmtique ) sur @xmath22 .
on normalise le corps de classes pour que @xmath23 senvoie sur @xmath21 ce qui permet de voir @xmath20 comme le caractre non - ramifi de @xmath24 qui envoie @xmath0 sur @xmath22 . on note enfin @xmath25 le caractre fondamental de serre de niveau @xmath6 .
on appelle @xmath26 le sous - groupe de borel suprieur de @xmath7 .
pour allger les notations , il nous arrivera dcrire @xmath27 pour @xmath7 et @xmath28 pour @xmath26 ainsi que didentifier @xmath24 au centre de @xmath7 .
si @xmath29 et @xmath30 sont deux @xmath13-reprsentations lisses de longueur finie de @xmath26 ( ou de @xmath7 ) , on crira @xmath31 ( ou bien @xmath32 ) pour signifier que les semi - simplifies de @xmath29 et de @xmath30 sont isomorphes .
si @xmath1 est une reprsentation @xmath0-adique de @xmath14 , on en choisit un @xmath11-rseau @xmath33 stable par @xmath14 et @xmath34 ne dpend pas du choix de @xmath33 .
de mme , si @xmath35 est une reprsentation de @xmath7 qui admet un rseau @xmath36 , on note @xmath37 la semi - simplifie de la rduction de @xmath36 qui , si elle est de longueur finie , ne dpend pas du choix du rseau dans une classe de commensurabilit .
si @xmath38 et si @xmath39 est un caractre continu , que lon identifie un caractre continu de @xmath14 via le corps de classes , alors on pose : @xmath40 on pose par ailleurs : @xmath41 o @xmath42 et @xmath33 est un certain oprateur de hecke .
thorme a si @xmath1 est une reprsentation trianguline irrducible , alors : @xmath43 , \lambda^{-1 } , \omega^{r+1 } \chi)\end{aligned}\ ] ] ici @xmath44 $ ] est lunique entier appartenant @xmath45 et qui est congruent @xmath46 modulo @xmath47 .
la dmonstration de ce thorme est fonde sur le lien entre @xmath2 et le @xmath48-module @xmath49 , dabord montr par colmez pour les reprsentations semi - stables puis par breuil et lauteur pour les reprsentations cristallines et enfin par colmez pour les reprsentations triangulines ( les cas exceptionnels ntant pas ( encore ) rdigs )
. le premier chapitre est consacr des rappels .
le deuxime chapitre est consacr ltude de reprsentations de @xmath26 construites partir de certaines reprsentations galoisiennes .
on trouve soit des restrictions dinduites paraboliques , soit des restrictions de supersingulires et le fait den avoir des modles explicites est important pour la suite .
nous dmontrons notamment les rsultats suivants dintrt indpendant .
thorme b si @xmath35 est une @xmath13-reprsentation lisse irrductible de @xmath7 admettant un caractre central , alors : 1 .
si @xmath35 est un caractre , ou la spciale , ou supersingulire , alors sa restriction @xmath26 est toujours irrductible ; 2 .
si @xmath35 est une srie principale , alors sa restriction @xmath26 est une extension du caractre induisant @xmath50 par une reprsentation irrductible @xmath51 . thorme c si @xmath29 et @xmath30 sont deux @xmath13-reprsentations lisses semi - simples et de longueur finie de @xmath7 ( do nt les composantes irrductibles admettent un caractre central ) et do nt les semi - simplifications des restrictions @xmath26 sont isomorphes , alors @xmath29 et @xmath30 sont dj isomorphes en tant que reprsentations de @xmath7 .
dans le troisime chapitre , on utilise les rsultats explicites du deuxime pour dmontrer le thorme a. on termine en faisant le point sur les rsultats que lon en dduit concernant la rduction modulo @xmath0 des reprsentations cristallines et semi - stables .
lobjet de ce chapitre est de rappeler certaines des constructions faites dans le cadre de la correspondance de langlands @xmath0-adique .
@xcite pour plus de dtails .
nous reprenons les notations de lintroduction .
les caractres @xmath52 sont tous de la forme @xmath53 pour @xmath54 et @xmath55 .
rappelons la classification des reprsentations absolument irrductibles de @xmath14 de dimension @xmath6 sur @xmath13 .
soit @xmath56 le caractre fondamental de serre de niveau @xmath6 . on suppose dsormais que @xmath57 .
si @xmath38 et si @xmath39 est un caractre continu , que lon identifie un caractre continu de @xmath14 via le corps de classes , alors on pose : @xmath40 on obtient ainsi toutes les reprsentations absolument irrductibles de @xmath14 de dimension @xmath6 sur @xmath13 , et les entrelacements entre les @xmath58 sont les suivants : @xmath59 la classification des reprsentations @xmath0-adiques de @xmath14 est , comme on le sait , beaucoup plus complique mais la thorie des @xmath48-modules permet de sy retrouver un petit peu .
soit @xmath60 et @xmath61 lanneau @xmath62 o @xmath63 et @xmath64 quand @xmath65 .
on munit cet anneau dun frobenius @xmath11-linaire @xmath8 dfini par @xmath66 et dune action @xmath11-linaire de @xmath67 donne par @xmath68 si @xmath69 .
un @xmath48-module tale est un @xmath61-module @xmath70 de type fini muni dun frobenius semi - linaire @xmath8 tel que @xmath71 et dune action de @xmath67 semi - linaire continue et commutant @xmath8 .
rappelons que fontaine a construit dans ( * ? ? ?
* a.3.4 ) un foncteur @xmath72 qui toute @xmath11-reprsentation de @xmath14 associe un @xmath48-module tale et que ce foncteur est une quivalence de catgories .
si @xmath73 est une reprsentation @xmath0-adique , alors on pose @xmath74
. lanneau @xmath61 est un @xmath75-module libre de rang @xmath0 , do nt une base est donne par @xmath76 .
si @xmath77 , alors on peut crire @xmath78 et on dfinit un oprateur @xmath79 par la formule @xmath80 si @xmath81 .
si @xmath70 est un @xmath48-module tale sur @xmath61 , alors colmez a dfini dans @xcite un sous-@xmath82$]-module @xmath83 de @xmath70 , qui est caractris par les proprits suivantes : 1 .
@xmath83 est un treillis de @xmath70 ( voir @xcite ) ; 2 .
quels que soient @xmath84 et @xmath85 , il existe @xmath86 tel que @xmath87 si @xmath88 ; 3 .
loprateur @xmath89 induit une surjection de @xmath83 sur lui - mme . le calcul de @xmath90 pour des reprsentations de dimension @xmath91 ne pose aucun problme et le calcul suivant sera utile dans la suite de cet article .
[ calcidiot ] si @xmath92 est une @xmath13-reprsentation de @xmath14 de dimension @xmath91 , donne par un caractre @xmath53 , alors @xmath93 \cdot e$ ] , avec @xmath94 o @xmath95 est tel que @xmath96 .
sous les hypothses du lemme ci - dessus , on pose alors @xmath97 \cdot e$ ] .
remarquons en passant que @xmath98 et que @xmath99 si @xmath69 .
la notation @xmath100 provient de la thorie des reprsentations de hauteur finie ( qui nintervient pas dans le reste de cet article ) . [ schur_pgmod ] si @xmath101 est une @xmath13-reprsentation irrductible de @xmath14 de dimension @xmath102 , et si @xmath103 est un sous-@xmath104$]-module non - nul stable par @xmath89
, alors @xmath105 .
de mme , si @xmath101 est une @xmath13-reprsentation de @xmath14 de dimension @xmath91 , et si @xmath106 est un sous-@xmath104$]-module non - nul stable par @xmath89 , alors @xmath107 .
faisons tout dabord le cas o @xmath101 est de dimension @xmath102 .
pour tout polynme @xmath108 $ ] , on a @xmath109 parce que ( voir la preuve du ( iii ) de la remarque 5.5 de @xcite ) on a : @xmath110 le lemme suit alors de la proposition 4.47 de @xcite ( ou plus exactement de sa dmonstration , en remarquant que la dmonstration nutilise pas le fait que @xmath111 est surjectif ) . faisons maintenant le cas o @xmath101 est de dimension @xmath91 .
il sagit de montrer quun idal de @xmath104 $ ] non - nul et stable par @xmath89 est gal @xmath104 $ ] .
les idaux de @xmath104 $ ] sont de la forme @xmath112 $ ] pour @xmath113 et un calcul facile montre que si @xmath114 , alors @xmath115)$ ] contient @xmath116 $ ] ce qui permet de conclure .
[ constb ] commenons ce paragraphe en rappelant que si @xmath35 est une @xmath13-reprsentation compacte dun groupe localement profini @xmath117 , alors sa duale @xmath118 , o @xmath119 parcourt les sous-@xmath13-espaces vectoriels ouverts de @xmath35 , est une @xmath13-reprsentation lisse de @xmath117 et que rciproquement si @xmath120 est une @xmath13-reprsentation lisse de @xmath117 , alors sa duale @xmath121 , o @xmath122 parcourt les sous - groupes ouverts compacts de @xmath117 , est une @xmath13-reprsentation compacte de @xmath117 . de plus , on a @xmath123 et @xmath124 et les sous - espaces ferms de @xmath35 stables par @xmath117 correspondent aux quotients de @xmath120 stables par @xmath117 , et vice - versa .
si @xmath70 est un @xmath48-module , alors @xmath125 dnote lensemble des suites @xmath126 telles que @xmath127 pour tout @xmath128 .
si @xmath70 est un @xmath48-module sur
@xmath9 , on demande en plus que la suite @xmath129 soit borne pour la topologie @xmath0-adique . on fixe un caractre lisse @xmath130 de @xmath24 et on munit
@xmath125 dune action de @xmath26 comme suit .
tout lment @xmath131 peut scrire
comme produit : @xmath132 o @xmath133 , @xmath134 , @xmath135 et @xmath136 .
si @xmath137 , alors on pose pour @xmath138 : @xmath139 [ defomega ] si @xmath101 est une @xmath13-reprsentation irrductible de @xmath14 de dimension @xmath102 , et si @xmath39 est un caractre lisse , alors on pose @xmath140 et si @xmath101 est une @xmath13-reprsentation de dimension @xmath91 , alors on pose @xmath141 .
[ omirred ] si on munit @xmath142 dune action de @xmath26 en utilisant les formules ci - dessus , alors @xmath142 est une reprsentation lisse irrductible do nt le caractre central est @xmath130 .
comme @xmath143 est compact , le fait que @xmath142 est une reprsentation lisse rsulte des rappels que lon a faits au dbut du paragraphe .
le fait que le caractre central est @xmath130 est vident .
enfin , pour montrer que @xmath142 est irrductible , il suffit de montrer que @xmath143 ( ou bien @xmath144
si @xmath145 ) nadmet pas de sous - espace ferm stable par @xmath26 et non - trivial , ce que nous faisons maintenant ( si @xmath145 , il faut remplacer @xmath146 par @xmath147 dans les calculs qui suivent ) .
soit @xmath148 la projection @xmath149 .
si @xmath150 est un sous - espace ferm et stable par @xmath26 de @xmath143 , on note @xmath151 limage de @xmath152 . on voit que @xmath151 est
un sous-@xmath104$]-module non - nul de @xmath146 stable par @xmath89 ce qui fait que , par le lemme [ schur_pgmod ] , @xmath153 . on en dduit
que @xmath150 est dense dans @xmath143 et donc finalement que @xmath154 ce qui fait que @xmath143 est bien topologiquement irrductible et donc que @xmath142 est irrductible .
[ schur ] si @xmath155 et @xmath156 sont deux @xmath13-reprsentations irrductibles de @xmath14 et si @xmath157 est une application @xmath26-quivariante et non - nulle , alors @xmath158 et @xmath159 et @xmath160 est scalaire .
il est immdiat que @xmath158 et on se ramne donc montrer que si @xmath155 et @xmath156 sont deux reprsentations irrductibles telles quil existe une application @xmath161-quivariante continue et non - nulle @xmath162 , alors @xmath159 .
la dmonstration est analogue celle de la proposition 3.4.3 de @xcite et nous la faisons pour @xmath163 ; quand @xmath164 , il suffit de remplacer @xmath165 par @xmath166 dans les calculs qui suivent .
notons comme ci - dessus @xmath167 la projection @xmath168 .
commenons par montrer que si @xmath169 , alors @xmath170 ne dpend que de @xmath171 .
pour cela , soit @xmath172 lensemble des @xmath173 do nt les @xmath174 premiers termes sont nuls , ce qui fait que pour @xmath175 , @xmath172 est un sous-@xmath104$]-module ferm et stable par @xmath89 de @xmath176 et que @xmath177 . on en dduit
que @xmath178 est un sous-@xmath104$]-module ferm et stable par @xmath89 de @xmath179 .
le lemme [ schur_pgmod ] implique alors que soit @xmath180 , soit @xmath181 .
enfin , on voit que @xmath182 dune part et que @xmath183 si @xmath184 par continuit dautre part .
cela implique que @xmath183 pour tout @xmath175 et donc que si @xmath185 , alors @xmath186 .
pour tout @xmath187 , soit @xmath188 un lment de @xmath189 tel que @xmath190 .
les calculs prcdents montrent que lapplication @xmath191 donne par @xmath192 est bien dfinie , et quelle est @xmath104$]-linaire et commute @xmath89 et laction de @xmath67 .
par les propositions 4.7 et 4.55 de @xcite , elle stend en une application de @xmath48-modules @xmath193 et par fonctorialit , on en dduit quil existe une application non - nulle et @xmath14-quivariante de @xmath155 dans @xmath156 , ce qui fait que par le lemme de schur , on a @xmath159 et que @xmath160 est scalaire .
si @xmath101 est une @xmath13-reprsentation de dimension @xmath91 , le lemme [ calcidiot ] montre que @xmath93 \cdot e$ ] . on appelle
@xmath194 lapplication de @xmath146 dans @xmath13 qui @xmath195 associe @xmath196 et on note toujours @xmath194 lapplication de @xmath143 dans @xmath13 qui @xmath197 associe @xmath198 .
si @xmath199 et @xmath200 sont deux caractres de @xmath24 , on note @xmath50 le caractre de @xmath26 dfini par la formule : @xmath201 [ plusetdiese ] si @xmath202 dnote le caractre de @xmath24 associ @xmath101 par le corps de classes , alors lapplication @xmath194 induit par dualit une suite exacte de @xmath26-reprsentations : @xmath203 il sagit de montrer que si @xmath204 , alors on a une suite exacte : @xmath205 si lon pose @xmath206 , alors @xmath207 et @xmath208 ce qui fait quil suffit de calculer @xmath209 pour @xmath210 . on trouve que : @xmath211 @xmath212 ce qui permet de conclure .
nous allons voir dans la suite de cet article que si @xmath213 , alors les reprsentations @xmath142 sont les restrictions @xmath26 de reprsentations de @xmath7 .
quand @xmath214 , ce nest plus le cas et on obtient des reprsentations lisses irrductibles propres @xmath26 .
obtient - on ainsi toutes les reprsentations lisses irrductibles de @xmath26 ? [ repg ] lobjet de ce paragraphe est de rappeler brivement certaines des constructions de breuil et colmez .
il sagit dassocier une reprsentation @xmath2 de @xmath7 une reprsentation @xmath0-adique @xmath1 irrductible de dimension @xmath6 .
cette reprsentation @xmath2 a t conjecturalement construite pour les reprsentations cristallines et semi - stables par breuil ( voir @xcite pour le cas cristallin et @xcite pour le cas semi - stable ) .
le fait que la reprsentation @xmath2 est non - nulle , irrductible et admissible ( au sens de @xcite ) a t dmontr par colmez pour
@xmath1 semi - stable ( voir @xcite ; des cas particuliers se trouvent dans @xcite ) puis par breuil et lauteur pour @xmath1 cristalline ( voir @xcite ; des cas particuliers se trouvent dans @xcite ) .
enfin , ces rsultats ont t tendus aux reprsentations triangulines par colmez ( voir @xcite ) .
le thorme ci - dessous rassemble les rsulats do nt nous nous servirons cristalline non @xmath8-semi - simple ou bien @xmath1 non - cristalline mais le devenant sur une extension ablienne ) . ] .
[ colmod ] si @xmath1 est une reprsentation trianguline irrductible de dimension @xmath6 , alors : 1 . on a @xmath215
si @xmath33 est un rseau de @xmath1 , alors @xmath216 est un rseau de @xmath2 stable sous laction de @xmath26 et commensurable aux rseaux de @xmath2 ; 3 . on a @xmath217 .
le ( 1 ) est dmontr dans @xcite pour une reprsentation semi - stable , dans @xcite pour une reprsentation cristalline et enfin dans @xcite dans le cas dune reprsentation trianguline .
le ( 2 ) est un rsultat facile ( voir @xcite par exemple ) et le ( 3 ) suit du ( 1 ) et du ( 2 ) ainsi que de la proposition 4.50 de @xcite et du principe de brauer - nesbitt ( ce qui explique quil faut semi - simplifier ) .
rappelons prsent la classification des @xmath13-reprsentations lisses irrductibles de @xmath7 qui admettent un caractre central .
si @xmath38 , si @xmath42 et si @xmath218 est un caractre continu , on pose : @xmath41 o @xmath33 est un certain oprateur de hecke ( voir @xcite ) .
si @xmath219 , alors @xmath220 est irrductible ( pour @xmath221 , cest un rsultat de @xcite et pour @xmath222 , cest un rsultat de @xcite ) .
pour @xmath223 , les entrelacements entre les @xmath224 sont les suivants : @xmath225 si @xmath226 , alors la semi - simplifie de @xmath220 est somme dun caractre et de la tordue dune reprsentation appele la spciale ( qui correpond @xmath227 , @xmath228 et @xmath229 ) .
enfin , toute reprsentation lisse irrductible de @xmath7 qui admet un caractre central que lon considre admettent un caractre central . ]
est dans la liste ci - dessous : 1 .
les caractres @xmath230 ; 2 .
la spciale tordue par un caractre @xmath231 ; 3 .
les sries principales @xmath220 avec @xmath221 et @xmath219 ; 4 .
les supersingulires @xmath224 .
nous rappelons maintenant le calcul de @xmath232 pour certaines reprsentations cristallines .
si @xmath233 et si @xmath234 , alors on dfinit un @xmath8-module filtr @xmath235 par @xmath236 o : @xmath237 ce @xmath8-module filtr est admissible , et on sait ( par le thorme principal de @xcite ) quil existe alors une reprsentation cristalline @xmath238 de @xmath14 telle que @xmath239 ( on passe au dual pour que les notations soient compatibles avec celles de @xcite et @xcite ; remarquons tout de mme que lon a @xmath240 ) .
soit @xmath241 la reprsentation localement algbrique construite par breuil dans @xcite : @xmath242 o @xmath33 est un certain oprateur de hecke ( voir @xcite ) .
il est montr dans @xcite que @xmath241 admet un @xmath11-rseau de type fini sur @xmath243 $ ] et que sa compltion est isomorphe @xmath244 .
breuil a dautre part dtermin @xmath245 pour @xmath246 ( voir le paragraphe [ calc ] ) et nous aurons besoin dans la suite du cas particulier ci - dessous ( cest le corollaire 5.1 de @xcite ) .
[ breuilmod ] si @xmath247 , alors @xmath248 .
lobjet de ce chapitre est de faire le lien entre les diffrentes constructions de reprsentations lisses irrductibles de @xmath26 ( soit comme restriction de reprsentations lisses irrductibles de @xmath7 , soit directement comme @xmath142 ) .
[ indpar ] on commence par faire ltude des induites paraboliques .
si @xmath199 et @xmath200 sont deux caractres de @xmath24 , alors linduite parabolique @xmath249 est par dfinition lensemble des fonctions localement constantes @xmath250 qui vrifient @xmath251 . si @xmath252 , alors lapplication @xmath253 nous donne une
application de @xmath249 dans @xmath50 do nt on note @xmath51 le noyau ce qui fait que lon a une suite exacte de @xmath26-reprsentations : @xmath254 soit @xmath255 lensemble des fonctions @xmath256 localement constantes support compact .
si @xmath252 , alors on associe @xmath257 une fonction @xmath258 via la recette : @xmath259 la dcomposition de bruhat montre que lapplication @xmath260 est une bijection de @xmath51 dans @xmath255 et on a alors : @xmath261 [ omdim1 ] si @xmath202 dnote le caractre de @xmath24 associ @xmath101 par le corps de classes , alors @xmath262 .
ecrivons @xmath204 , ce qui fait que lon doit montrer que : @xmath263 nous allons dfinir une application : @xmath264 et montrer que cest un isomorphisme @xmath26-quivariant .
rappelons tout dabord la dfinition de la transforme damice ( tout au moins la version do nt nous avons besoin ) .
si @xmath265 est une forme linaire @xmath266 ( on dira que @xmath265 est une mesure sur @xmath267 ) , alors sa transforme damice @xmath268 $ ] est dfinie par : @xmath269 et lapplication @xmath270 $ ] est alors un isomorphisme .
fixons une base @xmath271 de @xmath147 telle que @xmath272 et @xmath273 ( voir le lemme [ calcidiot ] ) ; si @xmath274 , alors @xmath275 et on a @xmath276 avec @xmath277 $ ] .
on voit que @xmath278 et on dfinit une mesure @xmath279 sur @xmath267 en demandant que @xmath280 ce qui nous permet de dfinir une mesure @xmath281 sur @xmath10 en imposant que si @xmath282 est support dans @xmath283 , alors : @xmath284 le terme de droite ne dpend pas de @xmath285 , en vertu de la formule : @xmath286 et @xmath287 est donc bien dfinie .
il est clair que lapplication @xmath288 est une bijection qui nous permet didentifier , via la recette @xmath260 rappele au dbut de ce paragraphe , @xmath142 avec le dual de linduite parabolique .
pour terminer la dmonstration du thorme , nous devons montrer que cette identification est @xmath26-quivariante .
laction de @xmath26 sur les mesures est donne par laction correspondante sur leurs transformes damice et on trouve les formules suivantes : @xmath289 en comparant ces formules avec ( eq1 ) , on trouve bien que : @xmath290 si @xmath131 , @xmath291 et @xmath292 .
pour terminer , rappelons le lien entre les induites paraboliques et les @xmath220 ( cest le thorme 30 de @xcite ) .
[ pietind ] on a @xmath293 .
[ spe ] en particulier , la spciale @xmath294 est isomorphe @xmath295 .
dans ce cas , la suite exacte de reprsentations de @xmath26 : @xmath296 est dailleurs scinde ( le @xmath13-espace vectoriel engendr par la fonction @xmath297 est un supplmentaire @xmath26-stable de @xmath294 ) .
il est amusant de remarquer que la suite exacte de reprsentations de @xmath7 : @xmath298 nest _ pas _ scinde ( voir la dmonstration du thorme 29 de @xcite ) .
[ supsing ] dans ce court paragraphe , nous dmontrons lanalogue du thorme [ omdim1 ] pour les reprsentations de dimension @xmath6 mais contrairement la dmonstration du thorme [ omdim1 ] , largument nest pas direct et repose de manire essentielle sur des calculs en caractristique @xmath299 .
il serait intressant de disposer dune dmonstration purement en caractristique @xmath0 .
[ omdim2 ] si @xmath300 , alors @xmath301 est isomorphe la restriction au borel de @xmath224 . un calcul facile montre que @xmath302 et le ( 3 ) du thorme [ colmod ] montre alors que @xmath303 ce qui fait que , comme le caractre central de @xmath304 est @xmath305 , on a : @xmath306 par la proposition [ breuilmod ] .
[ appli ] pour terminer ce chapitre , nous donnons deux applications des identifications prcdentes ltude des reprsentations de @xmath26 et de @xmath7 .
[ resbor ] si @xmath35 est une @xmath13-reprsentation lisse irrductible de @xmath7 admettant un caractre central , alors : 1 .
si @xmath35 est un caractre , ou la spciale , ou supersingulire , alors sa restriction @xmath26 est toujours irrductible ; 2 .
si @xmath35 est une srie principale , alors sa restriction @xmath26 est une extension du caractre induisant @xmath50 par la reprsentation irrductible @xmath51 .
montrons tout dabord le ( 1 ) .
comme la spciale est isomorphe @xmath295 on renvoie au ( 2 ) et le cas des caractres est trivial , ce qui nous laisse les supersingulires .
comme @xmath307 , le fait que cette reprsentation est irrductible en restriction au borel suit de la proposition [ omirred ] . passons prsent au ( 2 ) .
on a vu au paragraphe [ indpar ] que lon a une suite exacte : @xmath308 ce qui montre lassertion concernant lextension , et comme @xmath309 par le thorme [ omdim1 ] , le fait que cette reprsentation est irrductible suit encore de la proposition [ omirred ] .
[ sameresbor ] si @xmath29 et @xmath30 sont deux @xmath13-reprsentations lisses semi - simples et de longueur finie de @xmath7 ( do nt les composantes irrductibles admettent un caractre central ) et do nt les semi - simplifications des restrictions @xmath26 sont isomorphes , alors @xmath29 et @xmath30 sont dj isomorphes en tant que reprsentations de @xmath7 . une reprsentation semi - simple et de longueur finie de @xmath7 est une somme directe de caractres @xmath230 , de tordues de la spciale , de sries principales @xmath249 avec @xmath310 et de supersingulires
. la restriction @xmath26 du caractre
@xmath230 est @xmath311 et par le thorme [ resbor ] , celle de @xmath249 est une extension de @xmath50 ( avec ncessairement @xmath310 ) par une reprsentation de dimension infinie alors que les restrictions des tordues de la spciale et des supersingulires sont irrductibles et de dimension infinie .
on voit donc que si lon a deux reprsentations @xmath29 et @xmath30 qui satisfont les conditions du thorme , alors elles contiennent les mmes caractres et sries principales .
pour terminer , il faut voir que les tordues des spciales et les supersingulires sont dtermines par leur restriction au borel .
elles sont toutes
de la forme @xmath142 ( avec @xmath101 de dimension @xmath91 pour la spciale par le thorme [ omdim1 ] et @xmath101 de dimension @xmath6 pour les supersingulires par le thorme [ omdim2 ] ) et la proposition [ schur ] montre quil ny a pas de @xmath26-entrelacements non - triviaux entre deux telles reprsentations de @xmath7 diffrentes .
dans ce chapitre , nous dmontrons le rsultat principal de cet article ( la correspondance entre rduction modulo @xmath0 des reprsentations galoisiennes @xmath1 et des reprsentations @xmath2 de @xmath7 ) et ensuite nous rappelons les cas o le calcul de cette rduction a t fait .
supposons tout dabord que lon connaisse @xmath314 .
le thorme [ omdim1 ] et la proposition [ plusetdiese ] quand @xmath314 est rductible , ou bien le thorme [ omdim2 ] quand @xmath314 est irrductible , ainsi que la formule ( eq2 ) , montrent que @xmath232 est une reprsentation de @xmath7 do nt la restriction @xmath26 concide avec celle qui est prdite par le thorme , et le thorme [ sameresbor ] permet alors de conclure
. supposons maintenant que lon connaisse @xmath232 ; on peut conclure comme ci - dessus ( nous laissons cela en exercice ) mais il y a plus explicite .
si @xmath232 est une reprsentation supersingulire , alors par le thorme [ resbor ] sa restriction @xmath26 est toujours irrductible ce qui fait que ( cause de la formule ( eq2 ) ) @xmath314 est irrductible et le thorme [ omdim2 ] et la proposition [ schur ] permettent de conclure .
si @xmath232 est du deuxime type , alors @xmath314 est forcment rductible et la formule ( eq2 ) ainsi que la proposition [ plusetdiese ] montrent que lon peut lire @xmath314 sur la restriction @xmath26 de @xmath232 ( cest la tordue par @xmath16 de la partie de dimension finie de cette restriction , cest cela qui est explicite ) .
[ simpler ] la dmonstration ci - dessus montre
quen fait , la restriction @xmath26 de @xmath232 suffit dterminer @xmath314 .
dans les applications , cest plutt @xmath232 que lon calcule , et cela a pour effet de simplifier ces calculs .
[ ghost ] si @xmath101 est de dimension @xmath91 , on voit que la reprsentation @xmath142 est dune part un sous - objet de @xmath315 et dautre part un quotient de @xmath316 .
on pourrait donc croire que @xmath317 mais en fait on voit que : @xmath318 et dans la correspondance entre @xmath319 et @xmath320 , \lambda^{-1 } , \omega^{r+1 } \chi)$ ] , les termes se croisent et il y a donc une sorte dentrelacement fantme entre @xmath220 et @xmath321 , \lambda^{-1 } , \omega^{r+1 } \chi)$ ] ( et il y a dailleurs un vritable entrelacement en @xmath322-adique pour @xmath323 , voir la remarque 4.2.5 de @xcite ) . lobjet de ce paragraphe est de rappeler les rsultats des calculs qui ont t faits par diffrents auteurs .
il sagit soit du calcul de @xmath314 , soit du calcul de @xmath232 , pour certaines reprsentations @xmath1 .
commenons par les reprsentations cristallines @xmath238 dfinies la fin du paragraphe [ repg ] .
le calcul de @xmath324 remonte fontaine et serre ( en raison du lien avec les formes modulaires ) ; la thorie de fontaine - laffaille ( voir @xcite ) permet de faire ce calcul pour tout @xmath325 .
quand @xmath247 , cest un rsultat dedixhoven ( sous lhypothse que la reprsentation provienne dune forme modulaire ) .
enfin , quand @xmath326 , la thorie des modules de wach permet de calculer @xmath324 si @xmath327 est assez grand ( voir @xcite et @xcite ) .
le calcul de @xmath328 a t fait par breuil pour @xmath329 ( voir @xcite pour @xmath330 ; si @xmath331 et @xmath332 , cest un calcul non - publi de breuil ) .
voici la liste des rsultats que lon obtient . 1 . si @xmath333 , alors @xmath334 .
2 . pour @xmath335 : 1 .
si @xmath336 , alors @xmath337 .
si @xmath338 , et si @xmath339 est une racine du polynme @xmath340 , alors : @xmath341
3 . pour @xmath342 : 1 .
si @xmath336 , alors @xmath343 .
si @xmath344 , et si @xmath345 , alors : @xmath346 3 .
si @xmath347 , alors @xmath348 .
4 . pour @xmath331 ( et @xmath349 ) : 1 .
si @xmath350 , alors @xmath351 .
si @xmath352 , alors : @xmath353 5 .
pour @xmath354 , les rsultats ne sont que partiels : 1 .
si @xmath355 et si @xmath356 , alors @xmath348 .
si @xmath355 , si @xmath357 et si @xmath358 , alors : @xmath359 [ forkisin ] pour @xmath360 et @xmath361 , on a un rsultat plus prcis ( voir ( * ? ? ?
* thorme 2.2.5 ) ) concernant lextension des deux caractres pour @xmath362 : il existe un rseau de @xmath238 do nt la rduction modulo @xmath0 est : @xmath363 o @xmath364 est non - trivial et peu ramifi .
pour ce qui est des reprsentations semi - stables , le calcul de la rduction ct galois a t fait dans @xcite ( pour @xmath365 pair vrifiant @xmath366 ) et le calcul de la rduction ct @xmath7 a t fait dans @xcite ( pour @xmath365 pair vrifiant @xmath333 et en supposant de plus que linvariant @xmath367 est de valuation @xmath368 si @xmath369 ) .
nous ne rappelons pas ici les formules , que lon peut trouver dans @xcite ct galois et dans @xcite ( pour @xmath370 ) et @xcite ( pour @xmath371 ) ct @xmath7 .
* remerciements * : je remercie c. breuil pour ses encouragements tout au long de la rdaction de ce travail , et je le remercie de mme que p. colmez pour des discussions clairantes sur plusieurs points de cet article .
|
on montre la conjecture de breuil concernant la rduction modulo @xmath0 des reprsentations triangulines @xmath1 et des reprsentations @xmath2 de @xmath3 qui leur sont associes par la correspondance de langlands @xmath0-adique .
lingrdient principal de la dmonstration est ltude de certaines reprsentations lisses irrductibles de @xmath4 via des modles construits en utilisant les @xmath5-modules .
we prove breuil s conjecture concerning the reduction modulo @xmath0 of trianguline representations @xmath1 and of the representations @xmath2 of @xmath3 associated to them by the @xmath0-adic langlands correspondence .
the main ingredient of the proof is the study of some smooth irreducible representations of @xmath4 through models built using the theory of @xmath5-modules .
|
the transfer of electronic charge density through a hydrogen bond manifests in effects ranging from the stabilization and spectroscopy of hydrated ions and aqueous solutions @xcite to the charging of polar residues in proteins @xcite , and has even been implicated in explaining the observed zeta potential at hydrophobic interfaces @xcite . however , defining the amount of charge transfer ( ct ) remains a subject of significant debate , since partitioning the electron density and assigning it to particular atoms or molecules is not uniquely defined .
this has led to the introduction of a number of approaches to calculate the charge transfer and charge transfer energy , with different approaches exhibiting both different quantitative and qualitative behavior @xcite .
for example , methods based on natural bond orbitals suggest that without ct the 0 k water dimer would adopt a different structure @xcite .
however , studies using other methods suggest a much less substantial role of ct in stabilizing the water dimer hydrogen bond @xcite .
previous ct studies of hydrogen bonded systems have analyzed minimum energy structures @xcite or those obtained from simulations that treat the nuclei classically @xcite .
however , nuclear quantum effects ( nqes ) , such as zero - point energy and tunnelling , are known to significantly change the geometry of hydrogen bonded systems @xcite .
the zero - point energy along the hydrogen bond increases the ability of the proton to delocalize between the donor and acceptor , resulting in shorter and stronger hydrogen bonds . on the other hand
, the distortion of the hydrogen bond due to quantization of the librational modes perpendicular to the hydrogen bond leads to weaker hydrogen bonding .
the relative importance of these two competing quantum effects @xcite leads to cases where nqes strengthen @xcite or weaken @xcite hydrogen bonds , depending on the geometry and chemical environment .
this competition has been used to elucidate the seemingly anomalously small nqes and resulting isotope effects observed in many systems @xcite .
here we investigate the role of nqes in modulating the ct characteristics of a range of hydrogen bonded systems involving water and ions in both the gas and condensed phase and show how these can be understood based on the changes in the hydrogen bond geometry .
we demonstrate that nqes give rise to large changes in both the amount of ct and its contribution to the donor - acceptor interaction energy , independent of the method used to study charge transfer . in particular ,
including nqes increases ct in some cases , while decreasing it in others .
it also leads to a much milder temperature dependence of ct in all cases . to uncover the reasons for this , we show that the amount of charge transfer is almost exclusively determined by a scaled proton sharing coordinate .
we use this observation to demonstrate that the charge of a water molecule in liquid water can be accurately estimated from the values of this coordinate for each hydrogen bond the molecule forms , thus provides a straightforward approach to modelling ct in solution .
to evaluate nqes on electronic properties in the gas phase , ab initio path integral molecular dynamics ( ai - pimd ) simulations at and of the water dimer as well as the water - fluoride and water - iodide complex were performed .
these simulations were performed with a time step of with 256 and 32 replicas , at the two respective temperatures .
the b3lyp @xcite exchange correlation functional and the pople basis set @xcite for the water dimer and fluoride - water complex as well as the def2-svpd basis set @xcite for the iodide - water complex were used to calculate electronic interactions on the fly during the simulations .
an additional at the beginning of each simulation was discarded as equilibration after starting from the optimized minimum energy configuration .
langevin thermostats @xcite with optimal coupling to each staging mode @xcite were used to sample the canonical distribution .
two independent classical ab initio molecular dynamics ( aimd ) simulations of each system were carried out using the same electronic structure method , time step and temperature , together with massive nos hoover chains thermostatting @xcite with a chain length of 6 .
the simulations were performed with the in - house python program pymd @xcite that uses ipython parallelization @xcite and the atomic simulation environment @xcite to link to gaussian 09 @xcite for the electronic structure calculation .
our choice of basis set and number of replicas gives good convergence for geometrical properties ( see si ) .
since it is known that density functional theory ( dft ) can cause spurious charge delocalization effects @xcite , we validated our choice of methods by also conducting mp2 simulations of the water dimer , as detailed in the si .
the geometries were further validated by comparing to the very accurate mbpol force field @xcite ; a detailed analysis can be found in the si .
population analyses of the gas - phase configurations were performed with the b3lyp density functional and the augmented cc - pvtz dunning basis @xcite ( water dimer and fluoride - water complex ) or the def2-qzvppd basis @xcite ( iodide - water complex ) using gaussian 09 @xcite where the basis sets were chosen to give basis set converged results ( see si ) . for the aimd trajectories , configurations
every were analyzed , whereas for the ai - pimd simulations at configurations every of every 10@xmath0 bead were used .
configurations of the ai - pimd simulations at were selected from every 5@xmath0 bead every .
note that we applied a standard hydrogen bond criterion ( cutoff of 3.5 on the oxygen - oxygen distance and 130 on the o - h - o angle ) for the water dimer at to discard non - bonded configurations caused by thermal fluctuations .
bader s quantum theory of atoms in molecules ( qtaim ) charges @xcite were calculated with the aimall program @xcite from the wavefunction provided by gaussian 09 , hirshfeld charges were calculated directly with gaussian 09 and natural population analysis ( npa ) charges with calls to the nbo 6.0 program @xcite .
sapt(2 + 3 ) interaction energies , including the extension for charge transfer @xcite , were obtained using psi4 @xcite with the aug - cc - pvtz basis set and the corresponding auxiliary basis set for density fitting . to analyze electronic properties in liquid water including an explicit treatment of nuclear quantum effects , ai - pimd and aimd trajectories obtained in our previous study @xcite were used .
the electronic structure was evaluated with the quickstep module of the cp2k package @xcite using the blyp - d3 functional @xcite and a dzvp basis .
the i - pi program @xcite was used and in the ai - pimd simulations each atom was represented by 6 beads using the piglet algorithm @xcite . for the analysis of the fluoride ion solvated in water ,
ai - pimd and aimd simulations of the ion with 64 water molecules in periodic boundary conditions with a box size of 12.414 were run for and , respectively , with a plane wave cutoff of 280 ry and settings otherwise identical to those for the liquid water simulations . to obtain the molecular charges in the condensed phase ( both neat water and hydrated fluoride )
, the full electronic structure of the box was calculated for snapshots taken from our trajectories .
configurations of the classical aimd trajectory and of one bead of the ai - pimd trajectory were analyzed every using the quickstep module of the cp2k package @xcite .
stricter electronic structure settings were used , based on thorough testing of the convergence of the charges ( see si ) .
specifically , the blyp - d3 functional was kept and the scf convergence was increased to @xmath1= , while the cutoff for the plane - wave representation of the charge density was set to 900 ry . the m - tzv2p basis set optimized for molecular systems ( molopt ) @xcite was employed and the three population analyses also used for the gas - phase systems were performed in the following way . to obtain bader charges ,
the electronic density on a grid was analyzed with the bader charge analysis program of the henkelman group @xcite .
hirshfeld charges were determined directly with cp2k .
npa charges are not available directly from cp2k , but were obtained by writing out the atomic orbital overlap and density matrix of every configuration and preparing nbo 6.0 @xcite input . to enable the decomposition of charge populations on molecules in liquid water , individual charge transfer contributions were calculated for hydrogen bonded dimers extracted from the bulk geometries . for this ,
hydrogen bonded configurations were detected by applying a cutoff on the oxygen - oxygen distance @xmath2 of 3.5 , which corresponds to the first minimum of the radial distribution function , and a cutoff on the hydrogen bond angle @xmath3 of 130 .
overall , dimer configurations were randomly extracted from the quantum ensemble , since it features a wider range of structures than the classical ensemble .
population analyses for extracted dimer configurations were performed using the blyp - d3 functional with otherwise similar specifications as for the gas - phase dimers .
we repeated the population analyses also using mp2 to ensure that the ct is not an artifact of dft , obtaining results comparable to those from dft ( see fig .
s12 in the si for details ) .
in what follows we investigate ct in several hydrogen - bonded systems over a range of temperatures in both the gas and condensed phase .
the water dimer was chosen as it is the prototypical hydrogen bond , the iodide - water complex was chosen as an intermediate strength hydrogen bond that is dominated by the charge of the diffuse and polarizable iodide and the fluoride - water complex was studied as an example with a very strong hydrogen bond .
first we consider these systems in the gas phase at the low temperature of 25 k as well as at ambient temperature to isolate the ground state contribution to ct .
second , to understand molecular charge populations in the condensed phase , we investigate ct in liquid water and around the hydrated fluoride anion .
finally , we show that it is possible to construct a simple model of ct based on proton sharing in hydrogen bonds .
for clarity , in the main text we report only npa charge populations , but results obtained with qtaim and the hirshfeld method are reported in the si .
it should be noted that although the absolute amount of charge varies with the different population analyses , the trends and conclusions we demonstrate are universal .
we start by examining the amount of charge transfer in the three gas phase complexes water - water , water - iodide , and water - fluoride at temperatures of and . at , the properties of all the systems are dominated by the vibrational ground state . hence , the influence of nqes when minimal thermal fluctuations are present can be assessed .
the results of the npa ct analysis are shown in the top 6 panels of fig .
[ fig : dimer_ct ] , which compares the classical and quantum ensembles as well as the minimum energy structure .
the ct is largest for the strongly hydrogen - bonded fluoride - water complex , less substantial for the iodide - water complex , which is of intermediate strength , and smallest for the neutral water dimer . at ,
the aimd simulations , which include only classical thermal fluctuations , show a narrow symmetric distribution of ct around an average which is very close to the value obtained for the minimum energy structure .
in contrast , when nqes are included in the description of the system using ai - pimd simulations , the distributions become substantially broader and more asymmetric with a long tail of high ct values .
this shifts the average ct away from that obtained for the minimum energy structure .
however , the direction of this shift differs for the three complexes , with the water dimer average ct decreased by 27 % , while the fluoride - water complex ct is increased by 16 % .
these results thus show that , due to the asymmetry and the shift of the average in the quantum simulations , the minimum energy structure does not provide an adequate approximation to the quantum mechanical ground state . upon increasing the temperature to 300 k , the difference between the classical and quantum ct distributions markedly decreases .
this is largely due to changes in the classical distributions , while the quantum distributions are essentially unchanged from those at 25 k , suggesting that even at 300 k , ct is dominated by the ground state contribution .
while for the water dimer the classical and quantum distributions look essentially identical , in the case of the two halide - water complexes , there are small residual differences , with the quantum average matching the minimum structure in the case of iodide and the classical average matching it for fluoride .
it is important to note , however , that given the shape and spread of the distributions , these matches should be considered entirely coincidental and should not be relied on in general .
elliptical coordinate system for the hydrogen bond reduced to the two heavy atoms ( o and x ) and the hydrogen bonded hydrogen atom ( h ) .
the two orthogonal coordinates @xmath4 and @xmath5 fully describe the hydrogen bond in this system together with the distance between the two foci of the ellipses which corresponds to the heavy - heavy atom distance @xmath6 . the close relation between the usual proton
coordinate @xmath7 and @xmath4 should be noted . ] to assess the structural changes which modulate the ct , we consider the elliptical coordinates defined in fig .
[ fig : coord ] .
these describe the position of the proton in the hydrogen bond in terms of a proton sharing coordinate @xmath4 , a dimensionless version of another commonly used proton sharing coordinate @xcite , and an orthogonal coordinate @xmath5 which describes the deviation from linearity
. a higher value of @xmath4 corresponds to a hydrogen bond with the proton closer to the acceptor heavy atom , with @xmath4=0 representing a proton being shared equally between the two heavy atoms . as shown in fig .
s6 , this coordinate strongly correlates with the amount of charge transfer .
this arises since , when the proton is more shared , the electron density shifts more towards the acceptor molecule , resulting in an increase in charge transfer . as depicted in fig .
[ fig : dimer_ct ] , the changes of the distribution of @xmath4 ( bottom 6 panels ) are thus mirrored in the behaviour of the ct distribution ( top 6 panels ) . at the lower temperature , the distribution of @xmath4 is spread out considerably upon including nqes for all systems .
in addition , for the water - dimer , nqes decrease the average value of @xmath4 , while for iodide - water it remains unchanged and for fluoride - water the average of @xmath4 is increased . at ambient temperature ,
the classical and quantum distributions become almost identical for these gas phase systems , again matching the behavior of the magnitude of ct .
correlation of npa charge transfer with sapt(2 + 3 ) charge transfer energy for the water dimer ( top ) and fluoride - water complex ( bottom ) from aimd ( classical , red ) and ai - pimd ( quantum , blue ) simulations at . ] thus far , our analysis was based on the partitioning of the electronic density to determine the amount of electrons partially transferred through the hydrogen bond . to provide a measure of how this ct affects the binding energy of the complex , we performed calculations for the water dimer and fluoride - water complexes using the symmetry adapted perturbation theory ( sapt ) method with an extension to partition the induction energy into charge transfer and polarization contributions .
[ fig : cor_npa_sapt ] shows that the charge transfer energy is strongly correlated with the amount of ct .
overall , the ct energy is in the range of @xmath804 kj mol@xmath9 for the water dimer , which corresponds to 15 % of the total binding energy .
as already implied by the much larger amount of ct , the range of ct energies for the fluoride - water complex is over one order of magnitude larger , @xmath820100 kj mol@xmath9 , and makes up roughly 30 % of the total binding energy .
owing to this correlation , the ct energies are influenced by nqes in a way analogous to the amount of ct and the proton sharing coordinate .
we now consider molecular ct in the condensed phase for ambient temperature liquid water as well as the hydrated fluoride anion .
the distributions of molecular npa charge populations for these two systems with a classical and quantum description of the nuclei are depicted in fig .
[ fig : liquid_ct ] , while results for the other population analyses are shown in fig . s9 . in the case of the fluoride anion , ct only occurs in one direction electronic density is transferred from the fluoride anion to the surrounding water molecules through the hydrogen bonds that they donate to the ion .
any deviation of charge population on the fluoride anion from the formal charge of -1 is thus the amount of ct due to all hydrogen bonds it receives in the liquid .
the hydrated fluoride anion in the liquid phase has an average ct of 0.19 e in the classical case .
this is only @xmath850 % higher than for the monohydrated anion in the gas phase , despite the increase in the number of hydrogen bonds it receives from 1 to an average of @xmath84.55.0 .
nqes increase the average ct by 11 % ( slightly less than the 18 % observed in the gas phase ) and markedly increase the range of values observed towards higher charge transfer .
water molecules , unlike the fluoride anion , can both accept and donate hydrogen bonds .
due to the symmetry of the hydrogen bonding in the bulk liquid , the distribution of molecular charges is symmetric and has an average of zero and thus only the fluctuations can vary .
our simulations show that nqes increase these fluctuations around zero charge the quantum distribution in the top panel of fig .
[ fig : liquid_ct ] is considerably wider than the classical distribution .
distribution of the molecular charge in liquid water ( top ) and charge transfer of the fluoride - ion solvated in water ( bottom ) obtained with the npa method from aimd ( classical , red ) and ai - pimd ( quantum , blue ) simulations at .
the standard deviation ( top ) / average ( bottom ) of each distribution is marked with a solid line and stated in the respective color . ]
based on our observations above , we now demonstrate how one can provide a simple but accurate model of ct through the hydrogen bond .
previous investigations of charge transfer between water molecules in the liquid phase have modelled the molecular charge distribution by using a procedure in which formation of a hydrogen bond contributes a constant amount of charge transfer @xcite . in these studies ,
each molecule was assigned a hydrogen bond donor - acceptor balance @xmath10 , calculated as the difference between the total number of donor @xmath11 and the total number of acceptor hydrogen bonds @xmath12 of a given molecule , @xmath13 .
this direct charge transfer model then gives a linear relationship between @xmath10 and the charge population on the molecule @xcite .
however , as shown in ref . and in fig .
s10 for our liquid water trajectories , while the presence of a hydrogen bond shifts the charge population on a water molecule , that alone is not sufficient to explain the ct , as seen from the width and overlap of the distributions in fig .
indeed , using this criterion there would be essentially no change expected in the ct upon including nqes because the number of hydrogen bonds is only minimally perturbed .
this suggests that the ct is determined by other parameters not captured by the binary criterion of hydrogen bond existence .
motivated by the strong dependence of charge transfer on the proton sharing coordinate @xmath4 in our gas - phase results ( fig .
[ fig : dimer_ct ] and fig .
s6 ) we show how incorporating this coordinate in a condensed - phase model of ct yields much improved predictions .
the upper panel shows the correlation of the proton sharing coordinate @xmath4 with the charge transfer obtained with the npa method for extracted dimers of the ai - pimd simulation at , where the probability is given in blue in equidistant steps and the median of the distribution is plotted as a dashed line , while a fit to the data is shown dotted in blue .
the lower panel shows the charge transfer as a function of @xmath4 and the orthogonal coordinate @xmath5 again in equidistant steps over the charge transfer range given by the upper panel .
the dotted black lines spanning both panels should illustrate how the color map of the lower panel is to be interpreted . ]
to parametrize such a model for ct we begin by extracting dimers from our quantum simulations , which were chosen as they sample a wider range of ct values ( see top panel of fig . [
fig : liquid_ct ] ) .
we then calculated the ct for each of these dimers separately to quantify the ct between that pair of water molecules .
the resulting ct values obtained are plotted against @xmath4 in the top panel of fig .
[ fig : fit ] . as for the dimer configurations obtained from the gas phase simulations ,
these show a strong correlation of the ct with @xmath4 . on the other hand
, the ct shows no dependence on the coordinate @xmath5 orthogonal to @xmath4 , as shown in the bottom panel of fig .
[ fig : fit ] , suggesting that including this coordinate in the ct model would offer little benefit .
this relationship can be used to determine the expected amount of ct given a value of @xmath4 . to fit the dependence we used the following simple functional form that uses only three parameters @xmath14 where the values of the parameters @xmath15 , @xmath16 and @xmath17 listed in tab . [
tab : para ] were determined by a least squares fit to the data . when this function is used to predict the ct of the extracted dimer set , a correlation of the predicted charge with the charge from the population analysis of 96.4 % is observed .
this high correlation coefficient shows that the other degrees of freedom account only for under 4 % of the ct and that a simple functional form is thus able to describe the dependence appropriately .
correlation of the predicted molecular charge using the obtained relation between charge transfer and the instantaneous value of the proton sharing coordinate @xmath4 with the calculated molecular charge for the npa method of the pi - aimd simulations of liquid water at as well as the corresponding distributions of the molecular charge with stated standard deviation . ] with this model , the total charge on each molecule in the bulk liquid can be predicted by identifying the hydrogen bond donors and acceptors of each molecule and their @xmath4 values .
one can then calculate the charge @xmath18 of water molecule @xmath19 by summing the incoming and outgoing ct contributions , @xmath20 where @xmath21 is obtained from eq .
[ eq : ct ] .
[ fig : cor_pred_calc ] shows the predictions from this simple model compared to the molecular charges obtained from a full electronic structure calculation for the quantum liquid water ensemble .
excellent agreement is obtained for the npa data shown , with the model giving slightly smaller molecular charges than the full electronic structure calculations . for npa
, the prediction works the best with a correlation coefficient of 95.2 % , while qtaim and the hirshfeld method yield correlation coefficients of 86.2 % and 83.6 % .
the prediction for the classical ensemble shows comparably high correlation coefficients of 93 % , 80 % and 81 % , respectively .
this shows that the prediction of molecular charges based solely on the proton sharing coordinate @xmath4 in hydrogen bonds is a good approximation for the total charge population in the liquid phase . @ l * 3d @ & & & + @xmath15 & 0.6254 & 0.2107 & 0.4872 + @xmath16 & -0.5932 & -0.6464 & -0.6526 + @xmath17 & 2.5937 & 1.8677 &
1.4376 + since we have demonstrated that ct and the resulting molecular charge in liquid water are almost entirely dictated by a single geometric parameter of the hydrogen bond , we can now elucidate the difference between the quantum and classical ensembles .
[ fig : tau_bulk ] shows the distributions of this geometric parameter , the proton sharing coordinate @xmath4 for liquid water .
although the classical and quantum ensembles match closely at more negative values of the coordinate , as one approaches zero , corresponding to a highly shared proton , the quantum distribution exhibits significantly higher probability density .
this increase in proton sharing , observed in a number of previous studies @xcite , is vital in this case , as it solely determines the amount of ct .
this is in stark contrast to many other properties which exhibit a cancellation between quantum effects along @xmath4 and the orthogonal coordinate @xmath5 @xcite .
in conclusion , by comparing quantum and classical simulations over a wide range of hydrogen bonded structures in both the gas and condensed phases we have shown that nqes can make large qualitative and quantitative changes to the magnitude of ct .
we have demonstrated that these results are robust with respect to the choice of electronic structure method and population analysis approach . in particular
, the insensitivity of ct to the coordinates orthogonal to the hydrogen bond direction means that the higher ct due to increased proton sharing is not compensated by decreases due to other degrees of freedom . due to the large amount of zero point energy in the @xmath4
coordinate in the quantum case , the additional thermal energy at the higher temperature only changes the position distribution by a small amount , in contrast to the classical case .
this leads to a very small temperature dependence of the quantum ct distribution compared to that observed classically .
as such , one can easily be misled by considering classical configurations or minimum energy structures alone when assessing the amount of ct .
further , we were able to show that it is possible to accurately model the charge population on a water molecule in the condensed phase by considering only the sharing of protons with its immediate hydrogen bonding neighbors .
this approach is able to explain 8095 % of the charge population on a given molecule even though it lacks higher body effects , showing that the wider environment plays only a minor role .
indeed , recent work has shown that 2-body effects are sufficient to model the dipole moment surface of liquid water @xcite , further suggesting that the electrostatics of water are well captured at this level .
our analysis thus provides physical insight into the origin of charge transfer in hydrogen bonded systems as well as a simple basis for its accurate modelling .
this material is based upon work supported by the national science foundation under grant no .
t.e.m also acknowledges support from a cottrell scholarship from the research corporation for science advancement .
we would like to thank stanford university and the stanford research computing center for providing computational resources and support that have contributed to these research results .
c.s . gratefully acknowledges financial support from a scholarship from the german academic scholarship foundation ( _ studienstiftung des deutschen volkes _ ) .
|
upon hydrogen bond formation , electronic charge density is transferred between the donor and acceptor , impacting processes ranging from hydration to spectroscopy . here
we use ab initio path integral simulations to elucidate the role of nuclear quantum effects in determining the charge transfer in a range of hydrogen bonded species in the gas and liquid phase .
we show that the quantization of the nuclei gives rise to large changes in the magnitude of the charge transfer as well as its temperature dependence .
we then explain how a single geometric parameter determines the charge transfer through the hydrogen bond .
these results thus demonstrate that nuclear quantum effects are vital for the accurate description of charge transfer and offer a physically transparent way to understand how hydrogen bonding gives rise to it .
|
among substellar objects , the relationship between giant planets and brown dwarfs is unclear . the generally acknowledged dividing line between these two classes of objects
is based on mass .
specifically , objects above the minimum mass to burn deuterium are defined to be brown dwarfs , whereas objects less massive than this limit are defined to be planets .
the deuterium burning limit is roughly @xmath813@xmath9 , although in detail this depends on one s definition of `` burning deuterium , '' and on the detailed composition of the object @xcite .
lcc @xmath10&mass ( @xmath11 ) & @xmath12 + @xmath13&radius ( @xmath14 ) & @xmath15 + @xmath16&effective temperature ( k ) & @xmath17 + @xmath18&period ( days ) & @xmath19 + @xmath20&mass ( @xmath9 ) & @xmath21 + @xmath22&radius ( @xmath23 ) & @xmath24 + @xmath25&eccentricity & @xmath26 + @xmath27&argument of periastron ( degress ) & @xmath28 + @xmath29 & & @xmath30 + @xmath31 & & @xmath32 + @xmath33&surface gravity & @xmath34 + @xmath35&equilibrium temperature ( k ) & @xmath36 + @xmath37&incident flux ( 10@xmath38 erg s@xmath39 @xmath40 ) & @xmath41 + @xmath42&radius of the planet in stellar radii & @xmath43 + @xmath44&semi - major axis in stellar radii & @xmath45 + @xmath46&inclination ( degrees ) & @xmath47 + @xmath48&impact parameter & @xmath49 + @xmath50&transit depth & @xmath51 + @xmath52&fwhm duration ( days ) & @xmath53 + @xmath54&ingress / egress duration ( days ) & @xmath55 + @xmath56&total duration ( days ) & @xmath57 + [ tab : kelt-1b.finalfinal.median ] on the one hand , distinguishing between objects below and above 13@xmath9 is clearly arbitrary , particularly since after roughly a billion years deuterium burning is over and any evidence of this initial internal energy source is largely gone , i.e. , an old @xmath850@xmath9 object that never fused deuterium would be difficult to distinguish from one that did @xcite .
therefore , giant planets and brown dwarfs can properly be thought of as a continuum of objects , with masses and surface gravities that vary accordingly . by studying how the observable properties of these objects vary as a function of mass and surface gravity for controlled samples with similar compositions and in similar environments , we can gain insight into the uncertain physics at work in these bodies .
such insights will in turn constrain the origin of these bodies . particularly important in this regard are constraints on the atmospheric systems of brown dwarfs and giant planets , as these systems present not only the most uncertain physics , but are also the most amenable to empirical constraints .
on the other hand , giant planets and massive brown dwarfs likely have distinct origins , at least for companions to sunlike stars .
this is evidenced by the existence of the brown dwarf desert , the local minimum in the mass function of relatively close - in ( @xmath5810au ) companions to sunlike stars near @xmath830 to @xmath850@xmath9 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
presumably , objects below the brown dwarf desert were formed in a circumstellar disk , whereas objects above were formed in a manner more analogous to stars . however , this hypothesis is relatively untested , and even within this interpretation many questions remain .
for example : what is largest - mass object that can form in a circumstellar disk ? what is the smallest mass companion that can form like a star ?
do these masses overlap ? does the brown dwarf desert depend on the properties of the star , or the separation from the star ?
why is there apparently no brown dwarf desert for isolated objects ? by better understanding the physics of companions from the giant planet up through the brown dwarf regime , we may better understand their origins and thus provide answers to these questions .
unfortunately , obtaining empirical constraints on giant planets and brown dwarfs in similar environments has proven difficult .
the majority of our empirical constraints on brown dwarfs come from isolated brown dwarfs , brown dwarf binaries , or brown dwarfs as wide companions to stars @xcite .
these systems are often amenable to detailed study of their atmospheres , including spectra and time series photometry . however , in the vast majority of cases , these objects do not have masses , radii , or age measurements .
one exception is the eclipsing brown dwarf system 2m0535 - 05 @xcite , which allows for a direct measurement of the masses and radii of the two brown dwarf components .
furthermore , analogous isolated or wide - separation giant planets are considerably more difficult to study , due to their intrinsic faintness .
it is only recently , and only for relatively massive and young planetary - mass objects , that the first such empirical constraints have been obtained ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* among others ) .
as expected , obtaining the kinds of measurements routinely acquired for more massive brown dwarfs for these first planetary - mass objects has already provided important insight into the physics at work in these bodies .
for example , consider the the strong j - band brightening that brown dwarfs undergo as they cross the l- to t - dwarf transition .
this is widely interpreted as clouds clearing from the atmospheres of the l - dwarfs @xcite , though the precise mechanism for this process is not well understood @xcite .
direct imaging observations of the hr 8799 planets , which are giant planets at or past the l / t transition , do not show a similar j - band brightening @xcite .
this indicates that the atmospheric dynamics responsible for the l / t transition is highly dependent on surface gravity @xcite .
if we are able to understand what the differences are between atmospheric flows and forcing mechanisms in brown dwarfs and exoplanets , we may be able to understand what is precisely occurring at the l / t transition .
in contrast to brown dwarfs , the majority of our empirical constraints on giant planets comes from transiting systems .
these systems provide masses , radii , and crude ages for most systems , but because of selection biases , nearly all these planets are on short periods and so likely are tidally locked and subject to very strong stellar irradiation , dramatically altering and complicating their atmospheres and atmospheric dynamics .
therefore , the empirical constraints on these systems can not be directly interpreted in the same context as isolated brown dwarfs , hampering the ability to define the relationship between these two types of objects .
the existence of the brown dwarf desert , and the resulting paucity of transiting brown dwarfs , has prevented any direct comparison of observations of brown dwarfs under similarly irradiated environments as close - in giant planets .
kelt-1b @xcite provides the best opportunity to directly compare a brown dwarf to giant exoplanets under the same environmental condition of strong external irradiation .
the previously discovered transiting brown dwarf companions to stars , corot-3b @xcite , corot-15b @xcite , kepler-39b @xcite , wasp-30b @xcite , koi-415b @xcite , koi-205b @xcite , and lhs 6343c @xcite , all orbit relatively faint stars ( with the brightest being wasp-30 at @xmath59 and the others being significantly fainter ) . on the other hand
there are massive transiting giant planets like hat - p-2b @xcite , xo-3b @xcite , and wasp-18b @xcite , all with masses around 10 @xmath9 , that transit stars bright enough to allow for high quality follow - up observations , but their masses place them within the planetary regime .
kelt-1b is a 27@xmath9 object on a short 1.2 day orbit around a bright ( @xmath60 ) f5v star .
the close orbit places kelt-1b in a highly irradiated environment , with an incident stellar flux of @xmath61 erg s@xmath39 @xmath40 , that places it forty times above the empirical threshold for inflated planets determined by @xcite and @xcite .
in addition , based on the expected tidal synchronization timescale of @xmath810myr for kelt-1b @xcite we expect that kelt-1b is tidally locked to its orbital period .
kelt-1b thus allows us to study a brown dwarf where we know the mass and radius , in an irradiation and tidal environment similar to hot jupiters , and around a star bright enough to allow for precision follow - up observations . to take advantage of the opportunity offered by kelt-1b
, we observed several secondary eclipses of the kelt-1 system from the ground and from space during the fall of 2012 .
-0.0 in -0.0 in we observed secondary eclipses of kelt-1b in the 3.6@xmath1 m and 4.5@xmath1 m bands using the irac instrument on the spitzer space telescope .
we took the 3.6@xmath1 m observations over the course of ut 2012 september 10 and 11 and the 4.5@xmath1 m observations during ut 2012 september 11 and 12 .
both sets of observations lasted for 6.57 hours and used the subarray mode with 2.0 second exposures .
we additionally used the spacecraft s peak - up mode , with kelt-1 as the peak - up target , to enhance the pointing stability .
we executed two observing sequences in each band : the first was 0.5 hours long intended to allow for the detector ramp and for the spacecraft to settle its pointing .
we discarded this first observing sequence , since it was dominated by these two effects .
the second sequence in each band was the main science observing sequence .
this lasted for 6.07 hours and provided us with 10,944 images in each band .
we used the basic calibrated data ( bcd ) images from these two sequences for all our photometry .
-0.2 in -0.0 in we calculated the @xmath62 time of each image by using the fits header entries as follows .
the bcd images come in 64 image data cubes with header keywords that record the start time ( mbjd_obs ) for each sequence containing 64 images , and the total sequence duration ( aintbeg and atimeend ) .
we calculated the @xmath63 time at mid - exposure for each of the 64 images in each data cube by assuming the image sequences began at mbjd_obs and that the 64 images in each sequence were evenly spaced between aintbeg and atimeend . to adhere to the timing system used in the kelt-1 discovery paper , we converted to @xmath62 by adding 64.184 seconds to the derived @xmath63 times @xcite .
to calculate the background in each of the 32@xmath6432 pixel subarray images , we first excluded the light from the kelt-1 system by masking out a central circular area 12 pixels in radius . on each image , we then performed three rounds of 3@xmath65 clipping on the remaining non - masked area to remove outliers .
we estimate the background flux for each image by fitting a gaussian to a histogram of the values of the remaining pixels and used the fitted value for the mean as the background flux .
we subtracted this fitted mean background value from each image .
the mean background flux averaged over all the images used in the analysis was 0.05% of kelt-1
s flux at 3.6@xmath1 m and 0.02% at 4.5@xmath3 m .
we extracted lightcurves in each band using simple aperture photometry . to determine the position of the star in our images we fit a two - dimensional gaussian to the stellar psf using the entire image .
we also tried finding the star s position using flux - weighted centroiding following @xcite , but later found this provided inferior corrections for intrapixel sensitivity variations in our data . in doing the position determinations we used a modified set of our background - subtracted images
: we replaced any hot pixels with median flux values for that pixel to prevent spurious centroid shifts .
we identified pixels as `` hot '' in a particular image if their flux in the image was more than 3@xmath65 away from the median flux for that pixel over an entire 64 image data cube .
we then replaced the hot pixel with the pixel s 64 image median flux . for our aperture photometry
we used the original background subtracted images without the hot pixels replaced , so as to remain as close to the raw data as possible .
instead , we used 5@xmath65 clipping on the aperture photometry results to remove images where a hot pixel occurred within the photometric aperture .
-0.0 in -0.0 in -0.2 in -0.0 in to extract photometry from the images we chose to use variable apertures scaled to the fwhm of the stellar psf for both the 3.6@xmath1 m and 4.5@xmath1 m data .
other secondary eclipse measurements with spitzer usually use a variable aperture for 3.6@xmath1 m data and a fixed aperture for their 4.5@xmath1 m observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , but we found that using a variable aperture for both bands gave the lowest rms scatter in the residuals to the best fit lightcurves .
we estimated the width of the stellar psf in each image using the noise pixel parameter ( following * ? ? ?
* ) , which is defined in section 2.2.2 of the irac instrument handbook as @xmath66 where @xmath67 is the intensity of the @xmath46th pixel . to calculate the noise pixel parameter in our data we summed over the pixels within a radius of three pixels of kelt-1 s position in the image , including fractional pixels
the fwhm of the stellar psf is then @xmath68 @xcite .
the third panels in figures 1 and 2 show how the fwhm calculated from the noise pixel parameter ( i.e. @xmath68 ) varied as a function of time . at 3.6@xmath1 m
the median fwhm over the 9,775 images we used was 2.15 pixels , with a standard deviation of 0.04 pixels . at 4.5@xmath1 m
the median over the 10,307 images we used was 2.03 pixels with a standard deviation of 0.02 pixels . at 3.6@xmath1 m
the noise pixel parameter was nearly perfectly ( @xmath69 ) correlated with the y - position of the stellar centroid .
we set our photometric aperture size to be @xmath70 , which is the fwhm , @xmath68 , in a particular image plus some constant , @xmath71 .
we chose the optimum aperture by extracting photometric timeseries for a range of values of @xmath71 , fitting an eclipse model to each of these lightcurves , and then choosing the value of @xmath71 which resulted in the lowest rms residuals with respect to the model .
we tested values @xmath71 from @xmath72 pixels to @xmath73 pixels in steps of 0.05 .
this roughly corresponds to apertures with radii of 1.6 to 4.1 pixels ( figures 3 and 4 ) , but we remind the reader that @xmath68 varies with time .
the approximate radii plotted in figures 3 and 4 use the median values for @xmath68 over all images .
we fit the photometry using the first amoeba stage of the fitting procedure described in section 4 .
figures 3 and 4 show the fitted depth ( black ) and residual rms ( red ) as a function of aperture size for the 3.6@xmath1 m and 4.5@xmath1 m data .
the lowest residual rms occurred for an aperture size of @xmath74 pixels in the 3.6@xmath1 m data and @xmath75 pixels in the 4.5@xmath1 m data .
we therefore utilized these aperture sizes to extract the photometry that we employ in our final analysis .
the standard deviation on fitted depth as a function of aperture size in the 3.6@xmath1 m data was 0.003% over the entire trial range , which is below the 0.011% final uncertainty we find for the 3.6@xmath1 m eclipse .
for the 4.5@xmath1 m data this was not the case , so we considered the behavior of this dataset in more detail . at 4.5@xmath1
m we found the fitted eclipse depth varied by 0.06% , depending upon the aperture sized used to extract the photometry ( figure 4 ) , particularly for apertures smaller than @xmath76 .
this is significantly above the 0.012% final uncertainty we calculate for the 4.5@xmath1 m data .
there is also a `` bump '' in the rms and fitted depth around an aperture radius of @xmath77 . for reference
, we refer the reader to figure 6 of @xcite for an example of `` well - behaved '' 4.5@xmath1 m data .
we were not able to satisfactorily determine the cause of the variation , or the reason for the bump .
we first tried switching the lightcurve extraction to use a non - scaled aperture size , instead of an aperture scaled to the fwhm of the stellar psf , but these lightcurves showed a similar variability in the eclipse depth and a `` bump '' at a radius of 2.3 pixels .
next , we tested to see if the bump was caused by a bad pixel by setting individual pixels in all the images to zero one - by - one . by setting pixels ( 14,14 ) , ( 14,15 ) and
( 14,16 ) to zero we were able to remove the `` bump '' , but this almost doubled the variability in the fitted eclipse depth for the photometry using both non - scaled and scaled aperture sizes .
a visual inspection of these three pixels timeseries showed no obvious abnormalities . zeroing out other pixels had no discernible effect .
we ultimately decided to use an aperture of @xmath75 pixels to extract the 4.5@xmath1 m photometry from the unaltered bcd images . in figure 4
one can see that for apertures larger than @xmath76 ( @xmath78 ) pixels the fitted eclipse depth is nearly constant , and the lowest residual rms occurs at @xmath75 .
we therefore judged that whatever the cause of the systematic changes in the depth and rms variation for smaller aperture size is mitigated for apertures larger than @xmath76 , and thus , that the photometry at our chosen aperture of @xmath75 is representative of kelt-1b s true eclipse depth at 4.5@xmath3 m .
if this is not the case , then the uncertainty on the 4.5@xmath1 m depth we report is an underestimate .
we also trimmed out some of the initial images due to the remains of the initial photometric ramp .
the first 1,000 images of the 3.6@xmath1 m data and the first 600 4.5@xmath1 m images displayed a clearly discernible ramp feature , so we excluded them from our analysis . finally , we removed points that were more than 5@xmath65 away from the median flux .
the number of points that were clipped varied slightly depending on the exact aperture size ( i.e. the value of @xmath71 ) used . in the apertures we chose to use for our final analysis ,
this clipping removed 169 ( 1.7% ) 3.6@xmath1 m and 37 ( 0.4% ) 4.5@xmath1 m images .
for our chosen photometric aperture for each data stream , our final lightcurve contained 9,775 images at 3.6@xmath1 m covering -2.56 to + 3.01 hours around the center of secondary eclipse .
the final 4.5@xmath1 m lightcurve contained 10,307 images and spans from -2.98 to + 2.82 hours around the center of secondary eclipse .
over the summer and fall of 2012 we observed seven secondary eclipses of kelt-1b in @xmath4 at moore observatory , which is operated by the university of louisville .
we used the 0.6 m rcos telescope with an apogee u16 m 4k@xmath644k ccd , giving a @xmath79@xmath64@xmath79 field of view and a plate scale of @xmath80 pixel@xmath39 .
since kelt-1 is separated from its nearest detectable neighbor in dss2 imagery by @xmath8@xmath81 , we were able to defocus the telescope to allow for longer exposures without the risk of blending from the neighbor star .
we used the same observing parameters for the ground - based observations across all nights .
the exposure time was 240 seconds ( plus a 20 second readout time ) , and we slightly defocused telescope to give a toroid shaped point spread function ( psf ) .
the target and comparison stars were placed at the same detector locations , and the guiding maintained this placement within a few pixels across all nights .
our image calibration consisted of bias subtraction , dark subtraction , flat - field division , and detector non - linearity compensation .
we extracted differential aperture photometry from the calibrated images using astroimagej ( aij , k. a. collins , et al .
, in preparation ) .
the comparison stars were selected from sources on the detector which had @xmath4 band brightness similar to kelt-1 and which produced relatively flat light curves ( after airmass detrending ) when compared to the other stars in the ensemble .
the final comparison ensemble included four stars near kelt-1 ( tyc 2785 - 2151 - 1 , tyc 2781 - 2231 - 1 , ltt 17089 , and tyc 2785 - 1743 - 1 ) .
we chose to allow the photometric aperture radius to vary based on an aij estimate of the fwhm of the toroidal psf .
after testing values in the range of 1.0 to 1.4 times the estimated fwhm , we found that a factor of 1.25 minimized the scatter in the light curves .
this factor resulted in an aperture radius that varied between 20 - 30 pixels across the four nights .
the sky background was estimated from an annulus with inner radius 40 pixels and outer radius 80 pixels .
iterative 2@xmath65 clipping was first performed to remove outliers and stars from the background annulus .
the mean of the remaining pixels was adopted as the sky background value and subtracted from each pixel in the photometric aperture .
three of the events , on ut 2012 september 7 , 2012 october 5 and 2012 october 12 , suffered from abnormally poor seeing or interruptions by clouds .
we excluded these observations from consideration .
the other four secondary eclipses , on ut 2012 july 30 , 2012 august 16 , 2012 november 18 and 2012 november 29 , were high - quality , complete observations of the eclipses . the typical per point uncertainties on these nights were 0.10% to 0.13% .
the top four panels in figure 9 show the lightcurves from these four good nights plotted individually , after being detrended against airmass and time as described in section 3.3.2 .
we modeled the ir data as a combination of a @xcite eclipse lightcurve and a set of decorrelation parameters . to make the eclipse lightcurves we used the implementation of the @xcite lightcurves built into exofast @xcite .
we modeled the eclipse by assuming kelt-1b was a uniformly bright disk , with no limb - darkening , being occulted by the much larger kelt-1 .
compared to a transit lightcurve , this has the immediate effect that @xmath82 and the eclipse depth are no longer directly related .
-0.0 in -0.0 in 0.08 in in both channels the data showed strong correlations between the flux and the @xmath83- and @xmath84-position of the star .
these correlations persist regardless of the aperture size we used for the data reduction .
these light curve systematics are a result of the well - known intrapixel sensitivity variation in the 3.6@xmath1 m and 4.5@xmath1 m detectors ( e.g. , * ? ? ?
we fit for , and removed , these trends by including a decorrelation function that modifies the flux and contains three terms : a linear term each for the x- and y - pixel position of kelt-1 , and a linear time term .
we included this additional linear decorrelation against time as our initial fits using only the positional decorrelation showed a clear residual trend with time .
we considered using additional quadratic decorrelation terms in @xmath83- and @xmath84-position but found that the corresponding decorrelation coefficients varied substantially with choice of the aperture size used to extract the light curves ( figures 5 and 6 ) .
the linear terms , on the other hand , settled to specific values once the aperture size grew larger than approximately 2.6 pixels .
we therefore considered the quadratic terms to be poorly unconstrained by the data .
furthermore , we found that the fitted eclipse depth was strongly correlated with the quadratic fit parameters , such that the eclipse depth was artificially suppressed when these terms were included .
this is due to the relatively small amount of out - of - eclipse data , and the fact that the @xmath83- and @xmath84-position measurements are strongly correlated with time , which allowed the quadratic position terms to `` fit out '' the transit without significantly worsening the fit outside of eclipse .
-0.0 in -0.0 in the model we fit to the data was therefore @xmath85\end{aligned}\ ] ] where @xmath86 is the decorrelation function which describes the variation of the unocculted total flux due to systematic effects , and @xmath87 describes the fractional flux decrement during the eclipse as computed using the modified @xcite model .
in addition to a time dependence , the eclipse model s exact form depends upon : the time of the previous transit ( @xmath88 ) , the orbital period @xmath18 , @xmath89 , @xmath90 , the cosine of the orbital inclination ( @xmath91 ) , the radius of the planet in stellar radii ( @xmath82 ) , the semi - major axis in units of the stellar radii ( @xmath92 ) , the baseline flux level ( @xmath93 ) , and the eclipse depth ( @xmath50 ) .
note that we do not explicitly fit for the time of secondary eclipse .
instead , we calculate the secondary eclipse time based on the time of the previous transit ( @xmath88 ) , the orbital period @xmath18 , and @xmath89 and @xmath90 .
we begin by determining the eccentricity and orientation of the orbit via @xmath94 and @xmath95 .
this allows us to calculate the mean anomaly of kelt-1b during transit ( @xmath96 ) and eclipse ( @xmath97 ) .
then the eclipse time is @xmath98 we explain the motivation for using this parameterization below .
we fit for a total of twelve parameters : the nine eclipse parameters and the three decorrelation parameters . for seven of these parameters ( @xmath88 , @xmath18 , @xmath89 , @xmath90 , @xmath91 , @xmath82 , and @xmath44 )
, we had a prior expectation for their values from the kelt-1b discovery paper .
we did not have any prior expectations for the five remaining parameters ( @xmath93 , @xmath50 , and the decorrelation terms ) . to incorporate our priors into the fitting process
, we added a term for each parameter to the @xmath99 function of the form @xmath100 here @xmath101 is the trial value of an individual parameter @xmath102 , @xmath103 is the prior value , and @xmath104 is the @xmath105 uncertainty in that prior value .
note that this does not consider any possible covariance between the parameters .
we used central values and @xmath105 uncertainties for @xmath88 , @xmath18 , @xmath89 , @xmath90 , @xmath91 , @xmath82 , and @xmath44 from the discovery paper fit that was based on a free eccentricity ( see tables 4 and 5 in @xcite ) , which we list for convenience in table 1 . to calculate @xmath88 , the time of the previous transit
, we assumed no variation in the transit times , such that @xmath106 and @xmath107 .
we also derive values and uncertainties for @xmath89 and @xmath90 by using the values and uncertainties for @xmath25 , @xmath108 and @xmath109 in tables 4 and 5 of @xcite and dividing the two latter quantities by @xmath110 .
we chose to use @xmath89 and @xmath90 as the prior parameters in our mcmc fits , rather than @xmath108 and @xmath109 , as the former parameterization results in a uniform prior for @xmath25 , while the latter leads to a prior that is proportional to @xmath25 .
we chose to use @xmath88 , @xmath18 , @xmath89 and @xmath90 to calculate the secondary eclipse time , instead of using the predicted time of secondary eclipse , @xmath111 , from table 5 of @xcite , so that we could properly allow for the possibility of a non - zero eccentricity and calculate appropriate uncertainties . if we were to only fit for the time of secondary eclipse @xmath111 , and calculate the orbital eccentricity based on the eclipse time and duration
, then we would incorrectly be assuming a circular orbit in our modeling of the eclipse orbital geometry and lightcurve .
this would mean that our priors on @xmath91 and @xmath44 , the two terms that would then completely set the eclipse duration , would potentially dominate our measurement of the eclipse duration and thence @xmath90 .
on the other hand , if we allowed for eccentric eclipse geometries in the lightcurve modeling by instead fitting for @xmath111 , @xmath89 and @xmath90 we would be double - counting the uncertainties in @xmath89 and @xmath90 ; the uncertainty on @xmath111 in table 5 of @xcite already includes the uncertainties in @xmath112 , @xmath89 and @xmath90 .
using @xmath88 , @xmath89 and @xmath90 therefore allows us to correctly model possible eclipse geometries and compute proper uncertainties .
note again , though , that this makes the assumption that kelt-1b has a fixed orbital period and does not display transit or eclipse timing variations .
we do account for the 25 second rmer delay the eclipse has relative to the transit ephemeris .
as we have no prior expectation for the values of @xmath93 , @xmath50 , and the decorrelation terms ( @xmath113 and @xmath114 ) , we do not include a @xmath99 penalty for these terms , and implicitly assume a uniform prior for all five parameters .
of the twelve fitting parameters , these five are therefore the only ones set entirely by our secondary eclipse observations .
-0.0 in -0.0 in among the remaining seven terms , all of which had well defined priors , @xmath89 and @xmath90 are the only ones for which the data provide a tighter constraint than provided by their prior distributions . @xmath88 and @xmath18 are completely determined by their priors , as the data from an individual eclipse provides no constraint on the nearest transit time or the orbital period . similarly , our posterior constraints on @xmath91 , @xmath82 , and @xmath92 are dominated by our priors .
this is a result of the fact that , for an eclipse lightcurve , the only constraint on @xmath82 comes from the ingress and egress durations , in contrast to a transit lightcurve , where @xmath82 is heavily determined by the depth of the transit .
this means that the two major timing measurements an eclipse lightcurve provides ( the fwhm and ingress / egress durations ) now must be used to constrain three different parameters . as a consequence , unique values for @xmath91 , @xmath82 , and @xmath92 are poorly constrained by the secondary eclipse data alone , and their values and uncertainties remain nearly identical to those from the discovery paper . in principle this degeneracy is resolved by the exact shape of the ingress and egress portions of the lightcurve , but this is below the precision of our data .
we fit our data using the amoeba and mcmc routines packaged with exofast .
we chose to fit the 3.6@xmath1 m and 4.5@xmath1 m data separately .
our fitting process began by using amoeba to find initial @xmath99 minima to use as estimates of the best fits .
this is the same procedure we used previously , to fit the data for our determination of the optimum photometric aperture size , and allowed us to quickly find a likely minimum in the @xmath99 surface .
we then used mcmc to explore the parameter space around the @xmath99 minimum found by amoeba to determine uncertainties .
exofast uses the differential evolution markov chain implementation of the mcmc algorithm , which runs multiple , simultaneous , chains to determine the correct step size and direction .
in addition to verifying that we had found the global @xmath99 minimum , the mcmc analysis also provided us with appropriate uncertainties for all of the system and eclipse parameters .
the final system parameters and errors determined by the mcmc analyses are in tables 2 and 3 .
figure 7 shows our best fits to the detrended data .
-0.0 in -0.0 in we next conducted a prayer bead analysis on our data to assess the effects of correlated noise in the data .
the prayer bead analysis we conducted followed the general description of @xcite and @xcite . in each band
we took the residuals to our best fit model , shifted the residuals by one , added the incremented residuals back onto the original best fit model and then refit using amoeba .
we shifted through the residuals to the entire lightcurve this way , such that the @xmath46th model point had the @xmath46th+@xmath115th residual added to it , with @xmath115 going from one to the number of points in the lightcurve .
the remainder at the temporal end of the lightcurve was looped around and added to the beginning .
the goal of the prayer bead analysis is to appropriately account for the presence of correlated noise in the data .
this is done by examining the variation in the fitted parameters as a function of the shifting residuals .
for both the 3.6@xmath1 m and 4.5@xmath1 m observations the standard deviations in the lightcurve and decorrelation parameters output by the prayer bead chains were within 10% of the 1@xmath65 error bars from the mcmc analysis .
we therefore chose to use the mcmc errors as our final uncertainties for the spitzer observations .
another possible source of systematic uncertainty in our observations is the possible stellar companion to kelt-1 discovered by @xcite .
the companion is located 558 mas to the southeast of kelt-1 , and has @xmath116 and @xmath117
. this separation places the companion 0.47 pixels away from kelt-1 in our spitzer observations and 1.4 pixels away in our ground - based observations . in both cases
it is unresolved . assuming that the companion is associated with kelt-1 , its luminosity difference and its @xmath118 colors imply that it is a mid m - dwarf .
if we extrapolate from @xmath119 and @xmath120 to the spitzer bandpasses using a kurucz - based 6500k , @xmath121 , spectrum for kelt-1 and a 3200k blackbody spectrum for the companion , then we find that the companion contributes less than 1% of the total system light at 3.6@xmath1 m and 4.5@xmath3 m . in both bands
this is substantially below the fractional uncertainty we calculate for the measured eclipse depths ( about 10% in both cases ) .
this is also faint enough that the companion does not affect our measurements of kelt-1 s pixel position in the images .
we therefore chose to ignore its contribution to our observations .
-0.0 in -0.0 in we analyzed each of the four nights individually .
we fit only for the depth of a possible eclipse using a trapezoidal eclipse model that had the eclipse time , total duration and ingress / egress duration fixed .
we computed the expected eclipse times for each night by extrapolating from our measured 3.6@xmath1 m eclipse time and assuming a fixed period of 1.217514 days .
the choice of the 3.6@xmath1 m eclipse time is arbitrary ; we repeated our entire analysis of the @xmath4 using the 4.5@xmath1 m eclipse time and found no difference in our results . the total duration and ingress / egress duration we set to the average of our 3.6@xmath1 m and 4.5@xmath1 m results .
in addition to the eclipse model , we also included linear decorrelation terms for airmass and time .
we scaled the errors on each night so that a baseline zero depth fit had a reduced @xmath99 of one . in all cases
the scaling factor was within ten percent of unity .
we used the baseline fit to calculate the @xmath122 values for the non - zero depth fits .
combining the data from all four good nights , we find suggestive evidence for an eclipse depth of @xmath5% in @xmath4 .
figure 8 shows the @xmath122 as a function of eclipse depth for these four nights .
the black line in figure 8 is the @xmath122 of all the nights added together , and is valid under the assumption that our uncertainties are uncorrelated night to night .
we have adopted this joint constraint as our final determination of the @xmath4 eclipse depth of kelt-1b .
figure 9 shows the four complete @xmath4 lightcurves individually , and combined , phased , and overplotted with our best fit eclipse model .
if the detection in @xmath4 is real , then it is the result of thermal emission from kelt-1b , and not reflected light . in the case of an extreme bond albedo of one , and assuming kelt-1b reflects as a lambert sphere , then the eclipse depth due to reflected light alone would be @xmath123% .
a more realistic bond albedo of 0.1 would reduce this depth by a factor of ten , and place it far below our precision in @xmath4 .
we strongly detect the secondary eclipses of kelt-1b at both 3.6@xmath1 m and 4.5@xmath3 m , and weakly detect the eclipse in @xmath4 ( figures 7 , 9 , and 10 ) .
we measure eclipse depths of @xmath124% in @xmath4 , @xmath125% at 3.6@xmath1 m and @xmath126% at 4.5@xmath3 m .
these depths correspond to brightness temperatures of @xmath127k , @xmath128k and @xmath129k for the @xmath4 , 3.6@xmath1 m and 4.5@xmath1 m eclipses , respectively .
we derive the median and 68% confidence intervals for @xmath108 , @xmath109 , and @xmath25 from the final mcmc chain in both of the spitzer bands . at both 3.6@xmath1 m and 4.5@xmath1 m
we infer values of @xmath108 and @xmath109 that are consistent with zero , which signifies kelt-1b s orbit is consistent with circular .
while we formally calculate a value of @xmath25 that is greater than zero at @xmath1301@xmath65 in both bands , this is a result of the well - known lucy - sweeney bias @xcite .
our measurement of circular orbit lends credence to the strong circumstantial evidence that the orbit of kelt-1b has been tidally circularized and that the star kelt-1 has tidally synchronized to the orbital period ( see section 6.2 of the discovery paper ) .
lcc @xmath93&baseline flux & @xmath131 + @xmath132&x - position linear coefficient & @xmath133 + @xmath134&y - position linear coefficient & @xmath135
+ @xmath114&time linear coefficient & @xmath136 + @xmath112&time of nearest transit ( @xmath137 ) & @xmath138 + @xmath139&log orbital period ( days ) & @xmath140 + @xmath141 & & @xmath142 + @xmath143 & & @xmath144 + @xmath145&cosine of inclination & @xmath146 + @xmath42&radius of planet in stellar radii & @xmath147 + @xmath148&log semi - major axis in stellar radii & @xmath149 + @xmath50&eclipse depth & @xmath150 + @xmath18&orbital period ( days ) & @xmath19 + @xmath111&time of eclipse ( @xmath137 ) & @xmath151 + @xmath152&semi - major axis in stellar radii & @xmath153 + @xmath46&inclination ( degrees ) & @xmath154 + @xmath48&impact parameter & @xmath155 + @xmath52&fwhm duration ( days ) & @xmath156 + @xmath54&ingress / egress duration ( days ) & @xmath157 + @xmath56&total duration ( days ) & @xmath158 + @xmath159 & & @xmath160 + @xmath161 & & @xmath162 + @xmath25&orbital eccentricity & @xmath163 + @xmath164&argument of periastron ( degrees ) & @xmath165 [ tab : ch .
1 kelt-1 sec .
0.50 ( variable ) ] lcc @xmath93&baseline flux & @xmath166 + @xmath132&x - position linear coefficient & @xmath167 + @xmath134&y - position linear coefficient & @xmath168
+ @xmath114&time linear coefficient & @xmath169 + @xmath112&time of nearest transit ( @xmath137 ) & @xmath170 + @xmath139&log orbital period ( days ) & @xmath171 + @xmath141 & & @xmath172 + @xmath143 & & @xmath173 + @xmath145&cosine of inclination & @xmath174 + @xmath42&radius of planet in stellar radii & @xmath175 + @xmath148&log semi - major axis in stellar radii & @xmath176 + @xmath50&eclipse depth & @xmath177 + @xmath18&orbital period ( days ) & @xmath19 + @xmath111&time of eclipse ( @xmath137 ) & @xmath178 + @xmath152&semi - major axis in stellar radii & @xmath179 + @xmath46&inclination ( degrees ) & @xmath180 + @xmath48&impact parameter & @xmath181 + @xmath52&fwhm duration ( days ) & @xmath182 + @xmath54&ingress / egress duration ( days ) & @xmath183 + @xmath56&total duration ( days ) & @xmath184 + @xmath159 & & @xmath185 + @xmath161 & & @xmath186 + @xmath25&orbital eccentricity & @xmath187 + @xmath164&argument of periastron ( degrees ) & @xmath188 [ tab : ch .
2 kelt-1 sec .
0.80 ( variable ) ]
overall , kelt-1b is a unique object : it is a relatively high mass and high surface gravity object that orbits only 3.6 stellar radii away from its host star . among high mass sub - stellar objects ,
this places kelt-1b squarely in a radiation environment that until now has been populated solely by hot jupiters .
kelt-1b can therefore be interpreted in the context of a hot jupiter with very high surface gravity , or in the context of a brown dwarf subject to strong external irradiation .
we consider both perspectives .
if considered as a planet , kelt-1b stands out due to its extremely high surface gravity ( @xmath189 ) , which is thirty times higher than for a typical hot jupiter .
kelt-1b therefore allows us test theories of hot jupiter atmospheres at a very high surface gravity .
of particular interest are the amount of heat redistribution and the presence of a stratospheric temperature inversion within the atmosphere of kelt-1b . @xcite have noted that planets hotter than @xmath82000@xmath190k are observed to have extremely low amounts of heat redistribution from their day- to their nightsides , presumably because the shorter radiative timescales in hotter atmospheres cause these planets to reradiate the incident stellar flux , rather than advecting it through winds to the nightside . in kelt-1b , due to its high surface gravity , the theoretical radiative timescale is relatively longer , and the theoretical advection timescale relatively shorter , than in most hot jupiters . similarly , consider the presence of a stratospheric temperature inversion in the atmosphere of kelt-1b .
temperature inversions have been observed in several hot jupiters , predominantly among those with equilibrium temperatures higher than @xmath191k ( e.g. , * ? ? ?
* ) . @xcite and
@xcite have proposed that gas - phase tio in an atmosphere causes temperature inversions , since it is a strong optical absorber and condenses between @xmath192k and @xmath191k , depending on the pressure .
however , the ultimate cause of inversions , and their precise regulatory mechanisms , have not been definitively agreed upon .
our observations show that kelt-1b does not have a high heat redistribution efficiency .
the tio , complete redistribution atmosphere model is strongly excluded by our observations , and the highest allowable redistribution efficiency , presuming the presence of tio , would be for day - side redistribution . following the notation of @xcite , it is probable that @xmath193 , and possible that @xmath194 ( instantaneous re - emission of the incident stellar radiation ) .
the large difference between the equilibrium temperature of kelt-1b ( @xmath195k ) and the brightness temperatures we measure at all three wavelengths ( @xmath8@xmath196k ) is also indicative of an extremely low amount of heat redistribution occurring in the atmosphere .
this agrees with the trend noted by @xcite that hotter planets have lower heat redistribution inefficiencies . that being said , and
as noted previously , the increased photospheric pressure in kelt-1b may complicate this interpretation .
all other things being equal , the pressure level for the @xmath197 surface in an atmosphere is proportional to surface gravity . thus kelt-1b should have a photosphere at a pressure @xmath830 times deeper than on a typical hot jupiter . since radiative time constants tend to increase greatly with pressure @xcite , one would expect the radiative time constant at the photosphere would be larger for kelt-1b than for an otherwise identical low - mass hot jupiter .
similarly , the advection timescale scales inversely with surface gravity and scale height @xcite , which would make advection relatively quick in kelt-1b s atmosphere .
these changes in the radiative and advective timescales would allow more time for faster advection within kelt-1b s atmosphere , thereby lessening the day - night temperature difference and causing a large hotspot offset from the substellar point , for a given set of irradiation conditions .
for this reason it will be interesting to compare kelt-1b to lower - mass hot jupiters that have similar irradiation levels .
specifically , orbital phase curve observations of kelt-1b would allow one to directly measure the day - night temperature difference and test the effect of kelt-1b s greater photospheric pressure .
-0.0 in -0.0 in 0.1 in as a side note , our determination that @xmath193 would appear to be roughly inconsistent with the ground - based secondary eclipse observations undertaken for the kelt-1b discovery paper .
figure 14 of @xcite implies that @xmath193 would have been detectable by at least @xmath198 in those data .
however , when we examined this issue we discovered that the analysis in @xcite incorrectly dealt with the observations taken using the faulkes telescope north ( ftn ) .
the ftn data were reported as differential magnitudes , but we treated them as differential fluxes , inverting the sense of the changes in intensity . if we recalculate figure 14 of @xcite correctly , then atmospheres with @xmath193 would at best be detectable at @xmath199 .
our observations are not sufficient to conclusively determine whether a tio inversion exists in the atmosphere of kelt-1b , but we can nonetheless provide some useful constraints .
figure 10 shows our measured eclipse depths on top of atmosphere models from @xcite .
the models without tio ( dashed lines ) are for an atmosphere without an inversion , while the tio models ( solid lines ) have an inversion .
the best fit to the data is the no - tio , strong hotspot model , with @xmath200 for three degrees of freedom .
however , the no - tio , mild hotspot model has a @xmath122 relative to the best model of only 1.44 , while the tio , day - side redistribution model has a @xmath201 , and thus these models are also consistent with the data . on the other hand , the no - tio , day - side redistribution model is marginally excluded with @xmath202 , and while the tio , complete redistribution model is strongly excluded with @xmath203 . since the @xmath83100k brightness temperature that we measure for the day side is much hotter than the @xmath191k condensation temperature of tio at kelt-1b s photospheric pressure , the lack of a strong tio signal raises the possibility that a day - night tio cold trap exists in kelt-1b s atmosphere .
a day - night cold trap occurs for tio when a planetary day side is hot enough to allow for gaseous tio , but the night side is below the condensation temperature .
this allows for tio to condense and settle out of the atmosphere on the planetary nightside , removing it from the upper atmosphere .
this is distinct from the cold traps predicted by 1d atmosphere models , which exist between a hot upper atmosphere and a hot lower convective layer as a pressure band cold enough to allow for gaseous tio to condense @xcite .
day - night cold traps were suggested as an important mechanism in hot jupiter atmospheres by @xcite and more thoroughly modeled by @xcite , who specifically examined the role of a cold trap in hd 209458b .
@xcite found that tio could settle out of the nightside atmosphere rapidly enough to prevent an inversion if the condensate grain size was larger than a few microns , which they found unlikely due to the relative scarcity of tio .
@xcite did allow that if the tio combined and condensed with a more abundant gas , such as sio , then sufficiently large grains were much easier to form . in the case of kelt-1b ,
its high surface gravity may aid the efficiency of a cold - trap by increasing the particle settling velocity , and hence the settling efficiency .
the dynamics of a day - night tio cold trap may be very different on kelt-1b than on a lower surface gravity hot jupiter .
however , as an example of the complexity of the issue , consider the presence of temperature inversions in the two other planets with surface gravities higher than @xmath204 that have been observed with spitzer : wasp-18b and hat - p-2b .
wasp-18b does not show strong evidence for a temperature inversion @xcite versus @xmath205 ) .
their conclusion is partly based on wasp-18b s measured temperature of @xmath206k , and the trend for planets hotter than @xmath191k to have a stratospheric temperature inversion . ] , while @xcite find that hat - p-2b does have a strong inversion .
both planets have large day to night temperature contrasts @xcite , and both have eclipse brightness temperatures at 3.6@xmath1 m and 4.5@xmath1 m above 2100k @xcite .
this implies that both planets should have a day - night tio cold trap that inhibits inversions , though the atmospheric dynamics of hat - p-2b are complicated by its eccentric ( e=0.52 ) orbit , which causes the stellar insolation to vary considerably .
@xcite makes the intriguing suggestion that one could test for the presence and efficiency of a day - night tio cold trap by looking for a latitude dependence in the dayside temperature structure of a planet .
atmospheric gases at higher latitudes could have a shorter nightside crossing - time , lessening the amount of tio depletion that occurs in the cold trap .
if the high - latitudes retain tio while the equator does not , this would create a latitude - dependent inversion on the dayside of the planet .
a latitudinal variation in the dayside temperature could be directly observed by using the phase - mapping technique demonstrated by @xcite and @xcite , though this requires previous knowledge of the longitudinal temperature gradient of the planetary dayside .
this argues for obtaining 3.6@xmath1 m and 4.5@xmath1 m orbital phase curve observations of kelt-1b using spitzer , as those observations would be the only way to directly observe the longitudinal temperature gradient . as a brown dwarf
, kelt-1b is unusual because of the strong stellar insolation it receives , which is far in excess of its own internal luminosity .
if it were isolated , we estimate that kelt-1b s surface flux from internal heat should be approximately @xmath207 to @xmath208 erg s@xmath39 @xmath40 @xcite , assuming the discovery paper s age measurement of @xmath2092 gyr .
this heat flux is two to three orders of magnitude less than the incident stellar flux of @xmath61 erg s@xmath39 @xmath40 .
the dayside energy budget of kelt-1b is therefore dominated by the incident stellar radiation .
indeed , our measured brightness temperature of about @xmath196k corresponds to an mid m - dwarf , even though by its mass , age and surface gravity we would expect kelt-1b to be a mid t - dwarf if it were isolated , at about @xmath210k @xcite .
the large amount of stellar insolation relative to the internal heat flux of kelt-1b probably means that the atmospheric circulation in kelt-1b is driven by thermal forcing , similar to hot jupiters .
this is in contrast to the atmospheres of cold brown dwarfs , whose circulation is expected to by primarily driven by the breaking of upward - welling gravity waves generated at the radiative - convective boundary .
@xcite calculate that for internal energy fluxes of @xmath209 @xmath211 erg s@xmath39 @xmath40 , which is ten to one hundred times higher than what we expect for kelt-1b based on @xcite , these waves ought to generate atmospheric winds of tens to hundreds of meters per second . by comparison ,
the thermal forcing in a typical hot jupiter atmosphere is expected , and observationally implied , to cause wind speeds of several thousand meters per second @xcite .
interestingly , since kelt-1b s rotation is presumably tidally synchronized to its orbital period and thus relatively slow , the atmospheric rossby number is @xmath212(1@xmath23@xmath213 , where @xmath214 is the wind speed and @xmath215 is the characteristic length scale of atmospheric flows .
this is very high compared to the expected rossby numbers of cold brown dwarfs , which should range from @xmath216 to @xmath217 @xcite .
we therefore expect the atmospheric dynamics of kelt-1b to be similar to hot jupiters
. ultimately , we would like to join observations and theories of hot jupiters and brown dwarfs , by using kelt-1b as one link in that chain .
this sort of union is already occurring for brown dwarfs and directly imaged , internal energy dominated , giant planets using the systems discovered around hr 8799 @xcite and gj 504 @xcite as well as the isolated object pso j318 - 22 @xcite .
as of yet , however , there has been no chance to do the same with irradiated jupiters and brown dwarfs .
a comparison of irradiated jupiters and cold brown dwarfs already points to intriguing atmospheric differences between the two populations .
figure 11 shows the equilibrium temperature and @xmath6-[4.5]$ ] color of kelt-1b relative to other brown dwarfs and planets .
we measure a @xmath6-[4.5]$ ] color for kelt-1b of @xmath7 .
this is consistent with other brown dwarfs at a similar equilibrium temperature , which generally have @xmath6-[4.5]$ ] colors near zero for @xmath218k .
the strongly irradiated jupiters show no such clear behavior with temperature across any of the temperature range .
this striking diversity for the planets could be due to a variety of factors , perhaps most strongly influenced by the presence or absence of dayside temperature inversions .
other factors such as non - standard abundance ratios and scatter in bond albedos and dayside energy redistribution may be important as well .
the clear trends in brown dwarf colors with @xmath219 are now well understood in terms of atmospheric chemistry @xcite . at high temperatures h@xmath220o and
co absorption bands dominate the infrared spectrum .
when temperatures fall below @xmath8@xmath221k carbon transitions from co to ch@xmath222 , which absorbs strongly in the spitzer 3.6@xmath1 m band .
this chemistry change combined with the ever redward shift of the planck curve to longer wavelengths explains the trend to redder [ 3.6]-[4.5 ] spitzer colors in figure 11 .
-0.0 in -0.0 in the fact that kelt-1b so strikingly resembles other brown dwarfs , despite being more like a hot jupiter in terms of external forcing of the atmosphere by stellar irradiation , suggests that surface gravity is a very important factor in governing the ultimate atmospheric dynamics of these bodies .
the further population of this diagram across planet temperature and mass may help identify further differences or similarities between irradiated planets and self - luminous brown dwarfs .
we have measured the secondary eclipse of the highly irradiated transiting brown dwarf kelt-1b in three bands , and found that the object s high surface gravity , and not the high stellar irradiation , dominates kelt-1b s atmosphere .
this makes kelt-1b s atmosphere appear more similar to field brown dwarfs at the same effective temperature , rather than to strongly irradiated hot jupiters .
these observations are the first constraints on the atmosphere of a highly irradiated brown dwarf .
specifically , we measure secondary eclipse depths of @xmath0% at 3.6@xmath1 m and @xmath2% at 4.5@xmath3 m .
we also find tentative evidence for the secondary eclipse in the @xmath4 band with a depth of @xmath5% . from these measured eclipse depths ,
we conclude that kelt-1b does not have a high heat redistribution efficiency , and does not show strong evidence for a stratospheric temperature inversion .
importantly , our measurements reveal that kelt-1b has a @xmath6-[4.5]$ ] color of @xmath7 , identical to that of isolated brown dwarfs of similarly high temperature . in contrast
, hot jupiters generally show redder @xmath6-[4.5]$ ] colors of @xmath80.4 , with a very large range from @xmath80 to @xmath81 .
kelt-1b gives us the chance to study a high surface gravity atmosphere using all of the tools that have been developed to measure the dynamics in hot jupiter atmospheres ( e.g , hd 189733b by * ? ? ?
for the first time we will be able to directly observe large scale atmospheric dynamics and flows in a high surface gravity environment .
already , our secondary eclipse observations demonstrate that there is almost no heat redistribution from the day to the night side of kelt-1b , and suggest that there may be a global hotspot at the substellar point .
this implies that the radiative timescale in the atmosphere is extremely short relative to the relevant dynamic timescale in the atmosphere . by serving as a transitional object between cold brown dwarfs and hot giant planets , more detailed observations of kelt-1b
will have a strong ability to illuminate the similarities and differences between these two populations . in the near - term
, spitzer observations of kelt-1b s orbital phase curve would provide the best extension to our current understanding of the object s atmosphere , by directly measuring the day - night temperature contrast and the presence of large scale flows in the atmosphere .
transmission spectroscopy of kelt-1b is currently impossible due to kelt-1b s high surface gravity , though longer - term the james webb space telescope may be able to conduct these observations .
finally , our tentative detection of the eclipse in @xmath4 demonstrates the relatively unique opportunities for secondary eclipse observations from the ground that this system affords us particularly with modest telescopes .
work by t.g.b . and
was partially supported by nsf career grant ast-1056524 .
k.a.c . was supported by a nasa kentucky space grant consortium graduate fellowship .
this research has made use of nasa s astrophysics data system .
|
we present secondary eclipse observations of the highly irradiated transiting brown dwarf kelt-1b .
these observations represent the first constraints on the atmospheric dynamics of a highly irradiated brown dwarf , the atmospheres of irradiated giant planets at high surface gravity , and the atmospheres of brown dwarfs that are dominated by external , rather than internal , energy . using the spitzer space telescope , we measure secondary eclipse depths of @xmath0% at 3.6@xmath1 m and @xmath2% at 4.5@xmath3 m .
we also find tentative evidence for the secondary eclipse in the @xmath4 band with a depth of @xmath5% .
these measured eclipse depths are most consistent with an atmosphere model in which there is a strong substellar hotspot , implying that heat redistribution in the atmosphere of kelt-1b is low .
while models with a more mild hotspot or even with dayside heat redistribution are only marginally disfavored , models with complete heat redistribution are strongly ruled out .
the eclipse depths also prefer an atmosphere with no tio inversion layer , although a model with tio inversion is permitted in the dayside heat redistribution case , and we consider the possibility of a day - night tio cold trap in this object . for the first time
, we compare the irac colors of brown dwarfs and hot jupiters as a function of effective temperature .
importantly , our measurements reveal that kelt-1b has a @xmath6-[4.5]$ ] color of @xmath7 , identical to that of isolated brown dwarfs of similarly high temperature .
in contrast , hot jupiters generally show redder @xmath6-[4.5]$ ] colors of @xmath80.4 , with a very large range from @xmath80 to @xmath81 .
evidently , despite being more similar to hot jupiters than to isolated brown dwarfs in terms of external forcing of the atmosphere by stellar insolation , kelt-1b appears to have an atmosphere most like that of other brown dwarfs .
this suggests that surface gravity is very important in controlling the atmospheric systems of substellar mass bodies .
|
a problem of broad interest is the advection and diffusion of passive scalars in a turbulent flow .
the classical paradigm of a passive scalar is the temperature field @xmath0 when the heating is small .
the temperature increments @xmath1 have been studied in the literature @xcite in search of scaling in the intermediate range of scales that are unaffected directly by either the stirring mechanism or diffusive and viscous effects . for examples of early and recent experimental studies , see @xcite and @xcite , respectively .
the following two types of moments of @xmath2 , the so - called structure functions of order @xmath3 , have been employed : @xmath4 here , @xmath5 is the separation vector between two spatial positions , and @xmath6 defines a suitable ensemble average . for convenience we will call ( [ sn ] ) the _ normal _ structure functions and ( [ snabs ] ) the _ absolute _ structure functions . when @xmath7 is small compared to the large scale @xmath8 , both structure functions are homogeneous , i.e. , independent of @xmath9 .
clearly , ( [ sn ] ) and ( [ snabs ] ) coincide for even @xmath3 .
however , remembering that normal odd moments are zero in the absence of a mean temperature gradient while that is not so for absolute odd moments , one may expect , for odd @xmath3 , that there might be perceptible differences in the two classes of structure functions .
it is useful to quantify these differences and stress the reasons why they might be important .
the following two comments put the present work in perspective . the first concerns the extraction of scaling exponents of structure functions using the extended self - similarity ( ess ) @xcite . instead of examining the scaling of normal structure functions , @xmath10 , with respect to the scale separation @xmath11 directly
, the practice is to examine the scaling relative to another structure function , say @xmath12 , @xmath13 .
this usually leads to the extension of a possible algebraic scaling range , and the relative scaling exponent , @xmath14 where @xmath15 , can be obtained with greater confidence . in the literature ,
the implementation of the method has often mixed up normal structure functions and absolute structure functions without exploring the differences between them .
further , it was recently shown that in the presence of convection , ess fails to show the existence of two distinct scaling ranges the nearly passive behavior at small scales and the dominance of buoyancy at large scales @xcite .
this point is illustrated in fig .
[ s2 ] . in the top figure
, we show the second order structure function with clearly separated scaling ranges ; in the bottom figure , the corresponding ess plot is shown to result in a line of nearly constant slope , without making the necessary distinction between the regions marked a and b. ( the second set of data in the bottom figure corresponds to another set of measurements , with qualitatively similar conclusion . )
the second comment is that the structure functions , eqs.([sn ] ) and ( [ snabs ] ) , have different sign - symmetry properties .
one has to distinguish the sign - symmetry with respect to the reversal of increments @xmath16 from that with respect to spatial reflection ( also known as parity)that is , for the transformation @xmath17 . the absolute structure functions ( [ snabs ] ) remain unchanged for all orders @xmath3 under both transformations ; they have _ even _ sign - symmetry .
the same is true for even - order normal structure functions but the odd - order normal structure functions change sign under both transformations ; they have _ odd _ sign - symmetry or sign - antisymmetry .
note that @xmath18 does not follow from @xmath17 although , for homogeneous turbulence , both transformations have the same effect on ( [ sn ] ) .
we wish to comment on ess in the light of the sign - symmetry properties of structure functions . for this purpose ,
it is convenient to introduce new types of structure functions which explicitly emphasize the sign - symmetry with respect to the increment @xmath2 .
this is our basic goal .
these new quantities , along with ( [ sn ] ) and ( [ snabs ] ) , will be calculated from temperature data from a high - reynolds - number atmospheric boundary layer .
the taylor microscale reynolds number is about 3500 .
since the experiments have been described in some detail in @xcite , we shall mention only a few details here .
measurements were made in a boundary layer above salt flats of the dugway proving ground in utah , at a height of 1.75 m above the ground .
the ground was smooth on the order of a millimeter .
taylor hypothesis was used and @xmath19 was about 7% with the mean speed @xmath20 ms@xmath21 .
measurements were made at various times of the day , covering intensely convective motion in late afternoon , essentially neutral conditions in the evening hours and somewhat stable conditions until about 11 pm .
the data records chosen for analysis corresponded to constant wind conditions in magnitude as well as direction .
temperature fluctuations were measured by two cold wires of 0.6 @xmath22 m diameter and 1 mm length .
the data acquisition system was a standard constant - current anemometer system operated at a small enough current to minimize velocity sensitivity .
the paper is organized as follows . in sec.[sec : decomposition ] , we will construct the new types of structure functions , in line with the approach of @xcite for velocity increments .
since these structure functions are of intrinsic interest , they are presented first before considering ess .
section [ sec : ess ] contains the results of the ess data analysis for structure functions with even and odd sign - symmetry . in sec .
[ sec : explanation ] , we discuss a possible explanation of the effects observed in terms of the so(3 ) rotation group decomposition of structure functions of different sign - symmetry .
some concluding remarks are presented in sec.[sec : conclusions ] .
let the probability density function ( pdf ) of the temperature increment at fixed distance vector @xmath5 be given by @xmath23 .
first , we can define ( [ sn ] ) and ( [ snabs ] ) as @xmath24 the pdf can be decomposed into its symmetric and antisymmetric parts with respect to the increment as @xmath25 note that @xmath26 does not have the positive - definite property of a pdf .
we can also define the positive and negative parts of the pdf as @xmath27 and define moments of @xmath2 with respect to @xmath28 , @xmath26 , @xmath29 and @xmath3 , respectively , by the following relations : @xmath30 the following relations are now valid : @xmath31 for odd values of @xmath32 , eq .
( [ evenodd2 ] ) reduces to the normal odd - order structure function @xmath33 whereas , for even @xmath34 , eq .
( [ evenodd2 ] ) is a new structure function which is @xmath35 order , but sign-@xmath36 .
it is not possible to have this combination in terms of either normal or absolute structure functions . on the basis of the discussion of the sign - symmetries of @xmath28 and @xmath26 , using the definitions ( [ pos_def ] ) and ( [ neg_def ] ) , it is clear that @xmath37 is sign - symmetric , while @xmath38 is sign - antisymmetric . to illustrate the differences between normal and absolute structure functions , we plot in fig .
[ fig : fig1 ] the logarithmic local slopes for four types of moments : @xmath39 , @xmath40 , @xmath41 and @xmath42 .
figure [ fig : fig1a ] show these quantities for @xmath43 . here
@xmath44 is the normal ( and absolute ) second order structure function ; @xmath45 is the newly defined sign - antisymmetric second order structure function .
figure [ fig : fig1b ] shows the same four quantities for @xmath46 . although the scaling is not impeccable even at this high reynolds number ( see ref .
@xcite for comments in this regard on the scaling of velocity structure functions ) , a scaling tendency can be discerned in the range between a few mm and a few cm .
independent of this detail , it is clear that the exponents , if one were to assign nominal values in the scaling range , are not the same for all the different structure functions of the same order .
in particular , the second - order sign - antisymmetric structure function @xmath45 has a substantially larger exponent than the classical exponent of 2/3 . while @xmath47 has a scaling exponent close to the kolmogorov prediction of 1 , the other curves have measurably smaller scaling exponents . in particular , the absolute structure functions have smaller scaling exponents than the normal structure function . at the least
, these features are disconcerting to anyone interested in scaling exponents .
an understanding of differences in the exponents of the various types of structure functions may help in this regard .
this will be attempted in the next section .
in this section , we shall examine ess of scalar statistics in the light of the new objects defined in terms of their parity .
we first apply ess as is normally done for velocity statistics : plot structure functions of all orders against the absolute third - order structure function .
we shall follow this practice for illustrative purposes , even though the third - order does not have a comparably significant meaning for temperature . in fig .
[ fig : fig2a ] , the ess plot of @xmath48 versus @xmath49 shows a single relative scaling exponent outside the dissipative range .
the inset shows the local logarithmic slope of the extended self - similarity plot which is calculated via @xmath50 .
however , this same signal has two scaling regions with distinct exponents when plotted against @xmath11 ( corresponding to the passive range at small scales and the convective range at large scales , see top panel of fig . 1 ) .
this two - exponent scaling behavior is masked by ess .
this is essentially so even if we plot the sign - antisymmetric structure function of the second - order against @xmath51 ( which is also sign - antisymmetric , see sec .
2 ) , as seen in fig .
[ fig : fig2b ] .
let us now compare objects of the same order , but different sign - symmetry . in fig .
[ fig : fig3a ] , we show a plot of @xmath52 which is odd - parity ( made of the antisymmetric combination of the pdf functions ) versus @xmath53 , which is the corresponding even - parity object of the same order .
the two objects scale identically in the smallest ( dissipative / diffusive ) scales with relative exponent of unity , transitioning into a region of relative exponent 1.27 .
this corresponds to the inertial range scaling , as we already know from fig .
[ fig : fig1 ] . past this range
, the relative scaling exponent drops back to @xmath54 . in fig .
[ fig : fig3b ] , we repeat the ess comparison for the second - order . instead of plotting the second - order object against another order ,
we choose to compare the even- and odd - parity manifestations @xmath55 and @xmath56 , respectively .
once again , it is seen that there are two regions of scaling beyond the dissipative range .
the small - scale range has a relative scaling exponent of about 1 while the large - scale range has a slope of 1.42 .
these exponents are as expected from the direct scaling analysis of @xcite .
finally , when we compare objects of different order @xmath57 different parity as shown in fig . [ sna_vs_sms ] , the dual scaling range feature persists .
figure [ fig : fig4a ] shows @xmath58 versus @xmath59 while fig .
[ fig : fig4b ] shows @xmath44 vs. @xmath60 .
again , we recover two separate scaling ranges , one inertial and one convective . from these examples
, we can now make the following general statement . in a process with two distinct scaling ranges
, ess can not distinguish between them whenever a comparison is made between two like - parity objects .
however , if ess compares objects of @xmath61 order , but of opposite parity , a second scaling ( if it exists ) can be recovered .
this is our major qualitative conclusion .
this kind of comparison has been made possible by the introduction of the odd - symmetry , even - order objects @xmath62 ( @xmath3 even)which , to our knowledge , have not been considered before .
we shall now discuss the above property of ess from another perspective that contains the correct description of the parity property .
we consider an analytical expression for structure functions with two distinct scaling ranges , separated by a crossover scale @xmath63 .
since we do not know the proper analytical expression for the structure functions , we use an empirical interpolation formula .
batchelor s attempt @xcite in this direction has been extended variously , in particular in refs .
@xcite , for structure functions of all orders . here , we shall extend the specific form proposed in ref .
@xcite to the @xmath3th order as @xmath64^{\alpha_{n}-\beta_{n}}\ , , \label{bat1}\end{aligned}\ ] ] with the function @xmath65 where the assumed scaling exponents @xmath66 hold for the passive range , @xmath67 , and @xmath68 for the convective range , @xmath69 , and @xmath70 ; @xmath71 is a dimensionless function that is monotonically decreasing , and the exponent @xmath72 determines the width of the crossover from one scaling range to the other .
we make use of the @xmath73 @xmath74 of the signed structure functions by noting that , in homogeneous turbulence , @xmath75 is even - parity with respect to reflection under @xmath76 for all @xmath3 ; the odd - order normal structure functions @xmath77 are odd - parity ; the newly defined @xmath78 , even - order sign-@xmath36 statistics are odd - parity as well .
so , in the homogeneous case the sign - symmetry with respect to the increment is the @xmath79 as the sign - symmetry with respect to the spatial orientation of @xmath5 ( i.e. parity ) and this allows for the following decomposition of the objects defined in sec .
the recently developed so(3 ) group decomposition @xcite applies conveniently to the symmetric and antisymmetric structure functions . for the scalar case , the basis functions are spherical harmonics @xmath80 @xcite . the even or odd _ spatial _ parity is carried by the angular dependence in the spherical harmonics , in effect by the index @xmath81 , as @xmath82 perfectly isotropic objects would contain an @xmath83 sector only ,
i.e. , no angular dependence would be present .
sign - symmetric functions are composed of only even sectors while sign - antisymmetric functions are composed only of odd sectors .
thus , @xmath84 where the superscripts @xmath85 , @xmath86 , @xmath87 denote the even - parity contributions allowed , and @xmath88 where the superscripts @xmath89 , @xmath90 , @xmath87 denote the odd - parity contributions allowed .
we can substitute the algebraic scaling form of ( [ bat1 ] ) for each scaling term in ( [ symm1 ] ) and ( [ symm2 ] ) to obtain @xmath91^ { \alpha_{n}^{(2k)}-\beta_{n}^{(2k ) } } \nonumber\\ & \times & y_{2k , m}(\theta,\phi)\ , , \label{bat3}\end{aligned}\ ] ] and @xmath92^ { \alpha_{n}^{(2k+1)}-\beta_{n}^{(2k+1 ) } } \nonumber\\ & \times & y_{2k+1,m}(\theta,\phi)\ , .
\label{bat3a}\end{aligned}\ ] ] the finding that ess yields the same relative exponents , when comparisons are made of even - parity statistics , means that for all sectors and for any pair of orders @xmath93 we have @xmath94 equivalently , for @xmath95 , it follows that @xmath96 where the quantities @xmath97 are the relative increments between successive orders .
the constraint ( [ even_const ] ) means that the relative increments in both convex functions , @xmath98 as well as @xmath99 , are equal .
this was already shown in ref.@xcite , where physically different turbulent states such as the kolmogorov turbulence in three dimensions , thermal convection as well as magnetohydrodynamic turbulence showed the same relative ess exponents .
we suppose , therefore , that as long as the different regimes belong to the same universality class " i.e .
, if they have the same relative increments @xmath100 for all orders @xmath3ess will simply mask the presence of distinct scaling regimes .
the argument for the odd - parity ess comparisons is similar , now with the constraint @xmath101 one remains within a universality class when plotting @xmath102 as functions of @xmath103 , but it can be expected that the relative exponents will be different from those for the even parity case .
we now turn to the case where @xmath104 are compared to @xmath103 , which are objects with different symmetries .
when proceeding as in ( [ even_const ] ) , ( [ bat4 ] ) , and ( [ odd_const ] ) we obtain @xmath105 for insensitivity to different scaling regimes , the following must hold true : @xmath106 recall that the superscript @xmath107 corresponds to even parity objects while @xmath108 corresponds to odd parity .
we expect that several mechanisms , such as the breaking of reflection symmetry by a mean gradient for scales @xmath109 or by buoyancy effects for scales @xmath110 , will lead to different ratios on the left and right hand side of the eq .
( [ bat6 ] ) , respectively , thus leading to a `` discontinuity '' in scaling .
while this discussion can not be a proof of our main observation , it shows that the two scaling regimes belong to different universality classes when different parities are involved .
further , it provides hints for what kind of quantities have to be investigated in order to shed more light on the issue , in experiments as well as in numerical simulations .
we have analyzed moments of temperature increments with respect to their sign - symmetry properties , and defined sign - symmetric and sign - antisymmetric components for both even and odd - order structure functions .
ess analysis of all combinations of symmetry and order of structure functions indicates that this technique masks the convective scaling regime .
only if objects of opposite parity are compared can one recover the distinct scalings .
we have presented a model for how such scaling regimes behave in the ess analysis using even and odd orders of a spherical harmonic expansion .
unfortunately , this model can not be taken to its logical conclusion because the various numerical coefficients can not be obtained with any certainty from the existing data .
furthermore , a dependence on the strength of the mean temperature gradient might modify some of our results as well e.g .
, differences in the exponents of even order moments in comparison to odd order moments .
one purpose here has been to point out the possible pitfalls in using ess without proper consideration of the symmetry properties of the statistics being compared .
this observation does not detract from the merits of the method .
indeed , ess has proved to be a very useful tool in extracting exponents when the scaling range with respect to @xmath11 is short .
we would like to acknowledge fruitful discussions with a. bershadskii .
one of us ( j.s . ) acknowledges support by the deutsche forschungsgemeinschaft , the feodor - lynen fellowship program of the alexander von humboldt - foundation and yale university .
he also wishes to thank the institute for physical science and technology at the university of maryland , and the international centre for theoretical physics at trieste for hospitality .
|
new scalar structure functions with different sign - symmetry properties are defined .
these structure functions possess different scaling exponents even when their order is the same .
their scaling properties are investigated for second and third orders , using data from high - reynolds - number atmospheric boundary layer .
it is only when structure functions with disparate sign - symmetry properties are compared can the extended self - similarity detect two different scaling ranges that may exist , as in the example of convective turbulence .
|
recent surveys with high - resolution uv spectra have shown that mass outflow is common in seyfert 1 galaxies , appearing as blueshifted absorption in more than half of observed objects ( crenshaw et al .
detailed studies are needed to determine the physical conditions ( density , ionization structure , total gas column ) and geometry of the absorbers . here
, we focus on constraints derived on the absorbers in the uv spectrum of ngc 3783 to demonstrate some techniques that can be used to probe mass outflow in agns .
since the background agn light that the absorption features are imprinted on is comprised of distinct emission sources with different flux distributions , sizes , and geometries , the most general treatment of absorption features should take into account the different covering factors associated with each source .
the lyman series lines in the averaged stis and _ fuse _ spectra were used to separate the individual covering factors of the continuum and emission line sources in ngc 3783 ( gabel et al .
2003 ) . by incorporating variability of the emission - line profiles ,
that analysis can be extended to treat _ distinct emission - line regions_. figure 1 shows the c iv , ly@xmath2 , and n v emission - line profiles in both high - state ( solid line ) and low - state ( dotted line ) stis spectra .
the low - state profiles have been scaled to match the flux in the high velocity wings of the high - state spectrum .
the profiles diverge near line center due to the superposition of a varying broad line region component and a non - varying narrower component ( an intermediate line region , ilr ) .
the ilr flux profiles ( dashed lines ) were solved as described in gabel et al .
( 2004 , in preparation ) .
figure 2 shows the n v absorption profiles for two normalizations to demonstrate the effect of treating the covering factors of distinct emission components separately . in the left panel ,
the absorption has been normalized simply by dividing the total observed spectrum by our fit to the continuum plus total emission - line flux , i.e. , the blr + ilr . in the right panel ,
the ilr flux has first been subtracted out of both the data and emission - line fits , which is equivalent to assuming the ilr is completely unocculted by the absorbers . for each normalization , both the high - state ( top panel ) and low - state ( bottom panel ) spectra are plotted for comparison and the two doublet lines are overlayed , shifted to coincide in radial velocity .
this shows that if the covering factors of the emission components are assumed to be the same , then the n v absorption is not saturated and has effective covering factors in the range @xmath3 0.6 0.65 for components 1 3 .
the similarity in the absorption depths between the high and low - states ( for both doublet lines ) in the left panel implies there is little change in the column densities for this scenario .
conversely , if the ilr is unocculted , the equivalent absorption depths of the two doublet lines indicates the absorption is saturated in both the low and high states in some of the components .
this leads to significant differences in interpreting the physical conditions and variability in the absorption .
the c iii * @xmath51175 multiplet lines have been used as a density diagnostic for agn absorbers in several studies . however , as pointed out by behar et al .
( 2003 , apj , in press ) , the high densities derived in these studies ( e.g. , gabel et al .
2003 ) were based on calculations of level populations that only treated the @xmath0p@xmath6 level .
the @xmath70 and 2 levels have much lower radiative transition probabilities to the ground state and thus are populated at densities that are lower by several orders of magnitude ( bhatia & kastner 1993 ) .
we have computed the relative populations of the @xmath0p@xmath8 levels , extending the results of bhatia & kastner ( 1993 ) down to electron temperatures expected for the photoionized uv absorbers seen in agns .
collisional excitation and de - excitation and radiative decay between the six lowest terms / levels of the @xmath9 ion were included in our calculations .
the top panel of figure 3 shows the c iii*1175 absorption complex in ngc 3783 , with the location of the six multiplet lines marked and identified by the @xmath10 level of the transition .
measured column densities for each metastable level are given below the spectrum .
the bottom panel shows the computed populations of the metastable levels over a large range in density for @xmath11 16000 , 20000 , and 40000 k ( bottom to top ) .
thus , if the absorption is sufficiently narrow to resolve and measure individual lines in the c iii * multiplet , the _ ratios _ of the excited level populations give a very tight constraint on the electron density . for ngc 3783 , the @xmath72 : @xmath70 column density ratios , @xmath12 2.2 , gives @xmath13=3@xmath1410@xmath15 @xmath16 , which is largely insensitive to temperature .
combining this with a measurement of the ground state population provides a stringent temperature diagnostic .
the full implications of this measurement for the absorbers in ngc 3783 will be given in gabel et al .
( 2004 , in preparation ) .
|
results from an analysis of the intrinsic uv absorption in the seyfert 1 galaxy ngc 3783 are presented .
we focus on two new results that demonstrate techniques for deriving important constraints on the physical conditions and geometry of the absorbers .
first , using variability in the spectrum , the emission - line profile is separated into distinct kinematic components and the effect on the interpretation of covering factors and column densities is demonstrated .
second , measurements of the 2s2p @xmath0p metastable levels of c@xmath1 derived from the c iii*1175 - 76 absorption multiplet are presented .
new calculations of the metastable level populations are given and shown to provide a powerful diagnostic of the density ( and thus location ) and temperature in an absorber .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
|
one century after the discovery of cosmic rays the search for their sources is still ongoing .
astrophysical objects which are either confirmed or expected to be able to accelerate hadrons to the observed energies include supernova remnants@xcite , active galactic nuclei ( agn ) , gamma - ray bursts ( grbs ) , and shocks in star formation regions of galaxies .
the cosmic - ray nuclei interact with ambient matter and radiation fields close to their source@xcite .
charged pions produced in these interactions decay into neutrinos . therefore , the detection of high - energy neutrinos from such objects provides a unique possibility to identify individual astrophysical objects as cosmic - ray sources .
however , low fluxes and small interaction probabilities make the detection of high - energy neutrinos challenging . to date
no astrophysical object has been conclusively identified as a source of @xmath18 neutrinos .
previous searches have established limits enabling astrophysical models to be constrained@xcite . while individual neutrino sources might be too weak to be detectable with current instruments
, they would still contribute to a collective astrophysical neutrino flux .
fermi shock acceleration is thought to be the main acceleration mechanism for cosmic - ray nuclei and therefore a power - law spectrum with an index of about @xmath19 is expected for the nuclei in the interaction regions where the neutrinos are produced . based on the energy density of ultra - high - energy cosmic rays and assuming the cosmic - ray sources are transparent
, the all - flavor diffuse neutrino flux can be constrained theoretically to be lower than the waxman - bahcall bound of @xmath20@xcite . as neutrinos are assumed to originate mainly from pion decays , at the source
a flavor ratio of @xmath21 is expected .
this ratio would transform to @xmath13 on earth due to neutrino oscillations@xcite .
however , observing unequal or energy - dependent flavor contributions would be interesting , since for example the flavor ratio is sensitive to the assumed production mechanism at the source@xcite .
recently , evidence for this diffuse astrophysical neutrino flux was found .
its all - flavor intensity is estimated to be @xmath22 with indications for a cutoff at @xmath23 .
it is consistent with an isotropic flux and a flavor ratio of @xmath13 ( @xcite and fig .
[ figure : atmosphericdiffusefluxes ] ) . in order to measure the diffuse astrophysical neutrino flux at @xmath24 energies
, it has to be separated from two main sources of background , which both originate from the earth s atmosphere .
these are atmospheric muons and neutrinos produced in cosmic - ray air showers .
the atmospheric neutrino flux has two components .
the so - called conventional atmospheric neutrinos are produced in decays of pions and kaons .
their intensity is well - measured up to @xmath25 for @xmath26 and up to @xmath27 for @xmath28@xcite . at higher energies
the poor knowledge of the composition of the cosmic - ray flux , creating the neutrinos , causes significant uncertainties on the intensity .
the spectrum of the conventional atmospheric neutrinos is steeper than the cosmic - ray spectrum due to pion and kaon energy losses in the atmosphere .
the second component originates from the decay of charmed mesons , which have livetimes several orders of magnitude smaller than charged pions and kaons .
accordingly , neutrinos from these decays are called prompt atmospheric neutrinos . due to the short lifetime of the parent mesons the energy spectrum of the prompt atmospheric neutrinos
is expected to follow the spectrum of the cosmic rays that create them .
however , their intensity has never been measured and uncertainties in the relevant production cross sections lead to large uncertainties in the predicted flux .
the presence of a prompt neutrino component , like an astrophysical neutrino component , introduces a break into the neutrino energy spectrum .
given the large uncertainties in the prompt neutrino predictions , identification and separation of the astrophysical and prompt components needs to be made through their respective spectral signatures ( see fig . [ figure : atmosphericdiffusefluxes ] ) . .
the theoretical predictions and measurements for the atmospheric neutrino flux are shown , as well as the current estimate for the diffuse astrophysical neutrino flux.,scaledwidth=49.0% ] the icecube neutrino observatory is located at the south pole and is the first kilometer - scale cherenkov neutrino telescope .
an optical sensor array observes the cherenkov radiation from secondary charged particles produced in neutrino interactions deep in the ice .
these are dominantly neutrino - nucleon interactions except for the glashow resonance@xcite for electron anti - neutrinos at @xmath29 . based on the signature of the neutrino interaction , which depends on the flavor of the incident neutrino and the type of the interaction , two main detection channels exist .
searches in the muon channel look for charged - current muon neutrino interactions .
these have a muon in the final state whose direction is reconstructible with a resolution of about @xmath30 degree@xcite .
the large muon range also allows to detect neutrino interactions outside the instrumented volume .
the cascade channel comprises all other interaction scenarios which have particle showers in the final state . above energies charged current @xmath31 interactions exhibit more complex event signatures but at lower energies they appear in the cascade channel .
consequently , an astrophysical flux with equal neutrino flavor contributions would yield more cascade than track events starting inside of icecube .
if the neutrino interaction happens inside the detector , the cherenkov light yield of particle showers scales nearly linearly with the deposited energy , leading to an energy resolution that is better than in the muon channel . on the other hand
the angular resolution is rather poor ( @xmath32 for the completed icecube detector ) .
overall , the cascade channel is best suited for searches for diffuse astrophysical neutrinos in which the neutrino energy measurement is more important than pointing capabilities@xcite .
this paper presents searches for neutrino - induced cascades in one year of data taken during the construction phase of icecube , when about half the detector was operational ( icecube-@xmath0 ) .
the main objective of the searches was to identify an astrophysical flux of neutrinos .
in addition , a sensitivity to atmospheric neutrinos in the few @xmath18 energy range was maintained to allow a validation of the anticipated backgrounds in the data set .
a @xmath33 excess of events above @xmath14 was found , compatible with the all flavor astrophysical diffuse neutrino flux estimate obtained in icecube s high - energy starting events ( hese ) analysis@xcite . in comparison to that analysis , the icecube-@xmath0 cascade analysis provides an event sample with unprecedented low background contamination between @xmath34 and @xmath35 .
this is possible because both searches employed rather different event selection strategies .
methods outlined in this paper also prove powerful in cascade searches with later icecube configurations@xcite .
the paper is organized as follows : the icecube detector and icecube-@xmath0 dataset are described in section [ section : theicecubedetector ] .
the simulation used is presented in section [ section : simulation ] followed by a description of the cascade reconstruction in section [ section : reconstructionofparticleshowers ] .
the details of the event selection and expected sensitivity are presented in section [ section : eventselectionandanalysismethod ] .
a survey of the systematic uncertainties follows in section [ section : systematicuncertainties ] before the results and implications are discussed in section [ section : resultsanddiscussion ] .
a conclusion is given in section [ section : summary ] .
detector configuration .
the circles and squares are the positions of the strings .
the point @xmath36 is the centre of the complete @xmath37-string detector .
particle showers with reconstructed vertices inside the instrumented volume are called contained events .
the analyses presented in this work reject non - contained events in order to suppress atmospheric muons that enter the detector from the outside .
the blue dashed and solid lines show the two differently tight containment requirements that are used .
the strings denoted by red squares form the outer layer of the detector .
they are used to veto incident atmospheric muons.,scaledwidth=49.0% ] the icecube neutrino observatory@xcite consists of an in - ice array of optical sensors and a complementary surface air shower detector called icetop .
the analyses presented here utilized only the in - ice component so the following detector description will be limited to that .
the optical sensors , called digital optical modules ( doms)@xcite , are sensitive to cherenkov photons between @xmath38 and @xmath39 .
the doms are deployed between depths of @xmath40 and @xmath41 and are attached to strings that are formed by the readout cables .
each string has @xmath42 doms attached .
the vertical string spacing of the doms is approximately @xmath43 and the horizontal spacing between the strings is approximately @xmath44 .
the data for this analysis was collected between april 2008 and may 2009 with a total of @xmath45 days livetime . in this period @xmath0 strings
were deployed and operational .
the detector layout is shown in fig .
[ figure : ic40detector ] . before
icecube s completion in 2010 , the icecube-@xmath46 , and icecube-@xmath47 configurations took data with @xmath46 and @xmath47 deployed strings , respectively .
each dom consists of a @xmath48 diameter photomultiplier tube ( pmt)@xcite , made by hamamatsu photonics , and a data acquisition system housed within a pressure sphere made of @xmath49 thick borosilicate glass .
the pmt s dynamic range is @xmath50 photoelectrons per @xmath51 and it is designed to accurately record the amplitudes and widths of the pulses with a timing resolution of @xmath52 .
their peak quantum efficiency is approximately @xmath53 and they are operated at a gain of @xmath54 to resolve single photoelectrons .
the time - resolved pmt signal ( waveform ) is digitized in the dom . for this purpose
two digitization devices are available on the dom mainboard : two analog transient waveform digitizers ( atwd ) and a fast analog - to - digital converter ( fadc ) .
the atwds have three channels operated in parallel at different gains to provide a large dynamic range ( a fourth channel is used only for calibration purposes ) . due to scattering in the ice
the arrival times of photons emitted at the same point and time can vary by microseconds .
the atwds provide a sampling rate of @xmath55 over a time window of @xmath56 allowing them to record the earliest photons ( i.e. those least affected by scattering in the ice ) with high precision .
the second digitizer , the fadc , has a coarser sampling of @xmath57 recording data over a longer time period for photons with larger delays of up to @xmath58 . in order to reduce data readout volume due to noise , in icecube-@xmath0
a local coincidence criterion is required .
only if a neighbouring dom on the same string also detects light within the local coincidence time window of @xmath59 , the pmt response is digitized , time - stamped , and transmitted to the surface for analysis .
the surface data acquisition system combines the individual pmt responses and forms events when one of the several possible triggering criteria are fulfilled .
the trigger requirement for the icecube-@xmath0 cascade search was the simple multiplicity trigger , which requires that eight doms were hit within a @xmath60 time window .
the data rate for icecube-@xmath0 from this trigger was approximately @xmath61 .
interactions of all flavors of neutrinos were simulated to model atmospheric and astrophysical neutrinos .
the ` nugen ` software package maintained by the icecube collaboration was used .
it is based on the ` anis`@xcite neutrino generator , which produces neutrinos isotropically over the earth s surface and propagates them to interact in or near the detector volume .
neutrino attenuation and @xmath62 regeneration are accounted for using the prem earth model@xcite .
cteq5 structure functions@xcite were used to model the deep - inelastic neutrino - nucleon scattering cross section . throughout this paper
the diffuse astrophysical neutrino flux is simulated isotropically , with a flavor ratio of @xmath13 and , if not stated otherwise , with an unbroken power - law spectrum with index of @xmath19 and an all - flavor intensity of @xmath63 .
rate predictions for the atmospheric neutrinos are based on the hkkms07 model@xcite for conventional atmospheric neutrinos and the ers model@xcite for prompt atmospheric neutrinos .
extrapolations of the original calculations to higher energies provide rate predictions at the energy range relevant to this work .
the steepening of the cosmic - ray spectrum around a few @xmath64 ( the so - called `` knee '' ) causes a similar feature in the atmospheric neutrino spectrum which is not accounted for in the hkkms07 model .
a modification to the hkkms07 model@xcite was applied to account for the knee . for one of the presented analyses
the bartol model@xcite was used to estimate the conventional atmospheric neutrino flux .
compared to the modified hkkms07 model it predicts a higher @xmath26 contribution ( see fig .
[ figure : atmosphericdiffusefluxes ] ) .
the propagation of muons and taus through the detector and their energy losses were simulated using the ` mmc ` program@xcite and the cascade simulation inside the detector was handled by the ` cmc ` program@xcite .
neutrino - induced cascades below a threshold of @xmath65 were simulated as point - like light sources , emitting an angular cherenkov light profile typical of an electromagnetic shower@xcite .
cascades of higher energies are split into segments along the direction of the shower development .
each cascade segment is then approximated by a point - like sub - shower with a light yield proportional to the light yield in the corresponding segment of the electromagnetic cascade .
the elongation of electromagnetic cascades due to the suppression of bremsstrahlung and pair production cross sections above @xmath64 energies ( lpm effect@xcite ) is accounted for .
hadronic cascades are simulated as electromagnetic cascades with a smaller light yield per deposited energy to account for the neutral shower components which do not generate cherenkov light@xcite .
the contribution from atmospheric muon events is estimated from simulations done with a modified version@xcite of the ` corsika ` air shower simulation software@xcite .
a large number of background events must be generated due to the high background suppression that is necessary to reach an event sample dominated by neutrinos . providing
a large background sample is computationally challenging , mostly because of the sheer number of air showers needed but also due to the simulation of light propagation in the optically inhomogeneous ice .
the figure of merit used to quantify the statistics of a simulated data sample is the effective livetime @xmath66 , i.e. the time that one would have to run the real experiment to obtain the same statistical error as in the simulated dataset . .
bremsstrahlung cascades from muons originating in these showers are studied in order to assess the probability to obtain a bremsstrahlung cascade bright enough to pass event selection cuts from air showers below a given energy threshold .
the plot shows the fraction of proton air showers with primary energy @xmath67 below @xmath68 , @xmath34 , @xmath69 that exhibit catastrophic energy losses of @xmath70 above a given energy.,scaledwidth=49.0% ] the chemical composition of the cosmic rays is important for an estimate of the muon background .
previous cascade searches@xcite have shown that protons are of prime importance .
proton primaries produce lower multiplicities of muons with higher individual energies than heavier primaries . compared to air showers induced by heavier elements , which typically lead to muon bundles reaching the detector ,
proton showers are more likely to generate single high - energy muons .
if such a high - energy muon has a single catastrophic energy loss , then the relative light yield between shower and muon can make the resulting event signature look shower - like .
we use a simplified cosmic - ray composition model in our simulation that nevertheless reflects these qualitative differences between light and heavy elements .
this two - component model considers only the two extrema of cosmic - ray composition and comprises two broken power laws , one for the proton and one for the iron spectrum , respectively ( see@xcite and fig .
[ figure : eikecrspectrum ] ) .
the mean logarithmic mass of the all - particle spectrum formed from the sum of the two power laws is compatible with measurements of this quantity by air shower experiments@xcite .
the created background sample provides more than a year of effective livetime above an energy per nucleon threshold of about @xmath71 . below this threshold
the statistics drop rather quickly and these low energetic events are underrepresented in the event sample . however , the risk of underestimating the background is low , since the fraction of air showers with a primary energy below @xmath72 which exhibit bremsstrahlung losses of @xmath1 ( @xmath73 ) is only @xmath74 ( @xmath75 ) ( see fig . [
figure : ecascfraction ] ) .
cherenkov light , emitted by charged particles , traverses the optically inhomogeneous detector material and is then sampled by the three - dimensional icecube array with nanosecond precision . hence , the light s arrival times and amplitudes are the sole available information on any particle interaction in the detector .
previous analyses relied on reconstruction algorithms that used only parts of this information , i.e. by considering only the integrated amplitude of all light reaching a dom or by using only the arrival time of the first photon . a new algorithm ` credo ` was developed .
by using the waveform information ` credo ` is able to reconstruct not only the interaction vertex and the deposited energy but also the direction of the incident neutrino .
` credo ` is the first icecube cascade reconstruction algorithm which is able to estimate the direction of the neutrino .
the measurement process is described with a probabilistic model and a maximum likelihood estimator allows information on the vertex of the neutrino interaction , the neutrino direction and the deposited energy to be inferred .
` credo ` considers the information on an event in the form of time intervals @xmath76 in which a dom at position @xmath77 recorded a given amount of charge @xmath78 .
formally , the event is then described with the set @xmath79 , where the index @xmath80 runs over all such intervals .
it should be noted , that this set also contains time intervals in which no charge has been recorded , since these time intervals also carry constraining information .
the particle showers are modeled as point - like cherenkov emitters , in the same way as they are described in simulation .
a particle shower is then fully specified by @xmath81 parameters @xmath82 : the time and position of the vertex , two angles for the direction of the neutrino and the deposited energy .
the scattering and absorption of light in the antarctic ice are depth dependent .
the ` photonics ` package@xcite is used to calculate the light propagation in the ice and to create tables with light arrival times and amplitudes as a function of the depth and the relative position between particle shower and dom . in order to avoid binning effects the table values
are interpolated by multidimensional spline fits@xcite prior to being used in the reconstruction .
the ` photonics ` tables are used to calculate for each time interval @xmath80 an estimate of the mean expected charge @xmath83 . in this sum @xmath84
is the contribution due to the cascade @xmath85 and @xmath86 is a continuous noise contribution . in each time interval
a counting experiment is performed , and the probability of a given measurement @xmath87 under the condition @xmath85 can be calculated : @xmath88 from this probability one can construct a maximum likelihood estimator , yielding the parameters @xmath89 best supported by the measurement : @xmath90 the search for the global maximum in the @xmath81-dimensional likelihood space is performed by minimizing the negative logarithm of the likelihood using the simplex minimizer in the minuit software package@xcite . in order to avoid local minima the search is done iteratively , where in each step the minimizer starts at a different position in the parameter space .
the iterative minimization and the many necessary table lookups for each time interval form a time - consuming procedure that can not be applied to all events . however , this reconstruction finds the position of the particle shower with a resolution of @xmath91 horizontally and , due to the smaller dom spacing , @xmath92 vertically .
the energy resolution for an astrophysical neutrino spectrum is about @xmath93 and the angular resolution is about @xmath94 .
other variants of the likelihood reconstruction described here have been developed and been used in more recent analyses . by incorporating improvements in the understanding of the detector response and a better modeling of the ice they provide better resolutions and determine icecube s performance on cascades today@xcite .
a small number of neutrino generated showers need to be isolated from a large background of atmospheric muons .
this is achieved by finding and applying a set of conditions on the reconstructed event properties that neutrino induced showers fulfill , but atmospheric muons do not . in practice
this selection is implemented as a multi - step process .
the first steps in the process are aimed at conservatively reducing the background , allowing the use of computationally intensive reconstruction algorithms , that can be applied to a reduced set of events only .
each step in the event selection process is referred to as a `` level '' .
level @xmath30 , the trigger condition , was described in section [ section : theicecubedetector ] .
level @xmath95 , the online filter applied at south pole , and subsequent levels are outlined in this section .
the section will first introduce the classification schemes and cut variables used for separation of the signal from the background and then turn to a description of the individual samples .
the background rejection criteria used in neutrino cascade searches can be classified conceptually as belonging in the following four categories . * reconstructed particle parameters : * four different likelihood reconstruction routines are used in the searches described here .
the routines differ in runtime , precision , initial assumptions and in the number of neutrino parameters that they infer from the event .
three of the reconstruction routines return parameters with the hypothesis that the event is a particle - shower .
the fourth routine used assumes the event contains a muon track . at earlier cut levels
the ` cscd_llh ` likelihood reconstruction@xcite is used .
it does nt account for the optical inhomogeneities of the ice but provides a quick estimate for the vertex and a quality parameter @xmath96 that describes how well the event fits to the cascade hypothesis . using the vertex estimate from ` cscd_llh ` another estimate for the deposited energy , @xmath97 , is provided by the ` atmcscdenergyreco ` algorithm@xcite .
it is quick to compute and considers the optical inhomogeneities of the ice . at later levels the ` credo ` algorithm
described in section [ section : reconstructionofparticleshowers ] provides the best estimates for the reconstructed vertex @xmath98 , deposited energy @xmath99 and zenith angle @xmath100 .
the superscript @xmath101 distinguishes , where necessary , applications of the ` credo ` algorithm with differing numbers of iterations .
all events are also reconstructed with another likelihood reconstruction@xcite which assumes that the event contains a muon track .
this routine gives a zenith angle estimate @xmath102 and provides a quality of fit parameter @xmath103 for the track hypothesis .
the zenith angle estimator correctly identifies much of the muon background as downgoing .
particle showers are preferentially interpreted as either horizontally or diagonally passing tracks which allows for some signal - background separation .
* containment & vetoing : * a particularly problematic background are muon events which pass close to the edges of the detector producing a light distribution which is similar to that produced by cascades . to counter this background
various containment conditions are placed on the position of the reconstructed vertex , the first - hit dom and the dom with the highest collected charge .
events are excluded if the first hit dom or the dom with the highest collected charge is located in the outermost vertical layer ( see fig .
[ figure : ic40detector ] ) .
events are also vetoed if the depth of the first hit dom @xmath104 falls in the top or bottom @xmath105 of the detector .
the reconstructed vertex @xmath106 is required to lie inside the instrumented volume and not in the top or bottom @xmath105 of the detector .
the analyses presented in this paper found different containment conditions on the @xmath107-position of the vertex to be optimal , when combined with their particular selection cuts , for suppressing background while maintaining signal efficiency .
the two alternative containment conditions are illustrated by the solid and dashed lines in fig .
[ figure : ic40detector ] . * topological characteristics : * the hit patterns of particle showers in icecube are approximately spherical while the muon - track hit patterns are more elongated .
a number of different quantities can be calculated to characterize the different topology of cascade and muon events and used to preferentially select cascades .
for example , a quantity analogous to the tensor of inertia of a rigid body , is calculated for each event .
the collected charge on each dom takes the role of the rigid body s mass distribution .
the ratio of the smallest eigenvalue to the sum of all of the eigenvalues , @xmath108 , tends to @xmath109 for spherical events while muon tracks typically have smaller eigenvalue ratios@xcite .
another way to select spherical events is to construct a spherical volume surrounding the reconstructed event vertex @xmath106 and consider the proportion of hit doms , versus the total number of doms , in this sphere .
the radius of the sphere considered is chosen to scale with the average distance between reconstructed vertex @xmath106 and position of the hit doms a robust estimate for the overall size of the hit pattern .
the fill - ratio @xmath110 denotes the fraction of doms , falling within the sphere , on which light is recorded@xcite .
hence , fill - ratios close to @xmath111 are obtained for spherical hit patterns , while muon events yield lower values .
the fill - ratio is especially efficient to suppress coincident muon events two or more muons from different air showers that cross the icecube detector within microseconds of each other .
a second quantity , the difference @xmath112 of two fill - ratios with different radii , is also used .
this quantity provides further separation power due to the fill - ratio s dependency on the chosen radius being slightly different for the differently shaped hit patterns of cascades and tracks .
* time evolution and charge distribution : * below @xmath64 energies the cherenkov light of particle showers originates within a few meters of the interaction vertex and then propagates through the detector with the speed of light in ice @xmath113 . in contrast , muon tracks traverse the detector with velocities close to the speed of light in vacuum @xmath114 and emit cherenkov photons continuously along their track .
several approaches exploit this difference to separate cascade and track events .
a simple approach , that can be applied before the event vertex has been reconstructed , is the ` linefit ` algorithm .
the hit pattern is fitted with a straight line propagating with velocity @xmath115@xcite .
relativistic muons often yield ` linefit ` velocities close to @xmath114 whereas lower velocities are obtained if the fit is applied to cascades .
if the interaction time and vertex are well reconstructed then causality can be used to provide a strong constraint on whether an event is a neutrino - induced particle shower event . for each dom
, light is expected to arrive at the earliest after the time necessary to cover the distance between the dom and the vertex , at velocity @xmath113 .
while delayed photon arrivals are common due to light scattering in the ice , much earlier arrival times indicate a problem with the cascade hypothesis .
the difference between the expected ( @xmath116 ) and observed ( @xmath117 ) arrival time is calculated for all doms and @xmath118 is defined as the smallest such delay time @xmath119 .
events with large negative values of @xmath118 are removed .
another approach is to sort the doms by the time they recorded light and then consider the unit vector from one hit dom to a subsequent hit dom as an individual dipole moment .
the global dipole moment @xmath120 is obtained by averaging over all individual dipole moments .
larger moments are expected from tracks and smaller moments from cascades@xcite . yet another way to emphasize
the different hit pattern evolution of tracks and cascades is to divide the event into two parts based on the times of the hit doms .
this splits tracks into two disjunct segments and cascades into two mostly concentric shells .
the ` cscd_llh ` algorithm is used to obtain , for each half , the vertices @xmath121 and @xmath122 .
large radial and vertical distances @xmath123 and @xmath124 between the vertices as well as large differences in the reconstructed time @xmath125 are then indicative of tracks .
another feature of cascade events , which is also related to the fact that the particle showers are only a few meters in length , is that most of the light is recorded close to the vertex and hence early in the event .
in contrast , since muons emit light continuously along their track it is more likely to see later contributions to the total charge .
the variable @xmath126 denotes the fraction of the event length in which the total collected charge rises from @xmath127 to @xmath10 .
greater @xmath126 values indicate a longer time interval for the second half of the event s total charge to be collected and are more likely to occur for muon events .
some discrimination power comes from doms where just enough light arrives to trigger the readout and hence only a single pulse is reconstructed from the waveform . a combination of light yield , scattering in the ice and the geometric shape of the hit pattern results in a slightly higher number of these doms for muons .
the ratio @xmath128 of doms with only one reconstructed pulse over the total number of hit doms tends to smaller values for cascades .
finally , a useful variable is the ratio of the charge collected in the dom with the highest charge , compared to the total recorded charge @xmath129 .
low energetic muons passing very close to one dom can yield a high charge concentration in this dom compared to others .
these events can resemble cascade - like hit patterns and are prone to be overestimated in energy .
hence , requiring a low ratio @xmath129 is useful to reject this class of background events .
three event selections , named slowromancap1@a , slowromancap1@b and slowromancap2@ , were developed to search for neutrinos in the icecube-@xmath0 dataset .
the multiple event selections allowed sensitivity to both high - energy astrophysical and low - energy atmospheric neutrinos as well as providing a collaboration internal cross - check .
the different event selections rely on similar selection methods that are differently combined for the individual goals of the analyses .
the event selections share the first three filter levels but differ at later filter steps , since analysis slowromancap1@a aimed at measuring atmospheric neutrinos whereas analyses slowromancap1@b and
slowromancap2@ were optimized towards an astrophysical neutrino flux . in order to avoid experimenter s bias a blind analysis
was performed .
each event selection was developed and tested on simulation and a @xmath130 subset of the experimental data , sampled uniformly over the year .
a simple cut - and - count experiment was done on the remaining @xmath10 .
analyses slowromancap1@a and slowromancap1@b were developed in parallel to analysis slowromancap2@ , but only unblinded afterwards .
the crucial difference is that they utilized a significantly improved background simulation described in section [ section : simulation ] , resulting in the choice of tighter cuts and a purer neutrino sample .
consequently here the focus is put on samples slowromancap1@a and slowromancap1@b .
an overview of the cuts performed to obtain the three samples is given in table [ table : cutcomparison ] .
a comparison of the energy thresholds and event rates is given in table [ table : eikeunblindingresults ] .
sizable systematic uncertainties must be considered in interpreting the result .
therefore for samples slowromancap1@a and slowromancap1@b a bayesian approach was chosen ( described in appendix [ section : bayesianmethod ] ) and the result is reported in form of the posterior probability for the number of non - background events .
triggered events are filtered online at the south pole to reduce the data volume to the bandwidth available for data transfer from the pole via satellite .
the online filter , or level @xmath95 is a filter on easy - to - compute variables that retains the majority of the neutrino signal while removing a large fraction of the background .
it requires the ` linefit ` velocity to be @xmath131 and the tensor - of - inertia eigenvalue ratio to be @xmath132 .
the distributions of these variables are shown in figs .
[ subfigure : polefilter_1d_lfvel ] and [ subfigure : polefilter_1d_evalratio ] . at this early stage of the event selection , where no event quality cuts are done yet ,
the distributions show some disagreement between data and simulation .
further steps in the event selection will reduce this disagreement , mostly by removing events that are too low in energy or that are not contained in the instrumented volume .
the pole filter reduced the data rate by two orders of magnitude to approximately @xmath133 . due to the energy threshold of about @xmath134 and the steeply falling energy spectrum ,
the energy distribution of background events , passing level 2 is strongly peaked at lower energies . as the majority of the events
are still atmospheric muons most of the events are reconstructed as downgoing tracks ( see fig . [
figure : level3_2d_zenith_energy ] ) .
consequently , the level @xmath135 cuts concentrate on reducing the background at low energies .
all events with a reconstructed energy @xmath136 are kept , whereas lower energetic events are subject to two more cuts .
first , when reconstructed under a muon track hypothesis , events reconstructed as downgoing with zenith angles @xmath137 are rejected .
secondly , those events are removed , where a ` cscd_llh ` likelihood value of @xmath138 indicates a poorly fitted event . after application of the level @xmath135 filter the data rate is reduced by another order of magnitude to approximately @xmath139 , while contained astrophysical ( atmospheric ) neutrinos were kept with an efficiency of @xmath140 ( @xmath141 ) . after level @xmath135 ,
the analyses diverge .
event selection slowromancap1@a aims at the observation of atmospheric neutrinos with energies of a few @xmath18 .
previous cuts aimed to provide an optimal efficiency for cascades regardless of their energy and position inside the detector . as a consequence at level @xmath135 a significant number of events , that are very low in energy or happening at the border of the instrumented volume , remain in the sample . for these events no reliable separation between signal and background could be found .
therefore , a set of geometric conditions has been defined to remove such events : it is required that the first hit dom is neither on the outer layer of icecube-@xmath0 ( red squares in fig .
[ figure : ic40detector ] ) nor in the top or bottom @xmath142 of the detector . also doms on at least five different strings must have recorded light .
this cut reduces the data rate to @xmath143 , while contained astrophysical ( atmospheric ) neutrinos are kept with an efficiency of @xmath144 ( @xmath145 ) .
after these cuts , at the so - called level @xmath146 , additional reconstructions but no cuts were performed . in particular
the remaining sample was small enough to perform the ` credo ` likelihood reconstruction described in section [ section : reconstructionofparticleshowers ] with @xmath147 iterations .
muon background still dominates over atmospheric neutrino - induced showers by three orders of magnitude at selection level @xmath146 .
therefore , the vertex and energy estimates are used to apply another series of event selection conditions . at level @xmath148 ,
the energy threshold is set to @xmath149 . in order to remove events where low energetic muons pass very close to a dom and where their energy is likely to be overestimated ,
a cut on @xmath150 is applied .
furthermore , the containment requirement is enforced : the reconstructed vertex must lie inside the blue solid boundary shown in fig .
[ figure : ic40detector ] , its @xmath151-coordinate must be inside the instrumented depth interval and also the dom with the highest recorded charge in the event may not be on the outer layer . a constraint on the minimum delay time @xmath152 is used to remove events where causality rules out the cascade hypothesis .
together with requiring a fill - ratio @xmath153 , this set of cuts reduces the data rate by two orders of magnitude down to @xmath154 , while retaining efficiencies for contained astrophysical and atmospheric neutrinos of @xmath155 and @xmath156 , respectively .
overall agreement of event properties between simulation and experimental data is good ( see fig . [
figure : tmvaeike_inputvariables ] ) .
this is a necessary precondition for the final event selection step , which uses a multivariate algorithm to distinguish signal and background events .
the final step of the event selection is to remove the remaining background and obtain a sample dominated by neutrinos . for this purpose event properties that still provide separation power
are used as input to a multivariate algorithm to obtain a single quality parameter for each event .
the tmva package@xcite is used to train a boosted decision tree ( bdt ) . in total @xmath157 variables
are combined into one final event quality variable @xmath158 .
these are the ` cscd_llh ` algorithm s quality parameter @xmath96 , the vertical and radial distances @xmath159 of the time - split reconstructions , the fraction of doms with only one reconstructed pulse @xmath128 and the fraction of the event duration needed to accumulate the second half of the total charge @xmath126 .
the two zenith angle estimators @xmath160 and @xmath102 enter the bdt as well as the dipole moment @xmath120 and the fill - ratio difference for two radii @xmath112 . finally , the eigenvalue ratio @xmath161 and the minimum delay time @xmath162 are used , too .
they have already been used at earlier cut levels but still provide some discrimination power .
the four variables ranked most important by the classifier are shown in fig .
[ figure : tmvaeike_inputvariables ] . the distribution of the bdt output variable and its ability to separate signal from background is shown in fig .
[ figure : eikebdt ] . at low values ,
the distribution is dominated by the atmospheric muon background and the experimental data is adequately described by the ` corsika ` simulation .
the signal distribution centers at higher bdt scores , but even there a contribution from atmospheric muons is present .
closer inspection reveals that these events are muons with prominent bremsstrahlung cascades and little to no hint of the muon track .
the events resemble neutrino - induced particle showers rather well .
the geometry of icecube-@xmath0 with one dimension being much shorter than the others is obviously vulnerable to this class of background events . in the energy range below @xmath14 , with the given detector and the aforementioned background rejection methods , including their combination within a multivariate classifier , these events turn out to be irreducible background .
this limits the possible background suppression for sample slowromancap1@a .
atmospheric neutrino signal and muon background are similarly distributed at @xmath158 scores @xmath163 , so cutting at a higher value removes the atmospheric neutrino signal as much as the atmospheric muon background . as a consequence
a @xmath158 value of @xmath164 is the optimal separation point between background and signal that maximizes the detection potential@xcite for an atmospheric neutrino flux .
however , a rather large number of @xmath165 events is expected in sample slowromancap1@a from which @xmath3 are expected to be atmospheric muons and @xmath4 conventional and prompt neutrinos .
both numbers are affected by rather large systematic uncertainties , which are discussed in section [ section : systematicuncertainties ] . .
a cut at @xmath166 defines the sample slowromancap1@a .
the atmospheric muon and neutrino components are stacked on top of each other and describe together the recorded data well .
the experimental data shown comprises @xmath111 of the sample . for bins in which no event was observerd a @xmath167 c.l .
upper limit is shown .
the white hatched area shows the distribution of atmospheric muons and neutrinos , including systematic and statistical uncertainties .
the orange hatched area denotes the prediction of astrophysical neutrinos according to the flux estimate from@xcite.,scaledwidth=49.0% ] the remaining background events can be removed by increasing the energy threshold of the analysis .
because of their steeply falling spectrum conventional atmospheric neutrinos are severely reduced by an energy cut . however , the harder spectra of prompt neutrinos and the assumed diffuse astrophysical neutrino flux are not so much affected .
this fact is used to define the data sample slowromancap1@b with an energy threshold of @xmath168 . a comparatively loose requirement on the bdt score of @xmath169
is then sufficient to remove all simulated background events ( see fig .
[ figure : eikeparspace ] ) .
in contrast to sample slowromancap1@a , which was designed to find atmospheric neutrinos , sample slowromancap1@b has a significantly better performance for detecting a prompt or astrophysical neutrino flux . ) and the red shaded area ( @xmath170 ) denote the cuts of samples slowromancap1@a and slowromancap1@b , respectively .
the color scale denotes linearly the expected / seen number of events per bin in @xmath111 of the livetime.,scaledwidth=49.0% ] for this sample , the background contribution from atmospheric muons can not be determined from simulation anymore as all events have been removed . in order to estimate the muon rate ,
the reconstructed energy distribution is extrapolated from the low - energy region , where simulated events are still available , to higher energies . up
to @xmath64 energies , the energy spectrum of the brightest bremsstrahlung cascades along simulated muon tracks can be described by a power - law @xmath171 . for muon events , which pass the event selection , the corresponding energy spectrum is not necessarily the same , because of the energy - dependent performance of the cuts .
e.g. muon events with bright bremsstrahlungs cascades are suppressed with increasing energy , since the muon becomes more likely to be detected in the veto region . on the other hand ,
the selection efficiency for particle showers rises with increasing deposited energy .
hence , the model used to extrapolate the muon background into the signal region is the product of two functions : the selection efficiency for particle showers as a function of deposited energy ( derived from simulation ) and a power - law with the index and normalization as free parameters .
this model provides a reasonable fit to the energy distribution of background muons below the energy threshold ( see fig . [
figure : eikehighe ] , the same @xmath158 score @xmath172 is required for these events as for the events in the high - energy sample ) .
the fit is consistent with the expectation that the remaining background events are dominated by high bremsstrahlung cascades and that the energy estimator describes their energy spectrum .
this confirms the physical motivation of the extrapolation in the signal region . from the extrapolation
one can amount the muon contribution to @xmath173 events , where the errors are derived from varying the parameters within the uncertainties obtained from the fit .
an additional @xmath16 events are expected from conventional and prompt atmospheric neutrinos .
this event selection reduced the muon background by seven orders of magnitude from the rate after the online filter .
this high suppression comes at the price of a low total efficiency of @xmath174 ( @xmath175 ) for contained astrophysical ( atmospheric ) neutrinos .
the energy - dependent selection efficiencies are presented in the form of effective areas for both samples in fig .
[ figure : eikeeffectivearea ] .
table [ table : eikeefficiencies ] summarizes the performance of the event selection at different cut levels .
.event rates and cut efficiencies at the different levels leading to samples slowromancap1@a and slowromancap1@b . while the data column refers to all recorded events that pass the cuts , the neutrino rates refer to contained events , i.e. neutrinos which have their interaction vertex inside the area circumscribed by the blue solid line in fig .
[ figure : ic40detector ] . for
charged current @xmath176 interactions the muon must have its highest energy loss inside the area . [ cols="<,^,^,^ " , ] were found in the samples slowromancap2@ and slowromancap1@b .
an additional event was already found in the @xmath130 sample used to develop the cuts .
the event displays visualize the light distribution in the detector : each dom is shown as a sphere , which size scales with the recorded charge by the dom .
the color coding illustrates the arrival time of the light at the dom ranging from red ( early hits ) over green to blue ( late hits).,scaledwidth=49.0% ] credible interval of this work and the hatched area denotes the expected number of events according to the best fit flux in@xcite , including a cutoff at @xmath177 .
the solid and dashed lines show the @xmath10 cl upper limits derived from sample slowromancap2@ and the icecube-@xmath46 diffuse muon search@xcite , respectively .
the upper @xmath178-axis transforms an expected event count into an all - flavor flux normalization of an unbroken @xmath179 power - law . in transforming the flux estimate and cutoff from@xcite to an unbroken power - law ,
a slightly lower flux normalization is obtained.,scaledwidth=49.0% ] the results from the sample slowromancap1@a , slowromancap1@b and slowromancap2@ are consistent , with an excess appearing only towards larger energies .
in addition , one can compare the overlap between the samples as a crosscheck .
due to the different energy thresholds , one expects sample slowromancap2@ to be largely a subsample of slowromancap1@a .
however , due to the difference in selection criteria and general low efficiency of only @xmath180 with which signal events are selected , a perfect overlap can not be expected .
we find that of the @xmath6 events of sample slowromancap2@ , @xmath81 are also contained in slowromancap1@a , where the majority of the remaining have not passed the different containment cut .
all three high - energy events of sample slowromancap1@b are also contained in sample slowromancap2@. only two of them are also in sample slowromancap1@a .
we have reported on a search for astrophysical and atmospheric neutrino - induced particle showers in the icecube-@xmath0 detector .
the data was taken between april 2008 and may 2009 with a total of @xmath45 days livetime .
three different event selections , slowromancap1@a , slowromancap1@b and slowromancap2@ , each tailored to different energy ranges , were developed and applied to the data .
the low and intermediate energy samples slowromancap1@a and slowromancap2@ have an energy threshold of @xmath1 and @xmath5 , respectively , and have sensitivity to atmospheric neutrinos . in both samples , an excess over the atmospheric muon background
was observed that can be well explained by conventional atmospheric neutrinos . with an energy threshold of @xmath5
, analysis slowromancap2@ also has sensitivity to a high - energy astrophysical neutrino flux . in the absence of an excess of events above the combined background of atmospheric neutrinos and muons
we are left with an upper limit ( @xmath10 confidence level ) of @xmath181 on the all - flavor astrophysical neutrino flux , assuming an @xmath179 spectrum and a @xmath13 neutrino flavor ratio at the earth . for this limit
, @xmath10 of the expected signal events have energies in the range between @xmath5 to @xmath11 .
the upper limit is below that reported from cascade searches using the icecube-@xmath182 detector@xcite and approaches the waxman - bahcall limit@xcite .
finally , sample slowromancap1@b was optimized towards the largest sensitivity for high energy astrophysical neutrinos .
compared to sample slowromancap2@ , it profited from a larger sample of simulated muon background during the optimization of the event selection . in @xmath10 of the available data ,
three events were observed above the energy threshold of @xmath14 , with an expectation of only @xmath183 events from atmospheric neutrinos ( both conventional and prompt ) as well as atmospheric muons a @xmath17 excess .
the @xmath130 burn sample contains a fourth event , which has not entered the significance calculation .
the excess seen above @xmath14 in sample slowromancap1@b is noteworthy . although not significant enough to claim evidence for an astrophysical neutrino flux , it is consistent with similar excesses found in diffuse neutrino searches with icecube-@xmath46@xcite using muon neutrino events , the two @xmath64 events@xcite and the @xmath184 events found in@xcite , the strongest single evidence that icecube is seeing a high - energy neutrino flux of astrophysical neutrinos .
this analysis thereby provides three neutrino event candidates between @xmath185 and @xmath186 an intermediate energy scale with unprecedented low background contamination of conventional atmospheric neutrinos and muons .
the constraints on the all - flavor normalization of the high - energy neutrino flux of astrophysical neutrinos with equal flavor contributions are summarized in fig .
[ figure : posterior_highe ] .
a @xmath10 credible interval covers the range @xmath187 and is compatible with the more stringent flux estimate established by@xcite .
however , the sensitivity towards a diffuse flux of high - energy neutrinos was reached using data from only @xmath127 of the final icecube configuration .
it is the good energy resolution and the little intrinsic background associated with the signature of neutrino - induced cascades , that is thereby providing the large sensitivity to the diffuse flux@xcite .
the icecube detector is now completed with @xmath37 strings and an instrumented volume of @xmath188 .
future cascade searches benefit from the more favorable detector geometry , which allows for even better suppression of the background from atmospheric muons and improved cascade detection efficiency@xcite .
neutrino - induced cascades will hence continue to play a prime role in further exploring the high - energy astrophysical neutrino flux .
the experiment under consideration is a counting experiment in the presence of background . in order to incorporate systematic and statistical uncertainties into the interpretation of the result
, a bayesian approach was chosen . the probability to have observed @xmath189 events in the presence of @xmath190 events , signal selection efficiency @xmath191 and @xmath192 background events
is given by the poisson probability : @xmath193 from having observed @xmath189 we want to infer , whether a given value of @xmath190 is supported or ruled out by the experimental result .
this information is given by the posterior probability of @xmath190 , which can be obtained by applying bayes theorem . in order to use the theorem , available information on the expected signal as well as the uncertainty of the other parameters
must be quantified in the form of priors . by marginalizing over all parameters other than @xmath190 ,
remaining uncertainties are then incorporated into the final result .
the background uncertainties from model predictions and selection efficiencies are described by @xmath194 . for the signal we model all uncertainties with @xmath195 and the prior belief with @xmath196 .
the posterior probability can then be calculated : @xmath197 a constant is chosen for the signal prior to reflect no prior knowledge on the signal .
for the background prior a gaussian is used .
the mean is centered at the rate prediction , the width represents the modeled uncertainty and it is truncated at zero since rates have to be positive .
the uncertainty in the signal efficiency is modeled with the factor @xmath191 that is applied to the number of signal events after all cuts .
the prior for @xmath191 is modeled with a gaussian centered at @xmath30 and having a width corresponding to the systematic uncertainty .
it is also truncated at zero .
we acknowledge the support from the following agencies : u.s .
national science foundation - office of polar programs , u.s .
national science foundation - physics division , university of wisconsin alumni research foundation , the grid laboratory of wisconsin ( glow ) grid infrastructure at the university of wisconsin - madison , the open science grid ( osg ) grid infrastructure ; u.s .
department of energy , and national energy research scientific computing center , the louisiana optical network initiative ( loni ) grid computing resources ; natural sciences and engineering research council of canada , westgrid and compute / calcul canada ; swedish research council , swedish polar research secretariat , swedish national infrastructure for computing ( snic ) , and knut and alice wallenberg foundation , sweden ; german ministry for education and research ( bmbf ) , deutsche forschungsgemeinschaft ( dfg ) , helmholtz alliance for astroparticle physics ( hap ) , research department of plasmas with complex interactions ( bochum ) , germany ; fund for scientific research ( fnrs - fwo ) , fwo odysseus programme , flanders institute to encourage scientific and technological research in industry ( iwt ) , belgian federal science policy office ( belspo ) ; university of oxford , united kingdom ; marsden fund , new zealand ; australian research council ; japan society for promotion of science ( jsps ) ; the swiss national science foundation ( snsf ) , switzerland ; national research foundation of korea ( nrf ) ; danish national research foundation , denmark ( dnrf )
|
we report on the search for neutrino - induced particle - showers , so - called cascades , in the icecube-@xmath0 detector .
the data for this search was collected between april 2008 and may 2009 when the first @xmath0 icecube strings were deployed and operational .
three complementary searches were performed , each optimized for different energy regimes .
the analysis with the lowest energy threshold ( @xmath1 ) targeted atmospheric neutrinos .
a total of @xmath2 events were found , consistent with the expectation of @xmath3 atmospheric muons and @xmath4 atmospheric neutrino events .
the two other analyses targeted a harder , astrophysical neutrino flux .
the analysis with an intermediate threshold of @xmath5 lead to the observation of @xmath6 cascade - like events , again consistent with the prediction of @xmath7 atmospheric neutrino and @xmath8 atmospheric muon events .
we hence set an upper limit of @xmath9 ( @xmath10 c.l . ) on the diffuse flux from astrophysical neutrinos of all neutrino flavors , applicable to the energy range @xmath5 to @xmath11 , assuming an @xmath12 spectrum and a neutrino flavor ratio of @xmath13 at the earth .
the third analysis utilized a larger and optimized sample of atmospheric muon background simulation , leading to a higher energy threshold of @xmath14 .
three events were found over a background prediction of @xmath15 atmospheric muon events and @xmath16 events from the flux of conventional and prompt atmospheric neutrinos . including systematic errors this corresponds to a @xmath17 excess with respect to the background - only hypothesis .
our observation of neutrino event candidates above @xmath14 complements icecube s recently observed evidence for high - energy astrophysical neutrinos .
|
it is an intriguing problem with a rather long history to explore the possible existence of hadronic excitations in the quark - gluon plasma ( qgp ) phase@xcite .
the existence of the hadronic ( mesonic ) excitations in the light - quark sector was first predicted as being the soft mode associated to the chiral transition@xcite .
the recent lattice qcd simulations also suggest the existence of the mesonic bound states in heavy - quark sector in the qgp phase@xcite . in this work
, we investigate the effect of these bosonic excitations on the quark quasi - particle picture in the qgp phase@xcite .
we first explore how the soft mode of the chiral transition@xcite affects the quark spectrum in the qgp phase using the nambu - jona - lasinio(njl ) model , and show that a novel three - peak structure is formed in the quark spectrum near the critical temperature @xmath0@xcite .
we shall elucidate the mechanism and show the universality of the emergence of the three - peak structure in the fermion spectral function , employing a yukawa model composed of a massless fermion and a massive boson@xcite : we show that the complicated spectra originate from the mixing between a quark ( anti - quark ) and an anti - quark hole ( quark hole ) caused by a `` resonant scattering '' of the quasi - fermions with the thermally - excited massive boson .
to explore the quark matter near @xmath0 , we first employ the two - flavor nambu jona - lasinio ( njl ) model in the chiral limit @xmath1,\end{aligned}\ ] ] with the coupling constant @xmath2 gev@xmath3 and the three dimensional cutoff @xmath4 mev .
this model gives a second order phase transition at @xmath5mev for vanishing quark chemical potential , and the soft modes of the phase transition are dynamically formed in the scalar and pseudoscalar channels .
the spectral function in these channels shows a pronounced peak in the time - like region , the peak position and the width of which become smaller and vanishes eventually , as @xmath6 approaches @xmath0@xcite .
now notice that the soft modes composed of quarks can in turn contribute to the quark self - energy @xmath7 , which may be given in the random phase approximation ( rpa ) as shown diagrammatically in fig .
[ fig : diagram ] : in the imaginary time formalism it reads , @xmath8 where @xmath9 are the matsubara propagators of the scalar and pseudo - scalar channels and @xmath10 the free quark propagator .
the spectral functions of the quark and anti - quark are expressed as @xmath11 @xmath12 $ ] , with the retarded green function of the quark @xmath13 and the projection operators @xmath14 . in the left panel of fig .
[ fig : spectrum ] , we show the quark spectral function @xmath15 for @xmath16 , i.e. , slightly above @xmath0 .
one sees a three - peak structure in the spectral function .
our numerical calculation shows that the clear three - peak structure survives up to @xmath17@xcite . in the njl and yukawa models .
the left and right panels show @xmath15 in the njl model for @xmath18@xcite , and in the yukawa model for @xmath19@xcite , respectively .
there appears the three - peak structure in both models .
, scaledwidth=42.0% ]
we have seen that `` elementary '' bosonic excitations with a small width significantly modify the quark spectrum at finite temperature . to elucidate the mechanism of the modification of the fermion spectra on the more general ground
, we employ the yukawa model composed of a massless quark field @xmath20 and an elementary massive scalar boson @xmath21 ; @xmath22 in this model , the quark self - energy in the imaginary time formalism at the one - loop order is expressed as @xmath23 where @xmath24^{-1}$ ] is the matsubara green function for the scalar boson .
taking the matsubara summation and the analytic continuation , we obtain the self - energy in the real time . in the right panel of fig .
[ fig : spectrum ] , we show @xmath25 at @xmath26 and @xmath27 .
one sees a clear three - peak structure in the spectral function at intermediate temperatures around @xmath28@xcite , as in the njl model .
in the yukawa model for several values of @xmath6 . @xmath29 is the mass of the scalar boson .
right figures : parts of the physical processes which im@xmath30 describes .
the thick lines represent the quasi - quarks , the thin lines the on - shell free quarks and the wavy lines the scalar boson .
, scaledwidth=45.0% ] in order to understand the mechanism of the novel structure in @xmath31 , we show the imaginary part of @xmath32 in the yukawa model with several @xmath6 in the left panel in fig .
[ fig : sigma ] : there are two peaks in @xmath33 with a positive and negative energy , which grow rapidly as @xmath6 is raised .
it can be easily checked that these peaks in the positive and negative energy regions physically correspond to the landau dampings shown in the right panel in fig .
[ fig : sigma](a ) and ( b ) , respectively ; the wavy line represents the scalar boson .
we remark that the incident anti - quark line in fig .
[ fig : sigma](a ) describes a thermally excited antiquark , which disappears after the collision with the scalar boson .
the disappearance of the anti - quark means the creation of a hole in the thermally - excited anti - quark distribution@xcite .
figure [ fig : sigma](b ) describes the decay process of a quasi - quark which is a mixed state of quarks and antiquark - holes to an on - shell quark .
the two processes induce a respective quark-`antiquark hole ' mixing and hence two gaps are realized in @xmath34 at @xmath35 where im@xmath36 has a peak corresponding to each process .
the mixing mechanism of the quarks can be understood in terms of the notion of _ resonant scattering _ as in the case of the ( color-)superconductivity @xcite , although a crucial difference arises owing to the different nature of the bosonic modes . in the case of the superconductivity , the precursory soft mode which induces
the mixing is diffusion - mode like and has a strength around @xmath37@xcite .
the resonant scattering with the soft mode induces the mixing between a particle and a hole , and gives rise to a gap - like structure in the fermion spectrum around the fermi energy@xcite ; correspondingly , the imaginary part @xmath38 of the quark self - energy has a single peak around the fermi energy@xcite . in the present cases ,
the bosonic modes have a strength at finite energies and hence the resonant scattering of the quarks bring the two peaks in @xmath38 at finite energies and induces a mixing between a quark ( antiquark ) and an antiquark - hole ( quark - hole ) .
thus the two gaps in the quark spectrum are formed in the positive- and negative - energy regions . in summary , using the njl model
, we have shown that the quark spectrum in the qgp phase near the critical temperature of the chiral transition can have a three - peak structure in the low - momentum region owing to the effect of the hadronic soft modes of the chiral transition .
it has been shown that the three - peak structure in the fermion spectra also arises in the yukawa model with an elementary massive boson .
the detailed analysis of this model have elucidated the mechanism for the appearance the three - peak structure , which is realized as a result of the mixing between a quark(anti - quark ) and antiquark - hole ( quark - hole ) induced by the _ resonant scattering _ of the quarks . t.k .
is supported by grant - in - aid for scientific research by monbu - kagakusyo ( no .
17540250 ) .
is supported by the 21st century coe progaram of nagoya university and grant - in - aid for scientific research by monbu - kagakusyo ( no .
18740140 ) . t. matsui and h. satz , phys .
b * 178 * , 416 ( 1986 ) .
m. asakawa and t. hatsuda , phys .
lett . * 92 * , 012001 ( 2004 ) ; s. datta , et al . ,
d * 69 * , 094507 ( 2004 ) ; t. umeda , et al .
j. c * 39s1 * , 9 ( 2005 ) .
m. kitazawa , t. kunihiro and y. nemoto , phys .
lett . * b633 * , 269 ( 2006 ) .
m. kitazawa , t. kunihiro and y. nemoto , in preparation .
m. kitazawa , t. kunihiro and y. nemoto , phys .
b631 * , 157 ( 2005 ) .
m. kitazawa , t. koide , t. kunihiro and y. nemoto , phys .
d * 65 * , 091504 ( 2002 ) ; prog .
. phys . * 114 * , 205 ( 2005 ) .
m. kitazawa , t. koide , t. kunihiro and y. nemoto , phys .
d * 70 * , 056003 ( 2004 ) .
|
we investigate , using a chiral effective model , the quark spectrum in the critical region of the chiral transition focusing on the effect of the possible mesonic excitations in the quark - gluon plasma phase .
we find that there appears a novel three - peak structure in the quark spectra .
we elucidate the mechanism of the appearance of the multi - peak structure with the help of a yukawa model with an elementary boson .
|
a three - level laser is a quantum optical system in which light is generated by three - level atoms inside a cavity usually coupled to a vacuum reservoir . in one model of a such laser ,
three - level atoms initially prepared in a coherent superposition of the top and bottom levels are injected into a cavity and then removed after they have decayed due to spontaneous emission [ 1,2 ] . in another model ,
the top and bottom levels of the three - level atoms injected into a cavity are coupled by coherent light [ 3,4 ] .
the statistical and squeezing properties of the light generated by three - level lasers have been investigated by several authors [ 5 - 11 ] .
it is found that a three - level laser in either model generates squeezed light under certain conditions .
it appears to be quite difficult to prepare the atoms in a coherent superposition of the top and bottom levels before they are injected into the laser cavity .
in addition , it should certainly be hard to find out that the atoms have decayed spontaneously before they are removed from the cavity .
on the other hand , the degree of squeezing of the light generated by the three - level laser , with the top and bottom levels coupled by coherent light , is relatively large when the mean photon number is relatively small [ 4 ] .
moreover , the quantum analysis of a three - level laser is usually carried out by including the interaction of the atoms inside the cavity with the vacuum reservoir outside the cavity
. it may be possible to justify the feasibility of such interaction for a laser with an open cavity into which and from which atoms are injected and removed
. however , there can not be any valid justification for leaving open the laser cavity in which the available atoms are pumped to the top level by electron bombardment .
therefore , the aforementioned interaction is not feasible for a laser in which the atoms available in a closed cavity are pumped to the top level by means of electron bombardment .
we seek here to analyze the quantum properties of the light emitted by the three - level atoms available in a closed cavity and pumped to the top level at a constant rate . thus taking into account the interaction of the three - level atoms with a resonant cavity light and the damping of the cavity light by a vacuum reservoir
, we obtain the photon statistics of the cavity light and the quadrature squeezing of the cavity ( output ) light .
we also determine the quadrature squeezing of the cavity ( output ) light in any frequency interval .
we carry out our calculation by putting the noise operators associated with the vacuum reservoir in normal order and without considering the interaction of the three - level atoms with the vacuum reservoir outside the cavity .
we consider here the case in which @xmath1 degenerate three - level atoms in cascade configuration are available in a closed cavity .
we denote the top , middle , and bottom levels of these atoms by @xmath2 , @xmath3 , and @xmath4 , respectively .
in addition , we assume the cavity mode to be at resonance with the two transitions @xmath5 and @xmath6 , with direct transition between levels @xmath2 and @xmath4 to be dipole forbidden .
the interaction of one of the three - level atoms with the cavity mode can be described at resonance by the hamiltonian @xmath7,\ ] ] where @xmath8 and @xmath9 are lowering atomic operators , @xmath10 is the annihilation operator for the cavity mode , and @xmath11 is the coupling constant between the atom and the cavity mode .
we assume that the laser cavity is coupled to a vacuum reservoir via a single - port mirror .
in addition , we carry out our calculation by putting the noise operators associated with the vacuum reservoir in normal order .
thus the noise operators will not have any effect on the dynamics of the cavity mode operators .
we can therefore drop the noise operator and write the quantum langevin equation for the operator @xmath10 as @xmath12,\ ] ] where @xmath13 is the cavity damping constant . then with the aid of eq .
( [ 1 ] ) , we easily find @xmath14 furthermore , applying the relation @xmath15\rangle\ ] ] along with eq .
( [ 1 ] ) , one can readily establish that @xmath16 @xmath17 @xmath18 @xmath19 @xmath20 where @xmath21 @xmath22 @xmath23 @xmath24 we see that eqs .
( [ 7])-([12 ] ) are nonlinear differential equations and hence it is not possible to find exact time - dependent solutions of these equations .
we intend to overcome this problem by applying the large - time approximation [ 12 ] . then using this approximation scheme
, we get from eq .
( [ 5 ] ) the approximately valid relation @xmath25 evidently , this turns out to be an exact relation at steady state . now introducing eq .
( [ 17 ] ) into eqs .
( [ 7])-([12 ] ) , we get @xmath26 @xmath27 @xmath28 @xmath29 @xmath30 where @xmath31 we prefer to call the parameter defined by eq .
( [ 24 ] ) the stimulated emission decay constant .
the three - level atoms available in the cavity are pumped from the bottom to the top level by means of electron bombardment .
the pumping process must surely affect the dynamics of @xmath32 and @xmath33 .
if @xmath34 represents the rate at which a single atom is pumped from the bottom to the top level , then @xmath32 increases at the rate of @xmath35 and @xmath33 decreases at the same rate . in view of this , we rewrite eqs .
( [ 20 ] ) and ( [ 23 ] ) as @xmath36 and @xmath37 we next sum eqs .
( [ 18 ] ) , ( [ 19 ] ) , ( [ 22 ] ) , ( [ 25 ] ) , and ( [ 26 ] ) over the @xmath1 three - level atoms , so that @xmath38 @xmath39 @xmath40 @xmath41 @xmath42 in which @xmath43 @xmath44 @xmath45 @xmath46 @xmath47 with the operators @xmath48 , @xmath49 , and @xmath50 representing the number of atoms in the top , middle , and bottom levels .
in addition , employing the completeness relation @xmath51 we easily arrive at @xmath52 furthermore , using the definition given by ( [ 42 ] ) and setting for any @xmath53 @xmath54 we have @xmath55 following the same procedure , one can also check that @xmath56 @xmath57 @xmath58 @xmath59 @xmath60 where @xmath61 moreover , using the definition @xmath62 and taking into account eqs .
( [ 51])-([56 ] ) , it can be readily established that @xmath63 @xmath64 @xmath65 in the presence of @xmath1 three - level atoms , we rewrite eq .
( [ 5 ] ) as @xmath66 in which @xmath67 is a constant whose value remains to be fixed . applying the steady - state solution of eq .
( [ 5 ] ) , we get @xmath68_{k}={\gamma_{c}\over\kappa}(\hat{\eta}_{c}^{k}-\hat{\eta}_{a}^{k})\ ] ] and on summing over all atoms , we have @xmath69={\gamma_{c}\over\kappa}(\hat{n}_{c}-\hat{n}_{a}),\ ] ] where @xmath70=\sum_{k=1}^{n}[\hat{b},\hat{b}^{\dagger}]_{k}\ ] ] stands for the commutator of @xmath10 and @xmath71 when the cavity mode is interacting with all the @xmath1 three - level atoms . on the other hand , using the steady - state solution of eq . ( [ 61 ] ) , one can easily verify that @xmath72=n\bigg({2\lambda\over\kappa}\bigg)^{2}(\hat{n}_{c}-\hat{n}_{a}).\ ] ] thus on account of eqs .
( [ 63 ] ) and ( [ 65 ] ) , we see that @xmath73 and in view of this result , eq .
( [ 61 ] ) can be written as @xmath74
we next seek to calculate the mean and variance of the photon number at steady state .
however , we need first to establish several important relations . hence with the aid of ( [ 49 ] )
, one can put eq .
( [ 39 ] ) in the form @xmath75 and application of the large - time approximation scheme to eq .
( [ 40 ] ) yields @xmath76 thus on taking into account this result , we find the steady - state solution of eq .
( [ 68 ] ) to be @xmath77 we also notice that at steady state eq .
( [ 39 ] ) leads to @xmath78 we now proceed to calculate the expectation value of the atomic operator @xmath79 .
we assume that the state vector of a three - level atom , put in the form @xmath80 @xmath81 and @xmath82 in order to have a mathematically manageable analysis , we take @xmath83 to be real , so that @xmath84 and on subtracting @xmath85 from both sides of this equation , we have @xmath86 hence for all possible values of @xmath87 and @xmath88 , we see that @xmath89 and @xmath90 now on account of ( [ 71b ] ) and ( [ 71c ] ) together with ( [ 71 g ] ) and ( [ 71h ] ) , eq . ( [ 71d ] ) takes the form @xmath91 finally on summing over @xmath53 from 1 up to @xmath1 and taking into account eqs .
( [ 45 ] ) and ( [ 47 ] ) , we have @xmath92 furthermore , adding eqs .
( [ 36 ] ) and ( [ 37 ] ) , we have @xmath93 in order to include the effect of the pumping process , we rewrite this equation as @xmath94 in which @xmath95 is a noise operator with vanishing mean and @xmath96 is a parameter whose value remains to be determined .
employing the relation @xmath97 along with eq .
( [ 73 ] ) , we easily find @xmath98 from which follows @xmath99 on the other hand , on account of eqs .
( [ 39 ] ) and ( [ 40 ] ) , we see that @xmath100 this can be rewritten as @xmath101 hence comparison of eqs .
( [ 76 ] ) and ( [ 78 ] ) shows that @xmath102 and @xmath103 we observe that eq .
( [ 80 ] ) is equivalent to @xmath104 upon casting eq .
( [ 36 ] ) into the form @xmath105 one can also easily verify that @xmath96 has the value given by ( [ 79 ] ) and @xmath106 with the atoms considered to be initially in the bottom level , the expectation value of the solution of eq .
( [ 82 ] ) happens to be @xmath107 then on account of this result , the expectation value of the solution of eq .
( [ 73 ] ) turns out to be @xmath108 on the other hand , the expectation value of the solution of eq .
( [ 67 ] ) is expressible as @xmath109 now in view of eq .
( [ 85 ] ) and the assumption that the cavity light is initially in a vacuum state , eq .
( [ 86 ] ) goes over into @xmath110 we observe on the basis of eqs .
( [ 67 ] ) and ( [ 87 ] ) that @xmath10 is a gaussian variable with zero mean .
the steady - state solution of eq .
( [ 67 ] ) is given by @xmath111 now using eq .
( [ 88 ] ) along with ( [ 58 ] ) , the mean photon number of the cavity light is expressible as @xmath112 it proves to be convenient to refer to the regime of laser operation with more atoms in the top level than in the bottom level as above threshold , the regime of laser operation with equal number of atoms in the top and bottom levels as threshold , and the regime of laser operation with less atoms in the top level than in the bottom level as below threshold . thus according to eq .
( [ 71 ] ) for the laser operating above threshold @xmath113 , for the laser operating at threshold @xmath114 , and for the laser operating below threshold @xmath115 .
we note that for the laser operating well above threshold , eq .
( [ 89 ] ) reduces to @xmath116 and for the laser operating at threshold , we find @xmath117 in which @xmath118 .
the variance of the photon number for the cavity light is expressible as @xmath119 and using the fact that @xmath10 is a gaussian variable with zero mean , we readily get @xmath120 employing once more eq .
( [ 88 ] ) and taking into account ( [ 59 ] ) , we find @xmath121 in addition , taking into account ( [ 71 ] ) , one can express eq .
( [ 71i ] ) in the form @xmath122 hence with the aid of eqs .
( [ 88 ] ) , ( [ 60 ] ) , and ( [ 95 ] ) , we easily get @xmath123 now on account of eqs .
( [ 93 ] ) , ( [ 94 ] ) , and ( [ 96 ] ) along with ( [ 69 ] ) and ( [ 71 ] ) , we arrive at @xmath124 with @xmath125 .
we then see that for the laser operating well above threshold @xmath126 where @xmath127 is given by eq .
( [ 90 ] ) . and
for the laser operating at threshold , we get @xmath128 in which @xmath127 given by eq .
( [ 91 ] ) .
in this section we wish to calculate the quadrature squeezing of the cavity and output light in any frequency interval .
the squeezing properties of the cavity light are described by two quadrature operators defined by @xmath129 and @xmath130 it can be readily established that @xmath131=2i{\gamma_{c}\over\kappa}(\hat{n}_{a}-\hat{n}_{c}).\ ] ] it then follows that @xmath132 the variance of the quadrature operators is expressible as @xmath133^{2}\rangle { \mp}[\langle\hat{b}^{\dagger}\rangle{\pm}\langle\hat{b}\rangle]^{2},\ ] ] so that on account of ( [ 87 ] ) , we have @xmath134 now employing ( [ 89 ] ) , ( [ 94 ] ) , and ( [ 96 ] ) , we arrive at @xmath135 and @xmath136 we recall that the light generated by a two - level laser operating well above threshold is coherent , the quadrature variance of which is given by [ 12 ] @xmath137 we calculate the quadrature squeezing of the cavity ( output ) light relative to the quadrature variance of the cavity ( output ) coherent light .
we then define the quadrature squeezing of the cavity light by @xmath138 hence employing ( [ 108 ] ) , ( [ 109 ] ) , and ( [ 70 ] ) , one can put eq .
( [ 110 ] ) in the form @xmath139 a plot of eq .
( [ 111 ] ) versus @xmath140 .
we observe that , unlike the mean photon number , the quadrature squeezing does not depend on the number of atoms .
this implies that the quadrature squeezing of the cavity light is independent of the number of photons . the plot in fig .
1 shows that the maximum squeezing of the cavity light is 50 @xmath141 below the coherent - state level and this occurs when the three - level laser is operating below threshold at @xmath142 . on the other hand
, we define the quadrature squeezing of the output light by @xmath143 where @xmath144 is the quadrature variance of the output coherent light . since all calculations in this analysis are carried out by putting the vacuum noise operators in normal order , one can write @xmath145 we then easily see that @xmath146 and @xmath147 now in view of eqs .
( [ 112 ] ) , ( [ 114 ] ) , ( [ 115 ] ) , and ( [ 110 ] ) , we arrive at @xmath148 we observe that the quadrature squeezing of the output light is equal to that of the cavity light .
we finally seek to obtain the quadrature squeezing of the cavity ( output ) light in a given frequency interval . to this end
, we first determine the spectrum of quadrature fluctuations of the cavity light .
we define this spectrum for a single - mode light with central frequency @xmath149 by @xmath150 upon integrating both sides of eq .
( [ 117 ] ) over @xmath151 , we get @xmath152 in which @xmath153 is the quadrature variance of the light mode at steady state . on the basis of the result given by eq .
( [ 118 ] ) , we assert that @xmath154 is the steady - state quadrature variance of the light mode in the interval between @xmath151 and @xmath155 . in view of eq .
( [ 87 ] ) , we note that @xmath156 we now proceed to determine the two - time correlation function that appears in eq .
( [ 120 ] ) for the cavity light . to this end
, we realize that the solution of eq .
( [ 67 ] ) can also be written as @xmath157 on the other hand , the solution of eq .
( [ 73 ] ) is expressible as @xmath158 applying the large - time approximation scheme to eq .
( [ 82 ] ) , we get @xmath159 so that on introducing this into eq .
( [ 122 ] ) , there follows @xmath160 now combination of eqs .
( [ 121 ] ) and ( [ 124 ] ) yields @xmath161 + { g\over{\sqrt{n}}}e^{-\kappa\tau/2}\int_{0}^{\tau}d\tau'e^{(\kappa-\mu)\tau'/2}\nonumber\\ & & \times\int_{0}^{\tau'}d\tau''e^{\mu\tau''/2}\bigg(\hat{f}_{a}(t+\tau '' ) + \hat{f}_{m}(t+\tau'')\bigg).\end{aligned}\ ] ] hence using this result , we readily arrive at @xmath162.\end{aligned}\ ] ] applying once more the large - time approximation , one gets from eq .
( [ 67 ] ) @xmath163 with this substituted into eq .
( [ 126 ] ) , there emerges @xmath164 following a similar procedure , one can also readily establish that @xmath165 and @xmath166 therefore , on account of eqs .
( [ 128 ] ) , ( [ 129 ] ) , and ( [ 130 ] ) , there follows @xmath167 now on introducing eq .
( [ 131 ] ) into eq . (
[ 117 ] ) and carrying out the integration , we find the spectrum of the minus quadrature fluctuations for the cavity light to be @xmath168^{2}}}\bigg ) -{\mu\over{\kappa-\mu}}\bigg({\kappa/2\pi\over{(\omega-\omega_{0})^{2}+[\kappa/2]^{2}}}\bigg)\bigg],\end{aligned}\ ] ] in which @xmath169 is given by ( [ 108 ] ) .
we realize that the variance of the minus quadrature in the interval between @xmath170 and @xmath171 is expressible as @xmath172 in which @xmath173 hence taking into account eq .
( [ 132 ] ) , we readily get @xmath174 where @xmath175 is given by @xmath176 on account of the result described by ( [ 134 ] ) , the quadrature variance of the output coherent light can be written in the same frequency interval as @xmath177 we define the quadrature squeezing of the cavity light in the @xmath178 frequency interval by @xmath179 so that on account of ( [ 134 ] ) , ( [ 136 ] ) , and ( [ 110 ] ) , there follows @xmath180 this shows that the quadrature squeezing of the cavity light in a given frequency interval is equal to that of the cavity light in the entire frequency interval . finally , defining the quadrature squeezing of the output light in the aforementioned frequency interval by @xmath181 and taking into account the fact that @xmath182 and @xmath183 we arrive at @xmath184 we see that the quadrature squeezing of the output light in a certain frequency interval is the same as that of the output light in the entire frequency interval .
we have considered a three - level laser in which the three - level atoms available in a closed cavity are pumped from the bottom to the top level by means of electron bombardment .
we have carried out our analysis by putting the vacuum noise operators in normal order and applying the large - time approximation scheme .
the procedure of normal ordering the noise operators renders the vacuum reservoir to be a noiseless physical entity .
we maintain the standpoint that the notion of a noiseless vacuum reservoir would turn out to be compatible with observation .
based on the definition given by eq .
( [ 24 ] ) , we infer that an atom in the top or middle level and inside a closed cavity emits a photon due to its interaction with the cavity light .
we certainly identify this process to be stimulated emission .
we have seen that the light generated by the three - level laser operating under the condition @xmath185 is in a squeezed state , with the maximum quadrature squeezing being @xmath0 below the coherent - state level .
this occurs when the three - level laser is operating below threshold at @xmath142 .
we have also established that the quadrature squeezing of the laser light has the same value in any frequency interval . on the basis of this result , we come to the conclusion that the quadrature squeezing of the laser light is an intrinsic property of the individual photons . [ 1 ] m.o .
scully , k. wodkiewicz , m.s .
zubairy , j. bergou , n. lu , j. meyer ter vehn , phys .
rev . lett .
60 , 1832 ( 1988).fesseha kassahun , fundamentals of quantum optics ( lulu press inc .
, north carolina , 2010).n .
zhao , j. bergou , phys .
a 39 , 5189 ( 1989).eyob alebachew and k. fesseha , opt .
265 , 314 ( 2006).m.o .
scully and m.s .
zubairy , opt .
66 , 303 ( 1988).n . lu and s.y .
zhu , phys .
rev . a 41 , 2865 ( 1990).n.a .
ansari , j. gea - banacloche , and m.s .
zubairy , phys .
a 41 , 5179 ( 1990).c.a .
blockely and d.f .
walls , phys .
a 43 , 5049 ( 1991).n.a .
ansari , phys .
a 48 , 4686 ( 1993).j .
anwar and m.s .
zubairy , phys . rev . a 49 , 481 ( 1994).k .
fesseha , phys .
a 63 , 33811 ( 2001).fesseha kassahun , opt . commun .
284 , 1357 ( 2011 ) .
|
we analyze the quantum properties of the light generated by a three - level laser with a closed cavity and coupled to a vacuum reservoir .
the three - level atoms available in the cavity are pumped from the bottom to the top level by means of electron bombardment and we carry out our analysis by putting the noise operators associated with the vacuum reservoir in normal order .
the maximum quadrature squeezing of the light generated by the laser , operating far below threshold , is found to be @xmath0 below the coherent - state level .
we have also established that the quadrature squeezing of the output light is equal to that of the cavity light and has the same value in any frequency interval .
this implies that the quadrature squeezing of the laser light is an intrinsic property of the individual photons .
keywords : stimulated emission , photon statistics , quadrature squeezing , noiselessvacuum reservoir
|
the evaluation of the transport coefficients from the boltzmann equation for inelastic hard spheres ( ihs ) is quite involved .
in fact , to get explicit results one usually considers the leading order in a sonine polynomial expansion of the velocity distribution function .
these difficulties increase when one considers multicomponent systems since not only the number of transport coefficients is larger than for a single gas but they are also functions of more parameters such as composition , masses , sizes , and the coefficients of restitution . as in the elastic case , a possible way to partially overcome these problems
is to consider other interaction models that simplify the complex mathematical structure of the boltzmann collision integrals for ihs .
for this reason , the so - called inelastic maxwell models ( imm ) have been widely used in the past few years as a toy model to characterize the influence of the inelasticity of collisions on the physical properties of granular fluids . the imm share with elastic maxwell molecules the property that the collision rate is velocity independent but , on the other hand , their collision rules are the same as for ihs . in this sense , although these imm do not correspond to any microscopic potential interaction , it has been shown by several authors @xcite that the cost of sacrificing physical realism is in part compensated by the amount of exact analytical results .
most of the studies carried out by considering imm have been devoted to homogeneous states , especially in the analysis of the overpopulated high energy tails .
@xcite however , much less is known in the case of _ inhomogeneous _ situations and , more specifically , on the dependence of the transport coefficients on dissipation . for a monocomponent granular gas subjected to simple shear flow , the imm
have been used to calculate the rheological properties ( shear and normal stresses ) in three dimensions .
@xcite more recently , this study has been extended @xcite to multicomponent systems , the exact results of imm showing a close agreement with those obtained analytically for ihs in the first sonine approximation @xcite and by means of monte carlo simulations .
@xcite all these results are restricted to steady shear flow problems without any limitation on the strength of the shear rate . for general inhomogeneous problems and in the case of a monocomponent gas , the boltzmann equation for imm has been solved @xcite from the chapman - enskog method @xcite for states near the ( local ) homogeneous cooling state .
explicit expressions of the navier - stokes transport coefficients of imm in @xmath0 dimensions have been obtained for unforced systems as well as for systems driven by thermostats .
in contrast to the findings of ref . , the comparison with the transport coefficients of ihs @xcite shows that their dependence on inelasticity is captured by the imm only in a mild qualitative way .
this fact stimulates the determination of the exact expressions of the transport coefficients for inelastic granular mixtures .
the goal of this paper is to derive the hydrodynamic equations for a @xmath0-dimensional binary mixture of inelastic maxwell gases at low - density . as in the single gas case ,
normal _ solution to the coupled set of boltzmann equations for the two species is obtained by using the standard chapman - enskog method @xcite conveniently generalized to inelastic collisions . in the first order of the spatial gradients of the hydrodynamic fields , we get the corresponding navier - stokes hydrodynamic equations with explicit expressions for the relevant transport coefficients of the mixture . for molecular fluid mixtures , the specific set of gradients contributing to each flux
is restricted by fluid symmetry , time reversal invariance ( onsager relations ) , and the form of entropy production .
@xcite however , in the case of inelastic collisions , only fluid symmetry holds and so there is more flexibility in representing the fluxes and identifying the corresponding transport coefficients . in the case of the pressure tensor ,
fluid symmetry considerations implies that its form to first order in the gradients is the same as for the monocomponent gas . in the case of mass and heat fluxes , several different ( but equivalent ) choices of hydrodynamic fields
can be used and some care must be taken when comparing transport coefficients coming from different representations . as in the case of ihs,@xcite
we take the gradients of the mole fraction , the ( hydrostatic ) pressure , the temperature , and the flow velocity as the relevant ones . as a consequence , in this representation
there are seven independent scalar transport coefficients : the mutual diffusion , the pressure diffusion and the thermal diffusion associated with the mass flux , the shear viscosity corresponding to the pressure tensor and the dufour coefficient , the thermal conductivity , and the pressure energy coefficient associated with the heat flux .
all these coefficients are given in terms of the coefficients of restitution as well as on the ratios of concentration , masses and particle sizes .
in addition , as in the previous study for ihs , @xcite our theory takes into account the effect of temperature differences ( failure of energy equipartition ) on the transport coefficients , leading to additional dependencies of them on the concentration . the plan of the paper is as follows . in section [ sec2 ]
the boltzmann equation for imm and the macroscopic conservation laws are introduced .
the model includes average collision frequencies @xmath1 which can be freely fitted to get good agreement with ihs .
here , we fix @xmath1 by the criterion that the cooling rates @xmath2 of imm be the same as those obtained for ihs in the local equilibrium approximation . the homogenous solution of the boltzmann equation is analyzed in section [ sec3 ] , where the temperature ratio and the fourth cumulant ( kurtosis ) of the velocity distribution functions are exactly obtained .
comparison with the results obtained for ihs @xcite shows an excellent agreement for the temperature ratio but significant discrepancies with the fourth cumulant of ihs .
section [ sec4 ] deals with the application of the chapman - enskog method to get the transport coefficients of imm . in section [ sec5 ] ,
the dependence of some of these coefficients on the parameters of the system is illustrated and compared with known results derived for ihs .
@xcite the comparison shows in general a qualitative good agreement , especially for the transport coefficients defining the mass flux .
the paper ends in section [ sec6 ] with a brief discussion on the results reported in this paper .
let us consider a binary mixture of inelastic maxwell gases at low density .
the boltzmann equation for imm @xcite can be obtained from the boltzmann equation for ihs by replacing the rate for collisions between particles of species @xmath3 and @xmath4 by an average velocity - independent collision rate , which is proportional to the square root of the `` granular '' temperature @xmath5 .
this means that a random pair of colliding particles undergoes inelastic collisions with a random impact direction . with this simplification ,
the velocity distribution functions @xmath6 @xmath7 of each species satisfy the following set of nonlinear boltzmann kinetic equations : @xmath8 \;,\ ] ] where the boltzmann collision operator @xmath9 $ ] is @xmath10 & = & \frac{\omega_{rs}({\bf r},t;\alpha_{rs})}{n_s({\bf r},t)\omega_d } \int d{\bf v}_{2}\int d\widehat{\boldsymbol{\sigma } } \nonumber\\ & & \times \left [ \alpha_{rs}^{-1}f_{r}({\bf r},{\bf v}_{1}',t)f_{s } ( { \bf r},{\bf v}_{2}',t)-f_{r}({\bf r},{\bf v}_{1},t)f_{s}({\bf r},{\bf v}_{2},t)\right ] \;.
\label{2.2}\end{aligned}\ ] ] here @xmath11 is the number density of species @xmath3 , @xmath12 is an effective collision frequency ( to be chosen later ) for collisions of type @xmath3-@xmath4 , @xmath13 is the total solid angle in @xmath0 dimensions , and @xmath14 refers to the constant coefficient of restitution for collisions between particles of species @xmath3 with @xmath4 .
in addition , the primes on the velocities denote the initial values @xmath15 that lead to @xmath16 following a binary collision : @xmath17 where @xmath18 is the relative velocity of the colliding pair , @xmath19 is a unit vector directed along the centers of the two colliding spheres , and @xmath20 . the collision frequencies @xmath1 can be seen as free parameters in the model .
its dependence on the restitution coefficients @xmath21 can be chosen to optimize the agreement with the results obtained from the boltzmann equation for ihs .
of course , the choice is not unique and may depend on the property of interest .
there is another more refined version of the inelastic maxwell model @xcite where the collision rate has the same dependence on the scalar product ( @xmath22 ) as in the case of hard spheres .
however , both versions of imm lead to similar results in problems as delicate as the high energy tails.@xcite therefore , for the sake of simplicity , here we will consider the version given by eqs .
( [ 2.2 ] ) and ( [ 2.3 ] ) .
the relevant hydrodynamic fields in a binary mixture are the number densities @xmath11 , the flow velocity @xmath23 , and the granular temperature @xmath5 .
they are defined in terms of the distribution @xmath24 as @xmath25 @xmath26 @xmath27 where @xmath28 is the mass density of species @xmath3 , @xmath29 is the total number density , @xmath30 is the total mass density , @xmath31 is the peculiar velocity , and @xmath32 is the hydrostatic pressure . furthermore , the third equality of eq .
( [ 2.5 ] ) defines the kinetic temperatures @xmath33 of each species , which measure their mean kinetic energies .
the collision operators conserve the particle number of each species and the total momentum , but the total energy is not conserved : @xmath34=0,\ ] ] @xmath35=0,\ ] ] @xmath36=-\case{d}{2}nt\zeta \;,\ ] ] where @xmath37 is identified as the `` cooling rate '' due to inelastic collisions among all species . at a kinetic level , it is also convenient to introduce the `` cooling rates '' @xmath38 for the partial temperatures @xmath33 .
they are defined as @xmath39,\ ] ] where the second equality defines the quantities @xmath2 .
the total cooling rate @xmath37 can be written in terms of the partial cooling rates @xmath38 as @xmath40 where @xmath41 is the mole fraction of species @xmath3 . from eqs .
( [ 2.2.1 ] ) to ( [ 2.6 ] ) , the macroscopic balance equations for the binary mixture can be obtained .
they are given by @xmath42 @xmath43 @xmath44 in the above equations , @xmath45 is the material derivative , @xmath46 is the mass flux for species @xmath3 relative to the local flow , @xmath47 is the total pressure tensor , and @xmath48 is the total heat flux .
the balance equations ( [ 2.9])([2.11 ] ) apply regardless of the details of the model for inelastic collisions considered .
however , the influence of the collision model appears through the dependence of the cooling rate and the hydrodynamic fluxes on the coefficients of restitution . as happens for elastic collisions,@xcite
the main advantage of using imm is that a velocity moment of order @xmath49 of the boltzmann collision operator only involves moments of order less than or equal to @xmath49 .
@xcite this allows one to determine the boltzmann collisional moments without the explicit knowledge of the velocity distribution function .
the first few moments of the boltzmann collision operator @xmath50 $ ] have been explicitly evaluated in appendix [ appa ] . in particular , according to eq .
( [ 2.7 ] ) , the second moment of @xmath50 $ ] allows one to get a relationship between the collision frequencies @xmath1 and the cooling rates @xmath2 . from eq .
( [ a2 ] ) , one easily gets @xmath51,\ ] ] where @xmath52 @xmath53 is the partial pressure of species @xmath3 and @xmath54 . in order to get explicit results , one still needs to fix the parameters @xmath1 .
the most natural choice to optimize the agreement with the ihs results is to adjust the cooling rates @xmath2 for imm , eq . (
[ 2.15 ] ) , to be the same as the ones found for ihs .
@xcite although the cooling rates are not exactly known for ihs , a good estimate of them can be obtained by considering the local equilibrium approximation for the velocity distribution functions @xmath24 , i.e. , @xmath55 in this approximation , one has @xcite
@xmath56,\ ] ] where @xmath57 is a thermal velocity defined in terms of the temperature @xmath58 of the mixture .
thus , according to eq .
( [ 2.15 ] ) , the collision frequencies @xmath1 are given by @xmath59 where @xmath60 is an effective collision frequency given by @xmath61 upon deriving ( [ 2.32 ] ) use has been made of the fact that the mass flux @xmath62 vanishes in the local equilibrium approximation ( [ 2.16bis ] ) . in the remainder of this paper , we will take the choice ( [ 2.32 ] ) for @xmath1 .
as a previous step to determine the navier - stokes transport coefficients from the chapman - enskog method , @xcite one needs to analyze the _ homogeneous _ solution of the boltzmann equation ( [ 2.1 ] ) . in this case
( spatially isotropic homogeneous states ) , the boltzmann equation ( [ 2.1 ] ) becomes @xmath63.\ ] ] from eq .
( [ 2.16.1 ] ) , one has @xmath64 the time evolution of the temperature ratio @xmath65 follows from the second equality of eq .
( [ 2.17 ] ) : @xmath66 in addition , since the mass flux vanishes in the homogeneous state , eq .
( [ 2.15 ] ) gives the relation @xmath67,\ ] ] where @xmath1 is given by eq .
( [ 2.32 ] ) .
the so - called homogeneous cooling state ( hcs ) qualifies as a _
solution for which all the time dependence of @xmath68 is through the global temperature @xmath58 .
consequently , it follows from dimensional analysis that @xmath68 has the scaling form @xmath69 where @xmath70 is the thermal velocity previously introduced .
the fact that the time dependence of @xmath68 only occurs through the temperature @xmath58 ( which is the relevant one at a hydrodynamic level ) implies that the three temperatures @xmath71 , @xmath72 and @xmath58 are proportional to each other and their ratios are all constant .
one possibility is that all three temperatures are equal , as in the case of elastic collisions .
however , as we will see later , the partial temperatures are different . since the temperature ratio must be independent of time , eq .
( [ 2.19 ] ) leads to the equality of the cooling rates : @xmath73 when the expression ( [ 2.18bis ] ) is substituted into ( [ 2.21 ] ) , one gets a closed nonlinear equation for @xmath74 whose numerical solution gives the dependence of the temperature ratio on the parameters of the problem .
except for mechanically equivalent particles , our results show that although both species have a common cooling rate , their partial temperatures are clearly different .
this implies a breakdown of the energy equipartition .
the violation of energy equipartition in multicomponent granular systems @xcite has also been confirmed in computer simulations @xcite and even observed in real experiments in two @xcite and three @xcite dimensions .
it must be remarked that the fact that @xmath75 does not mean that there are additional hydrodynamic degrees of freedom since the partial temperatures @xmath33 can be expressed in terms of the granular temperature @xmath5 as @xmath76 the problem is therefore to solve the boltzmann equation for a distribution of the form ( [ 2.20 ] ) subject to the self - consistency constraint ( [ 2.21 ] ) . in terms of the reduced velocity @xmath77 , the boltzmann equation ( [ 2.16.1 ] ) for the reduced distribution @xmath78 defined in eq .
( [ 2.20 ] ) becomes @xmath79,\ ] ] with @xmath80 .
in addition , @xmath81=\frac{\omega_{rs}^*}{\omega_d } \int d{\bf v}_{2}^*\int d\widehat{\boldsymbol{\sigma } } \left [ \alpha_{rs}^{-1}\phi_{r } ( v_{1}'^*)\phi_{s } ( v_{2}'^*)-\phi_{r}(v_{1}^*)\phi_{s } ( v_{2}^*)\right],\ ] ] where @xmath82 . upon writing eq .
( [ 2.22 ] ) , we have accounted for the time dependence of @xmath24 , which implies that @xmath83 although the exact form of the distribution @xmath78 is not known , an _ indirect _
information on the behavior of @xmath78 is given through its velocity moments .
in particular , the deviation of @xmath78 with respect to its maxwellian form @xmath84 can be characterized through the fourth cumulant @xmath85,\ ] ] where @xmath86 and @xmath87 to get the fourth velocity moment , we multiply both sides of the boltzmann equation ( [ 2.22 ] ) by @xmath88 and integrate over the velocity .
the result is @xmath89.\ ] ] the collisional moment appearing on the right - hand side of ( [ 2.28 ] ) is evaluated in appendix [ appa ] with the result @xmath90&= & \frac{\mu_{sr}(1+\alpha_{rs})}{d(d+2)}\omega_{rs}^*\left\{3\mu_{sr}^3(1+\alpha_{rs})^3\langle v^{*4}\rangle_{s}\right.\nonumber\\ & & + \left[2d+3\mu_{sr}^2(1+\alpha_{rs})^2 - 6\mu_{sr}(1+\alpha_{rs})+4\right]\left[\mu_{sr}(1+\alpha_{rs})-2\right ] \langle v^{*4}\rangle_{r } \nonumber\\ & & + \frac{d(d+2)}{4}\mu_{sr}\theta_r^{-1}\theta_s^{-1 } ( 1+\alpha_{rs})\left[2d+4 - 12\mu_{sr}(1+\alpha_{rs})\right.\nonumber\\ & & \left.\left .
+ 6\mu_{sr}^2(1+\alpha_{rs})^2\right]\right\}.\end{aligned}\ ] ] substitution of eq .
( [ 2.29 ] ) into eq . ( [ 2.28 ] ) leads to a coupled set of linear equations for @xmath91 and @xmath92 ( or equivalently , for @xmath93 and @xmath94 ) .
the solution of this set gives @xmath93 and @xmath94 in terms of the parameter space of the problem .
in the one - dimensional case ( @xmath95 ) , our results reduce to the ones previously obtained by marconi and puglisi@xcite from the so - called maxwell scalar model ( i.e. , by taking @xmath96 ) .
moreover , for mechanically equivalent particles ( @xmath97 , @xmath98 , @xmath99 ) , the results obtained by santos @xcite for the single gas case are recovered , namely , @xmath100 and @xmath101 a full presentation of the results is difficult due to the many parameters involved in the problem : @xmath102 , @xmath103 , @xmath104 , @xmath105 , @xmath106 , and @xmath107 .
for the sake of concreteness , henceforth we will consider the case @xmath108 .
in fig .
[ fig2 ] we show the dependence of @xmath74 on @xmath109 in the three - dimensional case for @xmath110 , @xmath111 and three different values of the mass ratio @xmath105 .
we include the analytical results obtained for imm and ihs@xcite as well as the results ( symbols ) obtained @xcite by numerically solving the boltzmann equation by means of the direct simulation monte carlo ( dsmc ) method .
@xcite it is apparent that the analytical results for imm and ihs are practically indistinguishable over the range of values of @xmath109 considered and that the agreement of both approaches with simulation is excellent .
we also observe that the extent of the equipartition violation is greater when the mass disparity is large .
in addition , the temperature of the excess species is larger ( smaller ) than that of the defect species when the excess species is heavier ( lighter ) than the defect species .
the dependence of the coefficients @xmath93 and @xmath94 on the coefficient of restitution @xmath109 is plotted in fig .
[ fig2 ] in the three - dimensional case for @xmath112 , @xmath111 and @xmath113 .
it can be observed that the hcs of imm deviates from the gaussian distribution ( which corresponds to @xmath114 ) much more than the hcs of ihs .
this is consistent with the fact that the former models have a stronger overpopulated high energy tail @xcite than the latter .
@xcite for both interaction models , the deviation of @xmath78 from its gaussian form is more significant for the heavy species .
we also see that the value of the kurtosis @xmath115 predicted by the imm exhibits quantitative discrepancies with the one found for ihs , especially for strong dissipation .
in this section , the chapman - enskog method @xcite generalized to inelastic collisions will be applied to the set of boltzmann equations ( [ 2.1 ] ) for imm to get explicit expressions for the navier - stokes transport coefficients as functions of the coefficients of restitution and the parameters of the mixture ( masses , composition , and sizes ) .
the balance equations ( [ 2.9])([2.11 ] ) become a closed set of hydrodynamic equations for the fields @xmath11 , @xmath23 , and @xmath5 once the fluxes ( [ 2.12])([2.14 ] ) and the cooling rate @xmath37 are obtained in terms of the hydrodynamic fields and their gradients . as noted in sec .
[ sec1 ] , while the pressure tensor has the same form as for a one - component system , there is greater freedom in representing the heat and mass fluxes . here , as done in the ihs case ,
@xcite we take the gradients of the mole fraction @xmath116 , the pressure @xmath32 , the temperature @xmath5 , and the flow velocity @xmath23 as the relevant ones .
thus , in this representation , the phenomenological constitutive relations for the fluxes in the low - density regime have the forms @xcite @xmath117 @xmath118 @xmath119 the transport coefficients in these equations are the diffusion coefficient @xmath120 , the thermal diffusion coefficient @xmath121 , the pressure diffusion coefficient @xmath122 , the dufour coefficient @xmath123 , the thermal conductivity @xmath124 , the pressure energy coefficient @xmath125 , and the shear viscosity @xmath126 .
the chapman - enskog method assumes the existence of a _
normal _ solution in which all space and time dependence of the distribution function occurs through a functional dependence on the hydrodynamic fields @xmath127.\ ] ] this functional dependence can be made local in space and time by means of an expansion in gradients of the fields .
thus , we write @xmath128 as a series expansion in a formal parameter @xmath129 measuring the nonuniformity of the system , @xmath130 where each factor of @xmath129 means an implicit gradient of a hydrodynamic field . the local reference state @xmath131 is chosen to give the same first moments as the exact distribution @xmath24 , or equivalently , the remainder of the expansion must obey the orthogonality conditions @xmath132=0,\ ] ] @xmath133=0,\ ] ] @xmath134=0.\ ] ] use of the expansion ( [ 3.5 ] ) in the definitions of the fluxes ( [ 2.12])([2.14 ] ) and the cooling rate ( [ 2.7 ] ) gives the corresponding expansion for these quantities .
the time derivatives of the fields are also expanded as @xmath135 .
the coefficients of the time derivative expansion are identified from the balance equations ( [ 2.9])([2.11 ] ) after expanding the fluxes and the cooling rate @xmath37 . in particular , the macroscopic balance equations to zeroth order become @xmath136 here , we have taken into account that in the boltzmann equation ( [ 2.1 ] ) the effective collision frequency @xmath137 is assumed to be a functional of @xmath24 and @xmath138 only through the mole fraction @xmath139 , the pressure @xmath32 , and the temperature @xmath5 . as a consequence , @xmath140 , and @xmath141 . in the zeroth order
, @xmath131 obeys the kinetic equation @xmath142,\ ] ] where use has been made of the relation ( [ 2.25 ] ) with @xmath143 given by eq .
( [ 2.18bis ] ) .
the distribution @xmath131 is given by the scaling form ( [ 2.20 ] ) except that @xmath144 and @xmath145 are local quantities and @xmath146 .
since @xmath131 is isotropic , it follows that @xmath147 where @xmath148 is the hydrostatic pressure . in the first order , the distribution function @xmath149 verifies the kinetic equation @xmath150 where it is understood that @xmath151 . here ,
@xmath152 and we have introduced the linearized collision operators @xmath153 + j_{rr}[f_r^{(1)},f_r^{(0)}]+j_{rs}[f_r^{(1)},f_s^{(0)}]\right),\ ] ] @xmath154.\ ] ] the action of the material time derivatives @xmath155 on the hydrodynamic fields is @xmath156 in these equations use has been made of the results ( [ 3.9 ] ) and the identity @xmath157 .
the last equality easily follows from eq .
( [ 2.15 ] ) .
the right - hand side of eq .
( [ 3.10 ] ) can be evaluated by using eqs .
( [ 3.13 ] ) and so the equation for @xmath149 can be written as @xmath158 where @xmath159 @xmath160 , \label{3.16}\ ] ] @xmath161 { \bf v } , \label{3.17}\ ] ] @xmath162 it is worth noting that eqs . ( [ 3.14])([3.18 ] )
have the same structure as that of the boltzmann equation for ihs .
@xcite the only difference between both models lies in the explicit form of the linearized operators @xmath163 and @xmath164 .
now we are in conditions to get the expressions for the mass flux , the pressure tensor , and the heat flux in the first order of gradients .
these expressions allows one to identify the relevant transport coefficients of the mixture through eqs .
( [ 3.1])([3.3 ] ) .
this section is devoted to the determination of the navier - stokes transport coefficients associated with the irreversible fluxes .
we only display here the final expressions for the transport coefficients with technical details given in appendix [ appb ] .
to first order in the hydrodynamic gradients , the mass flux has the form given by eq .
( [ 3.1 ] ) .
the transport coefficients @xmath120 , @xmath122 , and @xmath121 are given by @xmath165 , \label{4.8}\ ] ] @xmath166 @xmath167 where @xmath168 since @xmath169 and @xmath170 , @xmath120 should be symmetric while @xmath171 and @xmath121 should be antisymmetric with respect to the exchange @xmath172
. this can be easily verified by noting that @xmath173 . for elastic collisions , @xmath174 , @xmath175 , and so eqs . ( [ 4.8])([4.10 ] )
become @xmath176 @xmath177 where use has been made again of eq .
( [ 2.32 ] ) . in the case of mechanically equivalent particles , @xmath100 , @xmath178 , so that eq . ( [ 4.8 ] ) gives the expression of the self - diffusion coefficient @xmath179 the pressure tensor has the form ( [ 3.3 ] ) with the shear viscosity coefficient @xmath126 given by @xmath180 the expression of the partial contributions @xmath181 is @xmath182 where the quantities @xmath183 and @xmath184 are defined by eqs .
( [ 4.14 ] ) and ( [ 4.15 ] ) , respectively .
a similar expression can be obtained for @xmath185 by just making the changes @xmath186 . for mechanically equivalent particles , eq .
( [ 4.22 ] ) yields @xmath187 , where @xmath188 and @xmath189 .
the expression ( [ 4.22.1 ] ) coincides with the one previously derived in the single gas case .
@xcite the case of the heat flux is more involved .
its form is given by eq .
( [ 3.2 ] ) where the coefficients @xmath123 , @xmath125 and @xmath124 are @xmath190 by using matrix notation , the coupled set of six equations for the unknowns @xmath191 can be written as @xmath192 here , @xmath193 is the column matrix defined by the set ( [ 4bis.1 ] ) and @xmath194 is the square matrix @xmath195 the column matrix @xmath196 is @xmath197 where @xmath198,\ ] ] @xmath199,\ ] ] @xmath200 @xmath201 @xmath202 @xmath203 the expressions of the quantities @xmath204 and @xmath205 are given in appendix [ appb ] .
the solution to eq .
( [ 4.36 ] ) is @xmath206 this relation provides an explicit expression for the coefficients @xmath207 , @xmath208 and @xmath209 in terms of the coefficients of restitution and the parameters of the mixture . from these expressions
one easily gets the transport coefficients @xmath123 , @xmath125 , and @xmath124 from eq .
( [ 4.34 ] ) .
as expected , eqs .
( [ 4.36])([4.42.1 ] ) show that @xmath123 is antisymmetric with respect to the change @xmath172 , while @xmath125 and @xmath124 are symmetric .
this implies that in the case of mechanically equivalent particles , the dufour coefficient @xmath123 vanishes .
furthermore , in this limit , eq . ( [ 4.43 ] ) leads to the following expression for the heat flux @xmath210 where @xmath211 @xmath212 with @xmath213 note that upon writing eq .
( [ 4.43.1 ] ) use has been made of the relation @xmath214 .
again , eqs .
( [ 4.43.1])([4.43.4 ] ) coincide with results derived for a single gas .
@xcite using eqs .
( [ 4.43.2])([4.43.4 ] ) , it can be seen that the coefficients @xmath215 and @xmath216 diverge for @xmath217 and @xmath218 when @xmath219 and so , both coefficients become negative for @xmath220 .
this unphysical behavior could be due to the failure of the hydrodynamic description for values of the coefficient of restitution smaller than @xmath221 or perhaps to the existence of a certain type of hydrodynamic instability .
however , given that the value of @xmath221 is quite small ( @xmath222 at @xmath217 and @xmath223 at @xmath218 ) this singular behavior could be interpreted as a hydrodynamic breakdown indeed .
elucidation of this point requires further analysis .
it must be remarked that the above drawback is absent for ihs since all the transport coefficients are regular functions of @xmath109 for all @xmath0 .
the expressions for the transport coefficients of ihs described by the boltzmann equation have been obtained by garz and dufty @xcite in the leading sonine approximation for a three - dimensional system .
these expressions have been then evaluated for a variety of mass and diameter ratios in the cases of the diffusion coefficient @xcite and the shear viscosity coefficient , @xcite showing quite a good agreement with monte carlo simulations . in this section ,
we compare the results derived here for imm for the transport coefficients entering in the mass and momentum fluxes , namely , @xmath224 , and @xmath126 with those obtained for ihs in the case @xmath218 . as in the cases studied in section [ sec3 ] , we consider for the sake of simplicity a common coefficient of restitution , i.e. , @xmath225 .
let us consider first , the diffusion coefficient @xmath120 in the limit cases of self - diffusion ( mechanically equivalent particles ) and tracer concentration ( @xmath226 ) . figure [ fig3 ] shows the reduced self - diffusion coefficient @xmath227 as a function of the coefficient of restitution @xmath109 . here
, @xmath228 refers to the self - diffusion coefficient ( [ 4.11bis.1 ] ) for elastic collisions .
we include the results obtained for ihs by using the first sonine approximation ( dashed line ) , @xcite the second sonine approximation ( dotted line ) and by monte carlo simulations ( symbols ) .
@xcite we observe that the agreement between the predictions of the first sonine approximation for ihs and for imm is excellent in the whole range of values of @xmath109 analyzed .
moreover , both theories compare quite well with computer simulations even beyond the quasielastic limit ( say for instance , @xmath229 ) .
however , as dissipation increases , the agreement between theory and simulation is improved when one considers the second sonine approximation . for mechanically different particles , in fig .
[ fig4 ] we plot @xmath227 versus @xmath109 in the tracer limit ( @xmath226 ) for @xmath230 and @xmath231 as given by the exact results for imm , the second sonine approximation for ihs @xcite and by monte carlo simulations .
we see that both theories are practically indistinguishable and present an excellent agreement with simulation data . beyond the above two special cases , figs .
[ fig7][fig5 ] show the dependence of the reduced coefficients @xmath227 , @xmath232 , and @xmath233 on the coefficient of restitution for @xmath97 , @xmath234 , and two values of the mass ratio ( @xmath235 and 4 ) . here
, @xmath236 is given by the first equality of eq .
( [ 4.11bis ] ) and @xmath237 .
the solid lines correspond to the imm results while the dashed lines refer to the results obtained for ihs in the first sonine approximation .
@xcite we observe that , in general , the qualitative behavior of ihs is well captured by the imm . at a quantitative level , for not strong dissipation ( say for instance , @xmath229 ) the agreement between the results derived for imm and ihs is again quite good , especially in the case of @xmath235 .
nevertheless , the discrepancies between both interaction models increase as the coefficient of restitution decreases .
for instance , at @xmath238 , the discrepancies for @xmath227 , @xmath232 , and @xmath233 are about 4% , 5% , and 8% , respectively , for @xmath239 , while they are about 1% , 3% , and 3% , respectively , for @xmath235 .
the dependence of the reduced shear viscosity coefficient @xmath240 on dissipation is plotted in fig .
[ fig8 ] for @xmath97 , @xmath234 , and for two values of the mass ratio ( @xmath235 and 4 ) . here
, @xmath241 is the corresponding value of the shear viscosity coefficient for elastic collisions .
although again the imm predictions compare qualitatively well with the ones derived for ihs , we observe that in general the discrepancies for both interaction models at the level of the shear viscosity are larger than those found for the transport coefficients of the mass flux .
the discrepancies observed in fig . [ fig8 ] are similar to those found in the single gas case .
the primary objective of this work has been to derive the hydrodynamic equations of a granular binary mixture from the boltzmann kinetic theory for _ inelastic _ maxwell models ( imm ) . in the boltzmann equation for inelastic maxwell models ( imm ) ,
the collision rate of inelastic hard spheres ( ihs ) is replaced by an effective collision rate independent of the relative velocity of the two colliding particles . as in the elastic case,@xcite
this property allows us to get exactly the velocity moments of the boltzmann collision integrals without the explicit knowledge of the velocity distribution function . here
, the chapman - enskog method has been applied to get a _
normal _ solution of the boltzmann equation for states near the local homogenous cooling state .
the derivation of the navier - stokes hydrodynamic equations consists of two steps . as a first step ,
the reference homogenous cooling state for a mixture of inelastic maxwell gases is analyzed to provide the proper basis for description of transport due to spatial inhomogeneities .
as in the case of inelastic hard spheres ( ihs ) , @xcite our solution for the homogeneous state shows that the kinetic temperatures for each species are clearly different so that the total energy is not equally distributed between both species ( breakdown of energy equipartition ) .
in addition , we also compute the fourth cumulant of the velocity distribution functions , which is a measure of the deviation of the distributions from the maxwellian form .
once the reference state is well characterized , as a second step we obtain exact expressions for the mass flux , the pressure tensor , and the heat flux in the first order of the hydrodynamic gradients ( navier - stokes order ) . from these expressions
we identify the seven relevant transport coefficients of the problem , namely the mutual diffusion @xmath120 , eq .
( [ 4.8 ] ) , the pressure diffusion @xmath122 , eq .
( [ 4.9 ] ) , the thermal diffusion @xmath121 , eq .
( [ 4.10 ] ) , the shear viscosity @xmath126 , eqs . ( [ 4.19 ] ) and ( [ 4.22 ] ) , the dufour coefficient @xmath123 , the pressure energy coefficient @xmath125 , and the thermal conductivity @xmath124 as given by eqs . ( [ 4.34])([4.43 ] ) ,
these expressions are _ exact _ ( within the context of imm ) and constitute the main goal of this paper .
this contrasts with the previous results derived for ihs , @xcite where the transport coefficients have been _ approximately _ obtained by considering the leading terms in a sonine polynomial expansion of the distribution function .
the purpose of our work is twofold .
first , the evaluation of the transport coefficients for mixtures of imm is worthwhile studying by itself as a simple model to gain some insight into the influence of collisional cooling on the transport properties of the system .
second , the comparison between the exact results for imm with the results obtained for ihs by using approximate analytical methods allows us to assess the degree of reliability of imm to reproduce the relevant behavior of ihs in the context of granular mixtures .
recent results@xcite derived for multicomponent systems in the simple shear flow problem show a good agreement between both interaction models over a wide range values of the parameter space . to make contact with the results obtained for ihs@xcite one needs to fix the collision frequencies @xmath1 .
these quantities can be seen as free parameters of the model to be adjusted to optimize the agreement with ihs . as in the study made in ref . ,
here we have chosen @xmath1 to reproduce the cooling rates @xmath2 of ihs in the local equilibrium approximation , eq .
( [ 2.31 ] ) .
an exploration of the dependence of the seven transport coefficients on the full parameter space ( mass ratio , diameters , concentrations , coefficients of restitution ) is straightforward but perhaps beyond the scope of this presentation . here
, we have focused on the dependence of the coefficients @xmath120 , @xmath122 , @xmath121 , and @xmath126 on the common restitution coefficient @xmath109 for different values of composition , sizes , and masses .
the comparison with known results for ihs@xcite has been illustrated in figs .
[ fig3][fig8 ] .
in addition , monte carlo simulations have been also included in some plots .
the comparison shows that , in general , the imm predictions are reasonably good for not too large dissipation ( say , @xmath242 ) , especially for the transport coefficients associated with the mass flux ( first - degree velocity moment ) .
the discrepancies between both interaction models increase in the case of the shear viscosity coefficient , which is related to a second - degree moment ( pressure tensor ) . however , the imm results still capture qualitatively well the dependence of @xmath126 on @xmath109 for ihs since the discrepancies between both models are about 4% at @xmath243 and about 9% at @xmath238 . as in the monocomponent case , @xcite more significant disagreement between imm and ihs are expected when one compares higher degree moments , such as the heat flux . to show these discrepancies , in fig .
[ fig9 ] we plot the reduced coefficients @xmath244 and @xmath245 for @xmath218 in the single gas case . here , @xmath246 is given by eq .
( [ 4.43.2 ] ) while the coefficient @xmath247 is given by eq .
( [ 4.43.3 ] ) . as was emphasized in ref . , the trends observed for ihs are strongly exaggerated by the imm , where @xmath215 and @xmath216 diverge at @xmath248
. a simple application of the results obtained in this paper would be the determination of the dispersion relations for the hydrodynamic equations linearized about a homogeneous state .
this analysis would allow us to identify the conditions for stability as functions of the wave vector , the dissipation , and the parameters of the mixture .
another possible direction of study would be to check the validity of onsager s reciprocal relations @xcite between the different transport coefficients associated with the mass and heat fluxes . since the system is not time reversal invariant , violation of onsager s relations
is expected for inelastic collisions .
the goal would be to assess the influence of dissipation on such violation .
we plan to work on these issues in the next future .
in this appendix we will determine the collisional integrals of @xmath249 , @xmath250 , and @xmath251 appearing in the evaluation of the transport coefficients .
the two first integrals were already obtained by one of the authors in a previous paper @xcite on the analysis of rheological properties .
now , for the sake of completeness , we display their explicit expressions : @xmath252=-\frac{w _ { rs}}{\rho_sd}\mu_{sr}(1+\alpha _ { rs } ) \left(\rho_s{\bf j}_r-\rho_r{\bf j}_s\right),\ ] ] @xmath253&= & -\frac{w _
{ rs}}{\rho_sd}\mu_{sr } ( 1+\alpha _ { rs})\left\{2\rho_s{\sf p}_r-\left ( { \bf j}_r{\bf j}_s+{\bf j}_s{\bf j}_r\right)\right . \nonumber\\ & & -\frac{2}{d+2}\mu_{sr}(1+\alpha _ { rs})\left[\rho_s{\sf p}_r+\rho_r{\sf p}_s- \left({\bf j}_r{\bf j}_s+{\bf j}_s{\bf j}_r\right)\right.\nonumber\\ & & \left.\left.+\left[\frac{d}{2}\left(\rho_rp_s+\rho_sp_r\right)-{\bf j}_r\cdot { \bf j}_s\right]\openone \right]\right\},\end{aligned}\ ] ] where @xmath254 is the @xmath255 unit tensor
. it only remains here to get the collisional integral @xmath256 corresponding to the heat flux : @xmath257.\ ] ] to simplify its calculation , a useful identity for an arbitrary function @xmath258 is given by @xmath259=\frac{w _
{ rs}}{n_s\omega_d } \int \,d{\bf v}_{1}\,\int \,d{\bf v}_{2}f_{r}({\bf v}_{1})f_{s}({\bf v}_{2 } ) \int d\widehat{\boldsymbol{\sigma}}\,\left[h({\bf v}_1'')-h({\bf v}_1)\right],\ ] ] with @xmath260 and @xmath18 .
now , we particularize to @xmath261 . in this case , using ( [ a3 ] ) one has @xmath262 from the scattering rule ( [ a4 ] ) it follows that @xmath263\widehat{\boldsymbol{\sigma } } \nonumber\\ & & \left .
+ \left[2(\widehat{\boldsymbol{\sigma } } \cdot { \bf v}_1)- \mu_{sr}(1+\alpha_{rs})(\widehat{\boldsymbol{\sigma } } \cdot { \bf g}_{12})\right]{\bf v}_1\right\}.\end{aligned}\ ] ] to perform the angular integration , one needs the results @xmath264 @xmath265 where @xcite @xmath266 taking into account eqs .
( [ a5 ] ) and ( [ a6])([a8 ] ) , the integration over @xmath19 in eq .
( [ a4bis ] ) leads to @xmath267.\end{aligned}\ ] ] the corresponding integrations over velocity give the relations @xmath268
@xmath269 @xmath270 @xmath271 @xmath272 from eqs .
( [ a10])([a13 ] ) one finally gets the expression of @xmath256 : @xmath273\rho_s{\bf q}_r\right.\nonumber\\ & & + 3\mu_{sr}^2(1+\alpha_{sr})^2\rho_r{\bf q}_s+\frac{d}{2}\left [ \mu_{sr}(1+\alpha_{rs})\left(3\mu_{sr}(1+\alpha_{rs})-4\right)+d+2\right]p_r { \bf j}_s\nonumber\\ & & + \frac{d}{2}\mu_{sr}(1+\alpha_{rs})\left[d+4 - 3\mu_{sr}(1+\alpha_{rs})\right]p_s { \bf j}_r\nonumber\\ & & + \left[\mu_{sr}(1+\alpha_{rs})\left(3\mu_{sr}(1+\alpha_{rs})-(d+6)\right)+d+2\right ] { \sf p}_r\cdot { \bf j}_s\nonumber\\ & & \left.+\mu_{sr}(1+\alpha_{rs})\left[2 - 3\mu_{sr}(1+\alpha_{rs})\right ] { \sf p}_s \cdot { \bf j}_r\right\},\end{aligned}\ ] ] where @xmath53 . in the absence of diffusion and for mechanically equivalent particles , the collisional moment @xmath274 reduces to @xmath275 { \bf q}.\ ] ] this expression coincides with the one previously derived in the monocomponent case .
@xcite finally , let us evaluate the ( dimensionless ) collision integral of @xmath276 in the hcs , eq .
( [ 2.28 ] ) : @xmath277\nonumber\\ & = & \frac{\omega_{rs}^*}{\omega_d}\int d{\bf v}_{2}^ * \int d\widehat{\boldsymbol{\sigma } } \phi_{r } ( v_{1}^*)\phi_{s } ( v_{2}^*)\left(v_1''^{*4}-v_1^{*4}\right),\end{aligned}\ ] ] where use has been made of the property ( [ a3 ] ) .
henceforth , we will use dimensionless quantities and , for the sake of simplicity , the asterisks will be deleted . the scattering rule ( [ a4 ] )
gives @xmath278\nonumber\\ & & -4\mu_{sr}(1+\alpha_{rs})(\widehat{\boldsymbol{\sigma } } \cdot { \bf g}_{12 } ) ( \widehat{\boldsymbol{\sigma } } \cdot { \bf v}_{1})\left[v_1 ^ 2 + \mu_{sr}^2(1+\alpha_{rs})^2(\widehat{\boldsymbol{\sigma } } \cdot { \bf g}_{12})^2\right].\end{aligned}\ ] ] equations ( [ a6])([a8 ] ) allows the angular integral to be performed with the result @xmath279\nonumber\\ & & -\frac{4b_2}{d+2 } \mu_{sr}(1+\alpha_{rs})({\bf v}_1\cdot { \bf g}_{12})\left[(d+2)v_1 ^ 2 + 3 \mu_{sr}^2(1+\alpha_{rs})^2g_{12}^2\right].\end{aligned}\ ] ] therefore , the collision integral @xmath280 can be written as @xmath281v_1 ^ 4\right.\nonumber\\ & & \left .
+ 3\frac{\mu_{sr}^2(1+\alpha_{rs})^2}{d+2}v_2 ^ 4+\frac{2}{d}\left[d+2 - 6\mu_{sr}(1+\alpha_{rs } ) + 3\mu_{sr}^2(1+\alpha_{rs})^2\right]v_1 ^ 2v_2 ^ 2\right\}.\end{aligned}\ ] ] finally , by taking into account that @xmath282 we get @xmath283\left[\mu_{sr}(1+\alpha_{rs})-2\right ] \langle v^{*4}\rangle_{r } \nonumber\\ & & + \frac{d(d+2)}{4}\mu_{sr}\theta_r^{-1}\theta_s^{1 } ( 1+\alpha_{rs})\left[2d+4 + 6\mu_{sr}(1+\alpha_{rs})\right.\nonumber\\ & & \left.\left.\times \left(\mu_{sr}(1+\alpha_{rs})-2\right)\right]\right\}.\end{aligned}\ ] ] in the one - dimensional case ( @xmath95 ) , eq .
( [ a20bis ] ) agrees with the results derived in ref . for the scalar maxwell model ( i.e. , with @xmath284 ) .
in addition , for mechanically equivalent particles , the results of the single gas are recovered , namely , @xcite @xmath285\right.\nonumber\\ & & \left .
+ \frac{d(d+2)}{4}(1+\alpha)(4d-1 - 6\alpha+3\alpha^2)\right\}.\end{aligned}\ ] ]
in this appendix we will provide some details on the calculation of the navier - stokes transport coefficients appearing in the expressions ( [ 3.1])([3.3 ] ) of the irreversible fluxes .
let us consider each flux separately .
to first order , the mass flux @xmath286 is defined as @xmath287 to get this flux from eq .
( [ 3.14 ] ) , we need the collisional integral of @xmath288 which has been evaluated in the appendix [ appa ] . from the linearization of eq .
( [ a1 ] ) , one has the result @xmath289 where @xmath290 is the collision frequency @xmath291 and use has been made of eq .
( [ 2.32 ] ) in the second equality .
next , we multiply both sides of eq .
( [ 3.14 ] ) by @xmath288 and integrate over @xmath292 .
the result is @xmath293 note that the temperature ratio @xmath294 depends on the hydrodynamic state through the concentration @xmath106 .
the functional dependence of @xmath294 on @xmath106 can be obtained from the hcs condition ( [ 2.21 ] ) by using the expressions ( [ 2.18bis ] ) for the partial cooling rates @xmath295 .
the mass flux has the structure given by eq .
( [ 3.1 ] ) .
dimensional analysis shows that @xmath296 , @xmath297 , and @xmath298 .
consequently , @xmath299\nabla x_1\nonumber\\ & & + \frac{\rho\zeta^{(0)}}{p}\left(\frac{3}{2}d_p+d'\right)\nabla p-\frac{\rho\zeta^{(0)}}{2t}d_p\nabla t.\end{aligned}\ ] ] upon deriving this expression use has been made of the identities @xmath300 , \end{aligned}\ ] ] @xmath301 , \end{aligned}\ ] ] where we have taken into account that @xmath302 the two latter terms coming from @xmath303 . inserting eq .
( [ 4.5 ] ) into eq .
( [ 4.4 ] ) , one gets the expressions ( [ 4.8 ] ) , ( [ 4.9 ] ) , and ( [ 4.10 ] ) for the coefficients @xmath120 , @xmath122 , and @xmath121 , respectively .
the pressure tensor @xmath304 can be written as @xmath305 where the partial contribution @xmath306 to the pressure tensor is @xmath307 the linearization of eq .
( [ a2 ] ) leads to the following expression for the collisional integral of @xmath308 : @xmath309 where @xmath310,\ ] ] @xmath311 now , we multiply both sides of eq .
( [ 3.14 ] ) ( with @xmath312 ) by @xmath308 and integrate over @xmath292 to get @xmath313 where @xmath314 and @xmath315 a similar equation can be obtained for @xmath316 from ( [ 4.16 ] ) by interchanging @xmath186 . the solution to eq .
( [ 4.16 ] ) ( and its corresponding counterpart ) has the form @xmath317 according to eq .
( [ 3.3 ] ) , the shear viscosity coefficient @xmath126 is given in terms of the coefficients @xmath181 by @xmath318 dimensional analysis requires that @xmath319 and so , @xmath320 insertion of this relation into eq .
( [ 4.16 ] ) yields the following set of coupled equations for the two coefficients @xmath181 : @xmath321 its solution is given by eq .
( [ 4.22 ] ) . the heat flux @xmath322 can be written as @xmath323 where the partial contribution @xmath324 is given by @xmath325 to get the explicit expressions for the fluxes @xmath324 we proceed in a similar way as in the case of the pressure tensor .
first , linearization of eq .
( [ a14 ] ) leads to @xmath326 where @xmath327-\omega_{12}\mu_{21}\frac{(1+\alpha_{12})}{d(d+2)}\nonumber\\ & & \times \left\{\mu_{21}(1+\alpha_{12})\left[d+8 - 3\mu_{21}(1+\alpha_{12})\right]-3(d+2)\right\},\end{aligned}\ ] ] @xmath328 @xmath329\frac{t_1}{m_1}-\frac{\omega_{12}}{2}\mu_{21}\frac{(1+\alpha_{12})}{d}\nonumber\\ & & \times \left\ { \mu_{21}(1+\alpha_{12})\left[d-3\mu_{21}(1+\alpha_{12})+2\right]\frac{t_2}{m_2}\right . \nonumber\\ & & \left .
-\frac{x_1}{x_2}\left[d+3\mu_{21}^2(1+\alpha_{12})^2 - 6\mu_{21}(1+\alpha_{12})+2\right]\frac{t_1}{m_2}\right\}.\end{aligned}\ ] ] upon writing eq .
( [ 4.25 ] ) use has been made of the relation @xmath330 .
the corresponding expressions for @xmath331 , @xmath332 , and @xmath333 can be easily obtained from eqs .
( [ 4.27])([4.29 ] ) by the change @xmath186 . from eq .
( [ 3.14 ] ) , one gets @xmath334\nabla x_1\nonumber\\ & & -\frac{d+2}{2}\frac{n_1t_1 ^ 2}{m_1p } \left(1-\frac{m_1p}{\rho t_1}+\frac{c_1}{2}\right)\nabla p\nonumber\\ & & -\frac{d+2}{2}\frac{n_1t_1 ^ 2}{m_1t}\left(1+\frac{c_1}{2}\right)\nabla t,\end{aligned}\ ] ] where @xmath335 the coefficients @xmath115 have been obtained in section [ sec3 ] .
the solution to eq .
( [ 4.31 ] ) can be written as @xmath336 the total heat flux defines the transport coefficients @xmath123 , @xmath125 , and @xmath124 through eq .
( [ 3.2 ] ) .
according to eqs .
( [ 4.23 ] ) and ( [ 4.33 ] ) , these transport coefficients are given in terms of their partial contributions @xmath207
, @xmath208 , and @xmath209 as @xmath337 from dimensional analysis , @xmath338 , @xmath339 , and @xmath340 .
consequently , @xmath341\nabla x_1\nonumber\\ & & + \zeta^{(0)}\left(\frac{5}{2}l_r+\frac{t\lambda_r}{p}\right)\nabla p+ \zeta^{(0)}\left(\lambda_r-\frac{pl_r}{2t}\right)\nabla t.\end{aligned}\ ] ] substitution of eq .
( [ 4.35 ] ) into eq . (
[ 4.31 ] ) and taking into account the expression ( [ 3.1 ] ) for the mass flux , one arrives at the coupled set of equations ( [ 4.36 ] ) for the partial contributions @xmath207 , @xmath208 , and @xmath209 .
partial support from the ministerio de ciencia y tecnologa ( spain ) through grant no .
fis2004 - 01399 ( partially financed by feder funds ) is acknowledged .
a. a. is grateful to the fundacin ramn areces ( spain ) for a predoctoral fellowship .
|
the boltzmann equation for inelastic maxwell models is used to determine the navier - stokes transport coefficients of a granular binary mixture in @xmath0 dimensions .
the chapman - enskog method is applied to solve the boltzmann equation for states near the ( local ) homogeneous cooling state .
the mass , heat , and momentum fluxes are obtained to first order in the spatial gradients of the hydrodynamic fields , and the corresponding transport coefficients are identified .
there are seven relevant transport coefficients : the mutual diffusion , the pressure diffusion , the thermal diffusion , the shear viscosity , the dufour coefficient , the pressure energy coefficient , and the thermal conductivity .
all these coefficients are _ exactly _ obtained in terms of the coefficients of restitution and the ratios of mass , concentration , and particle sizes .
the results are compared with known transport coefficients of inelastic hard spheres obtained analytically in the leading sonine approximation and by means of monte carlo simulations .
the comparison shows a reasonably good agreement between both interaction models for not too strong dissipation , especially in the case of the transport coefficients associated with the mass flux .
* key words * : navier - stokes transport coefficients ; granular mixtures ; inelastic maxwell models ; boltzmann equation . running title : inelastic maxwell mixtures
|
[ intro ] the remarkable property of critical phenomena is the universal scaling appearing in vast variety of systems ; e.g. , magnets and gases follow simple power laws for the order parameter , specific heat capacity , susceptibility , compressibility , etc . @xcite . in thermodynamic systems ,
phase transitions take place at a critical temperature @xmath2 when the coefficients that characterize the linear response of the system to external perturbations _
diverge_. then long range order appears , causing a transition to a new phase due to collective behavior of the entire system @xcite .
the condition for nonlinear saturation in the test case of the bump on tail instability in plasmas @xcite is @xmath3 @xcite , where @xmath4 is the linear growth rate of a weakly unstable langmuir wave according to the landau theory @xcite , @xmath5 is the frequency of oscillations for particles trapped by the wave .
these trapped particles generate a long range order of the wavelength @xmath6 , the saturated amplitude @xmath7 can be considered as the order parameter , and the condition of saturation can be rewritten as a power law , typical for the second - order phase transitions , @xmath8 .
however , unlike thermodynamics this scaling contains the nonthermal control parameter @xmath4 , which is determined by the slope of the distribution function @xmath9 at the phase speed @xmath10 of the perturbation near the electron plasma frequency @xmath11 .
[ for thermodynamic systems like magnets for which the magnetization @xmath12 below the curie point @xmath2 is the order parameter , the scaling is @xmath13 , where @xmath14 at @xmath15 . ]
another difference is the critical exponent itself relation ( [ scalb ] ) predicts the very unusual exponent @xmath16 ; in contrast , the hydrodynamic hopf bifurcation @xcite , also described by the same scaling between the saturated amplitude and the growth rate , has the mean field critical exponent @xmath17 .
an analysis @xcite assuming thermalization in a vlasov poisson plasma or gravitating system leads to a critical exponent @xmath18 , and
the exponent @xmath17 has also been hypothesized for the bump - on - tail instability in @xcite . however , detailed center - manifold analysis , which establishes the normal form for a weakly unstable perturbation in one - component collisionless vlasov - poisson system , confirms @xmath16 @xcite .
the exponent @xmath16 is also confirmed numerically @xcite .
the striking discrepancy between these exponents can be better understood if we consider the structure of the phase space corresponding to these cases .
the exponents @xmath17 @xcite and @xmath18 @xcite correspond either to saturation of a strongly _ dissipative _ instability or to a thermalized system . in both these cases
the distribution function can be factorized as @xmath19 , where @xmath20 can be assumed to be gaussian , and the system is described by its momenta .
the exponent @xmath16 corresponds to saturation due to nonlinear wave - particle interactions in a weakly unstable _ collisionless _ system , where correlations between coordinates and impulses are not destroyed by dissipative processes , so the description can not be reduced to moments of the distribution .
more formally , a dissipative and/or thermalized system is represented by a _
discrete _ set of momenta of the distribution function @xmath21 , which depend only on the coordinate @xmath22 , but not on the velocity @xmath23 , @xmath24 , where @xmath25 , @xmath26 , and @xmath27 are the local density , the velocity , and the temperature , respectively .
the evolution is a flow @xmath28 which maps @xmath12 onto itself , @xmath29 .
in fact , in a neighborhood of the threshold @xmath30 the evolution @xmath31 can be reduced to a normal form , which maps only the order parameter , @xmath32 , where @xmath33 ( or @xmath34 , where @xmath35 is the set of complex numbers ) , and therefore the evolution is a _ trajectory _ @xmath36 or in other words the set @xmath37 .
the phase space of a one - dimensional collisionless system is a _ continuous _
set @xmath38 and evolution can be represented as the flow @xmath39 .
( for periodic boundary conditions the phase space is isomorphic to a cylinder @xmath40 , where @xmath41 is isomorphic to a circle . )
the sets @xmath42 and @xmath43 ( or @xmath44 ) have different _ dimensionality _ , and therefore renormalization of collisionless system in a vicinity of threshold i.e. , transformation of the set @xmath43 _ and _ the mapping @xmath45 involves one or more additional dimension .
further it is shown below , that scaling transformations close to the threshold are interrelated with the additional velocity coordinate , which disappears in hydrodynamic description because of integration of the distribution function @xmath21 over @xmath23 . from the theory of critical phenomena
it is known that dimensionality @xmath46 is inseparable part of the threshold description along with the critical exponents ( e.g. , @xcite ) .
besides @xmath47 , other critical exponents : @xmath48 , @xmath49 , @xmath50 , @xmath51 , and @xmath52 , describe the following scalings of the ising universality class : ( i ) the specific heat capacity scales as @xmath53 ( ii ) the susceptibility as @xmath54 ( iii ) the response @xmath12 at @xmath55 as @xmath56 ( iv ) the correlation length as @xmath57 and ( v ) the two - point correlation function as @xmath58 these exponents are not independent , but are interrelated via scaling laws , e.g. , the widom equality @xmath59 @xcite .
these scaling laws also include _ hyperscaling _ laws such as josephson s law , @xmath60 @xcite which involves the dimensionality @xmath46 along with the exponents . for thermodynamics the mean
field exponents are of the landau - weiss set , @xmath61 , @xmath17 , @xmath62 , @xmath63 , @xmath64 , and @xmath65 , and the scaling laws hold at the formal dimensionality @xmath66 .
however , the possibilities of critical phenomena are not exhausted by the ising universality class the percolation critical exponents @xcite , which describe another vast class of critical phenomena , are different from those in thermodynamics and scaling laws hold at a different dimensionality . in particular , for the bethe lattice ( or cayley tree ) @xcite josephson s law holds at dimensionality @xmath67 . despite the description being the same
, this difference separates the cases into different _
universality classes _ with different _ upper critical dimensions _ : @xmath68 for the ising universality class @xcite and @xmath67 for percolation . for a collisionless gravitating system ,
where the saturation mechanism is the same as for the bump - on - tail instability in plasmas , the critical exponent @xmath69 , and the critical exponents @xmath70 , @xmath71 can be determined analogously to thermodynamics and calculated from the response to an external pump @xcite .
these exponents are very different from the thermodynamic set , but nevertheless satisfy the widom equality , thus suggesting the validity of scaling laws .
josephson s law also holds , but at a rather surprising dimensionality which is the fractal one @xmath72 @xcite . at the same time , the processes resulting in @xmath73 differ _ qualitatively _ from those resulting in @xmath16 , similar to thermodynamics where spatial fluctuations of the order parameter , neglected in mean field theories , result in @xmath74 , therefore suggesting other universality classes were not completely ruled out .
these could be the wave - wave interactions , responsible for the strong turbulence in plasma @xcite ) , which are next in dynamical importance and have fewer degrees of freedom @xcite . in this paper
, we use numerical simulations to study the threshold scalings in a weakly unstable collisionless vlasov - poisson system . depending on the sign of the poisson equation
this set of equations describes either a plasma system or a gravitating system .
the saturation mechanism in a collisionless gravitating system is the same as for the bump - on - tail instability in plasmas and threshold corresponds to the condition @xmath30 in both cases .
we show in section [ bas_eq ] that the eigenfrequency contains only an imaginary part , and therefore is the simplest model to study the threshold . section [ resls ] describes the results of computations of the critical exponents and demonstrates that the scaling laws describing saturation are the same for plasma and gravitation .
section [ sym_law ] addresses the scaling transformations of the phase space and the scaling law , which appears as a result of this symmetry .
the exponent which describes correlations are obtained in section [ hyp_d ] , where fisher s equality @xmath75 is also proved . in section [ rel_d_uc ]
we show that the criticality in the system is described by the dirac propagator for a fermionic field .
we obtain hyperscaling laws and calculate upper critical dimensionalities in section [ hyp_slaw ] .
[ bas_eq ] the eigenfrequencies and eigenvectors of oscillations in a vlasov - poisson system are given by dispersion relation @xmath76=0~,\ ] ] where @xmath77 is the permittivity ( dielectric permittivity in the plasma case ) .
the boundary between stable and unstable cases is determined by the condition @xmath78=0~,\ ] ] @xcite . for the bump - on - tail instability condition ( [ plasm ] )
simplifies to , @xmath79 , and criticality is related to the zero of the imaginary part of the eigenfrequency .
therefore we can employ a model which does not contain the real part ; i.e. , @xmath80 .
the simplest is the one - dimensional self - gravitating vlasov - poisson model which is described by the equations @xmath81 where @xmath21 is the distribution function , and @xmath82 is the gravitational potential .
boundary conditions are assumed to be periodic in the @xmath22-direction . for the eigenfunctions @xmath83 where @xmath84^t~,\ ] ] @xmath85^t~,\ ] ] are the spatial fourier components , the superscript @xmath27 stands for transpose , @xmath86 is the wavevector , and @xmath87 is the system length we find @xmath88 or , explicitly for the components @xmath89 and for @xmath90 , @xmath91 where @xmath92 is the fourier component of density , and @xmath93 . for a maxwellian distribution @xmath94 the dispersion relation is @xmath95 where @xmath96 is the plasma dispersion function @xcite , @xmath97 , @xmath98 , and @xmath99 is velocity dispersion . in eq .
( [ dr ] ) , @xmath100 is the critical ( jeans ) velocity dispersion for the mode @xmath101 , and @xmath102 is the background density . for small @xmath103 [ i.e. , @xmath104 the dispersion relation reads @xmath105 or @xmath106.\ ] ] which is remarkably simpler than in plasma case , where the bump on tail instability and landau damping appear due to the wave particle resonance at the phase velocity of the wave @xmath107 .
the frequency spectrum in the case of gravitation does not contain a real part , so the resonance occurs at @xmath108 ; i.e. , in the main body of the particle distribution . for all @xmath101 , @xmath109 ,
thus , if we write @xmath110 , the distance from the instability threshold is @xmath111 analogously to @xmath14 . using ( [ om ] ) and ( [ theta ] ) the dispersion relation can be rewritten for @xmath112 as @xmath113 time is measured in the units of the free - fall time @xmath114 . since @xmath115 and the density @xmath116 , where @xmath117 is the particle mass , and @xmath42 is the concentration , @xmath118 in the units assumed here . . the saturated level @xmath119 is shown by the horizontal segment.,width=325,height=151 ]
[ resls ] dispersion relation ( [ om ] ) shows that there are no unstable modes above the threshold @xmath120 , @xmath121 , and therefore the system remains invariant with respect to translations @xmath122 , where @xmath123 is any number . below the threshold
the mode @xmath124 becomes unstable , and therefore the @xmath125 symmetry breaks and reduces to a lower _ discrete _ one with respect to translations @xmath126 .
therefore @xmath120 can be considered as the critical point of a second - order phase transition , and the amplitude of the mode @xmath124 as the order parameter following the definition of landau @xcite . vs. @xmath127 for @xmath128 .
circles are calculated values , the curve is a spline approximation.,width=325,height=151 ] [ res_bt ] equations ( [ dst ] ) and ( [ pot ] ) with initial distribution @xmath129,\ ] ] were integrated numerically using the cheng - knorr method @xcite .
the amplitudes @xmath130 for @xmath131 are shown in figure [ f1 ] .
the perturbation @xmath124 grows exponentially with the growth rate predicted by dispersion relation ( [ om ] ) .
then , the growth saturates at some moment @xmath132 at the amplitude @xmath133 .
figure [ f1 ] also shows the ( exponential ) growth of perturbations with @xmath134 while ( [ om ] ) predicts exponential damping for these modes .
this growth occurs because of nonlinear coupling between modes since the term @xmath135 dominates over the term @xmath136 in equation ( [ dcm2 ] ) initially , when @xmath137 and therefore one has @xmath138 for the growth rate @xmath139 .
figure [ f2 ] shows that @xmath119 is independent on @xmath127 for small @xmath127 , but there exists some threshold value of the initial perturbation @xmath127 , when becomes dependent on @xmath127 .
this threshold amplitude @xmath140 corresponds to the trapping frequency @xmath141 . at @xmath142 the processes due to trapping become as important as of the resonance between wave and particles responsible for the linear landau damping ( or growth ) in collisionless media . therefore to rule out the influence of trapping processes on linear growth
the amplitude @xmath127 must be small to provide @xmath143 as a function of @xmath144 .
circles represent calculated data , the dashed line is the power law best fit.,width=325,height=151 ] the distribution function @xmath21 is plotted in figure [ f6 ] as a surface at the moment @xmath132 .
note that the distribution function @xmath21 becomes flat in the part of the @xmath145 domain @xmath146 separatrix , as predicted for the bump on tail instability @xcite . outside this area
the fourier component @xmath147 remains modulated by the background maxwellian distribution as assumed at @xmath148 , and the components with @xmath134 remain negligible @xcite , so the dynamically important area lies at @xmath149 , where @xmath150 . the width of the dynamically important area must be small compared to @xmath99 at maximum amplitude ( i.e. , @xmath119 , @xmath151 ) , otherwise the background distribution will be altered by evolution . assuming the above two criteria , @xmath119 is calculated as a function of @xmath144 and plotted in figure [ f3 ] . from figure [ f3 ] we see that this dependence can be approximated by the power law @xmath152 while @xmath153 for @xmath154 . rewritten in terms of the bounce frequency @xmath155 and the linear growth rate @xmath156 the power law ( [ orpar ] ) becomes @xmath157 and the coefficient @xmath158 .
these values are almost identical to @xmath16 , and to the coefficient @xmath159 in relation ( [ scalb ] ) as predicted and calculated for the bump on tail instability @xcite .
plane ) in configuration space at the moment of saturation @xmath132.,width=325,height=170 ] [ res_gm_dl ] subjecting the system to an external pump of the form @xmath160 allows one to calculate the other two critical exponents , @xmath49 and @xmath50 , which describe the response properties .
the index @xmath49 describes the divergence of the susceptibility , which can be written as @xmath161 for @xmath124 .
the results are shown by triangles ( @xmath162 ) and circles ( @xmath163 ) in figure [ f4 ] .
computation of @xmath164 at some @xmath144 requires at least 5 values of @xmath119 corresponding to the given @xmath165 . at the same time
, @xmath165 must small enough to avoid the effects of @xmath119 depending nonlinearly on @xmath165 .
again it requires extensive calculation of all quantities to high accuracy . in both cases @xmath166
is approximated by @xmath167 and @xmath168 for @xmath162 , @xmath169 for @xmath163 giving @xmath170 .
these exponents are very close to the corresponding results for @xcite , because @xmath164 is the only linear coefficient , and this is common to both wave - particle and wave - wave interactions .
the exponent @xmath49 is the same as for the mean - field thermodynamic models but , opposite to thermodynamics , the response is _ stronger _ at @xmath162 than at @xmath163 , as figure [ f4 ] shows .
the susceptibilities are @xmath171 this difference , as well as the appearance of scaling ( [ scalb ] ) and ( [ orpaw ] ) with @xmath16 instead of @xmath17 can be explained if one takes into account the difference between the landau - ginzburg hamiltonian @xmath172 [ where @xmath173 is the order parameter and @xmath174 , which describes the ising universality class , and the equation @xmath175~,\ ] ] which describes the amplitude of a weakly - unstable perturbation in a one - species vlasov - poisson system @xcite . according to ( [ craw ] ) the maximum amplitude @xmath176 at @xmath177 scales with @xmath4 as @xmath178 .
vs. @xmath179 .
circles ( @xmath163 ) and triangles ( @xmath162 ) represent experimental data , the dashed lines are the power
law best fits.,width=325,height=151 ] on the assumption that at @xmath180 the system responds linearly to an external pump @xmath181 , one can obtain the response @xmath182 to @xmath181 as @xmath183 however , equation ( [ craw ] ) is not valid at @xmath184 since it predicts unlimited growth instead of damping in this case . at small initial perturbation @xmath185 the correct evolution
is given by the linear equation @xmath186 , and the susceptibility @xmath187 is @xmath188 and at @xmath189 @xmath190 at the critical point @xmath191 ( or @xmath30 ) the response is described by another critical exponent @xmath50 @xmath192 the results of simulation are plotted in figure [ f5 ] , giving @xmath193 .
this exponent can not be obtained by the previous simple assumption from ( [ craw ] ) because of its singularity at @xmath30 .
[ sym_law ] the remarkable property of the critical exponents @xmath49 , @xmath47 , and @xmath50 is that they satisfy the widom equality ( [ wideq ] ) @xcite with high accuracy . in thermodynamics
the widom equality is a consequence of the scaling of the gibbs free energy under the transformation @xmath194 from which it can be derived straightforwardly @xcite .
the functions which comply the condition ( [ gibsym ] ) are called _ generalized homogeneous functions _ , and the condition itself is termed a _ homogeneity
_ condition . as a function of @xmath165 .
circles represent experimental data , the dashed lines are the power - law best fit.,width=325,height=151 ] the nature of this scaling for the marginally - stable vlasov - poisson system is clear from figure [ f6 ] , where the distribution function @xmath21 is plotted at the moment @xmath132 .
the remarkable property of the critical dynamics is the topological equivalence of the phase portraits for different @xmath144 : at the moments of saturation @xmath195 and @xmath196 corresponding to @xmath197 and @xmath198 we can write @xmath199 or @xmath200 for the fourier component .
transformation between @xmath195 , @xmath197 and @xmath196 , @xmath198 can be written as @xmath201 , @xmath202 , so @xmath203 a weak external pump in the form @xmath204 creates a similar topology in the phase space because of the same mechanism of the saturation and adds an additional variable to the distribution function .
the transformation can be written as @xmath205 . finally for @xmath206 we can write the homogeneity condition as @xmath207 the critical exponents @xmath47 , @xmath49 , and @xmath50 can be expressed via the scaling exponents @xmath208 , @xmath209 , and @xmath210 from which the widom equality for the vlasov - poisson system can be proved directly ( appendix [ appn1 ] ) .
they also provide a deep insight into symmetry properties of the system . according to expression ( [ beta_a ] )
@xmath211 rescaling the parameter of @xmath144 ( or growth rate ) also rescales the distribution function @xmath212 in the @xmath23-direction .
this situation differs significantly from thermodynamics where @xmath213 this expression rescales the normalized distance from the critical point with external field @xmath214 .
substituting @xmath119 according to the power law ( [ orpar ] ) for the order parameter to @xmath215 one can obtain ( assuming @xmath216 ) @xmath217 from which the scaling exponent is @xmath218 for @xmath16 ( @xmath219 for @xmath17 ) .
remarkably , the two different processes the linear growth of an unstable perturbation due to the resonant wave - particle interaction and the subsequent nonlinear saturation of this process due to particle trapping are interrelated .
while there is no thermodynamic equilibrium in the collisionless system considered here , one can define the quantity which describes the response of the system to external thermal perturbation , just as the specific heat capacity describes the response of a thermodynamic system to heat transfer , @xmath220 .
for the case , considered here @xmath221 where @xmath222 is the potential energy of the system . to calculate the specific heat capacity , @xmath223 corresponding to @xmath119 is used .
the critical exponent @xmath48 can be calculated straightforwardly from ( [ heat ] ) and ( [ orpar ] ) .
because perturbations @xmath134 are negligible for @xmath224 , @xmath225 , where @xmath226 ; i.e. , @xmath227 , and the heat capacity is given by @xmath228 where @xmath229 the scaling law ( [ alph ] ) can be proven using the homogeneity condition ( [ hom ] ) ( appendix [ appn1 ] ) .
remarkably , the critical exponent @xmath48 does not depend on the sign of poisson s equation , and result is the same for the plasma case . unlike thermodynamics where the relation between exponents @xmath47 , @xmath49 , and @xmath48 is given by rushbrooke s equality , @xmath230 , the scaling law ( [ alph ] ) does not contain the critical exponent @xmath49
nevertheless , the set of critical exponents @xmath231 , @xmath232 , and @xmath233 satisfy rushbrooke s equality with high accuracy .
[ hyp_d ] the correlation function of fluctuations for the field @xmath234 can be found from the fluctuation - dissipation theorem @xcite as @xmath235}{\varepsilon^2}~,\ ] ] @xcite , where the permittivity @xmath77 is given by ( [ drelat ] ) .
relation ( [ fdt ] ) can be integrated using the kramers - kronig dispersion relations , and in the static limit @xmath236 ( [ fdt ] ) becomes @xmath237~,\ ] ] where @xmath238 is the boltzmann constant and @xmath117 is the particle mass .
this equation can be rewritten as @xmath239 where @xmath240 the susceptibility @xmath164 can be written in terms of @xmath241 as @xmath242 @xmath62 .
the combination of @xmath99 and @xmath112 gives the characteristic length for the system from the dispersion relation ( [ om ] ) @xmath243 or , in terms of @xmath144 , @xmath244 as @xmath245 .
therefore the critical exponent that characterizes the correlation length is @xmath246 .
the correlation function @xmath247 can be rewritten in terms of @xmath248 as @xmath249 where @xmath52 is another critical exponent which characterizes the correlation function . on the other hand , using ( [ coorth ] ) one can rewrite this expression as @xmath250 and , taking into account ( [ suscc ] ) , @xmath251 from which finally we obtain the equality @xmath252 the last equality is known as fisher s equality and gives the last critical exponent , @xmath253 .
[ rel_d_uc ] the correlation function ( [ correl ] ) looks rather counterintuitive , since at @xmath254 ( damping waves ) , one has @xmath255 , and the noise is _
imaginary_. nevertheless , this unusual situation has an analog for particle - particle annihilation reactions of the type @xmath256 ( corresponds to equation @xmath257 ) , @xmath258 ( @xmath259 ) the correlation function is also negative because of _ anticorrelation _ of particles @xcite . in the case
@xmath254 the amplitude @xmath260 as @xmath261 .
it is also shown that the criticality due to these annihilation processes belongs to a certain universality class which is different from the ising universality class @xcite and therefore is not described by the landau - ginzburg hamiltonian ( [ lgw ] ) .
another unusual quantity is the correlation length @xmath262 and the wavevector @xmath248 , whose use allows us to establish the validity of fisher s equality for the collisionless system , studied here .
it is not related to the size of the system @xmath87 but to the @xmath263 in the system which determine an average path of correlated motion of particle in presence of these fluctuations . as the system approaches the threshold ,
fluctuations become correlated since the characteristic time of correlations @xmath264 diverges as @xmath245 .
this behavior is analogous to thermodynamic systems where the correlation length is the only relevant scale near the critical point as @xmath265 . to demonstrate that the criticality in the vlasov - poisson system belongs to a different class ,
let us compare the critical exponents corresponding to the jeans instability in a self gravitating _ hydrodynamical _ system @xcite , using the same approach .
the dispersion relation for this system is @xmath266 or @xmath267 where @xmath268 is the critical velocity of sound , corresponding to @xmath269 . as for kinetic case if @xmath270 , there are no unstable modes , and the correlation length is @xmath271 where @xmath272 is the reduced sound velocity in a fluid .
here we have the mean - field exponent @xmath273 .
assuming @xmath124 and dividing the both sides of dispersion relation ( [ jeansr ] ) on @xmath274 one can obtain the correlation function as @xmath275 this is the propagator of euclidean theory or of the scalar boson field @xcite from which the landau mean - field theory follows automatically . on the other hand dispersion relation ( [ om ] ) for the collisionless case
gives @xmath276 for collisionless systems the propagator thus corresponds to the vector _
fermionic _ field and describes a _ different _ class of critical phenomena . in the language of quantum field theory
the parameters @xmath277 and @xmath144 are _ bare masses_. since @xmath278 from ( [ corh ] ) and ( [ corcl ] ) one can obtain @xmath65 for the case of hydrodynamics and @xmath253 for collisionless system .
[ hyp_slaw ] the approach assumed in the previous section allows us to establish the hyperscaling law for the vlasov - poisson system which involves the dimensionality @xmath46 along with critical exponents like josephson s law ( [ josephl ] ) .
using propagator ( [ corcl ] ) which is the potential energy , the specific heat capacity @xmath44 in @xmath46-dimensional space at @xmath245 can be obtained as @xmath279 which gives @xmath280 with relation ( [ coorth ] ) , ( [ hcf ] ) becomes @xmath281 taking into account the scaling law ( [ heatcdef ] ) for the specific heat capacity @xmath44 one can obtain the hyperscaling relation which interrelates the exponents @xmath48 , @xmath51 , and the dimensionality @xmath46 @xmath282 the last equality reveals @xmath283 as the _ upper critical dimensionality _ for the vlasov - poisson system since the heat capacity becomes divergent if @xmath284 , thus indicating the importance of fluctuations in the critical area .
it also shows that the dimensionality corresponding to the critical exponents @xmath285 and @xmath246 is @xmath0 , fluctuations at @xmath286 are insignificant , and therefore @xmath285 , @xmath16 , @xmath62 , @xmath246 , and @xmath253 are the mean - field exponents .
the use of the scalar field propagator ( [ corh ] ) instead of ( [ corcl ] ) gives @xmath287 and at @xmath61 the upper critical dimensionality is @xmath68 which is the landau mean - field theory case for the ising universality class .
however , relation ( [ ivanov_h1 ] ) is not valid for the vlasov - poisson system because of its different propagator . on the contrary to relations ( [ ivanov_h ] ) and ( [ ivanov_h1 ] ) which are valid for specific propagators ( [ corh ] ) and ( [ corcl ] ) , josephson s law ( [ josephl ] ) is universal for all cases considered . with exponents @xmath246 and @xmath273 it gives @xmath1 and @xmath68 as the upper critical dimensionalities for the collisionless and hydrodynamic cases , respectively , and @xmath0 for the exponents of the vlasov - poisson system calculated here . without going into details here , we note that this universality appears because the fundamental description is given by the same functional integrals in both cases . in particular for the free scalar bosonic field ( no interactions ) the partition function is @xmath288~,\ ] ] where @xmath289 is the landau - ginzburg hamiltonian @xmath290 ( [ lgw ] ) without quadratic term . in the fermionic case the lagrangian for a dirac spinor field
is used instead of @xmath291 .
we have studied numerically and analytically a model vlasov poisson system near the point of a marginal stability .
the most important finding is that the criticality of the vlasov
poisson model studied here belongs to a universality class described by the propagator corresponding to a _ fermionic vector _ field .
this finding is in striking contrast with the previous critical phenomena studies concerning systems whose criticality belongs to universality classes corresponding to the scalar _ bosonic _ fields , like the ising universality class .
this fundamental discrepancy emerges from the _ qualitative _ difference between objects considered : the landau
ginzburg hamiltonian ( [ lgw ] ) takes into account spatial variations of the order parameter via the local differential operator @xmath292 , whereas the integro differential operator for the vlasov
poisson model acts on the distribution function containing the additional dimension of velocity .
we have calculated numerically the critical exponents which describe the critical state of the model and established analytically that these exponents and the dimensionality are interrelated by the scaling and hyperscaling laws like the widom , rushbrooke , and josephson laws at the formal dimensionality @xmath0 .
the upper critical dimensionality is @xmath1 and since @xmath293 the calculated exponents are the mean - field exponents , different from those which one might expect the landau weiss set of critical exponents corresponding to the ising mean field model where @xmath68 .
this is related to the higher dimensionality of the vlasov poisson kinetic problem associated with the velocity space and to the type of the criticality of the vlasov - poisson systems , which belongs to a universality class _ different _ from the ising universality class .
the critical exponents we have found here are @xmath285 , @xmath16 , @xmath62 , @xmath294 , @xmath246 and @xmath253 .
the difference between this set and the set @xmath295 , @xmath296 , @xmath297 , @xmath298 , @xmath246 and @xmath253 @xcite is because @xmath119 is about 50 times larger for the latter case , thus causing _ wave - wave _ interactions to dominate , thereby yielding a different universality class .
more important , the later exponents satisfy scaling laws at _ fractal _ dimension @xmath299 indicating _ reduced _ dimensionality because wave - wave interactions have fewer degrees of freedom than wave - particle ones @xcite .
this work was supported by the australian research council and a university of sydney sesqui grant .
[ appn1 ] from the homogeneity condition ( [ hom ] ) @xmath300 for @xmath301 components by integration over @xmath23 , one has @xmath302 for any two @xmath303 and @xmath304 one can write @xmath305 assuming @xmath306 , the critical exponent @xmath47 can be rewritten in terms of the scaling exponents @xmath208 and
@xmath307 as @xmath308 in the similar way for @xmath49 and @xmath50 one can write @xmath309 @xmath310 and the widom relation follows from ( [ gam_p ] ) straightforwardly : @xmath311 equations ( [ beta_a])-([delta_a ] ) can be rewritten in matrix form as @xmath312 where @xmath313^t$ ] , @xmath314^t$ ] , and the matrix @xmath315 is @xmath316 the determinant of @xmath315 is @xmath317 using ( [ delta_a ] ) to eliminate @xmath208 , the system ( [ mfr ] ) can be reduced to @xmath318 for which solution exists only if the widom equality @xmath319 holds .
therefore @xmath307 and @xmath210 can be formally considered as the eigenvectors of @xmath320 whose eigenvalue is @xmath321 . in particular
@xmath322 which indicates that rescaling of the distribution function under an external pump is equivalent to rescaling due to the field which appears for nonzero order parameter .
the heat capacity can be formally defined as @xmath323 where @xmath222 is the potential energy of the system . to calculate the specific heat capacity , @xmath223 corresponding to @xmath119 is used .
assuming @xmath327 and @xmath328 , equation ( [ cph1 ] ) can be rewritten as @xmath329= \frac{\partial}{\partial \theta}a_{sat}^2(\theta,0),\ ] ] or @xmath330 or @xmath331 equation ( [ ww2 ] ) has the form of the power law , @xmath332 , with @xmath333 the last relation corresponds to rushbrooke s equality @xmath230 at @xmath62 .
|
a model vlasov poisson system is simulated close the point of marginal stability , thus assuming only the wave - particle resonant interactions are responsible for saturation , and shown to obey the power
law scaling of a second - order phase transition .
the set of critical exponents analogous to those of the ising universality class is calculated and shown to obey the widom and rushbrooke scaling and josephson s hyperscaling relations at the formal dimensionality @xmath0 below the critical point at nonzero order parameter .
however , the two - point correlation function does not correspond to the propagator of euclidean quantum field theory , which is the gaussian model for the ising universality class .
instead it corresponds to the propagator for the fermionic _ vector _ field and to the _ upper critical dimensionality _ @xmath1 .
this suggests criticality of collisionless vlasov - poisson systems as representative of the _ universality class _ of critical phenomena of _ a fermionic _ quantum field description .
|
two - particle correlation triggered by high- particles allows us to study hard scattering phenomena when full jet reconstruction is challenging .
an analysis of the first alice pp data where charged and neutral particles isolated or not are used as trigger particles is presented . the two - particle correlation between the trigger ( @xmath0 ) and the associate ( @xmath1 ) particles
is studied as a function of the imbalance parameter @xmath2=-@xmath3 and interpreted in terms of jet fragmentation function .
two - particle correlation measurements provide a suitable tool to study the jet properties in proton - proton and heavy - ion collisions @xcite where the full jet reconstruction is challenging and even impossible for jet- below about 50 gev / c .
such a measurement can be further improved if the trigger particle is isolated . ultimately , selecting a direct photon as the trigger particle would provide the optimum choice since the direct photon sets the reference of the @xmath4 type of hard scattering kinematics and balances the jet emitted in the opposite azimuthal direction .
in addition , while purely hadronic observables are strongly biased with respect to surface emission , photons can sample the entire collision volume .
however , identifying direct photons is quite challenging because of their scarcity and the overwhelming contribution of hadron decay - photons , mainly from @xmath5 .
applying isolation criteria ( select trigger particles spatially isolated ) is the traditional approach to enrich the data sample with hadrons sampling a large fraction of the parton momentum from which they fragment .
we have studied the two - particle correlation with the alice experiment @xcite at lhc in proton - proton collisions at @xmath6@xmath7 tev .
the experimental technique consists in tagging events with a leading trigger and measuring the distribution of charged hadrons associated to this leading trigger from the same event .
such a measurement requires to reconstruct charged tracks and identify neutral particles with good momentum / energy resolution . in alice , the electromagnetic calorimeters , phos
( @xmath8 and @xmath9 = 100@xmath10 ) and emcal ( @xmath11 and @xmath12 = 100@xmath10 ) , allow us to measure and identify photons and neutral mesons with high efficiency and resolution .
particles in the calorimeters are detected as clusters of adjacent hit calorimeter cells .
the central tracking system ( its and tpc ) , covering the pseudo - rapidity @xmath13 and the full azimuth @xcite , is used for charged track measurements , and contributes to the single particle jet events identification by applying the isolation technique .
two different types of trigger particle have been selected for the correlation study : ( i ) charged triggers are selected as the charged track with the highest transverse momentum among all the tracks in the event , ( ii ) neutral cluster triggers are defined as the calorimeter cluster with the highest energy among all detected clusters and among all the charged tracks detected in the same trigger hemisphere . at this early stage of the analysis ,
no particle identification has been applied yet on the calorimeters data , thus the cluster sample may contain a sizable fraction of charged particles which develop a shower in the calorimeters or high-@xmath14 @xmath15 from which two decay photons merge into a single cluster .
a monte - carlo study based on the pythia @xcite event generator indicates that the dominant contribution to the selected neutral cluster triggers is due to merged @xmath5 in the region which could be statistically reached with the data available for our study .
the results which will be discussed are the azimuthal correlation between the two particles , and the per - trigger conditional yield of charged hadrons as a function of the imbalance variable @xmath2= -@xmath16 , where @xmath17 denote the amplitude of the transverse momenta of the trigger and associate particle , respectively .
the azimuthal correlation ( @xmath18 ) between the trigger particle and the associate particles with transverse momentum threshold of @xmath19gev/@xmath20 and @xmath21 for different trigger bins is shown in fig .
[ fig : clusterdphi ] .
the resulting distribution is finally corrected for the two - particle pair efficiency .
the structure of the azimuthal correlations show consistent results independent on the trigger type and of the calorimeter for the cluster triggers ( fig .
[ fig : clusterdphi ] ) .
the main feature of these distributions is the typical 2-jets structure with a near side ( @xmath22 ) and away side ( @xmath23 ) peak .
the peaks become stronger when the trigger- increases reflecting the increasing multiplicity of fragmented hadrons for increasing jet energies ( on average the leading particle of a jet carries about 50% of its energy at lhc energy regime ) .
the 2-jet structure sits on top of a flat background which originates from the correlation of the trigger particle with particles from the underlying event ( all particles which do not originate from a hard scattering process ) .
the per - trigger conditional yield as a function of the variable @xmath2is used to study the away side jet fragmentation function .
the accuracy of this description depends on how well the trigger particle momentum approximates the jet momentum .
the measured @xmath2distribution contains contributions from away side jet fragmentation signal as well as background . to estimate this background
, one assumes that the associate particles distribution from underlying events is isotropic in the full azimuth angle .
therefore , the background level can be calculated from the transverse azimuthal region where the jet signal is minimum .
the resulting @xmath2signal distributions after underlying event background subtraction for different bins exhibits remarkable uniformity of the slope . in order
to quantity this feature , an exponential function ( dn / d@xmath2@xmath24 ) is fitted to the @xmath2distribution in the range @xmath25 @xmath2 @xmath26 , the fitting parameter @xmath27 stands for the inverse @xmath2 slope . figure [ fig : xeslope ] shows the evolution of the inverse @xmath2 slope as a function of the mean trigger obtained from alice data .
the slopes extracted using charged and neutral triggers approach to each other with rising .
this behavior is expected as high cluster triggers are mainly merged @xmath5while low clusters are dominated by single decay photon clusters from @xmath5
. slope as a function of the mean trigger for charged and neutral triggers in pp collisions at @xmath6 = 7 tev.__,scaledwidth=90.0% ] slope as a function of the mean trigger for charged and neutral triggers in pp collisions at @xmath6 = 7 tev.__,scaledwidth=90.0% ] similar measurements performed at phenix @xcite conclude that the @xmath2distribution is not sensitive to the away side jet fragmentation function because the trigger particle carries only a small fraction of the full jet energy ( = @xmath28 ) .
this conclusion is confirmed by our measurement where the measured inverse slope is in the region where one expects ( yellow band ) the @xmath2 slope when the trigger particle samples the jet momentum at @xmath29 .
the next step of the analysis consists in selecting isolated trigger with the goal to enrich the trigger sample with single hadron jet events or direct photon ( @xmath30 ) . to select isolated particles ,
one studies the charged hadronic activity around the trigger candidate , i.e. , calculating the sum of the transverse momentum of all charged hadrons inside a cone with radius @xmath31 rad around the trigger candidate and tags the trigger to be isolated if the sum is less than 10 % of the trigger s transverse momentum .
the azimuthal correlation obtained with an isolated trigger ( fig .
[ fig : ictrigdphi ] ) , compared to the one without isolation , shows an apparent single jet structure with , by construction , the absence of the near side peak whereas the away side peak remains present .
the slight difference in the away side peak intensity before and after isolation can be explained by a bias which is different in the case with and without isolation . from a monte - carlo study based on the pythia event generator
we have demonstrated that non - isolated triggers carry on average 50% of the jet energy while isolated triggers carry on average 80% of the jet energy .
therefore , for the same trigger , the isolated one will sample a lower energy jet compared to the non - isolated trigger . the resulting inverse @xmath2 slope parameter @xmath27 obtained by the exponential fitting for the non - isolated and isolated triggers
are presented in fig .
[ fig : icslope ] as a function of the trigger transverse momentum .
one observes that the @xmath27-value is generally higher for the isolated triggers than for the non - isolated triggers .
the data points are also compared with prediction of a simple fragmentation model based on parent - child relation @xcite .
the slope approaches the region where one expects @xmath27 to represent the slope of the true fragmentation function for quarks and gluons ( blue band defined by @xmath32 ) .
slope as a function of the mean trigger for non - isolated and isolated charge triggers in pp collisions at @xmath6 = 7 tev.__,scaledwidth=85.0% ] slope as a function of the mean trigger for non - isolated and isolated charge triggers in pp collisions at @xmath6 = 7 tev.__,scaledwidth=85.0% ] the results demonstrated that two particle correlation is a suitable tool to study jet properties , where two particle correlations show a di - jet structure and the isolated trigger - hadron correlations make a better description of the away side jet fragmentation .
the future study will focus on the identified particle correlations which has potential to distinguish the quark jets ( direct photon trigger ) and gluon jets ( @xmath5trigger ) .
|
|
the crab nebula supernova remnant ( snr ) is one of the strongest radio sources in the sky and has long been an object of intense scrutiny by radio astronomers .
nevertheless , a number of controversies with important physical implications continue to surround the details of its radio continuum spectrum . while there is general agreement that the crab s integrated spectrum has a slope @xmath2 @xmath3 from 10 mhz to the synchrotron break frequency near 10@xmath4 ghz ( strom & greidanus 1992 , kovalenko , pynzar & udaltsov 1994 ) , there is considerable dispute over the details of its spatial distribution ( wilson 1972 , swinbank 1980 , velusamy , roshi & venugopal 1992 , bietenholz & kronberg 1992 , hereafter bk92 ) .
since the spectral index serves as a useful measure of the energy distribution of the relativistic electrons which generate the synchrotron emission , claims of _ spatial variations _ in spectral index across the nebula therefore must be interpreted in terms of differences in the underlying physical processes which influence the energy of the particles .
the following are the three main areas where small variations in the radio spectral index , @xmath5 , have been reported : in the first instance , bk92 report spectral index variations in the vicinity of the pulsar , similar to the wisps found in the optical ( scargle 1969 , hester 1995 ) .
although they can be seen as weak arcsecond - width features in the total intensity images , these structures are most prominent in the spectral index images .
they have a flatter spectral index than the nebula , suggesting that they are regions of renewed particle acceleration .
these spectral index ridges are located in the same region but are not positionally coincident with the optical wisps .
they have the same concave arc - like appearance as the optical wisps but are somewhat longer .
these ridges were interpreted as manifestations of the wind shock in the crab by bk92 , and kronberg ( 1993 ) modeled them as such with a tilted ring or torus geometry . secondly , velusamy ( 1992 ) have claimed that the spectrum of the filaments in the body of the nebula is different than that of the intra - filament plasma , and that there is a trend of steepening filament spectral index as a function of radius in the nebula .
bk92 , however , reported no such variations .
if true , these spectral variations must be reconciled with interpretations of the physical nature of the filaments and the source of their radio emission ( reynolds 1988 ) .
this would imply a different , or at least an additional acceleration process for the relativistic electrons in the filaments .
thirdly , several groups have reported a steepening of the spectral index near the edge of the nebula .
( velusamy & sarma 1977 , agafonov 1987 ) these measurements , however , are difficult to make , and other groups report no steepening of @xmath5 towards the edges of the nebula ( wilson 1972 , swinbank 1980 , trushkin 1986 , bk92 ) . if such a steepening is real , then it could be interpreted as evidence for a _ shell _ of emission around the crab , analogous to the steeper spectrum shell - type emission typically observed from classical shell - type snrs ( cas - a ) .
the question of whether or not such a shell exists superposed on the plerion is extremely important , since if it does , this would successfully bring to an end the search for the classical radio shell exterior to the presently observed crab snr .
the lack of such a shell has led previous workers to important conclusions regarding the energy of the supernova explosion , the efficiency of its blast wave in accelerating relativistic particles , and the nature of the circumstellar environment surrounding the sn precursor star ( nomoto 1985 ; nomoto 1987 ; frail 1995 , f95 hereafter ) .
if the existence of a shell were confirmed , all conclusions drawn from those negative searches would be in error . in order to tackle these controversies
, we bring together in this paper a comprehensive body of observational material , and present here important new sub - arcminute resolution 74 mhz observations .
this radio database , encompassing an almost two decade range in frequency , allows us to conduct the most sensitive study of the magnitude and spatial distribution of the crab s radio spectral index conducted to date .
we then use this unique body of evidence to address the observational controversies outlined above , and to address the physical implications of these findings .
we used several sets of radio observations as described in table 1 .
all data were taken at the nrao vla . these include 74 mhz observations which currently utilize only eight vla antennas .
the 74 mhz data comprised snapshot observations from three a , two b , and one c array configuration spanning several years in order to obtain sufficient coverage . the non - standard procedures employed to reduce these data are described in kassim 1993 .
the image obtained at 74 mhz is shown in fig .
[ 74img ] .
several problems which have not been addressed fully in the past complicate the determination of the spectral index from interferometer data , and especially from the vla .
firstly , the images at different frequencies need to be registered accurately .
most radio images have had the antenna phases improved by self - calibration ( see , for example pearson & readhead , 1984 ; or perley , 1989 ) once this has been done , the absolute position information is degraded . to make a spectral index map ,
however , the two maps must be registered accurately , and so , a way of recovering the absolute position is needed .
typically this problem can be ignored , for if the starting model is reasonable the shift in position is small , and registering maps by the nominal positions is sufficient .
nonetheless , shifts of up to 1/4 of a beamwidth can occur . in the case of a completely resolved source , a shift of only 1/10 of a beam can change the map amplitude at a particular pixel position by up to 25% , which can cause substantial spurious spectral index variations , especially near bright gradients such as those near the edges of snrs .
images of the crab are complicated , and registering them to less than 1/10 beamwidths is difficult in the presence of noise .
one means of accomplishing this is by using the miriad program imdiff ( dickel 1988 ) which uses the maximum entropy method of tan and gull ( 1985 ) to find the translation required to make the first image resemble the second one as closely as possible in a least squares sense ( see bietenholz , 1991 , for an application of this method to radio images of the crab ) .
the other approach is to self - calibrate the data in phase using a model of known positional offset , which will serve to bring the positions to the reference frame defined by the model ( i.e. using the image at one frequency to phase calibrate the data at another frequency ) .
a much less tractable problem , peculiar to the vla , is the correlator non - linearity ( kulkarni & heiles 1980 ) .
the correction needed to compensate for this is known as the quantization or van vleck correction , and it is _ not _ done at the vla . to our knowledge
, none of the previous vla results have addressed this problem .
the output of the digital correlator employed at the vla is a known function of the true analog correlation factor . for small correlated flux densities ,
this function is very nearly linear , and this linear approximation is what is used at the vla ( daddario 1989 ) .
however , for objects as strong as the crab , the relation between the digital correlation and the true analog correlation is no longer linear , and the derived correlated flux densities will be in error .
this error increases with the observed flux density .
it affects the real and imaginary part of the correlation coefficient independently , and since the uncalibrated phase , that of the correlated flux density , is random , it will cause random phase and amplitude errors .
the observed amplitudes , however , will always be lower than the true ones , so the resulting maps will be _
biased_. for our data , at 5 ghz , these errors can be as high as 4% in amplitude .
unfortunately , the full information needed to correct for this effect is not as yet preserved at the vla , so it is difficult to accurately recover the true correlation after the fact .
this is less likely to affect our images at 74 mhz because the correlator non - linearity is less severe at this frequency , and our signal - to - noise ratio is considerably lower .
for the data at other bands , we investigated the effects of this problem by making two sets of spectral index maps : first , the `` standard '' version which is made in the straightforward way , with the maps being registered by imdiff .
secondly we made `` careful '' spectral index maps , in which we deconvolved using only data at distances @xmath6k@xmath7 , where the correlated flux density is low enough that the error in neglecting the correlator bias is minor .
the registration was achieved by also self - calibrating in phase using a common model .
we describe the production of these images in more detail below . whether any significant bias is introduced into the maps by the correlator non - linearity
is difficult to estimate .
the process of deconvolution routinely recovers the flux density in the center of the plane starting from an estimate of zero , so one might think that it should have no problem recovering a few percent of missing flux density . on the other hand
the deconvolution process is completely free to adjust the flux density in the center of the plane , whereas in regions where there are measurements , deconvolution will attempt to fit the biased measurements .
we found that the apparent noise levels in the maps deconvolved by our `` careful '' process was lower than the `` standard '' ones .
this suggests that there _ is _ discernible noise introduced by the missing correction , since one does nt expect the rms to decrease when one reduces the amount of data .
this would suggest that the bias might also be significant .
we will now describe more fully the production of the `` careful '' spectral index images : because the measurements near the center of the plane contain the most correlated flux , they are most severely affected by the correlator non - linearity .
we made maps without using the data compromised in this way .
we chose to use a distance cutoff rather than a flux density cutoff because its effect on the images is easier to understand .
the resulting spectral index images will contain little information on angular scales larger than @xmath8 , half the size of the nebula . at 5 ghz , which is the worst case ,
the correlated flux density at this distance is @xmath9 jy and the remaining correlator non - linearity errors will be @xmath10% .
since the crab nebula is considerably larger than the largest scale of @xmath8 which we are sampling , we will recover the large scale structure by deconvolving with a default ( support ) in our maximum entropy deconvolutions ( aips task vtess ; see cornwell & evans 1985 , cornwell 1988 ) .
we will use the same default ( appropriately scaled ) for all our observing frequencies , which is justified because the spectral index is quite uniform over the nebula ( wilson 1972 , bk92 ) .
we will concentrate on mapping the _ difference _ to the nebula s average spectral index rather than its absolute spectral index , which can be more reliably determined using other methods : kovalenko ( 1994 ) have recently determined it to be @xmath11 . using
the same default map will bias our maps to look similar , and thus any spatial variation in spectral index will be _ required _ by the data ( this is the `` forced '' vtess technique of anderson & rudnick 1993 ) .
the default map was produced as follows : at each frequency ( except at 74 mhz ) we mapped the data without any restrictions .
the resulting restored maps ( convolved with an appropriate `` clean '' beam and with the residuals added ) were corrected for the primary beam pattern , and for the expansion of the crab ( 0.13% per year , bietenholz 1991 ) .
they were then convolved to a common resolution of @xmath12 , and flux - scaled to a common frequency using the baars ( 1977 ) spectral index of @xmath13 .
the recent kovalenko ( 1994 ) value of @xmath14 is consistent with the older baars value to within the uncertainty , so using the latter should not affect our results .
these maps were then averaged , and the result was used as the default map in all the subsequent deconvolution processes . for each deconvolution , it was again scaled back to the flux density and epoch appropriate for that frequency .
furthermore , in order to assure accurately registered maps , this same ( scaled and expanded ) default image was used to self - calibrate , in phase only , the data prior to deconvolution .
this process has the side effect of further biasing the maps to be as similar as is allowed by the data . in other words
it will minimize any spatial variations in the resulting spectral index map .
however , our dynamic range is high , and we will show that this bias is negligible , and that our `` careful '' spectral index maps can be used with confidence . once self - calibrated in this way , the data were transformed into the image plane , using only data at distances greater than 1.6 k@xmath7 , and then deconvolved using the aips task vtess and the appropriately scaled default map .
the resulting restored maps , were convolved to a common beamsize , contracted to a epoch 1987.4 and then spectral index maps were made in the usual way .
the uncertainties were derived from empty regions of the maps near the nebula ( especially at 5 ghz , the noise in a map corrected for primary beam pattern increases with distance from the phase center ) .
these noise estimates were consistent with the differences between 4625 and 4885 mhz maps . at 74 mhz
the pulsar is quite strong , and we first subtracted it from the fully calibrated data . furthermore , the 74 mhz observations have much lower signal to noise and fewer antennas , and the correlator non - linearity is less severe , so we did not make a `` careful '' spectral index map involving this frequency .
the relative spectral index between 327 mhz and 1.5 ghz ( ) is shown in fig .
[ spxpl ] while fig .
[ spxpc ] shows the relative spectral index between 327 mhz and 4.9 ghz ( ) ; both at a resolution of @xmath12 . these are spectral indexes relative to the mean spectral index of the nebula ( @xmath15 ) : we are not sensitive to the total flux densities , so our integrated spectral indexes are not constrained by our data .
our reduction process will make them very close to the value of @xmath16 which we assumed for the default maps .
in addition , fig .
[ profil ] shows two profiles through the spectral index map .
the profiles go through the pulsar position at position angles @xmath17 and + 45 .
we have included both the `` careful '' and the `` standard '' maps in order to assess the significance of the bias ( see 2 ) and the degree of success we have had in treating it .
both maps reveal that the spectral index of the nebula is quite uniform .
the rms over the whole nebula of , for example , is only 0.01 in @xmath5 when one excludes the region near the edge of the nebula where the signal - to - noise is low .
the differences between the `` careful '' and the `` standard '' maps are negligible in the case of : the rms of the difference in @xmath5 is 0.008 .
there is no evidence of spectral steepening of any significant magnitude near the edges of the nebula .
an `` apparent '' steepening is visible at some locations , near the nebular boundaries .
however , these occur at brightness levels less than 1% of the peak brightness of the crab nebula ( at 327 mhz ) , where the uncertainty in @xmath18 is large .
the pulsar is the strongest feature in these spectral index maps .
it appears as a point source of steeper spectral index visible in the and maps .
it shows up as a steepening of only 0.07 on the map .
this translates into a pulsar flux density of 1.9 jy at 327 mhz , which is consistent with the value of 1.3 jy , given by lorimer ( 1995 ) , since pulsar fluxes are known to be quite variable .
since the spectral index at the pulsar did not change noticeably between the `` careful '' and `` standard '' maps , any bias introduced by our `` careful '' procedure is small .
all the maps also show a slight flattening in a region of about @xmath19 to @xmath20 in size surrounding the pulsar , and extending slightly farther to the nw .
the magnitude of this effect ranges from about @xmath21 on the `` standard '' map to @xmath22 on the `` careful '' and the maps .
it is notably larger on the `` standard '' map than on the `` careful '' one .
since the angular scales involved are those which could be biased by the correlator non - linearity , which is worst at 5 ghz , we suspect that the estimate of @xmath22 is more reliable .
this effect is gradual , and remaining systematic sources of error are probably at about this level .
we note that the correlator non - linearity would tend to reduce the observed flux density especially at higher frequencies and at large angular scales , which could cause a slight spurious spectral index steepening to appear towards the edges of the nebula in vla images .
we emphasize that our conclusions above are largely independent of whether we examine the `` careful '' or the `` standard '' spectral index maps ; this suggests that the bias introduced by the correlator non - linearity is in fact small .
our relatively uniform spectral index images suggest that the visibilities at two frequencies should differ only by a scale factor .
we examined the gridded visibility data , and they show the expected linear relationship between 327 and 1.5 ghz at -distances @xmath23k@xmath7 .
the formal scatter is about @xmath24 , which is compatible with the small spectral index differences found ( since the above process accounted for neither residual calibration errors nor the primary beam pattern , we expect the scatter about a linear relationship to be somewhat larger than 1 even in the case of a completely uniform spectral index ) . the `` wisps '' which were visible in the higher resolution spectral index map of bk82 are barely discernible at this resolution .
this is simply because our beam area is about 70 times that of bk82 .
another feature _ is _ consistently visible on all the and maps : it is a lane of steeper spectral index about @xmath19 to the nw of the pulsar .
this feature has an about 0.03 steeper than the surrounding region .
the spectral index of the crab nebula s `` jet '' is insignificantly steeper than that of the body of the nebula .
formally , of the jet is @xmath25 relative to the rest of the nebula .
we conclude that the jet seems to have the same spectral index as the rest of the nebula .
this is consistent with the star trail model of cox , gull & greene ( 1991 ) , which explains the jet as the star trail of the crab s progenitor .
the star trail has cooled into a hollow tube , and the lower pressure inside the tube causes the relativistic plasma from the nebula to expand more rapidly there .
[ spxl4 m ] shows the map of the spectral index between 74 mhz and 1.5 ghz ( 4 m ) , this time at the slightly coarser resolution of @xmath26 .
because of the lower signal - to - noise of the 74 mhz map and because the correlator non - linearity is small at this lowest frequency , we expect the errors caused by not correcting for it will be less than the noise .
the pulsar is , of course , not visible in the 74 mhz map , nor the 4 m map because it has already been removed from the 74 mhz data .
however , some significant structure is visible in this map .
the most distinct feature on this map is at ra = 05 31 27.7 , dec = 21 58 42.0 and has spectral index @xmath27 _
flatter _ than the rest of the nebula .
it is most likely a low frequency _ absorption _ feature because no corresponding feature is visible at frequencies above 327 mhz .
its reality was tested by deconvolving the 74 mhz data using an appropriately scaled , translated and convolved version of the 1.5 ghz map as a default once again biasing the maps to be as similar as is allowed by the data , and it did not diminish in strength .
furthermore , since the 74 mhz data is derived from a number of observing runs , and the feature in question is not confined to only one of them , it is almost certainly neither an instrumental or a deconvolution effect .
we see no steepening of the spectral index near the edge of the nebula , confirming the result of bk92 .
we find no support for the hypothesis proposed by velusamy & sarma ( 1977 ) that the emission near the edge of the nebula comes from a supernova shell which is superimposed on the plerionic emission from the rest of the nebula . on our very sensitive spectral index maps between 74 mhz and 5 ghz
we find that the mean spectral index near the edge of the nebula per beamwidth over those regions .
] is @xmath28 _ flatter _ than that in the interior of the nebula .
the spectral index between 74 mhz and 1.5 ghz ( 4 m ) shows no steepening to the edge of the nebula either , so the spectrum near the edge maintains its power - law form at least down to 74 mhz .
the uniformity of the spectral index throughout the nebula suggests that all the synchrotron - emitting electrons have the same origin , implying that they are all accelerated by the pulsar .
velusamy ( 1992 ) claim that the spectral index of the _ filaments _ ( @xmath29 ) as opposed to that of the smooth continuum emission , steepens gradually towards the edge of the nebula , with @xmath29 increasing smoothly by 0.3 from the center to the edge of the nebula .
such a steepening would be apparent in our data even though we do not separate the filament emission from the continuum : velusamy ( 1992 ) find that the filaments account for 20% of the total flux density at 327 mhz , and since their filling factor is @xmath30 we estimate that their contribution to the surface brightness at filament locations should be @xmath31 .
this would cause the @xmath5 of the combined emission from the continuum and filaments to change by at least 0.1 which is considerably larger than any spectral index change we observe .
a very slight and gradual steepening of the spectrum _ is _ observed as one moves away from the pulsar .
the magnitude of this effect is about 0.02 in @xmath5 .
while the reality of this feature was questioned in 3 , we may expect some steepening to occur , since the radio - emitting electrons are moving outward and losing energy by synchrotron radiation .
however , unlike in much longer - lived extragalactic sources , where synchrotron aging can dominate ( carilli 1991 ) , the magnitude of this effect may be too small to account for this slight observed steepening of 0.02 in @xmath5 : if we assume an ( average ) nebular field of 5 g and that the electrons radiate all their power at @xmath32gauss@xmath33 mhz we can estimate the change in spectral index of only @xmath34 over the lifetime of the nebula . the visible crab nebula is an expanding bubble of thermal filaments and relativistic gas .
it must , at some level , interact with the exterior medium into which it expands .
this interaction would probably take the form of a shell .
if it is expanding into the ism , this shell would presumably be similar to other , shell - like supernova remnants . at 327 mhz , f95 detected no extended shell emission _ exterior _ to the nebula , and thus no evidence for a fast hydrogen envelope around the crab .
if there is no fast envelope around the visible nebula , however , then the interface between the nebula and the exterior medium is at the observed edge of the nebula .
this interface presumably involves a shock which could accelerate particles .
the emission from this interaction would thus be superposed on the visible nebula .
shock acceleration typically produces radio spectra steeper ( @xmath35 ) than that of the pulsar - accelerated interior of the nebula , so if there were a significant quantity of shock - accelerated electrons near the outer boundary , a steepening of the radio spectral index would be expected .
we find no such steepening or any `` rim '' around the outer edges of the nebula .
taking the limit of the change in @xmath5 near the edge to be @xmath36 , we find that the maximum residual surface brightness with @xmath37 near edge of the nebula is @xmath38 at 327 mhz
. this would represent @xmath39 jy if spread evenly over the nebula , which would entail a fraction of less than 2% of the total radio luminosity of the crab ( integrated from @xmath40 to @xmath41 hz ) ; and even less if the @xmath42 emission displayed a shell morphology .
this would correspond to a minimum relativistic particle energy in the putative shell of less than few @xmath43 ergs ( using standard assumptions from pacholczyk 1970 ) .
while shell - type supernova remnants typically convert between 1% and 10% of the blast energy into shock - accelerated synchrotron - emitting particles ( duric 1995 ) , the crab s blast wave is expected to have converted a far smaller fraction of its energy into particles for two reasons : firstly , the crab is not yet in the sedov phase of its evolution , and calculations show the efficiency of particle acceleration is very low in the free - expansion phase ( drury markiewicz & vlk 1989 , markiewicz , drury & vlk 1990 ) . in any case , since the shell is observed to be accelerating ( bietenholz 1991 ) it can not have converted a great deal of its original energy into relativistic particles .
secondly , the crab supernova may well have been an anomalously low - energy event ( nomoto 1985 , nomoto 1987 , pols 1994 , f95 ) . in the case that the visible crab is expanding into a halo of pre - sn wind or ejecta ,
as hypothesized by hester ( 1996 ) , we would also expect @xmath44 of the energy of the expanding bubble to have been converted in accelerated particles since the relative velocity of the synchrotron bubble is lower relative to the exterior medium .
again , we would then not expect to see synchrotron emission from the shock - accelerated electrons at the edge of the synchrotron bubble .
the lane of steeper spectral index mentioned in 2 is about @xmath19 to the nw of the pulsar .
this corresponds in position to the moving feature found in the radio by bk92 , and to the moving ` dark lane ' visible in the optical ( oort & walraven 1956 , scargle 1969 ) .
this feature is moving outward , and in the radio , it seems to have a sharp outward edge ( bk92 ) . since our 327 mhz data was taken several years _
after _ the data at 1.5 and 5 ghz , we would expect such a moving feature to generate spurious steeper spectral index feature .
we can roughly estimate the feature s proper motion to be about @xmath45 asec yr@xmath46 , representing a speed of @xmath47 kms@xmath46 .
this is marginally consistent with the optical proper motion of @xmath48 asec yr@xmath46 ( scargle 1969 ) .
possibly , the feature is slowing down , since bk92 find a marginally slower speed ( 0.6 asec yr@xmath46 ) in 1985 .
this is consistent with the feature being embedded in the large - scale flow in the nebula .
higher resolution observations would be needed to accurately determine the speed of the feature .
as discussed above , the significant feature sw of the pulsar on the 4 m spectral index map has an index 0.3 flatter than the rest of the nebula , and suggests there is low - frequency absorption towards this location .
this feature has not been seen in prior spectral index maps , and it is the availability of a sub - arcminute resolution image at 74 mhz that has enabled its detection .
comparison with the three - dimensional [ oiii ] data of lawrence ( 1995 ) indicates that it corresponds to the location of the brightest optical knot with negative radial velocity .
this places the knot on the near side of the nebula and thus supports our hypothesis that the flat spectral index feature is due to free - free absorption by thermal electrons in the line emitting filaments .
the rms variations of the spectral index relative to the mean are @xmath49 when excluding the aforementioned feature , indicating that there _ is _ real structure in the spectral index .
a plot of the intensity of the [ oiii ] emission from the front of the nebula , that with negative radial velocity , against the 4 m spectral index ( fig . [ oiii ] ) shows that there is a weak correlation between the two , in the sense that the 4 m is flatter , the low - frequency absorption is higher at the locations where there is more [ oiii ] emission .
this is what would be expected if the thermal material responsible for the [ oiii ] emission is causing free - free absorption at 74 mhz .
assuming an electron temperature of @xmath50k for the optically emitting thermal filaments , our measure of the free - free optical depth constrains the emission measure to be @xmath51 @xmath52pc , the inequality arising because not all of the synchrotron emission will originate from behind the absorbing feature . assuming a line - of - sight depth similar to the feature s width
, we calculate that the electron density in the filaments @xmath53 @xmath54 , is compatible with the values expected for the crab s filaments ( davidson & fesen 1985 ) .
this is also compatible with densities derived from the depolarization of the crab between 5 ghz and 1.5 ghz ( bietenholz & kronberg 1991 ) .
in fact , a prominent depolarization feature can be seen at this location .
we have presented new , high - resolution 74 mhz observations of the crab nebula . combining these with a large body of observational material at other radio frequencies up to 5 ghz
, we have made sensitive radio spectral index maps of the crab nebula .
we have carefully registered our maps at different frequencies , and we have taken steps to eliminate the effects of the correlator non - linearity , which will tend to contaminate data of bright objects taken at the vla . in agreement with some earlier work , we find that the radio spectral index , @xmath5 , of the crab is quite uniform over the nebula at frequencies above 300 mhz . 1 .
there is _ no _ steepening of the spectral index near the edge of the nebula .
there is thus no evidence for any steeper spectrum _ shell _ emission around the crab .
such emission is expected to be faint since the crab is still in the free - expansion phase .
2 . any gradual steepening towards the edge of the nebula is small , being on the order of 0.02 in @xmath5 .
this is much smaller than has been claimed in the past .
synchrotron radiation losses probably produce an effect somewhat smaller than this , although it is of about this order of magnitude , and hence they may be responsible .
the spectral index of the filaments is no different from that of the body of the nebula for frequencies above 300 mhz , which suggests that the origin of the relativistic electrons in the filaments is the same as that in the body of the nebula .
we confirm the existence of a moving radio feature in the nw quadrant of the nebula .
it corresponds in position to the dark lane noted by oort and walraven ( 1956 ; see also scargle 1969 ) .
it appears to be moving at @xmath47 km .
5 . at frequencies below 300 mhz , we see evidence of the free - free absorption caused by the thermal material in the optical line emitting filaments .
estimates of the electron density in the line emitting filaments derived from the amount of free - free absorption we measure are compatible with other estimates derived from the optical spectra of the filaments .
we would like to thank s. lawrence for kindly supplying us with his [ oiii ] data .
research at york university was partly supported by nserc .
nrao is operated under license by associated universities , inc .
, under cooperative agreement with nsf . basic research in radio astronomy at the naval research laboratory
is supported by the office of naval research .
finally , we thank the referee , dr . richard strom , for his useful comments .
agafonov , m. i. , 1987 , , 64 , 60 anderson , m. c. , & rudnick , l. 1993 , , 408 , 514 baars , j. w. m. , genzel , r. , pauliny - toth , i. i. k. , & witzel , a. 1977 , , 61 , 99 bietenholz , m. f. , & kronberg , p. p. 1991
, , 368 , 231 bietenholz , m. f. , & kronberg , p.
, , 393 , 206 ( bk92 ) bietenholz , m. f. , kronberg , p. p.,hogg , d. e. , & wilson , a. s. 1991 , , 373 , l59 carilli , c. l. , perley , r. a. , dreher , j. , w. , & leahy , j. p. 1991
, , 338 , 554 cornwell , t. j. 1988 , , 202 , 316 cornwell , t. j. , & evans , k. f. , 1985 , , 143 , 77 cox , c. i. , gull , s. f. , & greene , d. a. 1991 , , 250 , 750 daddario , l. r. 1989 , in asp conf .
proc . 6 , synthesis imaging in radio astronomy , ed .
r. a. perley , r. r. schwab , & a. h. bridle ( san francisco : asp ) , 59 davidson , k. , & fesen , r. a. 1985 , , 23 , 119 dickel , j. r. , sault , r. , arendt , r. g. , matsui , y. , & korista , k. t. 1988 , , 330 , 254 drury , l. oc .
, markiewicz , w. j. , & vlk , h. j. 1989 , , 225 , 179 duric , n. , gordon , s. m. , goss , w. m. , viallefond , f. , & lacey , c. 1995 , , 445 , 173 frail , d. a. , kassim , n. e. , cornwell , t. j. , & goss , w. m. 1995 , 454 , l129 ( f95 ) hester , j. j. .
1995 , , 448 , 240 hester , j. j. .
1996 , , 456 , 225 kovalenko , a. v. , pynzar , a. v. , & udaltsov , v. a. 1994 , astr .
reports , 38 , 110 kassim , n. e. , perley , r. a. , dwarakanath , k. s. , & erickson , w. c. 1993 , , 106 , 2218 kronberg , p. p. , lesch , h. , ortiz , p. f. , & bietenholz , m. f. 1993 , , 416 , 251 kulkarni , s. r. , & heiles , c. 1980 , , 85 , 1413 lawrence , s. s. , macalpine , g. m. , uomoto , a. , woodgate , b. e. , brown , l. w. , oliversen , r. j. , lowenthal , j. d. , & liu , c. 1995 , , 109 , 2653 lorimer , d. r. , yates , j. a. , lyne , a. g. , & gould , d. m. 1995 , , 273 411 markiewicz , w. j. , drury , l. oc . , & vlk , h. j. 1990 , , 487 , 502 nomoto , k. 1985 , in the crab nebula and related supernova remnants , eds .
m. kafatos & r. b. c. henry ( cambridge : cambridge university press ) , 97 nomoto , k. 1987 , in the origin and evolution of neutron stars , iau symp .
125 , ed .
d. j. helfland & j .- h .
huang ( dordrecht : reidel ) , 281 oort , j. h. & walraven , th .
1956 , bull .
netherlands , 12 , 285 .
pacholczyk , a. g. 1970 radio astrophysics ( san francisco : w. h. freeman )
perley , r. a. 1989 , in asp conf .
proc.6 , synthesis imaging in radio astronomy , ed .
r. a. perley , r. r. schwab , & a. h. bridle ( san francisco : asp ) , 287 .
pearson , t. j. , & readhead , a. c. s. 1984 , , 22 , 97 pols , o. r. 1994 , , 290 , 119 reynolds , s. p. 1988
, , 327 , 853 scargle , j. d. 1969 , , 156 , 401 strom , r. g. , & greidanus , h. 1992 , nature , 358 , 654 swinbank , e. 1980 , , 193 , 451 tan , s. m. , & gull , s. f. 1985 , , 216 , 949 trushkin , s. a. 1986 , soviet astron .
, 12 , 81 velusamy , t. , & sarma , n. v. g. 1977 , , 181 , 455 velusamy , t. , roshi d. , & venugopal v. r. 1992 , 255 , 210 wilson , a. s. 1972 , 157 , 229
|
we present the results of a new , comprehensive investigation of the radio spectral index of the crab nebula supernova remnant .
new data at 74 mhz are combined with data at 327 mhz , 1.5 ghz and 5 ghz .
in contrast to previous claims , little spatial variation in the spectral index is seen . in particular ,
between 327 mhz and 5 ghz we see no evidence of spectral steepening near the edge of the nebula , the `` jet '' or the ionized filaments .
the rms limits on any spectral index variations in these regions amount to no more than 0.01 .
we believe that earlier reports of large steepening were the result of correlator bias and image registration problems .
an elongated feature was detected 1 northwest of the pulsar which may be a continuation of the well - known wisp - like structures seen closer to the center of the nebula . at 74 mhz , we see for the first time evidence of free - free absorption by the thermal material in the crab nebula s filaments .
apart from some possible renewed acceleration occurring in the wisps , the dominant accelerator of relativistic electrons in the crab nebula is the pulsar itself .
4m@xmath0 # 1@xmath1 # 1#2 # 1 # 2 # 3.#4 # 1 # 2 # 3.#4
|
neutron stars are highly compact stars of typical radius @xmath2 km and mass @xmath3 made mostly of degenerate neutron - rich matter at densities up to several times nuclear matter saturation density @xmath4 g / cc . by tracking their long - term thermal and rotational evolution
, we can learn about the nature of matter under the crust .
for example , @xcite have proposed that the thermal history of the neutron star in cas a may be indicating the recent onset of neutron superfluidity deep in its interior .
recently , @xcite have shown how rotational evolution is linked to a reorganization of particle composition in the stellar interior , leading to switching on of exotic neutrino emission processes .
a neutron star is a complex system with intertwined physical properties that can change on relatively short astrophysical timescales . in this paper
, we will be concerned with the rotational evolution of a highly magnetized isolated neutron star .
the electromagnetic emissions of a neutron star derive from its rotational kinetic energy , and its spindown is usually measured in terms of a braking index , @xmath5 , which is dependent on the magnetic field configuration ( @xmath5=3 for a dipolar field ) .
neutron stars can also spindown through gravitational wave emissions associated to the r - mode @xcite .
in fact , the observational interest in r - modes comes from the fact that no neutron stars have been found to spin at rates near the maximum allowed frequency ( the `` break - up frequency '' ) .
r - modes offer a possible explanation of this fact : in rotating neutron stars , these modes lose energy through gravitational waves , which carry away angular momentum from the star and act as braking radiation . as the star spins down , its central density increases .
this increase could be sufficient to make baryonic matter undergo phase transitions to more exotic phases of strongly interacting matter , such as quark matter , with possible implications for gamma - ray bursts @xcite and the formation of quark stars , if theoretical conjectures about the absolute stability of strange quark matter are realized in nature @xcite . while it is almost certain that this quark phase , if it exists inside neutron stars , can not be a free gas of quarks @xcite , an interacting phase of quarks still appears to be consistent with the recent finding of a 2@xmath6 neutron star @xcite .
the main questions we seek to answer are : given a few initial parameters of the newly born neutron star , such as its spin period , magnetic field and baryonic mass , can we determine which neutron stars are likely to eventually manifest a quark phase in their interior ?
if so , how long before the transition to quark matter occurs ? in this work , we take a step towards answering these questions by taking a closer look at the rotational evolution of a newly - born hot neutron star as it spins down , and we focus on two main driving factors - magnetic braking and gravitational radiation from r - modes .
we consider different equations of state ( eos ) for neutron stars , since the eos at high density is uncertain and lacking strong empirical constraints ( although a recent statistical analysis shows that the equation of state stiffens at high densities and is consistent with the expected range in certain nuclear parameters ; * ? ? ?
furthermore , second generation gravitational wave detectors such as advanced ligo will soon be operational , and our work provides an update for a similar theoretical study @xcite performed several years ago with regard to the ligo detector .
we should note here that in contrast to the assumptions in @xcite , our work includes the effects of magnetic damping of r - modes in the spindown evolution , which leads to quantitatively different results . in section [ theorymag ] , we outline the approach and main equations that describe the neutron star s spindown .
the theoretical analysis follows in part the works of @xcite , as well as @xcite and @xcite .
section [ magresults ] collects our main conclusions from this analysis . in section [ gwevo ]
, we analyze the evolution of the gravitational wave frequency associated to the growing r - mode , along the lines of @xcite , and present results in section [ gwresults ] .
we conclude in section [ conclude ] with a preliminary calculation of the gravitational wave signal from an explosive deconfinement transition ( the quark - nova " ) .
in a previous paper @xcite , we had determined birth parameters of a neutron star which would support the transition to deconfinement driven solely by magnetic braking .
we found that neutron stars that are born with mass @xmath7 and spin - period @xmath8 ms are the best candidates to reach deconfinement density in their core ( assumed to be @xmath9 ) , given a range of neutron star magnetic fields @xmath10-@xmath11 gauss .
this small value of the initial spin - period ( or large spin frequency ) is required since magnetic braking from a rotating magnetic dipole is itself proportional to the third power of the spin frequency .
most of the increase in central density occurs soon after the rapidly rotating neutron star is born . in that work
, we neglected the role of gravitational radiation from r - modes in an adhoc way - by assuming that neutrons stars are born axisymmetric and remain so . in reality , non - axisymmetric perturbations are expected due to the violent process of a supernova , which leaves the new - born neutron star in a turbulent state @xcite .
what is then the role of r - modes and gravitational braking in comparison to magnetic braking as far as the time to deconfinement is concerned ?
previous studies on the interplay of r - modes and neutron star magnetic fields have focused on different questions - for example , @xcite has studied the r - mode in magnetized neutron stars with @xmath12 g with an aim to understand x - ray pulsations from local hot spots in accretion - powered pulsars . @xcite
have shown that magnetic fields of @xmath13 g can make magnetic braking as important as r - mode spindown , specially for slowly rotating stars , and that alfvn wave driving of the r - mode can also play a role .
0.2 cm meanwhile , other works @xcite have focused on the evolution of toroidal magnetic fields generated by the secular r - mode .
essentially , the toroidal field is generated by differential rotation of the stellar fluid associated with the r - mode .
the differential rotation is a key feature of the r - mode instability in the non - linear regime @xcite . @xcite
have shown that , for isolated neutron stars , this effect can amplify existing magnetic fields by two orders of magnitude within the time taken by the r - mode to saturate ( @xmath14 few hundred secs . ) .
back - reaction on the r - mode due to this toroidal field implies that there is an associated magnetic damping , which has to be factored into the spin evolution . for neutron stars that are accreting from a binary companion ,
the build - up of toroidal fields to sizeable values is rather slow , and according to @xcite , can take up to several hundred years .
however , we are interested in this effect on isolated neutron stars .
the magnitude of the magnetic damping term @xmath15 is proportional to the integrated time evolution of @xmath16 @xcite , with @xmath17 being the r - mode amplitude .
these authors obtained a strong damping effect by assuming that the background evolution of @xmath18 is identical to that given by the absence of any magnetic fields .
however , we solve the evolution equations for the star s angular frequency @xmath19 and @xmath18 with magnetic fields and magnetic damping from the start .
@xmath15 grows while the mode is unstable , and once it saturates @xmath15 remains constant .
this consistent inclusion of @xmath15 leads to a slower evolutionary path for @xmath18 towards saturation , hence the magnetic damping is not as effective in suppressing the r - mode .
consequently , for a range of magnetic fields and small r - mode saturation values , we find that this timescale can be much larger than the typical time taken by the neutron star to spindown to typical quark deconfinement density .
therefore , our results clearly indicate that the r - mode , for realistic saturation values , quantitatively affects the spindown evolution of an isolated neutron star .
we now discuss the relevant equations for the evolution of the star s rotation rate and r - mode evolution .
the spindown @xmath20 of a neutron star ( mass @xmath21 , radius @xmath22 ) is accompanied by a loss of energy as well as angular momentum . for the radiating star , conservation of total angular momentum @xmath23 with @xmath24 a dimensionless constant and @xmath25 ( where @xmath26 is the moment of inertia about the rotation axis ) yields @xcite @xmath27 @xmath28 is the canonical angular momentum of the r - mode to 1st order in @xmath19 @xcite .
@xmath29 is the rate of gravitational radiation associated to the @xmath30=@xmath31=2 current multipole taken for an @xmath5=1 polytropic star from eqn.(65 ) of @xcite .
this is expected to be a good approximation to a large part of the neutron star interior .
@xmath32 is the accretion rate onto the star ( assumed zero for our case since we consider isolated stars only ) and @xmath33 is the magnetic braking rate @xcite .
the dimensionless quantity @xmath34 is defined from @xmath35 where @xmath36 is the density profile ( assumed radially symmetric ) of the star significantly , which is the basis for our approximation of a symmetric profile . ] .
@xmath37 evolves in time as a result of competing influences from gravitational damping ( which feeds the r - mode ) , viscosity ( which damps the r - mode ) and the magnetic damping term @xmath15 .
both bulk and shear viscosities are included in our analysis although for simplicity we keep the star at a uniform temperature of @xmath38k for the duration of the evolution .
although viscosities ( especially bulk viscosity ) is strongly temperature - dependent , our approximation is not as drastic as it may seem .
a typical cooling profile of a neutron star , driven by neutrino emission from the modified urca process is given by @xcite @xmath39 where @xmath40 is the star s core temperature @xmath41 in units of @xmath38k , @xmath42 the initial temperature , and @xmath43 yr is the characteristic cooling time from the modified urca process .
if we begin with birth temperatures @xmath44k , we see that within a few seconds , we have @xmath45k . since the r - mode does not have a large impact until @xmath46 seconds , we can approximate @xmath47 .
subsequent cooling is on the timescale of years and is also a small perturbation on our results . using the evolution equation for @xmath37 ,
( see eqn.(4 ) of * ? ? ?
* ) , one can write @xmath48\right]-f_{\rm mag } \quad { \rm and}\\ \frac{\dot{\alpha}}{\alpha}&=&\left[f_g-[f_v+f_m]\right]-\frac{\dot{\omega}}{2\omega}\ , , \nonumber\end{aligned}\ ] ] where @xmath49 is the viscous damping rate @xcite .
we solve the two coupled equations above numerically .
we use the rns code @xcite to construct 2-dimensional models of rapidly rotating neutron stars . for a given eos and for a fixed baryonic mass
, the rns code outputs a sequence of neutron star models ( with specific gravitational mass , radius , spin frequency etc . )
that have increasing central density and decreasing angular velocity .
the fastest model in such a sequence spins at or near kepler frequency . note that the magnetic field does not appear in the rns code , and in any case , its effect on structure at the field values considered here is negligible .
the magnetic field only determines the time to deconfinement , unless the spindown is completely r - mode dominated . as in @xcite
we assume @xmath50 to be a critical density at which quarks in the interior of the neutron star become deconfined - this is not a well determined number , and may span a range from ( 4 - 8)@xmath51 if a mixed phase of quark and nuclear matter is favored @xcite . to counter this uncertainty ,
we have checked our numerical results for a higher putative deconfinement density ( @xmath52 ) and found that it does not change any of our quantitative conclusions by a significant amount . using the rns code
, we construct sequences of stars ( for a given eos ) with constant baryonic mass and decreasing spin such that the non - rotating configuration has a central density equal to @xmath50 .
we then calculate the time the rotating star takes until its central density is within @xmath53 of @xmath50 .
this is a practical way of using eqn.([spinevo ] ) to obtain the minimum mass required ( for a given eos ) to support deconfinement , since the time taken to reach zero spin is infinite from the magnetic braking being proportional to @xmath19 .
since we are interested in the time it takes for a star to spindown to the critical density for deconfinement , by choosing a sequence where the non - rotating model has a central density equal to the critical density , this time gives us the maximum time the star takes to reach the deconfinement density .
we assume a very small initial amplitude of the r - mode for the fastest spinning star in a sequence , with typical initial @xmath54 , although our results for those cases where the r - mode saturates are insensitive to this initial value .
for the case when the mode does not saturate , we do find mild sensitivity to the initial value of @xmath17 .
from the first of eqn.([spinevo ] ) , we can calculate the time - step between two consecutive neutron star models in the spin - sequence output by rns code @xmath55 - \omega(t_i ) f_{\rm mag } } \,\,.\ ] ] the inverse dependence on @xmath17 implies that for small magnetic fields , where @xmath33 is small , @xmath56 is large when the mode amplitude is just starting to grow .
the rns code then needs to generate two widely separated rotating configurations within a sequence , leading to low resolution in some parts of the @xmath57 vs @xmath58 curves for small magnetic fields ( eg . , the curve for @xmath59=@xmath10 g in fig.[pvst ] ) .
however , this is only a problem initially and does not affect the total time to deconfinement .
the above equation , along with the second of eqn.([spinevo ] ) is then used to determine the time evolution of @xmath17 . for our analysis
we have used four different magnetic field strengths equally spaced on a logarithmic scale from @xmath10 g to @xmath11 g. we display results for three different equations of state : eos a which is composed of only neutrons@xcite and uses a variational principle to determine the minimum energy state ; eos bbb2 which includes muons and uses a field - theoretic many - body approach @xcite ; and finally eos apr which is relatively stiff and admits a mixed phase of quarks and nuclear matter for the heaviest stars @xcite .
all three eos generate stable configurations for central densities exceeding 5@xmath51 and can go up to even @xmath60 .
they differ in the details of the density profile , with the softest ( eos a ) having higher interior densities for the same mass ( more compact ) .
the maximum gravitational mass exceeds @xmath61 only for the apr eos @xcite , with bbb2 providing a maximum mass of 1.92@xmath6 @xcite .
eos a is too soft to generate a 2@xmath6 neutron star and may seem inadequate to explain recent observations of such massive compact stars @xcite .
however , if such massive stars are really quark or hybrid stars , we can not rule out eos a in this way , and its inclusion is still useful simply to examine the trend of a relatively soft eos on the gravitational signal from r - mode driven spindown .
the following results are obtained on solving eqns.(4 ) numerically . in fig .
[ pvst ] , the curves show how the period , starting from the kepler rotation rate at birth , evolves as a function of magnetic field strength , for different equations of state .
note that these curves correspond to structural parameters of that _ particular _ stellar configuration which _ just _ reaches the putative quark deconfinement density at zero angular velocity - we may call this the _
critical configuration_. for any given eos , this is how we determine the minimum mass and spin period required for a quark phase to appear as a result of spindown . unlike fig . 4 of @xcite , spindown
is no longer always magnetic - dominated for arbitrary magnetic fields .
for example , for @xmath62 g , r - mode spindown takes over at a few hundred seconds .
as the instability develops , magnetic damping @xmath15 grows and limits the growth of the r - mode .
this leads to a plateau in the period that can last up to several years before magnetic damping once again becomes the dominant driver of spindown .
it is noteworthy that for the critical configuration , there is no change in the time to deconfinement for any value of @xmath59 .
however , if we choose a heavier mass at birth , the time to deconfinement would be shorter , and can occur while the star is still being spun down as a result of the r - mode . assuming that even the most massive neutron stars can be transformed to stable quark stars
if they reach the deconfinement density ( that is , the quark matter eos should be sufficiently stiff to support such a mass ) , we may expect that several such neutron stars would have already spun down to the point where they underwent a quark - hadron phase transition , and now contain quark matter in their core .
our findings here are different from the conclusions of @xcite who studied the r - mode of magnetized and slowly rotating neutron stars , and found that the r - mode alters the spindown if magnetic fields are less than @xmath63 g . due to the inclusion of the magnetic damping term , which was omitted in their work
, we find that r - mode driven effects on spindown are pronounced only for @xmath62 g , since for higher @xmath59 fields , magnetic damping effectively kills the r - mode .
we emphasize that the effects of magnetic damping for smaller @xmath59 fields are not as severe as might be expected from the work of @xcite due to the self - consistent evolution of @xmath17 in the presence of the magnetic damping term , as explained at the beginning of section [ theorymag ] .
we also find that the r - mode driven spindown is more pronounced in a relatively stiff equation of state such as eos apr .
this is because @xmath29 , the gravitational damping rate is larger for a stiffer eos and for stars with the same baryonic mass and central density at zero angular frequency , this quantity is systematically larger for a stiffer eos . in other words , for a given baryonic mass , a stiffer eos can support a larger gravitational mass and radius , and has a slightly smaller kepler frequency since its average density is smaller ( less compact ) . for eg .
, @xmath64 6 . ] , driving the r - mode unstable in a shorter amount of time ( effectively spinning the star down to deconfinement density quickly ) .
the saturation amplitude @xmath1 of the r - mode assumed in the foregoing analysis is @xmath65 .
such a large value is probably unrealistic due to multi - mode coupling and onset of non - linear effects @xcite .
if we use a smaller saturation value , we find that our conclusions change quantitatively .
the r - mode grows more slowly for a smaller @xmath1 , but continues to be important since the main damping agent @xmath15 , which is roughly proportional to @xmath66 , is also smaller . as a consequence ,
the spindown due to the r - mode is weaker , but also lasts longer , as seen in fig.[pvst0.01 ] .
this can lead to the star taking a shorter time to reach a given period in the case of smaller @xmath59 fields .
this can be seen clearly in fig.[pvslogalpha005 ] which compares the time taken by the star to spindown to a period @xmath57=3 ms , for @xmath1=0.01 and @xmath1=0.005 .
as shown in the previous section , fairly large magnetic fields of @xmath67 g or more are required to make the r - mode irrelevant for spindown . for smaller magnetic fields
, the r - mode evolution can be the main agent for spinning down rapidly rotating stars .
following @xcite , we can obtain expressions for the growth of the mode amplitude ( @xmath17 ) and the corresponding evolution of the frequency of the gravitational wave ( @xmath68 ) under certain approximations .
the r - mode amplitude grows rapidly until non - linear saturation occurs @xcite at which point @xmath69=0 .
we can obtain a description of this saturated phase from eqns.([spinevo ] ) , which imply that the rotation rate evolves according to @xmath70 where @xmath71 is the saturation amplitude of the r - mode .
note that @xmath15 does not appear since it can be eliminated using the second line of eqns.(4 ) .
its effect will however show up in the growth phase discussed below .
using the fact that @xmath72 for the @xmath73 r - mode , setting @xmath74 ( a typical value determined from eqn.([iandjbar ] ) ) , and using the frequency - dependent expression for the gravitational wave damping timescale @xmath75 , we find @xmath76 where we have used the expression for @xmath77 from @xcite and neglected @xmath78 in the denominator of eqn.([sateqn ] ) . @xmath79 and @xmath26 are expressed in terms of reduced dimensionless units .
0.5 cm in the initial stages of the mode growth phase , the viscosity controls the evolution of the r - mode , but the instability to gravitational waves soon takes over .
viscous damping from bulk viscosity can be large at @xmath80k , but as the neutron star cools rapidly on the order of seconds , our approximation of setting @xmath81 implies that viscosity does not affect the subsequent spindown behaviour or the gravitational waveform during the growth of the r - mode instability .
it follows from eqns.([spinevo ] ) that during this phase , since @xmath17 is small @xmath82 the angular velocity evolves according to the first line of eqns.(4 ) .
we assume @xmath83 for the growth phase , which is tantamount to including fully the canonical angular momentum of the r - mode in the star s physical angular momentum .
this is only justified if differential rotation is small @xcite , as is the case when the mode is still small but growing .
then , as before , the proportionality between @xmath68 and @xmath19 implies @xmath84 where we have used the expression for @xmath15 from eqn.(11 ) of @xcite .
we can now use eqns.([ftsat ] ) and ( [ ftgrow ] ) to obtain the gravitational strain amplitudes and waveforms for the saturated phase and growth phase respectively .
the strain amplitude @xmath85 corresponding to the @xmath30=@xmath31=2 mode is found by a standard multipole analysis @xcite .
including the angle - average over the position of sources in the sky , @xmath86 where @xmath87 is the current multipole as given by eqn.(3.9 ) of @xcite , the mode frequency @xmath88 and @xmath89 is the source distance , chosen henceforth to have a typical value of 20 mpc extending to the virgo cluster is chosen to encompass enough neutron stars to ensure a reasonable event rate @xcite . ] .
figure [ gwvst ] shows the time - evolution of the strain amplitude @xmath85 for various magnetic fields and for different eos for @xmath90 , while fig .
[ gwvst0.01 ] shows this for @xmath91 . for magnetic fields @xmath13 g ,
the r - mode withers much before it nears saturation , for any eos - the strain amplitude is consequently very small . for @xmath92
g , we notice a strong dependence on the equation of state .
the soft eos a does not lead to a saturating r - mode while relatively stiffer eos do . for weaker magnetic fields around @xmath93 g , the r - mode saturates and displays the behaviour shown in fig .
5 of @xcite .
we now examine the gravitational waveform in the frequency domain @xmath94 which is the fourier transform of the time signal .
this is useful in estimating the signal - to - noise ratio ( snr ) in matched filtering techniques . in the stationary phase approximation ,
@xmath95 from eqns.([ftsat ] ) , ( [ ftgrow ] ) and ( [ ht ] ) , we can find @xmath94 for both the growth phase and the saturated phase . from figs .
[ htilde ] and [ htilde0.01 ] ( for @xmath96 and @xmath97 respectively ) , we see that for magnetic fields @xmath0 g , the signal from the r - mode is similar to previous result obtained in the absence of magnetic fields @xcite .
a sharp peak at high frequency ( @xmath14 1.6 khz ) corresponding to the growth phase is seen , followed by a plateau signal at lower frequencies as the r - mode saturates .
note that the curves terminate on the left since at that point , deconfinement density is reached and the subsequent signal from the phase transition needs a detailed analysis beyond the scope of this work .
however , we refer the reader to the appendix for a preliminary treatment of this issue .
previous works @xcite indicate that the saturation amplitude @xmath1 may be quite small , around @xmath98 instead of order 1 .
still , it is useful to contrast both cases : when @xmath99 and @xmath100 . in fig .
[ hf - b ] we present our estimate of the continuous gravitational wave signal from the spindown ( assuming a point source at a distance of 20 mpc ) , compared to the anticipated sensitivity of advanced ligo and the einstein telescope @xcite .
the left panel of fig .
[ hf - b ] shows the weighted strain amplitude ( 1/@xmath101 ) versus frequency for the three eos studied in this paper at @xmath67 g , compared to the anticipated noise - weighted design sensitivities of advanced ligo and the einstein telescope .
the right panel shows the same for @xmath102 g. in both figures , we are assuming @xmath100 .
the saturated phase of the curve is at least an order of magnitude above the anticipated sensitivity for advanced ligo for all eos at @xmath67 g and could therefore be expected to be observed out to 20 mpc with advanced ligo , with a typical snr of @xmath103 20 . at @xmath102 g and for the softest eos , the snr in advanced ligo drops to @xmath103 2 .
the einstein telescope is expected to gain about an order of magnitude in sensitivity compared to advanced ligo , and hence all of the above cases should be detectable in principle .
our snr estimates only provide upper limits since matched filtering is probably impractical for such sources , where spin parameters can not be promptly measured , though the source location could be known if the supernova is observed .
however , we note that most of the signal comes from the saturated phase , which for the case of small @xmath1 occurs several years after a neutron star s birth in a supernova .
observations of the neutron star s spindown parameters may then be possible , making the matched filtering method more feasible . in the context of r - modes in newly - born neutron stars , methods other than matched filtering
have been suggested @xcite that would be less computationally intensive . as mentioned in @xcite ,
the sharp spike in the strain amplitude seen in certain cases only lasts for a brief period of time ( of the order of a minute ) and carries a small fraction of the total energy emitted in gravitational waves over the spindown epoch ( see fig .
[ gwvst ] ) .
however , for the softer eos , such as bbb2 , we see that for magnetic fields @xmath105 g , the spike is softened into an asymmetric hump .
in this case , the growth phase lasts about an hour and carries a larger fraction of the emitted energy - it may be detectable owing to its larger snr in second and third generation detectors .
furthermore , for the softest eos considered here ( eos a ) , there is no saturation regime for fields larger than @xmath106 g. these are potentially distinguishing feature of the equation of state in gravitational waves . in general , for very high magnetic fields @xmath107 g , the r - mode is strongly suppressed , and the signal in gravitational waves is weak , with little possibility of detection even in third generation detectors . for a more realistic value of @xmath1
, we find that the strain amplitude is about an order of magnitude smaller than for @xmath1=0.5 , but the signal persists for a much longer time ( several months / years for @xmath108 g ) .
this is clear from fig .
[ gwvst0.01 ] .
this is because of the dependence of @xmath15 on @xmath109 and @xmath18 . for smaller @xmath59 fields and @xmath1 ,
the weaker magnetic damping allows the r - mode to survive for a longer time , even though mode growth is slower .
this eventually leads to a _ faster _
spindown for the star , compared to a higher @xmath1 . to emphasize this effect , in fig .
[ pvslogalpha005 ] , we compare spindown with @xmath110=0.01 and 0.005 , for the same initial stellar configuration and the same magnetic field .
the crossing of the two curves in that figure illustrates this effect .
however , we note from comparing figs .
[ htilde ] and [ htilde0.01 ] that @xmath111 remains relatively independent of @xmath1 , although both the @xmath112 g and @xmath113 g cases shows a flat segment corresponding to the saturated phase for @xmath91 . essentially , this is because the decrease in the peak value of @xmath85 for smaller @xmath1 is compensated by the corresponding decrease in @xmath114 ( eg .
, as in eqn.([ftsat ] ) ) .
this implies that we require a larger interval of time integration to validate the stationary phase approximation for a smaller @xmath1 ( eqn .
[ htildeqn ] ) .
thus , @xmath111 remains almost unchanged .
we have studied the role of r - modes in spinning a rapidly rotating and magnetized neutron star down to typical quark deconfinement densities .
apart from the usual spindown associated to magnetic braking , the magnetic damping of r - modes plays an important role in the period evolution of the star .
we find that , for realistic ( small ) values of the r - mode saturation amplitude @xmath1 , the time to deconfinement is sped up by the r - mode instability when magnetic fields are of order @xmath10 g or less .
this result reflects the strong dependence of the magnetic damping effect on @xmath59 and the evolution of @xmath17 .
thus , in contrast to the result in @xcite , where the r - mode was seen to affect the spindown already for magnetic fields less than @xmath63 g , the inclusion of the magnetic damping term leads us to conclude that r - mode driven effects on spindown are pronounced only for @xmath62 g .
we also explored the gravitational wave signal generated during the spindown phase as the r - mode first grows then saturates - we follow the signal continuously until the quark deconfinement threshold is reached . for realistic values of @xmath1 ,
the r - mode saturates and leads to a strong signal in gravitational waves , except in case of very soft equations of state and very large magnetic field ( @xmath11 g or more ) .
there could be a sizeable fraction of the total rotational kinetic energy emitted as gravitational waves in this epoch , which can last several years , making it detectable in upcoming second and third generation detectors .
therefore , gravitational waves could be used in this manner to probe the equation of state inside neutron stars . as shown here , using r - modes
is an alternate way in which the eos can be probed through gravitational waves gravitational waves are going to open a new window of observation into our universe . among the many discoveries that will be made , we anticipate the exciting prospect that neutron star spindown will reveal signs of the elusive r - mode instability as well as signatures for the onset of quark matter in the core . in the appendix
we outline a first preliminary estimate of what the gravitational wave signal from a `` quark - nova '' would look like .
it would be interesting to examine this signal with detailed simulations , especially in cases when the quark - nova occurs very shortly after the neutron star is born in a supernova - the signature of this `` dual - explosion '' in gravitational wave detectors would be two very different signals coming from the same source but separated in time by a few days to few weeks , depending on the time delay between the two explosions .
such a signature would be unmistakeable in upcoming gravitational wave detectors . it remains an open question as to what happens to the neutron star when the deconfinement density is reached .
one possibility is that the entire star is converted to a quark star due to the inherent stability of strange quark matter @xcite .
if this conversion occurs in an explosive manner , a``quark - nova '' could result @xcite - what will the gravitational signal from such an event look like ?
the conversion involves a two - stage process - neutrons in the core dissolve into a @xmath115 and @xmath116 quark fluid that is more compact , causing the core to shrink , followed by combustion to @xmath117 quarks through leptonic as well as non - leptonic processes .
@xcite have studied the first stage with newtonian hydrodynamics and found that quadrupolar and quasi - radial modes are excited during the collapse , leading to gravitational wave emission with an energy output of @xmath118 ergs . however
, this work applies to a mixed phase of quarks and nuclear matter and does not consider an explosive transition that begins with non - premixed fluids . 0.2 cm a full numerical treatment of the second stage , an explosive phase transition taking into account fluid motion in 3d is a complex task and beyond the scope of this work .
however , preliminary steps have been taken in this direction .
@xcite solved hydrodynamical flow equations for the combustion of neutron matter to strange quark matter in the laminar approximation , including weak equilibrating reactions and strange quark diffusion across the burning front .
the numerical results suggest laminar speeds of @xmath119 times the speed of light , much faster than previous estimates derived using only a reactive - diffusive description @xcite .
turbulent combustion has been addressed in a recent work by @xcite who found that the combustion stops short of converting the entire star to quark matter ( the reaction is no longer exothermic ) .
this was also the conclusion in @xcite , though for a different physical reason .
neither of these works continue on to estimate the gravitational signal from the explosive combustion . following the hydrodynamical approach of these studies ,
we have estimated the gravitational wave signal @xmath85 from the following steps . * starting from a mechanically stable configuration where quark matter constitutes a small fraction of the stellar core , we initiate combustion at the speeds obtained in @xcite .
the quark fluid , described by a simple bag model equation of state is subsequently evolved using inviscid hydrodynamical equations for relativistic fluid flow ( relativistic euler equations ) coupled with newtonian gravity , assuming an axisymmetric rotating configuration of the star about the @xmath120-axis . in
the cylindrical coordinates @xmath121 , where @xmath122 is the polar angle and @xmath123 is the azimuthal angle , the relevant equations are given by @xmath124 , where @xmath125 and where @xmath126 , @xmath127 , and @xmath128 are introduced solely for the purpose of writing the relativistic euler equations in a form analagous to the more familiar non - relativistic euler equations . in the natural units where
@xmath129 , the fluid velocity , lorentz factor , gravitational vector field , and specific enthalpy are respectively given by @xmath130 , @xmath131 , @xmath132 , and @xmath133 , where @xmath134 is the specific internal energy , @xmath135 is the pressure , and @xmath122 is the gravitational potential . in the non - relativistic limit ( @xmath136 ) , we recover the non - relativistic euler equations . *
the quark matter has a density @xmath137 and is described by the bag model ( bag constant @xmath138 mev ) while the nuclear fluid has density @xmath139 and is described by the perfect caloric eos with an adiabatic index of @xmath140 . at @xmath58=0 , we initiate combustion with a density discontinuity @xmath141 , which is situated at a radial coordinate @xmath142 inside the star , and choose an initial burning front speed of @xmath143=0.01@xmath144 , which is in the range of burn velocities obtained in @xcite .
we then evolve the above equations using a weighted average flux relativistic hll ( harten , lax , and van leer ) solver on a cartesian grid with spatial resolution 0.5 km and a timestep determined by a courant number of @xmath145 .
the solver is coupled with the poisson equation @xmath146 . * solving for @xmath147 and @xmath148 , we follow @xcite to compute the quadrupole wave amplitude @xmath149 for our axisymmetric configuration , and from there the signal @xmath85 as given by eqn.(21 ) of @xcite . + @xmath150 + we assume a source distance @xmath151 mpc to plot the maximal signal strength @xmath85 and the corresponding luminosity @xmath152 in fig.[qn - signal ] . the origin of the peak in @xmath85 is a balance between increasing mass outflow and a decreasing density gradient .
the luminosity is proportional to the square of @xmath153 ( hence the sharp dip at the maximum of @xmath85 ) while the total emitted energy obtained from integrating the luminosity curve is @xmath154 ergs , which is about 5 orders of magnitude smaller than the binding energy of the neutron star ( @xmath155 ergs ) and 2 orders of magnitude less than the energy released in gravitational waves during the first stage of core - collapse , where the energy comes from coupling oscillations to rotational motion @xcite .
based on this result , the gravitational wave signal for this stage of the phase conversion would be hard to detect in advanced ligo , unless the source is galactic ( within few kpc ) . most of the energy released in the phase transition is in the form of latent heat and neutrinos .
however , our results are only a preliminary estimate , designed to provide a guide to the expected signal from a quark - nova . simulations of gravitational wave signals from realistic core - collapse supernova models indicate that convective flows driven by neutrino heating @xcite can drive a strong gravitational wave signal .
the typical time for neutrinos to diffuse out of the hot quark star ( @xmath14 0.1 - 1 sec according to * ? ? ?
* ) is about an order of magnitude larger than the duration of the signal due to explosive combustion to strange quark matter , therefore , we can expect neutrino heating to be important in our context as well .
we have not included such convective effects in our present simulation , nor the effect of magnetic fields , which can also modify rayleigh - taylor instabilities @xcite .
we are grateful to the anonymous referee who pointed out the importance of magnetic damping on the r - modes .
we also thank david shoemaker and joel tohline for comments on the manuscript , sharon morsink for help with the rns code , and sam koshy for helpful discussions on gravitational wave theory .
j. s. and p. j. would like to thank the hospitality of the department of physics and astronomy at the university of calgary , where part of this work was completed .
j. s. is supported , in part , by grant ast-0708551 from the u.s . national science foundation and , in part , by grant nnx07ag84 g and nnx10ac72 g from nasa s atp program .
p. j. and v. c. acknowledge support from from california state university long beach and the u. s. army high performance computing research center .
the research of r. o. is supported by an operating grant from the national science and engineering research council of canada ( nserc ) .
k. the curves from left to right are labelled for log(@xmath59 ) . for @xmath156 ,
magnetic braking starts to dominate the r - mode as far as spindown is concerned .
the curves terminate abruptly at the moment when the central density is within @xmath157 of the critical density ( see discussion above eqn.([deltat])).,title="fig:",scaledwidth=50.0% ] k. the curves from left to right are labelled for log(@xmath59 ) . for @xmath156 ,
magnetic braking starts to dominate the r - mode as far as spindown is concerned .
the curves terminate abruptly at the moment when the central density is within @xmath157 of the critical density ( see discussion above eqn.([deltat])).,title="fig:",scaledwidth=50.0% ] k. the curves from left to right are labelled for log(@xmath59 ) . for @xmath156 ,
magnetic braking starts to dominate the r - mode as far as spindown is concerned .
the curves terminate abruptly at the moment when the central density is within @xmath157 of the critical density ( see discussion above eqn.([deltat])).,title="fig:",scaledwidth=50.0% ] k. the curves for softer eos and larger @xmath59 fields have a much lower peak strain as in these cases the r - mode amplitude never saturates.,title="fig:",scaledwidth=50.0% ] k. the curves for softer eos and larger @xmath59 fields have a much lower peak strain as in these cases the r - mode amplitude never saturates.,title="fig:",scaledwidth=50.0% ] k. the curves for softer eos and larger @xmath59 fields have a much lower peak strain as in these cases the r - mode amplitude never saturates.,title="fig:",scaledwidth=50.0% ] k. the @xmath102 g and @xmath11 g curves are distinct from the rest , as in these cases the r - mode amplitude barely or never saturates.,title="fig:",scaledwidth=50.0% ] k. the @xmath102 g and @xmath11 g curves are distinct from the rest , as in these cases the r - mode amplitude barely or never saturates.,title="fig:",scaledwidth=50.0% ] k. the @xmath102 g and @xmath11 g curves are distinct from the rest , as in these cases the r - mode amplitude barely or never saturates.,title="fig:",scaledwidth=50.0% ] g and right panel with b=@xmath102 g. the saturation amplitude @xmath1=0.01.,title="fig:",scaledwidth=50.0% ] g and right panel with b=@xmath102 g. the saturation amplitude @xmath1=0.01.,title="fig:",scaledwidth=50.0% ] , but with r - mode saturation value @xmath91 .
note that the effect of the r - mode is still significant but appears later than for @xmath90 .
this is due to the nature of magnetic damping .
, title="fig:",scaledwidth=50.0% ] , but with r - mode saturation value @xmath91 .
note that the effect of the r - mode is still significant but appears later than for @xmath90 .
this is due to the nature of magnetic damping .
, title="fig:",scaledwidth=50.0% ] , but with r - mode saturation value @xmath91 .
note that the effect of the r - mode is still significant but appears later than for @xmath90 .
this is due to the nature of magnetic damping .
, title="fig:",scaledwidth=50.0% ] = @xmath10 g ) but with different r - mode saturation values : @xmath1=0.01 and 0.005 .
time to reach @xmath57=3 ms ( where supposed deconfinement density 5@xmath51 is reached ) is shorter for smaller @xmath1 . ,
scaledwidth=50.0% ] but with @xmath91 . for smaller values of @xmath59 , the signal reflects the fact that the r - mode evolves on much longer timescales than for @xmath90.,title="fig:",scaledwidth=50.0% ] but with @xmath91 .
for smaller values of @xmath59 , the signal reflects the fact that the r - mode evolves on much longer timescales than for @xmath90.,title="fig:",scaledwidth=50.0% ] but with @xmath91 . for smaller values of @xmath59 ,
the signal reflects the fact that the r - mode evolves on much longer timescales than for @xmath90.,title="fig:",scaledwidth=50.0% ] , but with @xmath91 .
the @xmath11 g ( and for eos a also the @xmath102 g ) are distinguished , as in these cases the r - mode amplitude never saturates even when we assume such a low saturation value.,title="fig:",scaledwidth=50.0% ] , but with @xmath91 .
the @xmath11 g ( and for eos a also the @xmath102 g ) are distinguished , as in these cases the r - mode amplitude never saturates even when we assume such a low saturation value.,title="fig:",scaledwidth=50.0% ] , but with @xmath91 .
the @xmath11 g ( and for eos a also the @xmath102 g ) are distinguished , as in these cases the r - mode amplitude never saturates even when we assume such a low saturation value.,title="fig:",scaledwidth=50.0% ] and luminosity @xmath152 as a function of time for combustion of neutron matter to strange quark matter following quark deconfinement inside a neutron star.,title="fig:",scaledwidth=50.0% ] and luminosity @xmath152 as a function of time for combustion of neutron matter to strange quark matter following quark deconfinement inside a neutron star.,title="fig:",scaledwidth=50.0% ]
|
we study the spindown of isolated neutron stars from initially rapid rotation rates , driven by two factors : ( i ) gravitational wave emission due to r - modes and ( ii ) magnetic braking . in the context of isolated neutron stars ,
we present the first study including self - consistently the magnetic damping of r - modes in the spin evolution .
we track the spin evolution employing the rns code , which accounts for the rotating structure of neutron stars for various equations of state .
we find that , despite the strong damping due to the magnetic field , r - modes alter the braking rate from pure magnetic braking for @xmath0 g . for realistic values of the saturation amplitude @xmath1 , the r - mode can also decrease the time to reach the threshold central density for quark deconfinement . within a phenomenological model
, we assess the gravitational waveform that would result from r - mode driven spindown of a magnetized neutron star . to contrast with the persistent signal during the spindown phase
, we also present a preliminary estimate of the transient gravitational wave signal from an explosive quark - hadron phase transition , which can be a signal for the deconfinement of quarks inside neutron stars .
|
the ads / cft correspondence @xcite provides a unique window into the dynamics of a class of strongly coupled gauge field theories . for large @xmath0 gauge theories , the dynamics in the planar limit
is expected to be effectively classical , with @xmath1 controlling the quantum corrections . for a class of superconformal field theories arising as world - volume theories on d - branes or m - branes ,
the ads / cft correspondence identifies this classical dynamics of the single trace sector with that of classical string theory in a higher dimensional spacetime .
furthermore , if the field theory is strongly coupled then one can truncate to the zero mode sector of the string theory , viz .
, classical gravity in this higher dimensional spacetime .
the correspondence therefore provides an avenue to explore the strong coupling dynamics of field theories by reformulating the physics in terms of an effective classical gravity theory .
in this paper we continue our investigation of strongly coupled field theories on black hole backgrounds using the ads / cft correspondence , generalizing the results of @xcite .
field theories in curved spacetime are known to exhibit a rich array of physical phenomena ranging from vacuum polarization and particle production to hawking radiation and its associated puzzles with information loss .
however , much of the investigation in the past has focussed on perturbative field theory due to the lack of access to the full non - perturbative quantum state , even in the context where gravity is non - dynamical .
our current interest lies in understanding relevant quantum states beyond perturbation theory ; this is where ads / cft comes into play .
consider a field theory on a non - dynamical curved spacetime ( which we denote as @xmath2 ) with metric @xmath3 .
we would like to know the behavior of interesting quantum states and in particular the expectation values of gauge invariant local operators at the non - perturbative level .
if we restrict attention to strongly coupled conformal fields which arise in low - energy limits of d - brane world - volume theories , then we can exploit the ads / cft correspondence to answer these questions .
this is achieved by identifying higher - dimensional asymptotically ads gravitational solutions dual to the desired field theory states on @xmath2 ; we will refer to these gravitational saddle points as @xmath4 .
such bulk spacetimes are found by solving the gravitational equations of motion subject to the boundary condition that @xmath4 has as its timelike boundary @xmath2 . of the bulk spacetime @xmath4
be in the same conformal class as @xmath2 .
we will however demand that @xmath5 in fact is isometric to @xmath2 . ] in particular , smooth static spacetimes are candidate duals for the field theory hartle - hawking states .
of course , it might turn out that the field theory in question has a non - trivial phase structure , which implies that there are multiple such static saddle points for the bulk gravity description @xcite .
in @xcite we examined in some detail the holographic duals of field theories on black hole backgrounds @xmath2 .
in particular , we have argued that there are new classes of black hole geometries in asymptotically ads spacetimes : i ) single connected horizon solutions which we called black funnels and ii ) solutions with two disconnected horizons ; see . in the latter case , we called the component connected to the boundary horizon a black droplet . on the boundary ( we will refer to this as @xmath6-compact " ) .
in contrast , when the boundary spacetime has a good asymptotic region describing physics far from the boundary black hole , a black funnel should asymptote to the bulk solution describing a deconfined plasma in this distant region of spacetime . for example , in the case of asymptotically flat boundaries , it should asymptote to the planar ads black hole . as we discuss below
, the distinction between droplets and funnels is more subtle for spatially compact boundary metrics . ]
the funnel solutions are dual to black holes coupling strongly to the field theory plasma .
in contrast , the droplet solutions describe black holes coupling only weakly . in particular , the second outer horizon present in these solutions is interpreted as the field theory plasma , while the droplet itself describes field theory vacuum polarization near the horizon .
the lack of connection between these two ads horizons is evidence of the weak coupling .
note that while the gravitational dynamics may allow solutions where the two horizons have different temperatures , only the equal temperature solutions can be dual to field theory hartle - hawking states . as evidence for this picture , ref .
@xcite exhibited funnel and droplet solutions dual to @xmath7 and @xmath8 dimensional conformal fields living on black hole backgrounds .
in the @xmath7 case , we constructed black funnel solutions in whose boundary is the two dimensional black hole @xcite . while black droplets do not arise for @xmath7 boundary black holes , by exploiting the known ads c - metric solutions we were also able to construct both black funnels and black droplets in four bulk spacetime dimensions .
however , the droplet solutions found in this way do not generically describe hartle - hawking states as they were accompanied by a second horizon of a different temperature , so that the solutions did not describe equilibria .
due to the properties of the c - metric , the relevant 2 + 1 boundary metrics described black holes which asymptote to @xmath9 . a natural question to ask
is whether further interesting solutions are hidden among the ads c - metrics . in the present work
, we analyze this issue within the family of uncharged , non - rotating ads c - metric solutions .
recall that the ads c - metric solutions have been useful in the past to construct localized black holes on a uv brane in the brane - world context @xcite and also more recently to construct plasma ball solutions on an ir brane @xcite ( see also @xcite ) our interest is to remove the uv brane and work with some prescribed boundary metric .
since we are not a - priori fussed about what metrics we have on the boundary ( apart from the fact that they be black hole like ) , it seems plausible that new interesting solutions will emerge .
as we shall see in the following , there is indeed a rich class of boundary black holes contained within the c - metric family . with this motivation
, we undertake an exhaustive search of the ads c - metric family of solutions and find an interesting class of black funnel and black droplet solutions . in all cases , by using standard holographic methods we are able to compute the boundary stress tensor which includes the contribution from the quantum dynamics of the field theory in curved spacetime .
we find that the stress tensor does indeed capture the thermal aspect of hawking radiation and is furthermore regular on the black hole horizon in the boundary . on the boundary we generally find black holes living in spatially compact universes ;
i.e. , there are no spatial asymptopia .
this class of solutions is therefore somewhat different from those which arise for the special choice of parameters examined in @xcite , where the boundary metric was asymptotically a hyperbolic cylinder @xmath9 . due to the absence of an asymptotic region
, the precise definitions of funnel and droplet given in @xcite do not apply .
we therefore extend these definitions in below .
our new definitions are sufficient for spacetimes such as the c - metric which have an appropriate rotational killing field , even if they lack a useful asymptotic region .
the organization of this paper is as follows : we begin in with a brief overview of the ads c - metric solutions .
while these geometries have been studied in the literature before , we find it useful to review and generalize some of the results , especially those pertaining to the precise coordinate domains .
we then analyze the c - metric family in detail in section , where we show that , apart from trivial cases that are exactly ads or a quotient , any uncharged , non - rotating asymptotically ads c - metric with vanishing nut charge can be interpreted in terms of funnels , droplets , and planar black holes .
the discussion can clearly be generalized to include additional charges , but we refrain from doing so here in order to keep the discussion simple .
we extract the boundary stress tensor for these solutions in which allows us to see the advertised thermal behavior of the field theories in black hole backgrounds .
we end with a discussion in and describe some subtle limits in appendix [ s : lambn1 ] .
the c - metric solution in four dimensions corresponds physically to a pair of black holes being uniformly accelerated by a cosmic string .
the most general solution was found in @xcite in the context of einstein - maxwell theory with a cosmological constant .
the general solution is specified by seven parameters , corresponding to the mass , angular momentum , an acceleration parameter , electric and magnetic charges , cosmological constant and a nut parameter .
we will be interested in a sub - class of these solutions which we will exploit in the context of the ads / cft correspondence to study four dimensional bulk spacetime duals of @xmath8 dimensional field theories living on a black hole background .
consider the sub - class of ads c - metrics @xcite whose line element is given in conventional c - metric coordinates as : @xmath10 where the functions @xmath11 and @xmath12 in are of the form : @xmath13 these metrics describe uncharged , non - rotating solutions with vanishing nut charge which are negatively curved , i.e. , they solve einstein s equations with negative cosmological constant @xmath14 the ads c - metric describes the geometry of accelerating black holes in an asymptotically ads spacetime . in the metric
, @xmath15 captures the ( inverse ) acceleration , while @xmath16 is related to the cosmological constant and @xmath17 is the mass parameter of the black hole(s ) .
since cases with @xmath18 are locally isometric to either flat space or ads@xmath19 , we use the symmetry @xmath20 to take @xmath21 .
the bulk ads scale is @xmath22 so that @xmath23 at fixed @xmath15 is the flat space limit ( and @xmath24 would give desitter - c metrics ) .
we therefore take @xmath25 .
finally , @xmath26 is a discrete variable taking values @xmath27 and corresponds to different allowed topologies for the black holes ; @xmath28 corresponds to topologically spherical horizons while @xmath29 corresponds to non - compact horizons with @xmath30 topology .
a detailed discussion of the ads - c metric properties can be found in @xcite for @xmath28 and in @xcite for other values of @xmath26 . here
we note only that by taking a suitable @xmath31 limit one can recover the standard black hole for @xmath28 . for @xmath32
one obtains the topological black hole of @xcite and one can get the planar ads black hole in the case when @xmath33 .
let us now examine the geometry of the ads c - metric , more carefully , to classify all the distinct possibilities as we vary the parameters @xmath16 , @xmath17 , and @xmath26 ( we can ignore the parameter @xmath15 as it merely provides an overall scale ) .
the key features ( boundaries , horizons , singularites ) are determined by the coordinate ranges and the roots of the functions @xmath11 and @xmath12 in .
below , we first discuss the coordinate ranges and motivate the physical properties of the solutions , and in we study the root structure in more detail ( summarized in for the case of @xmath34 and for @xmath35 ) . due to the conformal factor @xmath36 in the metric
, it is clear that the boundary of the spacetime is at @xmath37 .
furthermore , the spacetime has singularities at @xmath38 and at @xmath39 which are genuine curvature singularities ; the kretschmann scalar @xmath40 diverges as @xmath41 @xcite .
typically this has led previous analysis of the ads c - metric to restrict attention to the region @xmath42 .
however , we will see that it is also sensible to consider the region @xmath43 , at least for certain choices of parameters . of course
, in making these choices we have to ensure that the spacetime in question does nt have any naked singularities .
we will return to this issue after a short examination of the coordinate ranges . to determine the range of @xmath44 we need to examine the function @xmath45 .
being a cubic , @xmath45 generically has three roots and as a result we can have two cases : ( i ) either the roots are all real , ( ii ) or only one root is real . when @xmath28 the distinction between the two situations is controlled by the mass parameter @xmath17 , whereas for @xmath46 only the case ( ii ) occurs ( see )
. we will order the roots as @xmath47 , @xmath48 and @xmath49 , with @xmath47 taken to be the smallest root of @xmath45 in case ( i ) and we take @xmath49 to be the solitary root in case ( ii ) .
horizons arise at the values of @xmath50 for which @xmath51 .
we will denote the roots again by @xmath52 , @xmath53 and @xmath54 with @xmath55 . in the situations where @xmath56 has a single root , we will for simplicity
denote it by @xmath52 ( even when it is continuously connected to the @xmath54 root ) .
for instance , in the simple case @xmath57 with @xmath28 we have @xmath58 in general the formula for the roots @xmath59 as a function of @xmath17 and @xmath16 is messy and we will not write it down . however , it is useful to note that from @xmath21 , the relation @xmath60 , and the fact that @xmath61 is an extremum of both functions , one can deduce the ordering @xmath62 when all roots are real .
see for plots of the functions @xmath11 and @xmath12 for various values of parameters .
when some roots are complex , the remaining real roots generally still satisfy ( [ order ] ) with the missing roots removed from the list .
a rather trivial exception occurs for @xmath63 in which case we have @xmath64 , replacing two of the inequalities in ( [ order ] ) by equalities .
the only other exception occurs in certain cases where @xmath11 and @xmath12 both have only one root for which @xmath65 ; i.e. , only the relative order of @xmath52 and @xmath66 differs from ( [ order ] ) . in the limit @xmath23 ,
the roots of @xmath11 and @xmath12 coincide and satisfy @xmath67 for @xmath34 or @xmath68 for @xmath46 . and
@xmath45 for various values of @xmath16 with the different panels corresponding to varying @xmath17 and @xmath26 as indicated under the respective plots .
the solid curve is the function @xmath45 while the dashed curves correspond to @xmath69 which are plotted for various values of @xmath16 ( which corresponds to the value of @xmath70 ) .
the lowest ( dot - dashed ) curve has the limiting value @xmath71 .
, title="fig : " ] and @xmath45 for various values of @xmath16 with the different panels corresponding to varying @xmath17 and @xmath26 as indicated under the respective plots .
the solid curve is the function @xmath45 while the dashed curves correspond to @xmath69 which are plotted for various values of @xmath16 ( which corresponds to the value of @xmath70 ) .
the lowest ( dot - dashed ) curve has the limiting value @xmath71 .
, title="fig : " ] + and @xmath45 for various values of @xmath16 with the different panels corresponding to varying @xmath17 and @xmath26 as indicated under the respective plots .
the solid curve is the function @xmath45 while the dashed curves correspond to @xmath69 which are plotted for various values of @xmath16 ( which corresponds to the value of @xmath70 ) .
the lowest ( dot - dashed ) curve has the limiting value @xmath71 .
, title="fig : " ] + and @xmath45 for various values of @xmath16 with the different panels corresponding to varying @xmath17 and @xmath26 as indicated under the respective plots .
the solid curve is the function @xmath45 while the dashed curves correspond to @xmath69 which are plotted for various values of @xmath16 ( which corresponds to the value of @xmath70 ) .
the lowest ( dot - dashed ) curve has the limiting value @xmath71 .
, title="fig : " ] and @xmath45 for various values of @xmath16 with the different panels corresponding to varying @xmath17 and @xmath26 as indicated under the respective plots .
the solid curve is the function @xmath45 while the dashed curves correspond to @xmath69 which are plotted for various values of @xmath16 ( which corresponds to the value of @xmath70 ) .
the lowest ( dot - dashed ) curve has the limiting value @xmath71 .
, title="fig : " ] ( 0,0 ) ( -13,10.7)@xmath72 ( -5.5,10.7)@xmath73 ( -8.,5)@xmath74 ( -12.5,-0.5)@xmath75 ( -5,-0.5)@xmath76 the temperature of any horizon is easily computed by noting that the euclidean metric is regular when we identify the thermal circle with period given by the inverse temperature @xmath77 for @xmath78 below , we will allow @xmath44 to range over intervals of the form @xmath79 $ ] where @xmath80 is a root of @xmath12 . as a result , @xmath81 at @xmath82 and our temperature corresponds to a unit normalized killing field at @xmath80 . before proceeding
further let us also record the boundary metric given by setting @xmath37 and stripping off two powers of a conformal factor , which we choose to be @xmath83 .
the result is @xmath84 this is the metric of a black hole with killing horizons at @xmath85 .
curiously , as we move away from the horizon in the range @xmath86 , where @xmath87 is the root of @xmath45 immediately to the right of @xmath88 , we find that the size of the euclidean time circle grows while that of the spatial @xmath89 circle shrinks .
in particular , @xmath90 , @xmath91 , indicating that the @xmath89 circle always has finite size at the horizon .
the ads c - metric solution can be physically visualized in terms of a black hole pulled by a cosmic string which accelerates it .
once one identifies the location of the cosmic string one can ask whether it hits the boundary .
the place where it hits the boundary is a conical defect on the boundary and ideally one would like to hide this behind the black hole horizon or to choose the period of @xmath89 so that the defect disappears .
one also wishes to ensure that the bulk curvature singularities at @xmath92 are likewise hidden behind horizons . to probe this issue we need to consider all possible coordinate regions .
we will do so in below , but let us first record two important facts which will play a crucial role in our analysis . by examining the behavior of the spatial part of the metric near a simple root of @xmath45 ,
say @xmath87 , we learn that the spacetime will be regular provided we identify the coordinate @xmath89 with period @xmath93 secondly , depending on the range of the coordinates we choose , we will be interested in the proper distance between interesting points such as the event horizon and the location of the cosmic string on the boundary .
this is easy to compute using the induced boundary metric and it is straightforward to see that for @xmath94 $ ] one has @xmath95 as we scan through the parameter space of the ads c - metrics it is possible to encounter degenerate roots of the function @xmath45 . in that case one encounters a new spatial infinity , for near a double root @xmath47 of @xmath45 the spatial part of the metric reduces to @xmath96 which is the metric on a euclidean hyperboloid @xmath97 .
this situation arises when we take @xmath28 and @xmath98 and was examined in some detail in @xcite .
due to the presence of this internal infinity , geodesics approaching @xmath47 are complete and one has a well defined asymptotic region , viz .
, @xmath9 .
this was useful in the analysis of @xcite since one could disentangle the physics of the black hole horizon from any curved spacetime effects associated with lack of spatial asymptopia .
below , we will investigate all the possible situations that arise from the ads c - metric in some detail .
we now proceed to identify black funnels and black droplets in the c - metric spacetimes . we will find it convenient to separate the discussion into various cases determined by the discrete parameter @xmath26 , and into sub - cases depending on the range we allow for the coordinates @xmath44 and @xmath50 .
as we will see , for a given c - metric there are in general several different interesting coordinate domains to consider . in each case
, we analyze the situation for all @xmath21 and @xmath25 .
the special limit @xmath23 is treated separately in .
now , in @xcite , droplets and funnels were primarily distinguished by their behavior with respect to the boundary s asymptotic region .
black droplets had @xmath6-compact horizons ( by which we mean compact with respect to the conformally rescaled metric with the chosen boundary value @xmath3 ) , and so were well separated from the asymptopia of @xmath3 .
black droplets were also typically suspended above a second horizon which did not connect to the boundary , though this was a secondary feature .
in contrast , black funnel horizons were non - compact , and extended into the bulk region associated with the boundary @xmath3 asymptopia ; see . specifically , @xcite required black funnels to asymptote to the bulk black hole solution known to describe a deconfined plasma in the asymptotic region ; e.g. , a planar ads black hole for asymptotically flat boundaries or the hyperbolic ( aka topological ) black hole of @xcite for boundaries which approach @xmath99 .
below , most of our boundary metrics will describe spatially compact universes , with no useful asymptotic regions . as a result
, the definitions of black funnels and droplets given in @xcite do not apply and must be generalized .
it is not clear to us what is the right definition in the broadest possible setting , or even whether a sharp distinction between droplets and funnels would remain possible .
however , all the geometries we study below possess a rotational killing field @xmath100 which commutes with the static killing field @xmath101 . furthermore
, each solution contains two special loci defined by @xmath100 .
at least one of these ( but possibly both ) is an axis ( fixed point set ) of @xmath100 corresponding to a root @xmath102 or @xmath49 of @xmath45 where the norm vanishes . in the one - axis case ,
the other is the singularity at @xmath103 , at which the norm of @xmath100 diverges .
this structure may be used to give a useful definition as follows .
first , in somewhat of an abuse of language , we refer to any horizon which does not reach the boundary as a _
planar black hole _ , no matter what the geometry or topology . in our examples
below , these will vaguely resemble the planar black hole of ( b ) .
it then remains only to classify horizons which connect to the boundary .
second , we remark that we are interested only in the so - called outer horizons , which by definition are horizons visible from the static region of the boundary . in some cases ,
the boundary metric will have two static regions and we will be forced to first choose a particular such region as a reference point .
now , since the horizons lie at constant @xmath50 and the axes / singularities of @xmath100 lie at constant @xmath44 ( and since @xmath44 and @xmath50 are independent ) , each outer horizon will intersect either one or both of the axes / singularities of @xmath100 .
the case of zero intersections does not arise below .
in addition , it is convenient that each horizon below with two such intersections fails to reach the boundary .
when @xmath104 .
in this case , the @xmath54 horizon might be said to develop a second intersection precisely at the boundary .
however , taking the limit @xmath23 carefully so as to maintain the ads radius @xmath105 fixed leads to a solution with only a single intersection for each horizon which reaches the boundary .
] as such , these are planar black holes in the sense described above and we need not consider this case further .
it remains to classify the cases with a single intersection , which we do as follows : * * black funnel : * when following a @xmath100 axis or singularity _ outward _ from the horizon leads toward the boundary , we call the horizon a black funnel . unless a second horizon is encountered , the axis or singularity then connects the horizon to the boundary through the visible static region . in such cases , an artistic impression of the spacetime resembles ( a ) , with the axis or singularity located near the edge of the diagram .
below , this is always an axis as the singularity at @xmath106 is always hidden by a horizon on the boundary . in some sense , the point where the axis reaches the boundary plays the role of an asymptotic region for the boundary metric . when black funnels appear below , only one axis or singularity of @xmath100 will be visible from the static region of the boundary .
* * black droplet : * when following a @xmath100 axis or singularity inward from the horizon leads toward the boundary , we call the horizon a black droplet . unless a second horizon is encountered , the axis or singularity then connects the horizon to the boundary through a hidden region behind the horizon . in the cases that arise below , we will always find a second separate @xmath100 axis or singularity that intersects the boundary outside the droplet horizon .
one may think that the first axis or singularity plays the role of the origin while the second axis or singularity plays the role of an asymptotic region in the boundary metric . in this sense , an artistic impression of the spacetime resembles ( b ) . in cases that appear below
, the first axis or singularity also extends `` below '' the droplet to intersect a planar black hole , a black funnel , or a new singularity .
as the reader will note , the above definitions are based on geometric features of horizons which generalize those of funnels , droplets , and planar black holes as defined in @xcite and illustrated in .
we nevertheless conjecture that , as for the original definition in @xcite , the funnel solutions are dual to field theory black holes coupling strongly to a deconfined plasma , while the droplets are dual to black holes coupling weakly .
some evidence for this is provided by the fact that all droplet solutions below which are free of naked singularities also include a second disconnected outer horizon .
this second horizon can be interpreted as describing the plasma , and the lack of connection as a sign of the weak coupling .
see for further discussion .
we begin our discussion with the case @xmath28 . for @xmath108
we find that @xmath45 has three real roots and for @xmath109 we obtain a single real root .
precisely for @xmath110 we have a degenerate root of @xmath45 ( this condition determines @xmath111 ) this was the situation discussed in detail in @xcite . as @xmath112 from below ,
the roots @xmath47 and @xmath48 approach each other . for larger values of @xmath17 these roots move off into the complex plane . now that we have understood the roots of @xmath12 , we can answer an important basic question : what restriction should we impose on the ranges of the coordinates @xmath44 and @xmath50 ? in order to maintain the correct lorentzian signature of the metric , we require @xmath113 . for @xmath114 , this means that to avoid unwanted boundaries at finite distance we should consider either the region @xmath115 or alternately @xmath116 .
the former is the conventional choice that has been made in previous analyses of the ads c - metric @xcite , but the latter is a perfectly reasonable coordinate domain as well
. for @xmath109 we must take @xmath117 .
we should also determine whether we restrict attention to the region @xmath118 or @xmath119 . to this end
it is useful to introduce a new coordinate @xmath120 which keeps track of the distance from the boundary @xmath37 , so that we choose either @xmath121 or @xmath122 .
it is worth noting that the coordinate @xmath123 introduced in is not the conventional feffereman - graham coordinate , which would instead be defined via the gauge choice @xmath124 and @xmath125 .
plane which characterize the distinct possibilities for the root structure of @xmath126 and @xmath127 for @xmath28 .
the behavior of @xmath127 is simply controlled by the parameter @xmath17 , while @xmath126 has non - trivial behavior across the various domains as indicated .
see main text for a detailed explanation . ] as a final piece of preparation , we describe the root structure of @xmath56 , which will in turn determine the horizons . recalling that the roots of @xmath11 and @xmath12 are ordered by ( with the exceptions noted in ) ,
it is easy to convince oneself that @xmath56 behaves as follows for @xmath34 : for @xmath128 we have @xmath129 for @xmath130 .
thus from we have a single real root at @xmath131 .
for @xmath132 we encounter a double root at the origin , which is degenerate and corresponds to the bulk poincar horizon . for @xmath133
we have three real roots of the function @xmath126 , two of which are negative .
in the special case when @xmath134 ( and @xmath135 ) , we have a double root at @xmath136 in addition to a single root at @xmath137 . for @xmath138
there is only one real root occurring at some @xmath139 . of course , this domain is non - empty only for @xmath109 .
note that , due to our convention that the root be called @xmath52 whenever @xmath11 has only one real root , the function @xmath140 is discontinuous at @xmath141 .
the single root for @xmath142 actually continuously connects to @xmath143 for @xmath144 .
this is the cause of the main exception to ( [ order ] ) noted in .
we are now in a position to list the various possible coordinate domains and to analyze each in turn as a function of @xmath16 and @xmath17 . for a complete illustration of the horizons in each domain , see . for various values of @xmath17 .
horizons ( diagonal lines ) are plotted in the @xmath145 plane .
note that @xmath123 increases downward while @xmath44 increases to the right .
the allowed regions are indicated by the roman numerals and can be considered as a complete spacetime unto themselves . to maintain the correct lorentz signature , the allowed regions are @xmath116 and @xmath146 $ ] respectively which are indicated by the numbers .
the different panels for a given value of @xmath17 correspond to situations with different numbers of roots for @xmath69 ; for a detailed behavior of the roots see .,title="fig : " ] + * region a - i : * @xmath147\ ; \ & \;z \le0 $ ] : this situation occurs for @xmath148 ( panels ( a ) , ( b ) , ( c ) , and ( d ) of ) . on the boundary
, we encounter a horizon at @xmath149 .
this boundary horizon is akin to the cosmological horizon in desitter space , with the @xmath89-circle pinching off as one moves outward to @xmath150 .
however , from the boundary perspective , the horizon shields a curvature singularity at @xmath103 .
in contrast , the @xmath106 singularity is visible through the bulk and in fact intersects the bulk @xmath151 horizon .
following the singularity away from @xmath52 through the static region takes us away from the boundary , so @xmath52 describes a ( singular ) black droplet .
the droplet does not reach the axis at @xmath47 , and this axis is visible from the boundary .
however , for all @xmath114 we can pick the period of the angle @xmath89 to be @xmath152 to avoid a conical singularity . in the domain @xmath153 where @xmath56 has three real roots ( panels ( a ) and ( c ) of ) , the @xmath52 droplet is suspended above a ( singular ) planar black hole horizon at @xmath154 ,
i.e. , we have two disconnected outer horizons .
note that there is a third horizon at @xmath155 which also reaches to @xmath156 .
however , this is an inner horizon since it is always hidden from the boundary observer . the special case @xmath157 ( panels ( c ) and ( d ) of ) where we encounter a double root at @xmath47 ( @xmath48 coalesces with @xmath47 in this limit ) was studied in some detail in @xcite . in this case
the spatial metric on the boundary is locally @xmath97 near @xmath150 and one has a well behaved asymptotic region of the boundary metric . since the axis has moved off to infinity , conical singularities do not arise for any choice of @xmath158 , though the curvature singularity at @xmath106 remains visible . for @xmath153 ( panel ( c ) )
we have a ( singular ) black droplet suspended over a ( singular ) hyperbolic ads black hole in the bulk , though the latter horizon disappears for @xmath128 ( panel ( d ) ) .
* region a - ii : * @xmath159\ ; \ & \ ; z\le 0 $ ] .
this situation similarly arises when @xmath148 ( panels ( a ) - ( d ) ) .
since the @xmath89-circle shrinks as one moves away from the horizon , any static region of the boundary spacetime again vaguely resembles that of de sitter space . the case @xmath160 ( panels ( b ) and ( d ) ) is uninteresting , as it contains no horizons in either the bulk or boundary .
it has only a naked singularity at @xmath161 . for @xmath162 and @xmath114 ( panel ( a ) )
, the cubic @xmath56 has two roots which satisfy @xmath163 , indicating that there are two black hole horizons on the boundary .
this also leaves us with two static regions on the boundary : @xmath164 , or @xmath165 .
note that for a given choice of static region , only one horizon will be an outer horizon .
the singularity at @xmath161 is visible from the former static region , so we focus on the latter . from this perspective
, @xmath53 is an outer horizon and @xmath54 an inner horizon .
both horizons reach the axis at @xmath166 , so @xmath53 is a black funnel .
( curiously , the inner horizon of the black funnel would in fact look like a droplet from the other static region s perspective . )
the @xmath49 axis is hidden behind the horizon , so we avoid all naked singularities by taking @xmath167 .
the @xmath53 and @xmath54 horizons merge to form a smooth extreme horizon when @xmath132 . the spacetime is similar in the limit @xmath112 ( panel ( c ) ) , though the @xmath47 and @xmath48 axes merge and move off to infinite distance , creating a new asymptotic region near @xmath168 ; see @xcite for details . as a result , for @xmath153
no conical singularities are visible from the static region @xmath165 for any value of @xmath158 .
* region a - iii : * @xmath147\ ; \ & \ ; z \ge0 $ ] .
we again encounter this situation only for @xmath148 ( i.e. panels ( a ) - ( d ) ) .
we have a black hole on the boundary with @xmath169 , and we have a bulk horizon which starts from @xmath149 on the boundary and reaches the axis @xmath150 in the bulk .
this axis connects the horizon and boundary through a static region , so the @xmath52 horizon is a black funnel .
there are no visible curvature singularities , and the period of @xmath89 may be chosen to make the axis at @xmath47 regular .
the situation is similar when @xmath110 , though since @xmath47 now represents an asymptotic region there is no conical singularity for any value of @xmath158 .
* region a - iv : * @xmath159\ ; \ & \ ; z \ge 0 $ ] : the regime of parameter space where we encounter this possibility is as in * a - ii * given by @xmath148 .
the only difference is that we allow ourselves to consider a different range of the coordinate @xmath123 .
consider first @xmath114 .
for @xmath160 ( panel ( b ) ) there is only a planar black hole ; the boundary metric contains no horizon .
however , there are boundary black holes for @xmath170 ( panel ( a ) ) .
in fact , as in * a - ii * , the boundary has two static regions : @xmath164 , or @xmath165 . from the perspective of the former
, only @xmath54 is an outer horizon .
it forms a black funnel . since
only the @xmath49 axis is visible , the choice @xmath171 leaves no naked singularities . from the perspective of the latter region ,
the outer horizons are @xmath52 ( a planar black hole ) and @xmath53 ( a droplet ) .
since both axes are visible there is a naked conical singularity for any choice of @xmath158 , though all curvature singularities are hidden .
the @xmath53 and @xmath54 horizons merge to form a smooth extreme horizon as @xmath172 . the situation is similar for @xmath157 . for @xmath160 ( panel ( d ) )
there is only a planar black hole , but for @xmath173 ( panel ( c ) ) there are again two static regions of the boundary metric : @xmath164 , or @xmath165 .
in fact , from the perspective of the first static region , the situation is identical to that with @xmath114 .
the new asymptotic region at @xmath47 is visible only from the latter static region , where it replaces an axis and allows us to remove all conical singularities by choosing @xmath171 , so that no singularities are visible .
we have a black funnel at @xmath54 and a black droplet at @xmath53 , suspended above the planar horizon at @xmath52 .
as described in @xcite , this planar horizon asymptotes near @xmath47 to one of the hyperbolic black holes described in @xcite .
* region a - v : * @xmath174\ ; \ & $ ] @xmath175 : we encounter this possibility when @xmath109 since in that regime @xmath45 has a single root at some @xmath176 . from ( [ order ] ) , we see that the boundary always contains a black hole .
when @xmath11 has only one real root ( panel ( f ) ) , it describes a single bulk horizon at @xmath52 which intersects the singularity at @xmath103 .
it is a singular droplet .
the singularity at @xmath161 is also visible , as is the @xmath49 axis . for @xmath138
, @xmath11 has three roots ( panel ( e ) ) and there are two choices of static region on the boundary , @xmath177 and @xmath164 . from the perspective of the first , the @xmath52 horizon is a singular droplet suspended above a ( singular ) funnel at @xmath53 . from the perspective of the second ,
the @xmath54 horizon is again a singular droplet and , in addition , the @xmath49 axis is visible .
* region a - vi : * @xmath174\ ; \ & \ ; z \ge 0 $ ] : as in the preceding case , this pertains to @xmath109 .
when @xmath11 has a single real root ( panel ( f ) ) , the only horizon is @xmath52 .
it is a black funnel and the @xmath49 axis is visible . for @xmath138 , @xmath11 has three real roots ( panel ( e ) ) and we again find two static regions on the boundary . choosing @xmath177 , the @xmath53 horizon is a black droplet suspended above a black funnel at @xmath52 .
choosing @xmath164 , the @xmath54 horizon is a black funnel . in both cases
, setting @xmath171 is necessary and sufficient to avoid naked singularities .
we next consider the simple case @xmath75 , where it is clear that the only real root of @xmath179 is located at @xmath180 . likewise @xmath56 has a single real root at @xmath181 ,
so this is the only horizon . this root is non - degenerate for @xmath182 , but becomes triply degenerate at @xmath132 . from we
see that the boundary metric has a single static region @xmath183 , from which the @xmath49 axis is clearly visible .
however , @xmath52 can lie on either side of the origin , depending on the sign of @xmath16 . maintaining lorentz signature of
the metric requires that we restrict attention to @xmath117 .
the only choice for the coordinate range is whether we approach the boundary at @xmath37 from above or below ; both choices yielding the same boundary metric .
moreover , since there is only one root of the function @xmath45 , we can choose @xmath89 to have the correct period to get rid of the potential conical defect , i.e. , @xmath171 .
once again , the situation is analogous to that encountered in the static region of de sitter space : the spatial sections are compact , and the size of the @xmath89 circle decreases as one moves away from the horizon , shrinking to zero at the @xmath49-axis . .
same conventions are used here as in . ]
the various possible coordinate regions are as shown in and can be summarized as follows .
* region b - i : * @xmath184 @xmath175 .
we have a black hole horizon on the boundary which extends away towards large negative @xmath44 in the bulk , reaching the singularity at @xmath103 .
following the singularity away from the horizon through the static region one moves away from the boundary , so the horizon is a ( singular ) black droplet .
the @xmath49 axis and the singularity at @xmath161 are visible from the boundary .
* region b - ii : * @xmath184 @xmath185 .
the horizon is a black funnel . for @xmath171 ,
the only singularities occur at @xmath103 and @xmath186 .
both singularities are hidden behind the horizon .
finally , for @xmath76 one has only a single real root for @xmath45 at some @xmath187 .
this is clear from and the fact that now @xmath188 for all @xmath189 .
we must therefore allow the entire range @xmath117 .
whether to take @xmath190 or @xmath191 .
plane which characterize the distinct possibilities for the root structure of @xmath126 for @xmath32 .
note that the behavior of @xmath127 is universal ; it always has one positive real root . ]
once again we can analyze the behavior of @xmath56 and use to conclude that ( see for an illustration ) : for @xmath192 we have a single real root for some @xmath193 . for @xmath57 there is a degenerate root at the origin and a positive real root at @xmath194 . for @xmath195
there are three real roots , one of which is negative and the other two positive , which we order as @xmath196 . for @xmath197
, @xmath11 has a single root at @xmath198 and a double root at @xmath199 .
for @xmath200 we have a single negative real root at @xmath201 for @xmath56 . .
same conventions are used here as in .
the allowed regions are @xmath117 .
the left panel corresponds to the situation when @xmath56 has three real roots , while the right panel protrays the situations when it has just one , compare with . ]
note that , due to our convention that the root be called @xmath52 whenever @xmath11 has only one real root , the function @xmath140 is discontinuous at @xmath132 .
the single root for @xmath153 actually continuously connects to @xmath143 for @xmath202 .
we therefore find the following behaviors : * region c - i : * @xmath117 & @xmath175 .
the singularity at @xmath103 is always visible through the bulk . when @xmath11 has only one real root ( panel ( b ) of )
, we find a singular droplet .
the @xmath49 axis and the @xmath161 singularity are also visible .
when @xmath11 has three real roots ( panel ( a ) ) , we have a choice of static regions on the boundary . from the perspective of the region @xmath177 ,
the @xmath52 horizon is a singular droplet suspended above a ( singular ) funnel at @xmath53 .
however , the @xmath161 singularity and the @xmath49 axis are hidden . from the perspective of the region @xmath164 ,
the @xmath54 horizon is a singular droplet and the @xmath49 axis and @xmath161 singularity are visible .
* region c - ii : * @xmath117 & @xmath185 .
when @xmath11 has only one real root , there is a single horizon at @xmath52 .
it is a black funnel .
when @xmath11 has three real roots , we have a choice of static regions . from the perspective of the region @xmath177 ,
the @xmath53 horizon is a droplet suspended above a black funnel at @xmath52 . from the perspective of the region @xmath164 , the @xmath54 horizon is a black funnel . in all cases ,
the choice @xmath171 is necessary and sufficient to avoid naked singularities .
we conclude by computing the boundary stress tensor for the solutions described above . this will allow us to see whether we can interpret any region around the back hole as containing a thermal fluid , in which one might hope to in some sense separate the effects of hawking radiation from those of vacuum polarization .
to this end , we need to use either ( i ) an explicit coordinate transformation to fefferman - graham coordinates or ( ii ) an appropriate definition of the boundary and use the counter - term procedure to compute the stress tensor .
we will follow the latter strategy since it is more convenient to implement for our purposes .
the counter - term procedure outlined in @xcite adds a boundary term to the einstein - hilbert action so that the result provides a well - defined variational principle for asymptotically ads spacetimes @xcite .
the full action is @xmath203 here @xmath6 is the induced metric on the boundary , which we take to be a surface of constant @xmath204 .
similarly , @xmath205 is the extrinsic curvature , and @xmath206 the boundary ricci scalar .
it is important to note that this @xmath123 agrees with the fefferman - graham coordinate typically used in the holographic renormalization literature only to leading order , and not beyond .
variations of with respect to @xmath207 lead to the stress tensor - like object @xmath208 however , because the metric @xmath209 diverges on the boundary , some rescaling will be required to obtain the boundary stress tensor .
one notes that @xmath210 , where @xmath211 and @xmath212 is the inverse of the physical boundary metric .
since @xmath213 has conformal dimension five , we have @xmath214 , where @xmath215 is the physical stress tensor on the boundary . as a result ,
the stress tensor with lower components will have a single factor of @xmath216 : @xmath217 for the boundary metric we obtain @xmath218 \nonumber \\
t_{xx } & = & c\ , \mu\,\sqrt{1+\lambda } \ , \gamma_{xx}\ , \nonumber \\ t_{\phi\phi } & = & c\ , \frac { \mu}{\sqrt{1+\lambda } } \ , \gamma_{\phi\phi } \ , \left[f(x ) - 2 \ , g(x ) \right ] , \label{sten}\end{aligned}\ ] ] where we define a central charge @xmath219 , measuring the effective degrees of freedom . where @xmath0 indicates the number of m2-branes probing the singularity of the calabi - yau cone over the sasaki - einstein base .
] we have also simplified the expression using the fact that @xmath220 . in deriving this expression we used the relation to express the parameter @xmath15 in terms of the physical length scale @xmath105 .
note that the signs and factors of two flip between the @xmath221 and @xmath222 components as we pass from horizon to the axes , reflecting the symmetry between @xmath44 and @xmath50 in the euclidean - signature solution .
finally , since the boundary field theory is odd dimensional , there is no conformal anomaly and the stress tensor must be traceless .
happily , tracelessness follows from the relation @xmath223 . a more natural way to present our result which makes the traceleness obvious is given by @xmath224 in general ,
does not take the perfect fluid form @xmath225 due to vacuum polarization effects .
however , we note that does reduce to at any root of @xmath45 where @xmath101 is timelike ( @xmath226 ) ; i.e. , where an axis of @xmath100 intersects the static region of the boundary .
there we identify @xmath227 and @xmath228 .
this is an analogue of the fact that for @xmath110 and @xmath34 this occurs in the asymptotic region near @xmath150 , where we saw in @xcite that the c - metric solutions approach the hyperbolic black holes of @xcite and are dual to a thermal plasma .
note that for our funnel solutions any axis of the above type intersects the funnel horizon , while for droplet solutions such an axis can not intersect the droplet horizon . instead , in every droplet case it intersects either a naked singularity , a planar black hole , or a black funnel .
though there is no sharp argument , we take this as supporting our basic picture of funnels and planar black holes as describing plasmas near the above axes , while droplet horizons describe physics that is only weakly coupled to the plasma .
the ads c - metric has been the inspiration for many interesting solutions in the past , most notably the brane - world black holes constructed in @xcite and more recently the exact plasma ball solution of @xcite .
we have in this paper undertaken a detailed analysis of the static , uncharged ads c - metrics to infer the examples of black funnels and black droplet solutions hidden in this family .
this required a generalization of the definitions of funnel and droplet given in @xcite to boundary metrics describing spatially compact universes .
as in @xcite , we conjecture that such solutions correspond to states of strongly coupled cfts in black hole backgrounds , with the distinction of funnel vs. droplet corresponding to two different types of behavior for the field theory state .
black funnels appear to describe horizons in the boundary metric coupling strongly to field theory plasmas , while black droplets appear to describe weak such couplings . in the latter case ,
the weak coupling is signified by the fact that , for the cases without naked singularities , our droplets were always accompanied by a second disconnected horizon ( which we called a planar black hole ) . from the field theory perspective we interpret
the droplet itself as describing vacuum polarization around the horizon in the boundary metric , and we interpret the planar black hole as describing the plasma . the fact that these horizons do not meet in the bulk implies that they describe pieces of field theory physics that can be thought of as coupling very weakly in the large @xmath0 limit .
this is what allows such states to be stationary even though , as one may check , the two horizons always have different temperatures outside of the special @xmath23 limits described in .
the ads c - metric family contains a rich variety of such solutions .
the comprehensive set of all possibilities is summarized in , , and ( for @xmath229 , respectively ) , where distinct panels portray distinct ranges of @xmath17 and @xmath16 yielding qualitatively different behavior , while within individual panels distinct spacetimes are separated by dashed lines ( and labeled by roman numerals ) . indeed , with the definitions given in , for @xmath230 every outer horizon could be classified as a planar black hole , a black droplet , or a black funnel . for each choice of c - metric parameters ,
we have also identified the static regions from which all singularities are hidden behind horizons . in these cases
, the dual field theory states should be regular at least on the region of the boundary that lies in the given static region .
we also found many settings with naked singularities in the bulk , but for which both the boundary metric and the boundary stress tensor are smooth . while such solutions are likely to be dual to singular states of the conformal field theory , it would be interesting to understand whether such singularities could be resolved by stringy or quantum effects in the bulk or , more likely , by introducing either time - dependence or some deformation of the dual field theory ; e.g. , by some perturbation that causes the theory to confine at an energy scale high enough to hide the would - be bulk singularities .
as we have seen , generic values of the c - metric parameters describe boundary metrics which are spatially compact in a natural conformal frame .
one of the disadvantages of this feature is that in the absence of spatial asymptopia it is a - priori unclear how one can disentangle the physics of the black hole horizon from curved spacetime effects .
nevertheless , it is curious that the quasi - local stress tensor induced on the boundary reduces to a thermal perfect fluid form close to any axis ( fixed point locus of the spatial isometry @xmath231 ) , at least for certain choices of parameters . of course , for special values of parameters it is possible to ensure that the induced boundary metric has a well defined asymptotics , viz .
, a hyperbolic cylinder @xmath9 as studied in @xcite .
one may also introduce a new asymptotic region at any point @xmath232 by a change of conformal frame , though at least one of @xmath233 or @xmath234 will then diverge at @xmath232 . as a result , while this may be useful for studying asymptotically ads black holes on the boundary , it will not provide boundary black holes with other familiar asymptotic behaviors .
one of the interesting generalizations which we have not explored here is the case of rotating ads c - metrics .
an asymptotically flat rotating black hole spacetime does not have a hartle - hawking vacuum due to super - radiance effects @xcite .
for very similar reasons , it is hard to imagine rigidly rotating black funnels when the boundary has a well - defined asymptotic region .
this would seem to require the distant shoulders of the black funnel to in some sense rotate faster than the speed of light .
however , the situation is somewhat different in the spatially compact case or when the boundary metric itself is asymptotically ads .
the more general ads c - metrics found in @xcite do allow for rotation and it would be interesting to examine this issue in some detail . the related case of rotating btz boundary metrics will be studied in @xcite .
it is a pleasure to thank roberto emparan and rob myers for very interesting discussions .
veh and mr would like to thank the kitp for wonderful hospitality during the workshop
fundamental aspects of superstring theory " , as well as the pedro pascual benasque center of science and the aspen center for physics for excellent hospitality during the course of this project .
in addition , veh , dm and mr would like to thank the icts , tifr for hospitality during the monsoon workshop in string theory where this project was initiated . veh and mr are supported in part by stfc rolling grant and by the us national science foundation under the grant no .
nsf phy05 - 51164 .
dm was supported in part by the us national science foundation under grants phy05 - 55669 and phy08 - 55415 and by funds from the university of california .
in the main text , we have identified black funnels and black droplets in the ads c - metrics for @xmath25 . while taking @xmath71 with finite @xmath15 yields the flat - space c - metric
, one might wonder if new asymptotically ads metrics might be obtained by taking @xmath23 holding @xmath105 fixed instead . since this requires @xmath235 , we must also scale @xmath44 and @xmath50 to obtain a finite limiting metric .
this scaling means that we effectively zoom in on some point @xmath236 in @xmath237 space . for @xmath238 ,
one obtains only pieces of flat minkowski space .
we therefore focus on the case @xmath239 below .
note that for @xmath240 we have @xmath241 , so that the roots @xmath11 coincide in this limit with the ( @xmath16-independent ) roots of @xmath12 .
when @xmath232 is not a root of @xmath12 , the non - trivial scaling limit leads to pure ads space in a slightly twisted version of poincar coordinates .
however , the behavior is more interesting when @xmath232 is a root of @xmath12 . since triple roots of @xmath12
do not arise , there are only two cases to consider . as we will see , the scaling limit is independent of all parameters and cares only about the degree of the root @xmath232 .
@xmath232 is a single root of @xmath12 .
+ consider the scaling limit @xmath242 denoting @xmath243 , we find that the metric reduces to @xmath244 we take @xmath245 without loss of generality , using the freedom to redefine @xmath246 to change the sign of @xmath247 if necessary .
the result can then be simplified using the coordinate change @xmath248 , @xmath249 to yield @xmath250 where we have made some additional trivial coordinate changes .
this is the metric of in the ds@xmath251 slicing .
the boundary is at @xmath252 and we clearly see ds@xmath251 in static coordinates
. we can furthermore check that this solution has no stress tensor by passing to the fefferman - graham coordinate chart using the transformation @xmath253 to write the metric as @xmath254 the absence of odd powers in the small @xmath123 ( near boundary expansion ) implies that @xmath255 .
@xmath232 is a double root of @xmath12 .
+ this case requires @xmath34 and @xmath110 , and so was discussed in @xcite .
briefly , the limit @xmath256 leads to the simple metric @xmath257 we have a horizon at @xmath258 and infinities at @xmath259 , @xmath260 .
note that all singularities have disappeared .
the induced metric on the boundary becomes @xmath261 while there are no rotation axes , we can apply the definitions of black funnels and black droplets given in if replace the axes at @xmath262 by the infinities @xmath263 due to the symmetry under @xmath246 , there are only two distinct choices of coordinate domains . for @xmath264 ,
the horizon at @xmath265 is a black funnel . for @xmath266 ,
the horizon at @xmath267 is a ( non - compact version of ) a black droplet suspended above a planar black hole at @xmath268 . as for the case of a single root , the boundary stress tensor vanishes . to gain some perspective on this statement , recall from @xcite that ( [ adsclm1 m ] ) asymptotes near @xmath269 to the @xmath270 black hole of @xcite which
, although it describes a deconfined phase , also has @xmath271 .
the vanishing of @xmath272 is due to a precise cancelation between the stress - energy of the deconfined plasma in this state and the stress tensor induced by vacuum polarization .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' _ adv .
* 2 * ( 1998 ) 231252 , http://arxiv.org/abs/hep-th/9711200[arxiv:hep-th/9711200 ] . v.
e. hubeny , d. marolf , and m. rangamani , `` hawking radiation in large @xmath0 field theories at strong coupling , '' http://arxiv.org/abs/0908.2270[arxiv:0908.2270 [ hep - th ] ] .
e. witten , `` anti - de sitter space , thermal phase transition , and confinement in gauge theories , '' _ adv .
_ * 2 * ( 1998 ) 505532 , http://arxiv.org/abs/hep-th/9803131[arxiv:hep-th/9803131 ] .
e. witten , `` on string theory and black holes , '' http://dx.doi.org/10.1103/physrevd.44.314[_phys .
_ * d44 * ( 1991 ) 314324 ] .
r. emparan , g. t. horowitz , and r. c. myers , `` exact description of black holes on branes , '' _ jhep _ * 01 * ( 2000 ) 007 , http://arxiv.org/abs/hep-th/9911043[arxiv:hep-th/9911043 ] .
r. emparan , g. t. horowitz , and r. c. myers , `` exact description of black holes on branes .
ii : comparison with btz black holes and black strings , '' _ jhep _ * 01 * ( 2000 ) 021 , http://arxiv.org/abs/hep-th/9912135[arxiv:hep-th/9912135 ] .
r. emparan and g. milanesi , `` exact gravitational dual of a plasma ball , '' http://arxiv.org/abs/0905.4590[arxiv:0905.4590 [ hep - th ] ] .
m. anber and l. sorbo , `` new exact solutions on the randall - sundrum 2-brane : lumps of dark radiation and accelerated black holes , '' _ jhep _ * 0807 * ( 2008 ) 098 , http://arxiv.org/abs/0803.2242[arxiv:0803.2242 [ hep - th ] ] .
m. m. anber , `` ads4/cft3+gravity for accelerating conical singularities , '' _ jhep _ * 0811 * ( 2008 ) 026 , http://arxiv.org/abs/0809.2789[arxiv:0809.2789 [ hep - th ] ] .
j. f. plebanski and m. demianski , `` rotating , charged , and uniformly accelerating mass in general relativity , '' http://dx.doi.org/10.1016/0003-4916(76)90240-2[_ann .
* 98 * ( 1976 ) 98127 ] .
r. emparan , `` ads / cft duals of topological black holes and the entropy of zero - energy states , '' _ jhep _ * 06 * ( 1999 ) 036 , http://arxiv.org/abs/hep-th/9906040[arxiv:hep-th/9906040 ] .
o. j. c. dias and j. p. s. lemos , `` pair of accelerated black holes in anti - de sitter background : the ads c - metric , '' http://dx.doi.org/10.1103/physrevd.67.064001[_phys . rev .
_ * d67 * ( 2003 ) 064001 ] , http://arxiv.org/abs/hep-th/0210065[arxiv:hep-th/0210065 ] . v. balasubramanian and p. kraus , `` a stress tensor for anti - de sitter gravity , '' _ commun .
* 208 * ( 1999 ) 413428 , http://arxiv.org/abs/hep-th/9902121[arxiv:hep-th/9902121 ] .
i. papadimitriou and k. skenderis , `` thermodynamics of asymptotically locally ads spacetimes , '' _ jhep _ * 08 * ( 2005 ) 004 , http://arxiv.org/abs/hep-th/0505190[arxiv:hep-th/0505190 ] .
b. s. kay and r. m. wald , `` theorems on the uniqueness and thermal properties of stationary , nonsingular , quasifree states on space - times with a bifurcate killing horizon , '' http://dx.doi.org/10.1016/0370-1573(91)90015-e[_phys .
_ * 207 * ( 1991 ) 49136 ] .
v. e. hubeny , d. marolf , and m. rangamani , `` hawking radiation from ads black holes . '' to appear , 2009 .
|
we recently argued that the dynamics of strongly coupled field theories in black hole backgrounds is related via the ads / cft correspondence to two new classes of ads black hole solutions : black funnels , and black droplets suspended above a second disconnected horizon .
the funnel solutions are dual to black holes coupling strongly to a field theory plasma . in contrast , the droplet solutions describe black holes coupling only weakly .
we continue our investigation of these solutions and construct a wide variety of examples from the ads c - metric in four bulk spacetime dimensions .
the solutions we find are dual to field theories on spatially compact universes with killing horizons .
( 0,0)(0,0 ) ( 350 , 320)dcpt-09/59 ( 350,305)nsf - kitp-09 - 165
|
we study the one - to - many matching problem , also known as the hospitals / residents problem .
formally , this is given by a set of residents @xmath2 and a set of hospitals @xmath3 , where every hospital @xmath4 has a capacity @xmath5 .
every resident seeks to get matched to a hospital and every hospital @xmath4 seeks to get matched to @xmath8 residents .
moreover , every resident @xmath9 has a strict ranking @xmath10 over hospitals that are acceptable to @xmath11 and every hospital @xmath4 has a strict ranking @xmath12 over residents that are acceptable to @xmath4 .
the set of mutually acceptable pairs is given by @xmath13 .
thus our input is a bipartite graph @xmath0 and the preferences of a vertex are expressed as an ordered list of its neighbors , e.g. , @xmath14 denotes the preference @xmath15 , i.e. , @xmath16 prefers @xmath17 to @xmath18 .
a matching @xmath7 in @xmath19 is a subset of @xmath20 such that @xmath21 for each @xmath22 and @xmath23 for each @xmath24 , where @xmath25 .
, we assume @xmath26 if @xmath27 , whenever suitable . ] the goal is to compute an _
optimal _ matching in @xmath6 .
the usual definition of optimality in this setting has been _ stability _ @xcite .
a matching @xmath7 in @xmath6 is said to be stable if there is no resident - hospital pair @xmath28 that `` blocks '' @xmath7 .
we say a pair @xmath28 blocks @xmath7 if ( 1 ) either @xmath11 is unmatched in @xmath7 or @xmath11 prefers @xmath4 to @xmath29 and ( 2 ) either @xmath4 has less than @xmath8 partners in @xmath7 or @xmath4 prefers @xmath11 to its worst partner in @xmath30 .
the gale - shapley algorithm can be easily generalized to find a stable matching in @xmath0 as was shown by gale and shapley @xcite .
since a stable matching is a maximal matching in @xmath6 , its size is at least @xmath31 , where @xmath32 is a max - size matching in @xmath6 .
this bound can be tight as shown by the following simple example : let @xmath33 and @xmath34 where each hospital has capacity 1 and the edge set is @xmath35 @xmath36 .
the preferences are shown in the table below . here
the only stable matching ( red line ) is @xmath37 , which is of size @xmath38 .
however , the max - size matching ( dashed lines ) is @xmath39 , which is of size @xmath40 .
@xmath41 \(r ) at ( 0,.5 ) ; ( r ) at ( 0,-.5 ) ; \(h ) at ( 3,.5 ) ; ( h ) at ( 3,-.5 ) ; \(r ) to ( h ) ; ( r ) to ( h ) ; ( r ) to ( h ) ; it can be shown that all stable matchings have to match the same set of residents and every hospital gets matched to the same capacity in every stable matching .
this is popularly called the `` rural hospitals theorem '' @xcite .
more precisely , roth @xcite showed that not only is every hospital matched to the same number of residents in every stable matching , but moreover , every hospital that is not matched up to its capacity in some stable matching is actually matched to the same _ set of residents _ in any stable matching .
thus the usual notion of stability is very restrictive . from a social point of view
it seems desirable to have a higher number of residents matched to hospitals in order to keep fewer residents unemployed and guarantee sufficient staffing for hospitals .
the latter point particularly applies to rural hospitals that oftentimes face the problem of being understaffed with residents by the national resident matching program in the usa ( cf .
a similar point could be made for matching workers to firms or students to seminar seats .
this motivates relaxing the notion of `` absence of blocking edges '' to a weaker notion of stability so as to obtain matchings that are guaranteed to be significantly more than @xmath31 .
note that we do not want to wish to ignore the preferences of vertices and impose a max - size matching on them as such a way of assigning partners will be socially undesirable as it does not take into account the preferences of vertices .
thus we would like to use a notion of optimality that replaces the _ local _ stability notion of `` no blocking edges '' with a weaker notion that is democratic or _ globally _ stable and achieves more `` global good '' , i.e. , its size is always at least @xmath42 for some @xmath43 .
* popularity .
* popularity is such a notion .
the notion of popularity was introduced by grdenfors @xcite in the stable marriage problem , i.e. , the one - to - one setting ( each vertex has capacity 1 ) .
popular matchings have been well - studied here @xcite and we generalize them to the hospitals / residents setting now . in the one - to - one setting ,
given any two matchings @xmath44 in @xmath6 and a vertex @xmath16 , we say @xmath16 prefers @xmath45 to @xmath46 if @xmath16 prefers @xmath47 to @xmath48 , where we say `` @xmath49 null '' if @xmath16 is left unmatched in matching @xmath50 , for @xmath51 note that the null option is the least preferred state for any vertex @xmath16 in a matching .
we will use the function @xmath52 for any vertex @xmath16 and neighbors @xmath53 of @xmath16 that is defined as follows : @xmath52 is 1 if @xmath16 prefers @xmath17 to @xmath18 , it is @xmath54 if @xmath16 prefers @xmath18 to @xmath17 , else it is 0 ( i.e. , @xmath55 ) .
for any two matchings @xmath44 in @xmath6 and a resident @xmath11 , we define @xmath56 @xmath57 where as before , we adopt the convention that `` @xmath58 null '' for a resident @xmath11 left unmatched in @xmath50 , for @xmath59 .
we would now like to define @xmath60 for a hospital @xmath4 .
if @xmath4 is not matched up to its capacity in @xmath50 , we will say `` @xmath61 null '' with multiplicity equal to @xmath62 and the null state is the least preferred state for @xmath4 .
* voting by hospitals .
* when we compare the preference of a hospital @xmath4 with @xmath63 for @xmath64 versus @xmath65 ( where @xmath66 is the @xmath67-th ranked resident in @xmath4 s preference list ) , we would like @xmath4 s vote to capture the fact that @xmath4 is better - off by _ 3 residents _ in @xmath68 when compared to @xmath69 .
so while comparing one matching versus another , we allow a hospital to cast more than one vote .
let @xmath4 be any hospital and let @xmath70 be any subsets of the set of @xmath4 s neighbors where we add some occurrences of `` null '' to make @xmath70 sets of size @xmath8 .
we will view the sets @xmath71 and @xmath72 as arrays @xmath73,\ldots , s'_i[k]\rangle$ ] ( for @xmath51 ) , where @xmath74 .
the preference of hospital @xmath4 for @xmath68 versus @xmath69 , denoted by @xmath75 , is defined as follows : @xmath76}\sum_{i = 1}^k { \mathsf{vote}}_h(s'_0[i],s'_1[\sigma(i)]),\ ] ] where @xmath77 $ ] is the set of permutations on @xmath78 .
let @xmath79 .
so @xmath60 counts the number of votes by @xmath4 for @xmath80 versus @xmath81 when the sets @xmath82 and @xmath83 are being compared in the order that is most adversarial or _ negative _ for @xmath45 .
that is , this order @xmath84 $ ] of comparison between elements of @xmath85 and @xmath86 gives the least value for @xmath87 , where @xmath88 is the number of indices @xmath67 such that @xmath89 \succ_h s'_1[\sigma(i)]$ ] and @xmath90 is the number of indices @xmath67 such that @xmath89 \prec_h s'_1[\sigma(i)]$ ] .
for instance , when a hospital @xmath4 with @xmath63 compares two subsets @xmath91 and @xmath92 ( where @xmath66 is the @xmath67-th ranked resident in @xmath4 s preference list ) , we have @xmath93 since comparing the following pairs results in the least value of @xmath75 : this pairing is ( @xmath94 with @xmath95 ) , ( @xmath96 with @xmath97 ) , ( @xmath98 with @xmath99 ) .
this makes @xmath100 . while computing @xmath101 , the pairing would be ( @xmath97 with @xmath94 ) , ( @xmath99 with @xmath96 ) , ( @xmath95 with @xmath98 ) : then @xmath102 . for any two matchings @xmath45 and @xmath46 in @xmath6 , we compare them using the function @xmath103 defined below .
we say @xmath45 is at least as popular as @xmath46 if @xmath105 and @xmath45 is more popular than @xmath46 if @xmath106 .
if @xmath105 then for every hospital @xmath4 in @xmath3 : _ no matter in which order _ the elements of @xmath82 and @xmath83 are compared against each other by @xmath4 in the evaluation of @xmath60 when we sum up the total number of votes cast by all hospitals and residents , the votes for @xmath46 can _ never _ outnumber the votes for @xmath45 .
[ def : pop - match ] @xmath45 is a popular matching in @xmath107 if @xmath105 for all matchings @xmath46 in @xmath6 .
thus for a matching @xmath45 to be popular , it means that @xmath45 is at least as popular as every matching in @xmath6 , i.e. , there is no matching @xmath46 such that @xmath108 . if there exists a matching @xmath46 such that @xmath108 then this is taken as a certificate of _ unpopularity _ of @xmath45 .
it is not obvious whether popular matchings always exist in @xmath6 .
our definition of popularity may seem too strict and restrictive since for each hospital @xmath4 , we choose the most negative or adversarial ordering for @xmath109 versus @xmath110 while calculating @xmath60
. a more relaxed definition may be to order the sets @xmath82 and @xmath83 in increasing order of preference of @xmath4 and take @xmath111,\,s'_1[i])$ ] as @xmath4 s vote .
an even more relaxed definition may be to choose the most favorable or _ positive _ ordering for @xmath85 versus @xmath86 while calculating @xmath60 .
[ def : weakly - pop ] call a matching @xmath45 _ weakly popular _ if @xmath112 for all matchings @xmath46 in @xmath6 .
thus @xmath45 is a weakly popular matching if the sum of votes for @xmath45 is at least the sum of votes for any matching @xmath46 when each hospital @xmath4 compares @xmath109 versus @xmath110 in the ordering that is most favorable for @xmath45 .
thus `` weak popularity '' is a more relaxed notion than `` popularity '' and a max - size weakly popular matching could presumably be larger than a max - size popular matching .
* our results .
* we will show that every stable matching in @xmath0 is popular , thus our definition of popularity is a relaxation of stability .
this also shows that popular matchings always exist .
we will present a simple linear time algorithm for computing a max - size popular matching @xmath45 in @xmath6 and show that @xmath113 .
we also show that @xmath45 is more popular than every _
larger _ matching , i.e. , @xmath106 for any matching @xmath46 that is larger than @xmath45 .
thus @xmath45 is also a _ max - size weakly popular matching _ in @xmath6 as no matching @xmath46 larger than @xmath45 can be weakly popular due to the fact that @xmath106 .
akin to the rural hospitals theorem , we show that all max - size popular matchings have to match the same set of residents and every hospital gets matched to the same capacity in every max - size popular matching .
however every hospital that is not matched up to its capacity in some max - size popular matching does _ not _ need to be matched to the same set of residents in any max - size popular matching , which is in contrast to stable matchings @xcite .
* techniques . *
our algorithm is an adaptation of the 2-level gale - shapley algorithm from @xcite that finds a max - size popular matching in a stable marriage instance .
our main contribution here is to show that this algorithm computes a max - size popular matching in the hospitals / residents setting as well .
while the analysis of the 2-level gale - shapley algorithm in @xcite is based on a structural characterization of popular matchings ( from @xcite ) on forbidden alternating paths and alternating cycles , we will use linear programming to show the correctness of this algorithm here .
our linear programming techniques are based on a linear program that was used in @xcite to find a popular fractional matching in a bipartite graph with _ 1-sided preference lists_. the popular fractional matching polytope shown in @xcite was also used in @xcite to show the popularity of some half - integral matchings in the stable marriage problem .
we do not design a popular fractional matching polytope for the hospitals / residents problem instead we focus on the matching @xmath45 obtained via the 2-level gale - shapley algorithm and use linear programming on a certain weighted graph @xmath114 that we build here .
* background and related work . * the first algorithmic question studied in popular matchings was in the domain of 1-sided preference lists @xcite where it is only vertices on the left , who are _ agents _ , that have preferences ; the vertices on the right are _ objects _ and they have no preferences .
popular matchings need not always exist here , however fractional matchings that are popular always exist and can be computed in polynomial time via linear programming @xcite .
popular matchings always exist in any instance of the stable marriage problem with strict preference lists since every stable matching is popular @xcite .
efficient algorithms to find a max - size popular matching in a stable marriage instance are known @xcite and a subclass of max - size popular matchings called _ dominant matchings _ was studied in @xcite . while a max - size popular matching can be easily computed here , there is no polynomial time algorithm currently known for computing a min - cost popular matching when there is a cost function on the edge set .
a polynomial time algorithm was shown in @xcite to find a min - cost popular _ half - integral _ matching .
when preference lists admit ties , the problem of determining if the given instance admits a popular matching or not is np - hard @xcite .
the stable matching problem in a marriage instance has been extensively studied we refer to the books @xcite on this topic .
the problem of computing stable matchings or its variants in the hospitals / residents setting is also well - studied @xcite . the stable matching algorithm in the hospitals
/ residents problem has several real - world applications it is used to match residents to hospitals in canada @xcite and in the usa @xcite .
in this section we describe the `` 2-level gale - shapley algorithm '' ( earlier used in @xcite ) to compute a max - size popular matching in @xmath107 .
this algorithm works in the graph @xmath115 defined as follows : @xmath116 consists of two copies @xmath117 and @xmath94 of every resident @xmath11 in @xmath2 , i.e. , @xmath118 .
the set @xmath3 of hospitals in @xmath119 is the same as in @xmath6 and the edge set here is @xmath120 .
the preference list of @xmath66 ( for @xmath51 ) is exactly the same as the preference list of @xmath11 .
the elements in the set @xmath121 will be called _ level @xmath67 _ residents , for @xmath51 .
every @xmath24 prefers any level 1 neighbor to a level 0 neighbor : within the set of level @xmath67 neighbors ( for @xmath122 ) , @xmath4 s preference order is the same as its original preference order .
for instance , if a hospital @xmath4 has only 2 neighbors @xmath11 and @xmath123 in @xmath6 where @xmath124 , the preference order of @xmath4 in @xmath119 is : @xmath125 . at any point in time
, only one of @xmath117 and @xmath94 will be _ active _ in our algorithm . to begin with ,
all level 0 residents are active in our algorithm and all level 1 residents are inactive .
we keep a queue @xmath126 of all the active residents and they propose as in gale - shapley algorithm : * every active unmatched resident @xmath66 who has neighbors in @xmath119 proposes to its most preferred neighbor in @xmath119 * in case @xmath117 has no neighbors in @xmath119 , then @xmath117 becomes inactive and @xmath94 becomes active and it joins the queue @xmath126 .
_ input : @xmath127 ; output : a matching @xmath7 in @xmath119_[alg : max - size ] when a hospital @xmath4 receives a proposal from @xmath66 , the hospital @xmath4 accepts this offer .
if @xmath4 is now matched to more than @xmath8 partners then @xmath4 rejects its worst partner @xmath128 in the current matching and so @xmath128 will have to find a new partner now and it joins @xmath126 .
if @xmath4 is now matched to @xmath8 partners then we delete all edges @xmath129 from @xmath119 where @xmath130 is a neighbor in @xmath119 that is ranked worse than @xmath4 s worst partner in the current matching
so no such resident @xmath130 can propose to @xmath4 later on in the algorithm .
once @xmath126 becomes empty , the algorithm terminates . a description of our algorithm is given as algorithm [ alg : max - size ] .
let @xmath7 be the matching returned by this algorithm and let @xmath45 be the matching in @xmath6 that is obtained by projecting @xmath7 to the edge set of @xmath6 , i.e. , @xmath131 for every edge @xmath132 in @xmath7 , where @xmath133 .
we will prove that @xmath45 is a max - size popular matching in section [ sec : analysis ] .
in this section we show a sufficient condition for a matching @xmath134 in @xmath6 to be popular .
this is shown via a graph called @xmath135 : this is a bipartite graph constructed using @xmath134 such that @xmath134 gets mapped to a _ simple matching _
@xmath136 in @xmath135 , i.e. , @xmath137 for all vertices @xmath16 in @xmath135 .
the vertex set of @xmath135 includes @xmath138 where @xmath139 .
that is , for each hospital @xmath4 , there are @xmath8 many copies of @xmath4 in @xmath140 . for each edge
@xmath28 in @xmath6 such that @xmath141 , we will arbitrarily choose a distinct @xmath142 and call @xmath143 .
if @xmath4 was not fully matched in @xmath134 , i.e. , it has less than @xmath8 many partners in @xmath134 , then some @xmath144 s will be left unmatched in @xmath136 . * for each edge @xmath28 in @xmath6 such that @xmath145 , we will have edges @xmath146 in @xmath135 , for @xmath147 .
* for each edge @xmath28 in @xmath6 such that @xmath141 , we will have the edge @xmath148 in @xmath135 where @xmath143 .
thus for any edge @xmath149 , there are @xmath8 many copies of @xmath150 in @xmath119 : these are @xmath146 for @xmath147 . however for any edge @xmath151 , there is only _
one _ edge @xmath148 in @xmath135 where @xmath143 , in other words , for @xmath152 , the resident @xmath11 is not adjacent to hospital @xmath144 in @xmath135 .
there are also some new vertices called `` last resort neighbors '' in @xmath135 : for any @xmath153 , there is one such vertex @xmath154 and every vertex @xmath16 ranks @xmath154 at the bottom of its preference list .
the purpose of the vertex @xmath154 is to capture the state of @xmath16 being left unmatched in any matching so that every matching in @xmath6 gets mapped to an _
@xmath157-complete matching _ in @xmath135 , i.e. , one that matches all vertices in @xmath138 .
we will use these last resort neighbors to obtain an @xmath157-complete matching @xmath158 from @xmath136 .
@xmath159 thus every resident @xmath11 unmatched in @xmath134 gets matched to @xmath160 in @xmath158 and if a hospital @xmath4 was not fully matched in @xmath134 , then some @xmath144 s will be matched to their last resort neighbors in @xmath158 .
we now define edge weights in @xmath135 . * for any edge @xmath161 : the weight of edge @xmath150 is @xmath162 .
note that for each edge @xmath163 , we have @xmath164 and @xmath165 if and only if @xmath150 blocks @xmath134 . * for any edge @xmath166 : the weight of edge @xmath150 is @xmath167 .
thus @xmath168 if @xmath16 was matched in @xmath136 and @xmath169 otherwise ( in which case @xmath170 )
. observe that every edge @xmath171 satisfies @xmath172 .
thus the weight of the matching @xmath158 in @xmath135 is 0 .
we will show in theorem [ main - thm ] that if _ every _ @xmath157-complete matching in the graph @xmath135 has weight at most 0 , then @xmath134 is a popular matching in @xmath6 .
we will show later that the matching @xmath45 obtained from our algorithm satisfies this condition .
suppose not .
that is , let @xmath134 be an _ unpopular _ matching in @xmath6 .
so there exists a matching @xmath173 in @xmath6 such that @xmath174 .
we will show a realization @xmath175 of @xmath173 in @xmath135 such that @xmath175 is an @xmath157-complete matching .
that is , each edge @xmath176 for some @xmath177 such that @xmath178 is in @xmath135 and @xmath175 will also have some appropriate @xmath155 edges .
* for every edge @xmath179 do : set @xmath180 ; thus if @xmath181 then @xmath148 belongs to @xmath175 as well .
* for every @xmath11 that is matched in @xmath173 and @xmath182 , we have to decide the index @xmath177 such that @xmath183 where @xmath184 . in the evaluation of @xmath185 , while comparing the sets @xmath186 and @xmath187 : * * let @xmath188 be the resident that @xmath4 compares @xmath11 with .
so the matching @xmath158 contains the edge @xmath189 for some @xmath190 .
set @xmath191 .
* * if @xmath11 is compared with `` null '' by @xmath4 ( so @xmath4 is not fully matched in @xmath134 ) , then we set @xmath191 for some @xmath190 such that @xmath192 and @xmath193 is unmatched so far in @xmath175 . * for any vertex @xmath156 that is left unmatched in steps ( i)-(ii ) above , set @xmath194 .
we have @xmath198 from the definition of edge weights in @xmath135 . by grouping together for each @xmath4 , the edges @xmath146 for all @xmath199 and any possible @xmath200 edges
, we get the right side of eqn . .
crucially , eqn
. follows from how we constructed the matching @xmath175 : for each hospital @xmath4 , we have @xmath201 and so @xmath202 .
the total sum of all the terms @xmath203 for @xmath204 is @xmath205 .
since @xmath174 , it follows that the weight of @xmath175 is positive .
thus there is a matching with positive weight in @xmath135 that matches all vertices in @xmath138 .
this contradicts the fact that every @xmath157-complete matching in @xmath135 has weight at most 0 .
hence @xmath134 is a popular matching .
let @xmath206 be any stable matching in @xmath6 .
consider the graph @xmath207 : since @xmath206 has no blocking edge in @xmath6 , it follows that every edge @xmath150 in @xmath207 satisfies @xmath208 .
thus every matching in @xmath207 has weight at most 0 and so by theorem [ main - thm ] , we can conclude that @xmath206 is popular .
we will now use theorem [ main - thm ] to prove the popularity of the matching @xmath45 computed in section [ sec : hosp - res - algo ] .
we will construct the matchings @xmath209 and the graph @xmath114 corresponding to the matching @xmath45 as described in the early part of section [ sec : analysis ] .
our goal is to show that every @xmath157-complete matching in @xmath114 has weight at most 0 .
note that the matching @xmath210 has weight 0 in @xmath114 . in order
prove that every @xmath157-complete matching in the graph @xmath114 has weight at most 0 , we will partition the set @xmath2 into @xmath211 and the set @xmath140 into @xmath212 as follows .
initially @xmath213 .
for each edge @xmath214 do : recall that @xmath221 is the matching in the graph @xmath119 obtained at the end of the 2-level gale - shapley algorithm ( see algorithm [ alg : max - size ] ) and the projection of @xmath7 on to @xmath222 is @xmath45 .
the definition of the sets @xmath223 implies that @xmath224 . also add residents unmatched in @xmath225 to @xmath219 and hospitals unmatched in @xmath225 to @xmath217 .
thus we have @xmath226 and @xmath227 ( see fig . [ fig0:label ] ) . theorem [ thm : max - pop ] will show that the matching @xmath45 satisfies the condition of theorem [ main - thm ] and this will imply that @xmath45 is a popular matching in @xmath6 .
this proof is inspired by the proof in @xcite that shows the membership of certain half - integral matchings in the popular fractional matching polytope of a stable marriage instance .
let our primal lp be the max - weight @xmath157-complete matching problem in @xmath114 .
we want to show that the primal optimal value is at most 0 .
we will do this by showing a dual feasible solution with value 0 .
the primal lp is the following : @xmath228 where @xmath229 is the set of edges incident on @xmath16 in @xmath114 .
consider the following assignment of @xmath230-values for all @xmath156 : set @xmath232 for all @xmath16 unmatched in @xmath233 ( each such vertex is in @xmath234 ) and for the matched vertices @xmath16 in @xmath233 , we set @xmath230-values as follows : @xmath235 if @xmath236 and @xmath237 if @xmath238
. observe that inequality holds for all vertices @xmath156 .
this is because @xmath239 for all @xmath16 unmatched in @xmath233 ; similarly , for all @xmath16 matched in @xmath233 we have @xmath240 . in order to show inequality
, we will use claim [ lem : g - m - edges ] stated below ( its proof is given after the proof of theorem [ thm : max - pop ] ) .
* claim [ lem : g - m - edges ] ( i ) states that for every edge @xmath245 in @xmath114 , we have @xmath246 . since @xmath247 for all @xmath248 , inequality holds for all edges of @xmath114 in @xmath249 .
* claim [ lem : g - m - edges ] ( ii ) states that for every edge @xmath146 in @xmath250 , we have @xmath251 . since @xmath252 for all @xmath253 ( for @xmath254 ) , inequality holds for all edges of @xmath114 in @xmath250 . since @xmath255 for all edges @xmath150 in @xmath114 and we set @xmath235 for all vertices @xmath236
, inequality is satisfied for all edges of @xmath114 in @xmath256 .
thus inequality holds for all edges @xmath146 in @xmath114 and so these @xmath230-values are dual feasible . for every edge @xmath257 , we have @xmath258 and @xmath232 for vertices @xmath16 unmatched in @xmath233 .
hence it follows that @xmath259 .
so by weak duality , the primal optimal is at most 0 .
in other words , every matching in @xmath114 that matches all vertices in @xmath138 has weight at most 0 .
* proof of claim [ lem : g - m - edges ] .
* consider any edge @xmath241 in @xmath114 .
note that the matching @xmath45 does not contain the edge @xmath28 if it did , then @xmath114 would have only one copy of this edge , say @xmath148 , which being an edge of @xmath233 has to be in either @xmath260 or @xmath261 whereas we are given that @xmath241 .
the resident @xmath262 , i.e. , @xmath94 got activated in our algorithm and recall that every hospital prefers level 1 neighbors to level 0 neighbors in our algorithm .
so if @xmath94 had proposed to @xmath4 , then this offer would have been accepted since @xmath4 had at least one partner who was a level 0 resident ( since @xmath263 ) .
thus @xmath94 must have been accepted by a neighbor that @xmath11 prefers to @xmath4 .
hence @xmath11 prefers its partner in @xmath45 to @xmath4 , so @xmath264 .
the resident @xmath117 was rejected by all its neighbors ( including @xmath4 ) and became inactive .
when @xmath4 rejected @xmath117 , the hospital @xmath4 was matched to @xmath8 neighbors , each of which was preferred by @xmath4 to @xmath117 .
thereafter , @xmath4 may have received ( and accepted ) better offers from its neighbors and since @xmath263 , the hospital @xmath4 never received enough offers from level 1 neighbors to have all its partners as level 1 residents .
in particular , @xmath144 is matched to a level 0 neighbor that is preferred to @xmath117 .
thus @xmath144 prefers its neighbor in @xmath233 to @xmath11 , so @xmath265 .
so it follows that @xmath246 .
we will now show part ( ii ) of this lemma . in our algorithm , the preference order of each hospital @xmath4 ,
when restricted to level 0 neighbors , is its original preference order and similarly , its preference order when restricted to level 1 neighbors , is its original preference order .
thus for each edge @xmath146 in @xmath114 where @xmath243 , either ( 1 ) @xmath144 prefers its partner in @xmath233 to @xmath11 or vice - versa or ( 2 ) @xmath257 . in both cases , we have @xmath251 .
we need to show that @xmath45 is a max - size popular matching in @xmath6 and we now show that this follows quite easily from the proof of theorem [ thm : max - pop ] .
let @xmath173 be any matching in @xmath6 .
we can obtain a realization @xmath175 of the matching @xmath173 in @xmath114 that is absolutely analogous to how it was done in steps ( i)(iii ) in the proof of theorem [ main - thm ] .
thus @xmath175 is an @xmath157-complete matching in @xmath114 and @xmath266 . we know from theorem [ thm : max - pop ]
that @xmath267 .
suppose @xmath173 is a popular matching in @xmath6 .
then @xmath268 has to be @xmath269 , otherwise the popularity of @xmath173 is contradicted since @xmath270 implies that @xmath271 ( because @xmath266 ) .
so if @xmath173 is a popular matching in @xmath6 , then @xmath175 is an optimal solution to the maximum weight @xmath272)-complete matching problem in @xmath114 .
recall that this is the primal lp in the proof of theorem [ thm : max - pop ] .
we will use complementary slackness to prove the following lemma . consider the @xmath275-values assigned to vertices in @xmath138 in the proof of theorem [ thm : max - pop ] .
this is an _ optimal dual _ solution since its value is 0 which is the value of the optimal primal solution .
thus complementary slackness conditions have to hold for each edge in the optimal solution @xmath276 to the primal lp .
that is , for each edge @xmath277 , we have : @xmath278 let @xmath156 be a vertex such that @xmath273 , so @xmath232 . if @xmath16 is a resident , then @xmath279 ; observe that all of @xmath16 s neighbors are in @xmath220
otherwise @xmath16 would propose to its neighbor in @xmath217 who would accept its offer since any level 1 resident is preferred to its current partner ( who is a level 0 resident ) .
similarly , if @xmath16 is a hospital , then @xmath280 and all its neighbors are in @xmath216 ; otherwise @xmath16 would be adjacent to some resident @xmath281 and @xmath16 ( being unmatched ) would have accepted @xmath282 s proposal and @xmath283 would never have been activated in the 2-level gale - shapley algorithm . in both cases every edge @xmath284 that is incident on @xmath16 in @xmath114 is _ slack _ because @xmath285 : thus @xmath232 and @xmath286 while @xmath287 .
thus it follows from equation ( [ eqn : new ] ) that @xmath288 for @xmath289 .
since @xmath175 is @xmath157-complete , it follows that @xmath274 .
now it is immediate to see that @xmath45 is a max - size popular matching in @xmath6 .
let @xmath173 be any popular matching in @xmath6 .
consider the matching @xmath290 .
lemma [ lem : new - max - size ] implies that @xmath291 because every vertex @xmath16 left unmatched in @xmath233 has to be left unmatched in @xmath292 also . since @xmath293 and @xmath294
, we have @xmath295 .
as this holds for any popular matching @xmath173 in @xmath6 , we can conclude that @xmath45 is a max - size popular matching in @xmath6 .
it is known that the gale - shapley algorithm for the hospitals / residents problem can be implemented to run in linear time @xcite .
hence the 2-level gale - shapley algorithm also runs in linear time and we can conclude the following theorem .
lemma [ lem : max - size ] below shows that any matching @xmath173 larger than @xmath45 can not be weakly popular ( see definition [ def : weakly - pop ] ) as @xmath271 for such a matching .
this implies that @xmath45 is also a max - size weakly popular matching in @xmath6 .
if @xmath173 is a larger matching than @xmath45 , then some vertices left unmatched in @xmath45 have to be matched in @xmath173 .
thus for some @xmath156 such that @xmath297 , the matching @xmath175 contains an edge @xmath298 where @xmath289 . thus @xmath175 contains a _ slack _
edge @xmath284 since @xmath232 , @xmath286 while @xmath299 ( see the proof of lemma [ lem : new - max - size ] ) .
it now follows from equation ( [ eqn : new ] ) that @xmath175 can not be an optimal solution to the maximum weight @xmath272)-complete matching problem in @xmath114 .
thus @xmath270 , in other words , @xmath271 since @xmath266 .
_ the above result implies that the following interesting property : for any definition of popularity that is `` in between '' popularity and weak popularity , the size of a max - size popular matching is the same . to formalize the meaning of `` in between '' , consider the two relations on assignments @xmath300 and @xmath301 , where @xmath302 if @xmath303 and @xmath304 if @xmath305 , induced by popularity and weak popularity , respectively .
clearly , @xmath306 .
note that popular matchings and weakly popular matchings correspond to maximal elements of @xmath300 and @xmath301 , respectively . is a maximal element of a relation @xmath307 if for all elements @xmath46 we have : @xmath308 implies @xmath309 . ] our result that @xmath45 is also a max - size maximal element of @xmath301 implies that if @xmath307 is a relation on matchings ( induced by an alternative notion of popularity ) such that @xmath310 , then @xmath45 is also a max - size maximal element of @xmath307 . hence we can conclude the following proposition which even allows for different hospitals to compare sets of residents in different ways . the rural hospitals theorem for stable matchings @xcite does not necessarily hold for max - size popular matchings .
that is , a hospital that is not matched up to capacity in some max - size popular matching is not necessarily matched to the same set of residents in every max - size popular matching .
consider the instance @xmath19 with @xmath33 and @xmath311 and @xmath312 and @xmath313 .
the edge set is @xmath314 .
the preferences are shown in the table below .
the ( max - size ) popular matchings are @xmath315 ( in black ) and @xmath316 ( in red ) .
so @xmath317 is matched to a different resident in the two max - size popular matchings @xmath7 and @xmath318 .
note that @xmath318 is not stable , as @xmath28 is a blocking pair .
however lemma [ lem : rural - hosp ] holds here . such a result for max - size popular matchings in the one - to - one setting
was shown in @xcite .
our proof is based on linear programming and is different from the combinatorial proof in @xcite .
[ lem : rural - hosp ] let @xmath173 be a max - size popular matching in @xmath6
. then @xmath173 matches the same residents as @xmath45 and moreover , every hospital @xmath4 is matched in @xmath173 to the same capacity as it gets matched to in @xmath45 .
consider the realization @xmath175 of @xmath173 such that @xmath175 is an @xmath157-complete matching in @xmath114 and @xmath266 . since @xmath173 is popular , we know that @xmath175 has to include all the edges @xmath155 for vertices @xmath16 left unmatched in @xmath233 ( by lemma [ lem : new - max - size ] ) .
let @xmath290 .
every vertex in @xmath114 that is unmatched in @xmath233 is left unmatched in @xmath292 also .
we also have @xmath319 as both @xmath45 and @xmath173 are max - size popular matchings in @xmath6 .
hence @xmath292 and @xmath233 match the same vertices in @xmath114 , i.e. , @xmath173 matches the same residents as @xmath45 and every hospital @xmath4 is matched in @xmath173 to the same capacity as it gets matched to in @xmath45 .
we show a non - trivial lower bound on the size of a max - size popular matching in @xmath6 in lemma [ lem : size - s ] .
this proof is based on an analogous proof in @xcite which shows such a result in the one - to - one setting .
the size of the matching @xmath45 is exactly the same as @xmath233 .
consider the graph @xmath114 without last resort neighbors ( call this graph @xmath322 ) we will show that every augmenting path with respect to @xmath225 here has length at least 5 .
this will immediately imply that @xmath323 where @xmath324 is the size of the max - size matching in @xmath322 . the value @xmath325
since corresponding to any matching @xmath173 in @xmath6 , we have a matching @xmath292 in @xmath322 such that @xmath293 .
thus the size of a largest matching in @xmath322 is at least @xmath326 and so we get @xmath327 .
consider any augmenting path @xmath328 with respect to @xmath233 in the graph @xmath322 ( see fig .
[ fig3:label ] ) let the endpoints of @xmath328 be @xmath11 and @xmath144 . since these vertices are left unmatched in @xmath233 , it follows that @xmath262 and @xmath263 .
as seen in the proof of lemma [ lem : new - max - size ] , the vertex @xmath94 is adjacent only to vertices in @xmath220 in the graph @xmath322 and the unmatched vertex @xmath144 is adjacent only to vertices in @xmath216 in the graph @xmath322 .
every vertex in @xmath216 is matched in @xmath45 to a neighbor in @xmath217 and every vertex in @xmath220 is matched in @xmath45 to a neighbor in @xmath219 .
thus the shortest augmenting path with respect to @xmath233 has the following structure with respect to the sets in @xmath329 that its vertices belong to : @xmath219-@xmath220-@xmath219-@xmath217-@xmath216-@xmath217 , i.e. , its length is at least 5 .
lemma [ lem : min - size ] below states that a stable matching is a _ min - size _ popular matching in @xmath6 .
such a result in the one - to - one setting was shown in @xcite and our proof is inspired by their proof .
in fact , the proof of lemma [ lem : min - size ] will show that a stable matching is a _ min - size weakly popular matching _ in @xmath6 .
let @xmath173 be a matching in @xmath6 such that @xmath330 .
consider a realization @xmath175 of the matching @xmath173 in the graph @xmath207 as described in the proof of theorem [ main - thm ] such that @xmath175 is @xmath157-complete and @xmath331 . recall that @xmath206 is a stable matching in @xmath6
hence for each edge @xmath150 in @xmath207 , we have @xmath208 .
moreover , because @xmath330 , there is a resident @xmath11 that is matched to a genuine neighbor in @xmath206 , however @xmath175 contains the edge @xmath332 .
we have @xmath333 .
thus @xmath334 .
in other words , @xmath335 and @xmath173 is unpopular .
since @xmath206 is a popular matching in @xmath6 , it means that @xmath206 is a min - size popular matching in @xmath6 .
we would like to remark that the 2-level gale - shapley algorithm and our analysis extend to the _ many - to - many _ setting as well . here
every vertex @xmath16 in the instance @xmath336 has a capacity @xmath337 associated with it and seeks @xmath338 many partners in a matching .
thus a max - size popular matching in such an instance can also be computed in linear time .
_ acknowledgments . _
the first author wishes to thank larry samuelson for comments on the motivation for popular matchings .
the second author wishes to thank david manlove and bruno escoffier for asking her about popular matchings in the hospitals / residents problem .
|
our input is a bipartite graph @xmath0 where each vertex in @xmath1 has a preference list strictly ranking its neighbors .
the vertices in @xmath2 ( similarly , in @xmath3 ) are called _ residents _ ( resp . , _ hospitals _ ) : each resident seeks to be matched to a hospital while each hospital @xmath4 seeks @xmath5 many residents to be matched to it .
the gale - shapley algorithm computes a stable matching in @xmath6 in linear time .
we consider the problem of computing a _ popular _ matching in @xmath6 a matching @xmath7 is popular if @xmath7 can not lose an election to any other matching where vertices cast votes for one matching versus another .
our main contribution is to show that a max - size popular matching in @xmath6 can be computed by the _ 2-level gale - shapley _ algorithm in linear time .
this is a simple extension of the classical gale - shapley algorithm and we prove its correctness via linear programming .
|
the blazar cta102 ( z=1.037 ) shows a curved jet which exhibits apparent velocities up to 15.4c ( ( * ? ? ?
* lister et al . 2009 ) ) .
the analysis of single - dish light curves and 43@xmath1 vlbi observations of ( * ? ? ?
* hovatta et al . ( 2009 ) ) and ( * ? ? ?
* jorstad et al .
( 2005 ) ) yields bulk lorentz factors , @xmath2 between 15 and 17 and doppler factors , @xmath3 between 15 and 22 .
these results lead to the picture that cta102 harbors a highly relativistic jet .
the source underwent a major radio flare from millimetre to centimetre wavelength during spring 2006 .
we have analyzed the single - dish data available for this source and here we present results of our modeling of the spectral evolution during the radio flare .
for our analysis of the 2006 radio flare in cta102 we used radio / mm light curves spanning from 4.8@xmath1 to 340@xmath1 . in order to obtain simultaneous spectra , we interpolated the observed light curves and performed a spectral analysis . within this analysis we subtracted a quiescent spectrum from the obtained data and fitted a self - absorbed spectrum , defined by @xmath4\right\}$ ] , where @xmath5 is the flux density , @xmath6 is the frequency at which the opacity @xmath7 , and @xmath8 and @xmath9 are the spectral indices for the optically thick and optically thin parts of the spectrum , respectively . the turnover frequency , @xmath10 , and the turnover flux density , @xmath11 , can be calculated from the first and the second derivative of the synchrotron spectrum and they can be regarded as the characteristics of the spectrum .
the uncertainties on the derived spectral values were obtained by performing monte carlo simulations .
the derived spectral evolution showed a discrepancies from the standard evolution with in the shock - in - jet model ( ( * ? ? ?
* marscher & gear 1985 ) ) , visible as a second hump in the @xmath12 plane ( ( * ? ? ?
* fromm 2009 ) ) .
the shock - in - jet model assumes a power law relation between the turnover frequency , @xmath10 , and the turnover flux density , @xmath11 , where the exponent depends on the dominant energy loss mechanism ( compton , synchrotron and adiabatic ) and the evolution of the physical properties in the jet .
the model assumes a power law distribution for the relativistic electrons , @xmath13 and the evolution of the constant @xmath14 follows the relation @xmath15 , where @xmath16 is the radius of the jet .
furthermore the the evolution of the magnetic field can be expressed as @xmath17 , the evolution of the doppler factor is @xmath18 and of the jet radius , @xmath19 , where @xmath20 corresponds to the distance along the jet . applying this model to the derived spectral behaviour leads to the following results for the evolution of the physical quantities .
the evolution of the parameter @xmath21 ( jet opening ) could be an indication for a non - pressure matched jet and the behaviour of the power law index @xmath22 could be interpreted as a re - acceleration of relativistic particles . together with the variation in the magnetic field @xmath23 and the doppler factor @xmath24 our result could be associated with a traveling - standing shock interaction .
standing shocks are common features in non - pressure matched jets . at the position of the standing shock , the traveling shock will encounter a region of increased particle density and magnetic field strength and will generate , due to shock acceleration at the shock front , a higher emissivity ( ( * ? ? ?
* gmez et al . 1997 ) ) .
this increase in the emissivity could lead to the observed spectral behaviour .
a more detailed discussion will be presented elsewhere .
future work should include the analysis of available multi - frequency vlbi observations in combination with relativistic magneto - hyrdodynamic simulations which could confirm our scenario of a shock - shock interaction in cta102 .
|
we have found evidence for interaction between a standing and a traveling shock in the jet of the blazar cta 102 .
our result is based in the study of the spectral evolution of the turnover frequency - turnover flux density ( @xmath0 ) plane .
the radio / mm light curves were taken during a major radio outburst in april 2006 .
|
the nuclear structure far from the @xmath6-stability line has been studied very actively with the development of the new - generation radioactive - isotope beam techniques together with of the microscopic nuclear models applicable to drip - line nuclei carried out by the high performance computers .
the ground state properties and the dynamical properties such as low - energy excitation modes and giant resonances in the medium - mass to the heavier nuclei have been successfully described by the self - consistent mean - field theory or the nuclear density - functional theory ( dft ) @xcite .
the nuclear dft has been applied to the exotic modes of excitation in unstable nuclei @xcite , and developed towards description of nuclei in the whole chart . along this line ,
the self - consistent random - phase approximation ( rpa ) including the pairing correlation and the nuclear deformation has been recently developed by several groups @xcite .
presently , the small excitation energy of the @xmath7 state and large transition probability @xmath8 in @xmath9 mg has been discussed in connection with the breaking of the spherical magic number @xmath10 in neutron - rich systems @xcite .
the recent gyromagnetic - factor measurement of @xmath1al at ganil @xcite and the @xmath6-decay study of @xmath1 mg at nscl @xcite indicate that @xmath1al has a certain amount of @xmath11 intruder configuration .
the electric quadrupole moment ( @xmath12 moment ) representing the deviation from a sphere is directly related to the deformation property of the nucleus , and thus its investigation for neutron - rich nuclei at around @xmath10 is strongly desired both experimentally and theoretically @xcite .
quite recently , the @xmath12-moment measurement of @xmath1al has been performed at ganil @xcite . in order to investigate the ground - state @xmath12 moments of neutron - rich al isotopes in the vicinity of the island of inversion " @xcite ,
we carry out the particle - vibration coupling ( pvc ) calculation based on the skyrme density functional , on top of the self - consistent quasiparticle - rpa ( qrpa ) .
the article is organized as follows : in sec .
[ method ] , the method is explained . in sec .
[ results ] , we perform the numerical calculations and investigate the core polarization for the electric quadrupole moments in @xmath13al . sec . [ summary ]
contains the conclusions .
the nuclear hamiltonian of a pvc model @xcite on top of the skyrme - hartree - fock - bogoliubov ( hfb ) and qrpa is written as @xmath14 here @xmath15 is the quasiparticle energy obtained as a self - consistent solution of the skyrme - hfb equation , @xmath16 the quasiparticle creation and annihilation operators .
the nucleon creation operator @xmath17 is then represented using the quasiparticle wave functions as @xmath18 the phonon energy @xmath19 is a solution of the qrpa equation on top of the skyrme - hfb , and @xmath20 the phonon creation and annihilation operators .
we solve the skyrme - hfb+qrpa equations in the @xmath21-scheme .
details of the calculation scheme is given in ref .
@xcite .
let us now consider the change of the density @xmath22 due to the collective vibrations as @xmath22 @xmath23 @xmath24 .
the nuclear potential @xmath25 $ ] is accordingly changed as @xmath25 $ ] @xmath23 @xmath26 $ ] . to first order in the change of the density ,
the difference of the nuclear potential is evaluated to be @xmath27-u[\varrho(\boldsymbol{r})]= \int d\boldsymbol{r}^{\prime } \dfrac{\delta u[\varrho(\boldsymbol{r})]}{\delta \varrho(\boldsymbol{r}^{\prime } ) } \delta \varrho(\boldsymbol{r}^{\prime},t).\ ] ] then , the pvc hamiltonian has a form of @xmath28}{\delta \varrho(\boldsymbol{r}^{\prime } ) } \delta \varrho(\boldsymbol{r}^{\prime},t ) \hat{\psi}^{\dagger}(\boldsymbol{r } ) \hat{\psi}(\boldsymbol{r } ) .
\label{hami_pv}\ ] ] we introduce the vacuum defined by the product of the hfb vacuum and the qrpa vacuum ; @xmath29 the density variation @xmath30 can be written in a second quantized form using the qrpa modes as @xmath31,\ ] ] where @xmath32 is a transition density to the qrpa state @xmath33 . the @xmath34 and @xmath35 parts in @xmath36 are taken into account in the qrpa phonons @xcite .
consequently , the pvc hamiltonian in the leading order reads @xmath37}{\delta \varrho(\boldsymbol{r}^{\prime } ) } [ \delta \varrho_{\lambda}(\boldsymbol{r}^{\prime})\hat{b}^{\dagger}_{\lambda } + \delta \varrho^{*}_{\lambda}(\boldsymbol{r}^{\prime})\hat{b}_{\lambda } ] \notag \\ & \times [ \varphi_{1,i}(\boldsymbol{r})\varphi^{*}_{1,j}(\boldsymbol{r } ) -\varphi_{2,i}(\boldsymbol{r})\varphi^{*}_{2,j}(\boldsymbol{r})]\beta^{\dagger}_{i}\beta_{j}. \label{hami_pv3}\end{aligned}\ ] ] the higher order effects can be treated systematically in the nuclear field theory @xcite .
the coupling interaction in eq .
( [ hami_pv3 ] ) is derived from the skyrme density functional . in the present calculation ,
we approximate the momentum dependent terms in the skyrme interaction by the landau - migdal ( lm ) form .
this approximation is made only for the construction of the pvc hamiltonian as in refs .
the isoscalar ( is ) and the isovector ( iv ) coupling interactions are expressed as @xmath38}{\delta \varrho(\boldsymbol{r}^{\prime } ) } \delta \varrho_{\lambda}(\boldsymbol{r}^{\prime } ) = \begin{cases } v^{\tau=0}(\boldsymbol{r})\delta \varrho_{\lambda}^{\mathrm{is}}(\boldsymbol{r}^{\prime } ) \delta(\boldsymbol{r}-\boldsymbol{r}^{\prime } ) \\
v^{\tau=1}(\boldsymbol{r})\delta \varrho_{\lambda}^{\mathrm{iv}}(\boldsymbol{r}^{\prime } ) \delta(\boldsymbol{r}-\boldsymbol{r}^{\prime})\tau_{z}\tau^{\prime}_{z}. \end{cases}\ ] ] the explicit expressions for @xmath39 and @xmath40 are given in ref .
@xcite .
in order to describe the odd-@xmath41 nuclear systems , we diagonalize the hamiltonian ( [ hami ] ) within the subspace @xmath42 . then , the resulting state vector is written as @xmath43 the operator for the quadrupole moment can be written as @xmath44 where @xmath45 , \\ q_{\lambda}&=\langle
0|[\hat{b}_{\lambda},\hat{q}]|0\rangle = \int d\boldsymbol{r } ( 3z^{2}-r^{2})\delta\varrho_{\lambda}^{\pi}(\boldsymbol{r}),\end{aligned}\ ] ] and @xmath46 is the vacuum expectation value .
the electric @xmath12 moment of the eigenstate @xmath47 is then calculated as @xmath48 \notag \\ & + \sum_{\lambda , jk}c^{1}_{\lambda j}c^{1}_{\lambda k}q_{jk } \bigr\}. \label{q_mom}\end{aligned}\ ] ] we apply this model to odd-@xmath49 nuclei to calculate the proton polarization charge of the state @xmath50 for the @xmath12 moment , which is defined as @xmath51 where @xmath52 . for the mean - field hamiltonian
, we employ the skm * interaction @xcite in the present numerical applications .
we use the lattice mesh size @xmath53 fm and a box boundary condition at ( @xmath54 fm , @xmath55 fm ) .
the quasiparticle energy cutoff is chosen at @xmath56 mev and the quasiparticle states up to @xmath57 are included . the pairing strength parameter is determined so as to reproduce the experimental pairing gap for neutrons in @xmath58 mg ( @xmath59 mev ) obtained by the three - point formula @xcite . the strength @xmath60
mev @xmath61 for the mixed - type pairing interaction with the exponent of the density - dependence @xmath62 leads to the pairing gap @xmath63 mev in @xmath58 mg @xcite .
on top of the skyrme - hfb , we solve the qrpa equation within the space of the two - quasiparticle excitation of @xmath64 mev .
the momentum - dependent terms in the residual interaction are exactly treated .
the density - dependent landau parameters @xmath65 and @xmath66 for the pvc interaction are determined by the parameters of the skm * interaction . as shown in refs .
@xcite , the attraction of the lm interaction is stronger than that of the self - consistent interaction in which the momentum - dependent terms are treated exactly . therefore , we multiply the overall factor @xmath67 for the pvc interaction .
we use @xmath68 and 0.9 for comparison . accordingly
, the difference of the results can be considered as a theoretical uncertainty .
we describe the odd-@xmath49 neutron - rich al isotopes as a proton single - hole state coupled to the neighboring si isotopes because the pairing gaps of protons in si isotopes are zero .
we summarize here the ground state properties and the structure of quadrupole excitations in @xmath69si .
cccc & @xmath9si & @xmath58si & @xmath5si + @xmath70 ( mev ) & @xmath71 & @xmath72 & @xmath73 + @xmath74 ( mev ) & @xmath75 & @xmath76 & @xmath77 + @xmath78 ( mev ) & 1.56 & 1.67 & 1.94 + @xmath79 ( fm ) & 3.22 & 3.32 & 3.39 + @xmath80 ( fm ) & 3.10 & 3.13 & 3.16 + in table [ gs ] , the ground state properties are summarized .
the neutron - rich si isotopes under investigation are spherical although the calculated deformation parameters are not exactly zero ( @xmath81 in @xmath58si ) .
this is due to the artificial breaking of the spherical symmetry associated with the finite mesh size and the rectangular box , and thus it is considered as a numerical error .
the average pairing gaps of neutrons are finite , while those of protons are zero .
this indicates that the @xmath58si has a neutron @xmath11 configuration in its ground state .
the neutron occupation number of the @xmath82 orbital is 0.31 , 0.78 and 2.21 in @xmath69si and that of the @xmath83 orbital is 0.10 in @xmath5si .
the transition strengths are smeared by a lorentzian function with a width of @xmath84 mev . ]
figure [ si_response ] shows the response functions for the isoscalar ( is ) , isovector ( iv ) and proton quadrupole excitations .
we can see a prominent peak at around 3 mev in all of the isotopes under investigation .
the isoscalar transition strengths are @xmath85 and 1137 @xmath86 in @xmath69si , corresponding to 104 , 98 and 162 in weisskopf unit .
besides the low - lying collective @xmath87 state , we can see the giant quadrupole resonances ( gqr ) at around 20 mev and 30 mev for the is and iv excitations .
the @xmath7 state in @xmath9si is mainly generated by the @xmath88-shell configurations of neutron and proton .
the microscopic structure of the @xmath89 component is given by the two - quasiparticle excitations of @xmath90 with a weight of 0.04 , @xmath91 with 0.61 , and @xmath92 with 0.30 .
the microscopic structure of the @xmath93 and @xmath87 components is the same within the numerical accuracy as of the @xmath89 component because of the spherical symmetry .
the strength in the energy region @xmath94 mev exhausts 79.4% of the is energy - weighted sum rule ( ewsr ) value .
the is strength in the low energy region up to 10 mev exhausts 1.2% of ewsr . the iv strength is distributed in a wider energy range @xmath95 mev .
the summed strength in this energy region exhausts 75.2% of the iv - ewsr value . in @xmath58si
the neutron excitations into the @xmath4-shell become appreciable for the @xmath7 state .
the microscopic structure of the @xmath89 component is given by the @xmath91 excitation with a weight of 0.21 , @xmath96 with 0.10 , and @xmath92 with 0.63 .
the strength in the energy region @xmath94 mev exhausts 79.8% of the is - ewsr value , and the strength in the energy region @xmath95 mev exhausts 74.3% of the iv - ewsr value . the @xmath89 component of the @xmath7 state in @xmath5si is mainly generated by the @xmath96 excitation with a weight of 0.39 , @xmath97 with 0.08 , and @xmath92 with 0.47 .
the strength in the energy region @xmath94 mev exhausts 77.6% of the is - ewsr value , and the strength in the energy region @xmath95 mev exhausts 72.9% of the iv - ewsr value . due to the mixing of the is and iv modes in neutron - rich nuclei
, we can see an appreciable iv strength in the lower energy region @xmath98 mev .
the summed strength in this energy region exhausts 10.4% of the iv - ewsr value . for describing the @xmath99 state of @xmath13al
, we diagonalize the hamiltonian ( [ hami ] ) in the model space of the proton single - hole state of the @xmath3 orbital @xmath100 and the coupled states of @xmath101 , @xmath102 and @xmath103 . we take the qrpa states @xmath104 whose is or iv quadrupole transition strengths possessing greater than in 1 w.u .
the dimension of the hamiltonian ( [ hami ] ) is 225 for @xmath0al .
the @xmath99 state of @xmath105 is constructed mainly by the one hole state and the hole coupled to the @xmath7 state as @xmath106 the amplitude associated with the other components are smaller than 0.1 .
the coupled states in eq .
( [ 31al_wf ] ) correspond to the @xmath107 state in the @xmath108-scheme representation
. the ratios of the amplitudes are identical to those of the clebsch - gordan coefficients @xmath109 , @xmath110 and @xmath111 . the quadrupole moment of @xmath0al is then calculated using eq .
( [ q_mom ] ) as @xmath112 , where we use @xmath68 ( 0.9 ) . from this value , the polarization charge of eq .
( [ charge ] ) is calculated as @xmath113 .
the result of the calculation overestimates slightly the experimental value of the @xmath12 moment @xmath114 @xcite . for @xmath1al
, the dimension of the hamiltonian ( [ hami ] ) is 250 . as in @xmath0al , the wave function of @xmath1al
is written by mainly the one hole state with a weight ( the squared amplitude ) of 0.90 and the hole coupled to the @xmath7 state with a weight of 0.08 .
the calculated values of the @xmath12 moment and the polarization charge are @xmath115 and @xmath116 , respectively . the quadrupole moments and polarization charges of @xmath0al and @xmath1al are not very different .
the dimension of the hamiltonian ( [ hami ] ) is 285 for @xmath2al . as in the case of @xmath0al and @xmath1al ,
the wave function of @xmath2al is written by the one hole state with a weight of 0.85 and the hole coupled to the @xmath7 state with 0.14 .
the contribution of the coupling to the @xmath7 state is larger than in @xmath117al .
the calculated values of the @xmath12 moment and the polarization charge are @xmath118 and @xmath119 , respectively . in order to see separately the effects of coupling to the low - lying modes and to the giant resonances on the polarization charge , we diagonalize the hamiltonian ( [ hami ] ) in the model space containing only the rpa modes with their energies larger than 10 mev . the obtained polarization charges for the @xmath3 orbital are @xmath120 , @xmath121 and @xmath122 in @xmath0al , @xmath1al and @xmath2al , respectively .
these values are close to the systematic value obtained in ref .
@xcite ( @xmath123 and @xmath124 in @xmath13al ) , where the microscopic pvc calculations were performed including only the giant resonances on top of the self - consistent hf+rpa in light neutron - rich nuclei .
the enhancement of the polarization charge in @xmath2al is thus due to the strong collectivity of the low - lying quadrupole vibrational mode in the core nucleus @xmath5si because the effects of coupling to the giant resonances are not sensitive to the neutron number in al isotopes under investigation .
but for the rpa strengths in @xmath58si .
] we investigate the effects of coupling to the @xmath4-shell in @xmath1al .
the dominant correlation in the present case is the pairing correlation because the core nucleus @xmath58si is calculated to be spherical at the hfb level .
figure [ 34si_response ] shows the is , iv and proton quadrupole transition strengths in @xmath58si obtained by solving the skyrme hf+rpa equations without pairing correlations .
the collectivity of low - lying states are tremendously weakened , while the structure of gqr is not very different to that obtained by solving the skyrme hfb+qrpa equations shown in fig .
[ si_response ] .
the @xmath7 state is constructed dominantly by the @xmath92 excitation with a weight of 0.98 .
the strength @xmath125 has only 237 @xmath86 . using the rpa transition densities , we diagonalize the hamiltonian ( [ hami ] ) with a dimension of 241 .
the calculated wave function is mainly generated by the one hole state with a weight of 0.97 and the hole coupled to the @xmath7 state with a weight of 0.02 .
this state is thus dominantly described by the proton @xmath88-shell configurations because the @xmath7 state is generated by the proton excitation to the @xmath126 orbital .
the resulting @xmath12 moment is @xmath127 , and the polarization charge is @xmath128 . the @xmath12 moment and the polarization charge in @xmath1al are quite sensitive to the neutron pairing correlation at @xmath10 .
this is similar to the enhancement mechanism of the @xmath8 in @xmath9 mg because the neutron pairing correlation is indispensable for the strong collectivity of the @xmath7 state in @xmath9 mg @xcite .
but in @xmath9 mg .
] the wave function @xmath129al@xmath130 can also be constructed by the quasi - proton coupled to @xmath9 mg .
the pairing gaps of neutrons and protons are 1.80 and 1.42 mev .
figure [ 32mg_response ] shows the response functions for the is , iv and proton quadrupole excitations . at 0.9 mev
, we can see a prominent peak possessing 2414 @xmath86 and 430 @xmath131@xmath86 for the is quadrupole strength and @xmath132 , respectively .
this result agrees well the experimental value @xcite .
the @xmath7 state is constructed by the two - quasiparticle excitations of @xmath91 with a weight of 0.10 , @xmath96 with 0.09 , @xmath133 with 0.11 , and @xmath134 with 0.61 .
we diagonalize the hamiltonian ( [ hami ] ) within the space of the proton quasiparticle of the @xmath3 orbital , and the coupled states of a @xmath88-shell quasi - proton to the quadrupole modes in @xmath9 mg . in the present calculation with a dimension of 276 ,
@xmath129al@xmath130 is constructed by the quasi - proton of the @xmath3 level with a weight of 0.85 , @xmath135 with 0.10 , @xmath136 with 0.04 , and @xmath137 with 0.01 .
the electric @xmath12 moment of @xmath1al is then calculated as 12.4 ( 12.6 ) @xmath138@xmath139 .
this is consistent with the calculation in sec .
the polarization charges for the electric quadrupole moment of the neutron - rich al isotopes at around @xmath10 have been investigated by carrying out the microscopic particle - vibration coupling calculation in which the coordinate - space skyrme - hartree - fock - bogoliubov and quasiparticle - random - phase approximation are employed to calculate the single - quasiparticle wave functions and the transition densities .
it has been found that the neutron pairing correlations are crucial to generate collectivity of the @xmath7 state in @xmath58si , and that the polarization charge of the proton hole state of the @xmath3 orbital becomes small in the absence of the pairing correlation in @xmath1al .
the effect of the neutron pairing correlation at @xmath10 on the enhancement of the polarization charge in @xmath1al is very similar to the enhancement mechanism of the @xmath132 in @xmath9 mg @xcite .
effects of coupling to the giant resonances on the polarization charge are not very different in a small region of isotopes , and the low - lying collective modes have much effect on the polarization charge .
therefore , the polarization charge in @xmath2al is larger than in @xmath117al as a consequence of the stronger collectivity of the @xmath7 state in @xmath5si .
the author acknowledges k. matsuyanagi , t. nakatsukasa for valuable discussions and encouragement , and t. nagatomo , h. ueno for stimulating discussions .
he is supported by the special postdoctoral researcher program of riken .
the numerical calculations were performed on the nec sx-8 supercomputer at the yukawa institute for theoretical physics , kyoto university and the nec sx-8r supercomputer at the research center for nuclear physics , osaka university .
|
the core polarization effect for the electric quadrupole moment of the neutron - rich @xmath0al , @xmath1al and @xmath2al isotopes in the vicinity of the island of inversion are investigated by means of the microscopic particle - vibration coupling model in which the skyrme hartee - fock - bogoliubov and quasiparticle - random - phase approximation are used to calculate the single - quasiparticle wave functions and the excitation modes .
it is found that the polarization charge for the proton @xmath3 hole state in @xmath1al is quite sensitive to coupling to the neutrons in the @xmath4-shell associated with the pairing correlations , and that the polarization charge in @xmath2al becomes larger due to the stronger collectivity of the low - lying quadrupole vibrational mode in the neighboring @xmath5si nucleus .
|
the theory of hard elastic scattering in quantum chromodynamics ( qcd ) has evolved considerably over many years of work .
currently there exist two self - consistent perturbative descriptions , each with a specific factorization method for separating the hard scattering from non - perturbative wave functions . a well - known procedure using the `` quark - counting '' diagrams has been given by lepage and brodsky @xcite .
a consequence , and direct test , of the factorization defining this mechanism is the hadron helicity conservation rule @xcite @xmath4 where the @xmath5 s are the helicities of the participating hadrons in the reaction @xmath6 .
the fact that this rule is badly violated in almost every case tested suggests two alternatives .
one possibility , advocated by isgur and llewellyn - smith @xcite , is that the energy and momentum transfer ( @xmath0 ) in the data is not large enough for the formalism to apply .
however , the data also shows behavior close to the model s predicted power dependence on @xmath0 , indicating that hard scattering of a few pointlike quarks is being observed .
the apparent contradiction has led to much discussion , and has even caused some authors to suggest that perturbative qcd itself might be wrong .
the second alternative is that another power behaved process causing helicity flip is present . in fact
the `` independent scattering '' subprocess , introduced by landshoff @xcite , is actually the leading process at very high energies @xcite .
but it has been assumed that hadron helicity conservation would be the same in the independent scattering process as in the quark - counting one , since both involve exchange of hard gluons with large @xmath0 . in general
, terms proportional to a quark mass , @xmath7 , for example , have been understood to cause helicity flip in either model , but with amplitude suppressed by a power of @xmath8 relative to the leading term .
such terms seem to be quite small and are probably not a believable explanation of the persistent pattern of large violations of the helicity conservation rule . here
we show that the independent scattering mechanism predicts high - energy helicity _ non - conservation _ at a much higher rate .
the calculations in momentum space are sufficiently complicated that this phenomenon has been overlooked for almost twenty years . adopting a transverse position space formalism introduced by botts and sterman @xcite ,
we show that the details rest on non - perturbative wave functions that should be _ measured _ rather than calculated .
these wave functions measure non
zero orbital angular momentum not taken into account by short distance expansions .
we argue that the novel factorization properties of independent scattering processes can not practically be reduced to the same ingredients used in the quark counting scattering . in any case , it is not necessary to flip a quark helicity : the new mechanism proceeds unimpeded in the limit of arbitrarily small quark mass and perfect chiral symmetry in the hard scattering .
this paper is organized as follows . in section 2
we review the derivation of helicity conservation in genuine short distance processes .. in section 3 , we present the independent scattering mechanism with special emphasis on non - zero orbital angular momentum wave functions .
we compute the contribution of these components to a helicity conserving reaction in section 4 , then to a helicity violating reaction in section 5 .
these two sections contain our most important results at asymptotic and at accessible energies .
section 6 is an experimental outlook .
first we review the conventional derivation of hadron helicity conservation @xcite .
the quark - counting factorization introduces the distribution amplitude @xmath9 @xcite , an integral over the transverse momentum variables of the wave function for quarks to be found carrying momentum fraction @xmath10 in the hadron .
introduced for renormalization group analysis in ref .
either convention can be used without affecting our argument . ] for simplicity of presentation we specialize to a single pair of quarks , the meson case .
let @xmath11 be a light cone wave function to find quarks with relative transverse momentum @xmath12 and light cone momentum fraction @xmath10 .
in terms of the fourier conjugate transverse space variable @xmath13 separating the quarks , then @xmath14 in the second line we have expanded the wave function @xmath15 to exhibit the @xmath16 orbital angular momentum eigenvalues @xmath17 : a complete set for the ( nearly on - shell ) quarks consists of the @xmath18-axis orbital angular momentum , the energy , and the @xmath18-momentum .
now suppose the distribution amplitude @xmath9 is assumed to be a good description of a process .
then , whatever the angular momentum of the wave function , evaluating the integrals reveals the @xmath19 element is the sole surviving term in eq .
( [ orbital ] ) : @xmath20 this shows that use of @xmath9 imposes two things : as @xmath21 the scattering region is both `` small '' , since the region @xmath22 dominates in the bessel function , and the scattering is `` round '' , _ i.e. _ cylindrically symmetric ( fig . 1 ) . in the absence of orbital angular momentum
, the hadron helicity becomes the sum of the quark helicities .
the quark helicities being conserved at leading order , the total hadron helicities are conserved . the hadron helicity conservation rule ( 1 ) therefore represents an exact _ symmetry _ of the quark - counting factorization .
a crucial question is : does this symmetry of the model represent a property of the entire perturbative theory . or , can we simply assume `` s wave '' @xmath16 wave functions to be the main contribution as in a non relativistic picture ?
the answer to both questions is _ no_. in general , quark wave functions themselves are not particularly restricted in orbital angular momentum content , even in the high energy limit .
for example , in the pion ( pseudoscalar meson ) case the light - cone wave function may be expanded on four dirac tensors allowed by parity symmetry as @xmath23\gamma_5\right\}_{\alpha\beta},\ ] ] where @xmath24-@xmath25 are functions of the light cone fraction @xmath10 , the transverse separation @xmath13 and the total four momentum of the meson @xmath26 .
for the moment we do not discuss dependence on gauge fixing and a path - ordered exponential of the gauge potential used to make the wave function gauge invariant .
the @xmath25-term carries one unit of orbital angular momentum and yet scales with the same power of the `` big '' momentum @xmath27 as the @xmath24term , which is s - wave .
since the @xmath25 term has a @xmath28 factor , which can be written in terms of @xmath29 this term represents one unit of orbital angular mometum . in terms of power counting , then , the @xmath30 and @xmath31 amplitudes can be equally large .
we also note that wave functions are not objects to be derived in perturbation theory , but instead represent the long - time , non - perturbative time evolution proceeding inside a hadron .
the non - perturbative hamiltonian of qcd does not conserve spin and orbital angular momentum separately , but instead generates mixing between orbital and spin angular momentum .
finally , there is no simple relation between `` s wave '' non relativistic models of constituent quarks , and the pointlike quarks resolved in large@xmath0 collisions ,
so no statement can reliably be made about quark angular momentum content of hadrons .
thus _ if _ a non - zero orbital angular momentum component somehow enters the hard scattering and this is a crucial point _ then _ the long - time evolution before or after the scattering can convert this angular momentum into the observed hadron spin .
it is not necessary to flip a quark spin in the hard interaction , because the asymptotic hadron spin fails to equal the sum of the quark spins .
such a mechanism is totally consistent with the impulse approximation of perturbative qcd .
the challenge in high energy hadron scattering is therefore to find those large @xmath0 process(es ) in which non - zero orbital angular momentum enters , or in other words , to find those which are not `` round '' .
it turns out that in any treatment relevant to current energies the independent scattering process is not `` round '' but instead `` flat '' ( fig . 1 ) .
the subprocess is highly asymmetric , showing an extreme dependence on the scattering plane . rather than disappearing in the high energy limit ,
the dynamics of this process _ increases its asymmetry over a range of _ @xmath0 .
the origin of the asymmetry is kinematic .
let @xmath32 be the direction perpendicular to the scattering plane and @xmath33 be a vector in the scattering plane . in the independent scattering mechanism ,
let two ( or more ) uncorrelated scattering planes be separated at the collision point by a transverse out of plane distance @xmath34 .
going beams of quarks must coincide in direction well enough to make hadrons in the final state . conserving 4-momenta , for each pair of scatterings there are three delta functions of large momenta ( scaling like @xmath27 ) , and one delta function of the out - of plane transverse momentum , of the order of @xmath35 , where @xmath36 indicates a typical expectation value in the state .
on shell quark
quark scattering is dimensionless and scale invariant ( up to logarithmic corrections ) in qcd .
the overall amplitude therefore scales like the product of these factors , namely like @xmath37 .
the @xmath0 power - counting of this argument is quite well - known ; what we emphasise here is the role of the scattering plane , namely the breaking of rotational symmetry with the out - of - plane direction @xmath32 .
it is as if hadrons `` flatten '' under impact in the in - plane direction @xmath33 , forming a cigar - shaped hard scattering region @xcite .
this fact is very hard to see in covariant perturbation theory in momentum space , explaining why it has generally been overlooked . the kinematic violation of hadron helicity conservation by independent scattering raises several new questions .
it is clear that the usual association of leading twist ( short distance ) and large @xmath0 either breaks down or hinges on delicate dynamical details .
our approach will exploit the fact that leading approximations to any kinematically distinct amplitude are always perturbatively calculable .
for example , higher order corrections of non leading twist type in the distibution amplitude formalism can not violate the hadron helicity conservation symmetry and will not affect our approach .
the first non vanishing contributions to helicity violating amplitudes involve extra partons in the short distance formalism .
a gluon embedded in the hard scattering , for example , could transfer spin to an outgoing hadron .
we need not consider such processes , because , as discussed later , they are subleading by a power of @xmath0 and perturbatively small since such gluons are `` hard '' .
it remains to be shown , of course , that helicity violation from independent scattering is not suppressed by the same order .
that is the main technical task of this paper .
botts and sterman have considered @xcite the generic `` elastic '' reaction @xmath38 , where @xmath39 s are light pseudoscalar mesons , at high energy @xmath40 and large angle center of mass scattering angle @xmath41 .
their study has shown that the reaction is dominated by the 2 independent scatterings of the valence constituents with a kinematical configuration depicted in fig . 2 .
one has two scattering planes separated at the collision point by a transverse distance @xmath42 . to make it clear , let us consider a light cone basis @xmath43 attached to each meson @xmath39 , and chosen so that , in the center of mass frame where @xmath44 and @xmath45 , @xmath46 in the cm frame , one has @xmath47 ( @xmath48 ) .
each bethe - salpeter amplitude @xmath49 @xmath50 is assumed to eliminate @xmath51 configuration with relative @xmath52 transverse momentum ( along @xmath53 and @xmath32 ) . the relative minus momentum ( along @xmath54 )
is of @xmath55 .
then , momentum transfers in @xmath56 or @xmath57 are dominated by large invariants built with @xmath58-terms and a first approximation is to neglect all but these components of the quark or antiquark momenta , in the hard amplitudes @xmath56 and @xmath57 .
this is the impulse approximation .
a second observation follows from kinematics ; although momentum conservation at the hard scattering @xmath56 relates the internal momentum dependence of the @xmath59 s , a variation of one such momentum @xmath60 in its @xmath53 or @xmath54 direction induces negligible modifications in the three other @xmath59 s .
consequently , momentum components of @xmath60 along @xmath53 or @xmath61 only appear as a relevant variable in the wave function @xmath49 and integrals over these components can be carried out .
the components along @xmath32 , denoted @xmath62 , represent transverse momentum out of the scattering plane and do not share the same property .
thus , the vertex delta function may be reexpressed as @xmath63 which indicates that the 4 constituents which enter or leave each hard scattering carry the same light cone fraction ( @xmath64 or @xmath65 ) . introducing the
`` out of plane '' impact parameter @xmath42 through @xmath66 one may write the amplitude of the process @xcite as @xmath67 where @xmath68 is the amplitude @xmath69 with dirac indices @xmath70 .
color indices are suppressed in eq .
( [ a0 ] ) and sums over repeated indices are understood .
we consider unflavored quarks .
the implementation of flavor is straightforward by setting to 0 some of the graphs we are going to consider .
@xmath56 and @xmath57 are feynman amplitudes ( a sum over allowed diagrams is understood ) at the level of quarks computed with standard perturbative qcd vertices and internal propagators ; they consist , at lowest order in the coupling constant , of one - gluon exchange or @xmath51 annihilation for each quark pair .
note that @xmath68 and @xmath56 are not individually gauge invariant .
if the short distance @xmath71 limit is assumed , then the four general wave functions in eq .
( [ pion ] ) can be reduced to one @xmath24term , by taking the trace @xmath72 of @xmath73 with @xmath74 . for the case of helicity conserving amplitudes , this selection was shown to be self
consistent by botts and sterman@xcite .
then , the zeroth moment of @xmath72 is related to the decay constant of the corresponding meson @xmath75 where , _
e.g. _ for the pion , @xmath76 .
this zero distance quantity contains no information on the interesting dependence on the transverse variable @xmath42 .
here we are concerned with the leading order description of helicity violating terms .
thus , we will consider @xmath24type and @xmath25type amplitudes on an equal footing , and make no a priori assumption that the region @xmath71 dominates .
the development so far has been sufficient to isolate the kinematic region of interest , which as we have already noted is characterized by finite separation between the participating quarks in the out - of scattering plane direction .
the amplitude is thus a strong function of the spatial dependence of the wave function .
the bethe salpeter wave function is a bilocal matrix element and is not gauge invariant .
however , we will now discuss how gauge invariance of the description can be obtained .
the key is in how the perturbation theory is re - arranged . in the sterman and botts factorization
certain `` soft '' corrections are put into the wave functions , leaving other parts of feynman diagrams to go into the hard scattering kernel . in dressing the wave function in this way ,
it is no longer a quark correlation ( the bethe salpeter wave function ) , but the matrix elements of operators determined by the types of diagrams put in .
the operators chosen in @xcite are path ordered exponentials ( poe s ) , shown by collins and soper @xcite to be the generators of eikonal approximations to the gluon attachments .
the poe s are gauge covariant , leading to a gauge invariant amplitude .
this is partly forced by physics , and partly a convention . as a convention for the perturbation theory ,
subsequent diagrams must be evaluated with subtractions to avoid double - counting .
more generally , any operator functional of the @xmath77 fields which transforms properly could serve in place of the poe s , and creating a different subtraction procedure .
let us extract what we can that is independent of convention .
let the operator in the definition of the wave functions be called @xmath78 ; we will call it a gauge - dressing operator . under a gauge transformation at the position @xmath10 , we require @xmath78 to transform like an antiquark
. then products such as @xmath79 are gauge invariant .
that is , we have gauge invariant matrix elements to find a dressed quark @xmath80 it is obvious that this requirement does not determine @xmath78 uniquely , because one could always attach a factor which is gauge invariant without changing the gauge transformation properties .
the particular choice of what to attach is a prescription , _
i.e. _ a definition of what parts of the amplitude will be put in the wave function and what in the hard scattering , and it can not be determined by gauge invariance alone .
however , due to gauge transformations one must attach some kind of gauge - dressing operator to have a well defined matrix elements .
although the standard way to do gauge dressing with the poe is path dependent , no path dependence generally need be associated with @xmath78 and in particular the observable process does not determine or favor any path .
this important point can be seen with an elementary example from qed , where the @xmath81 gauge invariance is easier to control .
the straightforward qed analogue of our process involves equal time ( not light cone time ) correlation functions in the gauge @xmath82 .
this gauge choice eliminates a mode , but there still remains a lack of definition of the coordinates due to time independent gauge transformations @xmath83
@xmath84 this produces a change in the longitudinal modes .
these modes are sometimes also called unphysical , a very unfortunate choice of terminology . in free space and in the absence of coupling a gauge theory
has two transverse degrees of freedom and the third would be called unphysical .
however we are interested in the case that matter fields exist ( and the non - abelian coupling is turned on ) in which case the third mode is real , but special inasmuch as being constrained in terms of the other variables by gauge invariance . to see this , note that we can decompose into transverse and longitudinal parts , @xmath85(x ) \hskip .25 in \;\hbox{(transforming part ) } \nonumber\\ { \bf a}_t & = & { \bf a } - \nabla\phi \hskip .75in\;\hbox{(invariant part).}\end{aligned}\ ] ] since it is there for gauge transformations , the longitudinal part @xmath86 is not free to be varied in arbitrary dynamical ways , but must accompany the matter field in a prescribed , unique functional . at time @xmath87
this operator is @xmath88 from which one can verify that under a gauge transformation , @xmath89 we will call this the coulomb dressing because it creates a classical coulomb field around the matter particle , as the reader can check by calculation . since the hamiltonian commutes with the gauge transformation operator once @xmath82 has been set , the time evolution will maintain the invariance of the combination @xmath90 .
however , as noted already , one is not forced to accept this as the unique answer , but can opt for @xmath91 , which will transform in the same way for any @xmath92 .
the reader may still be curious to know the relation between coulomb dressing and the poe approach .
this can be very simply exhibited by noting that @xmath93 this expression is still path independent .
this is the choice @xmath94 .
a different choice is @xmath95 in which case we have @xmath96 which is the standard path dependent poe .
both procedures are equally acceptable , as far as satisfying gauge invariance , but the path ordered exponential can create a line of physical transverse gauge field particles between the charged matter fields , depending on how it is oriented .
such a line of gluons , existing only along the chosen part , can be interpreted as an arbitrary model for the transverse gauge field inside the state of interest .
similarly , if one boosts the coulomb dressed definition , the boost also creates a blast of transverse gauge fields as seen by the `` equivalent photon '' approximation . in perturbation theory ,
the lowest order approximation to non - abelian dressing is the abelian case .
it is possible in the non - abelian theory to write down expressions analogous to the coulomb dressing but care must be used to keep track of the color indices .
it is equally valid to use poe s , which definitely transform properly , as building blocks to generate an infinite number of different ways to dress the quarks .
the different choices are not relevant for a leading order calculation to which we restrict this study .
the next crucial step is to elaborate a factorized form for the amplitude , whose prototype is eq .
( [ a0 ] ) , regarding radiative corrections . generalizing the results of @xcite to the case of the helicity violating dirac projections
, a leading approximation to the soft region rearranges these corrections to obtain the following expression @xmath97 where @xmath56 and @xmath57 are evaluated at respective scales @xmath98 and @xmath99 which are assumed to be large .
large logarithmic corrections to the process , with the coexistence of the two scales @xmath100 and @xmath101 , are resummed in @xmath102 , in such a way that @xmath103 is now a soft object : it does not include perturbative corrections harder than @xmath101 .
it is the non perturbative object necessary to connect short and long range physics , both present in hard hadronic processes .
expressions for the above quantities will be given in the phenomenological study , but let us specify here some of their qualitative features .
when @xmath42 is small , smaller or of the order of @xmath104 , radiative corrections are also small and must be considered as perturbative corrections to the hard amplitudes @xmath56 or @xmath57 ; in this regime , @xmath103 is closely related to the distribution amplitude @xmath105 evaluated at the renormalization scale @xmath106 @xcite . depending on the definition ,
the distribution amplitude also includes resummed logarithmic corrections from hadronic scale up to @xmath107 @xcite .
when @xmath42 is large , this replacement is not legitimate but since @xmath108 is a strong suppression factor in this region the exact value of @xmath103 is irrelevant .
endpoints in the @xmath10 integral where hard subprocesses would indeed become soft may look problematic .
both the distribution amplitude and wave function approaches used here become self consistent because of the end
point zeroes , e.g. @xmath109 , where @xmath110 which should occur independently of spin projection .
before analyzing helicity violating processes let us examine the leading contribution from the various components in eq . ( [ pion ] ) to a standard helicity conserving reaction such as @xmath111 . in their study , botts and sterman were interested in identifying the asymptotic behavior of the amplitude @xmath112 ( @xmath113 fixed ) .
asymptotically , the sudakov mechanism contained in @xmath108 results in a suppression of large @xmath42 region in the integral of eq .
( [ af ] ) . in this limit
one can forget about tensorial components of @xmath114 and only the projection of @xmath115 onto the particular tensor @xmath116 survives . in the intermediate @xmath0 regime ,
configurations of the @xmath51 pair sitting in a light meson with transverse separation smaller than the meson charge radius are not strongly affected by the sudakov mechanism @xcite . as anticipated in section 2
, @xmath117 components of the wave function which form large invariants in @xmath56 ( as large as the s - wave term ) may give sizeable corrections to the interaction amplitude between pure s - wave states .
for contributions with leading power behavior in the pseudoscalar case , we must keep the tensorial decomposition @xmath118\right\}_{\alpha\beta}.\ ] ] we now explore the calculation with this assumption . to begin with ,
one forms the projection , denoted as @xmath119 and @xmath120 , of the hard amplitudes @xmath56 and @xmath57 on the relevant various dirac tensors coming from the wavefunction decomposition .
we follow botts and sterman s classification of graphs , with three fermionic flows for @xmath56 and two gluonic channels each ( see fig .
3 ) , @xmath121 color flow in this problem is simplified by noting that one gluon exchange between two quark lines gives a color tensor @xmath122 which we may reexpress with the color tensors listed above , in the form @xmath123_{\{a_i\ } } = \sum_{i=1}^2c_ic^f_{i\{a_i\}},\ ; \left[s\,\hbox{or}\,u3 \right]_{\{a_i\}}=\sum_{i=1}^2\tilde{c}_ic^f_{i\{a_i\}},\ ] ] with @xmath124 with this notation , one finds for the hard amplitude @xmath125 , containing no @xmath42 factor , @xmath126 times an overall common factor @xmath127 where @xmath128 is the qcd coupling constant which appears in feynman rules .
we have already indicated that the whole amplitude eq .
( [ af ] ) can be properly defined regarding renormalization and factorization , so that @xmath129 stands for @xmath130 .
there is no @xmath131 or @xmath132 terms , due to the odd number of @xmath133 matrices ; this is a consequence of chiral symmetry .
the second term is therefore a hard amplitude @xmath134 , containing @xmath135 , which is found to be @xmath136 with a common factor @xmath137 ( here @xmath42 is a distance so that @xmath138 ) and the third one a @xmath139 hard amplitude , @xmath140 , which is the combination of eq .
( [ a0 ] ) , but with an overall factor @xmath141 a check of the above expressions or a possibility to reduce the number of graphs one has to compute is provided by symmetries under meson exchange ; starting from the expression one gets with two @xmath142-gluon exchange and fermionic flow @xmath143 , see fig.3 , which we label @xmath144 , one can generate @xmath145 the reader will easily find the channels obtained from another starting point , say @xmath146 , and the combination of exchanges needed to determine graphs which do not appear in the above array , thus completing the whole amplitude .
we are now ready to evaluate the integral defined in eq .
( [ af ] ) with the above hard amplitudes . in a first step ,
we approximate the sudakov factor by its dominant expression at large @xmath100 @xcite @xmath147\,\exp\,[x\leftrightarrow \bar{x}],\ ] ] with @xmath148 and @xmath149 for @xmath150 .
we have introduced the variable @xmath151 which turns out to be the relevant one to describe the sudakov unsuppressed region in the @xmath152plane : for @xmath153 there is no suppression from the first exponential in eq .
[ ull ] ; as soon as @xmath142 departs from unity , one gets rapidly a strong suppression due to the large @xmath154 factor .
( [ ull ] ) is likely to be valid for any dirac projection because the leading logs factor away independent of spin . in this approximation
, @xmath102 is a scalar in color space and one easily performs the color traces , with @xmath155 @xmath156 @xmath157 and obtains the following hard amplitudes , labeled by the power of @xmath42 entering : @xmath158 which at 90@xmath159 is @xmath160 @xmath161 for the remaining soft wave function @xmath72 , one can adopt two different models .
first , one can approximate @xmath72 to its value at small @xmath42 ( see @xcite and the argument in section 3.1 ) @xmath162 with @xmath163 which are standard choices for pion distributions .
the asymptotic form eq .
( [ phias ] ) is derived in @xcite ; it has no evolution with @xmath107 and is indeed the limit as @xmath164 of all distributions .
because evolution with @xmath107 is slow , the effective distribution at non asymptotic regime may be very different from @xmath165 .
chernyak and zhitnisky @xcite have built from qcd sum rules the above cz - form eq .
( [ phicz ] ) , which evolves with the scale , but with quite small effects on any computation when one does not examine a huge interval of energy .
we will ignore these effects in the following .
the existence of the additional component @xmath166 complicates the problem .
next to nothing is known about it , but a reasonable ansatz is to adopt a form similar to @xmath72 .
one notices that @xmath72 has a mass dimension which appears to be set by @xmath167 . in the case of @xmath166 ,
the dimension is a squared mass and we do not know what the normalization constant has to be .
we will adjust the normalization in the way described below , and assume the same @xmath10dependence for @xmath72 as @xmath166 .
for each hard part @xmath168 , freezing the coupling at the scale @xmath100 to simplify the study , we get through eq .
( [ af ] ) a value @xmath169 and perform the ratios @xmath170 . to get the asymptotic behavior
, we analytically evaluate the @xmath42-integral @xmath171 with a saddle point approximation ; using the change of variable @xmath172 .
one has a maximum of the integrand at @xmath173 and finds @xmath174 performing the @xmath10-integration @xmath175 an asymptotic expression is found for the ratio @xmath176 from which one deduces @xmath177 roughly speaking , each power of @xmath42 is suppressed by @xmath178 ( not @xmath104 ) . at accessible energies ,
we expect deviations from the result given in eq .
( [ ras ] ) .
this comes first from an eventual failure of the saddle point approximation with too small @xmath179 .
we thus have numerically evaluated the amplitude eq .
( [ af ] ) with @xmath102 given by eq .
( [ ull ] ) .
results for our computation of the ratio of amplitudes are displayed in fig . 4 with cz distribution ( solid line ) .
we get similar results for the asymptotic distribution amplitude ( fig . 5 ) . to fix the normalization , we choose here and in the following to set arbitrarily the ratios to 1 at @xmath180gev
we observe that @xmath181 decreases by a factor around 7 from @xmath180gev to 20gev .
this is a much milder suppression than the naively expected @xmath182 factor ( dotted curve ) .
@xmath183 drops more drastically by a factor around 20 in the same energy interval .
a numerical study shows that logarithmic corrections ignored within the saddle point approximation are not negligible in the accessible range of energies . at larger values of the energy , @xmath184gev , the approximated result of eq .
( [ ras ] ) becomes accurate .
secondly , eq . ( [ ull ] ) should be supplemented with non leading terms which are known in the s - wave case @xcite but presumably different in the p - wave case .
the neglected logs in the expression of @xmath102 eq .
( [ ull ] ) may modify soomewhat the ratio over some intermediate range of energy . to modelize such an effect ,
we add a simple @xmath185-independent term like @xmath186 in the expression of @xmath181 with @xmath187 some constant .
the ratio modified by such a factor is shown in fig . 4 as a shaded area limited by the curves corresponding to @xmath188 and @xmath189 ; this measures in some way the theoretical uncertainty on p - wave contribution suppression .
further theoretical progress in the computation of these @xmath190 terms will allow to get rid of this uncertainty but we do not tackle this task here . a third effect may come from the intrinsic transverse dependence of the wave function . while the replacement of @xmath191 by @xmath192 discussed in section 3.1
is reasonable at large @xmath100 ( @xmath193gev ) , it is more questionnable at intermediate values where long range physics may be accounted for by including some intrinsic @xmath42-dependence @xcite as @xmath194 with parameters @xmath195 , @xmath196gev@xmath3 and @xmath197 for the asymptotic distribution amplitude .
the results are modified as depicted in fig . 5 where the curves from this wave function and from the asymptotic distribution are shown for comparison .
furthermore , we notice that the phenomenology may also be modified considerably by allowing variations of different distribution form , that is @xmath198 for @xmath72 and @xmath199 for @xmath166 or the reverse choice .
these three effects show that the power - like decrement eq .
( [ ras ] ) of the ratio is diluted at intermediate energies and consequently the amplitudes @xmath200 and @xmath201 are likely to compete over a rather large interval of @xmath202 , say 1gev@xmath3100gev@xmath3 .
to begin with , let us find the possible tensorial decomposition for the quark antiquark wavefunctions of the @xmath205 meson .
one specifies a one particle state by the momentum @xmath206 ( @xmath207 ) , which with the notation of section 3 , one can write : @xmath208 ( m is the @xmath205 mass which we do not neglect for the moment ) , and by the helicity @xmath209 .
the 3 helicity states are described in a covariant way with the help of the 3 vectors @xmath210 , satisfying @xmath211 and @xmath212 . for @xmath213
, @xmath214 is a direct trihedral , and we choose @xmath215 a consistent choice for @xmath216 is @xmath217 thus , demanding a parity @xmath218 state , the bethe - salpeter amplitude has the most general dirac - matrices expansion @xmath219+{\cal p}'_0\,/\!\!\!\varepsilon_h+ { \cal p}_1\,\varepsilon_h.b\,/\!\!\!p+{\cal p}'_1\,\varepsilon_h.b+ \tilde{\cal p}_1\,[/\!\!\!\varepsilon_h,/\!\!\!p]\,/\!\!\!b+ \tilde{\cal p}'_1\,[/\!\!\!\varepsilon_h,/\!\!\!b]+ { \cal p}_2\,\varepsilon_h.b\,[/\!\!\!p,/\!\!\!b]+ { \cal p}'_2\,\varepsilon_h.b\,/\!\!\!b.\nonumber\end{aligned}\ ] ] one can then extract the relevant components for the study of independent scattering processes .
that is , we isolate dominant high - energy tensors , those which contain one power of the large scale @xmath100 and we restrict to @xmath220 ( _ i.e. _ @xmath221 in the notation of section 3.1 ) .
we get for a longitudinally polarized @xmath205 @xmath222\right),\ ] ] and for a transversally polarized one @xmath223 + { \cal p}_1\,\varepsilon_h.b_t\,/\!\!\!v+ \tilde{\cal p}_1\,[/\!\!\!\varepsilon_h,/\!\!\!v]\,/\!\!\!b_t+ { \cal p}_2\,\varepsilon_h.b_t\,[/\!\!\!v,/\!\!\!b_t]\right).\ ] ] before going through the computation of an helicity violating process , let us exhibit what we call the _
double - flip _ rule .
it follows from the property of each dominant component that each power of @xmath42 changes the chirality ( which is @xmath224 if the component anticommutes with @xmath225 and @xmath218 if it commutes ) . with vector gluon couplings
, it follows that if the process is allowed by hadron helicity conservation , chiral symmetry requires that the @xmath42-components occur in pairs . with respect to the short distance contribution from valence states .
however the chirality of the component is the same as the p - wave we are considering ( that is opposite to the s - wave ) . the above argument then applies to ensure that when p - waves and one - gluon components are likely to compete , at least 2 such components are necessary : from section 4 , one counts @xmath226 for one p - wave and @xmath227 for one extra gluon ; this shows that considering one extra gluon in the fock state is indeed subdominant with respect to considering a p - wave state .
] we have already encountered this property in the @xmath228 elastic scattering .
furthermore , for an helicity 0-state , in both pseudoscalar and vector case , the chirality of the s - wave term is positive , whereas for an helicity odd state , it is negative . then with chiral symmetry and because the total power of @xmath42 has to be
even , one concludes that the number of helicity-0 states has to be even too . for a 2 to 2 process ,
only those processes that violate the helicity conservation rule by two units are expected to be non - zero in the present framework and hence important at large energy .
an example of interaction which satisfies neither the helicity conservation rule nor the double - flip rule is @xmath229 such an amplitude is thus vanishing by power counting .
we then turn to the study of an helicity violating process which , however , satisfies the double - flip rule , namely @xmath230 .
let us first verify of the vanishing of the hard amplitude using the s - wave components of the external mesons .
for this purpose , it is instructive to examine the connection between quark helicities and the s - wave dirac tensors we have used until now .
this is accomplished in the following way .
one considers the free massless spinors of a quark and an antiquark moving in the same direction , the quark having a momentum @xmath231 and the antiquark a momentum @xmath232 , so that the compound system has a momentum @xmath233 .
then one constructs the four possible helicity states of the system with solutions of the dirac equation and finds @xmath234 the helicity conservation rule is then easily verified with these combinations of spinor when one chooses the chiral representation @xcite . in this representation ,
the 2 diagonal blocks of each @xmath235 are equal to the null 2@xmath2362 matrix .
a feynman - graph fermion line , with vector ( or axial ) couplings and massless propagator , is an even number of @xmath133matrices between two spinors : @xmath237 inclusions of @xmath225 , which is diagonal in this representation , do not modify this property .
then , since to order @xmath238 the chirality of a spinor corresponds to the helicity of the state and , in this representation , chiral eigenstate spinors have either their 2 first components or their two last equal to 0 , states of different helicity always give a null product .
this property is algebraic and therefore independent of the representation chosen .
however , it is more difficult to observe its effects in the trace formalism described in section 2 .
let us examine how it works in the case of the reaction considered . among the twelve graphs discussed in section 2 ,
each @xmath239 and @xmath240 graph is 0 because they contain traces over the product of an odd number of @xmath235 matrices .
for the 8 remaining graphs , the sequence @xmath241 occurs ( the anticommutation to the left of every @xmath225 is understood and does not modify the reasoning ) .
the product of an odd number of @xmath235 being a linear combination of @xmath235 and @xmath242 , one is left with the evaluation of @xmath243 where @xmath244 and @xmath245 have been used and the index 3 dropped .
one can decompose each @xmath235 onto @xmath246 , which are such that their square is 0 and their anticommutation rules are @xmath247 to conclude that all graphs effectively vanish . adding the contribution from one p - wave function does not give any contribution ( even before the integration over @xmath42 ) , because the total number of @xmath235 is necessarily odd .
the first non zero term is therefore a @xmath135 hard amplitude @xmath248 and the computation , within the approximation of section 4.2 , leads to @xmath249 this amplitude @xmath248 has to be supplemented with @xmath102 ( eq .
( [ ull ] ) ) and integrated over @xmath42 and @xmath10 .
even though this combination involves several unknown objects , one notices that the angular dependence varies from one component to another and is rather different from the one obtained in @xmath228 elastic scattering .
therefore , it may be possible to analyze the contribution to helicity violation processes from different wave functions and use this information to deduce properties of the wave functions .
the numerical study of section 4 can then be used to understand the energy dependence of double helicity violating processes at accessible energies .
as explicitly shown in fig .
4a and 5 , the naive @xmath182 factor is replaced by a milder suppression , and this is primarily due to the specificity of the independent scattering mechanism supplemented by sudakov effects . even at very large energies
the @xmath250 ratio of eq .
( [ ras ] ) looks quite weak a suppression .
studying meson - meson scattering is an unrealistic simplifying assumption . including baryons is a necessary but quite intricate further step , because of the high number of feynman diagrams and internal degrees of freedom to be integrated over when more valence quarks are involved .
we can however still draw some conclusions from our analysis , leaving to future work a comprehensive study .
the mechanism we have explored occurs in several experimentally accessible circumstances .
indeed , there is a host of reactions involving hadronic helicity violation from which we could learn about the interface of perturbative and non - perturbative qcd .
the helicity density matrix of the @xmath205 meson produced in @xmath251 at 90@xmath252 is a nice measure of helicity violating components .
experimental data @xcite yield @xmath253 , at @xmath254gev@xmath3 , @xmath255 , for the non - diagonal helicity violating matrix element . without entering a detailed phenomenological analysis
, we may use the results of section 5 through the following line of reasoning . assuming that the presence of the third valence quark , which is not subject to a third independent scattering , does not alter much the results
, one may view @xmath256 as coming from the interference of an helicity conserving amplitude like @xmath257 with a double helicity flip amplitude like @xmath230 .
we then get a mild energy dependence of this matrix element
_ i.e. _ @xmath250 ( eq .
( [ ras ] ) ) at asymptotic energies or as shown in fig .
4a and fig . 5 at accessible energies .
this is at variance with the picture emerging from the diquark model @xcite .
measuring the energy dependence of this effect is thus highly interesting .
the most well - known example of hadron helicity violation occurs in @xmath258 scattering @xcite .
our demonstration of helicity violating contributions to meson - meson scattering has a bearing on this , because generalized meson scattering is embedded in the diagrams for proton scattering . without needing to make any dynamical assumption of `` diquarks '' ,
the perturbative qcd diagrams for @xmath258 scattering contain numerous diquark regions , convolved with scattering of an extra quark .
there is no known selection rule which would prevent the scattering of such a subprocess with helicity flip from causing helicity flip in pp elastic scattering .
this does not exhaust the possibilities , because there are other channels of momentum flow and color combinatorics which might have different interpretations .
the data for @xmath258 also reveals large oscillations about power - law behavior , a second piece of evidence that the short distance picture is inadequate .
elsewhere @xcite we have identified these oscillations as a sign of independent scattering . given the theoretical @xcite and experimental evidence , we therefore find no evidence that hadronic helicity conservation is a feature of perturbative qcd , and we believe that independent scattering is a main contender in explaining the observations .
since reactions of baryons are extremely complicated , and next to nothing is known about the various wave functions in the proton , a productive approach to the question is to ask for experimental circumstances in which the general mechanism we have outlined could be tested without requiring too much detail .
we believe that progress here will come from using nuclear targets , and studying the phenomenon of color transparency in hard ( as opposed to diffractive ) reactions .
this program has been outlined elsewhere @xcite ; it suffices to mention here that suppression of large @xmath135 regions is expected in reactions of large nuclei .
it follows that helicity conservation should be obtained in the same circumstances .
thus the mechanism we have outlined is testable .
it is time to go beyond the short - distance physics used to establish qcd as the correct theory , and we believe that a multitude of phenomena involving spin , color transparency , and detailed hadron structure , will play a major role in the future . * acknowledgments * : this work was supported in part under doe grant number de fg02 85 er 40214 and the kansas institute for theoretical and computational science , and by the eec program `` human capital and mobility '' , network `` physics at high energy colliders '' , contract chrx ct930357 .
we thank stan brodsky , steve heppelmann , al mueller , jacques soffer , george sterman , and joseph zhitnitsky for useful discussions .
centre de physique thorique is _ unit propre du cnrs_. s. j. brodsky and g. p. lepage , phys .
d * 22 * , 2157 ( 1980 ) .
s.j . brodsky and g.p .
lepage , phys .
d * 24 * , 2848 ( 1981 ) .
n. isgur and c. h. llewellyn smith , phys .
lett . * 52 * , 1080 ( 1984 ) , phys .
b217 * , 535 ( 1989 ) ; a. v. radyushkin , nucl . phys . *
a532 * , 141 ( 1991 ) .
p. v. landshoff , phys .
d * 10 * , 1024 ( 1974 ) .
a. h. mueller , phys .
rep . * 73 * , 237 ( 1981 ) .
j. botts and g. sterman , nucl . phys . *
b325 * , 62 ( 1989 ) ; j. botts , _ ibid . _ * b353 * , 20 ( 1991 ) . g. r. farrar and d. jackson , phys .
lett . * 43 * , 246 ( 1979 ) ; s. j. brodsky and g. p. lepage , phys . lett . * 87b * , 359 ( 1979 ) ; a. v. efremov and a. v. radyushkin , phys . lett . *
94b * , 245 ( 1980 ) ; v. l. chernyak , v. g. serbo and a. r. zhitnitsky , yad . fiz . *
31 * , 1069 ( 1980 ) ; a. duncan and a. h. mueller , phys .
d * 21 * , 1636 ( 1980 ) .
for a recent review , see s. j. brodsky and g. p. lepage in _ perturbative qcd _ , edited by a. h. mueller ( world scientific , singapore , 1989 ) .
j. c. collins and d. e. soper , nucl .
b193 * , 381 ( 1981 ) and * b194 * ( 1981 ) 445 .
ralston and b. pire , in _ polarized collider workshop _
( penn state university , november 1990 ) edited by j.c .
collins , _ et al . _
( aip conference proceedings no . 223 ) p. 228 and
university of kansas and ecole polytechnique preprint a175.0592 ( unpublished ) .
r. jakob and p. kroll , phys . lett .
* 315b * , 463 ( 1993 ) . v.
l. chernyak and a. r. zhitnitsky , phys . rep .
* 112 * , 173 ( 1984 ) .
g. r. farrar and f. neri , phys .
130b * , 109 ( 1983 ) .
s. heppelmann _
lett . * 55 * , 1824 ( 1985 ) .
m. anselmino , p. kroll and b. pire , z. phys .
c * 36 * , 89 ( 1987 ) .
et al . _ ,
rev . d*32 * , 3070 ( 1985 ) .
b. pire and j. p.
ralston , phys .
117b * , 233 ( 1982 ) ; j. p.
ralston and b. pire phys .
* 49 * ( 1982 ) 1605 ; _ ibid _ * 61 * ( 1988 ) 1823 ; c. e. carlson , m. chachkhunashvili and f. myhrer , phys .
d * 46 * , 2891 ( 1992 ) .
g. r. farrar , phys .
rev . lett .
* 56 * , 1643 ( 1986 ) ; j. p.
ralston and b. pire , phys .
lett . * 57 * , 2330 ( 1986 ) .
ralston and b. pire , phys .
lett . * 65 * , 2343 ( 1990 ) ; see also in _ electronuclear physics with internal targets _
, edited by r.g .
arnold ( world scientific 1990 ) .
figure 1 : coordinate space pictures of meson meson independent scattering .
( a ) trajectories of quarks .
in the scattering plane quarks approach each other within a distance of order @xmath104 , while the transverse separation of scattering planes ( indicated in perspective view ) is larger and set by the wave function .
( b ) contour map of real part of the product of @xmath30 leading sudakov wave functions ( from @xcite ) which give a model for the integrand of eq .
( [ a0 ] ) .
no other soft wave function has been introduced .
( c ) same wave function multiplied by a polynomial representing @xmath259 .
contours in ( b , c ) show flattening in the @xmath33-direction of the effective wave function imposed by a gaussian with @xmath260 at @xmath2612 gev@xmath3 ; higher @xmath0 increases the flattening . units of @xmath34 and @xmath262 are fm .
figure 4 : the energy dependence of the @xmath181 ( a ) and of the @xmath183 ( b ) ratios with cz distribution amplitude @xcite ( thick lines ) and their naive behaviors ( thin lines ) , respectively @xmath182 and @xmath263 .
the shaded area in ( a ) indicates uncertainty from neglected logs .
the ratios are normalized to 1 at @xmath2642gev .
figure 5 : the energy dependence of the @xmath181 ratio calculated with the asymptotic distribution amplitude with ( solid thick line ) and without ( dashed line ) intrinsic @xmath42-dependence .
the thin line is as in fig .
|
we study a new mechanism for hadronic helicity flip in high energy hard exclusive reactions .
the mechanism proceeds in the limit of perfect chiral symmetry , namely without any need to flip a quark helicity .
the fundamental feature of the new mechanism is the breaking of rotational symmetry of the hard collision by a scattering plane in processes involving independent quark scattering .
we show that in the impulse approximation there is no evidence for of the helicity violating process as the energy or momentum transfer @xmath0 is increased over the region @xmath1 . in the asymptotic region
@xmath2 gev@xmath3 , a saddle point approximation with doubly logarithmic accuracy yields suppression by a fraction of power of @xmath0 .
chirally odd " exclusive wave functions which carry non zero orbital angular momentum and yet are leading order in the high energy limit , play an important role .
hadron helicity violation in exclusive processes : + quantitative calculations in leading order qcd * thierry gousset *
|
significant efforts have focused on determining the sources responsible for cosmic reionization .
studies suggesting the apparent minor role of quasars in keeping the universe ionized at @xmath8 ( e.g. , @xcite but see @xcite ) have led the community to assess the contribution of star - forming galaxies to the ionizing photon budget at high redshift ( e.g. , @xcite ) .
these investigations have advocated broadly for the important role of hitherto undetected uv - faint galaxies with their high number densities as inferred from the steep faint - end slopes of the uv luminosity functions and their assumed large lyman continuum ( lyc ) escape fractions in dominating the ionizing background .
however , the direct detection of lyman continuum radiation , possible only at modest redshifts ( @xmath9 ) where the transmissivity of the inter - galactic medium ( igm ) allows such measurements , remains elusive .
locally , only a handful of objects have been detected with lyc emission @xcite .
there are a number of statistical detections in stacked spectra @xcite and individual detections of candidate lyc emission from both lbgs and ly@xmath6-selected samples @xcite at @xmath8 .
however , foreground contamination has been identified as a significant issue in interpreting the lyc emission from high - redshift galaxies @xcite .
nonetheless , at face value , lyc detections at @xmath10 , combined with upper limits on the escape fraction locally and at intermediate redshifts @xcite , have motivated an interpretation where the lyc escape fraction i.e . , the fraction of h - ionizing photons that escapes galaxies before being attenuated by either the igm or the circum - galactic medium(cgm)evolves towards larger values at higher redshift @xcite
. analyses of the limited number of lyc detections , as well as observations of the ly@xmath6 escape fractions in high - redshift galaxies , suggest that dust reddening and/or gas covering fraction play an important role in modulating the escape of lyc ( and ly@xmath6 ) photons ( e.g. , @xcite ) .
rest - frame uv spectroscopic studies of high - redshift galaxies have focused on the propensity of ly@xmath6/lyc escape based on an examination of the gas covering fraction as typically inferred from saturated low - ionization interstellar ( is ) absorption lines ( e.g. , , , ) that are relatively straightforward to measure ( e.g. , @xcite ) . however , these transitions may arise in gas of much higher optical depth
e.g . , if the ions are co - spatial with the densest gas than that required to significantly attenuate lyc photons ( @xmath11\approx 18 $ ] ) , leading to some ambiguity in the @xmath5 covering fraction .
obviously , one should directly measure the column density and covering fraction of @xmath5 where the ly@xmath6/lyc opacity originates .
the difficulty with this approach , until recently , has been that : ( a ) the detection of the @xmath5 absorption lines requires deep far - uv spectroscopy with blue - sensitive spectrographs , and ( b ) the proper interpretation of these lines requires careful corrections for differential atmospheric refraction relevant at the observed wavelengths , as well as large samples to average over variations in ly@xmath6 forest blanketing .
moreover , apart from the anecdotal evidence of a connection between ionizing escape fraction , _ neutral _ gas covering fraction , and reddening , a quantitative study of how these properties are related to each other has yet to be undertaken . to that end
, we have used a sample of @xmath0 lyman break galaxies ( lbgs ) to place constraints on the two mechanisms that act to suppress the escape of ionizing photons : photoelectric absorption by neutral hydrogen ( @xmath5 ) and dust obscuration . in reddy
( 2016a ) , subsequently referred to as paper i , we use our sample of @xmath0 lbgs to deduce the shape of the dust attenuation curve to wavelengths close to the lyman break .
the resulting curve rises less steeply towards bluer wavelengths than other commonly adopted attenuation curves ( e.g. , @xcite ) , implying a factor of @xmath12 lower dust attenuation of lyc photons for typical ( @xmath13 ) galaxies at @xmath0 . in this paper , we extend these results by modeling the composite ( stacked ) uv spectra of @xmath0 galaxies to assess the neutral gas column densities and covering fractions relevant for typical star - forming galaxies at these redshifts .
we then draw a connection between reddening , gas covering fraction , and ionizing escape fraction .
the outline of this paper is as follows . in section [ sec : sample ] , we briefly discuss the sample of galaxies and the procedure to construct their far - uv spectral composites .
the methodology of fitting models to the composite spectra is presented in section [ sec : methodology ] .
we discuss the evidence for a non - unity covering fraction of gas in section [ sec : nonunity ] .
the modeling of the stacked uv spectra and the correlations found between continuum reddening and gas covering fraction , column density , and line - of - sight extinction are considered in section [ sec : specfitting ] .
we discuss these correlations in the context of ly@xmath6 escape fractions and inferences of gas covering fractions from low - ionization interstellar absorption lines in section [ sec : lines ] . in section [ sec : escape ] , we discuss the implications of our results in the context of the escape of lyc photons from high - redshift galaxies , and how the framework established here can enable powerful constraints on the ionizing stellar populations of high - redshift galaxies .
all wavelengths are in vacuum .
all magnitudes are expressed in the ab system .
we adopt a cosmology with @xmath14kms@xmath15mpc@xmath15 , @xmath16 , and @xmath17 .
the imaging and spectroscopic survey of @xmath0 lbgs presented in @xcite forms the basis for the sample considered in both paper i and the present analysis . here
, we briefly recount a few of the salient aspects of our sample , and refer the reader to paper i for additional details .
candidate @xmath18 galaxies were selected to lie at @xmath0 based on their rest - frame uv colors as measured from ground - based @xmath19 imaging of 17 fields .
subsequent shallow spectroscopy of the candidates ( @xmath20hr ) with the low resolution imaging spectrograph ( lris ; @xcite ) on the keck telescope yielded a sample of 933 star - forming galaxies ( agn were excluded ; see paper i ) with secure spectroscopic redshifts in the range @xmath21 ( average redshift of @xmath22 ; see figure 1 of paper i for the histogram of redshifts ) .
the @xmath0 uv luminosity function indicates that galaxies in our sample are typical of those found at these redshifts , with an average luminosity corresponding to @xmath13 , or an absolute magnitude in the rest - frame uv ( @xmath23 ) of @xmath24 @xcite . of the 933 galaxies ,
121 were targeted with very deep ( @xmath25hrs ) spectroscopy with the blue arm of lris @xcite , providing rest - frame far - uv coverage at @xmath26 ( hereafter referred to as the `` lyc sample '' or the `` lyc spectra '' ) .
these additional observations were obtained with an atmospheric dispersion corrector ( adc ) to mitigate differential atmospheric refraction most prominent at blue wavelengths .
the deep spectroscopy is beneficial not only for the obvious reason of allowing for greater @xmath27 measurements close to the lyman break , but also enables the more efficient identification of lbgs that are contaminated by unresolved foreground objects .
the lyc sample excludes a handful of objects identified as spectroscopic blends ( see paper i and steidel et al .
, in prep . , for further discussion ) .
the rest - frame spectral resolutions of the shallow and deep spectra are @xmath28 and @xmath29 respectively .
the individual spectra were combined to produce a composite spectrum and associated error spectrum following the procedure presented in paper i. the signal - to - noise ( @xmath27 ) per resolution element of the final composite spectrum varies from @xmath30 at @xmath31 to @xmath32 at @xmath33 .
the composite of the 933 galaxies in our sample , including the shallow and deep spectra , is shown in figure [ fig : fcov_compall ] .
our analysis relies on the fitting of spectral models to the composite spectra of @xmath0 galaxies .
the composites enable us to average over discrete igm absorption features in order to accurately model the @xmath5 lyman absorption lines blueward of ly@xmath6 .
the methodology for fitting models to the stacked spectra is similar to that presented in paper i. here , we discuss some basic details of the adopted intrinsic template , the treatment of @xmath5 and @xmath34 absorption , and the fitting procedures .
models were fit to the composite spectra in order to deduce the reddening , average neutral gas column density , and covering fraction for galaxies in our sample . to accomplish this
, we constructed an intrinsic template from a combination of the @xcite models with the 2010 version of starburst99 far - uv spectral synthesis models @xcite .
we adopted a stellar metallicity of @xmath35 , based on current solar abundances @xcite , as this metallicity was found to best reproduce the uv stellar wind lines , including @xmath361176 and the @xmath361240 and @xmath361549 p - cygni profiles .
furthermore , a constant star - formation history and an age of @xmath37myr were assumed based on results from stellar population modeling of @xmath0 lbgs presented in @xcite .
nebular continuum emission was added to the stellar template to produce the final `` intrinsic template , '' which we denote by the symbol @xmath38 . to account for absorption by neutral hydrogen
, we added lyman series absorption to the intrinsic template assuming a voigt profile with a doppler parameter @xmath39kms@xmath15 ( the wings of the absorption lines are insensitive to the particular choice of @xmath40 ) and varying column densities @xmath41 .
, we consider a more realistic line profile for the @xmath5 derived from the composite spectrum itself . ]
the @xmath5 absorption was blueshifted by @xmath42kms@xmath15 to account for the observed positions of these lines in the composites .
the observed blueshift indicates that most of the neutral hydrogen seen in absorption is outflowing at several hundred kms@xmath15 .
models that include @xmath5 absorption are indicated by the symbol @xmath43 .
lastly , we included lyman - werner absorption with varying column densities and covering fractions of @xmath34 as described in paper i. this absorption , which is unresolved , decreases the level of the continuum at @xmath44 .
models that include both @xmath5 and @xmath34 are referred to by the symbol @xmath45 .
we considered two different cases when fitting the stacked spectra . in the first case
, we calculated the average reddening of the composite assuming a @xmath46 covering fraction of dust .
this is analogous to deriving the reddening of a galaxy assuming a starburst attenuation curve ( e.g. , @xcite ) . in the subsequent discussion ,
the `` updated '' versions of the @xcite and @xcite curves refer to those that are updated for our new measurements of the far - uv shape of the dust curve as presented in paper i. the intrinsic template was reddened by various amounts ( @xmath47 ) assuming the updated attenuation curves .
the best - fit @xmath1 was obtained through @xmath48 minimization by comparing the reddened templates i.e .
, @xmath49 where @xmath50 is the attenuation curve to the composite spectrum in the `` composite @xmath1 '' wavelength windows listed in table [ tab : waves ] .
these windows were chosen to avoid prominent interstellar and stellar absorption / emission features .
note that the windows used to measure @xmath1 are unaffected by @xmath5 or @xmath34 absorption , thus obviating the need to fit for this absorption when determining the reddening of the composite .
lc composite @xmath1 & 1268 - 1293 + & 1311 - 1330 + & 1337 - 1385 + & 1406 - 1521 + & 1535 - 1543 + & 1556 - 1602 + & 1617 - 1633 + + @xmath41 & 967 - 974 + & 1021 - 1029 + & 1200 - 1203 + & 1222 - 1248 + + @xmath51 & 941 - 944 + & 956 - 960 + & 979 - 986 + & 992 - 1020 [ tab : waves ] in section [ sec : nonunity ] , we present evidence from the composite spectra that indicates a non - unity covering fraction of gas for galaxies in our sample . in this second case , we fit models to the composite spectra assuming non - unity covering fractions of dust and neutral and molecular hydrogen .
the models were computed as : @xmath52 + \nonumber \\ & & ( 1-f_{\rm cov}({\text{\ion{h}{1}}}))m .
\label{eq : losfit}\end{aligned}\ ] ] here , @xmath2 is the covering fraction of dust and neutral gas , and @xmath53 is the covering fraction of @xmath34 .
this equation assumes that the @xmath34 is shielded by @xmath5 and that only the portion of the spectrum that emerges through the @xmath5 and @xmath34 gas is reddened .
as shown in sections [ sec : evidencefornonunity ] and [ sec : specfitting ] , this particular formalism is motivated by the fact that the ism is expected to be porous and that the covering fraction of @xmath34 is much smaller than that of @xmath5 ( see paper i and below ) .
the last term in equation [ eq : losfit ] corresponds to the portion of the light that exits the galaxy unabsorbed , which is simply the intrinsic template multiplied by @xmath54 .
finally , @xmath50 in this equation now refers to a _ line - of - sight extinction _ curve , in lieu of an attenuation curve , and @xmath55 refers to the corresponding line - of - sight reddening .
these adjustments assume that a foreground uniform dust screen describes the dust - covered portions of the galaxy .
the models were computed assuming a range of line - of - sight reddening , and covering fractions and column densities of @xmath5 and @xmath34 . in all cases ,
the models were smoothed to the spectral resolution of the composites , and modified by the cgm and igm opacity appropriate for the average redshift of each composite .
the average opacity was determined by randomly drawing line - of - sight absorber column densities from the distribution measured in the keck baryonic structure survey ( kbss ; @xcite ) , and taking an average over many lines - of - sight . covering fractions ,
as these fractions are sensitive to the ratio of the line depth to the continuum , both of which will be affected equally by any changes in the igm opacity . however , because the @xmath34 covering fraction is calculated by `` matching '' the model continuum to that of the composite spectrum in the @xmath34 windows ( table [ tab : waves ] ) , this quantity will suffer a systematic bias related to our choice of the igm+cgm opacity prescription ( see below ) . ] the best - fitting @xmath55 , @xmath2 , and @xmath41 were found by comparing the models to the stacked spectra in the `` composite @xmath1 '' and `` @xmath41 '' windows listed in table [ tab : waves ] .
as discussed in paper i , the `` @xmath41 '' windows were chosen to include the core and the wings of ly@xmath56 and ly@xmath57 , and the wings of ly@xmath6 , while excluding regions of metal line absorption ( e.g. , @xmath58 and @xmath59 ) .
the region between ly@xmath6 and @xmath60 is particularly sensitive to the gas column density and is included in the fitting windows .
the higher order lyman transitions ( e.g. , ly@xmath61 , ly@xmath62 , etc . ) were not used in the fitting due to line overcrowding .
furthermore , as noted above , the fitting windows do not include any interstellar metal absorption lines .
similarly , the best - fit values of @xmath53 and @xmath51 were determined by comparing the models to the stacked spectra in the `` @xmath51 '' windows listed in table [ tab : waves ] .
as discussed in paper i , the `` @xmath51 '' windows were chosen to avoid regions of strong metal line and @xmath5 absorption .
the apparent porosity of the ism in galaxies examples of which have been seen or suggested in the local universe ( e.g. , @xcite ) , and which may be enhanced by the high star - formation - rate surface densities characteristic of high - redshift galaxies ( e.g. , @xcite)motivates spectral models that allow for less than a complete covering of the continuum by dust and neutral gas .
notably , the composite spectrum ( figure [ fig : fcov_compall ] ) exhibits @xmath5 absorption lines that have damping wings , but which do not reach zero intensity at the line cores .
the residual flux density under the @xmath5 lines is @xmath63 of the median flux density in the range @xmath64 . for the normalization shown in figure [ fig : fcov_compall ] ,
the residual flux density is @xmath65@xmath66jy .
as shown in paper i , the @xmath5 damping wings are best reproduced by high column density gas , with @xmath67 \approx 21.0 $ ] . at face value , these results suggest a partial covering fraction of high column density gas . however , we sought first to rule out other possible sources for the residual flux observed at the @xmath5 line cores , namely uncertainties in redshifts and spectral resolution , and foreground contamination .
we determined the extent to which uncertainties in the redshifts and the resolution of the individual galaxy spectra lead to non - zero fluxes in the @xmath5 line cores in the stacked spectrum . to accomplish this
, we added @xmath5 absorption of varying column densities @xmath68}=20.0 - 22.0 $ ] to the intrinsic template assuming ( as above ) a voigt profile with a doppler parameter @xmath39kms@xmath15 , and a @xmath46 covering of gas , resulting in an `` @xmath5-absorbed template '' .
thus , this @xmath5-absorbed template has saturated black @xmath5 line cores .
the @xmath5-absorbed template was then smoothed to have spectral resolutions varying from @xmath29 ( the average rest - frame resolution inferred for the lyc spectra ) up to a factor of 2 times worse ( @xmath69 ) . for each value of the spectral resolution ,
we generated 121 versions of the smoothed @xmath5-absorbed templates
one for each of the 121 lyc spectra contributing to the composite at @xmath70 where each version was shifted in wavelength according to a normal distribution with @xmath71kms@xmath15 .
this velocity spread is based on the uncertainties in the systemic redshifts when they are derived using interstellar absorption lines ( see @xcite ) , and the vast majority of the lyc spectra have systemic redshifts based on the latter .
these 121 versions were then averaged together to produce a composite model . based on these simulations
, we would have had to systematically underestimate the average resolution of the lyc spectra by @xmath72 in order to match the @xmath5 line depths for @xmath68}>20.0 $ ] , a highly unlikely situation given that this @xmath73 systematic difference is much larger than the rms uncertainty ( @xmath74 ) in determining the spectral resolution from the widths of the night sky lines .
for the actual value of the resolution of our spectra and our typical redshift uncertainty , the cores of the @xmath5 lines remain black , implying that none of the observed residual flux density can be attributed to spectral resolution or redshift uncertainties .
we conclude that systematic uncertainties in spectral resolution can not account for the residual line flux density at the @xmath5 line cores .
we ruled out foreground contamination
, low - redshift objects that are unresolved from the lbgs in the ground - based photometry and spectroscopy as a possible origin for the residual line flux observed at the @xmath5 line cores based on multiple lines of reasoning .
first , we examined the contamination statistics for the sample of lbgs with spatially - resolved _ hubble space telescope _
( _ hst _ ) multi - filter ( @xmath75 ) and ground - based nb3420 imaging ( to target lyc emission at @xmath0 ) analyzed in @xcite .
their sample contains 16 lbgs with _
hst _ imaging , four of which have nb3420 detections .
of these four , two were identified as having foreground contaminants , such that the offending low - redshift interlopers would not have been resolved separately from the background lbgs in the ground - based photometry . of the 12 objects that were not detected in the nb3420 imaging ,
only one suffers from foreground contamination .
thus , the contamination rates for nb3420 detections and non - detections are @xmath76 and @xmath77 , respectively . extrapolating to the full lbg sample of @xcite , where five of 49 objects were detected in the narrow - band imaging
, implies that the fraction of lbgs with contaminated photometry is @xmath78 .
note that what we have just computed is the foreground contamination _
rate_. however , the quantity of relevance in the present context is not the contamination rate , but the actual fraction of _ flux _ contributed by foreground contaminants . to calculate this flux contamination , we examined the photometry for the 16 lbgs with _ hst _ imaging from @xcite .
we summed the @xmath75 flux densities for the 16 objects , both including and excluding foreground contaminants that would have been unresolved from the lbgs in the ground - based imaging .
the ratio of the total flux in each band including foreground contamination to that excluding this contamination is @xmath79 , @xmath80 , @xmath81 , and @xmath80 , respectively , for the @xmath75 bands .
accounting for the higher foreground contamination rate among the 16 lbgs ( @xmath82 ) relative to the rate among the full lbg sample ( @xmath83 , see above ) , and adopting the reasonable assumption that the flux contamination per object is roughly constant , the aforementioned ratios become @xmath84 , @xmath85 , @xmath80 , and @xmath85 , respectively , for the @xmath75 bands .
in other words , the flux contamination fraction for the rest - frame non - ionizing uv / optical flux is @xmath86 .
the flux contamination fraction in the @xmath87-band is substantially larger , a point that we revisit in section [ sec : escape ] .
based on these statistics , the contaminating flux density across the wavelength range spanning the ly@xmath56 , ly@xmath57 , and ly@xmath61 lines is a factor of @xmath88 lower than the residual flux densities measured at the aforementioned line centers .
note that the previously discussed contamination rates were calculated for lbgs that were spectroscopically confirmed using relatively shallow spectroscopic data @xcite .
spectroscopic blends may not be as readily identifiable in these shallow spectra compared to the substantially deeper lyc spectra analyzed here .
as indicated in section [ sec : sample ] , the @xmath25hrs spectroscopic integrations obtained with lris - b allowed us to remove a handful ( @xmath63 ) of lbgs that were confused with foreground interlopers .
in fact , this contaminating rate is consistent with the expected number of chance superpositions of foreground objects , as discussed in steidel et al .
( in prep ) , implying that the present sample has been expunged of most contaminants .
thus , the contamination rates calculated above should be regarded as conservative upper limits .
there is additional evidence that suggests that the residual line flux at the @xmath5 line centers is intrinsic to the lbgs .
namely , in the subsequent analysis , we show that this residual flux varies systematically with @xmath1 such that redder galaxies exhibit lower levels of residual emission ( section [ sec : specfitting ] ) , a trend that would not be expected if foreground interlopers were dominating the observed flux .
for example , of the three lbgs identified with foreground contamination from the sample of @xcite , two have interlopers that result in slightly _ redder _ colors compared to the corrected photometry , and one has an interloper that results in a color that is indistinguishable from the corrected color .
notwithstanding the small statistics , it follows that an increase in contamination fraction for galaxies with redder @xmath1 should lead to a corresponding _ increase _ in residual flux , a trend that is opposite to the one actually observed ( section [ sec : specfitting ] ) . to summarize our arguments
, we find that systematic uncertainties in the resolution of our spectra , combined with random uncertainties in systemic redshifts , can not explain the level of residual flux seen in the stacked spectra .
furthermore , calculations of the flux contamination fraction for the sample of lbgs examined in @xcite imply that @xmath89 of the residual flux must be intrinsic to the lbgs and not a result of foreground interlopers .
if the statistics of how the colors of lbgs are altered due to foreground interlopers extend to larger samples , then significant contamination would be expected to lead to a higher residual flux for redder galaxies , contrary to the trend that we observe , as discussed in section [ sec : specfitting ] .
we conclude with some words of caution regarding the covering fractions derived here .
first , while the composite may indicate @xmath90 , low resolution spectroscopy may mask the presence of very narrow absorption components that are opaque to lyc radiation . however , if the presence of such narrow ( and black ) components were the rule , rather than the exception , for galaxies in our sample , we would expect little correlation between the covering fraction derived from low resolution spectra and the ly@xmath6 escape fraction , counter to our observations ( sections [ sec : specfitting ] and [ sec : lines ] ) .
second , the method for computing @xmath2 assumes that the different velocity components of the @xmath5 gas are spatially co - incident when viewed in projection . if this is not the case
, if optically - thick gas that absorbs far from line center covers sightlines that are disjoint from those where the absorption at line center dominates then @xmath2 deduced based on the overall depth of the @xmath5 absorption will underestimate the true @xmath5 covering fraction . as we show in section [ sec : escape ] , the two effects just mentioned ( i.e. , unresolved narrow absorption and spatially disjoint gas at different velocities ) must be limited in scope .
even so , in light of this discussion , one must consider that the covering fractions derived in our subsequent analysis may be lower limits to the true fractions . to demonstrate the requirement of a non - unity covering fraction , we modeled the composite spectrum of all galaxies in our sample ( figure [ fig : fcov_compall ] ) using the same procedure as above in fitting @xmath1 and @xmath5 ( we ignored @xmath34 for this exercise ) , but where we now assumed a unity covering fraction of both dust and gas .
the depth of the ly@xmath56 line can be reproduced with @xmath68}=15.9 $ ] , but such a low column density does not match the combined red damping wing of ly@xmath6 and p - cygni absorption , nor the depths of the higher lyman series lines ( figure [ fig : covfracexample ] ) . alternatively , the assumption of @xmath68}=21.0 $ ] gas reproduces the combined red damping wing of ly@xmath6 and p - cygni absorption , but implies optically - thick gas and hence saturated @xmath5 lines that should reach zero intensity at the line cores for a unity covering fraction .
thus , it is clear that neither low or high column density gas with complete covering can simultaneously reproduce the residual flux at the @xmath5 line centers and the damping wings of ly@xmath6 .
rather , the roughly equal depths of the lyman series lines combined with the damping wings of ly@xmath6 suggest optically thick gas with @xmath68}\ga 21.0 $ ] and @xmath90 . consequently , the depths of the @xmath5 lines and the red damping wing of ly@xmath6 are most sensitive to changes in the covering fraction of optically thick @xmath5 . from a physical standpoint ,
a partial covering of dust and gas can result from supernovae chimneys or galactic fountains , whereby supernovae explosions heat the ism and drive hot gas into the galactic halo ( e.g. , @xcite and references therein ) . aside from direct observations of such hot halos in the milky way and other nearby galaxies ( e.g. , @xcite ) , the higher star - formation - rate surface densities of , and the ubiquitous presence of outflows in , high - redshift galaxies may increase the porosity of the ism .
furthermore , simulations of high - redshift galaxies show that the lyc escape fraction is essentially independent of dust obscuration , with the ionizing photons escaping the galaxies through channels that are essentially free of dust @xcite , as depicted in figure [ fig : cartoon ] . in this cartoon picture ,
the column of gas is sufficient to completely bound the ionized region by neutral gas , where the bulk of the neutral and ionized gas along the line - of - sight are entrained in an outflow ( see also figure 1 in @xcite ) .
using the procedure laid out in section [ sec : methodology ] , we fit a spectral model to the composite of all galaxies ( full composite ) , assuming a non - unity covering fraction of dust and gas , and an smc extinction curve .
the model was obtained by simultaneously varying @xmath91 , @xmath2 , @xmath41 , @xmath53 , and @xmath51 , until a best - fit was reached .
the best - fit parameters are @xmath92 , @xmath67=21.0 $ ] , @xmath93 , @xmath94=20.9 $ ] , and @xmath95 . by perturbing the composite spectrum by its errors and refitting these realizations
, we estimated the errors in @xmath55 , @xmath96 , and @xmath67 $ ] to be @xmath97 , @xmath98 , and @xmath99 ) \approx 0.20$]dex , respectively . nominally , these errors include the covariance between the parameters .
however , as @xmath2 is determined primarily by the depth of the @xmath5 lines , and @xmath41 is constrained primarily by the wings , there is negligible covariance between the covering fraction and column density of @xmath5 .
furthermore , as noted in paper i , the @xmath34 lines are unresolved and therefore we can not accurately determine the errors in the @xmath34 covering fraction and column density . in other words , for the models to match the continuum level of the stacked spectrum in the `` @xmath51 '' windows , the @xmath34 column density may be increased while the covering fraction is decreased , or vice versa . as noted above
, @xmath53 may also suffer a systematic bias depending on which igm opacity prescription we adopt .
lcccccccc @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & & @xmath105 & @xmath106 & @xmath107 + @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath104 & & @xmath112 & @xmath111 & @xmath113 + @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath113 & & @xmath118 & @xmath117 & @xmath119 + @xmath120 & @xmath121 & @xmath122 & @xmath123 & @xmath124 & & @xmath125 & @xmath123 & @xmath126 + @xmath127 & @xmath128 & @xmath129 & @xmath123 & @xmath124 & & @xmath130 & @xmath123 & @xmath131 [ tab : losfit ] as shown in figure [ fig : covfracexample ] and in paper i , a voigt profile provides a poor description for the shape of the @xmath5 lines ( i.e. , the voigt profile underestimates the width of the lines when simultaneously matching their depths ) . to obtain a better fit for these lines , we derived the shape of the @xmath5 line profile from the composite spectrum itself . to recover the @xmath5 line profile ,
the intrinsic spectrum was multiplied by @xmath54 , modified for the opacity of the igm , and subtracted from both the composite and its model fit , with the resulting spectra shown in figure [ fig : subtract ] .
this step allows us to remove the contribution of unabsorbed continuum prior to modeling the @xmath5 lines .
we then used the model fit subtracted by the intrinsic spectrum ( i.e. , red curve in figure [ fig : subtract ] ) to define `` continuum '' points by which the ly@xmath56 , ly@xmath57 , and ly@xmath61 line profiles were normalized .
there is a single continuum point each for ly@xmath56 and ly@xmath57 and we treated the continuum as constant across these line profiles . for ly@xmath61
, the continuum was estimated as a linear function passing through the two continuum points bracketing the line .
the normalized line profiles , transformed into velocity space ( where the minimum of each absorption trough is forced to zero velocity ) , and cast in terms of the optical depth , are shown in figure [ fig : subtract ] . the optical depth profile shown in figure [ fig : subtract ]
was assumed for all of the lyman series lines , except ly@xmath6 . because the damping wings are weaker and more difficult to measure for the higher lyman series lines due to line crowding , we modeled the damping wings of ly@xmath6 using a range of @xmath5 column densities with a doppler parameter @xmath39kms@xmath15as noted earlier , the wings of the absorption are insensitive to the particular choice of @xmath40 . in order to investigate trends between continuum color excess and covering fraction
, we divided the galaxies in our sample into five bins of @xmath1 .
specifically , the intrinsic template was reddened with different values of @xmath1 assuming the @xcite attenuation curve and a @xmath46 covering fraction of dust , shifted to the observed frame for the redshift of each galaxy , multiplied by the igm opacity relevant for that redshift , and then convolved with the @xmath132 and @xmath133 filters .
the best - fit @xmath1 was taken to be the value that resulted in a model - generated @xmath134 that best matched the observed @xmath134 color , where the latter was corrected for ly@xmath6 emission or absorption as measured from the spectra .
composite spectra for objects in each bin were constructed according to the procedures laid out in paper i and section [ sec : sample ] . in fitting the stacked spectra assuming a porous ism
, we considered both the smc and milky way extinction curves in the modeling ( i.e. , for the _ line - of - sight _ reddening ) , to span the possible shapes of the extinction curve in both low and high metallicity environments .
the fitting was accomplished by simultaneously varying @xmath135 , @xmath41 ( which affects the degree of damping of the ly@xmath6 line ) , @xmath2 ( which affects the depths of all the lyman series lines ) , and @xmath53 , until the differences between the composite spectra and the models were minimized in the wavelength regions indicated by the `` composite @xmath1 '' , `` @xmath41 '' , and `` @xmath51 '' entries in table [ tab : waves ] . as the @xmath34 lines are unresolved , we fixed @xmath136}=20.9 $ ] in the fitting .
we verified that the covering fractions of neutral gas obtained with and without including @xmath34 in the model fits were similar within @xmath137 .
the covering fractions of @xmath34 are significantly smaller , with a typical value of @xmath138 . additionally , each composite spectrum was fit assuming a @xmath46 covering fraction of dust in order to derive its @xmath1 .
the @xmath1 computed from the composite spectra agree well with the average @xmath1 of objects contributing to those spectra .
for simplicity in the following discussion , the color excess derived assuming a unity covering fraction of dust is referred to as the `` continuum reddening '' , @xmath1 , while the reddening deduced assuming a partial covering fraction of dust is referred to as the `` line - of - sight reddening '' , @xmath55 .
the continuum reddenings derived assuming the updated attenuation curves of @xcite and @xcite are denoted by `` @xmath139 '' and `` @xmath140 , '' respectively .
the fits assuming the smc extinction curve and a non - unity covering fraction of dust and neutral gas are shown in figure [ fig : confint ] , and the best - fit parameters and their errors are summarized in table [ tab : losfit ] .
the errors were calculated as the @xmath141@xmath142 dispersion in the values obtained by perturbing the composite spectra according to their error spectra and re - fitting these realizations .
not surprisingly , the new @xmath5 optical depth profile adopted for ly@xmath56 and the higher lyman series transitions ( section [ sec : newprofile ] ) provides a significantly better fit for these lines as well as the continuum level at @xmath143 than a single voigt profile , though the inferred covering fractions remain identical ( see figure [ fig : fcov_compall ] for the model fit to the full composite ) .
llcccc @xmath1 & @xmath2 & @xmath144 & @xmath145 & @xmath146 & @xmath147 + @xmath148 & @xmath68}$ ] & @xmath149 & @xmath150 & @xmath151 & @xmath152 + @xmath148 & @xmath153 & @xmath154 & @xmath155 & @xmath117 & @xmath156 [ tab : covfracfits ] several notable results from the fitting are summarized in figure [ fig : covfrac ] and table [ tab : covfracfits ] .
significant correlations are found between the continuum reddening and the line - of - sight reddening , column density of @xmath5 , and the covering fraction of @xmath5 .
for simplicity in the subsequent analysis , and unless indicated otherwise , we assume the updated @xcite curve for the continuum reddening , and the smc extinction curve for deriving the line - of - sight extinction and the neutral gas covering fraction and column density
. other combinations of the extinction / attenuation curves yield similar results to those obtained with the default choices specified above . to compute the significance of the correlations between @xmath1 and @xmath55 , @xmath41 , and @xmath2 , we calculated the pearson correlation coefficient and probability of a null correlation between the various quantities while perturbing the measured values many times according to their measurement uncertainties .
the values reported in table [ tab : covfracfits ] are the median pearson correlation coefficients and median probabilities of a null correlation between @xmath1 and @xmath2 , @xmath41 , and @xmath55 .
the inferred @xmath5 covering fractions assuming a partial covering fraction of both dust and gas are @xmath157 , substantially larger than those deduced assuming a unity covering fraction of dust ( @xmath158 ; see paper i ) . in the latter case ,
as the entire spectrum is reddened including the uncovered portion , the line depths can be reproduced with smaller covering fractions of @xmath5 . with a non - unity covering fraction of dust , larger @xmath5 covering fractions
are required to reproduce the depths when unreddened continuum is contributed from the uncovered portion of the galaxy spectrum . at any rate ,
our results imply that the reddening of the uv continuum slope reflects an increasing covering fraction of @xmath5 and dust , where the covering gas is also characterized by a higher line - of - sight reddening .
furthermore , the column density of @xmath5 increases with continuum reddening .
the ratio of the neutral gas column density and the line - of - sight reddening is only marginally correlated with continuum reddening with a pearson correlation coefficient of @xmath159 and a probability of null correlation of @xmath160 ( figure [ fig : nhiebmv ] ) .
the mean ratio of the neutral gas column density and the line - of - sight reddening for our sample is @xmath161 and @xmath162@xmath163mag@xmath15 for the smc and milky way extinction curves , respectively , where the errors represent the standard deviation in the five values contributing to the mean . for context ,
the average values found for smc and mw sightlines are @xmath164 and @xmath165@xmath163mag@xmath15 , respectively @xcite .
a note of caution is that the @xmath166 values calculated here are analogous to , but not the same as , the gas - to - dust ratio .
specifically , we do not make any accounting of either the molecular or ionized gas in our column density estimates , and much of the line - of - sight reddening can be attributed to dust in the outflowing _
ionized _ gas in high - redshift galaxies ( see section [ sec : isabslines ] ) .
the discussion in the previous section focused on the parameters obtained from fitting the composite spectra , including the continuum and line - of - sight reddening , and the covering fraction and column density of @xmath5 .
we now turn our attention to other features in the stacked spectra that can give further insight into the gas covering fraction and the distribution of metals and dust between the neutral and ionized ism . given the correspondence between ly@xmath6 strength and reddening / covering fraction noted in previous studies ( e.g. , @xcite ) ,
we calculated the absolute ly@xmath6 escape fraction , @xmath167 where @xmath168 and @xmath169 are the average observed and intrinsic ly@xmath6 line luminosities , respectively , for the galaxies contributing to each composite shown in figure [ fig : confint ] .
the observed line luminosity for each spectral stack was calculated as follows .
we first redshifted the composite to the mean redshift of objects contributing to that composite .
the spectrum was then converted to @xmath170 units .
the observed ly@xmath6 line luminosity was calculated by directly integrating the flux over just the ly@xmath6 emission profile ( i.e. , from the points where the emission line intersects with the underlying ly@xmath6 absorption profile ) using the iraf task _
splot_. naturally , this calculation will not account for the potentially large fraction of ly@xmath6 ( up to @xmath171 ; @xcite ) that may be resonantly scattered into a diffuse halo ( see discussion below ) .
the intrinsic ly@xmath6 line luminosity was estimated from the star - formation rate ( sfr ) .
the sfr was calculated using the uv luminosity density at @xmath172 , corrected for dust based on the average @xmath1 of objects contributing to each composite and assuming the @xcite attenuation curve , and converted to sfr assuming the @xcite relation for a @xcite imf .
the intrinsic ly@xmath6 luminosity was then estimated from the sfr using a modified form of the relation given in @xcite , where we assumed that the intrinsic ratio of the ly@xmath6-to - h@xmath6 line intensities is @xmath173 for case b recombination and @xmath174k .
this yields the following relationship between sfr and ly@xmath6 line luminosity , again assuming a @xcite imf : @xmath175 = \frac{{\rm sfr}\,[m_\odot\,{\rm yr}^{-1}]}{9.1\times 10^{-43}}.\end{aligned}\ ] ] the measurement error in the mean escape fraction was estimated by perturbing the composite many times according to its error spectrum , and calculating the observed and intrinsic ly@xmath6 luminosities for each of these realizations .
note that the observed ly@xmath6 line luminosity and sfr are calculated using the spectrum itself ; we made no corrections for slit loss , and thus the escape fractions derived here assume that the ly@xmath6 and uv continuum fluxes arise from spatially coincident regions .
the resulting @xmath176 are indicated for each composite in figure [ fig : confint ] .
not surprisingly , @xmath177 correlates with continuum reddening ( and hence with @xmath2 ; figure [ fig : lyafrac ] ; see also @xcite ) with a pearson correlation coefficient of @xmath178 and a probability of a null correlation of @xmath179 . for four of the five @xmath1 bins ,
the absolute escape fraction of ly@xmath6 photons agrees well with our expectation if such photons only escaped through clear sightlines in the ism ; namely , @xmath180 ( figure [ fig : lyafrac ] ) .
the bluest bin of @xmath1 exhibits a larger @xmath177 than this expectation .
this may suggest that a significant fraction of the ly@xmath6 flux may be resonantly scattered into the spectroscopic slit aperture and/or may arise from ly@xmath6 photons scattering out of resonance , though we note that the kinematic profile of the ly@xmath6 line for the galaxies in the bluest @xmath1 bin is not substantially different than that of the other bins .
moreover , it is unclear why the fraction of ly@xmath6 photons scattered into the spectroscopic slit would be much larger for galaxies with bluer @xmath1 . a more detailed assessment of _ how _
ly@xmath6 photons escape these galaxies ( as judged through the kinematic profile of the line ) must await more precise systemic redshifts derived from nebular emission lines . in the context of figure [ fig : cartoon ] , one might expect the fraction of ly@xmath6 emission exiting through low column density sightlines to perhaps correlate more directly with @xmath2 than the diffuse ly@xmath6 component ; the latter is not accounted for in our calculation .
however , based on the results of the previous section , both @xmath41 and @xmath55 increase in tandem with @xmath2 .
hence , ly@xmath6 photons that diffuse far from the sites of star formation face an increasing probability of absorption by dust with increasing @xmath2 , implying that the fraction of ly@xmath6 scattered into a diffuse halo and ultimately exiting the ism / cgm of the galaxy is also likely to correlate with @xmath2 .
this inference could have also been deduced from the results of @xcite , who find that when considering resonantly scattered ly@xmath6the attenuation of ly@xmath6 photons is consistently larger than that of the uv continuum , and the fraction of the total ly@xmath6 produced from star formation that diffuses into a halo correlates inversely with the spectroscopic ly@xmath6 flux . at any rate , the computed @xmath177
imply that the selection of galaxies based on strong ly@xmath6 emission will also tend to isolate those with lower gas covering and higher lyc escape fractions . in accord with expectations ,
the conditions that give rise to strong ly@xmath6 emission also facilitate the escape of lyc radiation .
given the prevalent use of low - ionization interstellar absorption lines as proxies for the neutral gas covering fraction in high - redshift galaxies ( e.g. , @xcite ) , it is instructive to directly compare the covering fractions deduced from the low - ionization lines with those computed from the depth of the @xmath5 lines . to accomplish this ,
the five composites shown in figure [ fig : confint ] were divided by their respective best - fitting models to produce normalized composites .
this method of normalization has the advantage of removing any stellar absorption component ( e.g. , @xmath181 ) that may affect equivalent width measurements of the interstellar absorption lines .
the normalized composites were transformed into velocity space around the @xmath182 , @xmath183 , @xmath181 , and @xmath184 is lines .
these include the strongest and most commonly - used is lines observed in the rest - frame uv spectra of high - redshift galaxies .
these features are shown in figure [ fig : lowions ] , with the different colors indicating the normalized spectrum in each bin of @xmath1 .
if the two transitions at @xmath185 and @xmath186 are optically thin , then the ratio of their equivalent widths will be @xmath187 .
the observed ratio is @xmath188 , implying that the lines are saturated and hence their depths are sensitive to the covering fraction of -enriched material .
other strong low - ionization is absorption features are shown in figure [ fig : lowions ] , including @xmath181 , and @xmath184 lines .
the similarity in the velocity widths and depths of these additional lines to those of the saturated transitions discussed above suggests that the former may be also arise from the same gas as the latter .
if this is the case , then collectively the depths of the low - ionization lines indicate a covering fraction that rises from @xmath189 to @xmath190 proceeding from the bluest to reddest @xmath1 bin ( see also @xcite ) .
the errors in @xmath191 were calculated by perturbing many times the normalized composite spectra by the normalized error spectra , and measuring the depth of the absorption lines for each of these realizations .
the dispersion in the absorption line depths yielded the measurement uncertainty in @xmath191 , but of course does not include systematic errors associated with the specific model for the continuum level . the covering fractions of the low - ionization species are correlated with , but are systematically lower than , those derived from the @xmath5 lines ( lower right panel of figure [ fig : lowions ] ) .
there are several possible explanations for why the low - ionization lines suggest lower covering fractions .
first , if most of the metals inhabit narrow and unresolved absorption systems , then the depth of the metal lines will underpredict the covering fraction . likewise ,
if the metal - enriched gas has a substantial velocity gradient across the face of the galaxy , then the derived covering fractions may be lower limits ( e.g. , see discussion in section [ sec : evidencefornonunity ] ) . while we can not definitively rule out either of these possibilities , the similarity in the blueshift ( @xmath192kms@xmath15 ) and velocity spread ( fwhm@xmath193kms@xmath15 ; e.g. , compare figure [ fig : lowions ] to the right panel of figure [
fig : subtract ] ) of the low - ionization and @xmath5 lines suggests that the two components are kinematically coupled , where the former covers a smaller portion of the galaxy s continuum . such a configuration may arise if dense and metal - enriched pockets of gas are embedded within the outflowing @xmath5 .
consequently , some substantial fraction of the outflowing gas may be metal - poor .
additionally , we note that @xmath194 increases by a factor of @xmath195 in response to a small ( @xmath63 ) increase in @xmath2 .
this behavior suggests that as @xmath2 increases , a larger fraction of the neutral gas is enriched with , implying a corresponding increase in the average dust - to - gas ratio of the outflowing gas .
notwithstanding the fact that the line - of - sight reddening will be sensitive to the total column of dust ( including that of the ionized gas ) , an increase in the dust - to - gas ratio of the outflowing gas for objects with redder continuum would result in a strong correlation between @xmath55 and @xmath1 , consistent with what we actually observe ( left panel of figure [ fig : covfrac ] ) .
figure [ fig : highions ] shows the profiles of the two @xmath196 transitions for each bin of reddening . as the ionization potential for ( @xmath197ev ) is much higher than that required for h - ionization , the associated transitions arise from h - ionized gas .
the lines are blueshifted by @xmath198kms@xmath15 , slightly smaller than the blueshifts derived from the low - ionization and @xmath5 lines , though a more detailed comparison of the kinematics of the neutral , and low- and high - ionization gas will rely on future precise measurements of the systemic redshifts of galaxies in our sample .
the ratio of the equivalent widths of the lines is @xmath199 , indicating that the absorption is consistent within the errors of being on the linear part of the curve of growth . as the depths of the lines
do not correlate obviously with the reddening of the continuum , we surmise that the outflowing ionized gas has an optical depth @xmath200 .
note that the errors in @xmath201 do not rule the possibility of the lines being saturated , in which case the covering fraction of the ionized material is @xmath202 .
more generally , the presence of these high - ionization is absorption lines suggests an ionized medium mixed with metals and dust . to summarize , we find that the absolute escape fraction of ly@xmath6 photons , @xmath177 , inversely correlates with @xmath2 , suggesting that the neutral gas covering fraction plays a significant role in modulating the escape of ly@xmath6 photons .
the low - ionization interstellar doublet @xmath203 is saturated , exhibits similar absorption depths and velocity profiles to those of other low - ionization is absorption lines , and suggests a covering fraction of low - ionization material , @xmath191 , that increases with continuum reddening ( and @xmath2 ) from @xmath204 for the bluest galaxies in our sample to @xmath205 for the reddest ones .
however , @xmath191 is systematically lower than @xmath2 , while the low - ionization lines have similar blueshifts and velocity profiles as the @xmath5 lines , suggesting that galaxy outflows consist in part of a neutral medium embedded with clumps of metal - enriched gas .
moreover , the rate of increase of @xmath191 relative to @xmath2 implies that the average column of dust increases with covering fraction , consistent with our observation that the line - of - sight reddening correlates with continuum reddening .
these results imply that a significant fraction of the neutral medium may be chemically pristine , and that low - ionization interstellar absorption lines may severely underpredict the neutral gas covering fraction .
a similar analysis for the high - ionization absorption lines , namely @xmath196 , suggests that the high - ionization gas is not optically thick and has @xmath202 .
lyman continuum ( lyc ) photons may be extinguished by dust absorption or through hydrogen photoelectric absorption . in the case of dust absorption
, our new measurement of the far - uv shape of the dust curve at high redshift , presented in paper i , implies a lower attenuation of lyc photons than those inferred from simple extrapolations of the @xcite and @xcite attenuation curves .
alternatively , the high column densities ( @xmath68}\ga 20.3 $ ] ) inferred from the fitting presented in section [ sec : specfitting ] suggest that the covered portions of a galaxy are optically thick to lyc radiation , with implied optical depths at @xmath206 of @xmath207 . thus , for the average galaxy , photoelectric absorption , rather than dust , dominates the depletion of lyc photons .
similar results have been suggested by simulations of escaping ionizing radiation from high - redshift galaxies @xcite .
accordingly , in the geometry illustrated in figure [ fig : cartoon ] , lyc photons will only escape from dust - free holes in the ism .
the very high optical depth of the covering neutral gas suggests that ly@xmath6 photons will also predominantly escape from regions cleared of dust and gas , an expectation that is verified by the strong correlation between @xmath177 and reddening / covering fraction ( section [ sec : lyaescape ] ) .
lyc extinction may be a more significant effect in galaxies where the radiation field is sufficient to completely ionize the hydrogen ( i.e. , corresponding to a `` density - bounded '' case ; see @xcite ) , but such galaxies must be relatively rare in our sample given the apparent frequency of high column density neutral gas as reflected in the composite spectra .
we can connect the gas covering fraction to the absolute escape fraction of lyc photons , where the latter is commonly written as : @xmath208 \times \nonumber \\ & & 10^{-0.4k({\rm lyc}){e(b - v)}}.\end{aligned}\ ] ] the first term in this expression accounts for the opacity of the ism , where @xmath209 is the optical depth at lyc wavelengths .
the second term accounts for the obscuration of lyc photons by dust , where @xmath1 is the reddening and @xmath210 is the dust attenuation curve evaluated at lyc wavelengths . in the case where the dust and neutral gas completely cover the continuum , the opacity of the ism can be combined with the dust attenuation curves calculated in paper i to compute @xmath211 . in the case of incomplete covering ,
the absolute escape fraction of lyc photons can be written as : @xmath212 \times \nonumber \\ & & 10^{-0.4k({\rm lyc}){e(b - v)}^{\rm cov}_{\rm los } } + \nonumber \\ & & ( 1-f_{\rm cov}({\text{\ion{h}{1}}}))\exp[-\tau_{\rm ism}^{\rm uncov}({\rm lyc } ) ] \times \nonumber \\
\nonumber \\ & & 10^{-0.4k({\rm lyc}){e(b - v)}^{\rm uncov}_{\rm los } } , \label{eq : twoterms}\end{aligned}\ ] ] where the first and second terms correspond to the covered and uncovered portions contributing to the galaxy spectrum , respectively . in this expression
, @xmath210 is a line - of - sight extinction curve evaluated at lyc wavelengths , and @xmath213 and @xmath214 indicate the line - of - sight reddening appropriate to the covered and uncovered continuum , respectively . similarly , the absolute escape fraction of uv continuum photons ( e.g. , typically measured between @xmath172 and @xmath23 ) is @xmath215 where @xmath216 refers to the attenuation curve relevant for the stellar continuum , assuming a unity covering fraction of dust . as two terms that refer to the covered and uncovered portions of the continuum , analogous to equation [ eq : twoterms ] : @xmath217 which , in the case of low dust column density for the uncovered portion , can be written as @xmath218 where @xmath216 now refers to a line - of - sight extinction curve . as the color excess is measured most conveniently assuming a unity covering fraction of dust with a stellar attenuation curve , these equations are of less practical use than equation [ eq : fuv1 ] . ]
these equations are cast most conveniently in terms of the `` relative '' escape fraction defined as @xmath219 @xcite .
the relationship between the relative escape fraction and the _ observed _ ratio of lyc - to - uv flux density is @xmath220\times \exp[\tau_{\rm cgm}({\rm cgm } ) ] , \label{eq : frel2}\end{aligned}\ ] ] where @xmath221 and @xmath222 are the observed flux densities of lyc and uv continuum photons , respectively ; @xmath223 and @xmath224 are the intrinsic flux densities of lyc and uv continuum photons , respectively , as typically estimated from stellar population synthesis models ; and the last two terms indicate the further depletion of lyc photons due to the opacity of the igm and cgm . combining equations [ eq : fuv1 ] , [ eq : frel1 ] , and [ eq : frel2 ] yields the following expression for the observed lyc - to - uv flux density ratio : @xmath225_{\rm obs } & = & \left [ \frac{s({\rm lyc})}{s({\rm uv})}\right]_{\rm int } \times \nonumber \\ & & \exp[-\tau_{\rm igm}({\rm lyc } ) ] \times \exp[-\tau_{\rm cgm}({\rm lyc } ) ] \times \nonumber \\ & & \left[\frac{f({\rm lyc})_{\rm abs}}{10^{-0.4k({\rm uv}){e(b - v)}}}\right ] , \label{eq : final1}\end{aligned}\ ] ] where @xmath211 is given by equation [ eq : twoterms ] . for the average @xmath0 galaxy in our sample , the first term of equation [ eq : twoterms ] is negligible owing to the large neutral gas column density and hence large value of @xmath226 , as noted above .
hence , if the uncovered portions of the galaxy continuum are characterized by very low gas and dust column densities , as would be expected in the simple geometry of figure [ fig : cartoon ] , then the absolute escape fraction of lyc photons is simply @xmath227 we noted in section [ sec : evidencefornonunity ] that @xmath228 may be lower than that indicated by equation [ eq : lycabs ] .
we discuss this further in section [ sec : applications ] . in the present context ,
equation [ eq : final1 ] becomes @xmath229_{\rm obs } & = & \left[\frac{s({\rm lyc})}{s({\rm uv})}\right]_{\rm int } \times \nonumber \\ & & \exp[-\tau_{\rm igm}({\rm lyc } ) ] \times \exp[-\tau_{\rm cgm}({\rm lyc } ) ] \times \nonumber \\ & & \left[\frac{1-f_{\rm cov}({\text{\ion{h}{1}}})}{10^{-0.4k({\rm uv}){e(b - v)}}}\right ] .
\label{eq : final1b}\end{aligned}\ ] ] in practice , @xmath2 is difficult to constrain in the absence of direct spectroscopy of the @xmath5 absorption lines e.g .
, as noted in section [ sec : isabslines ] , neutral gas covering fractions deduced from low - ionization is absorption lines may severely underpredict those derived from the @xmath5 lines .
thus , the utility of an empirical relation between @xmath2 and @xmath1 becomes obvious : the continuum reddening is relatively easy to measure , even with just a few broadband photometric filters covering the rest - frame uv . as discussed in section [ sec : correlations ] and shown in figure [ fig : covfrac ] and table [ tab : covfracfits ] , we find a significant correlation between @xmath2 and @xmath1 .
thus , a suitable functional relationship between these two variables can be used to cast equations [ eq : twoterms ] and [ eq : final1b ] in terms of @xmath1 . the functional form relating
@xmath2 and @xmath1 is the subject of the next subsection .
the middle and right panels of figure [ fig : covfrac ] show the relationship between @xmath2 and @xmath1 .
the pearson correlation coefficients when we assume the smc and milky way extinction curves in computing @xmath2 are @xmath230 and @xmath231 , respectively , with probabilities for no correlation between @xmath2 and @xmath1 of @xmath232 and @xmath233 , respectively . to uncover the functional form of @xmath2 vs. @xmath1
, we reddened the intrinsic template assuming different values of @xmath1 and the updated and original forms of the @xcite and @xcite attenuation curves . for each of these reddened templates ,
we determined which combination of the covering fraction and line - of - sight reddening assuming either the smc or milky way extinction curves provided the best match in the `` composite @xmath1 '' wavelength windows listed in table [ tab : waves ] , and one additional window covering the lyc region : @xmath234 .
the model expectations for various assumptions of the reddening and line - of - sight dust curves are shown in the middle and right panels of figure [ fig : covfrac ] . for clarity ,
not all of the combinations of attenuation / extinction curves are shown , as they all result in a similar functional behavior between @xmath2 and @xmath1 .
as expected , the covering fraction asymptotes to zero and unity for small and large values of the continuum reddening , respectively . motivated by this behavior , we adopted the following functional form for @xmath2 vs. @xmath1 : @xmath235 , \label{eq : functional}\end{aligned}\ ] ] where @xmath236 and @xmath40 are free parameters .
table [ tab : covfracfits ] summarizes the free parameters obtained by fitting the _ measured _ values of @xmath237 and @xmath1 for the updated @xcite attenuation curve and the smc extinction curve applicable to the continuum and line - of - sight reddening , respectively . to compute these fits
, we randomly perturbed the color excess and covering fraction measurements according to their respective errors and fit each of these realizations with equation [ eq : functional ] .
the average and standard deviation in the best - fit parameters among these realizations are the values reported in table [ tab : covfracfits ] .
the @xmath238 confidence intervals in the fits were computed as the @xmath141@xmath142 dispersion in the fits to the realizations at each value of @xmath1 .
these confidence intervals do not vary significantly over the range of @xmath1 probed in our study for different assumptions of the attenuation curve , as the fits are forced to go through the measured data .
finally , we can combine equations [ eq : final1b ] and [ eq : functional ] to obtain an expression for the observed lyc - to - uv flux density ratio in terms of @xmath1 for the geometry illustrated in figure [ fig : cartoon ] : @xmath239 \times \exp[-\tau_{\rm cgm}({\rm lyc } ) ] \rangle \times \nonumber \\ & & \left[\frac{\exp[a\times\langle{e(b - v)}\rangle^b]}{10^{-0.4k({\rm uv})\langle{e(b - v)}\rangle}}\right ] .
\label{eq : final2}\end{aligned}\ ] ] we have written this equation in terms of averages to emphasize that the relation applies for an ensemble of galaxies . adopting values for the igm and cgm opacities appropriate for the redshifts of interest allows us to use equations [ eq : twoterms ] , [ eq : final1 ] , and [ eq : final2 ] in several useful ways , as we discuss in section [ sec : applications ] . to uncover the extent to which the relationship between @xmath240_{\rm obs}$ ] and @xmath1 may change over a larger range of @xmath1 than accessible in the current sample
, we examined how several of the other factors relevant in the escape of lyc photons , namely @xmath41 and @xmath55 , vary with @xmath1 .
as noted in section [ sec : correlations ] and shown in figure [ fig : covfrac ] and table [ tab : covfracfits ] , several significant correlations are found between these factors and @xmath1 .
we fit @xmath241 vs. @xmath242 and @xmath153 vs. @xmath243 with linear functions .
the best - fitting intercepts and slopes , and their uncertainties , are summarized in table [ tab : covfracfits ] .
we used these linear fits to estimate @xmath41 and @xmath55 for galaxies with @xmath244 , and used equation [ eq : twoterms ] to calculate @xmath211 , and equation [ eq : final1 ] to calculate @xmath245_{\rm obs}$ ] . in the next section ,
we examine how @xmath245_{\rm obs}$ ] varies with continuum reddening . to calculate the predicted @xmath246 from our model ,
we first calculated the expected opacities of the igm and cgm by considering the neutral gas column density distribution in the ly@xmath6 forest at @xmath247 , as well as the excess @xmath5 absorption around @xmath248 galaxies at the same redshifts , as quantified in @xcite . using 1000 random realizations of these column density distributions , we calculate the attenuation of @xmath206 photons to be @xmath249 \times \exp[-\tau_{\rm cgm}(900)]\rangle = 0.37 $ ] for galaxies at @xmath250 , corresponding to the mean redshift of the 121 galaxies that contribute to the composite at @xmath70 .
the dispersion in the transmissivity among the 1000 realizations at this wavelength is @xmath251 , implying that the error in the mean transmissivity for the 121 galaxies is @xmath252 .
next , we use the observed relations listed in table [ tab : covfracfits ] to estimate @xmath2 , @xmath55 and @xmath41 , for different values of @xmath1 .
thus , equations [ eq : twoterms ] , [ eq : final1 ] and [ eq : final2 ] can be written purely in terms of @xmath1 , the line - of - sight extinction and attenuation curves evaluated at the wavelengths of lyc and uv photons , respectively , and the intrinsic lyc - to - uv flux density ratio . to examine the possible range of models given the constraints imposed by our data ,
we calculated an `` extrapolated '' model which assumes that the neutral gas is optically - thick to ionizing photons at all values of @xmath1 , and that the uncovered portions of the galaxies are effectively free of dust and gas . in this case , the model is given simply by equation [ eq : final2 ] . in the `` fiducial '' model
, we assumed that the neutral gas column density varies with @xmath1 according to the relations listed in table [ tab : covfracfits ] .
finally , in the `` fixed '' model , we assumed that the ratio @xmath253 is fixed , and that @xmath55 scales with @xmath1 according to the relations listed in table [ tab : covfracfits ] .
all three models were computed assuming the combination of the updated @xcite attenuation curve and either the smc or milky way extinction curves , and are shown in figure [ fig : models ] .
other combinations of the stellar attenuation and line - of - sight extinction curves yield quantitatively similar results ( see discussion below ) .
all models shown are computed by taking the uv luminosity density at @xmath172 and are cast in terms of the ratio of the observed and intrinsic 900-to-1500 flux density ratios . in all of the cases considered here , the extrapolated and fixed models are indistinguishable , because a fixed @xmath253 implies optically thick @xmath5 over the entire range of @xmath254 to @xmath255 shown in figure [ fig : models ] .
the fiducial model deviates from the other two at @xmath256 where the neutral gas becomes optically thin ( @xmath257@xmath163 ) , thus allowing the escape of additional lyc photons above the expectations from the extrapolated model ( i.e. , the covered portion of the galaxies continua , as illustrated in figure [ fig : cartoon ] , becomes optically thin ) .
for very blue @xmath1 , @xmath258 and the observed lyc - to - uv flux density ratio simply converges to the intrinsic value modulated by the combined igm and cgm opacities . for very red @xmath1 ,
the ratio of the observed and intrinsic lyc - to - uv flux density ratios increases to values larger than unity .
this can be explained as follows .
lyc photons escape the galaxies only through clear sightlines , and the modeling assumes that the dust attenuation of such photons in negligible along these sightlines .
thus , the only way in which the observed lyc flux density depends on the reddening is through the relationship between @xmath2 and @xmath1 . the absolute escape fraction of lyc photons , i.e. , @xmath54 , does not decrease as rapidly as the absolute escape fraction of non - ionizing uv photons as the covering fraction asymptotes to unity for @xmath259resulting in a steep rise of the model prediction of @xmath260/[s(900)_{\rm int}/s(1500)_{\rm int}]$ ] for @xmath261 .
there is negligible variation in the model curve shapes at @xmath262 for different assumptions of the stellar attenuation curve .
thus , in practice , one can simply assume a single model that is invariant to assumptions of the adopted attenuation curve .
on the other hand , the offset in the relation between @xmath2 and @xmath1 that arises from different assumptions of the line - of - sight extinction curve ( e.g. , figure [ fig : covfrac ] ) results in a factor of up to @xmath195 difference in the predicted @xmath263 at a given @xmath1 for @xmath264 . this systematic difference may be partly attributed to the lack of data for galaxies bluer than @xmath265 with which to constrain the fit between @xmath2 and @xmath1 . in the following ,
we discuss how these model curves may be used to evaluate the possible mechanisms of lyc escape and/or constrain the intrinsic ionizing spectrum of high - redshift galaxies .
we preface our discussion by noting that deep far - uv spectra of the kind used for our study would be useful to obtain over a broader dynamic range of @xmath1 , in particular focusing on the bluer galaxies that are more analogous to typical galaxies at @xmath266 and which may have larger ionizing escape fractions .
far - uv spectra may be used to directly measure the covering fraction of neutral gas at @xmath267 using the technique described in section [ sec : specfitting ] , and will enable more accurate estimates of the column densities and line - of - sight reddening for such galaxies .
nonetheless , in the following , we illustrate how such a model may be used to discern the mechanisms by which lyc photons escape from high - redshift galaxies .
the most direct application of our model is that it can be used to predict the average relative escape fraction of ionizing photons for a given stellar population model .
for example , if we assume @xmath268 , a value typical of that adopted in previous studies ( e.g. , @xcite ) , then a sample of galaxies at @xmath269 with @xmath270 would be predicted to have an observed relative escape fraction of @xmath271 .
the relative escape fraction corrected for attenuation by the igm and cgm is @xmath272 , and the absolute escape fraction is @xmath63 .
we emphasize that the model should only be applied to an ensemble of galaxies where the stellar populations and different configurations of the gas and stars from galaxy - to - galaxy , and the stochasticity in the ly@xmath6 forest along different sightlines , can be averaged out .
a second and potentially powerful application of our model is immediately obvious .
namely , with direct measurements of the average observed lyc - to - uv flux density ratio for an ensemble of galaxies with some average continuum reddening , @xmath273 , one may invert equation [ eq : final2 ] to calculate @xmath274 . the importance of direct empirical constraints on the intrinsic lyc - to - uv flux density ratio is underscored by the fact that the different flavors of stellar population models ( e.g. , arising from different ages and metallicities for a given star - formation history , and/or the treatment of stellar rotation / binarity ; @xcite ) predict a factor of @xmath88 variation in the intrinsic lyc - to - uv flux density ratios of @xmath275 ( e.g. , @xcite ) .
the actual range in the predictions of the stellar population models are larger than a factor of @xmath88 if we also consider variations in the stellar initial mass function ( imf ) and star - formation history .
clearly , independent information from the rest - frame uv continuum spectra , rest - frame optical nebular emission line spectra , and/or stellar population modeling may be used to narrow the range of possible @xmath276 @xcite . nevertheless , there is substantial utility in directly constraining the intrinsic lyc - to - uv flux density ratio without having to extrapolate the spectra of massive stars to wavelengths blueward of what is typically accessible from observations . to demonstrate how our model can be used to constrain the intrinsic lyc - to - uv flux density ratio , we considered the recent measurement of the relative escape fraction of galaxies in our sample by steidel et al .
( in prep ) . for the subsample of galaxies with deep lyc spectra where the sample has been cleaned of foreground contamination based on the ground - based spectroscopy
( see section [ sec : sample])these authors measured the ratio of the flux densities at @xmath206 and @xmath172 , @xmath277 . with this measurement
, we can invert equation [ eq : final2 ] to compute @xmath278 . to account fully for the measurement and random errors that enter into equation [ eq : final2 ]
, we performed a monte carlo simulation by perturbing each variable in the equation by its associated measurement or random uncertainty .
as noted above , the mean combined igm and cgm attenuation at the mean redshift of the 121 galaxies of @xmath279 is @xmath280 \times \exp[-\tau_{\rm cgm}({\rm lyc})]\rangle = 0.370 \pm 0.030 $ ] .
the mean reddening of these galaxies is @xmath281 , assuming the @xcite dust attenuation curve . for this attenuation curve , @xmath282
as calculated in paper i. we performed 10,000 simulations where , each time , we randomly perturbed the igm@xmath283cgm attenuation , the mean reddening , the dust attenuation curve , and the observed relative escape fraction according to normal distributions centered on the mean values with @xmath142 equal to the associated uncertainties .
in addition , we randomly perturbed @xmath2 and @xmath1 for each composite according to their respective uncertainties and fit equation [ eq : functional ] to these perturbed values .
we then calculated @xmath274 for each of the 10,000 trials .
the resulting distributions of @xmath278 are shown in figure [ fig : inthist ] .
we find @xmath284 and @xmath285 assuming the @xcite and @xcite attenuation curves , respectively ( with an smc line - of - sight extinction in both cases ) .
the simulations indicate that we can use our methodology to constrain the intrinsic 900-to-1500 flux density ratio to @xmath286 _ random _ error for the sample of galaxies considered here .
in addition to this random error , there are several sources of systematic error in our estimate of @xmath274 , including the choice of the line - of - sight extinction curve ( e.g. , smc or mw ) and the choice of the stellar reddening curve ( e.g. , @xcite or @xcite ) .
the magnitude of the bias in @xmath274 that stems from these effects is @xmath287 and @xmath288 , respectively .
additional systematic uncertainty may arise from the choice of igm prescription .
for example , @xcite find a mean igm transmission of @xmath289 for @xmath206 photons emitted from a source at @xmath269 , a value at the high end of published measurements . on the other hand ,
the _ combined _
igm+cgm opacity predicted by @xcite is @xmath290 .
this suggests a systematic uncertainty associated with the igm prescription of @xmath291 .
the largest source of systematic uncertainty relates to our inability to fully resolve the @xmath5 absorption .
specifically , our estimate of @xmath2 may be a lower limit as a result of several effects , some of which are discussed in section [ sec : evidencefornonunity ] .
first , the relatively low spectral resolution of our data may mask the presence of narrow absorption line systems with a higher covering fraction than that deduced from the spectra .
second , if the different velocity components of the optically - thick @xmath5 are not covering the same lines - of - sight , then the integrated covering fraction may be larger than the value found for the bulk of the absorbing gas .
third , we are unable to discern whether non - negligible columns of @xmath5 and dust exist in regions that we would otherwise refer to as `` clear '' sightlines .
fourth , the ionized gas may not be completely bounded by @xmath5 ( i.e. , the `` average '' configuration of the @xmath5 and may be intermediate between the radiation- and density - bounded scenarios ) . in this latter case ,
attenuation by dust in the ionized gas may deplete lyc photons before they escape the galaxy .
for these reasons , the _ effective _ covering fraction , @xmath292 , may be larger than the value derived from the composite spectrum and hence the calculated value of @xmath274 may be a lower limit . to further evaluate the range of possible @xmath274 allowed by the data , we compared our estimate with the predictions of several stellar population synthesis models ( figure [ fig : stelpop ] ) .
we considered @xcite @xmath293@xmath294 and @xmath294 models with a @xcite imf ( power law index of @xmath295 ) over the mass range @xmath296@xmath297 ; and the most recent version of the starburst99 ( s99 ) models which adopt weaker stellar winds relative to the the previous version @xcite .
we generated s99 models using the @xcite imf with different power law slopes of @xmath298 , @xmath299 , and @xmath300 , over the mass range @xmath301@xmath297 , and with a metallicity of @xmath302@xmath294 .
lastly , we consider the `` binary population and spectral synthesis '' ( bpass ; @xcite ) models that include binary evolution , with a power law slope of the imf of @xmath303 over the mass range @xmath304@xmath297 , with metallicities of @xmath152 and @xmath302@xmath294 .
all metallicities assume the solar abundances measured in @xcite . the effect of binary evolution ( as in the bpass models ) is to allow efficient mass transfer from a massive star to its ( secondary ) companion , causing the latter to spin up quickly and , for a low stellar metallicity where the stellar wind is sufficiently weak , retain its angular momentum @xcite .
consequently , the secondary becomes fully mixed and undergoes quasi - homogeneous evolution @xcite whereby the main sequence lifetime and effective temperature are increased . in such models ,
the ionizing photon flux per non - ionizing uv photon flux may increase by up to @xmath305 relative to single star evolution ( e.g. , @xcite ; see also @xcite ) .
a similar effect can be achieved by decreasing the power law index of the imf , which results in a larger mass fraction in more massive , hotter stars .
figure [ fig : stelpop ] shows the limit on @xmath274 based on @xmath306 derived from the composite spectrum of all 933 galaxies in our sample .
our inference of the gas covering fraction favors an intrinsic lyc - to - uv flux density ratio that is larger than the value predicted from the @xcite stellar population models , and one that is consistent with those obtained with the newer versions of the s99 models that adopt weaker massive star stellar winds , a low stellar metallicity , and/or a flatter slope of the imf ; or those models that include the effects of binary evolution . for the reasons stated above
, the covering fractions derived here are likely lower limits , and therefore it is instructive to gauge the degree to which the covering fraction may be higher and still be consistent with the predictions of stellar population synthesis models . in particular , if the effective covering fraction is @xmath307 or @xmath308 , instead of @xmath117 , then the inferred @xmath278 increases from @xmath309 to @xmath310 and @xmath311 , respectively .
note that the latter is @xmath312 larger than the value predicted by the low - metallicity bpass models and/or the low - metallicity and shallow-@xmath6 imf s99 models .
thus it is apparent , though ironic , that it is those galaxies with the largest covering fractions ( as is the case for the average galaxy at @xmath0 ) which provide the most sensitive constraints on the intrinsic lyc - to - uv flux density ratio , since even small increases in @xmath2 will lead to very large changes in the inferred @xmath278 ( figure [ fig : stelpop ] ) .
@xcite suggest that stellar population models that include the effects of binary evolution are required to simultaneously reproduce the rest - frame far - uv continuum , stellar , and nebular features , and the rest - frame optical nebular emission line strengths measured for galaxies at @xmath313 .
if we adopt such models as the most appropriate , then it implies that our estimate of @xmath96 for typical galaxies at @xmath0 is not biased low by more than @xmath137 .
stated another way , the combined effects of additional @xmath5 absorption and dust attenuation can not deplete more than @xmath314 of the photons exiting through what we would otherwise refer to as `` clear '' sightlines based on the depths of the @xmath5 lines in the composite spectra .
_ this analysis underscores the fact that if the simple geometry of figure [ fig : cartoon ] is valid on average for an ensemble of galaxies , then deep uv spectroscopy for large samples of typical star - forming galaxies at high redshift should enable tight constraints on the intrinsic lyc - to - uv flux density ratio . _
hence , our framework provides an alternative and powerful independent method of constraining the intrinsic lyc - to - uv flux density ratio , which is currently one of the most uncertain parameters entering into the calculation of the absolute escape fraction of ionizing photons in high - redshift galaxies .
the geometry depicted in figure [ fig : cartoon ] may not apply for all galaxies , and in reality more complicated configurations of the dust , gas , and stars may be expected . in this context , the model described here provides a useful benchmark to investigate changes in geometry and hence alternate pathways for the escape of ionizing radiation . to illustrate this , we show in figure [ fig : devgeo ] @xmath315 as a function of @xmath1 assuming the fiducial model for the smc extinction curve .
we have assumed @xmath316 , a value typical of the bpass models and the s99 models with a flat imf slope , and igm and cgm opacities relevant for galaxies at z=3.05 .
direct measurements of @xmath315 and @xmath273 for an _ ensemble _ of galaxies may be used to ascertain the mechanisms by which lyc photons escape galaxies .
for example , if these average values lie in the _ green _ region , then it suggests that , at least on average , those galaxies have either a smaller intrinsic 900-to-1500 flux density ratio than the one we have adopted here , or that the `` uncovered '' portions of the continuum are not completely free of dust and/or gas .
similarly , objects whose average values place them in the _ purple _ region are ones that may have a larger intrinsic 900-to-1500 flux density ratio or optically - thin neutral gas that allows the escape of additional lyc photons .
finally , galaxies whose average values lie in the _ cyan _ region _ must _ have a larger intrinsic 900-to-1500 flux density ratio on average than the one assumed here .
further discrimination of the mechanisms by which lyc photons escape galaxies , or of the value of the intrinsic lyc - to - uv flux density ratio , can be accomplished with deep spectroscopy , such as that presented here . from this spectroscopy
, we have found that typical @xmath0 galaxies can be characterized by optically - thick outflowing neutral @xmath5 , a finding that already narrows the range of possible avenues by which ionizing radiation can escape these galaxies .
in this paper , we have modeled the rest - frame far - uv composite spectra of 933 @xmath317 lbgs at @xmath0 , 121 of which have very deep spectroscopic covering at @xmath318 , in order to investigate the relationship between uv reddening , neutral gas covering fraction , and the ionizing photon escape fraction .
this analysis is facilitated by our large sample of galaxies which allows us to average over stochastic variations in the ly@xmath6 forest along different sightlines to @xmath0 galaxies , and thus securely analyze the @xmath5 absorption lines in these galaxies .
our main conclusions are as follows .
* the composite spectra of @xmath0 galaxies show evidence for a partial covering of optically - thick outflowing neutral hydrogen .
specifically , the gas is ( a ) blueshifted by @xmath319kms@xmath15 ; ( b ) exhibits damping wings in the lyman series lines , including ly@xmath6 , signifying high - column density , @xmath320 \ga 20.3 $ ] ; and ( c ) lyman series lines including ly@xmath56 , ly@xmath57 , and ly@xmath61 that do not reach zero flux density at the line cores , implying the presence of unabsorbed stellar continuum .
this result is robust against spectral resolution and redshift uncertainties , and is based upon a careful isolation of foreground contaminants . * to investigate trends between reddening and gas covering fraction
, we divided our sample into bins of @xmath1 ( measured from the stellar continuum ) and constructed composites for galaxies in each of these bins .
we modeled the composite spectra , allowing for a non - unity covering fraction of @xmath5 and dust a configuration that arises from an increase in ism porosity due to stellar feedback processes ; and including both stellar and nebular continuum emission , as well as unresolved absorption from the @xmath34 lyman - werner bands .
the derived @xmath5 covering fractions vary from @xmath321 to @xmath322 , over a range of continuum reddening from @xmath323 to @xmath324 .
the modeling results imply that galaxies with redder uv continua have a larger covering fraction of @xmath5 and dust , with a higher line - of - sight dust column density .
* the absolute escape fraction of ly@xmath6 emission , as measured through a spectroscopic slit , correlates inversely with neutral gas covering fraction .
this suggests that the ly@xmath6 photons escape preferentially through sightlines that have been cleared of gas and dust .
separately , the neutral gas covering fractions deduced from saturated low - ionization interstellar absorption lines , such as @xmath203 , correlate with , but systematically underpredict by a factor of @xmath325 , those derived from the @xmath5 lines .
thus , inferences of gas covering fractions based on low - ionization interstellar absorption lines must be viewed with caution . moreover
, this result suggests that a significant fraction of outflowing @xmath5 may be metal - poor . the fact that the `` covering fractions '' inferred from the low - ionization interstellar absorption lines appear to rise rapidly with a small increase in @xmath2 , together with our finding that the line - of - sight extinction increases in tandem with the reddening of the stellar continuum , suggests that the outflowing material becomes progressively enriched with metals for galaxies with larger @xmath2 and redder uv continua .
* we discuss our results in the context of the escape of ionizing radiation at high redshift .
the partial covering fraction of optically - thick @xmath5 implies that the depletion of ionizing photons is dominated by photoelectric absorption , rather than dust attenuation , and that ionizing photons escape typical galaxies at @xmath0 through clear sightlines in the ism . * using a physically - motivated functional form for the relationship between continuum reddening and neutral gas covering fraction , we establish a relationship between the absolute and relative escape fraction of ionizing radiation and continuum reddening .
our model may be used to predict the average escape fractions of ionizing photons for an ensemble of galaxies with some average @xmath1 and an assumed intrinsic lyc - to - uv flux density ratio . alternatively , direct measurements of the ionizing escape fraction may be used to place constraints on the intrinsic lyc - to - uv flux density ratio . based on a recent measurement of @xmath326 for the subsample of galaxies with deep far - uv spectra analyzed here , we infer @xmath327 .
this lower limit is larger than the prediction from the @xcite models , and generally favors stellar population models that include weaker stellar winds , a flatter slope of the imf , and/or binary evolution .
we have demonstrated how our model may be used to provide powerful constraints on the intrinsic ionizing spectrum of high - redshift galaxies and , additionally , how galaxies with large covering fractions and associated large reddening , may be used as sensitive probes of the intrinsic lyc - to - uv flux density ratio . aside from providing constraints on the production efficiency of ionizing photons , we have discussed how our model may also be used to assess the pathways by which such photons escape galaxies .
our analysis points to several future avenues of investigation .
coupling deep rest - frame uv spectroscopy of high - redshift galaxies ( such as that presented here ) with multi - band _ hst _ imaging will allow us to construct the cleanest possible sample of galaxies uncontaminated by foreground interlopers based on high spatial resolution multi - component photometric redshift analysis @xcite and the identification of spectroscopic blends with which to measure the ionizing escape fraction . the high spatial resolution provided by _
hst _ has other potential benefits in terms of clarifying the escape of ionizing radiation .
specifically , a consequence of a partial covering fraction of @xmath5 is an inhomogeneous spatial distribution of lyc emission relative to the non - ionizing uv continuum . while resolving individual sightlines in these high - redshift galaxies is far beyond _ hst _
s capabilities , a `` clumpy '' distribution of lyc emission that is resolved with _
hst _ to be disjoint from the morphology of the non - ionizing uv continuum may still indicate the signature of a porous ism where lyc photons only escape through clear lines - of - sight .
more precise measurements of , or limits on , the neutral gas covering fractions may be achieved with higher resolution spectroscopy for large samples of galaxies , possible with the next - generation of instruments and @xmath328 m telescopes . in the more immediate future
, rest - frame optical nebular emission line spectroscopy will enable systemic redshift measurements that can be used to analyze the @xmath5 line profiles from composite spectra in a way that is less affected by redshift uncertainties .
the combination of rest - frame optical nebular spectroscopy , rest - frame uv continuum spectroscopy , and our model for the escape of ionizing radiation together will make it possible to construct a consistent picture of the massive and ionizing stellar populations in high - redshift galaxies .
lastly , the models presented here will benefit greatly from future deep spectroscopy of bluer galaxies and/or those detected with strong ly@xmath6 emission to extend the dynamic range of continuum reddening probed in our study . with such spectroscopy
, we will be in a position to evaluate the column densities , line - of - sight reddening , and covering fractions for the types of galaxies that are likely to contribute significantly to the budget of ionizing photons at high redshift .
the framework presented in this paper will be central to the interpretation of the ionizing escape fractions of galaxies at the reionization epoch , where the prohibitively thick igm precludes the detailed far - uv measurements presented here .
nar is supported by an alfred p. sloan research fellowship , and acknowledges the visitors program at the institute of astronomy in cambridge , uk , where part of this research was conducted .
ccs acknowledges nsf grants ast-0908805 and ast-1313472 .
mb acknowledges support of the serbian mestd through grant on176021 .
nar acknowledges gwen rudie for useful discussions .
we are grateful to the anonymous referee whose comments led to significant improvements in the clarity , content , and presentation of the analysis and results .
we wish to extend special thanks to those of hawaiian ancestry on whose sacred mountain we are privileged to be guests . without their generous hospitality , the observations presented herein would not have been possible .
|
using a large sample of spectroscopically confirmed lyman break galaxies at @xmath0 , we establish for the first time an empirical relationship between reddening ( @xmath1 ) , neutral gas covering fraction ( @xmath2 ) , and the escape of ionizing ( lyman continuum , lyc ) photons at high redshift .
our sample includes 933 galaxies at @xmath0 , 121 of which have very deep spectroscopic observations ( @xmath3hrs ) in the rest - frame far - uv ( @xmath4 ) with the low resolution imaging spectrograph on the keck telescope .
based on the high covering fraction of outflowing optically - thick @xmath5 indicated by the composite spectra of these galaxies , we conclude that photoelectric absorption , rather than dust attenuation , dominates the depletion of lyc photons . by modeling the composite spectra as the combination of an unattenuated stellar spectrum including nebular continuum emission with one that is absorbed by @xmath5 and reddened by a line - of - sight extinction
, we derive an empirical relationship between @xmath1 and @xmath2 .
galaxies with redder uv continua have larger covering fractions of @xmath5 characterized by higher line - of - sight extinctions .
the absolute escape fraction of ly@xmath6 correlates inversely with @xmath2 , consistent with the escape of ly@xmath6 through gas- and dust - free lines - of - sight .
gas covering fractions based on low - ionization interstellar absorption lines systematically underpredict those deduced from the @xmath5 lines , suggesting that much of the outflowing material in these galaxies may be metal - poor .
we develop a model which connects the ionizing escape fraction with @xmath1 , and which may be used to estimate the ionizing escape fraction for an ensemble of high - redshift galaxies .
alternatively , direct measurements of the escape fraction for the composite spectra allow us to constrain the intrinsic ionizing - to - non - ionizing flux density ratio to be @xmath7 , a value that favors stellar population models that include weaker stellar winds , a flatter initial mass function , and/or binary evolution .
lastly , we demonstrate how the framework discussed here may be used to assess the pathways by which ionizing radiation escapes from high - redshift galaxies .
|
a family @xmath1 of hypersurfaces of @xmath2 is said to move by mean curvature flow if there is a map @xmath3 such that @xmath4 is an immersion for all @xmath5 with @xmath6 and @xmath7 solves @xmath8 where @xmath9 is a @xmath10-dimensional manifold , @xmath11 is the mean curvature of @xmath12 and @xmath13 its normal , such that @xmath14 is the mean curvature vector .
if @xmath15 for a family of functions @xmath16 , then @xmath12 moves by mean curvature flow if and only if @xmath17 solves the graphical mean curvature flow equation , which is the parabolic partial differential equation @xmath18 graphical mean curvature flow was studied in @xcite by ecker and huisken .
they proved long - time existence for the mean curvature flow of entire graphs and showed that the solution stays graphical for all time .
more recently , sez trumper and schnrer proved in @xcite a long - time existence result for complete graphs .
starting from an open set @xmath19 and a proper function @xmath20 , they showed the existence of a solution @xmath17 to graphical mean curvature flow with initial data @xmath21 , where @xmath22 is defined on an open set @xmath23 for @xmath24 .
this solution will not develop singularities on a finite level but it can disappear to infinity forming a singularity at infinity - level .
it was observed that the sets @xmath25 can be interpreted as a weak solution to mean curvature flow , starting from @xmath26 , and that it coincides almost everywhere with the level - set flow as long as the latter does not fatten ( fig.[graphball ] ) .
( 10,5 ) ( 2,0 ) ( 5.5,4)graph ( 5.6,0.4)shadow in @xcite and existence results analogous to that of sez and schnrer are proven for some fully nonlinear flows of non - compact convex surfaces . in this article
we consider graphical mean curvature flow with a dirichlet boundary condition .
the existence result of this article reads as follows .
[ thm existence ] let @xmath27 be open , smooth , and mean convex , that is , its boundary @xmath28 has nonnegative mean curvature @xmath29\geq 0 $ ] .
let @xmath30 $ ] be continuous and assume @xmath21 is locally lipschitz in the set @xmath31 and @xmath32 is of class @xmath33 in @xmath34 .
then there is a continuous function @xmath35 $ ] which is smooth on @xmath36 and solves @xmath37 where @xmath38 .
the condition of nonnegative mean curvature for the boundary @xmath39 is necessary to expect longtime existence for the graphical mean curvature flow when considering the dirichlet problem for general boundary data .
otherwise , the gradient of @xmath17 may become unbounded in finite time and the flowing surface may cease to be graphical . note that for @xmath40 theorem [ thm existence ] is a generalization of the result in @xcite because we do not need the assumptions @xmath21 proper and @xmath41 .
the proof of theorem [ thm existence ] is the subject of section [ sec existence ] .
it involves an approximation of the problem by bounded auxiliary problems and uses an arzel - ascoli - argument and a priori estimates to pass to a limit . in section [ sec shadow - flow ] , we define the `` shadow '' at time @xmath42 to be the set @xmath43 and interpret this as a weak solution to mean curvature flow of hypersurfaces in @xmath44 with dirichlet boundary condition on @xmath39 using the notion of avoidance principle as defined in section [ sec shadow - flow ] .
roughly , the definition of weak solution implies that any classical solution starting inside the shadow stays inside and any classical solution starting outside stays outside .
an intersection of two smooth open sets is in general not smooth at the intersection of the boundaries . in section
[ sec smoothing ] we provide a lemma to locally mollify intersections of smooth open sets at the intersections of their boundaries while preserving certain curvature conditions .
[ thm smoothing ] let @xmath45 be open , @xmath46 and suppose @xmath47 is bounded .
then for any @xmath48 there is an open set @xmath44 with @xmath49 such that @xmath50 where @xmath51 .
moreover , if the principal curvatures at every point of @xmath52 and @xmath53 lie in a symmetric , open or closed , convex cone @xmath54 which contains the positive cone @xmath55 , then @xmath44 can be chosen such that the principal curvatures at any point of @xmath39 lie in @xmath56 , too .
here symmetric means invariant when interchanging @xmath57 .
choosing @xmath58 corresponds to convex subsets .
choosing @xmath59 corresponds to mean - convex domains .
this is what we use in section [ sec existence ] .
the problem of mollifying inside convex curvature cones was an open problem from the problem section of the conference _ geometric evolution equations _ which took place in konstanz in 2011 .
the proof uses distance functions to the boundaries and a mollified version of the minimum of these .
it can be read independently of the other sections and is applied most noteably in section [ sec existence ] in the approximation process .
the author expects that this result is of great use and will be widely applicable .
this article emerged from the author s master thesis and is meant to gather the main results .
the author wishes to thank oliver schnrer for supervising the thesis and for his great support .
the author also likes to thank ben lambert for helpful advice .
we prove the existence result theorem [ thm existence ] by approximating by auxiliary problems and using a priori estimates .
we cut off the initial function @xmath21 in height by considering @xmath60 for @xmath61 , where @xmath62 and @xmath63 are mollified versions of @xmath64 and @xmath65 respectively ( defined analogously to setting @xmath66 there ) .
next , we cut off the domain of definition @xmath44 by intersecting with a ball @xmath67 and using theorem [ thm smoothing ] .
this gives smooth open sets @xmath68 whose boundaries have nonnegative mean curvature @xmath69\geq 0 $ ] . finally , we restrict the functions @xmath70 to @xmath71 and take a mollification to find smooth functions @xmath72 defined on @xmath71 satisfying @xmath73 and @xmath74 .
now define @xmath75 as the solution of the auxiliary problem @xmath76 by ( * ? ? ?
* theorem 2.1 ) , @xmath75 is well - defined and satisfies @xmath77 ( by de giorgi - nash - moser - estimates the spatial derivative @xmath78 is even hlder - continuous for some exponent . ) to be able to utilize the arzel - ascoli - theorem and to pass to a limit and obtain a solution of the initial problem , we are going to need local a priori estimates .
these estimates will be local in space , time , and height : due to the unboundedness in height we can only expect estimates at points @xmath79 depending on @xmath80 . since by lemma [ lem gradient bounds ] and lemma [ lem sup bound ] we have local gradient bounds at the boundary , local gradient estimates easily follow from the results of section 2 in @xcite . using spheres as barriers
we obtain hlder - estimates in time with exponent @xmath81 ( cf .
section 6 of @xcite ) .
this is sufficient to apply an arzel - ascoli argument . to use arzel - ascoli for unbounded functions , simply compose with a homeomorphism @xmath82\to [ -1,1]$ ] which is smooth on @xmath83 .
then the gradient and hlder - in - time estimates give locally uniform estimates for @xmath84 where @xmath85 , which is sufficient : locally in space - time there is for any @xmath48 a @xmath86 such that for almost all @xmath87 we have @xmath88 by arzel - ascoli a subsequence of @xmath89 converges locally uniformly to a continuous function on @xmath90 as @xmath91 .
this correspondes to pointwise convergence of a subsequence of @xmath92 to a continuous function @xmath35 $ ] and locally uniform convergence on @xmath93 .
then , using the interior estimates for higher derivatives in @xcite , one has locally smooth convergence on @xmath94 and we see that @xmath17 solves the same differential equation there as the @xmath75 , which completes the proof .
now we are going to establish the a priori estimates .
most importantly we need [ lem gradient bounds ] let @xmath44 be as in the statement of theorem [ thm existence ] , @xmath95 , @xmath96 , @xmath97 and @xmath98 .
let @xmath99)$ ] be a solution of @xmath100 for the initial- and boundary data assume @xmath101 and @xmath102
. assume @xmath103)$ ] .
then on @xmath104 $ ] we have @xmath105 a nonlocal version of this result can be found , for example , in ( * ? ? ?
* chapter 1.4 ) .
there , one uses the barriers @xmath106 where @xmath107 is the signed distance function to @xmath39 and the constants are chosen like @xmath108 .
this barrier works on an @xmath109-neighbourhood of @xmath39 , where @xmath109 is chosen as @xmath110 and so small , that we are in a tubular neighbourhood . because of @xmath111 on the boundary we find the desired gradient bound on the boundary by the comparison principle for dirichlet boundary conditions . to obtain a local version simply add @xmath112 to the barriers , where @xmath113 satisfies @xmath114 , @xmath115 on @xmath116 and @xmath117 on @xmath118 .
the proof then works as in the nonlocal case but the choice of @xmath119 and @xmath120 depends additionally on @xmath121 because the bounds on derivatives of @xmath122 depend on @xmath121 . to apply lemma [ lem gradient bounds ]
we need local @xmath123-estimates on @xmath17 . for this purpose
we have the following [ lem sup bound ] let @xmath44 be as before , @xmath124 and @xmath125 , @xmath126 and @xmath127 defined as in lemma [ lem gradient bounds ] .
let @xmath128)$ ] be a solution of on @xmath129 .
then there exists @xmath130 dependent on @xmath131 and @xmath132 , such that @xmath133 where @xmath134 depends on the same quantities as @xmath121 and additionally depends on @xmath135 . without loss of generality
we may assume @xmath136 and that the inner normal to @xmath39 at @xmath125 is @xmath137 .
we are going to write the boundary locally as a graph : there is an open ball @xmath138 in @xmath139 with center at the origin and @xmath140 , such that @xmath141 and @xmath142 for some function @xmath143 .
because of @xmath29\geq 0 $ ] , @xmath144 satisfies the differential inequality @xmath145 take @xmath146 , @xmath147 , such that @xmath148 fulfils @xmath149 .
finally , define @xmath150)$ ] to be the solution of @xmath151 this way we have @xmath152)$ ] . by the maximum principle we also have @xmath153 , and by the strong maximum principle @xmath154 holds for all @xmath155 and @xmath156 .
define @xmath157 by @xmath158 and choose @xmath130 from our assertion such that @xmath159 and @xmath160 has positive distance to @xmath161 .
now we construct a barrier on @xmath157 with the aid of the function @xmath162 . on @xmath157
define @xmath163 setting @xmath164 for @xmath165 .
one easily verifies that the level - sets of @xmath166 move by mean curvature , so that @xmath166 solves @xmath167 using this fact we calculate @xmath168 where @xmath169 is a constant that ultimately only depends on @xmath170 and @xmath132 through the above construction .
finally , @xmath171 are upper and lower barriers for @xmath17 on @xmath157 respectively . observe @xmath172.\ ] ] now , on @xmath173 $ ] , @xmath174 and therefore @xmath175 are bounded by a constant as in the assertion .
it is worth pointing out the solution is not unique in general : consider two so called _ grim reaper curves _ lying next to each other , i.e. the graph of the function @xmath176 we can write down a translating solution to with initial data @xmath21 : @xmath177 but the solution @xmath17 we have constructed in the proof of theorem [ thm existence ] differs from this translating solution @xmath178 : the two grim reapers get connected at infinity .
this is because we cut off the function @xmath21 at some height @xmath179 in the approximation process . to see that the approximating solutions @xmath75 do not converge to @xmath178
we may consider the integral of the difference : @xmath180 where @xmath181 are the respective curvatures . by the maximum principle the convergence @xmath182 is monotone and thus @xmath183 as @xmath184 .
we conclude @xmath185 and therefore @xmath186 .
in this section we are going to investigate the projections / shadows of graphical mean curvature flow , that is the sets @xmath187 where @xmath17 is as in theorem [ thm existence ] .
we show that this shadow is a weak solution of mean curvature flow , where we use the following notion of weak solution which is based on the avoidance principle . for this section we do not need to assume constant dirichlet boundary values .
let @xmath44 be as in the last section .
a family @xmath188 of open subsets of @xmath189 is called * a supersolution to mean curvature flow if the following holds : for any family @xmath190}$ ] of open sets , such that @xmath191 with @xmath192 , and such that @xmath193}$ ] is a classical solution to mean curvature flow , we have @xmath194 * a subsolution to mean curvature flow , if @xmath195 is a supersolution . * a weak solution of mean curvature flow , if @xmath188 is both a super- and subsolution . we shall call @xmath196 the boundary values of @xmath188 . in the above setting in the definition of supersolution we even have @xmath197 for all @xmath198 $ ] , when @xmath199 is a supersolution .
to see that the closure is contained one can use , for instance , the translation invariance of classical flows .
if @xmath200 is a family of open subsets of @xmath189 such that @xmath201 is a classical solution to mean curvature flow , then , by the avoidance principle , @xmath200 is a weak solution .
let @xmath202 and suppose that @xmath203 is smooth in a ball @xmath204 . if @xmath200 is a weak solution and @xmath203 is smooth in a spacetime - neighbourhood of @xmath79 , then @xmath201 solves mean curvature flow at @xmath79 classically . we may assume , that @xmath205 is the graph of a smooth function . using the result of the next section we find two smooth open sets , one lying inside @xmath206 , the other inside @xmath207 , and such that their boundaries coincide in a neighbourhood of @xmath208 with @xmath203 .
taking @xmath121 to be small enough , small translations of these two open sets away from @xmath203 serve as barriers . this way it
can be seen , that the normal velocity of @xmath203 at @xmath79 coincides with the mean curvature at that point .
similar weak notions of mean curvature flow are the set - theoretic subsolutions of ilmanen ( @xcite ) or more generally the barriers of de giorgi .
both of them were compared to the level - set flow ( see @xcite for a comparison of de giorgi s barriers to level - set flow ) .
( see @xcite for a definition of set - theoretic subsolutions including boundary values . )
note that our definition of weak solutions is not very useful where @xmath209 ( taking the boundary @xmath203 relative to @xmath189 ) .
this is because there is no space left for a classical solution , that could possibly push @xmath203 inwards into @xmath44 . to circumvent this ,
one could compare with classical solutions with boundary values that may not be written as the boundary of an open set and which can intersect @xmath203 .
but this would cause trouble in the methods we are going to use next .
another way would be to compare @xmath203 not only with classical solutions in @xmath44 but with classical solutions that are boundaries of open subsets in @xmath0 and which do not intersect the boundary values .
this viewpoint has the disadvantage of not being intrinsically in @xmath189 .
the methods presented in the following also work if one adopts this definition for weak solutions .
for any open set @xmath210 there exists a weak solution of mean curvature flow @xmath188 with @xmath211 .
furthermore there is a smallest such weak solution @xmath188 : for any weak solution @xmath212 with @xmath213 we have @xmath214 for all @xmath215 .
let @xmath216 be the set of all families @xmath217}$ ] of open sets such that @xmath218 and @xmath192 fulfils mean curvature flow .
define @xmath219}\in \mathcal{b}_0 \text { with } t\in [ 0,b]\}.\ ] ] then inductively define @xmath220 and @xmath221 by setting @xmath220 to be the set of all families @xmath190}$ ] of open sets such that @xmath222 and @xmath192 fulfils mean curvature flow . then set @xmath223}\in \mathcal{b}_k \text { with } t\in [ a
, b]\}.\ ] ] finally define the open sets @xmath224 . note that @xmath225 is a nondecreasing sequence of open sets . as a union of subsolutions
@xmath200 is again a subsolution . to see that @xmath200 is a supersolution
let @xmath190}$ ] be a classical solution and let @xmath226 .
then by compactness @xmath227 for some @xmath228 .
therefore @xmath229 and hence @xmath200 is a supersolution .
the second assertion is obvious from the construction .
the following lemma concerning weak solutions will be useful .
the result is trivial for classical solutions .
[ lem weak cylinder ] suppose @xmath188 is a family of open subsets of @xmath189 , such that @xmath230 is a weak solution of mean curvature flow in @xmath231 .
then @xmath199 is a weak solution of mean curvature flow in @xmath189 .
we only show , that @xmath200 is a supersolution using the fact that @xmath232 is a supersolution .
let @xmath190}$ ] be a family of open sets such that @xmath192 fulfils mean curvature flow , and @xmath226 .
we need to show @xmath233 .
in fact we prove @xmath234 .
for this we approximate @xmath235 by bounded sets . for @xmath61 take @xmath236 to be a smoothed intersection of @xmath235 with @xmath237 containing @xmath238 $ ] ( use theorem [ thm smoothing ] ) .
we may take the same closing ends for different @xmath239 , to give curvature bounds on @xmath240 independent of @xmath179 .
then define @xmath241 .
then we still have uniform curvature bounds for @xmath242 sufficiently large .
thus , by proposition 4.1 of @xcite there are classical solutions @xmath243}$ ] of mean curvature flow with @xmath244 for some @xmath245 independent of @xmath179 .
these solutions @xmath246 are written as graphs over the @xmath240 , which contain @xmath247 $ ] .
by interior estimates of @xcite and the uniqueness of the limit ( see below ) we find for @xmath91 local convergence as graphs over @xmath248 .
this gives a solution of mean curvature flow which is written as a graph over @xmath248 and starts from there .
hence it coincides with @xmath249}$ ] .
( to see the uniqueness for smooth cylinders write a non - cylindrical solution as a graph over the initial cylinder . by the strong maximum principle
the difference to the ordinary cylindrical solution attains no interior maximum . then translate the non - cylindrical solution along the cylinder and again use interior estimates to find convergence to a new solution , and do the translation in such a way that the difference of the limit to the cylindrical solution attains an interior maximum
this contradicts the strong maximum principle . )
thus , we have shown that the corresponding flows @xmath250}$ ] of the open sets starting from @xmath251 satisfy @xmath252.\ ] ] let @xmath253}$ ] be the smallest weak solution in @xmath2 with @xmath254 .
the argument above has shown @xmath255 for @xmath256 $ ] .
the same argument shows , that the maximal time - interval on which @xmath257 and @xmath258 coincide is open .
it is easy to see that this maximal interval is also closed : suppose @xmath255 for @xmath259 . since @xmath257 is the smallest solution @xmath260 . to show the reverse inclusion
let @xmath261 . since @xmath262 is open
we find a ball - solution of mean curvature flow centred at @xmath208 that is contained in @xmath263 and contains @xmath264 .
this shows @xmath265 since @xmath255 for @xmath259 and @xmath266 is a supersolution and hence contains the ball - solution .
summarizing we find @xmath255 for all @xmath198 $ ] , i.e. the smallest weak solution starting from @xmath235 is @xmath263 . as @xmath267 is a supersolution with @xmath268
we conclude @xmath234 .
a shadowflow is a weak solution : let @xmath269 $ ] be continuous and smooth in the set @xmath270 and suppose @xmath17 satisfies in this set . define @xmath271 to be the projection of @xmath272 onto @xmath189 , that is @xmath273 then @xmath188 is a weak solution of mean curvature flow with boundary values @xmath274 . by lemma [ lem weak cylinder ] it suffices to show that @xmath267 is a weak solution .
first we prove that @xmath275 is a subsolution , which is slightly easier to see .
so let @xmath190}$ ] be a family of open bounded subsets of @xmath2 such that the boundaries @xmath193}$ ] form a smooth solution of mean curvature flow .
suppose @xmath276 .
assume for contradiction that there is @xmath277 such that @xmath278 . by openness of @xmath279
we may assume @xmath280 .
taking a vertical translation of @xmath281 we may further assume @xmath282 .
this contradicts the avoidance principle . now to see that @xmath230 is a supersolution , suppose @xmath283 and assume for contradiction that there is @xmath277 such that @xmath284 .
we assume @xmath285 and taking a vertical translation of @xmath281 we may assume @xmath286 for all @xmath287 , i.e. @xmath288 is above @xmath289 .
but then by the avoidance principle @xmath290 would be above @xmath291 which leads to a contradiction .
we observe the following let @xmath56 be open .
( the case of closed @xmath56 is handled similar . )
let @xmath294 be the signed distance function to @xmath295 ( which we assume to be nonempty ) such that @xmath296 . by the convexity of @xmath56 , @xmath297 is concave . by the symmetry of @xmath56 , @xmath297 is symmetric . and
because @xmath56 contains the positive cone @xmath297 is increasing in each component of its argument .
then @xmath298 is a concave function on the set of symmetric matrices , where @xmath299 denotes the eigenvalues of @xmath300 ( see e.g. ( * ? ? ?
* end of 3 ) ) .
thus @xmath301 is a convex subset of the symmetric matrices , which finishes the proof . *
1 . altered distance functions and reference neighbourhoods . * first note that since @xmath303 is bounded it suffices to consider everything in a large ball .
let @xmath304 be such that @xmath305 in @xmath306 and in our large ball we have @xmath307 in @xmath300 and @xmath308 coincides with the signed distance function in a tubular neighbourhood of @xmath52 .
let @xmath309 be such that @xmath310 and @xmath311 for @xmath312 , where we choose @xmath134 later , and set @xmath313 .
we derive @xmath314 in a tubular neighbourhood ( in our large ball ) @xmath315 is an eigenvector of @xmath316 with eigenvalue @xmath317 .
the remaining eigenvalues are @xmath318 times the eigenvalues of @xmath319 which are @xmath320 here @xmath321 are the principal curvatures at the boundary and @xmath322 denotes the closest point projection onto the boundary .
note that inside @xmath323 these eigenvalues of @xmath319 are not greater than the negated principle curvatures of the boundary . because @xmath56 contains the positive cone , we find the negation of the eigenvalues of @xmath316 which correspond to eigenvectors orthogonal to @xmath315 lie in @xmath56 .
now choose @xmath134 from above such that the eigenvalue of @xmath316 which corresponds to the eigenvector @xmath315 is the smallest ( largest in absolute value ) eigenvalue of @xmath316 in a neighbourhood of @xmath52 ( still restricted to a large ball ) .
we refer to this neighbourhood restricted to @xmath323 as the reference neighbourhood of @xmath52 ( in our large ball ) . in the reference neighbourhood the negations of the eigenvalues of @xmath324 are in @xmath56 for any @xmath325-dimensional hyperplane @xmath326 .
define @xmath335 this is a mollified version of @xmath336 with parameter @xmath119 .
in fact @xmath337 for all @xmath338 and @xmath339 .
we will choose @xmath44 from the assertion of the form @xmath340 for an appropriate choice of @xmath119 . *
3 . inclusions*. because of @xmath341 it is obvious from that @xmath342 holds for all @xmath343 . to check the other inclusion
let @xmath344 .
by continuity there is @xmath345 independent of @xmath208 , such that @xmath346 now choose @xmath347 and distinguish two cases : suppose @xmath348 .
then by property ( ii ) of @xmath297 we find @xmath349 if on the other hand @xmath350 then by , , and @xmath351 @xmath352 in summary we find the claimed inclusions , provided that @xmath119 is sufficiently small .
* 4 . smoothness .
* we compute @xmath353 assuming @xmath354 implies @xmath355 and therefore @xmath356}(a - b ) = 0.\ ] ] this occurs only if one of the following is true ( note that @xmath354 already implies @xmath357 , by ) by applying sard s theorem to the mapping @xmath369 one can show , that for almost all @xmath343 , @xmath370 where @xmath371 , so that by the implicit function theorem again @xmath372 is smooth for almost all @xmath339 and @xmath373 * 5
. curvature condition . *
let @xmath95 and assume without loss of generality @xmath136 and further assume that the tangent space of @xmath39 at @xmath136 is orthogonal to @xmath376 .
we identify the tangent space and @xmath377 .
locally @xmath39 is a graph over the tangent space at @xmath125 : let @xmath378 and @xmath130 with @xmath379 differentiating twice and using and we obtain @xmath385 note that the eigenvalues of @xmath386 are the principle curvatures of @xmath39 at @xmath125 .
hence , it remains to prove that the negated eigenvalues of @xmath387 lie in @xmath56 .
we compute @xmath388 the matrix @xmath389 is positive semi - definite and @xmath334 ( property ( iv ) ) . since @xmath56 contains the positive cone it suffices to consider the second term in .
we distinguish three cases , below .
but first we observe the following : we may assume that @xmath48 is so small that @xmath390 is contained in the intersection of the reference neighbourhoods of @xmath52 and @xmath53 .
then @xmath391 is in @xmath52 or @xmath53 or in the reference neighbourhoods of both @xmath52 and @xmath53 .
a. @xmath392 : then @xmath393 the term in question becomes @xmath394 .
moreover , in this case @xmath395 holds .
that is why we are in the reference neighbourhood of @xmath53 , where the negations of the eigenvalues of @xmath394 lie in @xmath56 .
b. @xmath396 : this case is treated similiar .
c. @xmath397 : here , by property ( i ) of @xmath297 , @xmath398 . therefore @xmath399 must be in the reference neighbourhoods of @xmath52 and @xmath53 . as a convex combination of two matrices ,
whose negated eigenvalues lie in @xmath56 , the negated eigenvalues of @xmath400 the matrix in question , are in @xmath56 by lemma [ lem convex ] .
g. bellettini , m. novaga .
_ comparison results between minimal barriers and viscosity solutions for geometric evolutions
. _ annali della scuola normale superiore di pisa - classe di scienze 26.1 ( 1998 ) : 97 - 131 .
|
we consider mean curvature flow of an initial surface that is the graph of a function over some domain of definition in @xmath0
. if the graph is not complete then we impose a constant dirichlet boundary condition at the boundary of the surface .
we establish longtime - existence of the flow and investigate the projection of the flowing surface onto @xmath0 , the shadow of the flow .
this moving shadow can be seen as a weak solution for mean curvature flow of hypersurfaces in @xmath0 with a dirichlet boundary condition .
furthermore , we provide a lemma of independent interest to locally mollify the boundary of an intersection of two smooth open sets in a way that respects curvature conditions .
|
in biological systems , water is present in a wide range of environments . as a highly polar molecule
, water tends to trap low - energy electrons ( lees ) and is expected to play an important role in lee - induced processes @xcite .
electron collisions with water vapour being the simplest to study , extensive experimental @xcite and theoretical @xcite work has been carried out . in the more complicated condensed phase systems , experimental studies of electron scattering from amorphous solid water , porous amorphous solid water and crystalline ice
have been reported @xcite . in water clusters ,
formation of ( h@xmath0o)@xmath1 by attachment of slow electrons has also been investigated @xcite in order to elucidate the mechanism of electron solvation . to our knowledge ,
no theoretical investigation of electron scattering with clusters or condensed water have been reported ; available methods are mainly developed for electron collisions with gas - phase molecules .
it is therefore highly desirable to develop a new theoretical approach for condensed - matter and cluster environments especially since many electron - driven processes occur in these media .
the main idea behind the multiple scattering ( ms ) method is to separate the potential of a complex target into non - overlapping regions with each region taken as a single scatterer . the impinging wave on each scatterer is composed of the incident plane wave and the wave scattered from the other scattering centres .
the advantage of such an approach is that one can determine the cross section for large systems by combining information from its subunits .
this enables the study of larger targets ( macromolecules , molecular clusters , etc . )
than is possible using the methods currently available .
the ms approach to lee scattering was previously proposed by two of the authors ( l. caron and l. sanche ) @xcite and its use in conjunction with r - matrix calculations was recently validated @xcite ( hereafter referred to as i ) .
the proposed approach combines scattering gas - phase data , obtained from accurate _ ab initio _ r - matrix calculations , to derive scattering information for a condensed - matter situation .
the partitioning of the space in the r - matrix method makes it ideal for a multiple scattering approach of the `` muffin - tin '' type .
the method is forcibly less rigorous than the ones available for small gas - phase targets , but it is fast and , for the systems tested so far , yields very good results . in i
, we have used the ms theory in the condensed phase , using the equivalent kkr ( korringa , kohn , and rostoker ) approach , in combination with gas - phase r - matrix data in the form of t - matrices .
we were able to derive the band structure of a molecular crystal ( ice ) .
the present work constitutes a first attempt to determine elastic cross section ( cs ) for lee scattering from a molecular cluster using the ms approach .
the water dimer is an excellent choice for such an attempt , primarily because , making use of our previous experience in electron - h@xmath0o collisions @xcite , we were able to perform accurate r - matrix calculations of electron - dimer collisions and test the ms technique .
in addition , water is present in a large variety of environments and information on lee interactions with it is highly relevant .
for example , the characteristics of the water in some of the solvation layers surrounding the double - helix structure of dna are not that of bulk water @xcite , so may be explained in terms of clustered h@xmath0o .
the main question we address in this paper is whether a multiple scattering approach is a good alternative to the computationally heavy _
ab initio _
methods for large to very large targets . for this purpose , we investigate how to apply the cut - offs found in i in order to perform a ms calculation of lee collision with the water dimer .
in this section we present a multiple scattering method to calculate the total t - matrix , @xmath2 , describing lee collisions from a target consisting of two identical scatterers from their individual t - matrices @xmath3 .
the monomers , labelled with the index @xmath4 , are located at @xmath5 .
the asymptotic wave function far away from the centre of mass of the two molecules can be expressed as follows : @xmath6 f_{l}^{0}~. \label{eq_0}\ ] ] where @xmath7 is the relative coordinate of the scattered electron and the centre of mass of the target .
the @xmath8 are spherical harmonics with @xmath9 , @xmath10 denotes the spherical bessel functions and @xmath11 the spherical hankel functions of the first kind .
@xmath12 with @xmath13 the energy of the incoming electron and @xmath14 is an amplitude factor for the incident plane wave .
the first term on the right side of ( [ eq_0 ] ) corresponds to the total incident wave and the second to the total scattered wave .
we need to re - expand the incident wave in ( [ eq_0 ] ) around the centre of mass of each scatterer @xmath4 .
we use the general re - expansion formula from which , for the spherical harmonics defined by messiah @xcite , is @xmath15 where @xmath16 and@xmath17 ^{\frac{1}{2}}\left
( \begin{array}{ccc } l_{1 } & l_{2 } & l_{3 } \\ 0 & 0 & 0% \end{array}% \right ) \left ( \begin{array}{ccc }
l_{1 } & l_{2 } & l_{3 } \\ m_{1 } & m_{2 } & m_{3}% \end{array}% \right).\]]@xmath18 is the wigner 3-j symbol .
it follows that the incident wave on each scatterer is : @xmath19 where @xmath20or , in matrix form @xmath21with @xmath22 we can now incorporate the multiple scattering between the two scatterers to obtain a global expression for the total impinging wave on each molecule .
this has the form @xmath23one can identify our @xmath24 with the expression @xmath25 in the analysis of . by comparing eq .
( [ eq_g1 ] ) to the total impinging part of their eq .
( 1 ) , _ i.e. _ the first term on the right - hand side , one then gets from their eq .
( 2 ) @xmath26 with @xmath27 , @xmath28 .
@xmath29 expresses how a scatterer influences its neighbour .
note that for @xmath30 @xmath31 .
the second term on the right side of ( [ eq_g ] ) is the total scattered wave from the second scatterer @xmath32 . in matrix form , ( [ eq_g ] ) can be written as @xmath33 now , we have to combine the amplitudes scattered from each scatterer @xmath34around the common origin ( centre of mass ) using @xmath35one gets@xmath36where@xmath37combining ( [ eq_m2 ] ) , ( [ eq_m1 ] ) , ( [ eq_chi ] ) and referring to the general expression ( [ eq_0 ] ) , we have @xmath38 we thus obtain the t - matrix for two identical scatterers from their individual t - matrices . * m1 * and * m2 * account for the transformation from the monomer to the dimer s centre of mass and @xmath39 expresses the multiple scattering between the two scatterers .
no experimental or theoretical data on elastic lee collisions with the water dimer are available . therefore , in order to test the quality of our ms results , we have performed an _ ab initio _ scattering calculation on the gas - phase dimer using the r - matrix method and the uk polyatomic r - matrix suite @xcite .
the gas - phase water dimer in its ground state equilibrium geometry has @xmath40 symmetry ( figure [ fig : geometry ] ) .
the geometry parameters from were used in the r - matrix calculation .
the geometry of the water monomer is very slightly changed upon dimer formation ( see table [ tab : geometry ] ) . in the ms calculation ,
the dimer was formed by putting together two water molecules separated by @xmath41 bohr along the z - axis .
this means that there are very minor differences between the geometries used in the r - matrix and ms calculations .
the most notable is the hydrogen bond r@xmath42 which is at most @xmath43 bohr longer in the dimer than in the monomers we use to build the ms results .
a detailed description of the r - matrix method as applied to polyatomic molecules within the fixed - nuclei ( fn ) approximation can be found in and .
the method is based on splitting coordinate space into two regions separated by a sphere , henceforth called r - sphere , centred on the centre of mass of the molecule .
the radius of the r - sphere is chosen in such way that the electronic density of the target is negligible outside it . as a consequence ,
exchange and correlation effects can be neglected in this outer region and a long - range multipole expansion used to represent the electron - target interaction . inside the r - sphere ,
both effects are significant and are therefore taken into account using rigorous quantum chemistry methods .
the wavefunction for the target + electron system is then expressed in terms of a close - coupling expansion .
the calculation was performed following previous work on electron scattering from isolated h@xmath0o @xcite .
we used the equilibrium geometry from and the same basis set ( including the diffuse functions ) employed for the monomer work by .
since in this work we are concerned with the elastic scattering , only the ground electronic state was considered and included in the close - coupling expansion .
we therefore produced natural orbitals exclusively from this state and then used them in a casci ( complete active space configuration interaction ) calculation in which 8 electrons are frozen ; this generated around 7000 configurations . with this model
, we obtained a good value for the ground state energy : -152.18 hartree , compared to -152.67 hartree from the most accurate calculation .
we also obtained excellent agreement with the experimental dipole moment : 1.065 a.u . in our calculations compared to 1.041 a.u .. we used an r - sphere radius of @xmath44=13 bohr . in order to confirm that all the electronic density was contained inside the sphere ,
tests were performed for radii of @xmath44=14 and @xmath44=15 bohr . only small differences at very low energies were found between the cross sections calculated with the three radii @xcite .
the continuum orbitals describing the scattered electron were expanded in a basis of gtos with @xmath45 centred on the centre of mass of the water molecule . in i , we showed that gas - phase r - matrix data can be efficiently used in conjunction with ms theory in a non - gaseous environment by deriving the band structure of a molecular crystal ( ice ) .
the fundamental lesson learnt from this kkr study is that for the multiple scattering term , we need a trimmed t matrix @xmath46 , without dipole contribution for radii @xmath47 and with an angular momentum cut - off . thus two different monomer t - matrices are needed : @xmath46 for the multiple scattering term @xmath48 and a matrix @xmath49 for which no cut - offs are applied . with this in mind
[ eq : final ] can be rewritten : @xmath50 the first step in our multiple scattering calculation involves generating the monomer t - matrices @xmath46 and @xmath49 required in equation ( [ eq : used ] ) .
the description of lee scattering by an isolated h@xmath0o molecule is the same than in i. since a detailed description of the calculation was given there , we will limit ourselves here to a brief summary .
we have used the r - matrix method within the fixed - nuclei approximation . for the target description we used the dzp basis set @xcite for the oxygen and a tz basis @xcite for the hydrogen .
a casci model was used with only 2 frozen electrons and the calculation included only the ground state .
the scattering calculation was performed with an r - sphere radius of @xmath51 bohr .
a gto basis for the continuum with @xmath52 was generated for this radius using the program gtobas @xcite .
the obtained t - matrices were rotated from the r - matrix coordinates to their corresponding position in the dimer .
the procedure to do so was explained previously ( equation ( 24 ) in i ) and the reader is referred to that work for further details .
as mentioned above , two cut - offs are applied when generating @xmath46 . a first cut - off on the range of the molecular dipole
is achieved by removing the dipole field for @xmath47
. the cut - off radius @xmath53 depends only on the dipole moment of the molecular monomer and should therefore be the same for the ice and the dimer calculations . in i
, we have found that @xmath54 bohr yields the best value for the electron s effective mass in ice .
an analysis of the electron-(isolated ) water cross section obtained for different values of @xmath53 supports this finding ( the reader is referred to fig . 1 in i ) : for too small cut - off radii ( @xmath55 ) , the high energy cross section is unphysical whereas for higher radii ( @xmath56 ) the cross section exhibits a dipole - driven behaviour ( it increases rapidly at low energy ) .
although this behaviour is physical and should be expected , it is the aim of the cut - off to at least partially eliminate it .
this can be understood as an attempt to ensure that @xmath46 involves only the contribution of the target potential in the small region around the scatterer . that leaves @xmath57 bohr as the most satisfactory radius .
the second cut - off , in the angular momentum components @xmath58 of the scattering matrix , is critical and depends both on the electron kinetic energy @xmath13 and the intermolecular distance @xmath59 .
the restriction is related to the angular momentum energy barrier @xmath60 .
the kkr calculation predicts that only angular momentum values @xmath61 should be retained such that @xmath62 with @xmath63 ry .
that is , only electrons that scatter from one monomer with energy larger than @xmath64 can reach the other one and undergo multiple scattering . in the dimer ,
the two h@xmath0o molecules are separated by @xmath65 bohr ( almost the same intermolecular distance than in crystalline ice in i ) .
we expect then to have the same cut - off criteria than in the kkr calculation .
once the t - matrices for the monomer are calculated , the dimer t - matrix in equation [ eq : used ] is built . from it , we build the r - matrix at @xmath66 bohr using the approach given in appendix a and propagate it outwards @xcite using _ the exact dipolar field of the dimer_. this is necessary in order to incorporate the effect of the dipole moment of the dimer in the ms description . to choose @xmath67 we propagate the r - matrix of the dimer ( obtained with the r - matrix calculation ) inwards from the asymptotic region and then calculate the cs from it .
the low - energy part of this cs decreases the more of the dipole field we remove ( the further in we go ) and takes an unsuspected and incorrect upturn for a radial distance of 9 bohr .
we believe that for this distance the multipolar field of the dimer is no longer describable by its total dipole field .
we finally generate the final t - matrix @xmath2 which takes the dimer s dipole contribution into account .
this t - matrix is a function of energy and of the angular momentum cut - off @xmath68 used for @xmath46 in the multiple scattering term of eq .
[ eq : used ] .
the integral elastic cross section is then obtained using the well known formula : @xmath69 in figure [ fig : test_l ] , we present the cs for different @xmath68 values . as mentioned earlier , the partial wave expansion must be limited as the collision energy decreases .
for example , for electron energies between @xmath70 ev , the kkr calculation predicts that we can only include @xmath71 in the ms calculation .
it can be easily seen from figure [ fig : test_l ] , that the cs is ill behaved in this region ( below @xmath72 ev ) when @xmath73
. a closer look at the cs for the other values of @xmath68 shows the presence of a critical energy below which the cs does not exhibit the right behaviour .
these critical energies are at slightly lower energies than the value given by @xmath74 .
if we scrupulously restrict the calculation to integer values of @xmath68 for electron energies between @xmath74 and @xmath75 , the cs will be jagged _
i.e. _ discontinuous at @xmath74 .
smoothing is needed .
a two - point interpolation on the parameter @xmath58 between @xmath76 and @xmath77 , where @xmath58 is the solution of @xmath78 , would get rid of the discontinuities .
but this would admix the cs at @xmath76 with any singular part of the cs at @xmath77 .
a quick look at figure [ fig : test_l ] indicates , for instance , that at 6 ev when @xmath79 the interpolation would spuriously bring in a sizable part of the huge peak of the @xmath80 cs .
the procedure can be regularised by upward shifting @xmath74 using a parameter @xmath81 such that @xmath82 .
a value of @xmath83 would shift the cut - off energies to the new values shown in figure [ fig : test_l ] , beyond the threshold of the singular behaviour of the cs for @xmath77 .
the interpolation procedure can now be done safely with @xmath58 a solution of @xmath84 .
the cs is linearly interpolated between cs values corresponding to the closest lower and larger integer values of @xmath58 .
the elastic electron-(h@xmath0o)@xmath0 cross section versus impact energy is plotted in figure [ fig : xs_final ] . as a test of the accuracy of the ms calculation
, we compare our results to the cs calculated with the r - matrix codes .
the agreement between the two calculation is very good .
the multiple scattering cross section is slightly higher than the r - matrix one but remains within 5% of the the fully _ ab initio _
cs for energies larger than 2.5 ev .
we believe these differences are well within the range of usual experimental errors and ab initio calculation uncertainties . for targets with large dipole moments ,
it is customary to add to the r - matrix cross section a born based correction @xcite to account for the partial waves not included in the _ ab initio _ calculation .
this correction is the same for the r - matrix and ms calculations and for this reason has not been included in the cross sections plotted in figure [ fig : xs_final ] .
based on the results of i , the need for a `` trimmed '' t - matrix ( @xmath46 ) in the ms term of ( [ eq : used ] ) is obvious .
figure [ fig : test_l ] illustrates the drastic impact of a larger angular momentum basis ; including higher @xmath58 at low energies introduces resonances and/or premature divergences in the cs caused by the resonant nature of the ms @xmath85 term in eq . [ eq : used ] . the cut - off in the long - range dipolar interaction is also very important . in i , we have shown that even though the value of the cut - off @xmath53 is not critical for band structure calculations , a too large or too small cut - off radius has a noticeable effect on the value of the effective mass ( which was used for calibration purposes ) .
the two cut - offs are necessary and have to be applied only to the ms term .
the use of a full t - matrix @xmath49 in ( [ eq : used ] ) is , however , essential to get the right cs behaviour at low energies and to properly account for the potential produced by the dipole of a h@xmath0o molecule .
the need for @xmath49 to include the dipole contribution arises from the fact that for @xmath86 , the scattering electron feels the individual dipoles of the isolated water molecules .
the radius @xmath67 has to be chosen in a proper way .
it must allow the separation between the regions where the individual water dipoles dominate the electron - molecule interaction ( @xmath87 ) and the region where the total dipole of the dimer dominates ( @xmath88 ) . based on a geometrical analysis ( figure [ fig : r - spheres ] )
, one can see that a radius of 9 bohr is an adequate choice . in order to check the appropriateness of this value
, we have tried several r - sphere radii ranging from 6 to 13 bohr and confirmed that 9 bohr yields the best cross section .
it is evident from figure [ fig : r - spheres ] , that a choice of @xmath89 bohr in our calculations does not exactly correspond to a non - overlapping potentials picture . for the spheres not to overlap we would need a radius around @xmath90 bohr .
however , using such small radius does not provide a good representation of the target .
this was already clear in i , where we established that small radii introduce a discontinuity at the boundary of the sphere that is too important to lead to good results .
we have developed a technique to derive the total t - matrix of two scatterers from their individual t - matrices as described by eq .
( [ eq : final ] ) .
the technique first involves finding the t - matrix describing the scattering for an isolated molecular target .
these are then combined using eq .
[ eq : used ] .
our main conclusion is that two different t - matrices , with different constraints , are needed to represent the electron - dimer interaction : two cut - offs have to be applied only to the ms term ( @xmath46 but not @xmath49 ) . from the total t - matrix , the r - matrix at @xmath67
is determined and then propagated to an asymptotic distance using the dipole moment of the cluster .
this propagation is fundamental in order to incorporate the electron interaction with the true dimer dipole moment .
one can argue that the need for @xmath49 and @xmath67 is a surface effect not present in the infinite crystal situation in i. in fact , the dimer is mostly surface .
our results show that a multiple scattering calculation can efficiently replace a _ standard ab initio _ scattering calculation for this system . very good agreement is found between the ms and the r - matrix cross sections .
the ms method should allow us to calculate elastic cross sections for small water clusters that are nevertheless too big to be studied _ ab initio_. the same value of @xmath53 should be employed and @xmath67 could be determined using the same geometrical considerations .
how large a cluster could one attempt to treat using this procedure ? the cluster should have a large surface to volume ratio .
the arguments presented in appendix b indicate that the number of water molecules should be much less than 500 . that leaves a lot of room for fair sized clusters . for very large clusters , having a low surface to volume ratio , the general picture would be as follows .
the physical space should be divided into three parts . in the innermost part ,
the bulk of the cluster , ms theory using @xmath46 in a uniform background optical potential @xmath91 , measuring the average polarisation potential energy between the muffins , would be used as in i. note that the resonant ms @xmath92 term of eq . ( [ eq : used ] ) , which has been freed of any spurious peaks through the angular momentum cut - off algorithm , will yet show peaks that correspond to damped electronic modes of the cluster .
these are , for instance , the modes that were calculated for crystalline ice in i. then there would be a narrow surface region which adapts the bulk value of the optical potential to the vacuum level and in which @xmath46 would be used for the ms part and @xmath49 for the exit part .
finally , there would be an outer region in which the r - matrix of the two inner regions can be propagated to infinity using the multipolar potential of the cluster .
we have thus far discussed only elastic scattering .
how can one deal with inelastic collisions ?
although this is unchartered territory , the scenario will likely go as follows .
in such a situation , new energy channels open up . for each of these
, there will be energy - diagonal t - matrices @xmath93 and cross - energy ones @xmath94 .
these will combine into a super - matrix @xmath95 .
the rest would basically be a repeat of what we have done for a single channel .
one would define the ms part @xmath46 with cut - offs on the range of the dipole and on the angular momentum basis .
note that the latter is different for each energy channel .
moreover , as the energy increases , larger and larger angular momenta will be needed . some adjustments on @xmath53 might also be required in order to include inelastic channels involving more extended molecular wavefunctions .
formally , eq . ( [ eq : used ] ) still applies to these super matrices . the resonant ms @xmath48 term would not only exhibit peaks related to the damped electronic modes of the cluster but also structures caused by the inelastic contributions to @xmath46 which will in turn modulate the inelastic peaks in @xmath49 .
interpolation of the total cs would now involve linear interpolation of the type described above in a multi - dimensional energy - channel space .
the computer time consuming aspect comes from repeated cluster calculations of eq .
( [ eq : used ] ) for many values of the incoming electron energy @xmath96 and @xmath97 .
this research is supported by the canadian institutes of health research and the epsrc .
we suppose in the following , that the r - matrix defined at a radius @xmath98 describes the inner region @xmath99 and that the potential outside the sphere is zero . for elastic scattering
we have @xcite , @xmath100 _ { r = a}~. \label{eq_r}\]]with @xmath101 the reduced radial wavefunctions . for @xmath102 , the wavefunction representing the scattering electron is@xmath103 in the region @xmath104
, one can write@xmath105where @xmath106 and @xmath107 .
@xmath108 is a solution of energy @xmath109 and angular momentum @xmath110 of the schrdinger equation in a zero potential . substituting ( [ eq_f ] ) in ( [ eq_r ] ) , we have@xmath111 ~,\]]where @xmath112 and @xmath113 .
one can rearrange this equation and write it in matrix form@xmath114where @xmath115 , @xmath116 and @xmath117 are the amplitude column vectors .
this gives us : @xmath118 we introduce the k - matrix , @xmath119 substituting ( [ eq_k ] ) in ( [ eq_f ] ) and then in ( [ eq_f ] ) and ( [ eq_psi ] ) , we have for @xmath104 @xmath120 y_{l^{\prime } } ( \hat{r})\ ] ] and thus@xmath121 y_{l^{\prime } } ( \hat{r})~ , \label{eq_psik}\]]which is conform to the definition of the @xmath122 matrix .
( [ eq_k ] ) gives a passage for the transformation @xmath123 .
it is worth noting that the r - matrix depends on the radius but not the k - matrix ( nor the transmission matrix @xmath3 , or the scattering matrix @xmath124 ) .
so once we have the r - matrix for any radius @xmath44 we can build the k - matrix using eq . ( [ eq_k ] ) and inversely .
the t - matrix is then obtained from the k - matrix using the well known relation : @xmath125
one can estimate a surface to volume ration by the ratio of the number of surface molecules @xmath126 to the number of bulk molecules @xmath127 .
one can get an estimate of these numbers by considering a spherical cluster of radius @xmath128 .
the surface molecules are located in the shell between @xmath128 and @xmath129 where @xmath59 is , as above , the distance between molecules .
if the mid - shell surface area @xmath130 is occupied by molecules forming a close packed array , the area occupied by one molecule would be @xmath131 and thus @xmath132 . assuming the bulk of volume @xmath133 is occupied by close packed molecules occupying a volume of @xmath134
, one gets @xmath135 . a high surface to volume ratio
can thus be quantised by the quantity @xmath136 . with @xmath137 a.u .
one gets @xmath138 .
one thus deduces that the cluster must contain much less than @xmath139 molecules to have a large surface to volume ratio .
xx garrett b c , dixon d a , camaioni d m , chipman d m , johnson m a , jonah c d , kimmel g a , miller j h , rescigno t n , rossky p j , xantheas s s , colson s d , laufer a h , ray d , barbara p f , bartels d m , becker k h , bowen k h , bradforth s e , carmichael i , coe j v , corrales l r , cowin j p , dupuis m , eisenthal k b , franz j a , gutowski m s , jordan k d , kay b d , laverne j a , lymar s v , madey t e , mccurdy c w , meisel d , mukamel s , nilsson a r , orlando t m , petrik n g , pimblott s m , rustad j r , schenter g k , singer s j , tokmakoff a , wang l - s , wettig c zwier t s 2005 _ chem .
rev . _ * 105 * , 355 .
.[tab : geometry ] h@xmath0o and ( h@xmath0o)@xmath0 geometries used in our calculations .
the values of the dimer parameters ( see figure [ fig : geometry ] for identification ) are from .
length units are in bohr and angles in deg . [ cols="^,^,^,^",options="header " , ]
|
multiple scattering theory is applied to low - energy electron collisions with a complex target formed of two molecular scatterers .
the total t - matrix is expressed in terms of the t - matrix for each isolated molecule .
we apply the approach to elastic electron-(h@xmath0o)@xmath0 collisions . following the method developed in our previous work on crystalline ice @xcite ,
we impose a cut - off on the dipole outside the r - matrix sphere and an energy dependent cut - off on the angular momentum components of the monomer t - matrix .
an r - matrix calculation of electron - dimer collisions is performed in order to evaluate the accuracy of the multiple scattering approach .
the agreement between the two calculations is very good .
|
the ads / cft correspondence @xcite provides a powerful theoretical method to study the strongly coupled systems in various fields of physics .
one of the most widely investigated objects is the holographic superconductor ( superfluid ) @xcite .
the physical picture is that some gravity background will become unstable as one tunes some parameter , such as temperature for black hole and chemical potential for ads soliton , to develop some kind of hair " .
this condensation of the hair " induces the symmetry breaking , which results in the non - vanishing vacuum expectation value of the dual operator in the field theory side . on the other hand , as a measurement of how a given quantum system is entangled or strongly correlated , the entanglement entropy is also considered as a useful tool for keeping track of the degrees of freedom of strongly coupled systems . dividing a quantum system into a subsystem @xmath0 and its complement , the entanglement entropy of @xmath0 is known as the famous von neumann entropy @xmath1 where @xmath2 is the reduced density matrix for @xmath0 by tracing over its complement of the total density matrix .
however , the calculation of entanglement entropy for a given system is found to be very difficult except for the case in @xmath3 dimensions . inspired by the bekenstein - hawking entropy of black hole and motivated by the development of ads / cft correspondence , the authors in refs .
@xcite have presented a proposal to compute the entanglement entropy of conformal field theories ( cfts ) from the minimal area surface in gravity side .
precisely , in an asymptotically anti de - sitter ( ads ) spacetime , consider a slice at constant radial coordinate @xmath4 , which is dual to a field theory living " on this slice with a uv cutoff labeled by the small parameter @xmath5 .
the @xmath6 corresponds to the uv limit of the dual field theory , which is known as the uv - ir relation @xcite .
search for the minimal area surface @xmath7 in the bulk with the same boundary @xmath8 of a region @xmath0 on that slice .
the entanglement entropy of @xmath0 with its complement is defined as the area law " : @xmath9 where @xmath10 is the newton s constant in the bulk . a proof for this proposal from the basic principle of holographic duality is given in ref .
@xcite and further properties of the holographic entanglement entropy are discussed in ref .
@xcite .
this proposal provides an elegant and executable way to calculate entanglement entropy of a strongly coupled system which has a gravity dual .
since then a lot of works have been carried out @xcite for investigating and calculating the entanglement entropy in various gravity theories . in particular , refs .
@xcite presented the calculations of entanglement entropy for ads soliton geometry and ref .
@xcite considered the case with higher derivative corrections .
very recently , the authors of ref .
@xcite have studied the entanglement entropy in the holographic metal / superconductor phase transition .
it was found that the entanglement entropy in superconducting phase is always less than the one in the metal phase .
the entanglement entropy is continuous , but does not smoothly cross the second order phase transition .
the holographic insulator / superconductor phase transition was first studied in ref .
@xcite , where the ads soliton background @xcite was used to mimic the insulator / superconductor phase transition at zero temperature . motivated by ref .
@xcite we are going to explore the behaviors of the entanglement entropy in the insulator / superconductor phase transition .
the model of the holographic insulator / superconductor phase transition is constructed in the einstein - maxwell - scalar theory with a negative cosmological constant in five - dimensional spacetime .
after completely solving the set of coupled equations of motion , we calculate the entanglement entropy for two geometry configurations , one is a half space and the other is a belt . in both cases ,
we find a discontinuity in the slope of the entanglement entropy as a function of chemical potential across the phase transition .
but , unlike the case of the metal / superconductor phase transition @xcite , after condensate , the entropy first rises as we increase the chemical potential @xmath11 , arrives at its maximum at a certain chemical potential and then it decreases monotonously . for the belt geometry
, there exits a so - called confinement / deconfinement " transition controlled by belt width .
our results confirm the prediction in ref .
@xcite that this phase transition must appear in any consistent gravity dual of a confining theory .
so there are totally four phases " parameterized by chemical potential @xmath11 and belt width @xmath12 .
the complete phase diagram " is drawn in figure .. here we would like to address that in fact there does not exist the parameter belt width @xmath12 " in the dual field theory .
when one changes @xmath12 , that a new phase " occurs in the entanglement entropy manifests a length scale of dual field theory under consideration .
for example , in this paper the length scale is nothing , but the correlation length of the confining theory .
this paper is organized as follows . in section
, we briefly review the holographic insulor / superconductor phase transition presented in ref .
@xcite , and give the complete equations of motion to be solved . in section , the fully back - reacted system is solved by shooting method . in section
we explore the behaviors of the entanglement entropy in insulator / superconductor phase transition . the conclusions and some discussions are included in section .
let us start with the einstein - maxwell - complex scalar theory with a negative cosmological constant in five - dimensional spacetime @xmath13 where @xmath14 is the radius of ads spacetime , field strength @xmath15 and @xmath16 .
when the maxwell field and scalar field are vanishing , the above action admits an exact solution named as ads soliton @xmath17 this asymptotically ads solution approaches to @xmath18 near the boundary at @xmath19 .
moreover , the compactification @xmath20 needed to avoid a conical singularity at @xmath21 gives a dual picture of a three - dimensional field theory with a mass gap , which resembles an insulator in the condensed matter physics @xcite .
the geometry in the @xmath22 sector just looks like a cigar with its tip at @xmath21 .
the temperature associated with the soliton background is zero . in order to mimic the superconductor phase ,
we need the soliton solutions with nonvanishing @xmath23 .
furthermore , the back reaction of the matter fields must be taken into account in order to see the difference of the holographic entanglement entropy .
so we choose the ansatz as follows @xmath24 here we have set @xmath25 without lose of generality .
we require that @xmath26 vanishes at the tip of the soliton . and in order to obtain a smooth geometry at the tip @xmath27 , @xmath28 should be made with an identification @xmath29 the independent equations of motion under this ansatz are given as @xmath30 where a prime denotes the derivative with respect to @xmath31 .
the matter fields near the boundary @xmath32 should behave as @xmath33 @xmath34 where @xmath35 , up to a normalization constant , and @xmath36 are the corresponding dual operators of @xmath37 in the field theory side .
the conformal dimensions of the operators are @xmath38 .
@xmath11 and @xmath39 are the corresponding chemical potential and charge density in the dual field theory , respectively .
there are only four independent parameters at the tip , i.e. , @xmath40 .
however , the above equations of motion have two useful scaling symmetries @xcite @xmath41 @xmath42 these two symmetries allow us to pick any values for the position of the tip @xmath27 and @xmath43 . in our numerical calculations , we will simply choose @xmath44 .
furthermore , to recover the pure ads boundary , we also need @xmath45 and @xmath46 . in general , a solution will have a nonvanishing @xmath47 . in this case
we will use the scaling symmetry to shift @xmath47 to zero .
from the discussion in section [ sect : background ] , for given @xmath48 , we can solve the equations of motion for the system by choosing @xmath49 as a shooting parameter .
after solving the coupled equations , we can obtain the condensate @xmath50 from and read off the chemical potential @xmath11 and charge density @xmath39 through .
in this paper we only focus on the case @xmath51 and @xmath52 .
it is well known that in five - dimensional spacetime , when @xmath53 , the scalar field admits two different quantizations related by a legendre transform @xcite .
@xmath37 can either be identified as a source or an expectation value . in this paper
we set @xmath54 and consider @xmath55 acting as the vacuum expectation value of the operator @xmath56 .
it is convenient to make a coordinate transformation from @xmath57coordinate to @xmath58coordinate by defining @xmath59 .
therefore , the infinite boundary is now at @xmath60 and the soliton tip sits at @xmath61 .
figure exhibits a typical solution , where @xmath62 and @xmath63 are two metric functions , and @xmath64 and @xmath65 are the scalar field and static electric potential , respectively .
note that the metric functions @xmath62 and @xmath63 will be used in calculating the holographic entanglement entropy . a typical soliton solution with nonvanishing scalar hair .
here the value of the scalar at tip is @xmath66 and the corresponding identification @xmath67 . ] for different choices of @xmath68 , the identification length @xmath69 in @xmath28 coordinate will be different . in order to compare different solutions with the same boundary behavior
, we can use the symmetry to set the boundary geometry the same .
taking advantage of the scaling symmetry , the relevant quantities scale as follows @xmath70 the combinations @xmath71 , @xmath72 and @xmath73 are all scale invariant . by using these scale invariant quantities
, we can accomplish our purpose easily . as in our units setup , the identification length @xmath69 in the pure soliton is @xmath74 , we scale all @xmath69 for each solution to be @xmath75 from now on
. we should stress here that after the scaling transformation , the tip @xmath27 will be no longer at @xmath76 .
figure reproduces the results in ref .
@xcite by a slightly different manner .
it shows the condensate and the charge density as functions of chemical potential .
it is clear from the figure that as the chemical potential @xmath11 exceeds a critical value @xmath77 for given mass and charge , the condensation of the operators emerges .
this can be identified as a superconductor ( superfluid ) phase . however ,
when less than @xmath77 , the scalar field is vanishing and this can be thought of as the insulator phase , since this system has a mass gap @xmath78 due to the identification in @xmath28 direction .
the condensate of operator @xmath79 ( left plot ) and the charge density @xmath39 ( right plot ) versus chemical potential @xmath11 respectively . here @xmath80 and @xmath69 is scaled to be @xmath74 .
the critical chemical potential in this case is @xmath81 .
it is a typical second order phase transition which can be seen clearly from @xmath82 plot.,title="fig : " ] the condensate of operator @xmath79 ( left plot ) and the charge density @xmath39 ( right plot ) versus chemical potential @xmath11 respectively . here @xmath80 and @xmath69 is scaled to be @xmath74 .
the critical chemical potential in this case is @xmath81 .
it is a typical second order phase transition which can be seen clearly from @xmath82 plot.,title="fig : " ] note that it is a second order phase transition in our choice of parameters .
in addition , let us mention that in the theory described by the action ( [ action ] ) , except for the ads soliton solution ( insulator ) and hairy soliton solution ( superconducting soliton ) mentioned above , there do exist another two solutions : the black hole solution without hair ( metal phase ) and black hole solution with hair ( another superconducting phase ) .
the complete phase diagram has been drawn in @xcite .
the details depends on the coupling constant @xmath83 . in this paper
we just focus on the case with zero temperature insultor / superconductor phase transition .
as shown in @xcite , with our choice of parameters @xmath51 and @xmath52 , the phase transition from the insulator to the superconductor always happens .
in particular , for any values of chemical potential beyond the critical one , the superconducting phase always exists .
after finding the gravitational soliton solution with full back reaction of matter fields , we are now ready to calculate the holographic entanglement entropy in this holographic model . due to the fact that the choice of the subsystem @xmath0 is arbitrary , we can define infinite entanglement entropies accordingly . in this paper
we first consider a simple case where @xmath0 is chosen to be a half of the total boundary space .
we assume that the subsystem is defined by @xmath84 and extended in the @xmath85 and @xmath28 directions , where @xmath86 .
then the entanglement entropy can be deduced from the formula as @xmath87 for the pure ads soliton solution , @xmath88 and we can read @xmath89 from metric that @xmath90 .
so the entropy can be easily calculated as follows .
@xmath91 where @xmath92 is the uv cutoff .
note that in our units setup @xmath93 .
the first term is divergent ( @xmath94 ) and represents the area law " @xcite .
in contrast , the second term does not depend on the cutoff and thus is physical important .
in fact , this finite term is the difference between the entropy in the pure ads soliton and the one in the pure ads space , because in both cases the divergent part of the entanglement entropy is the same .
the result in ( [ eq18 ] ) implies that the entropy in the ads soliton is less than the one in the pure ads space .
note that the uv behavior of @xmath95 will not change after the operator condensation .
thus the divergent part of the entropy known as the area law will not change because this part is only sensitive to uv quantities .
this can be understood from the gravity solution side , the new solution after the operator condensation still asymptotically approaches to ads space near the ads boundary .
we can write the entanglement entropy in general in the half embedding case as @xmath96 where @xmath97 has no divergence and @xmath98 corresponds to the pure ads soliton .
the value of @xmath97 as a function of chemical potential @xmath11 in the superconductor phase is presented in figure .
the entanglement entropy as a function of chemical potential in the half embedding .
the solid red curve denotes the entropy in the superconductor phase , while the dashed blue line is for the pure ads soliton solution .
the entropy rises as the chemical potential @xmath11 is increased after the phase transition , arrives at its maximum at a certain @xmath11 , and then decreases monotonously . ]
one can see from the figure that the entanglement entropy is a constant in the insulator phase .
after condensate , the entropy first rises and arrives at its maximum as the chemical potential @xmath11 increases , then it decreases monotonously .
thus , from @xmath77 to a certain value of @xmath11 , there exits a region where the entropies in the superconductor phase are larger than the one in the insulator phase .
this behavior of entanglement entropy is quite different from the one discussed in ref .
@xcite , where the case for the holographic metal / superconductor phase transition has been studied in detail .
it is shown that the entanglement entropy after condensate is always lower than the one in the metal phase and decreases monotonously as temperature is lowered , and the entropy as a function of temperate is found to have a discontinuous slop at the transition temperature @xmath99 , which indicates a reduction in the number of degrees of freedom after condensate .
similarly , we here also find a discontinuity in the slope of the increasing entropy at the critical chemical potential @xmath77 indicated by the vertical dotted black line in figure , which may be considered as a significant reorganization of the degrees of freedom of the system .
furthermore , this discontinuity can be regarded as the signature of the second order phase transition .
we next consider a more nontrivial geometry which is a straight belt with a finite width @xmath12 along the @xmath100 direction .
the subsystem @xmath0 sites on the slice @xmath92 and expends in @xmath85 and @xmath28 directions . the holographic dual surface @xmath7 is defined as a three - dimensional surface @xmath101 @xmath102 is the regularized length in @xmath85 direction .
the holographic surface @xmath7 starts from @xmath103 at @xmath92 , extends into the bulk until it reaches @xmath104 , then returns back to the ads boundary @xmath92 at @xmath105 .
it is easy to obtain the induced metric on @xmath7 as @xmath106 using the proposal , we need to minimize the following functional @xmath107=\frac{r\gamma}{2g_n}\int_{r_*}^{\frac{1}{\epsilon}}re^{\frac{a(r)}{2}}\sqrt{1+r^4b(r)(dx / dr)^2}dr.\ ] ] we can deduce the equation of motion for the minimal surface from @xmath108 where @xmath109 is a constant .
we are interested in the case that the surface is smooth at @xmath104 , i.e. @xmath110 .
this condition ensures that @xmath111 .
the width @xmath12 of the subsystem @xmath0 and @xmath112 are connected by the relation @xmath113 finally , we obtain the entanglement entropy as @xmath114 where the uv cutoff has been taken into consideration .
however , there is also a disconnected solution describing two separated surfaces that are located at @xmath115 , respectively .
this configuration is just two copies of the half embedding solution that we discussed above .
so the entanglement entropy of the disconnected configuration is trivially twice of the half embedding case .
the entanglement entropy as a function of strip width @xmath12 for @xmath116 .
the dashed magenta and solid red curves come from the connected solutions , while the solid blue one comes from the disconnected one .
the lowest curve is physically favored compared with others because it has minimal entropy . ] we find the behaviors of the entanglement entropy as a function of the belt width @xmath12 are quite similar for various chemical potential @xmath11 or condensation @xmath117 .
the result for @xmath116 is shown in figure as a typical example .
when @xmath118 , there does not exist connected surface .
the physically favored solution is replaced by the trivial disconnected one in this case . on the other hand , when @xmath119 , we have three different branches of entropy for a given @xmath12 , i.e. , the upper connected branch ( dashed magenta curve ) , the lower connected branch ( solid red curve ) and the middle disconnected branch ( solid blue line ) . the physical entropy is determined by the choice of the lowest one . as we can see from figure , for small @xmath12 , the lower connected branch is favored . however , there is a critical value @xmath120 above which the lower connected branch is not a physical one since its entropy is larger than the one for the disconnected solution .
thus , when @xmath12 increases , a phase transition occurs as belt width @xmath121 .
this is just the so - called confinement / deconfinement " phase transition discussed intensively in refs .
the transition can be understood as the fact that no correlation could be contributed among a distance larger than @xmath69 due to the mass gap @xmath78 .
the entanglement entropy as a function of belt width @xmath12 for different chemical potential @xmath11 .
the dotted black curve is for the pure ads soliton solution . with a fixed @xmath12 , the entanglement entropy increases ( left plot ) and
then decreases ( right plot ) as @xmath11 becomes large.,title="fig : " ] the entanglement entropy as a function of belt width @xmath12 for different chemical potential @xmath11 .
the dotted black curve is for the pure ads soliton solution . with a fixed @xmath12 , the entanglement entropy increases ( left plot ) and
then decreases ( right plot ) as @xmath11 becomes large.,title="fig : " ] more numerical results for entanglement entropy are exhibited in figure with different chemical potential . for @xmath122 at fixed chemical potential , the entropy is dominated by the connected surface and exhibits a non - trivial dependence on @xmath12 .
on the contrary , it is governed by the disconnected configuration and is independent of @xmath12 for @xmath123 .
the fact that the phase transition always appears at critical length @xmath120 is observed for all values of @xmath11 matches the prediction in ref .
@xcite that this confinement / deconfinemnet " phase transition must be exhibited in any consistent gravity dual of a confining theory .
we find that @xmath124 and @xmath125 for various chemical potentials up to @xmath126 in our numerical calculation .
if the belt width is large enough ( @xmath127 ) , the disconnected branch is always favored for all values of @xmath11 .
thus , the behavior of @xmath97 as a function of @xmath11 for fixed @xmath12 is the same as the one shown in figure .
it is helpful to slice the data in superconducting phase differently .
we first study how the entropy of the system changes with chemical potential @xmath11 for fixing belt width @xmath12 , which is presented in the left plot of figure .
it can be seen clearly that the entanglement entropy as a function of chemical potential for fixed @xmath12 has similar behavior as in the half embedding case .
in particular , the purple curve in the confined phase is the same as that in figure , in which @xmath128 . as the chemical potential @xmath11 increases , the entropy first rises and reaches its maximum at a certain value of chemical potential denoted as @xmath129 ,
then it decreases monotonously .
the curve will flatten out as we decrease the belt width @xmath12 .
the curve will finally become a line in the limit @xmath130 , which can be understood from the geometric viewpoint .
the minimal surface with small belt length can only probe the bulk sufficiently near the boundary @xmath32 where the metric solution is nearly pure ads .
furthermore , the value of @xmath129 also decreases as we decrease the belt width @xmath12 .
the values of critical belt width and entropy at the confinement / deconfinemnet " transition point for different chemical potential are drawn in the right plot of figure .
the knot is a consequence of the non - monotonic behavior of the entanglement entropy .
the entanglement entropy of the system changes with chemical potential @xmath11 for fixing belt width @xmath12 ( left plot ) .
the red curve is for @xmath131 and the blue one for @xmath132 .
the purple curve is for sufficiently large @xmath133 in the confining phase .
the right plot presents the behavior of entropy as a function of belt width at critical point of the confinement / deconfinement " transition for different chemical potential.,title="fig : " ] the entanglement entropy of the system changes with chemical potential @xmath11 for fixing belt width @xmath12 ( left plot ) .
the red curve is for @xmath131 and the blue one for @xmath132 .
the purple curve is for sufficiently large @xmath133 in the confining phase .
the right plot presents the behavior of entropy as a function of belt width at critical point of the confinement / deconfinement " transition for different chemical potential.,title="fig : " ] in the deconfinement phase with @xmath134 , the connected branch appears , so the quantitative behavior of @xmath135 changes . in order to measure the degrees of freedom at energy scale @xmath136 in this situation
, we would like to calculate the so - called entropic c - function first introduced in ref .
@xcite @xmath137 the entropic c - function as a function of belt width @xmath12 for a fixed @xmath11 is plotted in figure . just as it is expected , the value of @xmath138 decreases monotonically as we increase the belt width @xmath12 .
due to the appearance of mass gap @xmath139 , the c - function jumps to zero at the critical width @xmath120 ( @xmath140 in figure ) and keeps vanishing for larger @xmath12 .
this is just the reason that one calls the phase transition mentioned above the confinement / deconfinement " phase transition @xcite .
we also show the change of entropic c - function as we tune the value of chemical potential for a fixed small @xmath12 in figure .
we can also see from the figure the non - monotonic behavior with the increment of chemical potential .
furthermore , the curve becomes flat as we decrease the belt width @xmath12 .
it can be expected that the curve will finally become a line in the limit @xmath130 from the perspective of gravity side .
the behavior of entropic c - function as a function of chemical potential for a fixed small @xmath12 is quite qualitatively similar to the entropy in figure .
the key difference here is that the entropy increases as we increase the belt width , while the c - function decrease for fixed chemical potential .
but an important feature we observed here is that in both phases , confinement and deconfinement phases " , the entanglement entropy is always non - monotonic in the superconducting phase .
the entropic c - function as a function of belt width @xmath12 for @xmath141 ( left plot ) and of chemical potential @xmath11 at fixed @xmath142 ( right upper curve ) and @xmath143 ( right down curve ) , respectively.,title="fig : " ] the entropic c - function as a function of belt width @xmath12 for @xmath141 ( left plot ) and of chemical potential @xmath11 at fixed @xmath142 ( right upper curve ) and @xmath143 ( right down curve ) , respectively.,title="fig : " ] to summarize , there do totally exist four phases probed by the entanglement entropy for a belt geometry for this holographic system , i.e. , the confinement / deconfinement insulator phase , and confinement / deconfinement superconductor phase .
these phases are characterized by the chemical potential and belt width . in particular
, the belt width controls the confinement / deconfinement " phase transition . strictly speaking ,
the term phase transition here is inappropriate since the system itself , i.e. , the state of the boundary field theory , does not change at all as one changes the belt width @xmath12 .
however , this behavior is reminiscent of many thermodynamic phase transitions in holographic calculations ( see , for example refs .
@xcite ) and so here we adopt the terminology phase transition " to convey this picture as in the literature @xcite .
then one can draw the phase diagram " as in figure for the system .
the left part of the diagram is not new but just the one for the pure ads soliton case which has been studied in ref .
@xcite . in the superconductor phase ,
we find a non - trivial phase boundary between the confinement and deconfinement phases . in this sense
our work is a generalization of refs .
@xcite to the superconducting soliton solution . the phase diagram " of entanglement entropy for a belt geometry in the holographic insulator / superconductor model for @xmath51 , @xmath52 , and @xmath75 .
the phase boundary between the confinement phase and deconfinement phase is denoted by the dotted blue line and solid red curve , while the insulator phase and superconductor phase are separated by the vertical dashed line . ]
the non - monotonic behavior of the entanglement entropy in the superconductor phase looks strange at first glance .
it is quite different from the case in the metal / superconductor transition @xcite , in which the result of the entanglement entropy decreasing as the temperature is lowered agrees with the fact that more degrees of freedom will be condensed in lower temperature . in our case
, one can see from figure and figure that the behavior of both entropy and entropic c - function can be affected by belt width .
however , the belt width @xmath12 does not play the essential role for the non - monotonic behavior . as we observed above
, the non - monotonic behavior occurs in both the confinement and deconfinement superconducting phases . due to the lack of the knowledge of the dual field theory in the bottom - up " approach
, we do not know the details of the microscopic degrees of freedom and thus it is quite difficult to give a definite interpretation of the non - monotonic behavior of the entanglement entropy in the holographic insulator / superconductor phase transition . in order to understand the numerical result ,
however , we here try to give a simple physical picture as follows .
note that the entanglement entropy as a function of temperature near the critical temperature @xmath99 shows the same behavior in s - wave and p - wave holographic metal / superconductor models @xcite .
the metal phase can be thought of as the one filled with free charge carriers , such as electrons . at the critical temperature
the condensate turns on and the free charge carriers are continuously condensed to a certain microstate as temperature is lowered . as a result ,
the entanglement entropy in the superconducting phase is always less than the one in the metal phase and decreases monotonically as temperature is lowered .
in contrast , the insulator phase can be considered as having a little degrees of freedom and almost all charge carriers are bound and can not move freely .
this is due to the appearance of a mass gap @xmath144 which is the crucial distinction between the holographic metal / superconductor and insulator / superconductor transition . at the beginning of the phase transition
the condensate emerges .
thus , a new kind of charge carriers ( some quasi - particles like cooper pairs ) that can move unconstrained make contribution to the degrees of freedom , which leads to the increase of the entanglement entropy . when the chemical potential is enough large , however , all bounded carriers are condensed to form cooper pairs .
the system in this case is full of cooper pairs " and thus is an ordered state seen from momentum space . as a result ,
the number of degrees of freedom now are even less than the one in the insulator phase .
the above argument may account for the non - monotonic behavior of the entanglement entropy in the insulator / superconductor phase transition . to further understand the non - monotonic behavior of the entanglement entropy in the insulator / superconductor phase transition
, it might be helpful to recall the microscopic model presented in @xcite , where the authors wanted to compare the holographic insulator / superconductor transition with high-@xmath99 cuprates , in particular with the rvb ( resonating valence bond ) scenario of high-@xmath99 electron doped cuprates ( e.g. , nd@xmath145ce@xmath146cuo@xmath147 ) .
the phase diagrams for both systems are quite similar .
the high-@xmath99 cuprate at @xmath148 ( in the under - doped region , here @xmath100 stands for the electron doping ) is the antiferromagnetic insulator . by doping holes ,
one frustrates the antiferromagnetic order and eventually destroys it . beyond some critical doping @xmath149
, the superconducting ground state emerges .
the physics of the under - doped cuprates is well - described by the t - j model @xcite . in the slave - boson ( su(2 ) slave - boson )
approach , the electron operators can be decomposed into bosonic and fermionc parts , called holons and spinons , respectively . in the confined insulator phase
, spinons and holons are completely glued together . at the beginning of phase transition ,
the holons get condensed and the degrees of spinons might be released .
this might be the reason that the entanglement entropy increases compared to the case in the insulator phase .
when the electron doping continues to increase ( chemical potential gets large enough in the holographic insulator / superconudctor phase transition ) , the spinons will form cooper pairs and the degrees of freedom associated with spinons get condensed . in that case , the entanglement entropy of the system is even less than the case in the insulator phase .
this model might explain the non - monotonic behavior of the entanglement entropy .
but one caveat should be stressed here that in the gravity setup of the holographic insulator / superconductor model , an extra scalar field which is dual to the fermion bilinear in the field side does not included @xcite .
but one expects that including the extra scalar field to the model will not change the background geometry very much , so that the entanglement entropy keeps the same behavior . of course ,
the above discussions are just some speculations , further more deep understanding for the non - monotonic behavior of the entanglement entropy is called for .
by using entanglement entropy as a probe , in this paper , we investigated the holographic insulator / superconductor phase transition in a five dimensional einstein - maxwell - complex scalar theory with a negative cosmological constant , we calculated the entanglement entropy of dual field theory with two kinds of geometry in the ads boundary : one is a half space , and the other is a belt geometry with width @xmath12 . in the former case , we found that there does exist a discontinuity of the slop of the entropy with respect to the chemical potential at the phase transition point .
this discontinuity could be thought of as a significant reorganization of the degrees of freedom of the system and a signature of the phase transition . beyond the critical chemical potential
, we found that the entanglement entropy is not monotonic in the superconductor phase : at the beginning of the transition , the entropy increases and arrives at its maximum at a certain chemical potential , and then decreases monotonically . in the belt geometry case , the so - called confinement / deconfinement phase transition " happens in both the insulator and superconductor phases . as @xmath122
the connected surface is favored and the entropy depends on belt width @xmath12 non - trivially , while as @xmath123 the disconnected surface is favored and the entropy is @xmath12 independent .
the results enhance the prediction made in ref .
@xcite that this phase transition must happen in any consistent gravity dual of a confining theory .
thus , as a result , there do exist four different phases " characterized by chemical potential @xmath11 and belt width @xmath12 , in the holographic insulator / superconductor phase transition .
the complete phase diagram " is drawn in figure . we observed the non - monotonic behavior of the entanglement entropy with respect to the chemical potential @xmath11 in both the confinement / deconfinement superconductor phases .
the entanglement entropy firstly rises and reaches its maximum at a certain chemical potential and then decreases monotonically ( see figure ( [ halfen ] ) and figure ) .
although the entanglement entropy and entropic c - function depend on the belt width , it does not play any essential role in the non - monotonic behavior of the entanglement entropy .
the non - monotonic behavior of the entanglement entropy could be thought of as a key characteristic of difference between the holographic insulator / superconductor transition and the metal / superconductor transition . in the latter case ,
the entanglement entropy as a function of temperature near the critical transition point always monotonically decreases as temperature is lowered @xcite .
clearly it would be of some interest to check whether the non - monotonic behavior of the entanglement entropy is universal in other holographic insulator / superconductor model , for example , the p - wave case @xcite .
in addition it is worth stressing here that we did not have a clear microscopic interpretation for the non - monotonic behavior of the entanglement entropy , but we did present some arguments which try to give a physical picture for this behavior . needless to say , it is quite necessary to deeply understand this behavior further . in this paper
, we have limited ourselves to the choice of parameters as @xmath54 , @xmath51 , and @xmath52 .
note that the additional complications appear for sufficiently small charge @xmath83 .
it is worthy studying the cases with other ( @xmath150 ) s and the case at finite temperature .
it is also interesting to study entanglement entropy in other holographic superconductor models , such as the holographic p - wave superconductor @xcite .
what s more , it is of some interest to see the effect on the entanglement entropy of the magnetic field in the holographic superconductors @xcite .
we will leave them for future studies .
rgc thanks t. takayanagi for helpful correspondences and the authors thank zhang - yu nie , hai - qing zhang and shu - hao zou for their helpful discussions and suggestions .
this work was supported in part by the national natural science foundation of china ( no.10821504 , no.10975168 and no.11035008 ) , and in part by the ministry of science and technology of china under grant no .
2010cb833004 .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv .
theor . math .
* 2 * , 231 ( 1998 ) [ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] . s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys .
b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] .
e. witten , `` anti - de sitter space and holography , '' adv .
theor . math .
* 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] .
o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , `` large n field theories , string theory and gravity , '' phys . rept .
* 323 * , 183 ( 2000 ) [ arxiv : hep - th/9905111 ] .
s. s. gubser , `` breaking an abelian gauge symmetry near a black hole horizon , '' phys .
d * 78 * , 065034 ( 2008 ) [ arxiv:0801.2977 [ hep - th ] ] .
s. a. hartnoll , c. p. herzog and g. t. horowitz , `` building a holographic superconductor , '' phys .
lett . * 101 * , 031601 ( 2008 ) [ arxiv:0803.3295 [ hep - th ] ] .
s. a. hartnoll , c. p. herzog and g. t. horowitz , `` holographic superconductors , '' jhep * 0812 * , 015 ( 2008 ) [ arxiv:0810.1563 [ hep - th ] ] .
s. s. gubser , `` colorful horizons with charge in anti - de sitter space , '' phys .
* 101 * , 191601 ( 2008 ) [ arxiv:0803.3483 [ hep - th ] ] .
t. nishioka , s. ryu and t. takayanagi , `` holographic superconductor / insulator transition at zero temperature , '' jhep * 1003 * , 131 ( 2010 ) [ arxiv:0911.0962 [ hep - th ] ] .
s. ryu and t. takayanagi , `` holographic derivation of entanglement entropy from ads / cft , '' phys .
lett . * 96 * , 181602 ( 2006 ) [ hep - th/0603001 ] .
s. ryu and t. takayanagi , `` aspects of holographic entanglement entropy , '' jhep * 0608 * , 045 ( 2006 ) [ hep - th/0605073 ] .
l. susskind and e. witten , `` the holographic bound in anti - de sitter space , '' hep - th/9805114 .
d. v. fursaev , `` proof of the holographic formula for entanglement entropy , '' jhep * 0609 * , 018 ( 2006 ) [ hep - th/0606184 ] .
t. hirata and t. takayanagi , `` ads / cft and strong subadditivity of entanglement entropy , '' jhep * 0702 * , 042 ( 2007 ) [ hep - th/0608213 ] .
t. nishioka , s. ryu and t. takayanagi , ` holographic entanglement entropy : an overview,' j. phys . a a * 42 * , 504008 ( 2009 ) [ arxiv:0905.0932 [ hep - th ] ] .
t. albash and c. v. johnson , `` holographic entanglement entropy and renormalization group flow , '' jhep * 1202 * , 095 ( 2012 ) [ arxiv:1110.1074 [ hep - th ] ] .
r. c. myers and a. singh , `` comments on holographic entanglement entropy and rg flows , '' arxiv:1202.2068 [ hep - th ] .
j. de boer , m. kulaxizi and a. parnachev , `` holographic entanglement entropy in lovelock gravities , '' jhep * 1107 * , 109 ( 2011 ) [ arxiv:1101.5781 [ hep - th ] ] . l. -y . hung , r. c. myers and m. smolkin , `` on holographic entanglement entropy and higher curvature gravity , '' jhep * 1104 * , 025 ( 2011 ) [ arxiv:1101.5813 [ hep - th ] ] .
t. nishioka and t. takayanagi , `` ads bubbles , entropy and closed string tachyons , '' jhep * 0701 * , 090 ( 2007 ) [ hep - th/0611035 ]
. i. r. klebanov , d. kutasov and a. murugan , `` entanglement as a probe of confinement , '' nucl .
b * 796 * , 274 ( 2008 ) [ arxiv:0709.2140 [ hep - th ] ] .
a. pakman and a. parnachev , `` topological entanglement entropy and holography , '' jhep * 0807 * , 097 ( 2008 ) [ arxiv:0805.1891 [ hep - th ] ]
. n. ogawa and t. takayanagi , `` higher derivative corrections to holographic entanglement entropy for ads solitons , '' jhep * 1110 * , 147 ( 2011 ) [ arxiv:1107.4363 [ hep - th ] ] .
t. albash and c. v. johnson , `` holographic studies of entanglement entropy in superconductors , '' arxiv:1202.2605 [ hep - th ] .
g. t. horowitz , r. c. myers , `` the ads / cft correspondence and a new positive energy conjecture for general relativity , '' phys .
* d59 * , 026005 ( 1998 ) .
[ hep - th/9808079 ] .
i. r. klebanov and e. witten , `` ads / cft correspondence and symmetry breaking , '' nucl . phys .
b * 556 * , 89 ( 1999 ) [ hep - th/9905104 ] .
a. chamblin , r. emparan , c. v. johnson and r. c. myers , `` charged ads black holes and catastrophic holography , '' phys .
d * 60 * , 064018 ( 1999 ) [ hep - th/9902170 ] .
d. mateos , r. c. myers and r. m. thomson , `` holographic phase transitions with fundamental matter , '' phys .
lett . * 97 * , 091601 ( 2006 ) [ hep - th/0605046 ] .
s. s. gubser and s. s. pufu , `` the gravity dual of a p - wave superconductor , '' jhep * 0811 * , 033 ( 2008 ) [ arxiv:0805.2960 [ hep - th ] ] .
a. akhavan and m. alishahiha , `` p - wave holographic insulator / superconductor phase transition , '' phys .
d * 83 * , 086003 ( 2011 ) [ arxiv:1011.6158 [ hep - th ] ] .
r. g. cai , z. y. nie and h. q. zhang , `` holographic p - wave superconductors from gauss - bonnet gravity , '' phys .
d * 82 * ( 2010 ) 066007 [ arxiv:1007.3321 [ hep - th ] ] .
r. -g .
cai , z. -y .
nie and h. -q .
zhang , `` holographic phase transitions of p - wave superconductors in gauss - bonnet gravity with back - reaction , '' phys . rev .
d * 83 * , 066013 ( 2011 ) [ arxiv:1012.5559 [ hep - th ] ] .
t. albash and c. v. johnson , `` a holographic superconductor in an external magnetic field , '' jhep * 0809 * , 121 ( 2008 ) [ arxiv:0804.3466 [ hep - th ] ] .
x. -h .
ge , b. wang , s. -f .
wu and g. -h .
yang , `` analytical study on holographic superconductors in external magnetic field , '' jhep * 1008 * , 108 ( 2010 ) [ arxiv:1002.4901 [ hep - th ] ] .
|
we investigate the behaviors of entanglement entropy in the holographical insulator / superconductor phase transition .
we calculate the holographic entanglement entropy for two kinds of geometry configurations in a completely back - reacted gravitational background describing the insulator / superconductor phase transition . the non - monotonic behavior of the entanglement entropy is found in this system . in the belt geometry case ,
there exist four phases characterized by the chemical potential and belt width .
|
there currently exist three experimental signatures for neutrino oscillations .
the two signatures seen originally in solar and atmospheric neutrinos have been verified by several experiments , including experiments carried out with accelerator and nuclear reactor sources .
the results on atmospheric neutrinos can be explained by @xmath0 disappearance due to oscillations @xcite , while those on solar neutrinos can be explained by @xmath1 oscillations @xcite .
the third signature is @xmath2 appearance in a @xmath3 beam , observed by the short - baseline , accelerator - based lsnd experiment at los alamos @xcite .
this signature is strong from a statistical point of view , being a 3.8@xmath4 excess , but further confirmation by an independent experiment is necessary .
the miniboone experiment at fermilab will be able to confirm or refute the lsnd signature in the near future , with an experimental setup providing different systematics and higher statistics than lsnd @xcite . + taken at face value , the three experimental signatures point to three independent mass splittings .
three neutrino masses do not appear to be able to explain all of the three signatures @xcite ( see , however , @xcite ) .
one way to solve this puzzle is to introduce different mass spectra for the neutrino and antineutrino sector , thereby requiring cpt - violation but no extra neutrino generations @xcite .
another possibility is to add additional neutrinos with no standard weak couplings , often called `` sterile neutrinos '' .
+ in this paper we assume cpt- and cp - invariance , and we explore the possibility of adding one or two neutrino generations beyond the three active flavors assumed by the standard model .
we focus on extensions of the neutrino sector where the addition of fourth and fifth mass eigenstates are responsible for the high @xmath5 lsnd oscillations , and the three lower mass states explain solar and atmospheric oscillations .
when only one sterile neutrino is added , these models are labelled as ( 3 + 1 ) .
the flavor content of the four neutrino mass eigenstates for these models is schematically shown in fig .
[ fig : fig1]a .
the ( 3 + 1 ) hierarchy in fig .
[ fig : fig1]a is as opposed to the ( 2 + 2 ) hierarchy , where the solar and atmospheric mass splittings are separated from each other by the lsnd @xmath5 .
the ( 2 + 2 ) models require a different global analysis from the one discussed in this paper .
the simplest ( 2 + 2 ) models appear to be only marginally consistent with neutrino oscillations data @xcite , even though more general ( 2 + 2 ) mass and mixing scenarios might represent a viable solution to explain solar , atmospheric , and lsnd oscillations @xcite .
+ flavor content of neutrino mass eigenstates in ( 3 + 1 ) models ( a ) , and ( 3 + 2 ) models ( b ) .
neutrino masses increase from bottom to top .
the @xmath1 fractions are indicated by right - leaning hatches , the @xmath0 fractions by left - leaning hatches , the @xmath6 fractions by crosshatches , and the @xmath7 fractions by no hatches .
the flavor contents shown are schematic only.,title="fig:",width=143 ] flavor content of neutrino mass eigenstates in ( 3 + 1 ) models ( a ) , and ( 3 + 2 ) models ( b ) .
neutrino masses increase from bottom to top .
the @xmath1 fractions are indicated by right - leaning hatches , the @xmath0 fractions by left - leaning hatches , the @xmath6 fractions by crosshatches , and the @xmath7 fractions by no hatches .
the flavor contents shown are schematic only.,title="fig:",width=143 ] the ( 3 + 1 ) models are motivated by the criterion of simplicity in physics , introducing the most minimal extension to the standard model that explains the experimental evidence .
however , theories invoking sterile neutrinos to explain the origin of neutrino masses do not necessarily require only one sterile neutrino .
indeed , many popular realizations of the see - saw mechanism introduce three right - handed neutrino fields @xcite .
in particular , ( 3 + 2 ) neutrino mass and mixing models can be obtained in several see - saw mechanisms @xcite . from the phenomenological point of view , it is our opinion that two- and three - sterile neutrino models should also be considered and confronted with existing experimental results . in this paper
, we consider the results from the short - baseline experiments bugey @xcite , ccfr84 @xcite , cdhs @xcite , chooz @xcite , karmen @xcite , lsnd @xcite , and nomad @xcite , and examine how well ( 3 + 1 ) and ( 3 + 2 ) models agree with data . a schematic diagram for ( 3 + 2 ) models is shown in fig .
[ fig : fig1]b
. we do not consider ( 3 + 3 ) models in this paper . from our initial studies
, we believe that the phenomenology of a ( 3 + 3 ) model is similar to a ( 3 + 2 ) model .
+ the paper is organized as follows . in section
[ sec : two ] , we specify the neutrino oscillations formalism used in this analysis to describe ( 3 + 1 ) and ( 3 + 2 ) , short - baseline , oscillations . in section [ sec : three ] and [ sec : four ] , we present the results obtained for the ( 3 + 1 ) and ( 3 + 2 ) models , respectively . for both models ,
we first derive the level of compatibility between the null short - baseline ( nsbl ) experiments and lsnd .
second , we perform a combined analysis of all seven short - baseline experiments ( including lsnd ) to derive the preferred regions in neutrino mass and mixing parameter space . in section
v , we discuss four statistical tests to compare the ( 3 + 1 ) and ( 3 + 2 ) hypotheses . in section
vi , we briefly mention other experimental constraints on ( 3 + 1 ) and ( 3 + 2 ) models . in appendix
[ sec : eight ] , we describe the physics and statistical assumptions used in the analysis to describe the short - baseline experiments .
under the assumptions of cp- and cpt - invariance , the probability for a neutrino , produced with flavor @xmath8 and energy @xmath9 , to be detected as a neutrino of flavor @xmath10 after travelling a distance @xmath11 , is @xcite : @xmath12 where @xmath13 ( @xmath14 being the sterile flavor ) ; @xmath15 is the unitary neutrino mixing matrix ; @xmath16 ; @xmath17 ; and @xmath18 is the number of neutrino generations . neglecting cp - violating phases , there are in general @xmath19 independent mass splittings , and @xmath20 independent mixing matrix elements .
the situation simplifies considerably by considering short - baseline ( sbl ) data only . in this case
, it is a good approximation to assume @xmath21 , and only @xmath22 independent mass splittings are present .
moreover , given the set of sbl experiments considered , the number of mixing matrix elements probed is only @xmath23 , as we show now for the ( 3 + 1 ) and ( 3 + 2 ) cases .
+ for ( 3 + 1 ) models , @xmath18=4 , and only one mass splitting @xmath24 appears in the oscillation formula : this is sometimes referred to as to the `` quasi - two - neutrino approximation '' , or `` one mass scale dominance '' @xcite . using the unitarity properties of the mixing matrix
, we can rewrite eq .
[ eq : osc_gen ] for ( 3 + 1 ) models in a more convenient way : @xmath25 which depends on the mass splitting ( @xmath26 ) and mixing parameters ( @xmath27 ) of the fourth generation only .
since the two - neutrino approximation is satisfied in the ( 3 + 1 ) case , we can express eq .
[ eq : osc_3 + 1 ] in the usual forms : @xmath28 @xmath29 where eq .
[ eq : osc_3 + 1app ] applies to an oscillation appearance measurement , eq .
[ eq : osc_3 + 1dis ] to a disappearance measurement .
+ in this paper , we use the data from the bugey , ccfr84 , cdhs , chooz , karmen , lsnd , and nomad experiments .
bugey and chooz data constrain @xmath2 disappearance ; ccfr84 and cdhs data constrain @xmath0 disappearance ; and karmen , lsnd , and nomad data constrain @xmath30 oscillations .
therefore , from eqs .
[ eq : osc_3 + 1],[eq : osc_3 + 1app ] , and [ eq : osc_3 + 1dis ] , the experiments constrain the following combinations of ( 3 + 1 ) mixing parameters : * bugey , chooz : @xmath31 ; * ccfr84 , cdhs : @xmath32 ; * karmen , lsnd , nomad : @xmath33 . in ( 3 + 1 ) models ,
the tension between the experimental results comes about because bugey , chooz , ccfr84 , cdhs , karmen , and nomad limit the two independent mixing matrix parameters @xmath34 and @xmath35 to be small , whereas lsnd demands nonzero values .
+ in ( 3 + 2 ) models , we introduce two sterile neutrinos .
using eq .
[ eq : osc_gen ] and the unitarity of the mixing matrix , the ( 3 + 2 ) neutrino oscillation probability formula can be written : @xmath36 = \nonumber \\ % = \delta_{\alpha\beta}+4[u_{\alpha 4}^2(u_{\beta 4}^2-\delta_{\alpha\beta } ) \sin^2 x_{41}+u_{\alpha 5}^2(u_{\beta 5}^2-\delta_{\alpha\beta})\sin^2 x_{51}+ u_{\alpha 4}u_{\beta 4}u_{\alpha 5}u_{\beta 5}(\sin^2 x_{41}+\sin^2 x_{51}- \sin^2 x_{54 } ) ] % \label{eq : osc_3 + 2}\end{aligned}\ ] ] which in our case depends on two independent mass splittings ( @xmath37 ) and four independent mixing matrix parameters ( @xmath38 with @xmath39 ) .
[ eq : osc_3 + 1 ] can be recovered from eq .
[ eq : osc_3 + 2 ] by requiring @xmath40 . in ( 3 + 2 ) models ,
the quasi - two - neutrino - approximation is not valid , since there are three distinct @xmath5 values contributing in the oscillation formula : @xmath26 , @xmath41 , and @xmath42 , and therefore three distinct oscillation amplitudes : @xmath43 , @xmath44 , and @xmath45 .
+ we now comment on the monte carlo method used to apply the above oscillation formalism to the analyses presented in this paper .
we require the neutrino mass splittings to be in the range @xmath46 , with @xmath47 .
each mass splitting range is analyzed over a 200 point grid , uniformly in @xmath48 . in section [ sec : six ] , we briefly discuss why large mass splittings are not necessarily in contradiction with cosmological ( and other ) data .
the values of the mixing parameters , @xmath34 , @xmath35 , @xmath49 , and @xmath50 , are randomly generated over a four - dimensional space satisfying the four requirements : @xmath51 , @xmath52 , where : @xmath53 , @xmath54 .
these four inequalities are introduced to account for the fact that large electron and muon flavor fractions in the fourth and fifth mass eigenstates are not allowed by solar and atmospheric neutrino data . in principle , since the cdhs constraint on @xmath0 disappearance vanishes for @xmath55 , as shown in appendix [ sec : eight ] , the upper limit on @xmath0 disappearance from atmospheric neutrino experiments above the atmospheric @xmath5 should be considered instead . in this paper
, we do not reconstruct the likelihood for atmospheric data that would give the exclusion region for @xmath0 disappearance in the range @xmath56 .
however , the effect that the atmospheric constraints would have on our results is expected to be small . for example , in ref .
@xcite , bilenky _ et al .
_ use the atmospheric up - down asymmetry to derive the upper limit @xmath57 at 90% cl for ( 3 + 1 ) models , which is satisfied by our initial requirements @xmath58 .
a more recent analysis @xcite of atmospheric neutrino data using the full zenith angle distribution provides a tighter constraint on @xmath59 than the one given in ref .
@xcite ; the impact of this additional constraint on our sbl analysis is discussed in sections [ sec : three ] and [ sec : six ] . finally , from eqs .
[ eq : osc_3 + 1 ] , it is clear that the relative sign of both @xmath34 and @xmath35 can not be inferred in ( 3 + 1 ) oscillations .
similarly , from eq .
[ eq : osc_3 + 2 ] , the only physically observable relative sign between mixing parameters in cp - conserving ( 3 + 2 ) models is @xmath60 ; therefore , this is the only sign related to mixing parameters that we randomly generate in the analysis .
+ throughout the paper , we make use of the gaussian approximation in determining allowed regions in parameter space . in general
, this means that the regions of quoted confidence level are the ones enclosed by contours of constant @xmath61 values , whose differences with respect to the best - fit @xmath61 value depend on the number of free parameters in the model @xcite . in the text
, we use the symbol @xmath62 to denote the values of the confidence levels derived in this way .
as pointed out in @xcite , this approach should be considered approximate , as it may provide regions in parameter space of both higher and lower confidence than the one quoted .
regions of higher confidence than the quoted value may result from the presence of highly correlated parameters .
regions of lower confidence may result from the presence of fast oscillatory behavior of the oscillation probability formula , eq .
[ eq : osc_gen ] .
this section , like the next one on ( 3 + 2 ) models , consists of two parts .
first , we quantify the statistical compatibility between the nsbl and lsnd results , following a method described in @xcite , originally proposed to establish the compatibility between the lsnd and karmen results .
second , we perform a combined analysis of the nsbl and lsnd datasets , to obtain the favored regions in neutrino mass and mixing parameter space .
many analyses of the nsbl experiments within ( 3 + 1 ) models have concluded that the allowed lsnd region is largely excluded @xcite .
here , we repeat this study with two purposes .
first , we use this study to give context to our discussion of the basic model and techniques which will be expanded in later sections .
second , we demonstrate that our fit , which forms the basis of our new results for ( 3 + 2 ) models , reproduces the expected ( 3 + 1 ) exclusion region . for a discussion of the physics and statistical assumptions used to describe the short - baseline experiments used in the analysis
, the reader should refer to appendix [ sec : eight ] .
+ in this section , the nsbl and lsnd datasets are analyzed separately , providing two independent allowed regions in @xmath63 space .
the level of statistical compatibility between the two datasets can be determined by studying to what degree the two allowed regions overlap , as will be quantified later in this section .
+ for each randomly generated ( 3 + 1 ) model , we calculate the values for the @xmath61 functions @xmath64 and @xmath65 , where @xmath64 is defined as : @xmath66 for the analysis described in this section , the nsbl and lsnd allowed regions are obtained using two different algorithms , reflecting the fact that the nsbl dataset provides upper limits on oscillations , while the lsnd dataset points to non - null oscillations .
+ the nsbl allowed regions at various confidence levels @xmath67 are obtained via a raster scan algorithm @xcite .
let @xmath64 be the @xmath61 value for the particular model and @xmath68 be the minimum @xmath61 for the @xmath5 value considered .
for example , our quoted 95% cl upper limit on @xmath69 is given by the maximum value for the product @xmath70 chosen among the models which satisfy the inequality @xmath71 .
the value of 5.99 units of @xmath61 is chosen because there are two free parameters @xmath34 , @xmath35 for ( 3 + 1 ) models with fixed @xmath5 .
we note that even for the nsbl dataset , the parameters @xmath34 , @xmath35 can be correlated , since the karmen and nomad results probe a combination of the two parameters .
+ the lsnd allowed regions at various confidence levels @xmath72 are obtained via a global scan algorithm @xcite .
for example , for @xmath73 we require @xmath74 , where @xmath75 is now the global lsnd @xmath61 minimum value , considering all possible @xmath5 values .
the lsnd allowed region is computed for two free parameters as for the nsbl case , but the parameters are now @xmath5 and @xmath76 , as opposed to @xmath35 and @xmath77 .
compared to the nsbl case , the number of free parameters is reduced by one because the lsnd @xmath78 search only probes the product @xmath76 and not the two mixing matrix elements individually , and it is increased by one because the allowed region is now obtained by scanning over all possible @xmath5 values .
+ the regions allowed in @xmath63 parameter space by both the nsbl and lsnd datasets are shown in fig .
[ fig : fig2]a .
the nsbl allowed regions shown are 2-dimensional projections of 3-dimensional allowed regions in @xmath79 space .
the nsbl results alone allow the regions to the left of the solid grey , dotted black , and solid black lines in the fig .
[ fig : fig2]a , at a confidence level @xmath80 respectively . in fig .
[ fig : fig2]a , the @xmath81 cl allowed regions obtained by our analysis for lsnd data are also shown , as dark grey shaded , light grey shaded , and white areas , respectively .
we find no overlap between the two individual 95% cl allowed regions ; on the other hand , there is overlap between the two 99% cl regions . + fig . [
fig : fig2]b shows the same ( 3 + 1 ) allowed regions as fig .
[ fig : fig2]a but in the @xmath82 plane , where @xmath83 is defined as the @xmath84 oscillation probability averaged over the lsnd @xmath85 distribution : @xmath86 where @xmath87 is given by eq .
[ eq : osc_gen ] for @xmath88 , and is a function of all the mass and mixing parameters of the oscillation model under consideration .
this has the obvious disadvantage of being a quantity dependent upon the specifics of a certain experiment , as opposed to a universal variable such as @xmath89 .
however , @xmath83 has the advantage of being unambiguously defined for any number of neutrino generations , and thus is useful in discussing ( 3 + 2 ) models later in this paper . as stated previously ,
the oscillation probability estimator @xmath90 can not be used when more than one @xmath5 value affects the oscillation probability , as is the case for ( 3 + 2 ) models .
a second advantage of using @xmath83 instead of @xmath69 as the oscillation probability estimator , is that the allowed values for @xmath83 inferred from the lsnd result tend to be almost @xmath5-independent ( see grey - shaded areas in fig . [ fig : fig2]b ) , as expected for an almost pure counting experiment such as lsnd .
the oscillation probability reported by the lsnd collaboration @xcite is @xmath91 , and agrees well with our result of fig .
[ fig : fig2]b . + fig .
[ fig : fig2]c shows the values for @xmath92 as a function of @xmath5 .
the number of degrees of freedom is 143 .
as discussed in appendix [ sec : eight ] , the dip in @xmath92 at @xmath93 is due to bugey data preferring @xmath94 values , while the minimum at @xmath95 is due to cdhs ( mostly ) and ccfr84 data , preferring @xmath96 values .
the @xmath61 value for no oscillations , @xmath97 , is the largest @xmath61 value in fig .
[ fig : fig2]c ; this means that the choice of parameters @xmath98 provides the best - fit to nsbl data , for the @xmath5 values satisfying the condition @xmath99 .
note that the @xmath100 , @xmath101 dips in @xmath64 are consistent with @xmath102 , and therefore with @xmath103 , but give better fits than the no - oscillations hypothesis , @xmath98 . in other words ,
the goodness of fit for the @xmath103 region depends on the @xmath5 value considered .
+ similarly , fig .
[ fig : fig2]d shows the values for @xmath104 as a function of @xmath5 , used to obtain the lsnd allowed regions drawn in figs . [ fig : fig2]a , [ fig : fig2]b .
+ we now present a slightly different approach to determine the statistical compatibility between the nsbl and lsnd datasets in ( 3 + 1 ) models , which will prove useful in comparing the ( 3 + 1 ) and ( 3 + 2 ) hypotheses .
+ in fig .
[ fig : fig3 ] , we show the values for the @xmath61 differences @xmath105 , @xmath106 , as well as the corresponding confidence levels @xmath67 , @xmath72 , as a function of the lsnd oscillation probability .
the curves are for the set of ( 3 + 1 ) models with the neutrino mass splitting @xmath5 fixed to the best - fit value obtained in a combined nsbl+lsnd analysis ( see section [ subsec : threeb ] ) , @xmath107 , and mixing matrix elements @xmath35 , @xmath34 treated as free parameters .
the value for @xmath5 is chosen in this way because it represents to a good approximation the value for which one expects the best compatibility between the two datasets , as can also be seen in fig .
[ fig : fig2]b . in fig .
[ fig : fig3]a , we map the @xmath108 allowed space into an the @xmath109 , @xmath110 spaces . for any given value of @xmath83 , the minima for the @xmath64 and @xmath65 functions are found in the two ( @xmath34 , @xmath35 ) and one ( @xmath111 ) free parameters available , respectively .
the process is repeated for several @xmath112 values , and the collection of these minima for all values of @xmath83 give the two curves in fig .
[ fig : fig3]a . the individual confidence levels @xmath67 , @xmath72 , shown in fig .
[ fig : fig3]b , are obtained from @xmath105 , @xmath106 in the usual way , by assuming one and two free parameters for the lsnd and nsbl datasets , respectively .
+ we now address how to extract areas in parameter space of a given combined confidence @xmath62 from two independent experimental constraints , in our case obtained via the nsbl and lsnd datasets , without assuming statistical compatibility _ a priori_. the most straightforward way ( described , for example , in @xcite ) is to assign a confidence level @xmath113 to the overlapping part ( if any ) between the two separate allowed regions in parameter space which are found with the constraint @xmath114 .
the extra factor @xmath115 is due to the fact that the lsnd allowed region in the oscillation probability is two - sided , and overlap with the nsbl result on the same probability is obtained only for downward fluctuations in the lsnd result , and not for upward ones .
+ from fig .
[ fig : fig3]b , we find overlapping allowed ranges in @xmath83 for @xmath116 .
we conclude that , in ( 3 + 1 ) models , the lsnd and nsbl datasets are incompatible at a combined confidence of @xmath117 . in our opinion , this value does not support any conclusive statements against the statistical compatibility between nsbl and lsnd data in ( 3 + 1 ) models , although it represents poor agreement between the two datasets .
the reader should also refer to section [ subsec : fived ] , where a different method to quantify the compatibility between the nsbl and lsnd results is discussed .
future short - baseline constraints on @xmath118 appearance , as well as on @xmath0 and @xmath1 disappearance , should be able to definitively establish whether ( 3 + 1 ) models are a viable solution to explain the lsnd signal .
the second analysis we perform is a combined nsbl+lsnd analysis , with the purpose of obtaining the ( 3 + 1 ) allowed regions in parameter space , in both @xmath63 and @xmath82 space .
a combined analysis of this sort assumes statistically compatible results . in section [ subsec : threea ] , we have shown that the lsnd and nsbl results are marginally compatible , for ( 3 + 1 ) models . in the following ,
we refer to the nsbl+lsnd dataset as the short - baseline ( sbl ) dataset , and we construct the @xmath61 function : @xmath119 where the two contributions @xmath64 and @xmath65 are now simultaneously minimized with respect to the same set of three oscillation parameters @xmath5 , @xmath34 , @xmath35 . figs . [
fig : fig4]a and [ fig : fig4]b show the 90% , 95% , and 99% cl 3-dimensional allowed regions in @xmath120 projected onto the @xmath63 and @xmath82 2-dimensional regions , respectively , from the combined ( 3 + 1 ) analysis of sbl data . in this combined analysis
, we use the same monte carlo method described in section [ subsec : threea ] .
we define the allowed regions in parameter space by performing a global scan . for example , the 95% cl allowed region in the three - dimensional space @xmath120 is obtained by requiring @xmath121 , where @xmath122 is the global minimum @xmath61 value .
[ fig : fig4]c shows the minimum @xmath123 values obtained in the combined fit , as a function of @xmath5 .
of course , the @xmath123 values shown in fig . [ fig : fig4]c for any given @xmath5 value are larger than the sum of the two contributions @xmath64 , @xmath65 , shown in figs .
[ fig : fig2]c , d for the same @xmath5 value , since the latter were separately minimized with respect to the oscillation parameters .
similarly , figs .
[ fig : fig5]a and [ fig : fig5]b show the projections of the 90% , 95% , and 99% cl allowed regions in @xmath120 onto the @xmath124 and @xmath125 space , respectively , from the combined ( 3 + 1 ) analysis of sbl data .
the zenith angle distribution of atmospheric muon neutrinos provides a constraint to @xmath59 that is not included in this sbl analysis ; mixings to the right of the dashed vertical line in fig.[fig : fig5]b are excluded at 90% cl by atmospheric neutrino results @xcite .
the global @xmath61 minimum is @xmath123=144.9 ( 148 d.o.f . ) .
this @xmath61 value indicates an acceptable fit , assuming that the goodness - of - fit statistic follows the standard @xmath61 p.d.f .
@xcite ; for an alternative goodness - of - fit test , the reader should refer to section [ subsec : fived ] .
the individual nsbl and lsnd contributions to the @xmath61 minimum are @xmath64=137.3 and @xmath65=7.6 , respectively .
this best - fit point corresponds to the mass and mixing parameters @xmath107 , @xmath126 , @xmath127 .
having introduced the relevant oscillation probability formula in eq . [ eq : osc_3 + 2 ] , and the statistical estimator @xmath83 to compare the nsbl and lsnd results in section [ subsec : threea ]
, we can now quantitatively address the statistical compatibility between the nsbl and lsnd datasets under the ( 3 + 2 ) hypothesis .
+ ideally , we would like to determine the nsbl upper limit for @xmath83 , for all possible combinations of the mass parameters @xmath26 , @xmath41 .
this entails performing a scan equivalent to the one described in the ( 3 + 1 ) case as a function of @xmath26 , shown in fig .
[ fig : fig2 ] . in practice ,
the cpu - time requirements to pursue this route were prohibitive .
+ an easier problem to tackle is to determine the statistical compatibility between the nsbl and lsnd datasets only for the ( 3 + 2 ) models with mass splittings @xmath26 , @xmath41 fixed to their best - fit values , as obtained by the combined nsbl+lsnd analysis that we present in section [ subsec : fourb ] . in sections [ subsec :
threea ] and [ subsec : threeb ] , we have demonstrated that , at least for ( 3 + 1 ) models , this choice is a good approximation for the best possible statistical compatibility ( see figs .
[ fig : fig2 ] and [ fig : fig4 ] ) . + in fig . [
fig : fig6 ] , we show the behavior of the @xmath61 values @xmath128 and @xmath129 , and of the confidence levels @xmath67 and @xmath72 , as a function of @xmath83 , for the set of ( 3 + 2 ) models satisfying the requirements @xmath130 , @xmath131 . by analogy with fig .
[ fig : fig3 ] , we map the four - dimensional space @xmath132 into the two - dimensional spaces @xmath109 and @xmath109 , and we plot the minimum @xmath61 values obtained for any given value of @xmath83 .
the confidence levels shown in fig . [ fig : fig6]b are obtained from fig . [ fig : fig6]a considering the four free parameters ( @xmath34 , @xmath35 , @xmath49 , @xmath50 ) in the @xmath64 minimization process , and the two free parameters ( @xmath111 , @xmath133 ) for @xmath65 .
+ from fig .
[ fig : fig6]b , we find that , in ( 3 + 2 ) models , the nsbl and lsnd datasets are incompatible at an individual confidence level of @xmath134 , and at a combined confidence level @xmath135 .
[ fig : fig6 ] should be compared to fig .
[ fig : fig3 ] , obtained for ( 3 + 1 ) models . a detailed comparison of the ( 3 + 1 ) and ( 3 + 2 ) hypotheses is presented in section [ sec : five ] .
+ we now turn to a combined analysis of the nsbl and lsnd results in ( 3 + 2 ) models , assuming statistical compatibility between the two datasets .
the purpose of this combined analysis is to obtain the allowed regions in the mass parameter space @xmath136 , regardless of the simultaneous values for the mixing parameters .
results will be shown for @xmath47 ; the case @xmath137 can be obtained by simply interchanging @xmath26 with @xmath41 .
the 95% cl allowed region is defined as the @xmath136 for which @xmath138 , where @xmath139 is the absolute @xmath61 minimum for all @xmath136 values . in the minimization procedure ,
the mixing matrix elements @xmath34 , @xmath35 , @xmath49 , @xmath50 , are treated as free parameters . allowed ranges in @xmath136 space for ( 3 + 2 ) models , for the combined nsbl+lsnd analysis , assuming statistical compatibility between the nsbl and lsnd datasets .
the star indicates the best - fit point , the dark and light grey - shaded regions indicate the 90 and 99% cl allowed regions , respectively .
only the @xmath140 region is shown ; the complementary region @xmath141 can be obtained by interchanging @xmath26 with @xmath41.,width=302 ] fig .
[ fig : fig7 ] shows the 90% and 99% cl allowed regions in @xmath136 space obtained in the combined ( 3 + 2 ) analysis . in light of the ( 3 + 1 ) analysis shown in previous sections ,
the result is not surprising , pointing to favored masses in the range @xmath142 , @xmath143 , at 90% cl . at 99% cl ,
the allowed region extends considerably , and many other @xmath136 combinations appear . the best - fit model ( @xmath123=135.9 , 145 d.o.f . )
is described by the following set of parameters : @xmath130 , @xmath144 , @xmath145 , @xmath131 , @xmath146 , and @xmath147 .
we note here that the best - fit is not obtained for fourth and fifth mass eigenstates with degenerate masses , that is for @xmath148 .
the best - fit model we found for sub - ev neutrino masses is : @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath154 , corresponding to @xmath123=141.5 ( 145 d.o.f . ) .
in this section , we discuss four statistical tests that can be used to quantify the better overall agreement of sbl data to a ( 3 + 2 ) hypothesis for neutrino oscillations , compared to a ( 3 + 1 ) one .
test 1 uses only nsbl data to establish the ( 3 + 1 ) and ( 3 + 2 ) upper bounds on the lsnd oscillation probability @xmath83 . from figs .
[ fig : fig3 ] and [ fig : fig6 ] , we obtain at a confidence level @xmath155 : ( 3 + 1 ) : : : @xmath156 ( 3 + 2 ) : : : @xmath157 therefore , we find that ( 3 + 2 ) models can enhance the lsnd probability @xmath83 by quite a large factor , compared to ( 3 + 1 ) models .
the increase in @xmath83 that we obtain is significantly larger than the 25% increase found in @xcite , which is based on a specific choice of mixing parameters , as opposed to the complete parameter scan performed in this work .
the value for the @xmath78 oscillation probability measured by lsnd @xcite is @xmath91 , where the errors refer to the 1@xmath4 statistical and systematic errors , respectively .
test 2 uses both the nsbl and lsnd datasets , and treats them independently to find the combined confidence level @xmath113 for which the datasets are incompatible , both in ( 3 + 1 ) and ( 3 + 2 ) models .
the combined confidence levels can also be read from figs .
[ fig : fig3 ] and [ fig : fig6 ] : ( 3 + 1 ) : : : @xmath158 ( 3 + 2 ) : : : @xmath159 therefore , we find that in ( 3 + 1 ) models the two datasets are marginally compatible , and the agreement is better in ( 3 + 2 ) models .
test 3 combines the nsbl and lsnd datasets into a single , joint analysis .
the likelihood ratio test @xcite provides a standard way to assess whether two hypotheses can be distinguished in a statistically significant way .
we define the maximum likelihood @xmath160 from the minimum @xmath61 values @xmath161 as @xmath162 , where the index @xmath163 refers to the ( 3 + 1 ) and ( 3 + 2 ) hypotheses , respectively .
we can then form the likelihood ratio @xmath164 .
if the ( 3 + 1 ) hypothesis were as adequate as the ( 3 + 2 ) hypothesis in describing sbl data , the quantity @xmath165 should be distributed as a @xmath61 distribution with three degrees of freedom , where the number of degrees of freedom is the difference in the number of mass and mixing parameters in the ( 3 + 2 ) and ( 3 + 1 ) hypotheses , 6 - 3=3 .
+ in our combined fits , we obtain ( see sections [ subsec : threeb ] and [ subsec : fourb ] ) : ( 3 + 1 ) : : : @xmath166 , ( 148 d.o.f . ) ( 3 + 2 ) : : : @xmath167 , ( 145 d.o.f . ) and therefore @xmath168 .
this value is significantly larger than 3 : the probability for a @xmath61 distribution with three degrees of freedom to exceed the value @xmath169 is only 2.9% . in other words , according to the likelihood ratio test , the ( 3 + 1 ) hypothesis should be rejected compared to the ( 3 + 2 ) one at the 97.1% cl .
therefore , based on this test , we conclude from test 3 also that ( 3 + 2 ) models fit sbl data significantly better than ( 3 + 1 ) models . test 4 uses both the results of the individual nsbl and lsnd analyses , as well as the results of the combined nsbl+lsnd analysis .
the test is based on the `` parameter goodness - of - fit '' @xcite to compare the compatibility of the nsbl and lsnd results under the ( 3 + 1 ) and ( 3 + 2 ) hypotheses .
the test avoids the problem that a possible disagreement between the two results is diluted by data points which are insensitive to the mass and mixing parameters that are common to both datasets .
the number of parameters common to both datasets is @xmath170 in ( 3 + 1 ) models , and @xmath171 in ( 3 + 2 ) models .
one possible choice of common parameters is @xmath172 for ( 3 + 1 ) models , @xmath173 for ( 3 + 2 ) models .
the test is based on the statistic @xmath174 , where @xmath175 and @xmath176 are the ( positive ) differences for the nsbl and lsnd @xmath61 values obtained by minimizing the entire sbl @xmath61 function , minus the @xmath61 values that best fit the individual datasets .
.[tab : pg]parameter goodness - of - fit pg , as defined in @xcite , to test the statistical compatibility between the nsbl and lsnd datasets under the ( 3 + 1 ) and ( 3 + 2 ) hypotheses .
the quantities @xmath177 and @xmath178 are the nsbl and lsnd contributions to the test statistic @xmath179 defined in the text ; @xmath180 indicates the number of parameters common to both datasets . [ cols="^,>,>,>,>,<",options="header " , ] table [ tab : pg ] gives the values for the parameter goodness - of - fit pg as defined in @xcite , based on the @xmath179 statistic , and the number of parameters common to the nsbl and lsnd datasets , @xmath180 .
this test shows a dramatic improvement in the compatibility between the nsbl and lsnd results in going from a ( 3 + 1 ) to a ( 3 + 2 ) model , raising the compatibility by nearly two orders of magnitude , from 0.03% to 2.1%
. it will be interesting to investigate if ( 3 + 3 ) models can improve the compatibility further .
the resulting compatibility levels obtained with the parameter goodness - of - fit method are lower than those found in section [ subsec : fiveb ] ; this , however , is not surprising , since the two statistical tests are quite different .
the ( 3 + 1 ) and ( 3 + 2 ) models discussed in this work should be confronted with additional experimental constraints , other than the ones discussed in detail in the previous sections .
we limit ourselves here to list and comment on some of these constraints , rather than address them in a quantitative way .
mostly , we will discuss the impact that such additional constraints may have on the best - fit ( 3 + 1 ) and ( 3 + 2 ) models found in sections [ sec : three ] and [ sec : four ] .
+ first , nonzero mixing matrix elements @xmath34 , @xmath35 , @xmath49 , and @xmath50 may cause observable effects in atmospheric neutrino data , in the form of zenith angle - independent suppressions of the @xmath0 and @xmath1 survival probabilities . since our analysis of sbl data tends to give larger values for muon , rather than electron , flavor content in the fourth and fifth mass eigenstate
, the effect should be larger on muon atmospheric neutrinos .
for example , the ( 3 + 1 ) and ( 3 + 2 ) best - fit models from sections [ subsec : threeb ] and [ subsec : fourb ] would give an overall suppression of the @xmath0 flux of 8% and 17% , respectively .
the size of the effect of @xmath181 oscillations at high @xmath5 is comparable to the current accuracy with which the absolute normalization of the atmospheric neutrino flux is known @xcite , which is approximately 20% . a more quantitative analysis using the full super - kamiokande and macro spectral information @xcite puts an upper bound of 16% at 90% cl on this high @xmath5 contribution to the atmospheric @xmath0 flux suppression ( in the notation of ref .
@xcite , this suppression is parametrized as @xmath182 , where @xmath183 at 90% cl ) .
therefore , it is expected that the inclusion of atmospheric neutrino data in this analysis would pull the best - fit muon flavor components in the fourth and fifth mass eigenstates to lower values , but not in a dramatic way ( see also fig .
[ fig : fig5]b ) .
+ second , models with large masses @xmath184 and @xmath185 , and with nonzero mixing matrix elements @xmath34 and @xmath49 , should be confronted with tritium @xmath10 decay measurements . the presence of neutrino masses @xmath184 and
@xmath185 introduces kinks in the differential @xmath10 spectrum ; the location in energy of the kinks is determined by the neutrino masses , and the size of the kinks is determined by the amount of electron flavor component in the fourth and fifth mass eigenstates . for a spectrometer integrating over the electron energy interval @xmath62 near the @xmath10-decay endpoint , the count rate is @xcite : @xmath186 where the quantity @xmath187 does not depend on the small neutrino masses and mixings , @xmath188 or @xmath189 for ( 3 + 1 ) or ( 3 + 2 ) models , respectively , and we have assumed @xmath190 , and cp invariance . from the experimental point of view
, tritium @xmath10 decay results are generally expressed in terms of a single effective mass @xmath191 : @xmath192 where @xmath191 is the fit mass parameter . in the limit @xmath193
the relation between the true masses and mixings to the fitted mass @xmath191 is independent from the integration interval @xmath62 : @xmath194 the condition @xmath195 is generally satisfied for the neutrino masses considered in this paper , in order to ensure sufficient @xmath10 decay count rate statistics in the experiments
. therefore , to a first approximation , we can consider the effect of heavy neutrino masses @xmath184 , @xmath185 only on the single mass parameter @xmath191 fitted by the experiments . a more general analysis assessing the sensitivity of current and future @xmath10 decay experiments to multiple fitted neutrino masses ,
although highly desirable , is beyond the scope of this work ; for further details , the reader should consult ref .
the current best measurements on @xmath196 come from the troitsk and mainz experiments @xcite , which have very similar @xmath196 sensitivities .
both found no evidence for a nonzero @xmath196 value ; the latest mainz result is @xmath197 , or @xmath198 ev at 95% cl , using @xmath199 ev @xcite .
now , assuming a normal hierarchy ( @xmath200 ) with @xmath201 , the @xmath10 decay neutrino mass in eq .
[ eq : tritium_largedeltalimit ] can be written as @xmath202 ; the best - fit ( 3 + 1 ) and ( 3 + 2 ) models found in this analysis would give @xmath203 and @xmath204 , respectively , that is @xmath196 values well below the current experimental sensitivity .
the planned tritium @xmath10 decay experiment katrin should be able to improve the sensitivity to @xmath191 by roughly an order of magnitude in the forthcoming years , thanks to its better statistics , energy resolution , and background rejection @xcite .
specifically , the systematic and statistical ( for @xmath205 ) uncertainties on the single fitted mass @xmath196 quoted for katrin are 0.018 and 0.016 ev@xmath206 , respectively @xcite , which should provide some sensitivity to the preferred ( 3 + 1 ) and ( 3 + 2 ) neutrino models with a normal mass hierarchy , @xmath200 .
we now consider mass spectra with an inverted hierarchy , defined here as @xmath207 for ( 3 + 1 ) models , and @xmath208 for ( 3 + 2 ) models .
we note that for ( 3 + 2 ) models other hierarchies are also possible , but those do not satisfy the implicit assumption @xmath209 taken in this analysis . the @xmath10 decay neutrino mass in eq .
[ eq : tritium_largedeltalimit ] can now be written as @xmath210 for inverted ( 3 + 1 ) models , and as @xmath211 for inverted ( 3 + 2 ) models .
clearly , in this case the @xmath10 decay constraints depend strongly on the values of @xmath212 , and models with @xmath213 mass splittings are already severely disfavored .
+ third , introducing sterile neutrinos may affect a number of cosmological predictions , which are derived from various measurements @xcite .
the standard cosmological model predicts that sterile neutrinos in the @xmath214 mass range with a significant mixing with active neutrinos would be present in the early universe with the same abundance as the active neutrino species , in disagreement with cosmological observations @xcite . on the other hand
, several models have been proposed that would reconcile sterile neutrinos with cosmological observations , for example suppressing thermal abundances for sterile neutrinos ( see , _ e.g. _ , ref.@xcite and references therein ) . in particular , active - sterile oscillations in the early universe may provide a natural mechanism to suppress the relic abundances of sterile neutrinos @xcite , and scenarios invoking multiple sterile neutrinos are being investigated @xcite .
we have performed a combined analysis of seven short - baseline experiments ( bugey , chooz , ccfr84 , cdhs , karmen , lsnd , nomad ) for both the ( 3 + 1 ) and the ( 3 + 2 ) neutrino oscillation hypotheses , involving one and two sterile neutrinos at high @xmath5 , respectively .
the motivation for considering more than one sterile neutrino arises from the tension in trying to reconcile , in a cpt - conserving , four - neutrino picture , the lsnd signal for oscillations with the null results obtained by the other short - baseline experiments .
multiple ( _ e.g. _ three ) sterile neutrinos can also be motivated on theoretical grounds .
+ we have described two types of analyses for both the ( 3 + 1 ) and ( 3 + 2 ) neutrino oscillation hypotheses . in the first analysis , we treat the lsnd and the null short - baseline ( nsbl ) datasets separately , and we determine the statistical compatibility between the two . in the second analysis , we assume statistical compatibility and we combine the two datasets , to obtain the favored regions in neutrino mass and mixing parameter space . + the main results of the analysis are summarized in section [ sec : five ] , where we compare the adequacy of the ( 3 + 1 ) and ( 3 + 2 ) hypotheses in describing neutrino short - baseline data , by means of four statistical tests .
first , we treat the lsnd oscillation probability as a parameter that can be measured with nsbl data alone , and find that the nsbl 90% cl upper limit on the lsnd oscillation probability can be significantly relaxed by going from ( 3 + 1 ) to ( 3 + 2 ) models , by about 80% .
second , the combined confidence level for which the nsbl and lsnd datasets are incompatible is determined to be 96.4% and 70.0% in the analysis , for the ( 3 + 1 ) and ( 3 + 2 ) hypotheses , respectively .
third , a likelihood ratio test of the two hypotheses is discussed , and shows that the ( 3 + 1 ) hypothesis should be rejected compared to the ( 3 + 2 ) one at the 97.1% cl .
fourth , the parameter `` goodness - of - fit '' defined in @xcite shows much better agreement between the nsbl and lsnd results for ( 3 + 2 ) models than for ( 3 + 1 ) models .
+ in conclusion , we find that ( 3 + 1 ) models are only marginally allowed when considering all of the seven short - baseline results , including lsnd , in agreement with previous analyses @xcite , and that ( 3 + 2 ) models can provide a better description of the data . only the simplest neutrino mass and mixing patterns
have been fully characterized in the literature so far , and the analysis described in this paper may be viewed as a simple attempt to explore more generic scenarios , which appear both experimentally and theoretically plausible . given the bright potential for precision measurements by neutrino oscillation experiments in the near future , a more general phenomenological approach may be needed .
we thank j. bouchez , l. camilleri , k. eitel , j.j .
gomez - cadenas , e. a. hawker , g. b. mills , e. nagy , v. valuev , and g. p. zeller for kindly providing data used in this analysis .
we thank k. n. abazajian , g. barenboim , s. j. brice , k. eitel , b. kayser , w. c. louis , m. maltoni , j. monroe , p. nienaber , t. schwetz , a. y. smirnov , and k. whisnant for valuable discussions and useful suggestions .
this work was supported by nsf .
in this section , we briefly describe the physics and statistical assumptions used to obtain the approximate characterizations of the short - baseline experiments used in the analysis . for the analysis of the bugey , cdhs , and karmen data
, we also refer to the excellent reference @xcite , which we followed closely .
the bugey experiment @xcite is sensitive to @xmath2 disappearance by measuring the charged - current interaction of @xmath2 s produced by two nuclear reactors at the bugey nuclear power plant .
two liquid scintillator detectors , located at different positions , are used .
the signature for an antineutrino interaction is a positron and a delayed light pulse produced by the subsequent neutron capture on @xmath215 .
data are given for three baselines : @xmath216 , @xmath217 , and @xmath218 m between neutrino production and detection .
we follow the `` normalized energy spectra '' analysis discussed in the bugey paper @xcite .
the data are presented as ratios of observed to predicted ( for no oscillations ) positron energy spectra , between @xmath219 and @xmath220 mev positron energy .
we use 25,25 , and 10 positron energy bins for the @xmath216 , @xmath217 , @xmath218 m baselines , respectively . in the @xmath61 analysis , fits included not only the mass and mixing parameters , but also five large scale deformations of the positron spectrum due to systematic effects . the experimental positron energy resolution and the neutrino baseline smearing are taken into account ; the neutrino cross - section energy dependence within a positron energy bin is not ( the energy bin widths are small )
. + similarly , the chooz experiment @xcite investigates @xmath2 disappearance by observing interactions of @xmath2 s produced by two nuclear reactors @xmath221 km away from the chooz detector .
the signature for a neutrino interaction is a delayed coincidence between the prompt @xmath222 signal and the signal due to the neutron capture in the gd - loaded scintillator .
we follow `` analysis a , '' as discussed in the chooz paper @xcite .
data are given as positron yields as a function of energy .
in this analysis , seven positron energy bins , between @xmath223 and @xmath224 mev , are considered , for which the chooz observations , as well as the predictions on the positron yields for the no - oscillation case from both reactors , are given in @xcite . because of the presence of two reactor sources , the @xmath61 analysis comprises 14 positron yield bins for a given energy / baseline .
we use the full covariance matrix to take into account the fact that the yields corresponding to the same energy bin are extracted for both reactors simultaneously , as is done in @xcite .
the analysis fits for the systematic uncertainty in the absolute normalization constant on the @xmath2 yield from the reactors , in addition to the mass and mixing parameters .
since we are interested in the @xmath225 range only , where no energy shape distortions are expected , we neglect the systematic uncertainty on the energy - scale calibration , and the effect of the positron energy resolution .
+ the ccfr84 experiment @xcite constrains @xmath0 and @xmath3 disappearance by measuring the charged - current interaction of muon neutrinos and antineutrinos , produced by a fermilab secondary , sign selected beam yielding @xmath226 gev neutrinos from @xmath227 and @xmath228 decays in the @xmath229 m long decay pipe .
we refer here to the 1984 ccfr experiment ( hence the label ccfr84 throughout the text ) , which operated with two similar detectors located at different distances from the neutrino source , @xmath230 and @xmath231 m from the mid - point of the decay region , respectively .
the two sampling calorimeter detectors consisted of steel plates and scintillation counters .
six secondary beam momentum settings were used : five for neutrino running , and one for antineutrino running . for each secondary beam momentum setting , the data are divided into three neutrino energy bins , for a total of eighteen energy bins , from ref.@xcite .
data are presented as double ratios : the far to near detector ratio of observed number of events , divided by the far to near ratio of events predicted for no oscillations . as in @xcite ,
only the mean neutrino energy for a given neutrino energy bin is used in the @xmath61 analysis .
the systematic and statistical uncertainties on the far to near ratio normalization are taken into account .
the systematic uncertainty is assumed to be energy - independent and totally correlated between any two energy bins .
the neutrino pathlength smearing , mostly due to the long decay region , is also taken into account .
+ the cdhs experiment @xcite is also sensitive to @xmath0 disappearance via the charged - current interaction of @xmath0 s , produced by a 19.2 gev / c proton beam from the cern proton synchrotron .
two detectors are located at @xmath232 and @xmath233 m from the target .
the detectors are sampling calorimeters , with iron and scintillator modules interspersed , to measure the range of a muon produced in a neutrino interaction .
fifteen muon range bins are used .
the data are presented as double ratios : the far to near detector ratio of the observed number of events , divided by the far to near ratio of the number of events predicted for no oscillations .
neutrino energy distributions are obtained for a given muon energy ( or range ) via the nuance @xcite neutrino cross - section generator . as for ccfr84 , the systematic uncertainty on the far to near ratio and the neutrino baseline smearing
are taken into account .
+ the karmen experiment @xcite investigates the @xmath78 appearance channel , from @xmath3 s produced in the @xmath234-@xmath235-decay at rest ( dar ) chain of the isis neutrino source .
karmen measures the charged - current interaction @xmath236 , with a liquid scinitillator detector located at an average distance @xmath237 m downstream of the neutrino source .
the @xmath2 signature is a spatially correlated delayed coincidence between a prompt positron and a delayed @xmath238 event from a @xmath239 neutron capture reaction . in this analysis , only the positron ( `` prompt '' ) energy distribution after all cuts is taken into account , given in @xcite .
the data are binned into nine prompt energy bins , between @xmath240 and @xmath241 mev ( all bins are @xmath242 mev wide , except the highest energy one , ranging from @xmath243 to @xmath241 mev ) . in predicting the prompt energy distribution for a set of mass and mixing oscillation parameters , the given monte carlo positron energy distribution , and the total number of events expected after all cuts for full mixing and @xmath244 , are used @xcite .
energy resolution and baseline smearing effects ( due to finite detector size ) are taken into account .
given the low statistics of the nine karmen prompt energy bins , we construct the @xmath61 function by first defining the likelihood ratio @xcite : @xmath245 where @xmath246 denotes schematically all mass and mixing parameters , @xmath247 , @xmath248 and @xmath249 are the data , expected signal , and expected background vectors with nine elements , and @xmath250 are the probabilities for a poisson process with known background : @xmath251 we define @xmath252 as : @xmath253 the lsnd experiment at los alamos @xcite is also sensitive to @xmath78 appearance , with a neutrino source and detection signature similar to that of karmen , but with better statistics .
the lsnd liquid scintillator detector is located at an average distance of @xmath254 m from the neutrino source . as for karmen , in this analysis we consider only the positron energy distribution arising from a @xmath2 interaction in mineral oil , published as five energy bins between @xmath255 and @xmath256 mev @xcite .
our analysis ignores the information arising from the higher - energy neutrinos from pions decaying in flight , which has a smaller ( but non - negligible ) sensitivity to oscillations compared to the decay at rest ( dar ) sample considered here . in our simulation , we take into account the expected energy distribution from @xmath235 decay at rest , the neutrino baseline distribution for the @xmath257 m long cylindrical detector , the neutrino energy dependence of the cross - section for the detection process @xmath236 ( including nuclear effects , simulated with the nuance @xcite neutrino cross - section generator ) , and the experimental energy resolution .
we use the published numbers for the background expectations , the number of @xmath2 events for 100% @xmath78 transmutation , and for the efficiency of the event selection criteria .
we construct the lsnd @xmath61 function in the same way as we construct the one for karmen , because of the low statistics of the data sample .
+ finally , the nomad experiment is sensitive to @xmath118 oscillations at @xmath258 by looking for charged - current muon neutrino and electron neutrino interactions in the nomad detector @xcite .
the detector consists of a large dipole magnet which houses drift chambers to measure the momenta of the charged particles produced in neutrino interactions ; transition radiation modules for lepton identification ; an electromagnetic calorimeter to measure the energy of electrons and photons ; a hadron calorimeter for particle identification ; and muon chambers for muon identification .
neutrinos are produced by impinging 450 gev protons extracted from the cern sps accelerator onto a thick beryllium target .
the secondary particles produced in the target are focused into a nearly parallel beam by two magnetic lenses , and decay in a 290 m long decay tunnel to produce a @xmath259 gev neutrino beam with about 1% @xmath1 contamination .
neutrino interactions are then observed in the nomad detector at an average distance of 625 m from the neutrino source .
the @xmath118 search is performed by comparing the measured ratio @xmath260 of the number of @xmath1 to @xmath0 charged - current neutrino interactions with the one expected in the absence of oscillations .
the data are binned into 30 bins , covering ten bins in visible energy between 3 and 170 gev , and three radial bins in the neutrino interaction vertex .
a @xmath61 analysis is performed , using the final nomad numbers on the observed and predicted electron - to - muon ratio , including statistical errors as well as the full error matrix describing systematic uncertainties and uncertainty correlations over different bins @xcite . in predicting the effect of @xmath118 oscillations under any mass and mixing hypothesis , the contribution to @xmath260 from oscillations with full mixing and @xmath261 expected in nomad after all cuts
is used @xcite .
energy resolution and baseline smearing effects ( due to the long decay region ) are taken into account @xcite .
+ in fig .
[ fig : fig8 ] , we show our calculations of the 90% cl upper limits on oscillations as a function of @xmath5 for the six nsbl experiments considered here , as well as the 90% cl allowed region for lsnd .
the @xmath262 values as a function of @xmath5 for all of the experiments are also shown .
all the solid curves shown are obtained from the simplified analysis described here , and compare well with the published results @xcite . + the lsnd region obtained in our analysis of dar neutrinos is slightly shifted to the right compared to the final lsnd area , shown in fig .
[ fig : fig8]f as a dashed line , reflecting the difference in the two datasets .
more detailed lsnd dar analyses give results in rough agreement with our allowed region @xcite . +
the @xmath262 values obtained for the bugey and cdhs experiments as a function of @xmath5 give details that might seem surprising , at first .
slightly better fits to the data are obtained under a neutrino oscillations hypothesis , as opposed to the no oscillations one .
therefore , we add a final comment to explain the results of these fits .
+ the bugey fit is driven by the data at the shortest baseline , 15 m , where the statistical errors on the observed positron spectrum from @xmath2 interactions are the smallest .
as explained in ref.@xcite , systematic uncertainties are taken into account by allowing for linear deformations , as a function of positron energy , of the ratio of observed to predicted positron yields .
the values of @xmath263 as a function of @xmath5 are explained by the fact that , for certain @xmath5 values , an oscillatory fit to the 15 m positron spectrum ratio describes the data marginally better than any straight line .
our best - fit oscillation hypothesis to bugey data only is @xmath107 , @xmath264 .
+ for cdhs , the @xmath265 curve in fig . [ fig : fig8]d has a minimum at @xmath266
. this minimum is due to the fact that the far / near @xmath0 rate ratio , corrected for the baseline and detector mass differences between the two detectors ( as well as other minor effects ) , is measured to be slightly greater than one @xcite : @xmath267 .
this marginal deviation from one causes the fit procedure to prefer more @xmath0 disappearance by oscillations in the near than in the far detector .
given the average @xmath0 energy ( 3.2 gev ) and pathlength ( 130 m ) for neutrinos interacting in the cdhs near detector , this condition is satisfied in the @xmath268 range .
our best - fit oscillation hypothesis to cdhs data only is @xmath269 , @xmath270 .
+ 99 y. fukuda _ et al .
_ [ kamiokande collaboration ] , phys .
b * 335 * , 237 ( 1994 ) ; r. becker - szendy _ et al .
_ , nucl .
suppl . * 38 * , 331 ( 1995 ) ; w. w. allison _ et al . _
[ soudan-2 collaboration ] , phys .
b * 449 * , 137 ( 1999 ) .
[ arxiv : hep - ex/9901024 ] .
y. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys . rev .
lett . *
81 * , 1562 ( 1998 ) [ arxiv : hep - ex/9807003 ] .
m. ambrosio _ et al . _
[ macro collaboration ] , phys .
b * 517 * , 59 ( 2001 ) [ arxiv : hep - ex/0106049 ] .
b. t. cleveland _ et al . _ ,
astrophys .
j. * 496 * , 505 ( 1998 ) ; p. anselmann _ et al . _
[ gallex collaboration ] , phys .
b * 285 * , 376 ( 1992 ) .
d. n. abdurashitov _ et al .
_ , phys .
b * 328 * , 234 ( 1994 ) .
q. r. ahmad _ et al .
_ [ sno collaboration ] , phys . rev .
lett . * 87 * , 071301 ( 2001 ) [ arxiv : nucl - ex/0106015 ] ; q. r. ahmad _ et al .
_ [ sno collaboration ] , phys . rev .
lett . * 89 * , 011301 ( 2002 ) [ arxiv : nucl - ex/0204008 ] ; s. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys .
b * 539 * , 179 ( 2002 ) [ arxiv : hep - ex/0205075 ] . c. athanassopoulos _ et al .
_ [ lsnd collaboration ] , phys .
lett . *
77 * , 3082 ( 1996 ) [ arxiv : nucl - ex/9605003 ] ; c. athanassopoulos _ et al . _
[ lsnd collaboration ] , phys .
c * 58 * , 2489 ( 1998 ) [ arxiv : nucl - ex/9706006 ] ; a.
aguilar _ et al .
_ [ lsnd collaboration ] , phys . rev .
d * 64 * , 112007 ( 2001 ) [ arxiv : hep - ex/0104049 ] .
a. o. bazarko [ boone collaboration ] , arxiv : hep - ex/9906003 .
m. maltoni , t. schwetz and j. w. f. valle , phys .
d * 65 * , 093004 ( 2002 ) [ arxiv : hep - ph/0112103 ] . c. giunti , m. c. gonzalez - garcia and c. pea - garay , phys .
d * 62 * , 013005 ( 2000 ) [ arxiv : hep - ph/0001101 ] . g. barenboim and f. scheck , phys .
b * 440 * , 332 ( 1998 ) [ arxiv : hep - ph/9808327 ] ; g. barenboim , a. dighe and s. skadhauge , phys .
d * 65 * , 053001 ( 2002 ) [ arxiv : hep - ph/0106002 ] .
h. murayama and t. yanagida , phys .
b * 520 * , 263 ( 2001 ) [ arxiv : hep - ph/0010178 ] ; g. barenboim , l. borissov , j. lykken and a. y. smirnov , jhep * 0210 * , 001 ( 2002 ) [ arxiv : hep - ph/0108199 ] ; g. barenboim , l. borissov and j. lykken , arxiv : hep - ph/0212116 .
m. maltoni , t. schwetz , m. a. tortola and j. w. f. valle , nucl .
b * 643 * , 321 ( 2002 ) [ arxiv : hep - ph/0207157 ] .
h. paes , l. song and t. j. weiler , phys .
d * 67 * , 073019 ( 2003 ) [ arxiv : hep - ph/0209373 ] .
p. ramond , arxiv : hep - ph/9809459 .
r. n. mohapatra and r. e. marshak , phys .
lett . * 44 * , 1316 ( 1980 ) [ erratum - ibid . *
44 * , 1643 ( 1980 ) ] .
r. r. volkas , prog .
* 48 * , 161 ( 2002 ) [ arxiv : hep - ph/0111326 ] .
g. j. .
stephenson , t. goldman , b. h. j. mckellar and m. garbutt , arxiv : hep - ph/0307245 ; k. s. babu and g. seidl , arxiv : hep - ph/0312285 ; k. l. mcdonald , b. h. j. mckellar and a. mastrano , arxiv : hep - ph/0401241
. y. declais _ et al .
_ , nucl .
b * 434 * , 503 ( 1995 ) .
i. e. stockdale _ et al .
_ , phys .
lett . * 52 * , 1384 ( 1984 ) .
f. dydak _ et al .
_ , phys .
b * 134 * , 281 ( 1984 ) .
m. apollonio _ et al .
_ , arxiv : hep - ex/0301017 . b. armbruster _ et al .
_ [ karmen collaboration ] , phys .
d * 65 * , 112001 ( 2002 ) [ arxiv : hep - ex/0203021 ] .
p. astier _ et al .
_ [ nomad collaboration ] , phys .
b * 570 * , 19 ( 2003 ) [ arxiv : hep - ex/0306037 ] ; d. gibin , nucl .
suppl . * 66 * , 366 ( 1998 ) ; v. valuev [ nomad collaboration ] , _ prepared for international europhysics conference on high - energy physics ( hep 2001 ) , budapest , hungary , 12 - 18 jul 2001 _ b. kayser , arxiv : hep - ph/0211134
. k. hagiwara _ et al .
_ [ particle data group collaboration ] , phys . rev .
d * 66 * , 010001 ( 2002 ) .
e. d. church , k. eitel , g. b. mills and m. steidl , phys .
d * 66 * , 013001 ( 2002 ) [ arxiv : hep - ex/0203023 ] .
k. eitel , new j. phys .
* 2 * , 1 ( 2000 ) [ arxiv : hep - ex/9909036 ] . o. l. g. peres and a. y. smirnov , nucl .
b * 599 * , 3 ( 2001 ) [ arxiv : hep - ph/0011054 ] .
a. strumia , phys .
b * 539 * , 91 ( 2002 ) [ arxiv : hep - ph/0201134 ] .
w. grimus and t. schwetz , eur .
j. c * 20 * , 1 ( 2001 ) [ arxiv : hep - ph/0102252 ] .
f. james and m. roos , comput .
commun .
* 10 * , 343 ( 1975 ) .
g. j. feldman and r. d. cousins , phys .
d * 57 * , 3873 ( 1998 ) [ arxiv : physics/9711021 ] .
s. m. bilenky , c. giunti , w. grimus and t. schwetz , phys .
d * 60 * , 073007 ( 1999 ) [ arxiv : hep - ph/9903454 ] .
m. maltoni , t. schwetz and j. w. f. valle , phys .
b * 518 * , 252 ( 2001 ) [ arxiv : hep - ph/0107150 ] .
w. t. eadie , d. drijard , f. e. james , m. roos , b. sadoulet , `` statistical methods in experimental physics '' ( north holland , amsterdam , 1971 ) .
m. maltoni and t. schwetz , phys .
d * 68 * , 033020 ( 2003 ) [ arxiv : hep - ph/0304176 ] ; m. maltoni , t. schwetz , m. a. tortola and j. w. f. valle , arxiv : hep - ph/0305312 . m. honda , t. kajita , k. kasahara and s. midorikawa , phys . rev .
d * 52 * , 4985 ( 1995 ) [ arxiv : hep - ph/9503439 ] ; y. liu , l. derome and m. buenerd , phys .
d * 67 * , 073022 ( 2003 ) [ arxiv : astro - ph/0211632 ] .
y. farzan and a. y. smirnov , phys .
b * 557 * , 224 ( 2003 ) [ arxiv : hep - ph/0211341 ] ; y. farzan , o. l. g. peres and a. y. smirnov , nucl .
b * 612 * , 59 ( 2001 ) [ arxiv : hep - ph/0105105 ] . v. m. lobashev _ et al .
_ , nucl .
suppl . * 91 * , 280 ( 2001 ) ; j. bonn _ et al .
_ , nucl .
suppl . * 91 * , 273 ( 2001 ) .
a. osipowicz _ et al .
_ [ katrin collaboration ] , arxiv : hep - ex/0109033 ; l. bornschein [ katrin collaboration ] , econf * c030626 * , frap14 ( 2003 ) [ arxiv : hep - ex/0309007 ] .
d. n. spergel _ et al .
_ , arxiv : astro - ph/0302209 ; w. j. percival _ et al .
_ , arxiv : astro - ph/0105252 . j. m. omeara , d. tytler , d. kirkman , n. suzuki , j. x. prochaska , d. lubin and a. m. wolfe , astrophys .
j. * 552 * , 718 ( 2001 ) [ arxiv : astro - ph/0011179 ] ; w. l. freedman _ et al .
_ , astrophys .
j. * 553 * , 47 ( 2001 ) [ arxiv : astro - ph/0012376 ] . k. n. abazajian , astropart .
* 19 * , 303 ( 2003 ) [ arxiv : astro - ph/0205238 ] .
p. di bari , arxiv : astro - ph/0302433 ; p. di bari , phys .
d * 65 * , 043509 ( 2002 ) [ arxiv : hep - ph/0108182 ] .
r. foot , m. j. thomson and r. r. volkas , phys .
d * 53 * , 5349 ( 1996 ) [ arxiv : hep - ph/9509327 ] .
i. e. stockdale , fermilab - thesis-1984 - 08 .
d. casper , nucl .
suppl . * 112 * , 161 ( 2002 ) [ arxiv : hep - ph/0208030 ] .
k. eitel _ et al . _ [ karmen collaboration ] , nucl .
suppl . * 77 * , 212 ( 1999 ) [ arxiv : hep - ex/9809007 ] .
v. valuev , private communication .
|
we investigate adding two sterile neutrinos to resolve the apparent tension existing between short - baseline neutrino oscillation results and cpt - conserving , four - neutrino oscillation models .
for both ( 3 + 1 ) and ( 3 + 2 ) models , the level of statistical compatibility between the combined dataset from the null short - baseline experiments bugey , chooz , ccfr84 , cdhs , karmen , and nomad , on the one hand ; and the lsnd dataset , on the other , is computed .
a combined analysis of all seven short - baseline experiments , including lsnd , is also performed , to obtain the favored regions in neutrino mass and mixing parameter space for both models .
finally , four statistical tests to compare the ( 3 + 1 ) and the ( 3 + 2 ) hypotheses are discussed .
all tests show that ( 3 + 2 ) models fit the existing short - baseline data significantly better than ( 3 + 1 ) models .
|
the masses of super massive black holes in the centers of galaxies are known to correlate with several properties of the host galaxy .
the most well known correlation is with the stellar velocity dispersion of the galaxy ( , e.g. , * ? ? ?
the black hole is thought to play an important role in the evolution of its host ( e.g. agn feedback ) , as the properties of galaxies are tightly linked to .
therefore it is important to be able to measure the accurately and understand whether the scatter in the relations is due to measurement error , or if it is intrinsic .
most of the estimates upon which these relationships are based were derived using dynamical ( edge - on ) axisymmetric ( or spherical ) dynamical models ( see @xcite for a review ) .
it has long been known from photometry that some elliptical galaxies are triaxial ( e.g. * ? ? ?
* ) , i.e. have intrinsic shapes with three distinct axes @xcite , and more recently @xcite and @xcite have shown using stellar kinematics that around a third of the most massive ellipticals are at least mildly triaxial .
it is thus relevant to rederive the in these galaxies with our triaxial instead of axisymmetric orbit super - position models .
@xcite have modelled mock triaxial merger remnants using axisymmetric geometry and found a correlation between viewing angle and the recovered total galaxy mass .
additionally , they found that the luminous mass to light ratio ( ) was underestimated by up to 20% confirming that triaxial modelling is preferred .
in this paper we explore the black hole recovery with the triaxial machinery from @xcite using galaxies that have previously been modeled with axisymmetric codes to verify the triaxial models .
we do this with 32 and ngc 3379 which have their black hole mass determined using , as well as and integral field unit ( ifu ) data .
both galaxies have state - of - the - art kinematics available over a large spatial range and are inside the sphere of influence of the black hole .
we describe our modeling technique and uncertainties in [ method ] and then derive our black hole estimates on galaxies 32 and ngc 3379 in [ sec : bh3 ] .
we briefly address the reliability in [ sec : tbh_reliability ] , and we end with discussion and conclusions in [ sec : tbh_discuss ] .
in this paper we use the triaxial @xcite orbit superposition technique as it is described in @xcite , of which we give a brief summary here .
it is a powerful tool to construct realistic dynamical models .
it allows for an arbitrary triaxial gravitational potential ( with possible contributions from dark components ) in which the equations of motion are integrated numerically for a representative library of orbits .
then the superposition of orbits is determined for which the combined density and velocity moments best fit the observed surface brightness and kinematics using least squares . by marginalising over parameter space , schwarzschild s method
not only provides the viewing direction and the with the dark matter contribution , but also allows the investigation of the intrinsic dynamical structures as well as the distribution function through the orbital mass weights ( cf .
these models have complete freedom : specifically no form of ( an-)isotropy is implied , within the limits of the observed photometry and stellar kinematics .
@xcite showed that the ( 3i ) schwarzschild method can recover the phase - space distribution function in the two - integral axisymmetric case .
additionally , @xcite confirmed that the same is true for triaxial schwarzschild models and extended this to show that we can recover the internal dynamics and three - integral distribution function of triaxial early - type galaxies given their viewing angles .
in @xcite we have done tests on several three - integral abel models , that simulate real early - type galaxies , and found we are also able to recover the viewing angles of triaxial early - type galaxies , as long as their kinematics show clearly defined gradients .
these final tests allowed us to firmly establish robustness of the shape recovery and phase space distribution of triaxial schwarzschild modeling . to construct a ( luminous ) mass model we assume that the three - dimensional mass distribution can be parameterized with multiple coaxial gaussians . the mass distribution used in the models is constructed by fitting two - dimensional gaussians to the broad band photometry of the galaxy .
these gaussians can be deprojected onto a coaxial triaxial shape by choosing three viewing angles ; ( @xmath0 , @xmath1 ) and the apparent misalignment ( @xmath2 ) , which are used to define the direction from which the galaxy is seen .
the shape of each deprojected gaussian depends on the viewing angles , its observed flattening and its isophotal twist .
the shape of each three - dimensional gaussian is characterized by @xmath3 , @xmath4 and @xmath5 , where @xmath6 , @xmath7 and @xmath8 are the long , intermediate and short axis lengths of each individual gauss @xmath9 , and @xmath10 is the length of longest axis of the gauss as observed on the sky . to do a full search of all possible mass models we first sample uniformly over a ( separate ) set of axis ratios , @xmath11 , @xmath12 and @xmath13 , which translates to a set of viewing angles .
these viewing angles are then used to construct all the mass models .
the light model is converted to a mass by assuming a constant ratio .
the black hole mass estimates determined using schwarzschild modeling have not been without debate : @xcite reported the existence of a @xmath14 plateau prohibiting a assessment .
this could be avoided by using some form of regularisation and enough orbits in the models .
the debate was settled by @xcite who showed that the results from standard superposition methods are accurate , if observational errors are taken into account .
currently , mcdermid is leading a joint effort in a comparative study of the recovery using three independent axisymmetric codes . using generalised cross validation
, he finds that regularisation should be used to avoid the known issue of these models over - fitting the data , and thus yield reliable error estimates on .
while these results are obtained with axisymmetric modeling , it is likely that they will hold in the triaxial case too , as the triaxial method is conceptually very similar . in this paper
we repeated all modeling with and without regularisation and found no difference in the recovered black hole masses . to determine the uncertainty on the derived shape we shall use the @xmath14 based confidence interval that was established in @xcite .
the intervals are based upon the expected standard deviation of the @xmath15(=@xmath16 , where @xmath17 is the number of kinematical observations used to constrain the model ) .
as the determination is only influenced by a few of the innermost kinematical observations , we shall use the standard formal 1@xmath18 and 3@xmath18 results based upon a @xmath19 distribution with two degrees of freedom .
this is based on the assumption that the determined shape is independent of the .
it is not obvious that this assumption holds , but as we shall see in the rest of this paper , it works and we discuss the validity of this assumption in [ sec : tbh_reliability ] .
here we describe the results of our dynamical modelling ( as described in [ method ] ) of 32 and ngc 3379 .
we compare the near - axisymmetric estimates with literature values and explore the effect of the possible triaxial shape upon the black hole mass estimate and orbital structure . to be able to compare our results directly to other studies it was important that we used galaxies that have a published obtained using axisymmetric schwarzschild models .
therefore we chose the nearby compact fast rotator e3 galaxy 32 @xcite , as it has been investigated by several independent authors ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is consistent with axisymmetry , as it shows regular rotation and has almost no isophotal twist .
for our modeling of 32 we assumed a distance of 0.79 mpc and used the surface brightness distribution , and the wide field data from @xcite and the data from @xcite , that probes well within the sphere of influence of the black hole .
the total number of kinematical observations used to constrain the models is 58 from and 964 from the observations , measured up to the gauss - hermite moment @xcite @xmath20 .
the kinematics were point - symmetrized as described in appendix [ pointsym ] . because 32 has a dispersion ( @xmath2190 ) below the instrumental dispersion of ( 120 ) the moments @xmath22 and higher are hard to measure , and thus have large associated errors @xcite .
however , they are still included in the models to ensure that the ( otherwise unconstrained ) reconstructed losvds do not deviate too far from a gaussian - like shape .
we did not use the high spatial resolution ( hr ) observations from @xcite , as the kinematic extraction of the low resolution data by @xcite is superior , due to a much better extraction method and significantly improved stellar templates . since we have high resolution data from , we chose not to re - reduce the hr data . and 189 .
the six upper - right panels show the intrinsic shape parameters @xmath23 and the ; the three lower left panels show the viewing angles @xmath24 . the combination of @xmath0 and @xmath1 is shown in a lambert equal area projection ( the half circle ) , seen down the north pole ( @xmath25-axis ) . in this panel the @xmath26 , @xmath27 and @xmath25 symbols
give the location of views down those axis .
the best - fitting models are very round and near oblate .
the preferred viewing angles are near the y - axis with nearly no intrinsic misalignment
. see [ method ] and [ m32shape],title="fig : " ] before we can get an estimate we need to explore the shape of 32 , because it is too expensive to make models and explore parameter space for the full range in shape and , as it would increase the computation time by a factor 10 .
therefore , we first fix the at @xmath28 , while varying the shape and the . by doing this
we assume that the derived shape does not depend on the specific we use .
we show in [ m32bhrecov ] and [ sec : tbh_reliability ] that this is a reasonable assumption .
the result of this triaxial modeling is shown in fig .
[ m32shape ] .
the best - fitting shape is nearly as round as the observed flattening allows and is consistent with an oblate axisymmetric spheroid with axis ratios @xmath29 .
this is fully consistent with expectations given the aligned bi - symmetric rotation in the observed velocity field .
curiously though , @xcite did find a best - fitting inclination @xmath30 ( equivalent to @xmath11=1 , @xmath12=0.73@xmath31 ) using axisymmetric modeling .
our results do agree , but our errors are different , due to our conservative error bars .
our ` axisymmetric ' models are still slightly triaxial ( @xmath32 ) and thus do allow for additional freedom from the triaxial orbital families .
the box orbits are important as our ( near ) axisymmetric models contain 8% box orbits within one . within the subset of axisymmetric models ,
the inclination is not well constrained at @xmath33@xmath34 .
essentially the full inclination range ( given the observed flattening ) is allowed at the 3@xmath18 level .
this result is interesting as it agrees with the observation made in @xcite , and in @xcite , that the axisymmetric models can not constrain the inclination well .
however , as is also shown in @xcite , we can marginalize over the allowed shapes to gain insight into the physical parameters - like and anisotropy - which tells us more about the galaxy than the inclination .
the is @xmath35 , identical to the value from the models of @xcite and @xcite ( when corrected for the different assumed distance and for galactic reddening ) .
( top ) and the ( bottom ) confidence levels as a function of shape for 32 .
the small red crosses indicate where models were computed and the big black cross indicates the best - fit model . from far left to far right the models change from prolate ( @xmath36 , @xmath37 ) , triaxial ( @xmath38 , @xmath37 ) and oblate ( @xmath39 , @xmath40 ) .
the vertical line denotes the transition from triaxial to oblate .
the models on the vertical line are the roundest possible models .
the contours show increasing level of confidence .
the inner and outer thicrk contour indicate a @xmath41 of 14.1 and 63 , which correspond to the 99.9% and 96% confidence on the and shape , respectively . the models with = 0 are relocated to 8@xmath42 , to place them within the plot.,title="fig : " ] to illustrate the effect of triaxiality , we show what happens when we sample different shapes while varying the and the .
we vary the shape from maximally prolate to maximally oblate ( 6 models ) and at different axisymmetric shapes ( 8 models ) .
these shapes continuously follow the @xmath12=0.76 and then the @xmath11=1 line in the upper leftmost plot ( @xmath11 vs. @xmath12 ) in fig .
[ m32shape ] .
this also fully encompasses the uncertainty in estimated shape . in fig .
[ m32bh ] we show the confidence levels ( @xmath41 ) of the and as a function of these shapes .
as expected the contours show that the roundest models are the best - fit ( @xmath430.66 , @xmath440.88 ) .
after marginalizing over the shapes , our mass estimate of @xmath45 is fully consistent with previous results of @xmath46 from @xcite , @xmath47 from @xcite and @xmath48 @xcite and @xmath28 from @xcite .
the two most nearly prolate models in this test are a significantly worse fit than the best - fit model . to illustrate this we show the kinematics of the observed stellar kinematics and four models , varying from oblate to prolate , in fig .
[ m32kin ] . while the oblate and round models reproduce the observations , the prolate models are unable to do so , and we can therefore conclude that 32 does not have prolate shape .
interestingly the best - fitting and do not change significantly for the individual triaxial models ( see fig .
[ m32bh ] ) , indicating that our initial assumption , i.e. that the recovered shape is not dependent on the initial , is probably reasonable in this case . to ensure that our models do not depend on the number of orbits
, we doubled the number of orbits , and repeated the above test and found no significant change on the best - fit error and the formal error bars .
overall , this shows that it is possible to recover with our triaxial method .
furthermore , we showed that , in this case , the recovered does not significantly depend on the intrinsic shape , and that the shape of 32is very close to oblate .
as the second test galaxy we chose the fast rotator ngc 3379 , which also has its black hole mass measured with orbit - based models , has a decoupled nuclear gas ring ( e.g. * ? ? ?
* ) and an estimate from the gas kinematics .
the velocity maps show regular rotation that is consistent with oblate axisymmetry , and probe to well within the sphere of influence .
detailed axisymmetric orbit - based and gas disc models are shown in . however this galaxy shows mild evidence for triaxiality : it has a small isophotal twist @xcite , an @xmath49 kinematical misalignment ( , * ? ? ? * ) and even shows hints of kinematical twist @xcite .
@xcite suggested it might be a triaxial s0 seen face - on , and @xcite has argued that this galaxy is mildly triaxial .
@xcite showed that ngc 3379 is consistent with being like all the other _ fast - rotators _ in the sample : nearly axisymmetric and with a disk , but seen almost face - on ( based on their v/@xmath18 diagram ) .
their interpretation would mean that the misalignment is due to a non - perfect circularity of the disk .
all this makes ngc 3379 an ideal test case for the black hole recovery of the triaxial schwarzschild machinery .
we use the surface brightness distribution ( based upon 2 and ground - based imaging ) , and stellar kinematics , and distance ( 10.28 mpc ) from for our modeling .
ngc 3379 is a very round galaxy , so it is difficult to establish the photometric pa with high precision , and thus the exact amount of misalignment is uncertain ( @xmath49 ) , and is in principle consistent with no misalignment ( 0 ) within 2@xmath18 .
as such this galaxy could be an axisymmetric oblate spheroid . exploiting this uncertainty , aligned the stellar kinematics with the photometry by rotating the velocity field by @xmath50 and bi - symmetrizing the kinematics .
this adjustment to the observed kinematics ensures that they are completely compatible with the axisymmetric modelling and these ` corrected ' observations were used for their determinations . to check our triaxial modeling machinery ,
we duplicate those models from to see if we recover the same given the same input parameters .
this test is important as our triaxial machinery is not capable of making a perfectly axisymmetric model , so even our _ axisymmetric model _ would be minutely triaxial ( @xmath32 ) and can still benefit from the additional orbital families , especially the strongly radial box orbits , which do not exist in a pure axisymmetric potential .
we reconstructed the parameter space in and from .
the results , presented in fig
. [ n3379axi ] , are identical to the estimate produced with the axisymmetric code , giving = @xmath51 and an of @xmath52 and is also similar to the results from @xcite .
the fit to the kinematics is not shown , as it is essentially identical to the figures in . to be completely confident we also tested with double the number of orbits and
this does not affect the recovered black hole mass and its confidence interval .
the confidence levels in the and the for an axisymmetric model of ngc 3379 made with the triaxial software . see
[ n3379axi ] .
the thick contour denotes @xmath41=14.1 , corresponding to a 99.9% confidence level .
the location of the minimum and the shape of the confidence contours match the results from ( see the rightmost panel in their fig .
8) , demonstrating that both codes provide identical results when given the same observables and shape.,title="fig : " ] to study the effect of kinematical misalignment on the estimate we modelled ngc 3379 using our triaxial machinery and with the kinematical misalignment intact and point - symmetrized ( see appendix [ pointsym ] ) kinematics .
we first searched through the shape distribution , while keeping fixed at @xmath53 .
the best - fitting kinematics is shown in fig .
[ n3379kin ] and [ n3379oasis ] .
the recovered shape , shown in fig . [ n3379shape ] is @xmath54 at 1 @xmath55 .
our shape is consistent with the result from @xcite of @xmath56 and @xmath57 . given
that his method is completely different , it is reassuring to see that we can even reproduce the shape of the confidence intervals ( but see * ? ? ?
the allowed viewing angles cover a large range , but prefer strongly inclined ( face - on ) views , which is also consistent with the results from @xcite .
it is important to notice that at the @xmath58 level the shape is not constrained well , allowing almost all viewing angles ( @xmath59 ) and a large allowed range in shapes ( see fig .
[ n3379shape ] ) .
a pure oblate axisymmetric spheroid is excluded at the @xmath60 level , and this happens because the axisymmetric model can not reproduce the twist in the zero velocity curve ( @xmath61 ) .
this is shown in fig .
[ n3379kin ] . the differences between the axisymmetric ( third panel from the left in fig .
[ n3379kin ] ) and triaxial ( second panel from the left in fig .
[ n3379kin ] ) model are not very prominent ; the most visible change can be seen in the ( twisting ) shape of the zero velocity curve .
since uses bi - symmetrized kinematics and the axisymmetric models use different intrinsic mass bins it is not possible to directly compare the @xmath14 of those models with ours . to do a direct comparison we recreated the original axisymmetric model from with the triaxial machinery , without bi - symmetrizing , but with the kinematic misalignment correction ( see [ n3379axi ] ) .
we expected that this _ original _ axisymmetric model would fit the data better than the axisymmetric model
because the latter does not correct for the misalignment but this is not what we found ( rightmost panel in fig .
[ n3379kin ] ) .
the kinematics of the best - fit triaxial axisymmetric model are statistically a significantly much better fit , with a difference in @xmath62 .
the differences show up in the twist of the zero velocity curve and the ` hexagonal shape ' of the velocity dispersion .
it seems that in this case the twist of this galaxy was overestimated due to the inaccurate determination of the photometric or kinematic pa .
both are possible because ngc 3379 is very round , has some isophotal twist and does not have a strong velocity field .
luckily , the triaxial modeling ( as opposed to the axisymmetric modeling ) does not depend on these measurements , as it only requires that the relative orientation between the photometry and kinematics be known .
given that the shape can not be constrained accurately our choice of the might influence the recovered shape . to test if this was the case we checked to see if the recovered shape would differ if we set @xmath63 .
the shape recovered was not different .
the intrinsic orientation of the central gas and dust disc in this galaxy is interesting due to its apparent misalignment of @xmath2150 degrees with the main body of the galaxy .
the only stable configurations in a stationary triaxial geometry are in the principal planes .
@xcite did a thorough analysis and concluded that , if the disc lies in a plane , the only option that he could not rule out was a polar ring .
@xcite showed that stellar photometry inside 1 arcsecond has a sudden pa twist of more then 20 degrees , towards the gas disc .
the gas disc has a size of 4 arcsec and is thus at larger radii than this stellar feature , but the two could be connected . also , found evidence that the gas disc might be warped using an ad - hoc model , indicating that a simple stable gas disc in a plane might actually not be a good description .
our current mass model does not include any isophotal twist and therefore does not predict if these two features are intrinsically aligned . to do that , a mass model with isophotal twist would be needed . as an added benefit
such a mass model with isophotal twist will lower the amount of possible deprojection , which would indirectly help constrain the shape of this galaxy . in our current modelling without isophotal twist ,
the allowed range in viewing angles is large , and it might thus be possible to place the ring in a principal plane . the ring is misaligned 45 from the photometric pa , so to place the ring in a principal plane we need a misalignment of the pa of 45
. our allowed models do include these extreme misalignments of @xmath64 at the 3@xmath18 level ( fig .
[ n3379shape ] ) , but the other viewing angles are then quite restricted : at @xmath64 only @xmath65@xmath66@xmath67@xmath68@xmath69 are within the 3@xmath18 contour , essentially disallowing the disc in either the @xmath70 or @xmath71 plane . from our modelling
the polar ring is the only possibility , as the inclination of our best - fit models lies below @xmath7243 , which is exactly what is needed for the polar ring according to @xcite
. recovered shape and viewing angles for the model of ngc 3379 with kinematic misalignment .
the thin and thick contour represent @xmath41=152 and 456 , which represent 1 and 3 @xmath18 confidence intervals .
best - fit model is nearly as round as observed and there is no strong constraint on the viewing angles .
figure layout identical to fig .
[ m32shape ] .
see [ method ] and [ sec:3379triaxmodels],title="fig : " ] now that we have a handle on the shape , we investigate whether the inferred shape affects the recovered .
we used the six best - fitting mass models , while changing the and the .
the results are shown in fig .
[ n3379bh ] . in this figure
the shape is parameterized as @xmath73 , which is completely arbitrary , but allows us to plot two - dimensional contours and show how the @xmath14 minimum is bracketed .
the best - fit shape is independent of the chosen , showing that the recovered shape does not depend on the fixed black hole mass that was used in the previous section .
the ( top ) and the ( bottom ) confidence levels as function of the different triaxial shapes ( see text ) .
the small red crosses represent the values for which models where computed and the big black crosses indicate the best - fit model , respectively .
the contours show increasing level of confidence .
the inner contour indicates a @xmath41 of 42.3 , further contours represent integer steps of 152 , which correspond to the 1,2,3 ( thick),4 , ...
@xmath18 confidence intervals on the shape.,title="fig : " ] the best fitting is @xmath74 and the is @xmath75 .
surprisingly , this estimate from the triaxial model is more than twice as large as @xmath76 from the edge - on axisymmetric estimate from . to show the quality of the models we show the kinematics and models with different in fig .
[ n3379oasis ] .
for all the different black hole masses , the mean velocity field is reproduced extremely well , but the dispersion is only properly reproduced by the @xmath77 model .
our result is also significantly above the axisymmetric ( near ) face - on model of ( * ? ? ?
* @xmath78 ) and .
our is just outside of the scatter of the relationship , as that predicts @xmath79 , placing our estimate on the heavy side of this relationship
. showed that the gas disc inside can not be fit by simple keplerian motion , implying that the gas disk is disturbed and may not be a good candidate for probing the dynamical bh mass .
nevertheless they constructed an ad hoc and non - unique model and estimated ( @xmath80 , which is lower than our estimate .
even with a mildly triaxial shape , the long axis and box orbits can contribute a significant fraction @xcite .
inside one intrinsic our model consists of 70% , 20% and 10% short axis , long axis tubes and box orbits respectively . the orbital structure ( fig .
[ fig : n3379structure ] ) of the triaxial model reveals that ngc 3379 is radially anisotropic inside the sphere of influence of the black hole ( ) and at most mildly radially anisotropic outside .
this is different from , which showed ngc 3379 to be isotropic inside the core radius . in our model ,
the box orbits contribute most of the mass inside , and thus the model becomes strongly radially anisotropic in the center ( fig .
[ fig : n3379structure ] ) .
this is very different from an axisymmetric model , in which box orbits can not exist .
the box orbits in the center could even be the cause of our high . in the face - on view of these models ( 28@xmath81 49 )
the stars on box orbits in the center have the highest dispersion in the direction perpendicular to the viewer and can therefore affect the central observed dispersion .
this is exactly opposite to a mechanism suggested by @xcite that essentially makes the black hole unnecessary by viewing the galaxy down the @xmath26-axis ( end - on ) the box orbits would then account for the high velocity dispersion in the center . orbital structure of 3379 .
top plot shows orbital anisotropy @xmath82 and the bottom plot shows the orbit type as a function of radius .
the model is mildly radially anisotropic outside and strongly anisotropic inside .
the short axis tubes ( blue solid line ) dominate the galaxy outside , while the box orbits ( black dotted line ) become more important inside .
the long axis tubes ( red dashed line ) are roughly constant at 20%.,title="fig : " ]
to get to our best fitting models of 32 and ngc 3379 , we had to first assume a , find the best - fitting shape and then find the best - fitting , because the alternative searching the full parameter space is computationally unpractical .
we search the shape parameter space first because an initial guess for can be done using and we know from @xcite that the influence of shape on the quality of the fit is much bigger than that of the black hole , usually more than a factor ten in @xmath15 .
however , this procedure is not guaranteed to find the global minimum . to ensure that we do find the minimum
, we marginalized over all the best - fitting shapes when searching for the .
this showed that for both galaxies we found that the best - fitting shape was unchanged by the improved and that even if we would have chosen a different beforehand we would still have found the same minimum .
surprisingly , this was also true for ngc 3379 , of which the shape is not constrained well and the estimate changes .
this is evidence that the recovery of the intrinsic shape and are independent , which simplifies future modeling .
[ m32bh ] and [ n3379bh ] also showed the reverse : that the estimate does not depend significantly on the shape and thus that our estimate is reliable , even if we do not get the intrinsic shape perfectly correct .
the addition of a dark halo in the models would be the next step .
@xcite showed that adding a dark halo can increase the if the constant is overestimated .
however this is unlikely to happen for 3379 as the best - fit dynamical is already very close to the derived from the stellar population @xcite .
the amount of dark matter in this galaxy has recently been constrained by @xcite , who measured stellar absorption line kinematics at large radii . by combining both this outer data plus the observations the effect of the dark matter on the black hole mass estimate
could now be determined .
the final test would be to model an analytical triaxial test galaxy with a realistic density profile and with a central black hole .
however , we are not aware of the existence of such self - consistent models distribution function , but the isodensity surfaces are then identical to the equipotentials , so that the model would be non - consistent . the value of using such models as test cases is then limited as we do not expect this arrangement of isodensity surfaces and equipotentials to arise in real spheroids , which moreover have anisotropic velocity distributions . ] . as an alternative ,
galaxies generated using n - body simulations could be used ( similar to * ? ? ?
. however we shall refrain from doing this test now , as we expect that it would not change the results for the two nearly oblate axisymmetric galaxies discussed here .
we have shown two applications of the triaxial orbit super - position method to galaxies that had their central black hole mass measured with axisymmetric models .
the reason for this was two - fold : first , we confirmed that we obtain the same when using our new triaxial implementation in the oblate axisymmetric limit , confirming that our method is reliable .
secondly , we explored what might happen when the assumption of axisymmetry is relaxed .
for the nearby elliptical galaxy 32 we obtained identical results to previous axisymmetric modeling finding a of ( @xmath83 .
our best - fitting shape of 32 is very close to oblate axisymmetric , excluding strongly triaxial shapes .
the viewing angles are not well constrained .
we also duplicated the models of ngc 3379 . after confirming the results from in the axisymmetric limit
, we expanded our search to include triaxiality .
while the shape is not strongly constrained , we found that this galaxy is seen almost face - on and can best be described with a triaxial model .
the best - fitting models are very round in the center and become fairly flattened at larger radii .
these results confirm the claims about the intrinsic shape from @xcite , @xcite and @xcite .
the black hole in this triaxial model weighs @xmath84 , which is two times bigger than the axisymmetric estimate from and @xcite .
we speculate that the difference in the estimate is due to the combined effect of changing from an edge - on to a nearly face - on view and the inclusion of the strongly radial box orbits in the modeling .
the significance of the change of the of ngc 3379 is unclear .
while this individual result is still within the scatter of the relation , it can greatly affect all empirical relations based on , like , if similar effects are seen in more galaxies .
if many intrinsically triaxial galaxies would have their black hole mass underestimated using axisymmetric modeling , the empirical relations would be systematically underestimating black hole masses at the top end , as we believe that triaxial galaxies dominate at the high mass end ( e.g. * ? ? ? * ; * ? ? ? * ) . to study this in more detail ,
it is necessary to study the nuclei of the most massive galaxies with triaxial models .
also , the axisymmetric models and their estimates have only been tested with purely axisymmetric ( and spherical ) test models .
this means that if axisymmetric models can not accurately recover black hole masses in triaxial galaxies , and if most galaxies are significantly triaxial , then this would also have an impact on black hole demography . a strong hint in this direction is given by @xcite , who showed that the axisymmetric models have difficulty estimating the in triaxial galaxies . based on their results and our ngc 3379 model
, we speculate that the estimates in triaxial galaxies derived using axisymmetric modeling will have systematic errors , because the recovery of the is strongly linked to the .
it is a pleasure to thank michele cappellari , eric emsellem , davor krajnovi , richard mcdermid , glenn van de ven and anne - marie weijmans for stimulating comments and lively discussions .
we would also like to thank kristen shapiro and richard mcdermid for the reduced kinematics , and eric emsellem and michele cappellari for the 32 kinematics .
this project is made possible through grant 614.000.301 from nwo , leids kerkhoven - bosscha fonds and the netherlands research school for astronomy nova .
part of this work is based on data obtained from the eso / st - ecf science archive facility , the canadian astronomy data centre operated by the national research council of canada with the support of the canadian space agency , 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 and the 1.3 m mcgraw - hill telescope of the mdm observatory .
the observations were obtained at the william herschel telescope , operated by the isaac newton group in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias .
most of the models presented in this paper were computed at the texas advanced computing center ( tacc ) at the university of texas at austin .
o. , parmeggiani g. , 1995 , , 109 , 572 j. , 1976 , , 177 , 19 j. , 1978 , , 183 , 501 m. , held e. v. , nieto j .- l .
, 1987 , , 94 , 1519 m. , vietri m. , held e. v. , lorenz h. , 1991 , , 371 , 535 m. , 2002 , , 333 , 400 m. , bacon r. , bureau m. , damen m. c. , davies r. l. , de zeeuw p. t. , emsellem e. , falcn - barroso j. , krajnovi d. , kuntschner h. , mcdermid r. m. , peletier r. f. , sarzi m. , van den bosch r. c. e. , van de ven g. , 2006 , , 366 , 1126 m. , copin y. , 2003 , , 342 , 345 m. , emsellem e. , 2004 , , 116 , 138 m. , emsellem e. , bacon r. , bureau m. , davies r. l. , de zeeuw p. t. , falcn - barroso j. , krajnovi d. , kuntschner h. , mcdermid r. m. , peletier r. f. , sarzi m. , van den bosch r. c. e. , van de ven g. , 2007 , , 379 , 418 f. , gerhard o. , coccato l. , arnaboldi m. , capaccioli m. , douglas n. g. , freeman k. c. , kuijken k. , merrifield m. r. , napolitano n. r. , noordermeer e. , romanowsky a. j. , debattista v. p. , 2009
, , 395 , 76 n. g. , napolitano n. r. , romanowsky a. j. , coccato l. , kuijken k. , merrifield m. r. , arnaboldi m. , gerhard o. , freeman k. c. , merrett h. r. , noordermeer e. , capaccioli m. , 2007 , , 664 , 257 e. , cappellari m. , krajnovi d. , van de ven g. , bacon r. , bureau m. , davies r. l. , de zeeuw p. t. , falcn - barroso j. , kuntschner h. , mcdermid r. , peletier r. f. , sarzi m. , 2007 , , 379 , 401 e. , cappellari m. , peletier r. f. , mcdermid r. m. , bacon r. , bureau m. , copin y. , davies r. l. , krajnovi d. , kuntschner h. , miller b. w. , de zeeuw p. t. , 2004 , , 352 , 721 e. , monnet g. , bacon r. , 1994 , , 285 , 723 l. , ford h. , 2005 , space science reviews , 116 , 523 k. , richstone d. , kormendy j. , lauer t. r. , ajhar e. a. , bender r. , dressler a. , faber s. m. , grillmair c. , magorrian j. , tremaine s. , 2000 , , 119 , 1157 k. , richstone d. , tremaine s. , lauer t. r. , bender r. , bower g. , dressler a. , faber s. m. , filippenko a. v. , green r. , grillmair c. , ho l. c. , kormendy j. , magorrian j. , pinkney j. , 2003 , , 583 , 92 k. , thomas j. , 2009 , , 700 , 1690 o. e. , 1988 , , 232 , 13p o. e. , 1993 , , 265 , 213 c. , de zeeuw p. t. , 1992 , , 389 , 79 c. l. , merritt d. , olling r. , valluri m. , bender r. , bower g. , danks a. , gull t. , hutchings j. , kaiser m. e. , maran s. , weistrop d. , woodgate b. , malumuth e. , nelson c. , plait p. , lindler d. , 2001 , , 550 , 668 j. , 1985 , , 295 , 73 j. , 2004 , in ho l. c. , ed .
, coevolution of black holes and galaxies the stellar - dynamical search for supermassive black holes in galactic nuclei .
j. , bender r. , 1996 , , 464 , l119
d. , bacon r. , cappellari m. , davies r. l. , de zeeuw p. t. , emsellem e. , falcn - barroso j. , kuntschner h. , mcdermid r. m. , peletier r. f. , sarzi m. , van den bosch r. c. e. , van de ven g. , 2008 , , 390 , 93 d. , cappellari m. , de zeeuw p. t. , copin y. , 2006 , , 366 , 787 d. , cappellari m. , emsellem e. , mcdermid r. m. , de zeeuw p. t. , 2005 , , 357 , 1113 t. r. , faber s. m. , ajhar e. a. , grillmair c. j. , scowen p. a. , 1998 , , 116 , 2263 t. r. , faber s. m. , gebhardt k. , richstone d. , tremaine s. , ajhar e. a. , aller m. c. , bender r. , dressler a. , filippenko a. v. , green r. , grillmair c. j. , ho l. c. , kormendy j. , magorrian j. , pinkney j. , siopis c. , 2005 , , 129 , 2138 j. , 2006 , , 373 , 425 g. , bacon r. , emsellem e. , 1992 , , 253 , 366 r. f. , 1993 , , 271 , 51 h. , de zeeuw p. t. , cretton n. , van der marel r. p. , carollo c. m. , 1997 , , 488 , 702 m. , 1979 , , 232 , 236 k. l. , cappellari m. , de zeeuw t. , mcdermid r. m. , gebhardt k. , van den bosch r. c. e. , statler t. s. , 2006 , , 370 , 559 ( s06 ) t. s. , 1994 , , 108 , 111 t. s. , 2001 , , 121 , 244 t. s. , dejonghe h. , smecker - hane t. , 1999 , , 117 , 126 j. , jesseit r. , naab t. , saglia r. p. , burkert a. , bender r. , 2007 , , 381 , 1672 s. , gebhardt k. , bender r. , bower g. , dressler a. , faber s. m. , filippenko a. v. , green r. , grillmair c. , ho l. c. , kormendy j. , lauer t. r. , magorrian j. , pinkney j. , richstone d. , 2002 , , 574 , 740 m. , merritt d. , emsellem e. , 2004 , , 602 , 66 g. , de zeeuw p. t. , van den bosch r. c. e. , 2008 , , 385 , 614 ( vdv08 ) g. , van den bosch r. c. e. , verolme e. k. , de zeeuw p. t. , 2006 , , 445 , 513 r. , de zeeuw t. , gebhardt k. , noyola e. , van de ven g. , 2006 , , 641 , 852 r. c. e. , van de ven g. , 2009 , , 398 , 1117 r. c. e. , van de ven g. , verolme e. k. , cappellari m. , de zeeuw p. t. , 2008 , , 385 , 647 r. p. , cretton n. , de zeeuw p. t. , rix h. , 1998 , , 493 , 613 r. p. , franx m. , 1993 , , 407 , 525 p. o. , 1984 , , 287 , 475 e. k. , cappellari m. , copin y. , van der marel r. p. , bacon r. , bureau m. , davies r. l. , miller b. m. , de zeeuw p. t. , 2002 , , 335 , 517 a .-
, cappellari m. , bacon r. , de zeeuw p. t. , emsellem e. , falcn - barroso j. , kuntschner h. , mcdermid r. m. , van den bosch r. c. e. , van de ven g. , 2009 , , 398 , 561
almost all orbit - based modellers usually symmetrize their input kinematics ( e.g. * ? ? ?
* ; * ? ? ?
* ) , as their model assumptions enforce bi- or point symmetry anyway .
there are numerous reasons to do this .
most commonly this is done to reduce the noise in the observations , facilitate @xmath85-by - eye comparison , force the observations to be bi- or point symmetric , or because they want to reduce the number of observables .
it is also a good method to remove systematic effects , including systematic offsets in the odd moments .
the symmetrization also improves the linear gauss - hermite moments reconstruction used in our models @xcite .
symmetrizing velocity maps is a degenerate problem and there are an infinite number of _ solutions _ and none are perfect . in this appendix
we describe a novel method to point - symmetrize velocity fields .
it is accurate , conserves the amount of spatial information and propagates the errors , without making unnecessary assumptions .
the idl - script itself is available in the electronic tarball on arxiv.org . at the very least
, this method is useful to determine the systematic offsets in the odd velocity moments .
the multi - step process is depicted in fig .
[ figsymm ] .
we make three assumptions : first we assume that the gauss - hermite moments are orthogonal and un - correlated .
this assumption is also enforced by the dynamical model itself and is therefore ` fair ' , but not true .
second , we assume that the velocity field varies linearly along spatial coordinates , which is generally true to first order .
third , we assume that we can safely symmetrize without worrying about the psf .
the kinematics are typically binned @xcite .
this means that most kinematic observations ( and associated error ) span several spaxels ( i.e. lenslets ) .
but still every spaxel has an individual flux and signal - to - noise measurement .
we assign every spaxel an error based on its relative flux within that bin and the error of the kinematic observation ( @xmath86 ) of that bin as follows . where @xmath88 is the total flux in that bin and @xmath89 is the flux in the spaxel @xmath90 ( * 2 * in fig .
[ figsymm ] ) .
this definition conserves the error of the original bin when the spaxels are combined back into a bin later .
the spaxels can be recombined into the bins using the weighted mean : @xmath92 are the kinematics of each individual spaxel , which are yet unknown . to estimate them we construct a linear interpolation of the velocity field over the individual spaxels , using the bin centers as the nodes in the linear interpolation ( * 2 * in fig .
[ figsymm ] ) . to be as conservative as possible we need to make sure that after this linear interpolation the combined spaxels still reproduce the original kinematics .
using eqn .
[ binrecomb ] we compute for each bin what the difference is between the original kinematics and the interpolated one . this difference @xmath93 is then used to self - calibrate the linear interpolation step , by adjusting the velocities at the nodes using @xmath93 and recomputing the interpolation and this is repeated as needed .
this self - calibration typically converges quickly and ensures that the velocity map interpolated over the spaxels reproduces the original binned velocity map when binned back ( * 3 * in fig .
[ figsymm ] ) . in a triaxial stellar system
the odd kinematic moments have to be anti - correlated across the center .
the reduction pipeline centers the central spaxel on the galaxy center .
therefore it is trivial to find the point symmetric counterpart of each spaxel .
we can now use this information to correct for systematic offsets ( e.g. recession velocity and template mismatch ) in the velocity map of the odd moments .
it is important to do this before symmetrizing , because bins that do not have a point symmetric counterpart will otherwise not get corrected for this offset .
for each spaxel that has a counterpart , we compute their weighted mean velocity and combined error .
we then take those values plus the value of the central spaxel and compute the total weighted average , which is the systematic offset .
we add this systematic offset to the velocities of the individual spaxels ( * 4 * in fig .
[ figsymm ] ) .
now we do the actual point - symmetrizing of the spaxels .
this step is almost identical to the previous step , but has some significant differences . for each spaxel that has a counterpart ,
we compute their weighted mean velocity and combined error .
we then replace the values of those spaxels with the weighted mean velocity that we just computed and replace their errors with the combined error multiplied by @xmath94 .
the error is corrected , because we just stored the velocity in two spaxels and we do not want to artificially decrease our errors ( * 5 * in fig .
[ figsymm ] ) .
after we have point - symmetrized all the spaxels ( once ) in this way , we combine the spaxels ( and their errors ) back into the original bins .
now the point symmetrization is complete ( * 6 * in fig .
[ figsymm ] ) .
the end result has exactly the same bins as the original observations and has thus conserved the spatial information . in principle
it is possible to adapt this routine to bi - symmetrization too .
however , in the bi - symmetric case no unique counterpart of the spaxels exist , because the position angle ( pa ) of the ( presumed axisymmetric ) galaxy does not have to coincide with the pixel grid . to extend our algorithm to bi - symmetrization another interpolation or supersampling needs to be added .
alternatively , appendix c in @xcite describes a way to bi - symmetrize velocity fields , which interpolates the kinematics between the bin centroids , without self - calibrating , using an un - weighted mean and without propagating the errors .
bi - symmetrization also critically depends on the existence and determination of the global pa . since we model ( axisymmetric ) galaxies here with our triaxial method
, we can avoid these issues by applying point - symmeterization instead of bi - symmeterization .
the method described here is not perfect and no method for velocity map symmetrizing ever will be . in principle
it should be possible to relax some of our assumptions and design a better algorithm . instead
, it would be much more useful to update the orbit - based models , so that they are robust against the systematics in the unsymmetrized kinematics .
|
most of the super massive black hole mass ( ) estimates based on stellar kinematics use the assumption that galaxies are axisymmetric oblate spheroids or spherical . here
we use fully general triaxial orbit - based models to explore the effect of relaxing the axisymmetric assumption on the previously studied galaxies 32 and ngc 3379 .
we find that 32 can only be modeled accurately using an axisymmetric shape viewed nearly edge - on and our black hole mass estimate is identical to previous studies .
when the observed 5 kinematical twist is included in our model of ngc 3379 , the best shape is mildly triaxial and we find that our best - fitting black hole mass estimate doubles with respect to the axisymmetric model . this particular black hole mass estimate is still within the errors of that of the axisymmetric model and consistent with the relationship .
however , this effect may have a pronounced impact on black hole demography , since roughly a third of the most massive galaxies are strongly triaxial .
[ firstpage ] black hole physics , galaxies : elliptical and lenticular , cd - galaxies : kinematics and dynamics - galaxies : individual : ngc 3379 , m32 galaxies : structure , galaxies : nuclei
|
a notable feature in the development of hierarchical and bayesian nonparametric models is the role of mixing measures , which help to combine relatively simple models into richer classes of statistical models @xcite . in recent years
the mixture modeling methodology has been significantly extended by many authors taking the mixing measure to be random and infinite - dimensional via suitable priors constructed in a nested , hierarchical and nonparametric manner .
this results in rich models that can fit more complex and high - dimensional data ( see , e.g. , @xcite for several examples of such models , as well as a recent book @xcite ) .
the focus of this paper is to analyze convergence behavior of the posterior distribution of latent mixing measures as they arise in several mixture models , including finite mixtures and the infinite dirichlet process mixtures .
let @xmath1 denote a discrete probability measure .
atoms @xmath2 s are elements in space @xmath3 , while vector of probabilities @xmath4 lies in a @xmath5-dimensional probability simplex . in a mixture setting , @xmath6 is combined with a likelihood density @xmath7 with respect to a dominating measure @xmath8 on @xmath9 , to yield the mixture density : @xmath10 . in a clustering application ,
atoms @xmath2 s represent distinct behaviors in a heterogeneous data population , while mixing probabilities @xmath11 s are the associated proportions of such behaviors . under this interpretation , there is a need for comparing and assessing the quality of mixing measure @xmath12 estimated on the basis of available data .
an important work in this direction is by chen @xcite , who used the @xmath13 metric on the cumulative distribution functions on the real line to study convergence rates of the mixing measure @xmath6 .
chen s results were subsequently extended to a bayesian estimation setting for a univariate mixture model @xcite .
these works were limited to only univariate and finite mixture models , with @xmath14 bounded by a known constant , while our interest is when @xmath14 may be unbounded and @xmath3 is multidimensional or even an abstract space .
the analysis of consistency and convergence rates of posterior distributions for bayesian estimation has seen much progress in the past decade .
key recent references include @xcite .
analysis of specific mixture models in a bayesian setting has also been studied @xcite .
all these works primarily focus on the convergence behavior of the posterior distribution of the data density @xmath15 . on the other hand ,
results concerned with the convergence behavior of latent mixing measures @xmath6 are quite rare .
notably , the analysis of convergence for mixing ( smooth ) densities often arises in the context of frequentist estimation for deconvolution problems , mainly within the kernel density estimation method ( e.g. , @xcite ) .
we also note recent progress on consistent parameter estimation for certain finite mixture models , for example , in an overfitted setting @xcite or with an emphasis on computational efficiency @xcite .
the primary contribution of this paper is to show that the wasserstein distances provide a natural and useful metric for the analysis of convergence for latent mixing measures in mixture models , and to establish convergence rates of posterior distributions in a number of well - known bayesian nonparametric and mixture models .
wasserstein distances originally arose in the problem of optimal transportation @xcite .
although not as popular as well - known divergence functionals such as kullback leibler , total variation and hellinger distances , wasserstein distances have been utilized in a number of statistical contexts ( e.g. , @xcite ) . for discrete probability measures , they can be obtained by a minimum matching ( or moving ) procedure between the sets of atoms that provide support for the measures under comparison , and consequentially are simple to compute .
suppose that @xmath3 is equipped with a metric @xmath16 .
let @xmath17 .
then , for a given @xmath18 the @xmath19 wasserstein metric on the space of discrete probability measures with support in @xmath20 , namely , @xmath21 , is @xmath22^{1/r},\ ] ] where the infimum is taken over all joint probability distributions on @xmath23 \times[1,\ldots , k']$ ] such that @xmath24 and @xmath25 .
as clearly seen from this definition , wasserstein distances inherit directly the metric of the space of atomic support @xmath3 , suggesting that they can be useful for assessing estimation procedures for discrete measures in hierarchical models .
it is worth noting that if @xmath26 is a sequence of discrete probability measures with @xmath14 distinct atoms and @xmath27 tends to some discrete measure @xmath28 in the @xmath29 metric , then @xmath27 s ordered set of atoms must converge to @xmath28 s atoms in @xmath16 after some permutation of atom labels .
thus , in the clustering application illustrated above , convergence of mixing measure @xmath6 may be interpreted as the convergence of distinct typical behavior @xmath2 s that characterize the heterogeneous data population .
a hint for the relevance of the wasserstein distances can be drawn from an observation that the @xmath13 distance for the cdfs of univariate random variables , as studied by chen @xcite , is in fact a special case of the @xmath30 metric when @xmath31 .
the plan for the paper is as follows .
section [ sec - basic ] investigates the relationship between wasserstein distances for mixing measures and well - known divergence functionals for mixture densities in a mixture model .
we produce a simple lemma which gives an upper bound on @xmath0-divergences between mixture densities by certain wasserstein distances between mixing measures .
this implies that @xmath29 topology can be stronger than those induced by divergences between mixture densities .
next , we consider various identifiability conditions under which convergence of mixture densities entails convergence of mixing measures in a wasserstein metric .
we present two key theorems , which provide upper bounds on @xmath32 in terms of divergences between @xmath15 and @xmath33 .
theorem [ thm - bound - supnorm ] is applicable to mixing measures with a bounded number of atomic support , generalizing a result from @xcite .
theorem [ thm - convolution ] is applicable to mixing measures with an unbounded number of support points , but is restricted to only convolution mixture models .
section [ sec - convergence ] focuses on the convergence of posterior distributions of latent mixing measures in a bayesian nonparametric setting . here
, the mixing measure @xmath6 is endowed with a prior distribution @xmath34 .
assuming an @xmath35-sample @xmath36 that is generated according to @xmath37 , we study conditions under which the postetrior distribution of @xmath6 , namely , @xmath38 , contracts to the `` truth '' @xmath28 under the @xmath39 metric , and provide the contraction rates . in theorems [ thm - convergence-1 ] and [ thm - convergence-2 ] of section [ sec - convergence ] , we establish the convergence rates for the posterior distribution for @xmath6 in terms of the @xmath39 metric .
these results are proved using the standard approach of ghosal , ghosh and van der vaart @xcite .
our convergence theorems have several notable features .
they rely on separate conditions for the prior @xmath34 and likelihood function @xmath0 , which are typically simpler to verify than conditions formulated in terms of mixture densities .
the claim of convergence in wasserstein metrics is typically stronger than the weak convergence induced by the hellinger metric in the existing work mentioned above . in section [ sec - examples ] posterior consistency and convergence rates of latent mixing measures
are derived , possibly for the first time , for a number of well - known mixture models in the literature , including finite mixtures of multivariate distributions and infinite mixtures based on dirichlet processes . for finite mixtures with a bounded number of atomic support in @xmath40 , the posterior convergence rate for mixing measures is @xmath41 under suitable identifiability conditions .
this rate is optimal up to a logarithmic factor in the minimax sense . for dirichlet process mixtures defined on @xmath40 ,
specific rates are established under smoothness conditions of the likelihood density function @xmath0 . in particular , for ordinary smooth likelihood densities with smoothness @xmath42 ( e.g. , laplace ) , the rate achieved is @xmath43 for any @xmath44 . for supersmooth likelihood densities with smoothness @xmath42 ( e.g. , normal ) ,
the rate achieved is @xmath45 . for ease of notation
, we also use @xmath46 in place of @xmath47 and @xmath48 in place of @xmath49 for likelihood density functions . divergences ( distances ) studied in the paper include the total variational distance : @xmath50 , hellinger distance : @xmath51 and kullback
leibler divergence : @xmath52 these divergences are related by @xmath53 and @xmath54 .
@xmath55 denotes the covering number of the metric space @xmath56 , that is , the minimum number of @xmath57-balls needed to cover the entire space @xmath3 .
@xmath58 denotes the packing number of @xmath56 , that is , the maximum number of points that are mutually separated by at least @xmath57 in distance .
they are related by @xmath59 .
@xmath60 denotes the diameter of @xmath3 .
let @xmath56 be a space equipped with a nonnegative distance function @xmath61 , that is , a function that satisfies @xmath62 if and only if @xmath63 . if , in addition , @xmath16 is symmetric ( @xmath64 ) and satisfies the triangle inequality , then it is a proper metric .
a discrete probability measure @xmath6 on a measure space equipped with the borel sigma algebra takes the form @xmath65 for some @xmath66 , where @xmath67 denotes the proportion vector , while @xmath68 are the associated atoms in @xmath3 .
@xmath69 has to satisfy @xmath70 and @xmath71 . [ with a bit abuse of notation , we write @xmath72 when @xmath73 has countably infinite support points represented by the infinite sequence of atoms @xmath74 and the associated sequence of probability mass @xmath75 . ]
likewise , @xmath76 is another discrete probability measure that has at most @xmath77 distinct atoms .
let @xmath78 denote the space of all discrete probability measures with at most @xmath14 atoms .
let @xmath79 , the set of all discrete measures with finite support .
finally , @xmath80 denotes the space of all discrete measures ( including those with countably infinite support ) .
let @xmath81^{k\times k'}$ ] denote a joint probability distribution on @xmath82 that satisfies the marginal constraints : @xmath83 and @xmath84 for any @xmath85 .
we also call @xmath86 a coupling of @xmath75 and @xmath87 .
let @xmath88 denote the space of all such couplings .
we start with the general transportation distance : [ def - cd ] let @xmath16 be a distance function on @xmath3 .
the transportation distance for two discrete measures @xmath89 and @xmath90 is @xmath91 when @xmath3 is a metric space ( e.g. , @xmath40 ) and @xmath16 is taken to be its metric , we revert to the more standard notation of wasserstein metrics , @xmath92 and @xmath93 . however , @xmath94 will be employed when @xmath16 may be a general or a nonstandard distance function or metric . from here on , probability measure
@xmath95 plays the role of the mixing distribution in a mixture model
. let @xmath96 denote the density ( with respect to a dominating measure @xmath8 ) of a random variable @xmath97 taking values in @xmath9 , given parameter @xmath98 .
for the ease of notation , we also use @xmath99 for @xmath100 . combining @xmath6 with the likelihood
function @xmath0 yields a mixture distribution for @xmath97 that takes the following density : @xmath101 a central theme in this paper is to explore the relationship between wasserstein distances of mixing measures @xmath102 , for example , @xmath103 , and divergences of mixture densities @xmath104 .
divergences that play important roles in this paper are the total variational distance , the hellinger distance and the kullback
leibler distance .
all these are in fact instances of a broader class of divergences known as the @xmath0-divergences ( csiszr @xcite ; ali and silvey @xcite ) : let @xmath105 denote a convex function .
an @xmath0-divergence ( or ali silvey distance ) between two probability densities @xmath46 and @xmath48 is defined as @xmath106 .
likewise , the @xmath0-divergence between @xmath15 and @xmath33 is @xmath107 .
@xmath0-divergences can be used as a distance function or metric on @xmath3 .
when @xmath16 is taken to be an @xmath0-divergence , @xmath108 , for a convex function @xmath109 , we call the corresponding transportation distance a _ composite _ transportation distance : @xmath110 for @xmath111 we obtain the squared hellinger ( @xmath112 ) , which induces the composite transportation distance @xmath113 . for @xmath114
we obtain the variational distance ( @xmath115 ) , which induces @xmath116 . for @xmath117
, we obtain the kullback leiber divergence ( @xmath118 ) , which induces @xmath119 .
[ lem - gen - ineq ] let @xmath120 such that both @xmath121 and @xmath122 are finite for some convex function @xmath109 .
then , @xmath123 .
this lemma highlights a simple direction in the aforementioned relationship : any @xmath0-divergence between mixture distributions @xmath15 and @xmath33 is dominated by a transportation distance between mixing measures @xmath6 and @xmath124 . as will be evident in the sequel , this basic inequality is also handy in enabling us to obtain upper bounds on the power of tests .
it also proves useful for establishing lower bounds on small kullback
leibler ball probabilities in the space of mixture densities @xmath15 in terms of small ball probabilities in the metric space @xmath56 .
the latter quantities are typically easier to obtain estimates for than the former .
[ example - gaussian ] suppose that @xmath125 , @xmath16 is the euclidean metric , @xmath96 is the multivariate normal density @xmath126 with mean @xmath127 and identity covariance matrix , then @xmath128 .
so , @xmath129 .
the above lemma then entails that @xmath130 .
similarly , for the kullback
leibler divergence , since @xmath131 , by lemma [ lem - gen - ineq ] , @xmath132 .
for another example , if @xmath96 is a gamma density with location parameter @xmath127 , @xmath3 is a compact subset of @xmath133 that is bounded away from 0 . then @xmath134 .
this entails that @xmath135 .
lemma [ lem - gen - ineq ] shows that for many choices of @xmath16 , @xmath94 yields a stronger topology on @xmath21 than the topology induced by @xmath0-divergences on the space of mixture distributions @xmath15 . in other words ,
convergence of @xmath15 may not imply convergence of @xmath6 in transportation distances . to ensure this property ,
additional conditions are needed on the space of probability measures @xmath21 , along with identifiability conditions for the family of likelihood functions @xmath136 .
the classical definition of teicher @xcite specifies the family @xmath137 to be identifiable if for any @xmath138 , @xmath139 implies that @xmath140 .
we need a slightly stronger version , allowing for the inclusion for discrete measures with infinite support : the family @xmath137 is finitely identifiable if for any @xmath141 and @xmath142 , @xmath143 for almost all @xmath144 implies that @xmath140 . to obtain convergence rates ,
we also need the notion of strong identifiability of @xcite , herein adapted to a multivariate setting .
[ def - identifiability ] assume that @xmath145 and @xmath16 is the euclidean metric .
the family @xmath137 is strongly identifiable if @xmath96 is twice differentiable in @xmath127 and for any finite @xmath14 and @xmath14 different @xmath146 , the equality @xmath147 implies that @xmath148 , @xmath149 for @xmath150 . here
, for each @xmath151 , @xmath152 and @xmath153 denote the gradient and the hessian at @xmath2 of function @xmath154 , respectively .
finite identifiability is satisfied for the family of gaussian distributions for both mean and variance parameters @xcite ; see also theorem 1 of @xcite .
chen identified a broad class of families , including the gaussian family , for which the strong identifiability condition holds @xcite .
define @xmath155 if @xmath156 and @xmath157 otherwise . also define @xmath158 if @xmath156 and @xmath157 otherwise .
the notion of strong identifiability is useful via the following key result , which generalizes chen s result to @xmath3 of arbitrary dimensions .
[ thm - bound - supnorm ] suppose that @xmath3 is a compact subset of @xmath40 , the family @xmath159 is strongly identifiable , and for all @xmath160 , the hessian matrix @xmath161 satisfies a uniform lipschitz condition @xmath162 for all @xmath163 and some fixed @xmath164 and @xmath165 .
then , for fixed @xmath166 , where @xmath167 , @xmath168 the assertion also holds with @xmath169 being replaced by @xmath170 .
suppose that @xmath28 has exactly @xmath14 distinct support points in @xmath3 ( i.e. , @xmath1 where @xmath171 for all @xmath150 ) .
then , an examination of the proof reveals that the requirement that @xmath3 be compact is not needed .
indeed , if there is a sequence of @xmath172 such that @xmath173 , then it is simple to show that there is a subsequence of @xmath27 that also has @xmath14 distinct atoms , which converge in the @xmath16 metric to the set of @xmath14 atoms of @xmath28 ( up to some permutation of the labels ) .
the proof of the theorem proceeds as before . for the rest of this paper , by strong identifiability we always mean
conditions specified in theorem [ thm - bound - supnorm ] so that equation ( [ eqn - bound - supnorm ] ) can be deduced .
this practically means that the conditions specified by ( [ eqn - uniform ] ) and ( [ eqn - lipschitz ] ) be given , while the compactness of @xmath3 may sometimes be required .
next , we state a counterpart of theorem [ thm - bound - supnorm ] for @xmath174 , that is , mixing measures with a potentially unbounded number of support points .
we restrict our attention to convolution mixture models on @xmath40 .
that is , the likelihood density function @xmath96 , with respect to lebesgue , takes the form @xmath175 for some multivariate density function @xmath0 on @xmath40 .
thus , @xmath176 and @xmath177 .
the key assumption is concerned with the smoothness of density function @xmath0 .
this is characterized in terms of the tail behavior of the fourier transform @xmath178 of @xmath179 .
we consider both ordinary smooth densities ( e.g. , laplace and gamma ) and supersmooth densities ( e.g. , normal ) .
[ thm - convolution ] suppose that @xmath102 are probability measures that place full support on a bounded subset @xmath180 .
@xmath0 is a density function on @xmath40 that is symmetric ( around 0 ) , that is , @xmath181 for any borel set @xmath182 .
moreover , assume that @xmath183 for all @xmath184 .
suppose that @xmath185 as @xmath186 @xmath187 for some positive constants @xmath188 and @xmath42 .
then for any @xmath189 , there is some constant @xmath190 dependent only on @xmath191 and @xmath192 such that @xmath193 as @xmath194 .
suppose that @xmath195 as @xmath186 @xmath187 for some positive constants @xmath196 .
then there is some constant @xmath197 dependent only on @xmath198 and @xmath42 such that @xmath199 as @xmath194 .
the theorem does not actually require that mixing measures @xmath102 be discrete .
moreover , from the proof of the theorem , the condition that the support points of @xmath6 and @xmath124 lie in a bounded subset of @xmath40 can be removed and replaced by the boundedness of a given moment of the mixing measures .
the upper bound remains the same for the supersmooth likelihood case .
for the ordinary smooth case , we obtain a slightly weaker upper bound for @xmath200 . for the standard normal density on @xmath40 , @xmath201
, we obtain that @xmath202 as @xmath203 [ so that @xmath204 , by lemma [ lem - gen - ineq ] ] . for a laplace density on @xmath133 , for example , @xmath205 , then @xmath206 for any @xmath207 , as @xmath208 .
we turn to a study of convergence of mixing measures in a bayesian setting .
let @xmath36 be an i.i.d .
sample according to the mixture density @xmath209 , where @xmath0 is known , while @xmath210 for some unknown mixing measure in @xmath78 .
the true number of support points for @xmath6 may be unknown ( and/or unbounded ) . in the bayesian estimation framework , @xmath6 is endowed with a prior distribution @xmath34 on a suitable measure space of discrete probability measures in @xmath21 .
the posterior distribution of @xmath6 is given by , for any measurable set @xmath211 , @xmath212 we shall study conditions under which the posterior distribution is consistent , that is , it concentrates on arbitrarily small @xmath39 neighborhoods of @xmath28 , and establish the rates of the convergence .
we follow the general framework of ghosal , ghosh and van der vaart @xcite , who analyzed convergence behavior of posterior distributions in terms of @xmath0-divergences such as hellinger and variational distances on the mixture densities of the data . in the following
we formulate two convergence theorems for the mixture model setting ( which can be viewed as counterparts of theorems 2.1 and 2.4 of @xcite ) .
a notable feature of our theorems is that conditions ( e.g. , entropy and prior concentration ) are stated directly in terms of the wasserstein metric , as opposed to @xmath0-divergences on the mixture densities .
they may be typically separated into independent conditions for the prior for @xmath6 and the likelihood family and are simpler to verify for mixture models .
the following notion plays a central role in our general results .
fix @xmath213 .
let @xmath214 .
define the hellinger information of the @xmath39 metric for subset @xmath215 as a real - valued function on the real line @xmath216 : @xmath217 note the dependence of @xmath218 on the ( fixed ) @xmath28 , but this is suppressed for ease of notation .
it is obvious that @xmath218 is a nonnegative and nondecreasing function .
the following characterizations of @xmath218 are simple consequences of theorems [ thm - bound - supnorm ] and [ thm - convolution ] : [ thm-1 ] suppose that @xmath166 , and both @xmath78 and @xmath215 are compact in the wasserstein topology .
in addition , assume that the family of likelihood functions is finitely identifiable .
then , @xmath219 for all @xmath220 .
suppose that @xmath180 is compact , and the family of likelihood functions is strongly identifiable as specified in theorem [ thm - bound - supnorm ] .
then , for @xmath14 there is a constant @xmath221 such that @xmath222 for all @xmath223 .
suppose that @xmath180 is bounded , and the family of likelihood functions is ordinary smooth with parameter @xmath42 , as specified in theorem [ thm - convolution ] .
then , for any @xmath224 there is some constant @xmath225 such that @xmath226 for all . for the supersmooth likelihood family
, we have @xmath227 $ ] for all @xmath220 .
a main ingredient in the analysis of convergence of posterior distributions is through proving the existence of tests for subsets of parameters of interest .
a test @xmath228 is a measurable indicator function of the i.i.d .
sample @xmath36 . for a fixed pair of measures @xmath229 such that @xmath230 , where @xmath215 is a given subset of @xmath21 , consider tests for discriminating @xmath28 against a closed wasserstein ball centered at @xmath231 .
write @xmath232 the following lemma highlights the role of the hellinger information : [ lem - convexball ] suppose that @xmath56 is a metric space .
fix @xmath213 and consider @xmath214 . given @xmath230 ,
let @xmath233 .
suppose that either one of the following two sets of conditions holds : @xmath215 is a convex set , in which case , let @xmath234 .
@xmath215 is nonconvex , while @xmath3 is a totally bounded and bounded set . in addition , for some constants @xmath235 , @xmath236 for any likelihood functions @xmath237 in the family . in this case , define @xmath238 then , there exist tests @xmath239 that have the following properties : @xmath240 , \\
\label{eqn - type-2 } \sup_{g\in{\mathcal{g}}\cap b_w(g_1,r/2 ) } p_{g}(1-\varphi_n ) & \leq & \exp \bigl[-n \psi_{\mathcal{g}}(r)/8\bigr].\end{aligned}\ ] ] here , @xmath241 denotes the expectation under the mixture distribution given by density @xmath242 .
the set of conditions
( ii ) is needed when @xmath215 is not convex , an example of which is @xmath243 , the space of measures with at most @xmath14 support points in @xmath3 .
it is interesting to note that the loss in test power due to the lack of convexity is captured by the local entropy term @xmath244 .
this quantity is defined in terms of the packing by small @xmath39 balls whose radii are specified by the hellinger information .
hence , this packing number can be upper bounded by exploiting a lower bound of the hellinger information . next
, the existence of the test can be shown for discriminating @xmath28 against the complement of a closed wasserstein ball : [ lem - ballcomplement ] assume that conditions of lemma [ lem - convexball ] hold and define @xmath245 as in lemma [ lem - convexball ] .
suppose that for some nonincreasing function @xmath246 , some @xmath247 and every @xmath248 , @xmath249 then , for every @xmath248 , for any @xmath250 , there exist tests @xmath228 ( depending on @xmath251 ) such that @xmath252 , \\ \label{eqn
- type-2b}\quad \sup_{g\in{\mathcal{g}}\dvtx { w_2}(g_0,g ) > t_0\varepsilon } p_{g } ( 1- \varphi_n ) & \leq & \exp\bigl[-n \psi_{{\mathcal{g}}}(t_0 \varepsilon)/8\bigr].\end{aligned}\ ] ] it is interesting to observe that function @xmath246 is used to control the packing number of thin layers of wasserstein balls ( a similar quantity that also arises via the peeling argument in @xcite ( theorem 7.1 ) ) , in addition to the packing number @xmath253 of small wasserstein balls in terms of smaller wasserstein balls whose radii are specified in terms of the hellinger information function . as in the previous lemma , the latter packing number appears intrinsic to the analysis of convergence for mixing measures .
the preceeding two lemmas provide the core argument for establishing the following general posterior contraction theorems for latent mixing measures in a mixture model .
the following two theorems have three types of conditions .
the first is concerned with the size of support of @xmath34 , often quantified in terms of its entropy number .
estimates of the entropy number defined in terms of wasserstein metrics for several measure classes of interest are given in lemma [ lem - entropy ] .
the second condition is on the kullback
leibler support of @xmath34 , which is related to both the space of discrete measures @xmath21 and the family of likelihood functions @xmath96 .
the kullback leibler neighborhood is defined as @xmath254 the third type of condition is on the hellinger information of the @xmath39 metric , function @xmath255 , a characterization of which is given above .
[ thm - convergence-1 ] fix @xmath213 , and assume that the family of likelihood functions is finitely identifiable .
suppose that for a sequence @xmath256 that tends to a constant ( or 0 ) such that @xmath257 , and a constant @xmath258 , and sets @xmath259 , we have @xmath260 , \\
\label{eqn - support-2s } \pi\bigl(b_k(\varepsilon_n)\bigr ) & \geq&\exp\bigl(-n\varepsilon_n^2c\bigr).\end{aligned}\ ] ] moreover , suppose @xmath261 is a sequence such that @xmath262 & \rightarrow&0.\end{aligned}\ ] ] then , @xmath263 in @xmath264-probability .
the following theorem uses a weaker condition on the covering number , but it contains an additional condition on the likelihood functions which may be useful for handling the case of nonconvex sieves @xmath265 .
[ thm - convergence-2 ] fix @xmath213 .
assume the following : the family of likelihood functions is finitely identifiable and satisfies @xmath266 for any likelihood functions @xmath237 in the family , for some constants @xmath267 .
there is a sequence of sets @xmath259 for which @xmath268 is defined by ( [ eqn - m ] ) .
there is a sequence @xmath269 such that @xmath270 is bounded away from 0 or tending to infinity , and a sequence @xmath261 such that @xmath271\\[-8pt ] & & \qquad { } + \sup_{g_1\in{\mathcal{g}}_n } \log m({\mathcal{g}}_n , g_1,\varepsilon ) \leq n\varepsilon_n^2\qquad \forall\varepsilon\geq\varepsilon_n,\nonumber\\[-12pt]\nonumber\end{aligned}\ ] ] @xmath272 \qquad\forall j\geq m_n,\\[-12pt]\nonumber\end{aligned}\ ] ] @xmath273 & \rightarrow&0.\end{aligned}\ ] ] then , we have that @xmath263 in @xmath264-probability .
\(i ) from the theorem s proof , the above statement continues to hold if conditions ( [ eqn - support-2 ] ) and ( [ eqn - rate - cond ] ) are replaced by the following condition : @xmath274 \rightarrow0.\ ] ] \(ii ) in theorem [ thm - convergence-2 ] and in theorem [ thm - convergence-1 ] augmented with condition ( a ) of theorem [ thm - convergence-2 ] , it is simple to deduce the posterior convergence rates for the mixture density @xmath15 .
we obtain that for a sufficiently large constant @xmath275 , latexmath:[\[\pi\bigl(g\dvtx h(p_{g_0},p_{g } ) \geq m\varepsilon_n @xmath264-probability . before moving to specific examples
, we state a simple lemma which provides estimates of the entropy under the @xmath29 metric for a number of classes of discrete measures of interest . because @xmath29 inherits directly the @xmath16 metric in @xmath3 , the entropy for classes in @xmath277
can typically be bounded in terms of the covering number for subsets of @xmath56 .
[ lem - entropy ] suppose that @xmath3 is bounded .
let @xmath278 .
@xmath279 .
@xmath280 .
let @xmath281 .
assume that @xmath282 and @xmath283 .
then , @xmath284 where the supremum in the right - hand side is taken over all bounded subsets such that @xmath285 .
in this section we derive posterior contraction rates for two classes of mixture models , for example , finite mixtures of multivariate distributions and infinite mixtures based on the dirichlet process .
let @xmath3 be a subset of @xmath40 , @xmath16 be the euclidean metric , and @xmath34 is a prior distribution for discrete measures in @xmath78 , where @xmath286 is known .
suppose that the `` truth '' @xmath287 . to obtain the convergence rate of the posterior distribution of @xmath6 , we need the following : ( a1 ) @xmath3 is compact and the family of likelihood functions @xmath7 is strongly identifiable .
( a2 ) for some positive constant @xmath288 , @xmath289 for any . for any @xmath290 , @xmath291 for some constant @xmath292 .
( a3 ) under prior @xmath34 , for small @xmath165 , @xmath293 and @xmath294 for some constants @xmath295 . ( a4 ) under prior @xmath34 , all @xmath11 are bounded away from 0 , and all pairwise distances @xmath296 are bounded away from 0 .
assumptions ( a1 ) and ( a2 ) hold for the family of gaussian densities with mean parameter @xmath127 . assumption ( a3 ) holds when the prior distribution on the relevant parameters behaves like a uniform distribution , up to a multiplicative constant .
[ thm - finite ] under assumptions , the contraction rate in the @xmath297 wasserstein distance metric of the posterior distribution of @xmath6 is @xmath298 .
let @xmath299 .
combining lemma [ lem - gen - ineq ] with assumption ( a2 ) , if @xmath300 and @xmath301 for @xmath302 , then @xmath303 , for any @xmath304 .
thus , @xmath305 . hence , under prior @xmath34 , @xmath306 in view of assumptions ( a2 ) and ( a3 ) , we have @xmath307 .
conversely , for sufficiently small @xmath57 , if @xmath308 , then by reordering the index of the atoms , we must have @xmath309 and @xmath310 for all @xmath302 [ see the argument in the proof of lemma [ lem - entropy](c ) ] .
this entails that under the prior @xmath34 , @xmath311 let @xmath312 be a sufficiently large multiple of @xmath313 .
we proceed by verifying conditions of theorem [ thm - convergence-2 ] .
let @xmath314 .
then @xmath315 , so equation ( [ eqn - support-1 ] ) trivially holds .
next , we provide upper bounds for @xmath316 , where @xmath317 , and @xmath268 so that ( [ eqn - entropy-2 ] ) is satisfied . indeed , for any @xmath251 , @xmath318 .
by lemma [ lem - entropy](c ) and assumption ( a4 ) , @xmath319 is bounded in terms of @xmath320 , which is bounded above by a constant when @xmath321 s are subsets of @xmath3 whose diameter is bounded by a multiple of @xmath57 . turning to @xmath322 , due to strong identifiability and assumption ( a2 ) , @xmath323 for some constant @xmath324 . by lemma [ lem - entropy](a ) , for some constant @xmath325 ,
thus , equation ( [ eqn - entropy-2 ] ) holds . by proposition [ thm-1](b ) and assumption ( a4 ) , we have @xmath327 for some constant @xmath328 .
to ensure condition ( [ eqn - rate - cond ] ) , note that ( constants @xmath329 change after each bounding step ) @xmath330 & \lesssim&\exp\bigl(2n\varepsilon_n^2\bigr)\sum _ { j\geq m_n } \exp\bigl[-nc(j\varepsilon_n)^{4 } \bigr ] \\ & { \lesssim } & \exp\bigl(2n\varepsilon_n^2 -ncm_n^4 \varepsilon_n^{4}\bigr).\end{aligned}\ ] ] this upper bound goes to zero if @xmath331 , which is satisfied by taking @xmath261 to be a large multiple of @xmath332 .
thus , we need @xmath333 . under the assumptions specified above , @xmath334 on the other hand , for @xmath335 , we have @xmath336 \geq \exp[nc(j\varepsilon_n)^4/16]$ ] which is bounded below by an arbitrarily large constant by choosing @xmath261 to be a large multiple of @xmath332 , thereby ensuring ( [ eqn - support-2 ] ) .
thus , by theorem [ thm - convergence-2 ] , rate of contraction for the posterior distribution of @xmath6 under the @xmath39 distance metric is @xmath337 , which is up to a logarithmic factor the minimax optimal rate @xmath338 as proved for the univariate finite mixtures by @xcite . given the `` true '' discrete mixing measure , @xmath339 , where @xmath3 is a metric space but @xmath340 is unknown . to estimate @xmath28 , the prior distribution @xmath34 on discrete measure @xmath341
is taken to be a dirichlet process @xmath342 that centers at @xmath343 with concentration parameter @xmath344 @xcite . here ,
parameter @xmath343 is a probability measure on @xmath3 . for any @xmath18
, the following lemma provides a lower bound of small ball probabilities of metric space @xmath277 in terms of small ball @xmath343-probabilities of metric space @xmath56 .
[ lem - dpdense ] let @xmath345 , where @xmath343 is a nonatomic base probability measure on a compact set @xmath3 .
for a small @xmath251 , let @xmath346 denote the @xmath57-packing number of @xmath3 under the @xmath16 metric . then , under the dirichlet process distribution , @xmath347 here , @xmath348 denotes the @xmath349 disjoint @xmath350-balls that form a maximal @xmath57-packing of @xmath3 .
the supremum is taken over all such packings .
@xmath351 is the gamma function .
since every point in @xmath3 is of distance at most @xmath57 to one of the centers of @xmath352 , there is a @xmath349-partition @xmath353 of @xmath3 , such that @xmath354 , and @xmath355 for each @xmath356 .
let @xmath357 , @xmath358 , and @xmath359 . from the definition of dirichlet processes , @xmath360 .
to obtain an upper bound for @xmath361 , consider a coupling between @xmath28 and @xmath6 , by associating @xmath362 probability mass of supporting atoms for @xmath28 contained in subset @xmath363 with the same probability mass of supporting atoms for @xmath6 contained in the same subset , for each @xmath364 .
the remaining mass ( of probability @xmath365 ) for both measures are coupled in an arbitrary way .
the expectation under this coupling of the @xmath366 distance provides one such upper bound , that is , @xmath367^{r}.\ ] ] due to the nonatomicity of @xmath343 , for @xmath57 sufficiently small , @xmath368 for all @xmath356 .
let @xmath369 .
then , under @xmath34 , @xmath370 the second inequality in the previous display is due to the fact that @xmath371 .
the third inequality is due to @xmath372 , since @xmath373 and @xmath374 almost surely .
the last inequality is due to the fact that @xmath375 for @xmath376 .
this gives the desired claim .
( b1 ) the nonatomic base measure @xmath343 places full support on a bounded set @xmath180 .
moreover , @xmath343 has a lebesgue density that is bounded away from zero .
( b2 ) for some constants @xmath377 , @xmath378 for any . for any @xmath290 , @xmath379 for some constants @xmath380 . given assumptions and and the smoothness conditions for the likelihood family as specified in theorem [ thm - convolution ]
, there is a sequence @xmath381 such that @xmath382 in @xmath264 probability . specifically : for ordinary smooth likelihood functions , take @xmath383 for any constant @xmath384 . for supersmooth likelihood functions ,
take @xmath385 .
the proof consists of two main steps .
first , we shall prove that under assumptions ( b1)(b2 ) , conditions specified by ( [ eqn - entropy-2s ] ) , ( [ eqn - support-1s ] ) and ( [ eqn - support-2s ] ) in theorem [ thm - convergence-1 ] are satisfied by taking @xmath386 , which is a convex set , and @xmath312 to be a large multiple of @xmath387 .
the second step involves constructing a sequence of @xmath261 and @xmath388 for which theorem [ thm - convergence-1 ] can be applied .
_ step _ 1 : by lemma [ lem - gen - ineq ] and ( b2 ) , @xmath389 .
also , @xmath390 ^ 2 { \lesssim}c_2 d_{\rho ^{m_1}}(g_0,g)^{m_2}$ ] .
assume without loss of generality that @xmath391 ( the other direction is handled similarly ) .
we obtain that @xmath392 for some constant @xmath393 . from ( b1 ) , there is a universal constant @xmath394 such that for any @xmath57 and any @xmath349-partition @xmath395 specified in lemma [ lem - dpdense ] , there holds @xmath396 moreover , the packing number satisfies @xmath397^d$ ] . combining these facts with lemma [ lem - dpdense ] , we have @xmath398 , where the approximation constant is dependent on @xmath399 .
it is simple to check that condition ( [ eqn - support-2s ] ) holds , @xmath400 , by the given rate of @xmath312 , for any constant @xmath258 . since @xmath401 , ( [ eqn - support-1s ] ) trivially holds .
turning to condition ( [ eqn - entropy-2s ] ) , by lemma [ lem - entropy](b ) , we have @xmath402 by the specified rate of @xmath312 .
_ step _ 2 : for any @xmath403 , let @xmath404 be the inverse of the hellinger information function of the @xmath405 metric .
specifically , for any @xmath406 , @xmath407 note that @xmath408 .
@xmath409 is nondecreasing because @xmath410 is .
let @xmath256 be the sequence determined in the previous step of the proof .
let @xmath411 , and @xmath412 .
condition ( [ eqn - rate - conds-2s ] ) holds by definition of @xmath413 , that is , @xmath414 . to verify ( [ eqn - hellinger - sum ] ) , note that the running sum with respect to @xmath415 can not have more than @xmath416 terms , and , due to the monotonicity of @xmath417
, we have @xmath418 \\ & & \qquad\leq{\operatorname{diam}}(\theta)/\varepsilon_n \exp\bigl(2n\varepsilon_n^2 - n\psi_{{\mathcal{g}}_n}(m_n\varepsilon_n)/8\bigr ) \rightarrow0.\end{aligned}\ ] ] hence , theorem [ thm - convergence-1 ] can be applied to conclude that @xmath419 in @xmath264-probability . under the ordinary smoothness condition ( as specified in theorem [ thm - convolution ] ) , @xmath420 , where @xmath421 is an arbitrarily positive constant .
so , @xmath422 on the other hand , under the supersmoothness condition , @xmath423 .
so , @xmath424 .
proof of theorem [ thm - bound - supnorm ] suppose that equation ( [ eqn - bound - supnorm ] ) is not true , then there will be sequences of @xmath425 and @xmath426 tending to @xmath28 in the @xmath39 metric , and that .
we write @xmath427 , where @xmath428 for indices @xmath429 greater than @xmath430 , the number of atoms of @xmath27 .
similar notation is applied to @xmath431 . since both @xmath27 and @xmath431 have a finite number of atoms , there is @xmath432 so that @xmath433 .
let @xmath434 .
this set exists because there are pairs of atoms @xmath435 such that @xmath436 is bounded away from zero in the limit . since @xmath432
, we can express @xmath437 and , by taylor s expansion , @xmath438 where @xmath439 due to ( [ eqn - lipschitz ] ) and the definition of @xmath440 .
so @xmath441 .
the quantities @xmath442 and @xmath443 are linear combinations of elements of @xmath96 , @xmath444 and @xmath161 for different @xmath127 s , respectively . since @xmath3 is compact , subsequences of @xmath27 and @xmath431 can be chosen so that each of their support points converges to a fixed atom @xmath445 , for @xmath446 . after being rescaled
, the limits of @xmath447 and @xmath448 are still linear combinations with constant coefficients not depending on @xmath151 .
we shall now argue that not all such coefficients vanish to zero .
suppose this is not the case .
it follows that for the coefficients of @xmath448 we have @xmath449 this implies that @xmath450 .
since @xmath3 is bounded , there exists a pair @xmath451 such that @xmath452 does not vanish to zero .
but then , one of the coefficients of @xmath453 does not vanish to zero , which contradicts the hypothesis .
next , we observe that some of the coefficients of @xmath454 and @xmath448 may tend to infinity . for each @xmath35 ,
let @xmath455 be the inverse of the maximum coefficient of @xmath453 , @xmath456 , and @xmath448 . from the conclusion in the preceding paragraph
, @xmath457 is uniformly bounded from above by a constant for all @xmath35 .
moreover , @xmath458 converges to @xmath459 and @xmath460 converges to @xmath461 , and @xmath462 converges to @xmath463 , for some finite @xmath464 and @xmath465 , not all of them vanishing ( in fact , at least one of them is 1 ) .
we have @xmath466\\[-8pt ] & & \qquad\rightarrow\biggl|\sum_{j=1}^{k^*}\alpha_{j } f \bigl(x|\theta^*_{j}\bigr ) + \beta_j^t df
\bigl(x|\theta^*_j\bigr ) + \gamma_j^t d^{2}f\bigl(x|\theta^*_j\bigr ) \gamma_j \biggr| \nonumber\end{aligned}\ ] ] for all @xmath151 .
this entails that the right - hand side of the preceding display must be 0 for almost all @xmath151 . by strong identifiability
, all coefficients must be 0 , which leads to contradiction . with respect to @xmath467 ,
suppose that the claim is not true , which implies the existence of a subsequence @xmath468 such that that @xmath469 .
going through the same argument as above , we have @xmath470 , not all of which are zero , such that equation ( [ eqn - conv - pointwise ] ) holds .
an application of fatou s lemma yields latexmath:[$\int|\sum_{j=1}^{k^*}\alpha_{j } f(x|\theta_{j } ) + \beta_j^t df(x|\theta_j ) + \gamma_j^t d^{2}f(x|\theta_j ) \gamma_j
@xmath151 , leading to contradiction .
proof of theorem [ thm - convolution ] to obtain an upper bound of @xmath472 in terms of @xmath473 under the condition that @xmath204 , our strategy is approximate @xmath6 and @xmath124 by convolving these with some mollifier @xmath474 . by the triangular inequality , @xmath475 .
the first two terms are simple to bound , while the last term can be handled by expressing @xmath476 as the convolution of the mixture density @xmath15 with another function .
we also need the following elementary lemma ( whose proof is given section [ sec - aux ] ) . [ lem - ineq - moment ] assume that @xmath477 and @xmath478 are two probability density functions on @xmath40 with bounded @xmath479-moments . for @xmath480 such that @xmath481 , @xmath482 let @xmath483 denote the volume of the @xmath198-dimensional unit sphere
. then , @xmath484 take any @xmath485 , and let @xmath486 be a symmetric density function on @xmath40 whose fourier transform @xmath487 is a continuous function whose support is bounded in @xmath488^d$ ] .
moreover , @xmath489 has bounded moments up to order @xmath490 .
consider mollifiers @xmath491 for @xmath165 .
let @xmath492 and @xmath178 be the fourier transforms for @xmath474 and @xmath0 , respectively .
define @xmath493 to be the inverse fourier transform of @xmath494 : @xmath495 note that function @xmath496 has bounded support .
so , @xmath497 , and @xmath498 is the fourier transform of @xmath493 . by the convolution theorem , @xmath499 . as a result ,
@xmath500 from the definition of @xmath474 , the second moment under @xmath474 is @xmath501 . consider a coupling @xmath6 and @xmath476 under which we have a pair of random variables @xmath502 where @xmath57 is independent of @xmath127 , the marginal distributions of @xmath127 and @xmath57 are @xmath6 @xmath474 , respectively . under this coupling , @xmath503 , which entails that @xmath504 . by the triangular inequality , @xmath505 ,
so for some constant @xmath506 dependent only on kernel @xmath489 , @xmath507 theorem 6.15 of @xcite provides an upper bound for the wasserstein distance : for any two probability measures @xmath8 and @xmath508 , @xmath509 , where @xmath510 is the total variation of measure @xmath511 .
thus , @xmath512 we note that since density function @xmath489 has a bounded @xmath490th moment , @xmath513 \\ & = & 2^{s } \biggl[\int\|\theta\|^s \,dg(\theta ) + \delta^s \int\|x\|^s k(x ) \,dx \biggr ] < \infty,\end{aligned}\ ] ] because @xmath6 s support points lie in a bounded subset of @xmath40 . applying lemma [ lem - ineq - moment ] to ( [ eqn - bound - wasser - variational ] )
, we obtain that for @xmath514 , @xmath515\\[-8pt ] & \leq & c(d , k , s ) \bigl\|g*k_\delta - g'*k_\delta \bigr\|_{l_2}^{2(s-2)/(d+2s)}.\nonumber\end{aligned}\ ] ] here , constants @xmath516 are different in each line , and they are dependent only on @xmath517 and the @xmath490th moment of density function @xmath489 .
next , we use a known fact that for an arbitrary ( signed ) measure @xmath8 on @xmath40 and function @xmath518 , there holds @xmath519 , where @xmath520 denotes the total variation of @xmath8 : @xmath521 by plancherel s identity , @xmath522^d } { \tilde{f}}(\omega ) ^{-2}\,d\omega.\end{aligned}\ ] ] the last bound holds because @xmath523 has support in @xmath488^d$ ] and is bounded by a constant . collecting equations ( [ eqn - perturb ] ) , ( [ eqn - bound - wasser - variational ] ) , ( [ eqn - bound - l1 ] ) and ( [ eqn - bound - l2 ] ) and the preceding display , we have @xmath524^d } { \tilde{f}}(\omega)^{-2}\,d\omega \biggr]^{({s-2})/({d+2s } ) } \biggr\}.\end{aligned}\ ] ] if @xmath525 as @xmath186 @xmath526 for some positive constant @xmath188 , then @xmath527 the exponent tends to @xmath528 as @xmath529 , so we obtain that @xmath530 , for any constant @xmath531 , as @xmath532 .
if @xmath533 as @xmath186 @xmath187 for some positive constants @xmath534 , then @xmath535 taking @xmath536 , we obtain that @xmath537 proof of lemma [ lem - gen - ineq ] we exploit the variational characterization of @xmath0-divergences ( e.g. , @xcite ) , @xmath538 where the infimum is taken over all measurable functions defined on @xmath9 .
@xmath539 denotes the legendre
fenchel conjugate dual of convex function @xmath109 [ @xmath539 is again a convex function on @xmath133 and is defined by @xmath540 . by the variational characterization , @xmath541 is a convex functional ( jointly of its two arguments ) .
thus , for any coupling @xmath542 of two mixing measures @xmath6 and @xmath124 , @xmath543 , where the inequality is obtained via jensen s inequality . since this holds for any @xmath542 , the desired bound follows .
proof of proposition [ thm-1 ] ( a ) suppose that the claim is not true , and there is a sequence of @xmath544 such that @xmath545 always holds and that converges in @xmath39 metric to @xmath546 and @xmath547 , respectively .
this is due to the compactness of both @xmath78 and @xmath215 .
we must have @xmath548 , so @xmath549 . at the same time , @xmath550 , which implies that @xmath551 for almost all @xmath144 . by the finite identifiability condition , @xmath552 , which is a contradiction .
\(b ) is an immediate consequence of theorem [ thm - bound - supnorm ] , by noting that under the given hypothesis , there is @xmath553 depending on @xmath14 , such that @xmath554 for sufficiently small @xmath555 .
the boundedness of @xmath3 implies the boundedness of @xmath555 , thereby extending the claim for the entire admissible range of @xmath555 .
( c ) is obtained in a similar way to theorem [ thm - convolution ] .
we outline in this section the proofs of theorems [ thm - convergence-1 ] and [ thm - convergence-2 ] .
our proof follows the same steps as in @xcite , with suitable modifications for the inclusion of the hellinger information function and the conditions involving latent mixing measures .
the proof consists of results on the existence of tests , which are turned into probability bounds on the posterior contraction .
proof of lemma [ lem - convexball ] we first consider case ( i ) .
define @xmath556 . since @xmath16 is a metric in @xmath3 , it is a standard fact of wasserstein metrics that @xmath557 is a convex set . since @xmath215 is also convex , so is the set @xmath558 .
it follows that @xmath559 is a convex set of mixture distributions .
now , applying a result from birg @xcite and le cam ( @xcite , lemma 4 , page 478 ) , there exist tests @xmath228 that discriminate between @xmath264 and convex set @xmath559 such that @xmath560 , \\ \label{eqn - power-2 } \sup_{g\in{\mathcal{g}}\cap b_w(g_1,r/2 ) } p_{g}(1-\varphi_n ) & \leq & \exp \bigl[-n \inf{h}^2(p_{g_0},p_1)/2\bigr],\end{aligned}\ ] ] where the exponent in the upper bounds are given by the infimum hellinger distance among _ all _ @xmath561 . due to the triangle inequality , if @xmath562 and @xmath563 , then @xmath564 .
so , @xmath565 this completes the proof of case ( i ) . turning to case ( ii ) , for a constant @xmath566 to be determined ,
consider a maximal @xmath567-packing in the @xmath39 metric in set @xmath558 .
this yields a set of @xmath568 points @xmath569 in @xmath570 .
[ in the following we drop the subscripts of @xmath571 . ]
we note the following fact : for any @xmath572 , if @xmath573 and @xmath574 , by lemma [ lem - gen - ineq ] we have @xmath575 ( the second inequality is due to the condition on the likelihood functions ) ; and so it follows that @xmath576 choose @xmath577 so that the previous bounds become @xmath578 and @xmath579 . for each pair of @xmath580 ,
there exist tests @xmath581 of @xmath37 versus the closed hellinger ball @xmath582 such that @xmath583,\\ \sup_{g\in{\bar{\mathcal{g}}}(\theta ) : { h}(p_g , p_{\tilde{g}_t } ) \leq{h}(p_{g_0},p_{\tilde{g}_t})/2 } p_{g}\bigl(1-\omega_n^{(t ) } \bigr ) & \leq & \exp\bigl[-n { h}^2(p_{g_0},p_{\tilde{g}_t})/8 \bigr].\end{aligned}\ ] ] consider the test @xmath584 , then @xmath585 , \\
\sup_{g\in{\mathcal{g}}\cap b_w(g_1,r/2 ) } p_{g}(1-\varphi_n ) & \leq & \exp \bigl[-n \psi_{\mathcal{g}}(r)/8\bigr].\end{aligned}\ ] ] the first inequality is due to @xmath586 , and the second is due to the fact that for any @xmath573 there is some @xmath587 such that @xmath588 , so that @xmath589 .
proof of lemma [ lem - ballcomplement ] for a given @xmath590 choose a maximal @xmath591-packing for set @xmath592 .
this yields a set @xmath593 of at most @xmath594 points .
moreover , every @xmath595 is within distance @xmath591 of at least one of the points in @xmath593 .
for every such point @xmath596 , there exists a test @xmath597 satisfying equations ( [ eqn - type-1 ] ) and ( [ eqn - type-2 ] ) . take @xmath228 to be the maximum of all tests attached this way to some point @xmath596 for some @xmath598 .
then , by the union bound and the fact that @xmath246 is nonincreasing , @xmath599 \\ & \leq & d(\varepsilon)\sum_{t\geq t_0}\exp\bigl[-n \psi_{{\mathcal{g}}}(t\varepsilon)/8\bigr ] , \\ \sup_{g\in\bigcup_{u\geq t_0}s_{u } } p_{g}(1-\varphi_n ) & \leq & \sup_{u\geq t_0 } \exp\bigl[-n\psi_{{\mathcal{g}}}(u\varepsilon)/8\bigr ]
\leq \exp \bigl[-n\psi_{{\mathcal{g}}}(t_0\varepsilon)/8\bigr],\end{aligned}\ ] ] where the last inequality is due the monotonicity of @xmath410 . proof of theorems [ thm - convergence-1 ] and [ thm - convergence-2 ]
the proof for theorem [ thm - convergence-1 ] proceeds in a similar way to theorem 2.1 of @xcite , while the proof for theorem [ thm - convergence-2 ] is similar to their theorem 2.4 .
the main difference is that the posterior distribution statements are made with respect to mixing measure @xmath6 rather than mixture density @xmath15 . by a result of ghosal et al .
@xcite ( lemma 8.1 , page 524 ) , for every @xmath251 and probability measure @xmath34 on the set @xmath600 defined by ( [ eqn - kl - neighborhood ] ) , we have , for every @xmath258 , @xmath601 this entails that , for a fixed @xmath602 , there is an event @xmath603 with @xmath264-probability at least @xmath604 , for which there holds @xmath605 let @xmath606 , @xmath607 for each @xmath608 .
the conditions specified by lemma [ lem - ballcomplement ] are satisfied by setting @xmath609 ( constant in @xmath57 ) .
thus , there exist tests @xmath228 for which equations ( [ eqn - type-1b ] ) and ( [ eqn - type-2b ] ) hold . then , @xmath610 + p_{g_0 } \bigl[(1-\varphi_n)\pi(g\in{\mathcal{o}}_n|x_1,\ldots , x_n)\bigr ] \\
\hspace*{-3pt}&&\qquad \leq p_{g_0}\bigl[\varphi_n \pi(g\in { \mathcal{o}}_n|x_1,\ldots , x_n)\bigr ] + p_{g_0}{\mathbb{i}}\bigl(a_n^c\bigr ) \\ \hspace*{-3pt}&&\qquad\quad { } + p_{g_0 } \bigl[(1-\varphi_n)\pi(g\in { \mathcal{o}}_n|x_1,\ldots , x_n ) { \mathbb{i}}(a_n)\bigr].\end{aligned}\ ] ] due to lemma [ lem - ballcomplement ] , the first term in the preceding display is bounded above by @xmath611 \rightarrow0 $ ] , thanks to ( [ eqn - rate - cond ] ) .
the second term in the above display is bounded by @xmath612 by the definition of @xmath603 .
if @xmath257 , let @xmath613 .
if @xmath270 tends to a positive constant away from 0 , we let @xmath164 be arbitrarily large so that this probability in the second term vanishes to 0 .
to show that the third term in the display also vanishes as @xmath614 , we exploit the following expression : @xmath615 and then obtain a lower bound for the denominator by ( [ eqn - denominator ] ) . for the nominator , by fubini s theorem , @xmath616,\nonumber\end{aligned}\ ] ] where the last inequality is due to ( [ eqn - type-2b ] ) .
in addition , by ( [ eqn - support-1 ] ) , @xmath617 combining bounds ( [ eqn - t1 ] ) and ( [ eqn - t2 ] ) and condition ( [ eqn - support-2 ] ) , we obtain @xmath618}{\exp ( -2n\varepsilon_n^2)\pi(b_k(\varepsilon_n ) ) } \\ & & \qquad \leq o(1 ) + \exp\bigl(2n\varepsilon_n^2\bigr)\sum _ { j\geq m_n } \exp\bigl[-n\psi_{{\mathcal{g}}_n}(j \varepsilon_n)/16 \bigr].\end{aligned}\ ] ] the upper bound in the preceding display converges to 0 by ( [ eqn - rate - cond ] ) , thereby concluding the proof of theorem [ thm - convergence-2 ] .
the proof of theorem [ thm - convergence-1 ] proceeds similarly .
proof of lemma [ lem - entropy ] to simplify notation , we give a proof for @xmath619 . the general case for @xmath620
can be carried out in the same way .
\(a ) suppose that @xmath621 forms an @xmath57-covering for @xmath3 under metric @xmath16 , where @xmath622 denotes the ( minimum ) covering number .
take any discrete measure @xmath623 .
for each @xmath2 there is an approximating @xmath624 among the @xmath625 s such that @xmath626 .
let @xmath627 be a @xmath14-dim vector in the probability simplex that deviates from @xmath75 by less than @xmath421 in @xmath628 distance : @xmath629 .
. we shall argue that @xmath631 [ to see this , consider the following coupling between @xmath6 and @xmath124 : associating @xmath632 probability mass of @xmath2 ( from @xmath6 ) with the same probability mass of @xmath624 ( from @xmath124 ) , while the remaining mass from @xmath6 and @xmath124 ( of probability @xmath633 ) are coupled in an arbitrary way .
the right - hand side of the previous display is an upper bound of the expectation of the @xmath16 distance between two random variables distributed according to the described coupling . ]
it follows that a @xmath634-covering for @xmath78 can be constructed by combining each element of a @xmath421-covering in the @xmath628 metric of the @xmath5-probability simplex and @xmath14 @xmath635-coverings of @xmath3 .
now , the covering number of the @xmath5-probability simplex is less than the number of cubes of length @xmath636 covering @xmath637^{k}$ ] times the volume of @xmath638 , that is , @xmath639 .
it follows that @xmath640 .
take @xmath369 toachieve the claim .
\(b ) as before , let @xmath621 be an @xmath635-covering of @xmath3 . take any @xmath641 , where @xmath14 may be infinity .
the collection of atoms @xmath146 can be subdivided into disjoint subsets @xmath642 , some of which may be empty , so that for each @xmath643 , @xmath644 for any @xmath645 .
define @xmath646 , and let @xmath647 , then we are guaranteed that @xmath648 by using a similar coupling argument as in part ( a ) .
let @xmath649 be a @xmath650-dim vector in the probability simplex that deviates from @xmath87 by less than @xmath421 in the @xmath628 distance : @xmath651 .
take @xmath652 .
it is simple to observe that @xmath653 . by the triangle inequality , @xmath654 the foregoing arguments
establish that an @xmath634-covering in the wasserstein metric for @xmath21 can be constructed by combining each element of the @xmath421-covering in @xmath628 of the @xmath655 simplex and a single covering of @xmath3 . from the proof of part ( a ) , @xmath656 .
take @xmath657 to conclude .
\(c ) consider a @xmath1 such that @xmath658 . by definition ,
there is a coupling @xmath659 so that @xmath660 . since @xmath661
, this implies that @xmath662 .
thus , for each @xmath150 there is a @xmath415 such that @xmath663 . without loss of generality , assume that @xmath664 for all @xmath150 . for sufficiently small @xmath57 , for any @xmath429
, it is simple to observe that @xmath665 .
thus , @xmath666 .
now , an @xmath667 covering in @xmath94 for @xmath668 can be constructed by combining the @xmath669-covering for each of the @xmath14 sets @xmath670 and the @xmath636-covering for each of the @xmath14 sets @xmath671 $ ] .
this entails that @xmath672^k ( 8\varepsilon k / m\delta)^k$ ] .
take @xmath673 to conclude the proof .
proof of lemma [ lem - ineq - moment ] ( a ) for arbitrary constant @xmath674 , we have @xmath675 , choosing @xmath676^{1/{s}}\ ] ] to conclude .
\(b ) for any @xmath674 , we have @xmath677^{1/2 } \\ & \leq & v_d^{1/2}r^{d/2}\bigl\|p - p'\bigr\|_{l_2}.\end{aligned}\ ] ] we also have @xmath678 thus , @xmath679 which gives the desired bound .
the author wishes to thank arash amini , the associate editor and , in particular , an anonymous reviewer for valuable comments and suggestions .
the shortened proof of lemma [ lem - gen - ineq ] was pointed out by the reviewer , who also helped to spot a gap in a previous version of lemma [ lem - convexball ] .
|
this paper studies convergence behavior of latent mixing measures that arise in finite and infinite mixture models , using transportation distances ( i.e. , wasserstein metrics ) .
the relationship between wasserstein distances on the space of mixing measures and @xmath0-divergence functionals such as hellinger and kullback leibler distances on the space of mixture distributions is investigated in detail using various identifiability conditions .
convergence in wasserstein metrics for discrete measures implies convergence of individual atoms that provide support for the measures , thereby providing a natural interpretation of convergence of clusters in clustering applications where mixture models are typically employed .
convergence rates of posterior distributions for latent mixing measures are established , for both finite mixtures of multivariate distributions and infinite mixtures based on the dirichlet process .
|
the search for new physics at the lhc has covered enormous ground so far , constraining the parameters of many possible new particles that appear in a variety of extensions of the standard model ( sm ) .
there are still exceptions , models that have not been carefully confronted with the lhc data , amongst them an extension of the scalar sector of the sm with a color octet electroweak doublet scalar .
this model was introduced some time ago by manohar and wise ( mw ) @xcite , and it is motivated by minimal flavor violation @xcite .
there are several phenomenological papers concerning lhc studies of the mw model available , but all of them predate the lhc .
the original mw paper already gave analytic expressions for the pair production of the new scalars through the dominant gluon fusion channel , finding production cross - sections that vary many orders of magnitude , between @xmath3 fb and @xmath4 fb for @xmath5 between @xmath6 and @xmath7 tev respectively at lhc14 @xcite .
shortly after , gresham and wise considered the production of a single neutral scalar , a one loop process similar to the production of the higgs boson , which becomes dominant over the two scalar production at around 1 tev @xcite .
et al _ @xcite considered pair production of scalars leading to final states with four heavy quarks ( top or bottom ) at the lhc concluding that discovery would be possible for masses up to about a tev .
finally arnold and fornal considered constraints that can be imposed by studying high @xmath8 four - jet events at lhc @xcite , and concluded that 10 fb@xmath9 would be enough to discover a scalar as heavy as 1.5 tev at lhc14 . in this paper
we partially address the existing gap in the extraction of lhc constraints for the mw model by comparing the cross section for production of a single neutral scalar and a pair of neutral scalars to limits obtained by atlas and cms in the dijet , top - pair , dijet - pair , @xmath0 and @xmath1 channels .
we defer the study of charged scalar production and their different decay modes to a future publication .
we restrict this paper to a parton - level study because its main weakness , not knowing the relevant acceptance of lhc detectors , can only be addressed by the experimental collaborations .
for this reason , the constraints we obtain will be stronger than is actually possible .
a similar study for pseudo - scalar color octets , common to composite higgs models , exists in the literature @xcite .
the mw model contains many new parameters , several of which have been studied phenomenologically before . in particular the new color octet scalars can have a large effect on loop level higgs production and decay @xcite and this has resulted in multiple studies of one - loop effective higgs couplings @xcite .
they are also constrained by precision electroweak measurements @xcite , flavor physics @xcite , unitarity and vacuum stability @xcite and other lhc processes @xcite . in the mw model , the new field
@xmath10 transforms as @xmath11 under the sm gauge group @xmath12 and this gives rise to the gauge interactions responsible for its pair production at the lhc .
numerous new couplings appear in the self - interactions of the scalars as well as in the yukawa couplings to fermions .
the latter are restricted by minimal flavor violation to two complex numbers @xcite , @xmath13 where @xmath14 are left - handed quark doublets , @xmath15 and the @xmath16 generators are normalized as @xmath17 .
the matrices @xmath18 are the same as the higgs couplings to quarks , and @xmath19 along with their phases @xmath20 , are new overall factors .
non - zero phases signal @xmath21 violation beyond the sm and contribute for example to the edm and cedm of quarks @xcite .
the most general renormalizable scalar potential is given in ref .
@xcite and contains several terms . of these
, our study will only depend on the following ( with @xmath22 gev the usual higgs vacuum expectation value ) @xmath23 the number of parameters in eq.([potential ] ) can be reduced for our study as follows : first @xmath24 can be chosen to be real by a suitable definition of @xmath10 ; then custodial @xmath25 symmetry can be invoked to introduce the relations @xmath26 ( and hence @xmath27 ) @xcite and @xmath28 @xcite ; and finally , requiring cp conservation removes all the phases , @xmath29 , @xmath30 , and @xmath31 .
after symmetry breaking , the non - zero vev of the higgs field in eq.([potential ] ) gives the physical @xmath32 its usual mass @xmath33 , and in addition it splits the octet scalar masses as , @xmath34 these relations , combined with the use of custodial and cp symmetries result in the following independent input parameters : @xmath35 the parameter @xmath36 controls the split between the two neutral resonances @xmath37 and the parameter @xmath38 controls the strength of scalar loop contributions to single neutral scalar production through gluon fusion .
the parameters @xmath19 control respectively the strength of the @xmath39 and @xmath40 interactions . under the above assumptions
, the effective one - loop coupling @xmath41 can be written in terms of two factors @xmath42 as , @xmath43 where @xmath44 is the gluon field strength tensor and @xmath45 .
at one - loop level these factors receive their main contributions from top - quark and scalar loops and are given by @xcite @xmath46\;,\nonumber\\ f_i & = & ( \sqrt{2}g_f)^{1/2 } \frac{\alpha_s}{8\pi}\frac{1}{2 } \left[\frac{m^2_t } { m^2_i } \eta_u \ , f\left(\frac{m^2_t}{m_i^2}\right)+\frac{m^2_b } { m^2_i } \eta_d \ , f\left(\frac{m^2_b}{m_i^2}\right)\right ] .
\label{loopc}\end{aligned}\ ] ] we have already simplified eq .
[ loopc ] with the relations @xmath47 , @xmath48 , and we have allowed for the mass @xmath49 , to be different from @xmath50 through @xmath36 as discussed above . throughout the calculation , the scalars @xmath51 and @xmath52 ( when not in a loop ) , are taken to be on - shell , in keeping with the narrow width approximation .
the loop functions @xmath53 and @xmath54 are well known and given by : @xmath55 from this it follows that the top - quark contribution starts as large as @xmath56 for a resonance with twice the top - quark mass , and goes to zero as @xmath57 . additionally , the factor @xmath58 , generates some suppression in contributions from scalar loops relative to top - quark loops . for this reason
only relatively large values of @xmath38 make a significant contribution to single scalar production .
we have also included the bottom - quark loop , which is important only for regions of parameter space where @xmath59 .
this situation is completely analogous to higgs production in type ii two higgs doublet models where the @xmath60 quark contribution is important for @xmath61 .
we will consider this case for the dijet final state study .
an early paper constraining new particles at the lhc introduced several benchmarks @xcite , including a color octet scalar @xmath10 , that have since been pursued by both atlas @xcite and cms @xcite .
both these studies appear to rule out the new color octet scalars with masses up to 3 tev .
unfortunately , the original benchmark does not cover interesting possibilities for additional color octet scalars like the mw model .
this can be easily seen from their coupling to two gluons , @xmath62 , which has been used with @xmath63 in the cms study . comparing this to @xmath64 in eq.([loopc ] ) , one sees that realistic couplings in the mw model result in cross - sections that are three orders of magnitude below the benchmark . to proceed with our numerical study
we implement the lagrangian of eq.([yukc ] ) and eq.([loopc ] ) in feynrules @xcite to generate a universal feynrules output ( ufo ) file , and then feed this ufo file into mg5_amc@nlo @xcite . to extract constraints on the parameters of the new resonances we will compare our predicted cross - sections times branching ratios with existing lhc studies .
we divide our study according to the different channels that dominate different regions of parameter space .
we will use in particular the following channels : two jets ; top pairs ; four jets ; four top - quarks ; and four bottom quarks .
we illustrate all cases with production and decay of @xmath51 or @xmath65 but the limits are identical for @xmath66 and similar for @xmath52 provided we prevent the decays @xmath67 ( which is done by a choice of @xmath36 ) .
if those decays are allowed , different limits apply from channels involving final state leptons which we do not consider in this paper . in the limit of @xmath21 conservation that we consider here
, the @xmath52 is a pseudo - scalar and therefore its gluon fusion production mechanism is a bit different from @xmath51 , being independent of @xmath38 .
this channel has been studied before in the context of the mw model for the tevatron . a comparison with the dijet cross - section at the tevatron revealed that signatures from the mw model are orders of magnitude below the qcd background and no constraint emerged for masses below 350 gev with @xmath68 as large as 10 @xcite .
existing lhc searches that can be used to compare the mw model with the two jets channel are : atlas-8 tev @xcite , atlas-13 tev @xcite , cms-8 tev @xcite , cms-13 tev @xcite . for example , atlas used its 8 tev data to conclude that color octet scalars are ruled out below 2.8 tev @xcite , but as explained above , their benchmark color octet model , that of ref .
@xcite , results in cross - sections many orders of magnitude larger than the mw model .
the dijet decay mode is dominant when one of the two neutral color octet resonances @xmath51 or @xmath52 is kinematically suppressed from decaying to the charged one through processes of the form @xmath67 , and its decay into top quark pairs is suppressed due to a small parameter @xmath69 ( we will assume the mass is above threshold for top - pair production ) . under these conditions the dominant decay rate is into two jets .
we illustrate this scenario with the choices : @xmath70 , which makes @xmath71 ; and @xmath72 , which places the mass , @xmath73 above @xmath74 .
the decay of @xmath51 in this case is almost 100% into two jets , with the dominant channels being @xmath75 and @xmath76 as illustrated in figure [ br - case1 ] . with the parameters @xmath70 , @xmath72 .
the solid lines are for @xmath77 gev and the dashed lines lines are for @xmath78 gev .
the higher ( red ) curves are for @xmath79 and the lower ( green ) curves for @xmath80 .
the panel on the left is for the case @xmath81 whereas the panel on the right is for @xmath82 .
[ br - case1],title="fig : " ] with the parameters @xmath70 , @xmath72 .
the solid lines are for @xmath77 gev and the dashed lines lines are for @xmath78 gev .
the higher ( red ) curves are for @xmath79 and the lower ( green ) curves for @xmath80 .
the panel on the left is for the case @xmath81 whereas the panel on the right is for @xmath82 .
[ br - case1],title="fig : " ] the left panel on figure [ br - case1 ] corresponds to @xmath81 where the resonance couples most strongly to @xmath75 pairs .
this coupling is weak enough that the resonance width does not exceed @xmath83 , a very narrow resonance which we compare to the experimental curve that assumes a width set at the detector resolution .
this figure illustrates the dependence on @xmath38 , showing it is almost symmetric but with a slight enhancement ( from constructive interference between quark and scalar loops ) for negative values of @xmath38 .
for this reason we will show negative values of @xmath38 in what follows , but remind the reader that results are almost identical for the corresponding positive values of @xmath38 .
the panel on the right in figure [ br - case1 ] explores the dependence on @xmath2 for a fixed value @xmath82 .
we see how the dominant decay mode turns from @xmath76 to @xmath75 for values of @xmath2 in the @xmath84 range . in this case
@xmath85 can reach the percent level for @xmath86 . the most restrictive data available for this channel comes from 13 tev and we use it to extract bounds in figure [ fig:2jets ] .
the figure on the upper left panel , covers a resonance mass up to 1.6 tev , and is best explored by cms .
we have fixed @xmath81 and present results for different values of @xmath38 . to compare with the cms limits we show their results for resonances produced by ( light ) @xmath87 and @xmath76 with 12.9 fb@xmath9 at 13 tev @xcite .
figure [ br - case1 ] suggests that this type of resonance would be dominantly produced by a @xmath75 initial state so that it does nt match either of the two cases explicitly considered by cms .
for example , with @xmath88 and @xmath89gev , production from @xmath90 accounts for 98.7% of the cross - section , with only about 1.1% originating in gluon fusion . for @xmath91 ,
the contribution from gluon fusion has increased to about 10% . from this comparison
we conclude that for values of @xmath92 , its perturbative unitarity bound @xcite , the cross - sections are too small for the lhc to place any meaningful limit .
the figure also shows that the resonance mass is constrained to be larger than about @xmath93 gev for larger values of @xmath38 , when we ignore the perturbative unitarity constraint .
in the upper right panel we repeat the exercise using the results available for a higher range of resonance mass .
we show two experimental curves , first the atlas result with 15.7 fb@xmath9 @xcite for a hypothetical resonance with a gaussian distribution and a width set at the detector resolution ( 2 - 3% of the mass ) .
the second curve is the cms result with 2.4 fb@xmath9 also at 13 tev for a quark - quark type resonance @xcite .
we see that the high energy region does not yet constrain this set of parameters . on the bottom panels we repeat the comparisons , but this time we fix @xmath94 ( somewhat over its perturbative unitarity limit ) and vary @xmath2 .
the lower energy region shown on the left panel indicates that resonances with @xmath95 gev are excluded for values of @xmath96 .
pushing @xmath38 to @xmath97 ( much higher than its perturbative bound ) pushes the lower bound on the octet mass to around 1.3 tev . ) and their decays into 2 jets in the low mass region with cms 13tev preliminary results @xcite .
the panels on the right correspond to the high mass region and they are compared also with the atlas 13tev preliminary results @xcite.,title="fig : " ] ) and their decays into 2 jets in the low mass region with cms 13tev preliminary results @xcite .
the panels on the right correspond to the high mass region and they are compared also with the atlas 13tev preliminary results @xcite.,title="fig : " ] ) and their decays into 2 jets in the low mass region with cms 13tev preliminary results @xcite .
the panels on the right correspond to the high mass region and they are compared also with the atlas 13tev preliminary results @xcite.,title="fig : " ] ) and their decays into 2 jets in the low mass region with cms 13tev preliminary results @xcite .
the panels on the right correspond to the high mass region and they are compared also with the atlas 13tev preliminary results @xcite.,title="fig : " ] a comparison with tevatron cross - sections in the mass range @xmath98 gev found that this model could have a large enough cross - section to be relevant for @xmath99 which is well beyond the perturbative unitarity bound @xcite .
in this section we use lhc data from atlas with 20.3 fb@xmath9 at 8 tev @xcite and from cms with 2.6 fb@xmath9 at 13 tev @xcite to obtain limits from top - pair production .
once the resonance mass is above the top - pair threshold , its branching ratio into @xmath100 will be completely dominant if we suppress decays into @xmath101 or @xmath102 by choosing @xmath72 as in the two - jet case and if @xmath69 is of order one .
we perform two comparisons with data , in the first one we set @xmath103 and explore the sensitivity to @xmath38 and in the second one we set @xmath104 and explore the sensitivity to @xmath69 .
the @xmath105 is at least 95% for all of the parameters studied ( with the second most dominant decay mode being two gluons ) , and the width of the resonance is close to 1% of its mass for @xmath103 but grows to about 70% by the time @xmath106 so we do not consider any larger values . the value of @xmath2 is set to 0 for illustration but it is irrelevant for this channel . ) and its decays into 2 tops at 8tev compared with figure 11-d of atlas preliminary results @xcite ( red curve ) .
the right figure is the same with the left but at 13tev and it is compared with figure 8 of cms preliminary results @xcite ( black curves).,title="fig : " ] ) and its decays into 2 tops at 8tev compared with figure 11-d of atlas preliminary results @xcite ( red curve ) .
the right figure is the same with the left but at 13tev and it is compared with figure 8 of cms preliminary results @xcite ( black curves).,title="fig : " ] ) and its decays into 2 tops at 8tev compared with figure 11-d of atlas preliminary results @xcite ( red curve ) .
the right figure is the same with the left but at 13tev and it is compared with figure 8 of cms preliminary results @xcite ( black curves).,title="fig : " ] ) and its decays into 2 tops at 8tev compared with figure 11-d of atlas preliminary results @xcite ( red curve ) .
the right figure is the same with the left but at 13tev and it is compared with figure 8 of cms preliminary results @xcite ( black curves).,title="fig : " ] in the left panel we compare to the atlas result @xcite for their benchmark spin 0 resonance produced by gluon fusion , which is the closest match to our discussion .
their resonance width is lower than 15% which is adequate for @xmath107 , when the width of @xmath51 is 17% of its mass . from the upper panel
we see that @xmath108 gev is obtained for values of @xmath38 within its perturbative unitarity limit .
the bottom panel indicates that for @xmath109 the lhc constrains @xmath110 gev . for the larger value of @xmath106
the width of @xmath51 is much larger than that assumed by atlas , but if we still use this result , we find @xmath111 tev . on the right hand panel
we have used the cms results @xcite presenting three of their benchmark studies : a generic @xmath112 produced by @xmath87 with decays to @xmath100 and a width equal to 1% or 10% of its mass ; and a kk gluon with about 94% branching ratio to top - pairs . since
none of these scenarios closely resembles @xmath51 , we can only obtain a very rough limit of @xmath113 gev for large values of @xmath69 and no constraints for @xmath103 .
the relevant mechanism for this channel corresponds to tree - level pair production of the neutral scalars through gluon fusion , followed by decays into light quarks or gluons .
the production cross - section depends only on the mass of the scalars and the different decay channels can be selected as before .
nlo calculations for the pair production of scalar color octets exist @xcite but we restrict ourselves to a lo calculation here .
scalar pair production becomes dominant over single scalar production when the one - loop diagrams responsible for the latter are suppressed .
this can happen for low enough values of @xmath69 , @xmath38 and @xmath2 as we illustrate in figure [ svsss ] .
production vs @xmath114 scalar pair production cross - sections at lhc13 for representative values of the parameters . ]
we begin with both scalars decaying to dijets , for which we suppress the decay into top - pairs with @xmath70 and the decay into @xmath101 with @xmath72 .
we use the results from atlas with 15.4 pb@xmath9 at 13 tev @xcite corresponding to their benchmark scenario that most closely resembles our case , pair production of colorons . on the left panel of figure [ fig:4jets ]
we fix @xmath115 , for which figure [ br - case1 ] tells us the jets will be dominantly gluon jets .
the figure explores the sensitivity to @xmath38 , finding that resonances as light as 500 gev are still allowed for @xmath116 ( the values allowed by perturbative unitarity ) , but that masses @xmath117 gev are excluded if we allow for larger values of @xmath38 . in the panel on the right we fix @xmath94 in which case resonances with @xmath118 gev are excluded with @xmath115 . for larger @xmath2 the dominant decay is into four bottom quarks , which we treat as a separate case below . ) and their decays into 4 jets at 13tev compared with figure 8-b of atlas preliminary results @xcite ( red curve).,title="fig : " ] ) and their decays into 4 jets at 13tev compared with figure 8-b of atlas preliminary results @xcite ( red curve).,title="fig : " ] the @xmath119 , @xmath120 or @xmath121 final states have been discussed in the literature before .
they have lower cross - sections than the respective heavy - quark pair production , but suffer from a much smaller qcd background and this makes them ideal to search for new resonances .
they have studied both for pair produced resonances @xcite , as we do here , and for a new resonance produced in association with a heavy quark pair @xcite . as in the previous sub - section we consider pair production of scalars via their qcd couplings and adjust the model parameters so that they decay predominantly into top - pairs as we did in [ sectp ] .
we thus fix @xmath122 , @xmath123 and @xmath124 . for lhc data we use
the atlas results with 20.3 fb@xmath9 at 8 tev @xcite and their benchmark model of scalar gluon pair production with subsequent decays to two top - quark anti - quark pairs . in the left side panel of figure [ fig:4ts ]
we fix @xmath125 and explore the dependence on @xmath38 .
we find that a new resonance with @xmath126 gev is excluded for values of @xmath38 within their perturbative unitarity limits .
as @xmath38 increases the constraints disappear because the decay of @xmath51 into two gluons becomes dominant over @xmath100 pairs . on the right side panel
we fix @xmath88 and allow @xmath69 to vary . for @xmath127
the exclusion @xmath128 gev is the same as before .
as @xmath69 increases further the constraint again gets weaker because the resonance width becomes very large , going from 16% to 64% of the mass for @xmath129 gev as @xmath69 is increased from 5 to 10 , for example . ) and their decays into 4 tops at 8tev compared with figure 26 of atlas preliminary results @xcite ( red curve ) .
, title="fig : " ] ) and their decays into 4 tops at 8tev compared with figure 26 of atlas preliminary results @xcite ( red curve ) .
, title="fig : " ] finally we consider pair production of scalars via their qcd couplings and adjust the model parameters so that they decay predominantly into bottom - pairs .
we study this case by comparing it to atlas results from 13.3 fb@xmath9 at 13 tev and their search for higgs pair production in the @xmath130 mode @xcite . in the left panel of figure [ fig:4bs ]
we set @xmath115 and vary @xmath38 . for values of @xmath131
this lhc study excludes @xmath132 tev . at larger values of @xmath38
the constraint disappears as the decay of @xmath51 into two gluons is dominant over @xmath75 pairs . on the right side panel
we fix @xmath91 and see that for @xmath133 the region @xmath132 tev is excluded . for lower values of @xmath2
the resonance decays to two gluons as can be seen in figure [ br - case1 ] and the constraint disappears . ) and their decays into 4 bottoms at 13tev compared with figure 11 of atlas preliminary results @xcite ( red curve ) .
, title="fig : " ] ) and their decays into 4 bottoms at 13tev compared with figure 11 of atlas preliminary results @xcite ( red curve ) .
, title="fig : " ]
we have extracted constraints on the parameter space of the mw model by comparing the cross - sections for dijet , top - pair , dijet - pair , @xmath0 and @xmath1 production at the lhc with the strongest available experimental limits from atlas or cms at 8 or 13 tev .
our comparisons offer rough estimates as the experimental collaborations have not considered this particular model and their results are somewhat model dependent through factors like detector acceptance .
overall we find that masses below 1 tev have not been excluded for color octet scalars as is often claimed in the literature .
* dijet channel : @xmath95 gev are excluded for values of @xmath96 with @xmath94 . *
top - pair channel : @xmath109 the lhc constrains @xmath110 gev * dijet pairs : @xmath134 gev are excluded for @xmath131 * four tops : @xmath135 gev are excluded for @xmath136 and @xmath125 or @xmath109 and @xmath137 . *
four bottom quarks : @xmath138 gev are excluded for @xmath131 with @xmath115 . in summary , color octet masses in the mw model below 1 tev , and in some cases substantially lower , are allowed by current lhc constraints for parameters within the region where they satisfy perturbative unitarity .
a. v. manohar and m. b. wise , phys .
d * 74 * , 035009 ( 2006 ) [ hep - ph/0606172 ] .
m. i. gresham and m. b. wise , phys .
d * 76 * , 075003 ( 2007 ) [ arxiv:0706.0909 [ hep - ph ] ] . m. gerbush , t. j. khoo , d. j. phalen , a. pierce and d. tucker - smith , phys .
d * 77 * , 095003 ( 2008 ) [ arxiv:0710.3133 [ hep - ph ] ] .
j. m. arnold and b. fornal , phys .
d * 85 * , 055020 ( 2012 ) [ arxiv:1112.0003 [ hep - ph ] ] .
a. belyaev , g. cacciapaglia , h. cai , g. ferretti , t. flacke , a. parolini and h. serodio , jhep * 1701 * , 094 ( 2017 ) doi:10.1007/jhep01(2017)094 [ arxiv:1610.06591 [ hep - ph ] ] .
x. -g .
he and g. valencia , phys .
b * 707 * , 381 ( 2012 ) [ arxiv:1108.0222 [ hep - ph ] ]
. b. a. dobrescu , g. d. kribs and a. martin , phys .
d * 85 * , 074031 ( 2012 ) [ arxiv:1112.2208 [ hep - ph ] ] .
y. bai , j. fan and j. l. hewett , jhep * 1208 * , 014 ( 2012 ) [ arxiv:1112.1964 [ hep - ph ] ] .
g. cacciapaglia , a. deandrea , g. d. la rochelle and j. -b .
flament , arxiv:1210.8120 [ hep - ph ] .
j. cao , p. wan , j. m. yang and j. zhu , arxiv:1303.2426 [ hep - ph ] .
x. g. he , y. tang and g. valencia , phys .
d * 88 * , 033005 ( 2013 ) doi:10.1103/physrevd.88.033005 [ arxiv:1305.5420 [ hep - ph ] ] .
c. p. burgess , m. trott and s. zuberi , jhep * 0909 * , 082 ( 2009 ) [ arxiv:0907.2696 [ hep - ph ] ] .
b. grinstein , a. l. kagan , j. zupan and m. trott , jhep * 1110 * , 072 ( 2011 ) doi:10.1007/jhep10(2011)072 [ arxiv:1108.4027 [ hep - ph ] ] .
x. d. cheng , x. q. li , y. d. yang and x. zhang , j. phys .
g * 42 * , no .
12 , 125005 ( 2015 ) doi:10.1088/0954 - 3899/42/12/125005 [ arxiv:1504.00839 [ hep - ph ] ] .
r. martinez and g. valencia , arxiv:1612.00561 [ hep - ph ] .
m. reece , arxiv:1208.1765 [ hep - ph ] .
x. g. he , h. phoon , y. tang and g. valencia , jhep * 1305 * , 026 ( 2013 ) doi:10.1007/jhep05(2013)026 [ arxiv:1303.4848 [ hep - ph ] ] .
l. cheng and g. valencia , jhep * 1609 * , 079 ( 2016 ) doi:10.1007/jhep09(2016)079 [ arxiv:1606.01298 [ hep - ph ] ] .
l. cheng and g. valencia , arxiv:1703.03445 [ hep - ph ] .
x. -g .
he , g. valencia and h. yokoya , jhep * 1112 * , 030 ( 2011 ) [ arxiv:1110.2588 [ hep - ph ] ] . g. d. kribs and a. martin , arxiv:1207.4496 [ hep - ph ] .
t. han , i. lewis and z. liu , jhep * 1012 * , 085 ( 2010 ) doi:10.1007/jhep12(2010)085 [ arxiv:1010.4309 [ hep - ph ] ] .
g. aad _ et al . _ [ atlas collaboration ] , phys . rev .
d * 91 * , no . 5 , 052007 ( 2015 ) doi:10.1103/physrevd.91.052007 [ arxiv:1407.1376 [ hep - ex ] ] .
a. m. sirunyan _ et al .
_ [ cms collaboration ] , [ arxiv:1611.03568 [ hep - ex ] ] . n. d. christensen and c. duhr , comput .
commun .
* 180 * , 1614 ( 2009 ) [ arxiv:0806.4194 [ hep - ph ] ] . c. degrande , c. duhr , b. fuks , d. grellscheid , o. mattelaer and t. reiter , comput .
phys . commun .
* 183 * , 1201 ( 2012 ) [ arxiv:1108.2040 [ hep - ph ] ] .
v. khachatryan _ et al . _
[ cms collaboration ] , phys .
lett . * 117 * , no . 3 , 031802 ( 2016 ) doi:10.1103/physrevlett.117.031802 [ arxiv:1604.08907 [ hep - ex ] ] .
v. khachatryan _ et al . _
[ cms collaboration ] , phys .
* 116 * , no .
7 , 071801 ( 2016 ) doi:10.1103/physrevlett.116.071801 [ arxiv:1512.01224 [ hep - ex ] ] .
cms collaboration [ cms collaboration ] , cms - pas - exo-16 - 032 .
cms collaboration [ cms collaboration ] , cms - pas - b2g-15 - 003 .
d. goncalves - netto , d. lopez - val , k. mawatari , t. plehn and i. wigmore , phys .
d * 85 * , 114024 ( 2012 ) doi:10.1103/physrevd.85.114024 [ arxiv:1203.6358 [ hep - ph ] ] . c. degrande , b. fuks , v. hirschi , j. proudom and h. s. shao , phys .
d * 91 * , no .
9 , 094005 ( 2015 ) doi:10.1103/physrevd.91.094005 [ arxiv:1412.5589 [ hep - ph ] ] .
l. beck , f. blekman , d. dobur , b. fuks , j. keaveney and k. mawatari , phys .
b * 746 * , 48 ( 2015 ) doi:10.1016/j.physletb.2015.04.043 [ arxiv:1501.07580 [ hep - ph ] ] .
t. han , g. valencia and y. wang , phys .
d * 70 * , 034002 ( 2004 ) doi:10.1103/physrevd.70.034002 [ hep - ph/0405055 ] .
|
we extract constraints on the parameter space of the mw model by comparing the cross - sections for dijet , top - pair , dijet - pair , @xmath0 and @xmath1 productions at the lhc with the strongest available experimental limits from atlas or cms at 8 or 13 tev .
overall we find that masses below 1 tev have not been excluded for color octet scalars as it is often claimed in the literature .
the constraints that can be placed on coupling constants are typically weaker than those from existing theoretical considerations , with the exception of the parameter @xmath2 .
|
most common computationally intensive tasks encountered in practice may be formulated as combinatorial optimization problems ( cops ) , many of which are found to belong to the algorithmic class _ nondeterministic - polynomial complete _ ( np - complete ) @xcite .
the np - complete problems are computationally hard - they are characterized ( in the worst cases ) by exponential scaling of the running time or memory requirements with the problem size .
a special property of the class is that any np - complete problem can be converted into any other np - complete problem in polynomial time on a classical computer ; therefore , it is sufficient to find a deterministic algorithm that can be guaranteed to solve all instances of just one of the np - complete problems within a polynomial time bound .
an instance of a cop of size @xmath0 may be encoded using bit strings @xmath1 , @xmath2 , with a corresponding value of the cost function ( or `` energy '' ) @xmath3 for each string . the objective is to find the bit string(s ) with the minimum cost ( and the corresponding cost value ) . in quantum computation ,
bits @xmath4 are replaced by spin-@xmath5 _ qubits _ ; the qubit states @xmath6 and @xmath7 are eigenstates of the @xmath8 component of the @xmath9-th spin , respectively .
the hilbert space of a quantum _ register _ with @xmath0 qubits is spanned by @xmath10 basis vectors @xmath11 .
following @xcite , we consider a quantum evolution of duration @xmath12 based on the time - dependent hamiltonian @xmath13 here , @xmath14
@xmath0-bit strings , and @xmath15 is a `` driver '' hamiltonian that is constructed in such a way as to cause transitions between those states - essentially an ising - type spin hamiltonian corresponding to 1- and 2-gate operations : @xmath16 coefficients @xmath17 and @xmath18 vary in time in such a way that at the initial instant of time @xmath19 and at the final instant @xmath20 .
a particular choice of the coefficients is @xcite @xmath21 the total hamiltonian ( [ farhi ] ) produces a nontrivial quantum evolution from some initial ( superposition ) state @xmath22 to a final ( solution ) state @xmath23 .
if no knowledge about the solution is available _ a priori _ , then the initial state may be chosen as the symmetric state ( cf .
@xcite ) @xmath24this choice is appropriate provided ( [ ini ] ) is a ground state of @xmath15 ( e.g. , @xmath25 ) .
now , if @xmath12 is sufficiently large , then functions @xmath26 and @xmath27 vary in time slowly and the system will remain in the instantaneous ( adiabatic ) ground state of @xmath28 during its entire evolution @xmath29 ( cf .
@xcite ) . accordingly
, @xmath23 will be a superposition of states @xmath30 corresponding to the ground state of @xmath31 .
it is clear that in this case a measurement performed on the quantum register at @xmath32 will find with certainty one of the solutions of cop . in this case
the complexity of the quantum algorithm is determined by its duration @xmath12 .
if we expand the wavefunction of the system @xmath33 in the basis of the adiabatic eigenfunctions @xmath34 of the hamiltonian @xmath28 @xmath35 then adiabatic approximation corresponds to @xmath36 ( up to the oscillating phase factor ) .
coefficients @xmath37 with @xmath38 correspond to nonadiabatic corrections .
using perturbation theory in the basis of eigenfunctions @xmath34 the total probability @xmath39 of _ not _ finding the system at the instant @xmath40 in its adiabatic ground state equals @xmath41 ^ 4 } \label{nonad } \\ & & v_{0k}=\langle \psi_0|v|\psi_k\rangle.\nonumber\end{aligned}\ ] ] here we used the explicit form of coefficients @xmath42 given in ( [ alpha ] ) .
it is seen that during the quantum evolution coefficients @xmath43 and the largest admixture of the exited states into the total superposition occurs at the instant of time when one of the exited levels @xmath44 closely approaches the ground state ( avoided - crossing ) . from here
the overall criterion for the adiabatic evolution can be expressed in the well - known form @xmath45 where @xmath46 is the closest approach of the ground state to one of the excited states during the evolution - a minimum gap- and @xmath47 is the characteristic energy scale for the matrix elements of @xmath15 .
we note that although instantaneous nonadiabatic corrections ( [ nonad ] ) are quadratic in the parameter @xmath48 ( near the avoided crossing ) the probability of nonadiabatic transitions @xmath49 away from the ground state is exponentially small in @xmath48 @xcite .
this probability is defined on an infinite time axis and its logarithm is proportional to the imaginary part of the integral along the contour in the complex plane of @xmath40 that begins and ends on the real time axis and loops around the complex branching point @xmath50 @xmath51 here @xmath50 correspond to one of the roots of the equation @xmath52 that provides the smallest value for the exponential in ( [ imagin ] ) ( out of all possible complex solutions of ( [ imagin ] ) for different excited states @xmath53 ) . in a standard ( landau - zener ) theory of nonadiabatic transitions the value of
the exponent is approximately of the order of the parameter @xmath48 in ( [ criterion ] ) , and therefore it is the size of the minimum gap @xmath46 that determines the condition for _ t _ and hence the complexity of the quantum adiabatic search algorithm according to @xcite .
we note finally that , as pointed out in @xcite , the improved complexity of the adiabatic algorithm is determined by the instantaneous rate @xmath54 of the variation of the control parameter @xmath55 near the avoided crossing
. we will not discuss in this paper such modifications and focus primarily on intrinsic properties of the quantum system in question .
in this paper , we will analyze the complexity of the adiabatic quantum optimization for the _ set partition problem _ ( spp ) , which is one of the basic np - complete problems of theoretical computer science @xcite .
the optimization version of spp is to partition a set of @xmath0 positive integers @xmath56 into two disjoint subsets @xmath57 and @xmath58 such that the `` residue '' @xmath59 is minimized .
the complexity of the problem substantially depends on the size of the integers @xmath60 ( see below ) .
it is often customary for the analysis of the random instances of the problem to introduce finite - precision rational numbers @xmath61 that are independently and identically distributed ( i.i.d . ) in the unit interval @xmath62 $ ] .
@xmath63\ ] ] here @xmath64 is the total number of bits used to represent the numbers @xmath61 .
the values of the partition can be encoded in binaries by attaching `` sign '' bits @xmath65 to the numbers @xmath61 .
the partition residue can be defined as @xmath66 where @xmath67 here @xmath68 is a _ signed _ partition residue .
we note that by definition the problem is symmetric : two bit strings that can be obtained from each other by flipping all the bits ( @xmath69 ) correspond to two values of @xmath68 that differ only in sign .
we note that the minimum - residue partition(s ) may be thought of as the ground state(s ) of the following spin hamiltonian @xcite @xmath70 this is an infinite range ising spin glass with mattis type antiferromagnetic coupling , @xmath71 .
infinite range coupling clearly represents a major problem with direct ( ` analog ' ) physical implementation of this hamiltonian on a quantum computer .
therefore one can consider using an oracle - type cost function @xmath72 to implement the problem hamiltonian in ( [ farhi ] ) for spp .
the corresponding unitary transformation will multiply the basis states @xmath30 by phase factors @xmath73 during the elementary discrete steps of the ` continuous - time ' adabatic quantum optimization ( [ farhi ] ) .
although this approach is natural for the satisfiability problem @xcite it has a serious limitation for spp ( as well as some other np - complete problems like integer programming , where the precision of integers is of central importance ) . to demonstrate this point we need to consider the density of states of the partition residues .
we define the density of states for a given instance of spp as follows @xmath74 the exact form of @xmath75 depends on a given instance of spp ( i.e. , a particular set of numbers @xmath61 ) .
however we introduce a _ coarse - grained _ density of states @xmath76 where averaging is over an interval of @xmath77 whose size will be determined below . using ( [ omega ] ) and
( [ density ] ) we can rewrite this expression in the form @xmath78 note that @xmath79 and @xmath80 has very sharp maxima ( minima ) at those points . in their vicinities the integral in ( [ cos ] ) can be evaluated by steepest descent method for any given problem instance .
the sum over the contributions from different saddle points was obtained by mertens @xcite in his derivation of the partition function for the corresponding spin glass model .
we emphasize however that @xmath80 can have multiple sharp resonances at the intermediate points @xmath81 .
the positions of these resonances are at the multiples of @xmath82 where @xmath83 is an _ approximate _ greatest common divisor ( g.c.d . ) of the set of @xmath0 numbers @xmath61 such that @xmath84 where @xmath85 are integers and @xmath86 are residues of the division . provided that most of the residues are sufficiently small @xmath87 the function @xmath80 will have steep peaks at those points .
it can be shown that in the general case the value of the approximate g.c.d . for a set of @xmath0 @xmath88 numbers inside
the unit interval scales as @xmath89 for @xmath90 .
obviously it equals @xmath91 for @xmath92 . in
what follows we will be interested in the high - precision case @xmath90 .
we choose the size of the averaging window @xmath93 and this introduces a cut - off in the integral ( [ cost0 ] ) at @xmath94 it follows from above that in this case the values of the g.c.d
. will lie outside the cutoff and corresponding resonances will not contribute to the integral .
the value of the integral can be estimated near the single remaining maximum at @xmath95 .
the width of the maximum near that point is @xmath96 and therefore the window function in ( [ omega ] ) works as a step function in that region .
finally we obtain @xmath97 here the variance @xmath98 is a `` self - averaging '' quantity , and the coarse - graining is performed over an interval much larger than the characteristic separation between neighboring partition residue values @xmath99 yet much smaller than the scale of variation of @xmath100 : @xmath101 .
we note that in the high - precision regime ( @xmath102 ) , partition residues @xmath68 are irregularly spaced and well separated from each other ( on the scale of @xmath91 ) .
however this structure is being averaged out in ( [ density1 ] ) and the result indicates that , in general , no more structure exists on a scale @xmath103 other than that given by the gaussian distribution in ( [ density1 ] ) .
we note that this distribution is usually obtained for the spp using averaging over different instances of the problem ( cf .
@xcite , @xcite(_b _ ) ) ; here we recovered it as a coarse - grained distribution for a _ given _ instance which is more consistent with our goal of studying the complexity of the adiabatic quantum optimization algorithm @xcite .
computational complexity of spp depends critically on the number of bits @xmath64 : numerical simulations with independent identically distributed ( i.i.d . )
random @xmath64-bit numbers @xmath61 show @xcite that the solution time grows exponentially with @xmath0 for @xmath104 ( high - precision , computationally ` hard phase ' ) , and polynomially for @xmath105 ( low - precision , computationally ` easy phase ' ) , exhibiting a behavior similar to a phase transition @xcite(_a _ ) . in the low - precision phase , values of @xmath68 are equally spaced ( in @xmath91 ) and strongly degenerate each corresponding to ( roughly ) @xmath106 number of bit - strings .
this degeneracy grows exponentially with
n if @xmath64 remains fixed .
the total number of solutions with zero residues accumulate correspondingly and this is why the complexity eventually becomes polynomial in @xmath0 .
the quantum algorithm suggested in @xcite directly computes the density of states ( [ density ] ) of the spp and is efficient in finding the number of solutions in the low - precision case . in this case
it is also feasible to use a cost function @xmath72 ( provided the number of possible values does not grow exponentially with @xmath0 ) .
the situation is qualitatively different in the high - precision case .
implementation of the approach based on the above cost function will require a quantum computer using exponentially high precision physical parameters ( external fields , etc ) to control small differences _ in the phases _ of unitary transformations on the scale at least @xmath107 .
this is a technical difference from the constraint satisfaction problem in which the cost function generally takes only the set of values that scales polynomially with @xmath0 ; the size of the set equals the total number of constraints @xmath108 ( the computationally most difficult case corresponds to @xmath109 and the case of @xmath110 is not of general interest there ) . to avoid the above restriction in the implementation of the adabatic quantum optimization algorithm for spp
, we introduce a cost function @xmath111 based on a logarithmic scale of the partition residue values : @xmath112 since the density of states is linear at @xmath113 number of states @xmath114 per energy level @xmath115 will grow exponentially with @xmath116 in that range ( @xmath117 ) . the total number of levels @xmath118 depends on the value of @xmath119 . using the density of states ( [ density1 ] ) for @xmath120
one can set @xmath121 where @xmath122 is some fixed number ( a few dozen ) independent of @xmath0 .
the number of ground states of the problem hamiltonian @xmath123 approximately equals @xmath122 , and the total number of energy levels @xmath118 is close to @xmath0 .
it can be estimated from ( [ cost2 ] ) that @xmath124 .
the distribution of the low - lying states @xmath114 with cost function ( [ cost2 ] ) is somewhat similar to that in the slightly underconstrained cases of the satisfiability problem .
to study the complexity of the adiabatic quantum optimization algorithm for spp , we numerically integrate the time dependent schr " odinger equation with the hamiltonian @xmath28 ( [ farhi ] ) , ( [ driver ] ) in which we set @xmath125 and @xmath126 .
we start from the symmetric initial state ( [ ini ] ) and integrate the schr " odinger equation in the interval @xmath127 . unlike the approach adopted in @xcite we do not set _ a - priori _ a value of success probability .
instead we introduce a complexity metric for the algorithm @xmath128 here @xmath129 is the total probability of finding the system in its ground level ( with @xmath130 ) at the end of the algorithm , @xmath131 , and @xmath132 is the number of states at the ground level .
the algorithm has to be repeated on average @xmath133 number of times to reach success probability @xmath134 1 . a typical plot of @xmath135 for an instance of spp with @xmath0=15 numbers is shown in fig .
[ fig : complexity15 ] . at
very small @xmath12 the wavefunction is close to the symmetric initial state and the complexity is @xmath136 .
the extremely sharp decrease in @xmath135 with @xmath12 is due to the buildup of the population @xmath129 in the ground level as quantum evolution approaches adiabatic limit . at certain @xmath137
the function @xmath135 goes through the minimum : for @xmath138 the decrease in the number of trials @xmath133 does not compensate anymore for the overall increase in the runtime @xmath12 for each trial .
the minimal complexity @xmath139 is defined via one dimensional minimization over @xmath12 for a given problem instance @xcite . in fig.[fig : complexity ] we plotted the data for optimal complexities @xmath140 at different values of @xmath0 on a logarithmic scale . vertical sets of points on the plot indicate the results for _ all _ simulation data we currently have for each @xmath0 .
the results indicate that the median value of complexity @xmath140 scales _ exponentially _ with @xmath0 ; linear fit to the graph gives @xmath141 .
this corresponds to the scaling law @xmath142 .
the exponential behavior of the algorithm clearly manifests itself for the larger values of @xmath143 11 .
the scatter in the values of @xmath144 appears to decreases with @xmath0 however this result is probably due to the smaller number of data points available for larger @xmath0 values . in fig .
[ fig : probabilities ] we show the distribution of the probabilities @xmath145 for different values of @xmath12 for an instance of spp with @xmath0=15 ( plots for different @xmath12 shown with different colors)and precision b=25 bits .
values of @xmath146 are ordered with respect to the corresponding values of the partition residues @xmath147 .
it is clearly seen on logarithmic scale that probability distribution forms steps corresponding to different values of the cost function @xmath148 defined in ( [ cost2 ] ) . within each step
, the distribution of probabilities does not reveal any structure .
the same property holds also for intermediate times ( @xmath149 ) .
detailed analytical results @xcite indicate that it is this absence of structure in @xmath150 that is responsible for the exponential complexity of the algorithm .
in addition to solving the time - dependent schr " odinger equation we also analyzed the adiabatic solutions of the stationary schr " odinger equation with the same form of the hamiltonian @xmath28 ( [ farhi ] ) as above .
our preliminary results were obtained using mathematica for modest values of @xmath151 10 .
the results for n=10 are shown in figs .
[ fig : eigenvalues1 ] and [ fig : eigenvalues2 ] .
figure [ fig : eigenvalues2 ] represents the magnified part of fig .
[ fig : eigenvalues1 ] near the avoided crossing region .
adiabatic eigenvalues were computed for different values of the scaled time parameter @xmath152 .
the solid line represents the evolution of the ground state eigenvalue between @xmath153 and @xmath154 .
the vertical sets of points correspond to excited adiabatic levels for a given @xmath155 . at the beginning ( @xmath153 ) eigenvalues correspond to those of the hamiltonian @xmath15 : equally spaced levels -n ,- n+2 , @xmath156 , n , corresponding to different number of spin excitations along the @xmath157 quantization axis .
the first excited state is @xmath158fold degenerate , the second is @xmath159 , the k - th exited state is @xmath160-fold degenerate , etc . for @xmath161 the degeneracy is removed .
for @xmath162 the eigenvalue spectrum is the one for the problem hamiltonian @xmath163 ( in [ hp ] ) .
we have shifted the energy reference in the hamiltonian @xmath31 by @xmath164 ( cf . also ( [ farhi ] ) ) to match the energy scale for the symmetric case which emphasizes the avoided crossing region . in our case
the ground state was 13-fold degenerate and the corresponding eigenvalues merge at @xmath162 .
the close approach of these eigenvalues is not relevant for the minimum - gap analysis since they all end up in the same final level .
however the minimum separation of the instantaneous adiabatic ground state eigenvalue from the excited state eigenvalues that _ do not _ end up on the same ground level at @xmath154 is clearly seen in the figures .
note that the size of this separation is much greater than the separations between the excited states .
this behavior clearly departs from the standard 2-level avoided crossing picture and is due to the contributions from the exponential number of terms in ( [ nonad ] ) as will be analyzed elsewhere @xcite .
we also note that the value n=10 does not correspond to the exponential scaling regime for the algorithmic complexity that appears to start for greater @xmath0 values as follows from the discussion above . in conclusion
, we have performed numerical simulations of the adiabatic quantum optimization for spp using a step - like density of states defined on a logarithmic scale of partition residues .
the results indicate an exponential scaling of the algorithmic complexity as a function of the problem size .
the apparent reason is the loss of structure in spp during the effective coarse - graining over the intervals of partition residues corresponding to the same cost function values .
( _ a _ ) e. farhi , j. goldstone , s. gutmann , and m. sipser , `` quantum computation by adiabatic evolution , '' arxiv : quant - ph/0001106 ; ( _ b _ ) e. farhi , j. goldstone , s. gutmann , j. lapan , a. lundgren , and d. preda , `` a quantum adiabatic evolution algorithm applied to random instances of an np - complete problem '' , _ science _ * 292 * , 472 ( 2001 ) .
e. farhi , j. goldstone , and s. gutmann , `` a numerical study of the performance of a quantum adiabatic evolution algorithm for satisfiability , '' arxiv : quant - ph/0007071 ; a. m. childs , e. farhi , j. goldstone , and s. gutmann , `` finding cliques by quantum adiabatic evolution '' , arxiv : quant - ph/0012104 .
s. mertens , ( _ a _ ) `` phase transition in the number partitioning problem , '' physical review letters * 81 * , 42814284 ( 1998 ) ; ( _ b _ ) `` random costs in combinatorial optimization , '' _ physical review letters _ * 84 * , 13471350 ( 2000 ) .
we also note that occasionally for certain `` singular '' instances of the partition problem values of approximate g.c.d
. can be rather large ( e.g. @xmath165 ) , i.e. most of the numbers are nearly commensurate with each other .
in this case intermediate resonances will be of interest and the density of states will have an additional structure at low frequencies @xmath166 .
we do not consider those instances in a present paper .
we note that an additional optimization can be done if one defines a complexity @xmath167 similar to ( [ complexity ] ) but uses an intermediate time instance @xmath40 instead of @xmath12 there . in this case quantum algorithm
is terminated at the instance @xmath168 for each @xmath12 when minimal complexity is reached and then additional minimization over @xmath12 is performed .
our results indicate that this method does provide further improvement for the overall complexity but takes a prohibitely long time to perform numerical optimization of @xmath167 .
|
we analyze the complexity of the quantum optimization algorithm based on adiabatic evolution for the np - complete set partition problem .
we introduce a cost function defined on a logarithmic scale of the partition residues so that the total number of values of the cost function is of the order of the problem size .
we simulate the behavior of the algorithm by numerical solution of the time - dependent schr " odinger equation as well as the stationary equation for the adiabatic eigenvalues .
the numerical results for the time - dependent quantum evolution indicate that the complexity of the algorithm scales exponentially with the problem size .
this result appears to contradict the recent numerical results for complexity of quantum adiabatic algorithm applied to a different np - complete problem ( farhi et al , _ science _ 292 , p.472 ( 2001 ) ) .
|
the two - photon state generated via spontaneous parametric down - conversion ( spdc ) is one of the most well - known examples of two - particle entangled states .
the spdc process can be briefly explained as a spontaneous splitting or decay of a pump photon into a pair of daughter photons ( typically called signal and idler photons ) in a nonlinear optical crystal @xcite .
this spontaneous decay or splitting only occurs when energies and momentum of the interacting photons satisfy the conservation condition , which is known as the phase matching condition .
signal and idler photons have the same polarization in type - i phase matching and have orthogonal polarization in type - ii phase matching . if the phase matching is perfect , assuming monochromatic plane wave pump , it is not hard to see that the state of spdc should be written as @xmath0 ( @xmath1 ) are the frequency and the wave vectors of the signal , idler , and the pump , respectively , and @xmath2 is the creation operator for the signal photon .
the @xmath3 functions in state ( [ eq : perfect ] ) ensure that the signal and the idler photons satisfy the phase matching condition , i.e. , there is only one energy or wave vector for the idler photon corresponding to a given energy or wave vector for the signal photon . in other words ,
the signal - idler photon pair is perfectly entangled in energy and momentum .
such a perfectly entangled two - photon state ( in energy and momentum ) naturaly has an infinite two - photon coherence time . in reality , such a strict one - to - one correspondence does not happen , because perfect phase matching can never occur , most notably , due to pump beam divergence , limited pump beam size , and limited thickness of the nonlinear crystal @xcite .
therefore , the above delta functions should be replaced by two - photon spectral functions that are sharply peaked around @xmath4 and @xmath5 and have some bandwidths .
this means that given an energy or momentum of the signal photon , there are some ranges of energies and momenta available for the idler photon : the entanglement between the photon pair is less - than - perfect . as a result , the two - photon state of spdc has finite coherence time and the shape of the correlation function is determined by the two - photon spectral functions , which depend on the types of phase matching .
( note that monochromatic pumping condition is assumed in this paper . )
we may then define the one - photon and the two - photon wavepackets of the spdc as the envelope of the first - order interference observed in the single - detector count rate and the envelope of the second - order interference observed in the coincidence count between two detectors , respectively . in this paper , we present theoretical calculation and experimental measurements of the one - photon and the two - photon wavepackets of spdc for both type - i and type - ii phase matching conditions .
first , in section [ sec : coh1 ] , we calculate the first - order ( @xmath6 ) and the second - order ( @xmath7 ) correlation functions for the quantum state of spdc .
we then calculate , in section [ sec : coh2 ] , ( i ) the one - photon wavepacket : the envelope of first - order interference fringe due to a michelson interferometer and ( ii ) the two - photon wavepacket : the envelope of second - order interference due to a shih - alley / hong - ou - mandel interferometer using the spdc state as the input @xcite .
it is found that the two - photon wavepacket measured this way is not related to the second - order correlation function @xmath8 of the state but related to the first - order correlation function ( @xmath6 ) of the state @xcite .
these predictions are experimentally tested in section [ sec : exp ] using type - ii and type - i spdc .
it turns out that , the experimental results for the type - i spdc case do not quite agree with the predictions in a realistic experimental setup .
possible reasons for such deviations are discussed .
the quantum state of spdc can be calculated using first - order perturbation theory @xcite , @xmath9 @xmath10 is the interaction hamiltonian which takes the form @xmath11 where @xmath12 is the pump laser field which is considered monochromatic ( cw ) and classical . assuming that it is propagating in @xmath13 direction and has frequency @xmath14 , @xmath15 $ ] .
@xmath16 , ( @xmath17 ) , is the quantized field operator for the signal and the idler photons . assuming also that @xmath16 is propagating in @xmath13 direction and has central frequency @xmath18 , it can be written as @xmath19 $ ] .
the state of spdc can then be calculated as @xcite , @xmath20 and @xmath21 is the thickness of the nonlinear crystal .
( we have ignored the normalization constant for simplicity . )
since the condition @xmath22 has to be satisfied at all times , we can simplify the above equation further by introducing the detuning frequency @xmath23 or , equivalently , @xmath24 where @xmath25 .
we therefore obtain @xmath26 where @xmath27 .
@xmath28 is the joint spectral function for the signal and the idler photons which determines the coherence properties of the state and @xmath29 is the frequency dependent phase term .
note also that this expression clearly shows the frequency anti - correlated feature of the signal and the idler photons .
the specific forms of @xmath28 depend on the phase matching condition used in an experiment and are well - known @xcite .
for type - ii spdc , @xmath30 where the group velocity difference ( in the crystal ) @xmath31 and @xmath21 is the thickness of the nonlinear crystal . in type - i spdc
, it is given as @xmath32 where the group velocity dispersion ( in the crystal ) @xmath33 .
( note @xmath34 for type - i spdc . )
let us now introduce the density operator as it is convenient to use the reduced density operator to calculate the first - order correlation function of the state .
since the two - photon state ( [ eq : state ] ) is a pure state , the density operator for the two - photon state is simply given as @xmath35\nonumber\\ & = & \int_{-\infty}^{\infty } d\nu \ , |s(\nu)|^2 \ , a_s^\dagger(\omega+\nu)|0\rangle\langle0|a_s(\omega+\nu).\end{aligned}\ ]
] first - order correlation function of the state can be calculated using the reduced density operator obtained in eq .
( [ eq : den ] ) . for stationary fields ,
the first - order correlation function can be written as @xcite @xmath36,\ ] ] where @xmath37 $ ] .
the first - order correlation function of the signal photon can then be calculated as @xmath38 where @xmath39 . as we can see clearly ,
first - order correlation function of the signal photon is simply a fourier transform of the power spectrum of the signal photon .
second order correlation function can be calculated rather simply by using state ( [ eq : state ] ) , @xmath40 where @xmath41 $ ] and @xmath42 $ ] . by using @xmath43 and @xmath44 , it is straightforward to obtain @xmath45 note that @xmath6 and @xmath7 can have quite different shapes even though they are associated with the same @xmath28 .
for example , @xmath6 does not get affected by the introduction of group velocity dispersion between the source and the detector , but @xmath7 gets broadened by it @xcite .
it is because any dispersion introduced in @xmath46 simply cancels when calculating @xmath6 . in the case of @xmath7
, this cancellation does not happen because two different fields are involved @xcite .
we have so far calculated first- and second - order correlation functions of the quantum state of spdc . in this section ,
we study how these correlation functions are actually linked to one - photon and two - photon wavepackets in simple interference experiments . for one - photon wavepacket measurement ,
we consider the output of a simple michelson interferometer , in which either the signal or the idler photons are the input .
for the two - photon wavepacket measurement , we consider a well - known shih - alley / hong - ou - mandel interferometer setup , in which the signal - idler photon pair is made to interfere at a beamsplitter and the coincidence counts between detectors , which are placed at the output ports of the beamsplitter , are measured @xcite .
let us first calculate the single count rates at the output port of a michelson interferometer when the signal photon of spdc is the input . in this case , the single count rate is proportional to @xmath47 , @xmath48,\ ] ] where the reduced density operator for the signal photon , @xmath49 , is given in eq .
( [ eq : den ] ) , @xmath50 + a(\omega ) \exp[-i\omega(t+\tau)]\right\}d\omega$ ] , and @xmath51 is the delay between the two arms of the interferometer .
it is then easy to show that @xmath52 which can be re - written as @xmath53 where @xmath54 .
therefore , the envelope of the interference fringe or the one - photon wavepacket observed at the output of the michelson interferometer directly corresponds to the first - order correlation function @xmath6 . for a two - photon wavepacket
, we need to calculate the coincidence count rates for a shih - alley / hong - ou - mandel interferometric setup @xcite .
consider the following setup : signal and idler photons are generated at the crystal , propagate at different directions , reflect off at mirrors , and made to interfere at a beamsplitter .
if both photons have the same polarization , polarization of one of the photons is rotated by @xmath55 before reaching the beamsplitter .
the delay between the two paths are @xmath51 .
a detector package which consists of a single photon detector and a polarization analyzer is placed at each output port of the beamsplitter and the coincidence counts between the two detectors are recorded ( see experimental setup shown in fig .
[ fig : type2fig ] and fig .
[ fig : type1fig ] ) .
the coincidence count rate is then proportional to @xmath56 , @xmath57 the quantized electric fields @xmath58 and @xmath59 at the detectors @xmath60 and @xmath61 can be written as @xmath62 where @xmath63 and @xmath64 are the angles of the polarization analyzers placed before the detectors , @xmath51 is the delay introduced between the two arms of the interferometer , @xmath65 and @xmath66 are the times at which the detectors @xmath60 and @xmath61 click , and the phase factor @xmath67 comes from the reflection at the beamsplitter . to trace the envelope of the interference , @xmath63 and @xmath64 values should be chosen so that the maximum and minimum of the interference can be observed for a certain value of @xmath51
. these values are @xmath68 for the interference minima and @xmath69 for the interference maxima @xcite . further evaluating eq .
( [ eq : rc1 ] ) for @xmath68 , we get @xmath70 where @xmath71 , @xmath72 . in approximating the above equation ,
we ve used the fact that @xmath28 is an even function and @xmath73 is in optical frequency .
the above equation can be re - written as @xmath74 where @xmath75 sign is for @xmath68 and @xmath76 sign is for @xmath69 .
it is interesting to note that the two - photon wavepacket , eq .
( [ eq : rc2 ] ) , does not contain the second - order correlation function @xmath7 .
this result has interesting implications : ( i ) @xmath56 has the same envelope as @xmath47 except that the @xmath56 envelope is half of the @xmath47 envelope , ( ii ) any dispersion element introduced in a shih - alley / hong - ou - mandel interferometer can not affect the shape of the interference envelope since @xmath77 is not affected by group velocity dispersion @xcite .
in this section , we describe two experiments which are designed to test the predictions made in section [ sec : coh2 ] . for both type - i and type - ii spdc experiments ,
the pump laser was a argon ion laser operating at 351.1 nm .
coincidence counts were measured using a time - to - amplitude converter ( tac ) and multi - channel analyzer ( mca ) set .
the coincidence window used for second - order interference measurement was 3 nsec .
let us first describe the wavepacket measurement of type - ii spdc .
the experimental setup can be seen in fig .
[ fig : type2fig ] .
a 2 mm thick type - ii bbo crystal was pumped by a 351.1 nm laser beam generating 702.2 nm collinear type - ii spdc photons .
the residual pump beam was removed by two pump reflecting mirrors ( pm ) . instead of using usual quartz delay line for introducing fine delay between horizontal and vertically polarized photons @xcite , a set of pbs , @xmath78 plates ( oriented at @xmath79 ) , and mirrors ( shown in the shaded area ) was used .
the inset containing a pbs and a @xmath80 plate ( oriented at @xmath81 ) was used to remove vertically polarized photons when measuring first - order interference . the delay between the two arms of the interferometer
was introduced by moving mirror m1 with an encoder driver .
photons were finally detected with detector packages which consist of a single - photon counting module and a polarization analyzer .
the distance from the bbo crystal to the detector was approximately 218 cm and all apertures used in this experiment were about 3 mm in diameter .
[ fig : dataii ] shows the experimental data for the type - ii spdc experiment .
[ fig : dataii](a ) shows first - order interference of the horizontal polarized photon . as discussed above ,
vertical polarized photons are removed with a pbs and a @xmath80 plate is used to rotate the polarization direction to @xmath79 . for this measurement , we used a 20 nm fwhm filter to suppress the white - light interference which occurs around @xmath82 @xcite . the observed triangular one - photon wavepacket agrees well with the theoretical prediction shown in fig .
[ fig : theory](a ) . to measure the two - photon wavepacket ,
we first removed the pbs-@xmath80 plate set used for first - order interference measurement .
the e - ray of the crystal ( vertically polarized photons ) could then be delayed with respect to the o - ray by moving m1 . for this measurement ,
only uv - cut off filters ( cut - off at 550 nm ) were used to suppress any residual pump noise .
[ fig : dataii](b ) shows typical triangular two - photon wavepacket observed in coincidence counts in which the dip or peak occurs when the e - polarized photons are delayed by @xmath83fs with respect to the o - polarized photons before reaching the beamsplitter bs @xcite .
again , the observed triangular two - photon wavepacket agrees well with the theoretical prediction shown in fig .
[ fig : theory](b ) . as we have predicted in the previous section ,
the one - photon and the two - photon wavepackets have the same shapes and the one - photon wavepacket is twice bigger than the two - photon wavepacet .
although we here have used collinear type - ii spdc for one - photon and two - photon wavepacket calculation and measurements , recent experimental results confirm that non - collinear type - ii spdc gives the same result for the two - photon wavepacket measurement @xcite . finally , we note that the power spectrum function , in type - ii spdc , includes parameters of both the signal and the idler photons ( in @xmath31 ) even though only one of them is actually measured @xcite .
let us now discuss the measurement of one - photon and two - photon wavepackets for type - i spdc .
the experimental setup can be seen in fig .
[ fig : type1fig ] .
the pump laser beam was centered at 351.1 nm and 702.2 nm centered spdc photons were generated from a 2 mm thick type - i bbo crystal .
the propagation angle of the signal - idler photon pair was about @xmath84 with respect to the pump beam propagation direction .
a @xmath80 plate was used to rotate the signal photon s polarization from horizontal polarization to vertical polarization .
the signal - idler photons were then made to interfere at a beamsplitter and the delay @xmath85 was varied by an encoder driver driven trombone prism .
detectors d1 and d2 placed at the output ports of the beamsplitter were used for second - order interference measurement .
for first - order interference measurement , a flipper mirror ( fm ) was used to direct the idler photon to the secondary michelson interferometer .
interference was measured by detector d3 as a function of the arm length difference @xmath86 .
the crystal to the d3 distance was about 200 cm and to d2-d1 was about 280 cm .
as before , all apertures used in this experiment were about 3 mm in diameter . for one - photon wavepacket measurement , we used the flipper mirror ( fm ) to direct the idler photon to the secondary michelson interferometer as shown in fig .
[ fig : type1fig ] .
first - order interference was observed by moving one of the mirror by an encoder driver and to reduce any residual pump noise , a uv - cutoff filter was used in front of the detector .
the experimental data for this measurement can be seen in fig .
[ fig : datai](a ) and the envelope of first - order interference or one - photon wavepacket closely follows the predicted curve shown in fig .
[ fig : theory](c ) .
this confirms that the power spectrum of either the signal or the idler photon of type - i spdc is indeed given by @xmath87 . for two - photon wavepacket measurement
, we removed the flipper mirror ( fm ) and let the signal - idler photons interfere at the beamsplitter .
the coincidence counts were measured as a function of both the delay @xmath85 and the analyzer angles .
we also measured the true coincidence counts as well as accidental coincidence counts by integrating additional 3 nsec window which was located about 10 nsec away from the true coincidence peak in the mca spectrum .
we first measured the two - photon wavepacket with 3 nm fwhm spectral filters .
the data for this experiment can be seen in fig .
[ fig : datai](b ) .
the visibility for this measurement is quite high ( @xmath88% ) , however , the two - photon wavepacket has quite different shape than the one - photon wavepacket , mostly due to narrow - band filtering of the spdc photons by the 3 nm filters .
we can roughly estimate the contribution of the spectral filters to the broadened coherence time by using @xmath89 fsec .
considering that the filters do not necessarily have perfect gaussian shape transmission curve and they may have different fwhm values than the specified values , this rough estimation gives a pretty good idea on the origin of the broadened two - photon wavepacket .
note also that the level of accidental coincidence is nearly negligible .
the same measurement was repeated with two more sets of spectral filters : 20 nm fwhm and 80 nm fwhm . with 20 nm filters ,
see fig .
[ fig : datai](c ) , we still observe quite good visibility of 83% .
notice that the level of accidental contribution has risen slightly and the two - photon wavepacket is now narrower . by using the simple picture again , we estimate @xmath90 fsec . this value is quite close to the observed two - photon wavepacket shown in fig .
[ fig : datai](c ) and it means that we are still observing a spectrally filtered , by the spectral filters , two - photon wavepacket .
finally , 80 nm fwhm filters were used for the same measurement , see fig .
[ fig : datai](d ) .
we find that the raw visibility has now dropped to 32% with almost no change in the two - photon wavepacket .
according to fig .
[ fig : theory](d ) , the unfiltered two - photon wavepacket should be about 15 fsec in fwhm .
note also that the contribution of accidental coincidence is now significant , unlike fig .
[ fig : datai](b ) and fig .
[ fig : datai](c ) .
this rather unexpected behavior of the two - photon wavepacket with broadband spectral filters can be understood as follows .
it is well - known that type - i spdc in general has much bigger bandwidth than type - ii spdc .
especially , for the case considered in this paper , the calculated fwhm of the type - i spdc spectrum is more than 80 nm which is much bigger than roughly 3 nm calculated fwhm bandwidth of type - ii spdc .
this calculation actually agrees quite well with the observed first - order interference shown in fig .
[ fig : dataii](a ) and fig .
[ fig : datai](a ) . based on this observation alone
, it may seem , at first , that the bandwidth of type - i spdc is not limited at all at the detectors . to see what really is happening , however , it is necessary to consider the type - i spdc tuning curve , which shows how spdc spectrum is distributed as functions of propagation angles and wavelengths .
[ fig : tuning ] shows the tuning curve of non - collinear type - i spdc used in this experiment .
the left curve shows the angle - spectrum distribution for the signal photons and the right curve shows the same for the idler photons .
note that the signal and the idler photons have the same angle - spectrum distribution as both signal and idler photons have the same polarization .
two vertical bars represent the angles defined by the apertures used in the experiment . for one - photon wavepacket measurement , only the signal or the idler photons
are measured .
if is clear from fig .
[ fig : tuning ] that the signal or the idler photons indeed have quite broad bandwidth ( more than 80 nm ) even if we only consider the small angle defined by the aperture .
it is because the slope of the tuning curve for type - i spdc is not steep , unlike type - ii spdc . for second - order interference measurement ,
however , we need to consider signal - idler photon pair detections and the fact that their frequencies are anti - correlated , i.e. , @xmath91 and @xmath92 . if spectral filers have narrow bandwidths around 702.2 nm ( shown as two circles at 702 nm in fig . [ fig : tuning ] ) , we just need to consider the cross - sections of the vertical bars and small area around 702.2 nm .
it is then easy to see that only frequency anti - correlated photon pairs can be detected almost all the time .
as we use spectral filters with broader bandwidths , the possibility of detecting uncorrelated photons gets bigger and bigger .
it is because when the filter bandwidth is big , uncorrelated photons with large frequency difference can result in significant accidental coincidence counts because of the apertures used in the experiment .
let us consider an example : the signal photon is at 662 nm and , due to the energy conservation condition , the idler photon is at 748 nm ( shown as two circles at 662 nm and 748 nm ) .
the 662 nm signal photon can be detected .
however , as we can see in fig .
[ fig : tuning ] , the 748 nm idler photon simply can not be detected because it lies outside the detectable area in the tuning curve .
therefore , when broadband filters are used , the two - photon wavepacket is mainly determined by spatial filtering by apertures rather than spectral filters . also , increased non - pair detection or uncorrelated photon detection due the use of broadband filters greatly increases the level of accidental coincidence counts , which in turn reduces the raw quantum interference visibility .
this effect is clearly demonstrated in fig .
[ fig : datai](b ) @xmath93 fig .
[ fig : datai](d ) . up
to 20 nm filters , uncorrelated detection evens are still quite negligible as accidental coincidence counts do not reduce visibility much . with 80 nm filters , which accepts nearly full bandwidths of type - i spdc ,
the accidental coincidence contribution is huge and this reduces the raw visibility significantly .
since such accidental coincidence counts from uncorrelated events produce a flat background , they can be subtracted from the overall coincidence counts ( the corrected visibility increases to 86% ) .
it may be possible to remove such spectral filtering effect by removing the apertures altogether or opening them completely .
however , this makes it nearly impossible to align and use the interferometer because spatial modes can not be well defined and the accidental coincidence will increase significantly , just as in this experiment .
we have indeed observed slight narrowing of two - photon wavepacket by opening the apertures , but , due to limited collection angles of our detection system , it was not possible to observe very short two - photon wavepacket predicted in section [ sec : coh2 ] . if the bandwidth of type - i spdc is inherently narrow and the angle - wavelength tuning curve slope is steep , for example , by using a different non - linear crystal or using different phase matching scheme , such an effect may nevertheless be observed .
for example , burlakov _ et .
. observed a similar effect using collinear type - i spdc from lio@xmath94 crystals , but the interferometric two - photon wavepacket measurement scheme involved two nonlinear crystals instead of one : i.e. , the two - photon wavepacket was measured by interfering two - photon amplitudes from different crystals .
we have measured one - photon and two - photon wavepackets of type - i and type - ii spdc generated from a cw laser pumped bbo crystal . in the case of type - ii spdc ,
the measured wavepackets agreed well with the theory .
although we used collinear type - ii spdc for these measurements , it was observed elsewhere that non - collinear type - ii spdc gives the same result @xcite . in experiments involving type - i spdc , even though the one - photon wavepacket measurement agreed well with the theory , the two - photon wavepacket was much bigger than the expected value . upon studying the tuning curve of type - i spdc from a bbo crystal
, we found that spatial filtering limits the two - photon pair detection bandwidth even though one - photon bandwidth is not limited .
such spatial filtering is specific to the tuning curve characteristics of the nonlinear crystal and the geometry of the experiment .
it may be avoided with the use of a nonlinear crystal which has sharp ( non - collinear ) angle - wavelength tuning characteristics or with the use of the collinear multi - crystal geometry @xcite . in this paper , we have used well - known one - dimensional approximation when calculating one - photon and two - photon wavepackets .
this approximation works well with both collinear and non - collinear type - ii spdc experiments with a bbo crystal .
however , for broadband non - collinear type - i spdc , it became clear that tuning curve characteristics of the crystal needs to be considered seriously .
our results imply that one should be careful using non - collinear type - i spdc in applications in which variables other than polarizations , such as energy and momentum , are important .
one example of such possible applications is quantum metrology ; care must be taken not to over - estimate the two - photon bandwidth .
our results also imply that generating high - purity polarization - entangled states or bell - states using type - i non - collinear spdc from a bbo crystal almost always rely on strong spectral post - selection , as increased detection bandwidths reduce the raw visibility significantly , even with a cw pump .
the author wishes to thank m.v .
chekhova for many enlightening discussions and w.p .
grice for helpful comments .
this research was supported in part by the u.s .
doe , office of basic energy sciences , the national security agency , and the ldrd program of the oak ridge national laboratory , managed for the u.s .
doe by ut - battelle , llc , under contract no .
de - ac05 - 00or22725 .
y.h . shih and c.o .
alley , in _ proceedings of the second international symposium on foundations of quantum mechanics in the light of new technology , tokyo , 1986 _ , edited by m. namiki ( physical society of japan , tokyo , 1987 ) .
burlakov , m.v .
chekhova , o.a .
karabutova , and s.p .
kulik , `` collinear two - photon state with spectral properties of type - i and polarization properties of type - ii spontaneous parametric down - conversion : preparation and testing , '' phys .
a * 64 * , 041803(r ) ( 2001 ) .
it is possible to completely remove the group velocity dispersion effect from @xmath7 if the positive dispersion introduced in @xmath95 is matched with a negative dispersion introduced in @xmath96 , see ref . .
|
one - photon and two - photon wavepackets of entangled two - photon states in spontaneous parametric down - conversion ( spdc ) fields are calculated and measured experimentally . for type - ii spdc ,
measured one - photon and two - photon wavepackets agree well with theory . for type - i spdc ,
the measured one - photon wavepacket agree with the theory .
however , the two - photon wavepacket is much bigger than the expected value and the visibility of interference is low .
we identify the sources of this discrepancy as the spatial filtering of the two - photon bandwidth and non - pair detection events caused by the detector apertures and the tuning curve characteristics of the type - i spdc .
|
a family of @xmath0-(bedt - ttf)@xmath3 ( bedt - ttf = bis(ethylenedithio)-tetrathiafulvalene , @xmath4 = cu[n(cn)@xmath1]br , cu[n(cn)@xmath1]cl , cu(ncs)@xmath1 , etc . ) salts is a charge - transfer group with quasi - two - dimensional electronic properties .
@xcite these salts have attractive physical properties , including a range of antiferromagnetic insulating ( afi ) to superconducting ( sc ) ground states depending on the pressure , @xcite deuteration , @xcite cooling rate , @xcite and magnetic field strength .
@xcite in these salts , the @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]br ( abbreviated @xmath5 $ ] hereafter ) is in the sc state in the vicinity of a sc afi boundary , named the mott boundary .
@xcite additionally , a perfectly deuterated salt ( @xmath6 $ ] ) formed under slow cooling conditions is believed to exist on the mott boundary . in the case of the fast cooling conditions , the ground state changes to be in the afi phase because the disorder originating from the mixing of the eclipsed and staggered conformations of the ethylene end groups in the bedt - ttf molecules of the sample affects the ground state characteristics .
@xcite in intermediate materials in which the deuterium is randomly distributed in the ethylene groups , the ground state is on the boundary between the sc and afi phases . in these cases , a partial sc state in the material
is observed .
@xcite for instance , a 75 % -deuterated material ( @xmath7 $ ] ) that exhibits a uniform metallic ( or sc ) state under slow cooling conditions changes to an insulating state with fast cooling conditions .
@xcite the spatial distribution of the center of the spectral weight ( @xmath8 ) of the reflectivity spectra [ @xmath9 in the 10 @xmath10m - scale region on the sample surface indicates that the inhomogenity contained in the @xmath7 $ ] sample itself is the origin of the metal insulator phase separation .
@xcite in the case of the presence of magnetic fields , the 50 % -deuterated @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]br ( @xmath11 $ ] ) under fast cooling conditions directly changes from the sc to afi states in spite of the fact that the sc state of the material formed with slow cooling conditions normally changes to a paramagnetic metallic ( pm ) state with increasing magnetic field strength .
@xcite in this paper , we report the magnetic field dependence of the spatial image of the sc afi transition in @xmath11 $ ] using the infrared magneto - optical imaging spectroscopy .
in addition , we found a magnetic field - induced sc afi
metal transition ( simt ) in a tiny region on the sample surface that appeared under the appropriate conditions of both the inhomogenity and disorder of the ethylene groups caused by fast cooling conditions .
the origin of the simt and the conceptual phase diagram of the magnetic field ( @xmath12 ) to @xmath13 in which @xmath14 and @xmath15 are the intradimer coulomb repulsion energy and the band width , respectively , of @xmath11 $ ] at 5 k are also discussed .
single crystals of @xmath11 $ ] were grown using a conventional electrochemical oxidation method .
the unpolarized infrared reflectivity spectra and the spatial image were measured at the infrared magneto - optics station of the beam line 43ir of a synchrotron radiation facility , spring-8 .
@xcite to observe the cooling rate dependence of the samples , two cooling rates , -17 k / min for fast cooling , and -0.05 k / min for slow cooling , were used in the temperature range from 90 to 70 k. to check the reproducibility of the spectra and the spatial images , the same fast cooling experiment was repeated , and also another @xmath11 $ ] sample was tested .
a strong reproducibility was obtained , indicating that the @xmath16 and the spatial image mainly originate from the inhomogenity of the sample itself , which has been demonstrated in previous studies .
@xcite the infrared @xmath16 imaging was measured in the wave number range from 870 to 8,000 @xmath17 at a temperature of 5 k and magnetic fields of 0 , 5 , and 10 t. the measurements were done only with increasing magnetic field strength only . ] in the superconducting ( @xmath5 $ ] , @xmath18 = 5 k , @xmath12 = 0 t ) and antiferromagnetic insulating ( @xmath6 $ ] , @xmath18 = 5 k , @xmath12 = 0 t ) phases and their centers of spectral weight ( @xmath8 ) . , width=226 ] to acquire the spatial imaging data , a total of 931 spectra were acquired in a 360@xmath19360 @xmath10m@xmath20 region at steps of 12 @xmath10 m with a spatial resolution of less than 20 @xmath10 m .
the spatial image of the reflectivity spectra were plotted using @xmath8 obtained from the following function : @xmath21 for each @xmath16 .
@xcite here , @xmath22 is the lowest accessible wavenumber , 870 @xmath17 , and @xmath23 is set to 5,000 @xmath17 , above which there is no difference between the metallic and non - metallic spectra .
@xcite the reason for this convergence in the spectra is that the change in the electronic structure due to the mott transition @xcite results in a shift of the optical spectral weight , because the optical spectra indicate the relative energy difference between the occupied and unoccupied states .
for instance , typical @xmath16 spectra in the sc ( @xmath5 $ ] , @xmath18 = 5 k , @xmath12 = 0 t ) and afi ( @xmath6 $ ] , @xmath18 = 5 k , @xmath12 = 0
t ) phases and their respective @xmath8 values are shown in fig .
although the origin of the spectral shape between 2,000 and 4,000 @xmath17 has not been concluded yet , the difference in the electronic structure is directly reflected in @xmath16 as well as @xmath8 .
the metal
insulator character in @xmath16 mainly appears below 1000 @xmath17 as shown in fig .
[ fig1 ] . however the spatial resolution becomes poorer than the present method because of the diffraction effect of light .
therefore , the @xmath8 was employed as being the representative of the shape of @xmath16 as well as the electronic structure . to check the wavenumber distribution of @xmath8 ,
the pixel number of @xmath8 as a function of @xmath8 was also derived . from our previous work , the rough boundary ( @xmath24 ) between metallic and insulating characteristics in the same analysis method
was evaluated to be around 2,350 @xmath17 .
@xcite the other analysis methods , the reflectivity ratio between at the lowest accessible wavenumber of 870 @xmath17 and at the shoulder structure of 3,500 @xmath17 and the wavenumber shift of the @xmath25 molecular vibration mode , @xcite gave the similar result .
note that this method can not separate an sc state from a normal metallic state because sc behavior does not effect @xmath16 down in the accessible lowest wavenumber of 870 @xmath17 . then the obtained metallic @xmath16 below the upper critical field ( @xmath264 t ) and below the critical temperature ( @xmath2711 k ) is regarded to be in an sc state .
) of 50 % -deuterated @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]br ( @xmath11 $ ] ) at @xmath18 = 5 k. the blue circles in ( b ) indicate the points of the different magnetic field dependencies on the sample surface are shown in fig .
the black lines indicate the rough m
i boundary ( @xmath24 ) of 2,350 @xmath17 and the lower and higher wavenumbers indicate the insulating and metallic ( superconducting ) reflectivity spectra , respectively . the hatched area is the outside of the sample .
( c ) magnetic field and cooling rate dependences of the normalized resistivity of @xmath11 $ ] at 5.5 k for the reference . though the cooling rate of -0.73 k / min is different from that of ( b ) , the physical character is same as shown in ref .
5 . , width=283 ] for @xmath11 $ ] derived from fig .
the gaussian fittings to the histograms are also plotted using lines .
the fitting results are discussed in the text .
, width=283 ] the spatial images of @xmath8 obtained under the slow cooling ( a ) and fast cooling ( b ) conditions are shown in fig . [ fig2 ] .
in the case of slow cooling at 0 t , the spatial image is nearly monochromatic , and then the @xmath8 value over the whole sample surface is lower than @xmath24 .
this indicates the whole sample surface is in the metallic state .
the metallic state does not change with increasing magnetic field strength up to 10 t. these results are consistent with the electrical resistivity data under magnetic fields as shown in fig .
[ fig2]c .
@xcite in the case of the fast cooling condition at 0 t , @xmath8 shifts to the higher wavenumber side .
this indicates that the insulating region expands compared with the results obtained under the slow cooling condition . the distribution of @xmath8 values shown in fig .
[ fig3]b also crosses @xmath24 , _
i.e. _ , the insulating state mixes with the metallic ( or sc ) state at the sample surface . with increasing magnetic field strength up to 5 t , @xmath8 shifts to the higher wavenumber side and the insulating region expands in contrast to the constant @xmath8 distribution observed in the slow cooling experiment .
this result is consistent with the electrical resistivity data if the following explanation is accurate . at 0 t
, the sc domains connect to one another resulting in the electrical resistivity dropping to zero even if the afi domains exist .
actually , the percolation of the sc state appears in the top figure of fig .
the coexistence of sc and afi states is also consistent with the ac - susceptibility data that does not show perfect diamagnetism .
@xcite when magnetic fields are applied , the afi domains expand and the sc ( or metallic ) domains disconnect as shown in the middle figure of fig .
[ fig2]b . as a result ,
the electrical resistivity drastically changes to reflect an insulating state .
in fact , the re - entrant sc phase appears at the boundary of the afi
sc transition in @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]cl ( denoted @xmath0-cl hereafter ) @xcite and @xmath6 $ ] @xcite with increasing pressure .
the origin of this observation has been revealed to be a mixture of the sc and afi domains resulting from the pressure dependent electrical resistivity of @xmath0-cl @xcite and from the pressure dependent infrared @xmath16 of @xmath6 $ ] .
@xcite therefore , the s afi transition observed in the electrical resistivity measurement originates from the increase in the size of the afi region in the sample .
the coexistence of sc and afi domains has been studied using the half - filled hubbard model in the limit of infinite dimensions . @xcite this paper pointed out that the magnetic field - induced metal
insulator transition occurs in the vicinity of @xmath28 .
therefore , the origin of the sc afi transition in @xmath11 $ ] might be the predicted metal
insulator transition in which the critical magnetic field of the mott transition is below @xmath29 .
spectra ( a ) and of @xmath8 ( b ) at points @xmath30 , and @xmath31 in fig .
the @xmath16 spectra at points @xmath32 and @xmath33 exhibit the magnetic field - independent insulating ( i ) and metallic ( m ) characteristics , respectively . on the other hand , the spectra at points @xmath34 and @xmath31 change from m ( or sc)@xmath35i and m ( or sc)@xmath35i@xmath35 m , respectively , with increasing magnetic field . , width=264 ] with increasing magnetic field strength up to 10 t , part of the sample was observed to change from an insulating to metallic state as shown in fig [ fig2]b .
however , the metallic ( or sc ) states do not connect to one another , then the macroscopic property must be in an insulating state .
this is consistent with the electrical resistivity data of such samples under magnetic fields as shown in fig .
[ fig2]c .
@xcite to investigate this in detail , the magnetic field dependencies of @xmath16 and @xmath8 at the characteristic four points marked in fig .
[ fig2 ] are shown in fig .
the points @xmath32 and @xmath33 have no magnetic field - dependent insulating and metallic reflectivity spectra , respectively .
this behavior is the same as that in @xmath6 $ ] and @xmath5 $ ] , respectively ( data not shown ) . in comparison with these points , the magnetic field dependencies at points @xmath34 and @xmath31 show characteristic behaviors . at both of these points ,
the metallic @xmath16 at 0 t changes to an insulating state at 5 t , _ i.e. _ , the magnetic field - induced mott transition occurs between 0 and 5 t. in contrast to the insulating @xmath16 at 10 t at point @xmath34 , @xmath16 at point @xmath31 changes to the metallic state again at 10 t , _ i.e. _ , simt occurs at the point @xmath31 .
the area in which the same field dependence as point @xmath31 appears is 6 % of the whole measured region .
phase diagram of @xmath11 $ ] .
the points @xmath30 , and @xmath31 refer to the same points in fig .
the solid circles and diamonds indicate the metallic ( m ) and insulating ( afi ) points , respectively , and the phase boundary is represented by a solid line . below the upper critical field ( @xmath264 t , dashed line ) , the superconducting state ( sc ) is regarded to appear in the metallic state ( m ) .
, width=283 ] the conceptual @xmath36 phase diagram from the magnetic field - dependent @xmath16 expected from fig .
[ fig4]b is shown in fig .
@xmath13 is roughly proportional to @xmath8 based on ref.14 .
the @xmath8 at @xmath33 , @xmath34 , and @xmath31 are similar to one anther at 0 t. however , @xmath8 at 10 t indicates that the @xmath13 values fall in the order of @xmath33 , @xmath31 , @xmath34 , and @xmath32 .
the metallic and insulating characteristics at each point under the magnetic fields are plotted with solid circles and diamonds , respectively , and the boundary is indicated by a solid line , as shown in fig .
the figure indicates that the boundary between the metallic ( or sc ) and insulating states is not linear , but rather has a re - entrant shape .
a similar re - entrant m(sc)i phase boundary has been observed in the @xmath37 phase diagram of @xmath0-cl .
@xcite in one - dimensional bechgaard salts , on the other hand , a transition from metallic to field - induced spin - density wave state has been observed .
@xcite however , to our best knowledge , our result is the first example of the magnetic field - induced simt .
the re - entrant phase diagram originates from the combination of the inhomogenity of the sample and the disorder of the ethylene end groups in the bedt - ttf molecules .
the inhomogenity also affects the width of the distribution of @xmath8 under the slow cooling conditions , as shown in fig .
the fwhm resulting from the gaussian fitting under slow cooling conditions is almost constant over the different magnetic fields ( @xmath38 @xmath17 at 0 t , @xmath39 @xmath17 at 5 t and @xmath40 @xmath17 at 10 t ) . under fast cooling conditions , however , fwhm = @xmath41 @xmath17 at 0 t , @xmath42 @xmath17 at 5 t and @xmath43 @xmath17 at 10 t for a sample possessing not only the inhomogenity but also disorder effects .
the width due to the disorder effect can be evaluated to be @xmath44 @xmath17 at 0 t , @xmath45 @xmath17 at 5 t and @xmath46 @xmath17 at 10 t based on these values under the slow and fast cooling conditions .
the width due to the disorder effect monotonically increases with increasing magnetic field strength .
based on these results , the disorder of the ethylene end groups must affect the electrical characteristics at high magnetic fields .
the simt appearing in the fast cooling experiment is regarded to have primarily originates from the disorder effect on the appropriate condition of the degree of deuteration and the inhomogenity contained in the sample itself .
in conclusion , the infrared reflectivity spectra and the spatial image of the center of the spectral weight of 50 % -deuterated @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]br were measured as functions of magnetic field and cooling rate at 5 k. under fast cooling conditions , the material exhibited a phase coexistence of the metallic ( or superconducting ) and insulating states on the sample surface in spite of the almost metallic phase in the slow cooling condition .
parts of the superconducting state change to the insulating state at 5 t and to the metallic state at 10 t. the superconductor insulator metal transition is concluded to primarily originate from the disorder effect of the ethylene end groups of the bedt - ttf molecules combined with the degree of deuteration and the inhomogenity contained in the sample itself .
the authors would like to thank staff members at bl43ir of spring-8 for their technical supports , and thank f. kagawa for fruitful discussion .
this work was performed at spring-8 with the approval of the japan synchrotron radiation research institute ( proposal nos .
2003b0061 , 2005a0012 ) and was partially supported by grants - in - aid for scientific research ( grant nos .
16038226 , 18340110 ) from mext of japan and by the research foundation for opto - science and technology .
f. kagawa , t. itou , k. miyagawa , and k. kanoda , phys .
. lett . * 93 * , 127001 ( 2004);f .
kagawa , t. itou , k. miyagawa , and k. kanoda , phys .
b * 69 * , 064511 ( 2004);f .
kagawa , k. miyagawa , and k. kanoda , nature ( london ) * 436 * , 534 ( 2005 ) .
a. kawamoto , k. miyagawa , and k. kanoda , phys .
b * 55 * , 14140 ( 1997 ) .
h. taniguchi , a. kawamoto , and k. kanoda , phys .
b * 59 * , 8424 ( 1999 ) .
h. taniguchi , k. kanoda , and a. kawamoto , phys .
b * 67 * , 14510 ( 2003 ) .
k. kanoda , hyperfine interact .
* 104 * , 235 ( 1997 ) ; physica c * 282 - 287 * , 299 ( 1997 ) .
a. kawamoto , h. taniguchi , and k. kanoda , j. am .
chem . soc . * 120 * , 10984 ( 1998 ) .
u. geiser , a. j. schults , h. h. wang , d. m. watkins , d. l. stupka j. m. williams , j. e. schirber , d. l. overmyer , d. jung , j. j. novoa and m. -h .
whangbo , physica c * 174 * , 475 ( 1991 ) .
t. nishi , s. kimura , t. takahashi , t. ito , h.j .
i m , y.s .
kwon , k. miyagawa , h. taniguchi , a. kawamoto and k. kanoda , solid state commun . * 134 * , 189 ( 2005 ) .
t. sasaki , n. yoneyama , n. kobayashi , y. ikemoto and h. kimura , phys .
lett . * 92 * , 227001 ( 2004 ) .
s. kimura , t. nanba , t. sada , m. okuno , m. matsunami , k. shinoda , h. kimura , t. moriwaki , m. yamagata , y. kondo , y. yoshimatsu , t. takahashi , k. fukui , t. kawamoto and t. ishikawa , nucl .
instrum . meth .
a * 467 - 468 * , 893 ( 2001 ) .
|
the magnetic field - induced superconductor insulator metal transition ( simt ) in partially deuterated @xmath0-(bedt - ttf)@xmath1cu[n(cn)@xmath1]br , which is just on the mott boundary , has been observed using the infrared magneto - optical imaging spectroscopy . the infrared reflectivity image on the sample surface revealed that the metallic ( or superconducting ) and insulating phases coexist and they have different magnetic field dependences .
one of the magnetic field dependence is simt that appeared on part of the sample surface .
the simt was concluded to originate from the balance of the inhomogenity in the sample itself and the disorder of the ethylene end groups resulting from fast cooling .
|
the discovery of neutrino oscillations provides the first evidence of physics beyond the standard model ( sm ) , marking the beginning of a new era in particle physics .
thanks to their brilliant confirmation by reactor and accelerator experiments , oscillations constitute the only viable explanation for the observed flavor conversion of `` celestial '' neutrinos @xcite , requiring both neutrino mass and mixing , as expected in theories without conserved lepton number @xcite . even in its simplest @xmath0 unitary form ,
the lepton mixing matrix @xmath1 @xcite differs from the quark mixing matrix in that each @xmath2 factor carries a physical phase : one is the km - analogue and appears in oscillations , while the other two are majorana phases and appear in lepton number ( l)-violating processes .
current experiments are insensitive to cp violation , so that oscillations depend only on the three mixing angles @xmath3 and on the two squared - mass splittings @xmath4 and @xmath5 characterizing solar and atmospheric transitions .
setting @xmath6 in the analysis of atmospheric and accelerator data , and @xmath7 to infinity in the solar and reactor data analysis one obtains the neutrino oscillation parameters , as summarized in figs .
[ fig : global ] and [ fig : th13 ] . fig .
[ fig : global ] gives the allowed values of `` atmospheric '' and `` solar '' oscillation parameters , @xmath8 & @xmath7 , and @xmath9 & @xmath10 , respectively .
the dot , star and diamond in the left panel of fig .
[ fig : global ] indicate the best fit points of atmospheric , minos and global data , respectively .
similarly the `` solar '' oscillation parameters are obtained by combining solar and reactor neutrino data , as shown in the right panel .
the dot , star and diamond indicate the best fit points of solar , kamland and global data , respectively . in both cases
minimization is carried out with respect to the undisplayed parameters .
current neutrino oscillation parameters , from a global analysis of the world s data @xcite .
, title="fig:",height=181 ] current neutrino oscillation parameters , from a global analysis of the world s data @xcite .
, title="fig:",height=181 ] one sees that data from artificial and natural neutrino sources are clearly complementary : reactor and accelerators give the best determination of squared - mass - splittings , while solar and atmospheric data mainly determine mixings .
constraints on @xmath11 from different neutrino oscillation data sets @xcite.,title="fig:",height=181 ] constraints on @xmath11 from different neutrino oscillation data sets @xcite.,title="fig:",height=181 ] the right panel in fig .
[ fig : th13 ] shows how data slightly prefer a nonzero @xmath12 value , though currently not significant , leading to an upper bound at 90%c.l .
( 3@xmath13 ) : + @xmath14 given for 1 dof , while the regions in fig .
[ fig : th13 ] ( left ) correspond to 90% cl for 2 dof .
the confirmation of a non - zero @xmath12 would strongly encourage the search for cp violation in upcoming neutrino oscillation experiments @xcite . note that the small parameter @xmath15 is currently well - determined experimentally as @xmath16 before closing let us note that many effects may distort the `` celestial '' neutrino fluxes reaching our detectors , such as regular @xcite and random @xcite solar magnetic fields .
these would induce spin - flavor precession in the convective zone @xcite as well as density fluctuations deep inside the sun s radiative zone @xcite .
although these can modify the solar neutrino survival probabilities @xcite , they can not have an important impact on the determination of oscillation parameters , thanks to the kamland reactor neutrino spectrum data .
the result is that oscillations remain robust against astrophysical uncertainties , and of all oscillation solutions allowed by solar data @xcite , only the large mixing angle solution survives kamland s measurements @xcite .
often the generation of neutrino mass in gauge theories ( left panel in fig .
[ fig : d-5-nsi ] ) is accompanied by effective sub - weak strength ( @xmath17 ) flavour - changing ( fc ) or non - universal ( nu ) dimension-6 operators , as seen in the right panel .
such non - standard neutrino interactions ( nsi ) are expected in low - scale seesaw schemes , such as the inverse @xcite and the linear @xcite seesaw . in such schemes nsi
would arise from the effectively non - unitary form of the corresponding lepton mixing matrix @xcite@xcite .
relatively sizeable nsi strengths may also be induced in models with radiatively induced neutrino masses @xcite .
current determination of solar neutrino parameters is not yet fully robust against the presence of large nsi , allowing for a new `` dark side '' solution that survives the inclusion of reactor data @xcite .
in contrast , thanks to the large statistics of atmospheric data over a wide energy range , the determination of atmospheric parameters @xmath7 and @xmath18 is fairly robust even in the presence of nsi , at least within the 2neutrino approximation @xcite , a situation likely to improve with future neutrino factories @xcite . however , nsi operators may have dramatic consequences for the sensitivity to @xmath12 at a neutrino factory @xcite and may affect the interpretation of future supernova neutrino data in an important way @xcite .
improved nsi tests will also shed light on the origin of neutrino mass , helping discriminate between high and low - scale schemes .
neutrino oscillations can not probe absolute neutrino masses , for this we need cosmic microwave background and large scale structure observations @xcite , high sensitivity beta decay and ( neutrinoless double beta decay ) studies @xcite .
the observation of neutrino oscillations suggests that light majorana neutrino exchange will induce as illustrated in the left panel of fig .
[ fig : bbox ] .
this nuclear process would hold the key to probe the nature ( dirac versus majorana ) of neutrinos @xcite since , in a gauge theory , it would imply a majorana mass for at least one neutrino @xcite , as illustrated in the middle panel of fig .
[ fig : bbox ] .
.1 cm such `` black - box '' theorem @xcite holds in any `` natural '' gauge theory , though its implications are rather difficult to quantify in general @xcite .
the detection prospects were discussed in @xcite and are summarized in the right panel in fig .
[ fig : bbox ] .
the broad bands are allowed by current neutrino oscillation data @xcite for normal and inverse neutrino mass hierarchy , while the narrow bands correspond to the case of lepton mixing angles fixed to the tri - bimaximal values @xcite .
note that for normal hierarchy there is in general no lower bound on as there can be a destructive interference among the three neutrinos . in contrast , the inverted neutrino mass hierarchy always gives a generic `` lower '' bound for the amplitude . on the other hand ,
quasi - degenerate neutrino models @xcite give the largest possible signal . taking into account state - of - the - art nuclear matrix elements
@xcite one can determine the best current limit , which comes from the heidelberg - moscow experiment , as well as future experimental sensitivities @xcite , summarized in the right panel in fig .
[ fig : bbox ] for gerda , majorana and cuore .
underpinning the origin of neutrino mass remains a challenge despite the tremendous progress we have achieved .
neutrino masses are markedly different from those of charged fermions in the sm .
the latter arise by coupling the two chiral species to the higgs scalar , hence linear in the electroweak symmetry breaking vacuum expectation value ( vev ) @xmath19 of the higgs scalar doublet @xmath20 . by contrast , being electrically neutral , neutrinos may get mass with just one chiral species : in other words , on general grounds they are of majorana type @xcite .
the lowest - dimensional lepton number violating ( lnv ) operator has @xmath21 , namely @xmath22 , where @xmath23 denotes a lepton doublet @xcite , see left panel in fig .
[ fig : d-5-nsi ] .
neutrino mass @xcite and non - standard neutrino interactions ( nsi ) operators @xcite.,title="fig:",height=83 ] neutrino mass @xcite and non - standard neutrino interactions ( nsi ) operators @xcite.,title="fig:",height=83 ] irrespective of their specific origin , the smallness of neutrino masses would come from the fact that they violate lepton number . the big issue is to identify which _ mechanism _ gives rise to the l - violating operator , its associated mass _ scale _ and its
_ flavor structure_. as for its magnitude , it may naturally be suppressed either by a high - scale @xmath24 in its denominator , or may involve a low - mass - scale in its numerator .
it is often argued that gravity breaks global symmetries @xcite .
this would induce the l - violating operators suppressed by the planck scale .
the resulting majorana neutrino masses are too small , hence the need for physics beyond the sm @xcite .
it is usual to assume that this physics lies at a large sub - planck scale , say , associated to unification .
however , @xmath25 may be suppressed by @xmath26 scales , yukawas and/or loop - factors @xcite .
there are three classes of mechanisms : ( i ) tree level , ( ii ) radiative , and ( iii ) hybrid , all of which may have high- or low - scale realizations , suggesting a fair chance that the origin of neutrino mass may be probed at accelerator experiments like the lhc . depending on the nature of spontaneous lepton - number breaking
there may be an extra neutral gauge boson @xcite or a nambu - goldstone boson coupled to neutrinos @xcite .
+ weinberg s dimension-5 operator @xcite may arise from the exchange of heavy fermion states with masses close to the `` unification '' scale .
type - i and iii ( left ) and type - ii ( right ) realizations of the seesaw mechanism.,title="fig : " ] type - i and iii ( left ) and type - ii ( right ) realizations of the seesaw mechanism.,title="fig:",height=68 ] depending on whether these are @xmath27 singlets or triplets the mechanism is called _ type - i _ @xcite , or _
type - iii _
seesaw @xcite , respectively .
neutrino masses may also arise from the exchange of heavy triplet scalars , now called _ type - ii _
seesaw @xcite @xcite , as seen in the right panel in fig .
[ fig : seesaw ] .
the hierarchy of vevs required to account for the small neutrino masses @xmath28 is consistent with the minimization of the scalar potential .
the resulting perturbative diagonalization of the seesaw mass matrix was given in ref .
@xcite in the most general form that may be used in any model . from a phenomenological viewpoint , however , the most basic effective description of _ any _ seesaw is in terms of the sm 21 gauge structure with explicit l - violation @xcite .
the seesaw mechanism is not any particular _ theory _ but rather represents a broad _ language _ in terms of which to phrase neutrino mass model - building and many are its pathways .
indeed , it can be implemented non - minimally with lepton number broken explicitly or spontaneously , over a wide range of energy scales , in a variety of models with different gauge groups and multiplet contents , with or without supersymmetry .
there is no point to attempt giving here a full taxonomy of seesaw schemes .
however one should stress that any model must ultimately reduce to the sm .
hence what is phenomenologically most relevant , for example to describe neutrino oscillation data , is the effective structure of the sm seesaw mixing matrix , given in @xcite .
in addition to the mixing angles characterizing oscillations , the latter includes non - unitarity effects which will become an important topic in the agenda of future studies probing neutrino propagation beyond oscillations .
an attractive class of non - minimal seesaw schemes employs , in addition to the left - handed sm neutrinos @xmath29 , two 21 singlets @xmath30 @xcite ( see also e.g. @xcite ) .
the basic parameter characterizing the violation of lepton - number can be small @xcite and may be calculable from supersymmetric renormalization group evolution effects @xcite . + neutrino masses may be radiatively calculable @xcite , with no need for a large scale . in this case
the suppression comes from small loop - factors and yukawa couplings . the same way as low - scale seesaw schemes
, the states responsible for generating neutrino masses in radiative models may lie at the weak - scale , opening the door to phenomenology at the lhc @xcite .
there is no reasonable doubt that flavor is violated in neutrino propagation @xcite .
current oscillation experiments indicate solar and atmospheric mixing angles which are unexpectedly large when compared to quark mixing angles . to a good approximation
they are given by @xcite , + @xmath31\nonumber\end{aligned}\ ] ] understanding the pattern of lepton mixing angles from first principles constitutes a big challenge to unified theories of flavor where quarks and leptons are related .
many less ambitious schemes have been suggested in order to reproduce the tri - bi - maximal pattern , at least partially , using various discrete flavor symmetry groups containing mu - tau symmetry , e. g. @xcite . in general
one expects the flavor symmetry to be valid at high energy scales .
deviations from tri - bi - maximal ansatz @xcite may be calculable by renormalization group evolution @xcite .
a simple possibility is that , as a result of a given flavor symmetry such as a4 @xcite , neutrino masses unify at high energies @xcite , the same way as gauge couplings do , due to supersymmetry .
such quasi - degenerate neutrino scheme predicts maximal atmospheric angle and vanishing @xmath12 , @xmath32 leaving the solar angle @xmath9 unpredicted , but cabibbo - unsuppressed , @xmath33 if cp is violated @xmath12 becomes arbitrary and the dirac phase is maximal @xcite .
the lower bound on the absolute majorana neutrino mass scale @xmath34 ev ensures that the model will be probed by future cosmological data and searches .
it is natural to expect that , at some level , lepton flavor violation will also show up as transitions involving the charged leptons , since these sit in the same electroweak doublets as neutrinos .
rates for lepton flavour violating processes @xmath35 often lie in the range of sensitivity of coming experiments , providing an independent test .
there are two basic mechanisms : ( i ) neutral heavy lepton exchange @xcite and ( ii ) supersymmetry @xcite .
both exist in supersymmetric seesaw - type schemes of neutrino mass , the interplay of both depends on the seesaw scale , and has been considered in @xcite . barring fine - tunings , high - scale seesaw models require supersymmetry in order to have sizeable lfv rates .
the most interesting feature of these models is that they bring in the possibility of direct in the production of supersymmetric particles . as seen in fig .
[ fig : prodxbr ] this provides the most direct way to probe lfv at the lhc in high - scale seesaw models . [ cols="^,^ " , ] versus the lnv scale for inverse seesaw ( top : red color ) and linear seesaw ( bottom , blue color ) . in both low - scale seesaw models we fix @xmath36 ( continuous line ) , @xmath37 ( dashed line ) and
@xmath38 ( dot - dashed line ) , from @xcite .
the right panel shows typical correlation between mu - e conversion in nuclei and @xmath39 , from @xcite.,title="fig:",width=264,height=151 ] versus the lnv scale for inverse seesaw ( top : red color ) and linear seesaw ( bottom , blue color ) . in both low - scale seesaw models we fix @xmath36 ( continuous line ) , @xmath37 ( dashed line ) and
@xmath38 ( dot - dashed line ) , from @xcite .
the right panel shows typical correlation between mu - e conversion in nuclei and @xmath39 , from @xcite.,title="fig:",width=264,height=151 ] in low - scale seesaw schemes , by contrast , the sizeable admixture of `` right - handed '' ( rh ) neutrinos in the charged current ( rectangular nature of the lepton mixing matrix @xcite ) induces potentially large lfv rates even in the absence of supersymmetry @xcite .
indeed , an important point to stress is that lfv @xcite and cp violation @xcite can occur in the massless neutrino limit , hence their attainable magnitude is unrestricted by the smallness of neutrino masses . in fig .
[ fig : lfv - low - scale ] we display @xmath39 versus the small lnv parameters @xmath40 and @xmath41 for two different low - scale seesaw models , the inverse and the linear seesaw , respectively .
clearly the lfv rates are sizeable in both cases , the different slopes with respect to @xmath40 and @xmath41 follow from the fact that @xmath21 in the first case and @xmath42 , in the second .
similarly @xcite in low - scale seesaw models the nuclear @xmath43 conversion rates lie within planned sensitivities of future experiments such as prism @xcite .
note that models with specific flavor symmetries , such as those in @xcite relate different lfv rates . to conclude we mention that some seesaw schemes , like type - iii with tiny yukawas @xcite or inverse type - iii @xcite , may be directly probed at the lhc by directly producing the tev states with gauge strength @xcite .
in supersymmetric models lepton number can be broken together with the so - called r parity , leading to an intrinsically supersymmetric origin for neutrino masses @xcite .
this may happen spontaneously , driven by a nonzero vev of an 21 singlet sneutrino @xcite , leading to an effective model with bilinear violation of r parity @xcite .
the latter provides the minimal way to break r parity and add neutrino masses to the mssm @xcite .
decay length versus @xmath44 for @xmath45 gev , @xmath46 , @xmath47 , and various @xmath48 values .
the three shaded bands around @xmath49 gev correspond to the variation of the brpv parameters in such a way that the neutrino masses and mixing angles fit the required values within @xmath50 .
the right panel gives the ratio of branching ratios , br@xmath51 over br@xmath52 in terms of the atmospheric angle in bilinear r parity violation @xcite.,title="fig:",width=226,height=132 ] decay length versus @xmath44 for @xmath45 gev , @xmath46 , @xmath47 , and various @xmath48 values .
the three shaded bands around @xmath49 gev correspond to the variation of the brpv parameters in such a way that the neutrino masses and mixing angles fit the required values within @xmath50 .
the right panel gives the ratio of branching ratios , br@xmath51 over br@xmath52 in terms of the atmospheric angle in bilinear r parity violation @xcite.,title="fig:",width=226,height=136 ] one finds that , typically , the atmospheric scale is generated at tree level by neutralino - exchange _ weak - scale seesaw _ , while the solar scale is radiatively induced @xcite . unprotected by any symmetry , the lightest supersymmetric particle ( lsp ) decays .
given the masses indicated by neutrino experiments these decays will happen inside typical detectors @xcite but with a decay path that can be experimentally resolved , leading to a so - called displaced vertex @xcite , see left panel in fig .
[ fig : ntrl ] . more strikingly , lsp decay properties correlate with the neutrino mixing angles . indeed , as seen in the right panel in fig .
[ fig : ntrl ] the lsp decay pattern is predicted by the low - energy measurement of the atmospheric angle @xcite , allowing for a clear test at the lhc @xcite , namely a high - energy redetermination of @xmath8 .
similar correlations hold in variant models based on alternative supersymmetry breaking mechanisms , where other states appear as lsp @xcite .
neutrinos can probe very early epochs in the evolution of the universe , previous to the electroweak phase transition .
for example , the high - scale generation of neutrino masses through the seesaw mechanism may induce the observed baryon asymmetry of the universe , as well as the dark matter , as i now discuss .
+ the observed cosmological matter - antimatter asymmetry in the universe may arise from the c / cp - violating out - of - equilibrium decays of the heavy rh neutrinos present in the seesaw .
these take place before the electroweak phase transition @xcite through the diagrams in the left panel in fig .
[ fig : lep - g ] .
the lepton asymmetry thus produced gets converted , through sphaleron processes , into a baryon asymmetry .
this so - called leptogenesis scenario @xcite is a framework to explain just one number , namely the baryon asymmetry , currently well - determined by wmap @xcite .
it turns out that , as displayed in the right panel in fig .
[ fig : lep - g ] , from ref .
@xcite , after taking into account carefully washout and flavor effects one finds that the required asymmetry can be achieved for a given range of model parameters which fits neutrino masses indicated by neutrino oscillation experiments .
it did not have to be so , _ a priori _ , hence this may be taken as a success .
the figure also shows how the inclusion of flavor effects lowers the minimum value of the lightest rh neutrino mass required for successful leptogenesis .
nevertheless , in order to avoid gravitino overproduction @xcite , which would destroy the standard big bang nucleosynthesis predictions , one also requires an upper bound on the reheat temperature @xmath53 after inflation , incompatible with fig .
[ fig : lep - g ] @xcite .
one way to prevent such _ gravitino crisis _ is to assume enhancement coming from resonant leptogenesis @xcite .
alternatively , there are many ways to go beyond the minimal type - i supersymmetric seesaw @xcite .
for example one may add a small r - parity violating superpotential term @xmath54 , where @xmath55 are rh neutrino supermultiplets @xcite . in the presence of this term
the produced asymmetry can be enhanced . in extended so(10 ) supersymmetric seesaw schemes leptogenesis can occur at relatively low scales , through the decay of a new singlet , as illustrated in the left panel in fig .
[ fig : lg10 ] .
this not only avoids the gravitino crisis but also opens the possibility of detecting the new neutral gauge boson at the lhc @xcite .
the right panel illustrates how a sizeable asymmetry may be achieved just with the leptonic cp violation parameter @xmath56 that characterizes neutrino oscillations .
neutrinos may get mass through the spontaneous breaking of ungauged lepton number .
due to quantum gravity effects the associated goldstone boson - the majoron - is likely to pick up a mass , and play the role of late - decaying dark matter , decaying mainly to neutrinos @xcite .
cosmic microwave background observations place constraints on the majoron lifetime and mass , illustrated in left and middle panels of fig .
[ fig : kev - maj ] .
this decaying dark matter scenario arises in type - ii seesaw models , where the majoron couples to photons through the higgs triplet and may be _ tested _ through the mono - energetic emission line from its sub - dominant decay to two photons , as illustrated in the right panel in fig .
[ fig : kev - maj ] .
neutrino masses may also open new possibilities for `` conventional '' supersymmetric dark matter .
for example , within the inverse seesaw mechanism minimal supergravity is more likely to have a _ sneutrino _ as the lightest superparticle than the conventional neutralino .
such schemes naturally reconcile the small neutrino masses with the correct relic sneutrino dark matter abundance and accessible direct detection rates in nuclear recoil experiments @xcite . + * acknowledgments * + work supported by spanish grants fpa2008 - 00319/fpa , multidark consolider csd2009 - 00064 and prometeo/2009/091 , and by european network unilhc , pitn - ga-2009 - 237920 .
|
a summary of neutrino oscillation results is given along with a discussion of neutrino mass generation mechanisms , including high and low - scale seesaw , with and without supersymmetry , as well as recent attempts to understand flavor .
i argue that if the origin of neutrino masses is intrinsically supersymmetric , it may lead to clear tests at the lhc .
finally , i briefly discuss thermal leptogenesis and dark matter .
|
since many systems in science and engineering are approximately linear , linear inverse problems have attracted great attention in the signal processing community . an input signal @xmath0
is recorded via a linear operator under additive noise : @xmath1 where @xmath2 is an @xmath3 matrix and @xmath4 denotes the noise .
the goal is to estimate @xmath5 from the measurements @xmath6 given knowledge of @xmath2 and a model for the noise @xmath7 .
when @xmath8 , the setup is known as compressed sensing ( cs ) and the estimation problem is commonly referred to as recovery or reconstruction ; by posing a sparsity or compressibility requirement on the signal and using this requirement as a prior during recovery , it is indeed possible to accurately estimate @xmath5 from @xmath6 @xcite . on the other hand , we might need more measurements than the signal length when the signal is dense or the noise is substantial .
wu and verd @xcite have shown that independent and identically distributed ( i.i.d . )
gaussian sensing matrices achieve the same phase - transition threshold as optimal nonlinear encoding , for any discrete continuous mixture .
hence , in cs the acquisition can be designed independently of the particular signal prior through the use of randomized gaussian matrices @xmath2 .
nevertheless , the majority of ( if not all ) existing recovery algorithms require knowledge of the sparsity structure of @xmath5 , i.e. , the choice of a _ sparsifying transform _ @xmath9 that renders a sparse coefficient vector @xmath10 for the signal .
the large majority of recovery algorithms pose a sparsity prior on the signal @xmath5 , e.g. , @xcite .
a second , separate class of bayesian cs recovery algorithms poses a probabilistic prior for the coefficients of @xmath5 in a known transform domain @xcite .
given a probabilistic model , some related message passing approaches learn the parameters of the signal model and achieve the minimum mean squared error ( mmse ) in some settings ; examples include em - gm - amp - mos @xcite , turbogamp @xcite , and amp - mixd @xcite . as a third alternative , complexity - penalized least square methods @xcite can use arbitrary prior information on the signal model and provide analytical guarantees , but are only computationally efficient for specific signal models , such as the independent - entry laplacian model @xcite .
for example , donoho et al .
@xcite relies on kolmogorov complexity , which can not be computed @xcite . as a fourth alternative
, there exist algorithms that can formulate dictionaries that yield sparse representations for the signals of interest when a large amount of training data is available @xcite . when the signal is non - i.i.d .
, existing algorithms require either prior knowledge of the probabilistic model @xcite or the use of training data @xcite . in certain cases , one might not be certain about the structure or statistics of the source prior to recovery .
uncertainty about such structure may result in a sub - optimal choice of the sparsifying transform @xmath9 , yielding a coefficient vector @xmath11 that requires more measurements to achieve reasonable estimation quality ; uncertainty about the statistics of the source will make it difficult to select a prior or model for bayesian algorithms .
thus , it would be desirable to formulate algorithms to estimate @xmath5 that are agnostic to the particular statistics of the signal . therefore , we shift our focus from the standard sparsity or compressibility priors to _ universal _ priors @xcite .
such concepts have been previously leveraged in the kolmogorov sampler universal denoising algorithm @xcite , which minimizes kolmogorov complexity @xcite .
related approaches based on minimum description length ( mdl ) @xcite minimize the complexity of the estimated signal with respect to some class of parametric sources .
unfortunately , mdl can provide a suitable algorithmic recovery framework primarily for parametric sources @xcite .
alternative approaches for non - parametric sources based on kolmogorov complexity are not computable in practice @xcite . to address this computational problem , we confine our attention to stationary ergodic sources and develop an algorithmic framework for universal signal estimation in cs systems , which can be applied to general linear inverse problems where more measurements might be needed .
our framework leverages the fact that for stationary ergodic sources , both the per - symbol empirical entropy and kolmogorov complexity converge asymptotically almost surely to the entropy rate of the source @xcite .
we aim to minimize the empirical entropy ; our minimization is regularized by introducing a log likelihood for the noise model , which is equivalent to the standard least squares under additive white gaussian noise .
other noise distributions are readily supported .
we make several contributions toward our universal cs framework .
first , we apply a specific quantization grid to a maximum _ a posteriori _
( map ) estimator driven by a universal prior , providing a finite - computation universal estimation scheme ; our scheme can also be applied to general linear inverse problems where more measurements might be needed .
second , we propose a recovery algorithm based on markov chain monte carlo ( mcmc ) @xcite to approximate this estimation procedure .
third , we prove that for a sufficiently large number of iterations the output of our mcmc recovery algorithm converges to the correct map estimate .
fourth , we identify computational bottlenecks in the implementation of our mcmc estimator and show approaches to reduce their complexity .
fifth , we develop an adaptive quantization scheme that tailors a set of reproduction levels to minimize the quantization error within the mcmc iterations and that provides an accelerated implementation .
sixth , we propose a framework that adaptively adjusts the cardinality ( or size for short ) of the adaptive quantizer to match the complexity of the input signal , in order to further reduce the quantization error and computation .
we note in passing that averaging over the outputs of different runs of the same signal with the same measurements will yield lower mean squared error ( mse ) for our proposed algorithm . to showcase the potential of our universal estimation approach , fig .
[ fig : m4 ] illustrates recovery results from gaussian measurement matrices for a four - state markov source of length @xmath12 that generates the pattern @xmath13 with 3% errors in state transitions , resulting in the signal switching from @xmath14 to @xmath15 or vice versa either too early or too late ( see section [ sec : numerical ] for details on the simulation setup ) .
note that the reconstruction algorithm does not know that this source is a binary source . while it is well known that sparsity - promoting recovery algorithms @xcite can recover sparse sources from linear measurements , the aforementioned switching source is not sparse in conventional sparsifying bases ( e.g. , fourier , wavelet , and discrete cosine transforms ) , rendering such sparsifying transforms not applicable .
signals generated by this markov source can be sparsified using an averaging analysis matrix @xcite whose diagonal and first three lower sub - diagonals are filled with @xmath15 , and all other entries are @xmath16 ; this transform yields @xmath17 non - zeros in the sparse coefficient vector . however , even if this matrix is known _ a priori _ , existing algorithms based on analysis sparsity @xcite do not perform satisfactorily , yielding mean signal - to - distortion ratios below @xmath18 db ( cf . section [ sec : numerical ] ) .
in contrast , our size- and level - adaptive mcmc ( sla - mcmc , cf .
section [ sec : adaptive ] ) algorithm estimates this source with high fidelity when a moderate number of measurements @xmath19 are available .
we provide more experimental results in section [ sec : numerical ] to show that the performance of mcmc is comparable to and in many cases better than existing algorithms . for different signal - to - noise ratio ( snr ) values ( @xmath12 ) .
existing cs algorithms fail at reconstructing this signal , because it is not sparse . _
[ fig : m4],width=302 ] this paper is organized as follows .
section [ sec : setting ] provides background content .
section [ sec : theory ] overviews map estimation , quantization , and introduces universal map estimation .
section [ sec : mcmc ] formulates an initial mcmc algorithm for universal map estimation , section [ sec : adaptive ] describes several improvements to this initial algorithm , and section [ sec : numerical ] presents experimental results .
we conclude in section [ sec : conclusions ] .
the proof of our main theoretical result appears in the appendix .
consider the noisy measurement setup via a linear operator ( [ eq : def_y ] ) .
the input signal @xmath20 is generated by a stationary ergodic source @xmath21 , and must be estimated from @xmath6 and @xmath2 . _
the distribution @xmath22 that generates @xmath5 is unknown . _
the matrix @xmath23 has i.i.d .
gaussian entries , @xmath24 .
is not dependent on a particular choice for the matrix @xmath2 . ] these moments ensure that the columns of the matrix have unit norm on average . for concrete analysis , we assume that the noise @xmath25 is i.i.d .
gaussian , with mean zero and known variance @xmath26 for simplicity .
we focus on the setting where @xmath27 and the aspect ratio is positive : @xmath28 similar settings have been discussed in the literature @xcite . when @xmath29 , this setup is known as cs ; otherwise , it is a general linear inverse problem setting .
since @xmath5 is generated by an unknown source , we must search for an estimation mechanism that is agnostic to the specific distribution @xmath22 . for a scalar channel with a discrete - valued signal @xmath5 , e.g. ,
@xmath2 is an identity matrix and @xmath30 , donoho proposed the kolmogorov sampler ( ks ) for denoising @xcite , @xmath31 where @xmath32 denotes the kolmogorov complexity of @xmath5 , defined as the length of the shortest input to a turing machine @xcite that generates the output @xmath5 and then halts , , kolmogorov complexity ( kc ) can be approximated using a fine quantizer .
note that the algorithm developed in this paper uses a coarse quantizer and does not rely on kc due to the absence of a feasible method for its computation @xcite ( cf .
section [ sec : adaptive ] ) . ] and @xmath33 controls for the presence of noise .
it can be shown that @xmath32 asymptotically captures the statistics of the stationary ergodic source @xmath34 , and the per - symbol complexity achieves the entropy rate @xmath35 , i.e. , @xmath36 almost surely ( * ? ? ?
* , theorem 7.3.1 ) . noting that universal lossless compression algorithms @xcite achieve the entropy rate for any discrete - valued finite state machine source @xmath34
, we see that these algorithms achieve the per - symbol kolmogorov complexity almost surely .
donoho et al . expanded ks to the linear cs measurement setting @xmath37 but did not consider measurement noise @xcite .
recent papers by jalali and coauthors @xcite , which appeared simultaneously with our work @xcite , provide an analysis of a modified ks suitable for measurements corrupted by noise of bounded magnitude .
inspired by donoho et al .
@xcite , we estimate @xmath5 from noisy measurements @xmath6 using the empirical entropy as a proxy for the kolmogorov complexity ( cf . section [ sec : compressor ] ) .
separate notions of complexity - penalized least squares have also been shown to be well suited for denoising and cs recovery @xcite .
for example , minimum description length ( mdl ) @xcite provides a framework composed of classes of models for which the signal complexity can be defined sharply . in general ,
complexity - penalized least square approaches can yield mdl - flavored cs recovery algorithms that are adaptive to parametric classes of sources @xcite .
an alternative universal denoising approach computes the universal conditional expectation of the signal @xcite .
this section briefly reviews map estimation and then applies it over a quantization grid , where a universal prior is used for the signal . additionally , we provide a conjecture for the mse achieved by our universal map scheme . in this subsection , we assume for exposition purposes that we know the signal statistics @xmath22 .
given the measurements @xmath6 , the map estimator for @xmath5 has the form @xmath38 because @xmath7 is i.i.d .
gaussian with mean zero and known variance @xmath26 , @xmath39 , where @xmath40 and @xmath41 are constants , and @xmath42 denotes the euclidean norm .
norm to an @xmath43 norm and adjust @xmath44 and @xmath45 accordingly . ]
plugging into ( [ eq : map ] ) and taking log likelihoods , we obtain @xmath46 , where @xmath47 denotes the objective function ( risk ) @xmath48 ; our ideal risk would be @xmath49 . instead of performing continuous - valued map estimation ,
we optimize for the map in the discretized domain @xmath50 , with @xmath51 being defined as follows . adapting the approach of baron and weissman @xcite
, we define the set of data - independent reproduction levels for quantizing @xmath5 as @xmath52 where @xmath53 . as @xmath54 increases
, @xmath51 will quantize @xmath5 to a greater resolution .
these reproduction levels simplify the estimation problem from continuous to discrete .
having discussed our reproduction levels in the set @xmath51 , we provide a technical condition on boundedness of the signal .
[ cond : tech1 ] we require that the probability density @xmath22 has bounded support , i.e. , there exists @xmath55 $ ] such that @xmath56 for @xmath57 .
a limitation of the data - independent reproduction level set ( [ eq : def : replevels ] ) is that @xmath51 has infinite cardinality ( or size for short ) .
thanks to condition [ cond : tech1 ] , for each value of @xmath58 there exists a constant @xmath59 such that a finite set of reproduction levels @xmath60 will quantize the range of values @xmath61 to the same accuracy as that of ( [ eq : def : replevels ] ) .
we call @xmath62 the _ reproduction alphabet _ , and each element in it a ( _ reproduction _ ) _ level_. this finite quantizer reduces the complexity of the estimation problem from infinite to combinatorial .
in fact , @xmath63 $ ] under condition [ cond : tech1 ] . therefore , for all @xmath64 and sufficiently large @xmath54 , this set of levels will cover the range @xmath65 $ ] .
the resulting reduction in complexity is due to the structure in @xmath66 and independent of the particular statistics of the source @xmath34 . now that we have set up a quantization grid @xmath67 for @xmath5 , we convert the distribution @xmath22 to a probability mass function ( pmf ) @xmath68 over @xmath67 .
let @xmath69 , and define the pmf @xmath70 as @xmath71 .
then @xmath72 gives the map estimate of @xmath5 over @xmath67 .
as @xmath54 increases , @xmath68 will approximate @xmath22 more closely under ( [ eq : def : replevels2 ] ) .
we now describe a universal estimator for cs over a quantized grid .
consider a prior @xmath73 that might involve kolmogorov complexity @xcite , e.g. , @xmath74 , or mdl complexity with respect to some class of parametric sources @xcite .
we call @xmath73 a _ universal prior _ if it has the fortuitous property that for every stationary ergodic source @xmath34 and fixed @xmath75 , there exists some minimum @xmath76 such that @xmath77 for all @xmath78 and @xmath79 @xcite .
we optimize over an objective function that incorporates @xmath73 and the presence of additive white gaussian noise in the measurements : @xmath80 resulting in corresponds to a lagrangian relaxation of the approach studied in @xcite . ]
our universal map estimator does not require @xmath29 , and @xmath82 can be used in general linear inverse problems .
donoho @xcite showed for the scalar channel @xmath30 that : ( @xmath83 ) the kolmogorov sampler @xmath84 ( [ eq : x_ks ] ) is drawn from the posterior distribution @xmath85 ; and ( @xmath86 ) the mse of this estimate @xmath87 $ ] is no greater than twice the mmse .
based on this result , which requires a large reproduction alphabet , we now present a conjecture on the quality of the estimation @xmath88 .
our conjecture is based on observing that @xmath88 also samples from the posterior distribution ; some experimental evidence to assess this conjecture is presented in figs .
[ fig : ber ] and [ fig : sparsel ] .
[ conj : double_mmse ] assume that @xmath89 is an i.i.d .
gaussian measurement matrix where each entry has mean zero and variance @xmath90 .
suppose that condition [ cond : tech1 ] holds , the aspect ratio @xmath91 in ( [ eq : delta ] ) , and the noise @xmath25 is i.i.d .
zero - mean gaussian with finite variance .
then for all @xmath92 , the mean squared error of the universal map estimator @xmath88 satisfies @xmath93}{n } < \frac{2 e_{x , z,\phi}\left[\|x - e_{x}[x|y,\phi]\|^2\right]}{n } + \epsilon\ ] ] for sufficiently large @xmath54 .
although the results of the previous section are theoretically appealing , a brute force optimization of @xmath82 is computationally intractable .
instead , we propose an algorithmic approach based on mcmc methods @xcite .
our approach is reminiscent of the framework for lossy data compression @xcite .
we propose a universal lossless compression formulation following the conventions of weissman and coauthors @xcite .
we refer to the estimate as @xmath94 in our algorithm .
our goal is to characterize @xmath95 , cf .
( [ eq : psidef ] ) .
although we are inspired by the kolmogorov sampler approach @xcite , kc can not be computed @xcite , and we instead use empirical entropy . for stationary ergodic sources , the empirical entropy converges to the per - symbol entropy rate almost surely @xcite . to define the empirical entropy
, we first define the empirical symbol counts : @xmath96 \triangleq | \
{ i \in [ \q+1,n ] : w_{i-\q}^{i-1}=\alpha , w_i=\beta \
} | , \label{eq : nq}\ ] ] where @xmath97 is the context depth @xcite , @xmath98 , @xmath99 , @xmath100 is the @xmath101 symbol of @xmath94 , and @xmath102 is the string comprising symbols @xmath83 through @xmath103 within @xmath94 .
we now define the order @xmath97 conditional empirical probability for the context @xmath104 as @xmath105 \triangleq \frac { n_\q(w,\alpha)[\beta ] } { \sum_{\beta ' \in \replevels_f } n_\q(w,\alpha)[\beta ' ] } , \ ] ] and the order @xmath97 conditional empirical entropy , @xmath106 \log_2\left ( \mathbb{p}_\q(w,\alpha)[\beta ] \right),\ ] ] where the sum is only over non - zero counts and probabilities .
allowing the context depth @xmath107 to grow slowly with @xmath54 , various universal compression algorithms can achieve the empirical entropy @xmath108 asymptotically @xcite . on the other hand , no compressor can outperform the entropy rate .
additionally , for large @xmath54 , the empirical symbol counts with context depth @xmath97 provide a sufficiently precise characterization of the source statistics .
therefore , @xmath109 provides a concise approximation to the per - symbol coding length of a universal compressor .
having approximated the coding length , we now describe how to optimize our objective function .
we define the energy @xmath110 in an analogous manner to @xmath111 ( [ eq : psidef ] ) , using @xmath112 as our universal coding length : @xmath113 where @xmath114 .
the minimization of this energy is analogous to minimizing @xmath111 .
ideally , our goal is to compute the globally minimum energy solution @xmath115 .
we use a stochastic mcmc relaxation @xcite to achieve the globally minimum solution in the limit of infinite computation . to assist the reader in appreciating how mcmc is used to compute @xmath116 , we include pseudocode for our approach in algorithm [ alg : mcmc ]
. the algorithm , called basic mcmc ( b - mcmc ) , will be used as a building block for our latter algorithms [ alg : mcmcal ] and 3 in section [ sec : adaptive ] .
the initial estimate @xmath94 is obtained by quantizing the _ initial point _
@xmath117 to @xmath118 .
the initial point @xmath119 could be the output of any signal reconstruction algorithm , and because @xmath119 is a preliminary estimate of the signal that does not require high fidelity , we let @xmath120 for simplicity , where @xmath121 denotes transpose . we refer to the processing of a single entry of @xmath94 as an iteration and group the processing of all entries of @xmath94 , randomly permuted , into super - iterations .
the boltzmann pmf is defined as @xmath122 where @xmath123 is inversely related to the temperature in simulated annealing and @xmath124 is a normalization constant .
mcmc samples from the boltzmann pmf ( [ eq : def_boltzmann ] ) using a _ gibbs sampler _ : in each iteration , a single element @xmath125 is generated while the rest of @xmath94 , @xmath126 , remains unchanged .
we denote by @xmath127 the concatenation of the initial portion of the output vector @xmath128 , the symbol @xmath98 , and the latter portion of the output @xmath129 .
the gibbs sampler updates @xmath125 by resampling from the pmf : @xmath130 \right ) } \label{eqn : gibbs},\end{aligned}\ ] ] where @xmath131 is the change in empirical entropy @xmath112 ( [ eq : def : h ] ) when @xmath132 is replaced by @xmath133 , and @xmath134 is the change in @xmath135 when @xmath132 is replaced by @xmath133 . the maximum change in the energy within an iteration of algorithm [ alg : mcmc ]
is then bounded by @xmath136 note that @xmath5 is assumed bounded ( cf
. condition [ cond : tech1 ] ) so that ( [ eq : deltad][eq : deltaq ] ) are bounded as well . in mcmc , the space @xmath137 is analogous to a statistical mechanical system , and at low temperatures the system tends toward low energies .
therefore , during the execution of the algorithm , we set a sequence of decreasing temperatures that takes into account the maximum change given in ( [ eq : deltaq ] ) : @xmath138where @xmath139 is a temperature offset . at low temperatures , i.e. , large @xmath140 ,
a small difference in energy @xmath110 drives a big difference in probability , cf .
( [ eq : def_boltzmann ] ) .
therefore , we begin at a high temperature where the gibbs sampler can freely move around @xmath118 . as the temperature
is reduced , the pmf becomes more sensitive to changes in energy ( [ eq : def_boltzmann ] ) , and the trend toward @xmath94 with lower energy grows stronger . in each iteration , the gibbs sampler modifies @xmath125 in a random manner that resembles heat bath concepts in statistical mechanics .
although mcmc could sink into a local minimum , geman and geman @xcite proved that if we decrease the temperature according to ( [ eq : st ] ) , then the randomness of gibbs sampling will eventually drive mcmc out of the local minimum toward the globally optimal @xmath82 . in order to help b - mcmc approach the global minimum with reasonable runtime
, we will refine b - mcmc in section [ sec : adaptive ] . :
initial estimate @xmath94 , reproduction alphabet @xmath62 , noise variance @xmath26 , number of super iterations
@xmath141 , temperature constant @xmath142 , and context depth @xmath143 compute @xmath144,~\forall~\alpha \in ( \replevels_f)^\q$ ] , @xmath98 @xmath145 draw permutation @xmath146 at random let @xmath147 be component @xmath148 in permutation compute @xmath149 [ sudo : fixedr_deltah ] compute @xmath150 [ sudo : fixedr_deltad ] compute @xmath151 [ sudo : fixedr : fs ] generate @xmath125 using @xmath152 update @xmath144,~\forall~\alpha \in ( \replevels_f)^\q$ ] , @xmath98 [ sudo : fixedr : update ] return approximation @xmath94 of @xmath88 the following theorem is proven in [ ap : th : conv ] , following the framework espoused by jalali and weissman @xcite .
let @xmath34 be a stationary ergodic source that obeys condition [ cond : tech1 ] .
then the outcome @xmath153 of algorithm [ alg : mcmc ] in the limit of an infinite number of super - iterations @xmath141 obeys @xmath154 .
[ th : conv ] theorem [ th : conv ] shows that algorithm [ alg : mcmc ] matches the best - possible performance of the universal map estimator as measured by the objective function @xmath155 , which should yield an mse that is twice the mmse ( cf .
conjecture [ conj : double_mmse ] ) . to gain some insight about the convergence process of mcmc , we focus on a fixed arbitrary sub - optimal sequence @xmath137 .
suppose that at super - iteration @xmath156 the energy for the algorithm s output @xmath110 has converged to the steady state ( see [ ap : th : conv ] for details on convergence ) .
we can then focus on the probability ratio @xmath157 ; @xmath158 because @xmath159 is the global minimum and has the largest boltzmann probability over all @xmath137 , whereas @xmath94 is sub - optimal .
we then consider the same sequence @xmath94 at super - iteration @xmath160 ; the inverse temperature is @xmath161 and the corresponding ratio at super - iteration @xmath160 is ( cf .
( [ eq : def_boltzmann ] ) ) @xmath162 that is , between super - iterations @xmath156 and @xmath160 the probability ratio @xmath163 is also squared , and the gibbs sampler is less likely to generate samples that differ from @xmath159 .
we infer from this argument that the probability concentration of our algorithm around the globally optimal @xmath159 is linear in the number of super - iterations .
studying the pseudocode of algorithm [ alg : mcmc ] , we recognize that lines [ sudo : fixedr_deltah][sudo : fixedr : fs ] must be implemented efficiently , as they run @xmath164 times .
lines [ sudo : fixedr_deltah ] and [ sudo : fixedr_deltad ] are especially challenging . for line [ sudo : fixedr_deltah ] ,
a naive update of @xmath165 has complexity @xmath166 , cf .
( [ eq : def : h ] ) . to address this problem , jalali and weissman @xcite recompute the empirical conditional entropy in @xmath167 time only for the @xmath168 contexts whose
corresponding counts are modified @xcite .
the same approach can be used in line [ sudo : fixedr : update ] , again reducing computation from @xmath166 to @xmath167 .
some straightforward algebra allows us to convert line [ sudo : fixedr_deltad ] to a form that requires aggregate runtime of @xmath169 .
combined with the computation for line [ sudo : fixedr_deltah ] , and since @xmath170 ( because @xmath171 , and @xmath172 ) in practice , the entire runtime of our algorithm is @xmath173 .
the practical value of algorithm [ alg : mcmc ] may be reduced due to its high computational cost , dictated by the number of super - iterations @xmath141 required for convergence to @xmath159 and the large size of the reproduction alphabet .
nonetheless , algorithm [ alg : mcmc ] provides a starting point toward further performance gains of more practical algorithms for computing @xmath159 , which are presented in section [ sec : adaptive ] . furthermore
, our experiments in section [ sec : numerical ] will show that the performance of the algorithm of section [ sec : adaptive ] is comparable to and in many cases better than existing algorithms .
while algorithm [ alg : mcmc ] is a first step toward universal signal estimation in cs , @xmath54 must be large enough to ensure that @xmath62 quantizes a broad enough range of values of @xmath174 finely enough to represent the estimate @xmath159 well . for large @xmath54
, the estimation performance using the reproduction alphabet ( [ eq : def : replevels2 ] ) could suffer from high computational complexity . on the other hand , for small @xmath54 the number of reproduction levels employed is insufficient to obtain acceptable performance .
nevertheless , using an excessive number of levels will slow down the convergence .
therefore , in this section , we explore techniques that tailor the reproduction alphabet adaptively to the signal being observed .
to estimate better with finite @xmath54 , we utilize reproduction levels that are _ adaptive _ instead of the fixed levels in @xmath62 .
to do so , instead of @xmath175 , we optimize over a sequence @xmath176 , where @xmath177 and @xmath178 denotes the size .
the new reproduction alphabet @xmath179 does not directly correspond to real numbers .
instead , there is an adaptive mapping @xmath180 , and the reproduction levels are @xmath181 .
therefore , we call @xmath179 the _ adaptive _ reproduction alphabet . since the mapping @xmath182 is one - to - one , we also refer to @xmath179 as reproduction levels . considering the energy function ( [ eq : mcmc_energy ] ) , we now compute the empirical symbol counts @xmath183 $ ] , order @xmath97 conditional empirical probabilities @xmath184 $ ] , and order @xmath97 conditional empirical entropy @xmath185 using @xmath176 , @xmath186 , and @xmath187 , cf .
( [ eq : nq ] ) , ( [ eq : def : pcond ] ) , and ( [ eq : def : h ] ) .
similarly , we use @xmath188 instead of @xmath189 , where @xmath190 is the straightforward vector extension of @xmath182 .
these modifications yield an adaptive energy function @xmath191 .
we choose @xmath192 to optimize for minimum squared error , @xmath193_m)^2\right],\end{aligned}\ ] ] where @xmath194_m$ ] denotes the @xmath195 entry of the vector @xmath196 .
the optimal mapping depends entirely on @xmath6 , @xmath2 , and @xmath197 . from a coding perspective
, describing @xmath198 requires @xmath199 bits for @xmath197 and @xmath200 bits for @xmath192 to match the resolution of the non - adaptive @xmath62 , with @xmath201 an arbitrary constant @xcite .
the resulting coding length defines our universal prior . *
optimization of reproduction levels : * we now describe the optimization procedure for @xmath192 , which must be computationally efficient .
write @xmath202 where @xmath203 is the entry of @xmath2 at row @xmath204 and column @xmath147 .
for @xmath205 to be minimum , we need zero - valued derivatives @xmath206 where @xmath207 is the indicator function for event @xmath208 .
define the location sets @xmath209 for each @xmath187 , and rewrite the derivatives of @xmath205 , @xmath210 let the per - character sum column values be @xmath211 for each @xmath212 and @xmath213 .
we desire the derivatives to be zero , cf .
( [ eq : derivative2 ] ) : @xmath214 thus , the system of equations must be satisfied , @xmath215 for each @xmath187 .
consider now the right hand side , @xmath216 for each @xmath187 .
the system of equations can be described in matrix form @xmath217 } ^{\omega } \overbrace { \left[\begin{array}{c } \map(\beta_1)\\ \vdots\\ \map(\beta_{|\alphabet| } ) \end{array}\right ] } ^{\map(\z)}= \overbrace { \left[\begin{array}{c } \sum_{m=1}^m y_m \mu_{m\beta_1}\\ \vdots\\ \sum_{m=1}^m y_m \mu_{m\beta_{|\alphabet| } } \end{array } \right ] } ^{\theta}.\end{aligned}\ ] ] note that by writing @xmath218 as a matrix with entries indexed by row @xmath204 and column @xmath219 given by ( [ eq : mu_def ] ) , we can write @xmath220 as a gram matrix , @xmath221 , and we also have @xmath222 , cf .
( [ eq : system_equation ] ) .
the optimal @xmath182 can be computed as a @xmath223 vector @xmath224 if @xmath225 is invertible .
we note in passing that numerical stability can be improved by regularizing @xmath220 .
note also that @xmath226 which can be computed in @xmath227 time instead of @xmath228 . *
computational complexity : * pseudocode for level - adaptive mcmc ( l - mcmc ) appears in algorithm [ alg : mcmcal ] , which resembles algorithm [ alg : mcmc ] .
the initial mapping @xmath182 is inherited from a quantization of the initial point @xmath119 , @xmath229 ( @xmath139 takes different values in section [ sec : adaptive_size ] ) , and other minor differences between b - mcmc and l - mcmc appear in lines marked by asterisks .
we discuss computational requirements for each line of the pseudocode that is run within the inner loop .
* line [ algo2:deltah ] can be computed in @xmath230 time ( see discussion of line [ sudo : fixedr_deltah ] of b - mcmc in section [ sec : b - mcmc_complexity ] ) .
* line [ algo2:mu ] updates @xmath231 for @xmath232 in @xmath233 time .
* line [ algo2:omega ] updates @xmath220 .
because we only need to update @xmath234 columns and @xmath234 rows , each such column and row contains @xmath235 entries , and each entry is a sum over @xmath233 terms , we need @xmath227 time .
* line [ algo2:map_opt ] requires inverting @xmath220 in @xmath236 time .
* line [ algo2:ell2 ] requires @xmath227 time , cf .
( [ eqn : compute_ell2 ] ) .
* line [ algo2:distribution ] requires @xmath235 time . in practice
we typically have @xmath237 , and so the aggregate complexity is @xmath238 , which is greater than the computational complexity of algorithm [ alg : mcmc ] by a factor of @xmath235 . :
initial mapping @xmath182 , sequence @xmath197 , adaptive alphabet @xmath179 , noise variance @xmath26 , number of super - iterations @xmath141 , temperature constant @xmath142 , context depth @xmath143 , and temperature offset @xmath139 compute @xmath183,~\forall~\alpha \in \z^\q$ ] , @xmath187 initialize @xmath220[algo2:init_omega ] @xmath239 draw permutation @xmath146 at random let @xmath147 be component @xmath148 in permutation compute @xmath240 [ algo2:deltah ] compute @xmath241 [ algo2:mu ] update @xmath220 [ algo2:omega ] compute @xmath192 [ algo2:map_opt ] compute @xmath242 [ algo2:ell2 ] compute @xmath243 [ algo2:distribution ] @xmath244 [ algo2:save_sequence ] generate @xmath245 using @xmath246 update @xmath247 $ ] at @xmath168 relevant locations update @xmath248 , @xmath249[algo2:update_mu ] update @xmath220 [ algo2:update_omega ] : return approximation @xmath190 of @xmath250 , and temperature offset @xmath251 while algorithm [ alg : mcmcal ] adaptively maps @xmath197 to @xmath252 , the signal estimation quality heavily depends on @xmath253 . denote the true alphabet of the signal by @xmath254 ; if the signal is continuous - valued , then @xmath255 is infinite .
ideally we want to employ as many levels as the runtime allows for continuous - valued signals , whereas for discrete - valued signals we want @xmath256 . inspired by this observation , we propose to begin with some initial @xmath253 , and then adaptively adjust @xmath253 hoping to match @xmath255 .
hence , we propose the size- and level - adaptive mcmc algorithm ( algorithm 3 ) , which invokes l - mcmc ( algorithm [ alg : mcmcal ] ) several times .
* three basic procedures : * in order to describe the size- and level - adaptive mcmc ( sla - mcmc ) algorithm in detail , we introduce three alphabet adaptation procedures as follows .
* _ merge _ : first , find the closest adjacent levels @xmath257 . create a new level @xmath258 and add it to @xmath179 .
let @xmath259 .
replace @xmath260 by @xmath258 whenever @xmath261 .
next , remove @xmath262 and @xmath263 from @xmath179 .
* _ add - out _ : define the range @xmath264 @xmath265 $ ] , and @xmath266 . add a _ lower _ level @xmath258 and/or _ upper level _
@xmath267 to @xmath179 with @xmath268 .
note that @xmath269 , i.e. , the new levels are empty . *
_ add - in _ : first , find the most distant adjacent levels , @xmath262 and @xmath263 . then , add a level @xmath258 to @xmath179 with @xmath259 . for @xmath270 s.t .
@xmath271 , replace @xmath260 by @xmath258 with probability @xmath272 , where @xmath273 is given in ( [ eq : def_boltzmann ] ) ; for @xmath270 s.t .
@xmath274 , replace @xmath260 by @xmath258 with probability @xmath275 . note that @xmath276 is typically non - zero , i.e. , @xmath258 tends not to be empty .
we call the process of running one of these procedures followed by running l - mcmc a _ round_. * size- and level - adaptive mcmc : * sla - mcmc is conceptually illustrated in the flowchart in fig .
[ fig : sla - mcmc ] .
it has four stages , and in each stage we will run l - mcmc for several super - iterations ; we denote the execution of l - mcmc for @xmath141 super - iterations by l(@xmath141 ) .
the parameters @xmath277 , and @xmath278 are the number of super - iterations used in stages 1 through 4 , respectively .
the choice of these parameters reflects a trade - off between runtime and estimation quality . in stage 1 ,
sla - mcmc uses a fixed - size adaptive reproduction alphabet @xmath179 to tentatively estimate the signal . the initial point of stage 1
is obtained in the same way as l - mcmc .
after stage 1 , the initial point and temperature offset for each instance of l - mcmc correspond to the respective outputs of the previous instance of l - mcmc .
if the source is discrete - valued and @xmath279 in stage 1 , then multiple levels in the output @xmath179 of stage 1 may correspond to a single level in @xmath280 . to alleviate this problem , in stage 2
we merge levels closer than @xmath281 , where @xmath282 is a parameter .
however , @xmath253 might still be larger than needed ; hence in stage 3 we tentatively merge the closest adjacent levels . the criterion
@xmath283 evaluates whether the current objective function is lower ( better ) than in the previous round ; we do not leave stage 3 until @xmath283 is violated .
note that if @xmath284 ( this always holds for continuous - valued signals ) , then ideally sla - mcmc should not merge any levels in stage 3 , because the objective function would increase if we merge any levels .
define the outlier set @xmath285 . under condition [ cond :
tech1 ] , @xmath286 might be small or even empty . when @xmath286 is small , l - mcmc might not assign levels to represent the entries of @xmath286 . to make sla - mcmc more robust to outliers , in stage 4a we add empty levels outside the range @xmath287 and then allow l - mcmc to change entries of @xmath197 to the new levels during gibbs sampling ; we call this _ populating _ the new levels .
if a newly added outside level is not populated , then we remove it from @xmath179 .
seeing that the optimal mapping @xmath192 in l - mcmc tends not to map symbols to levels with low population , we consider a criterion @xmath288 where we will will add an outside upper ( lower ) level if the population of the current upper ( lower ) level is smaller than @xmath289 , where @xmath290 is a parameter .
that is , the criterion @xmath288 is violated if both populations of the current upper and lower levels are sufficient ( at least @xmath289 ) ; in this case we do not need to add outside levels because @xmath192 will map some of the current levels to represent the entries in @xmath286 .
the criterion @xmath291 is violated if all levels added outside are not populated by the end of the round .
sla - mcmc keeps adding levels outside @xmath287 until it is wide enough to cover most of the entries of @xmath5 .
next , sla - mcmc considers adding levels inside @xmath287 ( stage 4b ) .
if the signal is discrete - valued , this stage should stop when @xmath256 .
else , for continuous - valued signals sla - mcmc can add levels until the runtime expires . in practice ,
sla - mcmc runs l - mcmc at most a constant number of times , and the computational complexity is in the same order of l - mcmc , i.e. , @xmath238 . on the other hand ,
sla - mcmc allows varying @xmath253 , which often improves the estimation quality .
donoho proved for the scalar channel setting that @xmath84 is sampled from the posterior @xmath85 @xcite .
seeing that the gibbs sampler used by mcmc ( cf .
section [ sec : b - mcmc ] ) generates random samples , and the outputs of our algorithm will be different if its random number generator is initialized with different _ random seeds _ , we speculate that running sla - mcmc several times will also yield independent samples from the posterior , where we note that the runtime grows linearly in the number of times that we run sla - mcmc . by mixing
( averaging over ) several outputs of sla - mcmc , we obtain @xmath292 , which may have lower squared error with respect to the true @xmath5 than the average squared error obtained by a single sla - mcmc output .
numerical results suggest that mixing indeed reduces the mse ( cf .
[ fig : munif_algos ] ) ; this observation suggests that mixing the outputs of multiple algorithms , including running a random reconstruction algorithm several times , may reduce the squared error .
in this section , we demonstrate that sla - mcmc is comparable and in many cases better than existing algorithms in reconstruction quality , and that sla - mcmc is applicable when @xmath293 . additionally , some numerical evidence is provided to justify conjecture [ conj : double_mmse ] in section [ sec : conjecture ] .
then , the advantage of sla - mcmc in estimating low - complexity signals is demonstrated .
finally , we compare b - mcmc , l - mcmc , and sla - mcmc performance .
we implemented sla - mcmc in matlab and tested it using several stationary ergodic sources .
except when noted , for each source , signals @xmath5 of length @xmath12 were generated .
each such @xmath5 was multiplied by a gaussian random matrix @xmath2 with normalized columns and corrupted by i.i.d .
gaussian measurement noise @xmath7 .
except when noted , the number of measurements @xmath19 varied between 2000 and 7000 .
the noise variance @xmath26 was selected to ensure that the signal - to - noise ratio ( snr ) was @xmath18 or @xmath294 db ; snr was defined as @xmath295)/(m\sigma_z^2)\right]$ ] .
we set the context depth @xmath296 , the number of super - iterations in different stages of sla - mcmc @xmath297 and @xmath298 , the maximum total number of super - iterations to be @xmath299 , the initial number of levels @xmath300 , and the tuning parameter from section [ sec : adaptive_size ] @xmath301 .
sla - mcmc was not given the true alphabet @xmath280 for any of the sources presented in this paper ; our expectation is that it should adaptively adjust @xmath253 to match @xmath255 .
the final estimate @xmath292 of each signal was obtained by averaging over the outputs @xmath302 of @xmath18 runs of sla - mcmc , where in each run we initialized the random number generator with another random seed , cf .
section [ sec : mix ] .
these choices of parameters seemed to provide a reasonable compromise between runtime and estimation quality .
we chose our performance metric as the mean signal - to - distortion ratio ( msdr ) defined as @xmath303/\mbox{mse}\right).\ ] ] for each @xmath19 and snr , the mse was obtained after averaging over the squared errors of @xmath292 for 50 draws of @xmath5 , @xmath2 , and @xmath7 .
we compared the performance of sla - mcmc to that of ( _ i _ ) compressive sensing matching pursuit ( cosamp ) @xcite , a greedy method ; ( _ ii _ ) gradient projection for sparse reconstruction ( gpsr ) @xcite , an optimization - based method ; and ( _ iii _ ) message passing approaches ( for each source , we chose best - matched algorithms between em - gm - amp - mos ( egam for short ) @xcite and turbogamp ( tg for short ) @xcite ) .
typical runtimes are @xmath304 hour ( for continuous - valued signals ) and @xmath305 minutes ( discrete - valued ) per random seed for sla - mcmc , @xmath306 minutes for egam @xcite and tg @xcite , and @xmath294 minutes for cosamp @xcite and gpsr @xcite . among these baseline algorithms designed for i.i.d .
signals , gpsr @xcite and egam @xcite only need @xmath6 and @xmath2 , and cosamp @xcite also needs the number of non - zeros in @xmath5 . only tg @xcite is designed for non - i.i.d .
signals ; however , it must be aware of the probabilistic model of the source .
finally , gpsr @xcite performance was similar to that of cosamp @xcite for all sources considered in this section , and thus is not plotted . as a function of the number of gaussian random measurements @xmath19 for different snr values ( @xmath12 ) .
_ [ fig : ber],width=302 ] ) distribution as a function of the number of gaussian random measurements @xmath19 for different snr values ( @xmath307 ) .
_ [ fig : densemrad],width=302 ] * bernoulli source : * we first present results for an i.i.d .
bernoulli source .
the bernoulli source followed the distribution @xmath308 , where @xmath309 is the dirac delta function .
note that sla - mcmc did not know @xmath310 and had to estimate it on the fly .
we chose egam @xcite for message passing algorithms because it fits the signal with gaussian mixtures ( gm ) , which can accurately characterize signals from an i.i.d .
bernoulli source .
the resulting msdrs for sla - mcmc , egam @xcite , and cosamp @xcite are plotted in fig .
[ fig : ber ] .
we can see that when @xmath311 , egam @xcite approaches the mmse @xcite performance for low to medium @xmath19 ; although sla - mcmc is often worse than egam @xcite , it is within @xmath312 db of the mmse performance .
this observation that sla - mcmc approaches the mmse for @xmath311 partially substantiates conjecture [ conj : double_mmse ] in section [ sec : conjecture ] .
when @xmath313 , sla - mcmc is comparable to egam @xcite when @xmath314 .
cosamp @xcite has worse msdr . for different snr values ( @xmath12 ) .
_ [ fig : sparsel],width=302 ] $ ] as a function of the number of gaussian random measurements @xmath19 for different snr values ( @xmath12 ) . _
[ fig : munif],width=302 ] * dense markov - rademacher source : * considering that most algorithms are designed for i.i.d .
sources , we now illustrate the performance of sla - mcmc on non - i.i.d .
sources by simulating a dense markov - rademacher ( mrad for short ) source .
the non - zero entries of the dense mrad signal were generated by a two - state markov state machine ( non - zero and zero states ) .
the transition from zero to non - zero state for adjacent entries had probability @xmath315 , while the transition from non - zero to zero state for adjacent entries had probability @xmath316 ; these parameters yielded @xmath317 non - zero entries on average .
the non - zeros were drawn from a rademacher distribution , which took values @xmath318 with equal probability . with such denser signals , we may need to take more measurements and/or require higher snrs to achieve similar performance to previous examples .
seeing that the runtime of cosamp @xcite is cubic in @xmath19 and the number of non - zeros in the signal , we only generated dense mrad signals with length @xmath307 to save runtime .
the number of measurements varied from @xmath319 to @xmath320 , with @xmath321 .
although tg @xcite does not provide an option that accurately characterize the mrad source , we still chose to compare against its performance because it is applicable to non - i.i.d . signals .
the msdrs for sla - mcmc , tg @xcite , and cosamp @xcite are plotted in fig .
[ fig : densemrad ] .
we can see that cosamp @xcite has poor performance because it does not consider the source memory ; although tg @xcite is designed for non - i.i.d .
sources , it is nonetheless outperformed by sla - mcmc .
this example shows that sla - mcmc reconstructs non - i.i.d .
signals well and is applicable to general linear inverse problems .
we now discuss the performance of sla - mcmc in estimating continuous sources .
* sparse laplace ( i.i.d . ) source : * for unbounded continuous - valued signals , which do not adhere to condition [ cond : tech1 ] , we simulated an i.i.d .
sparse laplace source following the distribution @xmath322 , where @xmath323 denotes a laplacian distribution with mean zero and variance one .
we chose egam @xcite for message passing algorithms because it fits the signal with gm , which can accurately characterize signals from an i.i.d .
sparse laplace source .
the msdrs for sla - mcmc , egam @xcite , and cosamp @xcite are plotted in fig .
[ fig : sparsel ] .
we can see that egam @xcite approaches the mmse @xcite performance in all settings ; sla - mcmc outperforms cosamp @xcite , while it is approximately @xmath324 db worse than the mmse .
recall from conjecture [ conj : double_mmse ] that we expect to achieve twice the mmse , which is approximately @xmath312 db below the signal - to - distortion ratio of mmse , and thus sla - mcmc performance is reasonable .
this example of sla - mcmc performance approaching the mmse further substantiates conjecture [ conj : double_mmse ] .
* markov - uniform source : * for bounded continuous - valued signals , which adhere to condition [ cond : tech1 ] , we simulated a markov - uniform ( munif for short ) source , whose non - zero entries were generated by a two - state markov state machine ( non - zero and zero states ) with @xmath325 and @xmath316 ; these parameters yielded @xmath326 non - zeros entries on average .
the non - zero entries were drawn from a uniform distribution between @xmath16 and @xmath304 .
we chose tg with markov support and gm model options @xcite for message passing algorithms .
we plot the resulting msdrs for sla - mcmc , tg @xcite , and cosamp @xcite in fig .
[ fig : munif ] .
we can see that the cosamp @xcite lags behind in msdr .
the sla - mcmc curve is close to that of tg @xcite when @xmath313 , and it is slightly better than tg @xcite when @xmath311 . when the signal model is known , the message passing approaches egam @xcite and tg @xcite achieve quite low mse s , because they can get close to the bayesian mmse .
sometimes the model is only known imprecisely , and sla - mcmc can improve over message passing ; for example , it is better than tg @xcite in estimating munif signals ( fig .
[ fig : munif ] ) , because tg @xcite approximates the uniformly distributed non - zeros by gm .
when the source is continuous ( figs . [ fig : sparsel ] and [ fig : munif ] ) , sla - mcmc might be worse than the existing message passing approaches ( egam @xcite and tg @xcite ) .
one reason for the under - performance of sla - mcmc is the @xmath312 db gap of conjecture [ conj : double_mmse ] .
the second reason is that sla - mcmc can only assign finitely many levels to approximate continuous - valued signals , leading to under - representation of the signal .
however , when it comes to discrete - valued signals that have finite size alphabets ( figs .
[ fig : ber ] and [ fig : densemrad ] ) , sla - mcmc is comparable to and in many cases better than existing algorithms .
sla - mcmc promotes low complexity due to the complexity - penalized term in the objective function ( [ eq : mcmc_energy ] ) .
hence , it tends to perform well for signals with low complexity such as the signals in figs .
[ fig : ber ] and [ fig : densemrad ] ( note that the bernoulli signal is sparse while the mrad signal is denser ) . in this subsection
, we simulated a non - sparse low - complexity signal .
we show that complexity - penalized approaches such as sla - mcmc might estimate low - complexity signals well . as a function of the number of gaussian random measurements @xmath19 for different snr values ( @xmath12 ) . _
[ fig : munif_algos],width=302 ] * four - state markov source : * to evaluate the performance of sla - mcmc for discrete - valued non - i.i.d . and non - sparse signals
, we examined a four - state markov source ( markov4 for short ) , which has been discussed in section [ sec : intro ] and fig .
[ fig : m4 ] .
we tried conventional sparsifying transforms , but none of them yielded coefficients that were reasonably sparse .
as pointed out in section [ sec : intro ] , there exists a special transform that yields a sparse representation of the markov4 signals , but using specialized transforms requires prior knowledge of the source .
despite knowing the sparsifying matrix , all our baseline algorithms performed poorly ( yielding msdrs below @xmath18 db ) .
thus , we did not include the results for these baseline algorithms in fig .
[ fig : m4 ] .
on the other hand , markov4 signals have low complexity in the time domain , and hence , sla - mcmc successfully reconstructed markov4 signals with reasonable quality even when @xmath19 was relatively small .
this markov4 source highlights the special advantage of our approach in reconstructing low - complexity signals .
we compare the performance of b - mcmc , l - mcmc , and sla - mcmc with different numbers of seeds ( cf .
section [ sec : mix ] ) by examining the munif source ( cf .
section [ sec : bvsub ] ) .
we ran b - mcmc with the fixed uniform alphabet @xmath62 in ( [ eq : def : replevels2 ] ) with @xmath327 levels .
l - mcmc was initialized in the same way as stage 1 of sla - mcmc .
b - mcmc and l - mcmc ran for @xmath328 super - iterations before outputting the estimates ; this number of super - iterations was sufficient because it was greater than @xmath329 in stage 1 of sla - mcmc .
the results are plotted in fig .
[ fig : munif_algos ] .
b - mcmc did not perform well given the @xmath62 in ( [ eq : def : replevels2 ] ) and is not plotted .
we can see that sla - mcmc outperforms l - mcmc . averaging over more seeds provides an increase of @xmath304 db in msdr .
db . ] finally , we tried a good " reproduction alphabet in b - mcmc , @xmath330 , and the results were close to those of sla - mcmc . indeed ,
b - mcmc is quite sensitive to the reproduction alphabet , and stages @xmath324@xmath331 of sla - mcmc find a good set of levels .
example output levels @xmath181 of sla - mcmc were : @xmath332 for bernoulli signals , @xmath333 for dense mrad signals , @xmath334 levels spread in the range @xmath335 $ ] for i.i.d .
sparse laplace signals , @xmath336 levels spread in the range @xmath337 $ ] for munif signals , and @xmath338 for markov4 signals ; we can see that sla - mcmc adaptively adjusted @xmath253 to match @xmath255 so that these levels represented each signal well .
also , we can see from figs .
[ fig : ber][fig : sparsel ] that sla - mcmc did not perform well in the low measurements and high snr setting , which was due to mismatch between @xmath253 and @xmath255 .
this paper provides universal algorithms for signal estimation from linear measurements . here , universality denotes the property that the algorithm need not be informed of the probability distribution for the recorded signal prior to acquisition ; rather , the algorithm simultaneously builds estimates both of the observed signal and its distribution .
inspired by the kolmogorov sampler @xcite and motivated by the need for a computationally tractable framework , our contribution focuses on stationary ergodic signal sources and relies on a map estimation algorithm .
the algorithm is then implemented via a mcmc formulation that is proven to be convergent in the limit of infinite computation .
we reduce the computation and improve the estimation quality of the proposed algorithm by adapting the reproduction alphabet to match the complexity of the input signal .
our experiments have shown that the performance of the proposed algorithm is comparable to and in many cases better than existing algorithms , particularly for low - complexity sources that do not exhibit standard sparsity or compressibility . as we were finishing this paper , jalali and poor @xcite
have independently shown that our formulation ( [ eq : mcmc_energy ] ) also provides an implementable version of rnyi entropy minimization .
their theoretical findings further motivate our proposed universal mcmc formulation .
our proof mimics a very similar proof presented in @xcite for lossy source coding ; we include all details for completeness .
the proof technique relies on mathematical properties of non - homogeneous ( e.g. , time - varying ) markov chains ( mcs ) @xcite . through the proof
, @xmath339 denotes the state space of the mc of codewords generated by algorithm [ alg : mcmc ] , with size @xmath340 .
we define a stochastic transition matrix @xmath341 from @xmath342 to itself given by the boltzmann distribution for super - iteration @xmath156 in algorithm [ alg : mcmc ] .
similarly , @xmath343 defines the stable - state distribution on @xmath342 for @xmath341 , satisfying @xmath344 .
@xcite _ dobrushin s ergodic coefficient _ of a mc transition matrix @xmath345 is denoted by @xmath346 and defined as @xmath347 , where @xmath348 denotes row @xmath83 of @xmath345 . from the definition , @xmath349 .
moreover , the ergodic coefficient can be rewritten as @xmath350 where @xmath351 denotes the entry of @xmath345 at row @xmath83 and column @xmath103 .
we group the product of transition matrices across super - iterations as @xmath352 .
there are two common characterizations for the stable - state behavior of a non - homogeneous mc .
@xcite a non - homogeneous mc is called _ weakly ergodic _ if for any distributions @xmath353 and @xmath354 over the state space @xmath342 , and any @xmath355 , @xmath356 , where @xmath357 denotes the @xmath43 norm .
similarly , a non - homogeneous mc is called _ strongly ergodic _ if there exists a distribution @xmath358 over the state space @xmath342 such that for any distribution @xmath353 over @xmath342 , and any @xmath355 , @xmath359 .
we will use the following two theorems from @xcite in our proof .
@xcite a mc is weakly ergodic if and only if there exists a sequence of integers @xmath360 such that @xmath361 .
[ th : block ] @xcite let a mc be weakly ergodic .
assume that there exists a sequence of probability distributions @xmath362 on the state space @xmath342 such that @xmath344 .
then the mc is strongly ergodic if @xmath363 . [ th : weak ] the rest of proof is structured as follows .
first , we show that the sequence of stable - state distributions for the mc used by algorithm [ alg : mcmc ] converges to a uniform distribution over the set of sequences that minimize the energy function as the iteration count @xmath156 increases .
then , we show using theorems [ th : block ] and [ th : weak ] that the non - homogeneous mc used in algorithm [ alg : mcmc ] is strongly ergodic , which by the definition of strong ergodicity implies that algorithm [ alg : mcmc ] always converges to the stable distribution found above .
this implies that the outcome of algorithm [ alg : mcmc ] converges to a minimum - energy solution as @xmath364 , completing the proof of theorem [ th : conv ] .
we therefore begin by finding the stable - state distribution for the non - homogeneous mc used by algorithm [ alg : mcmc ] . at each super - iteration @xmath156 ,
the distribution defined as @xmath365 satisfies @xmath366 , cf .
( [ eqn : gibbs ] ) .
we can show that the distribution @xmath343 converges to a uniform distribution over the set of sequences that minimize the energy function , i.e. , @xmath367 where @xmath368 . to show ( [ eq : stable ] )
, we will show that @xmath369 is increasing for @xmath370 and eventually decreasing for @xmath371 . since for @xmath370 and @xmath372 we have @xmath373 , for @xmath374 we have @xmath375 which together with ( [ eq : pidist ] ) implies @xmath376 . on the other hand ,
if @xmath371 , then we obtain @xmath377^{-1}. \label{eq : hlower}\end{aligned}\ ] ] for sufficiently large @xmath140 , the denominator of ( [ eq : hlower ] ) is dominated by the second term , which increases when @xmath140 increases , and therefore @xmath369 decreases for @xmath371 as @xmath156 increases . finally , since all sequences
@xmath370 have the same energy @xmath110 , it follows that the distribution is uniform over the symbols in @xmath378 . having shown convergence of the non - homogenous mc s stable - state distributions
, we now show that the non - homogeneous mc is strongly ergodic .
the transition matrix @xmath341 of the mc at iteration @xmath156 depends on the temperature @xmath140 in ( [ eq : st ] ) used within algorithm [ alg : mcmc ] .
we first show that the mc used in algorithm [ alg : mcmc ] is weakly ergodic via theorem [ th : block ] ; the proof of the following lemma is given in [ app : lemm : ergcoefbound ] .
the ergodic coefficient of @xmath341 for any @xmath379 is upper bounded by @xmath380 , where @xmath381 is defined in ( [ eq : deltaq ] ) .
[ lemm : ergcoefbound ] we note in passing that condition [ cond : tech1 ] ensures that @xmath381 is finite .
using lemma [ lemm : ergcoefbound ] and ( [ eq : st ] ) , we can evaluate the sum given in theorem [ th : block ] as @xmath382 and so the non - homogeneous mc defined by @xmath383 is weakly ergodic .
now we use theorem [ th : weak ] to show that the mc is strongly ergodic by proving that @xmath384 .
since we know from earlier in the proof that @xmath369 is increasing for @xmath370 and eventually decreasing for @xmath371 , there exists a @xmath385 such that for any @xmath386 , we have @xmath387 since the right hand side does not depend on @xmath388 , we have that @xmath389 .
this implies that the non - homogeneous mc used by algorithm [ alg : mcmc ] is strongly ergodic , and thus completes the proof of theorem [ th : conv ] .
let @xmath390 be two arbitrary sequences in @xmath342 .
the probability of transitioning from a given state to a neighboring state in an iteration within iteration @xmath148 of super - iteration @xmath156 of algorithm [ alg : mcmc ] is given by ( [ eqn : gibbs ] ) , and can be rewritten as @xmath391 where @xmath392 . therefore , the smallest probability of transition from @xmath393 to @xmath394 within super - iteration @xmath156 of algorithm [ alg : mcmc ] is bounded by @xmath395 using the alternative definition of the ergodic coefficient ( [ eq : ergcoef ] ) , @xmath396 proving the lemma .
preliminary conversations with deanna needell and tsachy weissman framed our thinking about universal compressed sensing .
phil schniter was instrumental in formulating the proposed framework and shepherding our progress through detailed conversations , feedback on our drafts , and probing questions .
gary howell provided invaluable guidance on using north carolina state university s high performance computing resources .
final thanks to jin tan , yanting ma , and nikhil krishnan for thoroughly proofreading our manuscript .
m. zhou , h. chen , j. paisley , l. ren , l. li , z. xing , d. dunson , g. sapiro , and l. carin , `` nonparametric bayesian dictionary learning for analysis of noisy and incomplete images , '' , vol .
1 , pp . 130144 , jan .
|
we study the compressed sensing ( cs ) signal estimation problem where an input signal is measured via a linear matrix multiplication under additive noise . while this setup usually assumes sparsity or compressibility in the input signal during recovery ,
the signal structure that can be leveraged is often not known _ a priori_. in this paper , we consider _ universal _ cs recovery , where the statistics of a stationary ergodic signal source are estimated simultaneously with the signal itself .
inspired by kolmogorov complexity and minimum description length , we focus on a maximum _ a posteriori _ ( map ) estimation framework that leverages universal priors to match the complexity of the source .
our framework can also be applied to general linear inverse problems where more measurements than in cs might be needed .
we provide theoretical results that support the algorithmic feasibility of universal map estimation using a markov chain monte carlo implementation , which is computationally challenging .
we incorporate some techniques to accelerate the algorithm while providing comparable and in many cases better reconstruction quality than existing algorithms .
experimental results show the promise of universality in cs , particularly for low - complexity sources that do not exhibit standard sparsity or compressibility .
compressed sensing , map estimation , markov chain monte carlo , universal algorithms .
|
quandle homology @xcite was defined from rack homology @xcite as the quotient by a subcomplex corresponding to the idempotency , for invariance under the type i reidemeister move .
similar subcomplexes have been considered for various identities of racks and moves on diagrams .
typically a certain change of knot diagrams requires a condition for quandle @xmath0-cocycles to satisfy to obtain desired cocycle invariant , and the condition leads to a subcomplex .
for example , for defining quandle cocycle invariants for unoriented knots , a good involution was defined in @xcite , and a corresponding condition for cocycles and a subcomplex were defined . for a telephone cord move for racks , a condition for rack @xmath0-cocycles and a subcomplex
were defined in @xcite .
certain moves on handlebody - links were considered in @xcite , and corresponding subcomplexes were defined . in this paper , we give a construction of a @xmath0-cycle @xmath1 from a given identity of quandles in a certain form , show that the abelian extension with a @xmath0-cocycle @xmath2 such that @xmath3 inherits the identity , and construct a subcomplex from the identity .
the homology and cohomology of these subcomplexes remain to be investigated . preliminary material and definitions
are provided in section [ sec : prelim ] , and an outline of the method is explained using type 3 racks in section [ sec : type3 ] .
the main results are presented in section [ sec : main ] with proofs , and applied to type @xmath4 quandles and identities similar to the burnside relations . in section [ sec : id ] , specific identities are examined among small connected quandles , and a large class of alexander quandles is given that satisfies many identities .
in this section , we provide preliminary material , definitions and notation . more details can be found , for example , in @xcite . a _ rack _ @xmath5 is a non - empty set with a binary operation @xmath6 satisfying the following conditions .
\(1 ) for any @xmath7 , the map @xmath8 defined by @xmath9 for @xmath10 is a bijection . ( 2 ) for any @xmath11 , we have @xmath12 the map @xmath13 in the first axiom is called the _ right translation by @xmath14_. by axioms @xmath13 is a _
rack isomorphism_. a _ quandle _ @xmath5 is a rack with idempotency : @xmath15 for any @xmath10 . a _
quandle homomorphism _ between two quandles @xmath16 is a map @xmath17 such that @xmath18 , where @xmath19 and @xmath20 denote the quandle operations of @xmath5 and @xmath21 , respectively .
generalized alexander quandle _ is defined by a pair @xmath22 where @xmath23 is a group , @xmath24 , and the quandle operation is defined by @xmath25 . if @xmath23 is abelian , this is called an _
alexander quandle_. let @xmath5 be a rack . for brevity
we sometimes omit @xmath26 and parentheses , so that for @xmath27 , @xmath28 , @xmath29 , and inductively , @xmath30 .
we also use the notation @xmath31 where @xmath32 is repeated @xmath4 times .
a rack @xmath5 is said to be of _ type @xmath4 _ ( cf .
@xcite ) if @xmath4 is the least positive integer such that @xmath33 holds for all @xmath34 , and we write @xmath35 .
a type 1 quandle is called _ trivial _ , and a type 2 quandle is called a _
kei _ or an _ involutory _ quandle .
the subgroup of @xmath36 generated by the permutations @xmath37 , @xmath10 , is called the _ inner automorphism group _ of @xmath5 , and is denoted by @xmath38 .
a rack is _ connected _ if @xmath38 acts transitively on @xmath5 .
the rack chain group @xmath39 for a rack @xmath5 is defined to be the free abelian group generated by @xmath4-tuples @xmath40 , @xmath27 for @xmath41 .
let @xmath42 be defined by @xmath43 respectively , where @xmath44 denotes deleting the entry .
then the boundary map is defined by @xmath45 $ ] .
the degeneracy subcomplex @xmath46 was defined @xcite for a quandle @xmath5 with generating terms @xmath47 with @xmath48 for some @xmath49 , and the quotient complex @xmath50 was defined @xcite as the quandle homology .
the corresponding @xmath0-cocycle is formulated as follows .
a quandle @xmath0-cocycle is regarded as a function @xmath51 for an abelian group @xmath52 that satisfies @xmath53 for any @xmath54 and @xmath55 for any @xmath56 . for a quandle @xmath0-cocycle @xmath2
, @xmath57 becomes a quandle by @xmath58 for @xmath34 , @xmath59 , denoted by @xmath60 or simply @xmath61 , and it is called an _ abelian extension _ of @xmath5 by @xmath52 . the second factor , in this case , is written in additive notation of @xmath52 .
see @xcite for more details .
computations using @xcite significantly expanded the list of small connected quandles .
these quandles , called _ rig _ quandles , may be found in the package rig @xcite .
rig includes all connected quandles of order less than 48 , at this time .
properties of some of rig quandles , such as homology groups and cocycle invariants , are also found in @xcite .
we use the notation @xmath62 for the @xmath63-th quandle of order @xmath4 in the list of rig quandles , denoted in @xcite by smallquandle@xmath64 .
note , however , that in @xcite quandles are left distributive , so that as a matrix , @xmath62 is the transpose of the quandle matrix smallquandle@xmath64 in @xcite .
before presenting the main theorem and proof , we describe the properties of rack identities through the example of type @xmath65 quandles in this section .
we note that a subcomplex for type 2 quandles , or keis , is defined in @xcite as a special case of their subcomplex .
recall that a rack @xmath5 is of type 3 if it satisfies the identity @xmath66 : @xmath67 for any @xmath34 .
we observe the following three properties .
\(i ) from this identity @xmath66 we form a @xmath0-chain @xmath68 it is checked that @xmath1 is a @xmath0-cycle : @xmath69 + [ \ ( x * y ) - ( x * y * y ) \ ] + [ \ ( x * y*y ) - ( x * y * y * y ) \ ] = 0,\ ] ] using the identity @xmath66 .
\(ii ) let @xmath70 be a rack @xmath0-cocycle with the coefficient abelian group @xmath52 such that @xmath3 .
then for @xmath71 , one computes @xmath72 hence @xmath73 is of type 3 .
\(iii ) define , for each @xmath4 , a subgroup @xmath74 generated by @xmath75 for a fixed @xmath76 , @xmath32 is positioned at @xmath77-th entry .
then @xmath78 is a subcomplex , which will be proved in the general case in section [ sec : main ] . in this section , to illustrate the idea of the proof , we compute the image under the boundary map for a specific generator of @xmath79 .
we simplify the notation and use @xmath80 for @xmath81 , @xmath82 for @xmath83 , etc .
let @xmath84 that is a generator for the case @xmath85 and @xmath86 .
we compute @xmath87 and show that the image is in @xmath88 .
first we compute @xmath89 which is a generator of @xmath88
. then we have @xmath90 one computes that @xmath91 , and @xmath92 [ ( 23 ) 3 ] = [ ( 13 ) ( 23 ) ] 3 = [ ( 12 ) 3 ] 3,\ ] ] so that @xmath93 which is a generator of @xmath88 .
we also compute @xmath94 ( c ) & = & [ ( 1,2,4 ) - ( 13,23,4 ) ] + [ ( 13,23,4)-(133,233,4 ) ] \\ & & \quad + \ [ ( 133,233,4)- ( 1,2,4)]\ = \ 0,\end{aligned}\ ] ] in this case .
next we have @xmath95 which is a generatior .
finally we have @xmath96 which is a generator .
this concludes the computation that @xmath97 .
let @xmath5 be a rack . for brevity
we omit @xmath26 and take the left - most parenthesis as before . fix a surjection @xmath98 , where @xmath99 are positive integers .
we consider identities of the form @xmath100 for @xmath101 , @xmath102 .
the expression @xmath103 is a word of length @xmath104 from the alphabet @xmath105 .
we assume that @xmath106 , since otherwise the quandle is trivial .
for example , for a type @xmath4 rack @xmath5 , there is an identity of the form @xmath107 for any @xmath108 .
another example is @xmath109 .
we call an identity @xmath66 of the form @xmath100 as described above a @xmath110 _ inner identity_. if an inner identity @xmath66 above holds for any @xmath111 , @xmath112 , then we say that @xmath5 satisfies the @xmath110 inner identity @xmath66 .
a rack @xmath5 satisfies a @xmath113 inner identity @xmath66 of the form @xmath114 if and only if @xmath115 for all @xmath116 , @xmath112 .
let @xmath66 be an inner identity @xmath117 .
set @xmath118 , when @xmath119 .
define a @xmath0-chain @xmath120 by @xmath121 let @xmath66 : @xmath122 be an inner identity .
set @xmath118 , when @xmath119 .
let @xmath123 , @xmath124 , be subgroups generated by @xmath125 we use the following lemma in the proof of theorem [ thm : main ] .
[ lem : reduce ] let @xmath5 be a quandle .
* for any @xmath126 , it holds that @xmath127 . * for any @xmath128 , it holds that @xmath129 .
\(i ) one computes inductively , using self - distributivity , @xmath130 [ ( b c_1 \cdots c_{i-1 } ) c_i ] \\ & = & [ ( a c_1 \cdots c_{i-1 } ) ( b c_1 \cdots c_{i-1 } ) ] c_i \\ & = & [ [ ( a c_1 \cdots c_{i-2 } ) ( b c_1 \cdots c_{i-2 } ) ] c_{i-1 } ] c_i \\ & = & \cdots \\ & = & ( ab ) c_1 \cdots c_i .\end{aligned}\ ] ] ( ii ) one computes inductively @xmath131 ( a_i b ) \\ & = & \cdots \\ & = & ( a_1 b ) \cdots ( a_i b ) \end{aligned}\ ] ] as desired . ,
title="fig:",width=432 ] + the proof of lemma [ lem : reduce ] has an diagrammatic interpretation as depicted in figure [ lemma ] .
colorings of knot diagrams by quandles are well known and extensively used to construct knot invariants . at a crossing , for positive crossings in the figure with all arcs oriented downwards ,
the coloring condition is as depicted at the top left crossing of figure [ lemma ] ( b ) , where @xmath132 and @xmath14 are assigned on the left top under - arc and the over - arc , respectively , and @xmath133 ( simply denoted by @xmath134 ) is required to be assigned on the other under - arc
. then the bottom right arc of ( a ) receives @xmath135 . in ( b )
the corresponding arc receives @xmath136 .
thus the fact that the colorings are in bijection under reidemeister moves shows the equality ( i ) .
similarly , ( c ) and ( d ) represents the equality ( ii ) , with @xmath137 and @xmath138 .
one also sees that the inductive calculations in the proof can be represented by step by step moves .
[ thm : main ] let @xmath5 be a rack .
let @xmath66 be a @xmath110 inner identity @xmath139 that @xmath5 satisfies .
then the following holds .
the @xmath0-chain @xmath120 is a @xmath0-cycle , @xmath140 . for an abelian group @xmath52 and a @xmath0-cocycle @xmath2
, @xmath141 satisfies @xmath66 if and only if @xmath142 .
the sequence of subgroups @xmath123 form a subcomplex @xmath143 , @xmath124 .
\(i ) set @xmath118 when @xmath119 as before .
one computes @xmath144 \\ & = & ( x ) - ( x \omega_k ) \ = \ 0 \end{aligned}\ ] ] as desired .
\(ii ) for @xmath145 , @xmath146 , one computes , inductively , @xmath147 which is equal to @xmath148 if and only if @xmath142 , as desired .
\(iii ) we check the following three cases . case ( 1 ) : @xmath149 . in this case
, each term of @xmath150 is obtained from the original term by deleting the @xmath151-th entry , hence the image is an element of @xmath152 .
each term of @xmath153 is obtained from the original by replacing the first @xmath151 entries of the form @xmath154 by @xmath155 by lemma [ lem : reduce ] ( i ) , we obtain @xmath156 hence the image is in @xmath152 .
case ( 2 ) : @xmath157 .
one computes @xmath158 for the @xmath63-the term , and @xmath159 for the @xmath160-th term , so that these terms cancel in pairs by opposite signs .
the first term before the sum over @xmath63 and the last term of the sum over @xmath63 , @xmath161 and @xmath162 are equal and cancel by opposite signs .
hence the image of this case is zero .
case ( 3 ) : @xmath163 .
each term of @xmath164 is obtained from the original term by deleting the @xmath151-th entry , hence the image is an element of @xmath152 .
each term of @xmath165 is computed as @xmath166 by lemma [ lem : reduce ] ( ii ) , we obtain @xmath167 we note that it holds that if @xmath168 then @xmath169 . hence we can set @xmath170 for @xmath171 , and @xmath172 , then @xmath173 so that the above sum @xmath174 is an element of @xmath152 as desired .
the construction of 2-cycles in ( i ) is a generalization of @xcite .
the following are immediate corollaries of theorem [ thm : main ] for specific identities .
let @xmath5 be a type @xmath4 quandle , with the identity @xmath66 : @xmath33 for all @xmath34 for a fixed @xmath175 .
then the following properties hold . *
the @xmath0-chain @xmath176 is a @xmath0-cycle , @xmath177 .
* for an abelian group @xmath52 and a @xmath0-cocycle @xmath2 , @xmath141 satisfies @xmath66 if and only if @xmath178 for any element @xmath34 .
* let @xmath179 be the subgroup of @xmath180 generated by @xmath181 then the sequence of subgroups @xmath143 is a subcomplex .
the following is motivated from burnside relations discussed in @xcite .
we consider the identity @xmath182 for @xmath183 for simplicity denote @xmath184 by @xmath185 where @xmath21 denotes @xmath186 and the exponent @xmath104 represents the number of repetitions of @xmath186 ( but each @xmath186 is not parenthesized ) .
let @xmath5 be a quandle that satisfies the identity @xmath66 : @xmath182 for @xmath187 as above , for all @xmath188 for a fixed @xmath189 .
then the following properties hold . *
the @xmath0-chain @xmath190 $ ] is a @xmath0-cycle , @xmath177 .
* for an abelian group @xmath52 and a @xmath0-cocycle @xmath2 , @xmath141 satisfies @xmath66 if and only if @xmath178 for any element @xmath188 .
* let @xmath179 be the subgroup of @xmath180 generated by @xmath191 \ \ } .\end{aligned}\ ] ] then the sequence of subgroups @xmath143 is a subcomplex .
a quandle @xmath5 is _ medial _ , or _
abelian _ , if the identity @xmath66 : @xmath192 holds for any @xmath193 .
this property is well known , see @xcite for some discussions .
we note that this is not in the form of inner identity , but point out that procedures analogous to those in the proof of theorem [ thm : main ] ( i ) and ( ii ) still apply .
let @xmath194 - [ \ , ( x , u ) + ( x*u , y*v ) \ , ] $ ] .
it holds that if @xmath5 is medial , then ( i ) @xmath177 for any @xmath193 , and ( ii ) for a @xmath0-cocycle @xmath2 with a coefficient abelian group @xmath52 , @xmath73 is medial if @xmath195 .
proof is direct computations . for ( i ) , we compute @xmath196 \\ & & - [ \ , ( x ) - ( x*u ) + ( x*u ) - ( ( x*u)*(y*v ) ) \ , ]
\quad = \quad 0 , \end{aligned}\ ] ] and for ( ii ) , we compute @xmath197
in this section we examine specific inner identities , as well as quandles that satisfy these identities .
first we present the number of type @xmath4 rig quandles for possible values of @xmath4 .
the following list of vectors @xmath198 $ ] represent that there are @xmath199 rig quandles ( all connected quandles of order @xmath200 ) of type @xmath104 .
@xmath201 & [ 3 , 38 ] & [ 4 , 90 ] & [ 5 , 16 ] & [ 6 , 117 ] & [ 7 , 15 ] & [ 8 , 38 ] & [ 9 , 13 ] \\ { } [ 10 , 31 ] & [ 11 , 10 ] & [ 12 , 52 ] & [ 13 , 4 ] & [ 14 , 19 ] & [ 15 , 14 ] & [ 16 , 9 ] & [ 18 , 27 ] \\ { } [ 20 , 19 ] & [ 21 , 14 ] & [ 22 , 11 ] & [ 23 , 22 ] & [ 24 , 9 ] & [ 26 , 5 ] & [ 28 , 17 ] & [ 30 , 15 ] \\ { } [ 31 , 6 ] & [ 36 , 12 ] & [ 40 , 16 ] & [ 42 , 12 ] & [ 46 , 22 ] & & & \end{array}\ ] ] let @xmath5 be a quandle .
the subgroup of @xmath36 generated by right transformations @xmath202 is called the _ inner automorphism group _ and denoted by @xmath38 . for a quandle @xmath5 , the map @xmath203 defined by @xmath204 for @xmath10
is called the _
inner representation_. computer calculations show that for over 3000 non - faithful quandles @xmath5 ( mostly generalized alexander quandles ) it holds that @xmath205 .
for the great majority of these quandles , @xmath206 are abelian extensions . in @xcite
, it was conjectured that if a quandle @xmath5 is a kei , then any abelian extension of @xmath5 is a kei .
thus we make the following conjectures .
let @xmath5 be a quandle .
let @xmath182 be an inner identity , where @xmath184 is a word in the alphabet @xmath211 .
let @xmath212 denote the length of @xmath184 , that is , the number of letters in @xmath184 .
we note that if the length of @xmath184 is @xmath104 , then the quandle is of type at most @xmath104 , since the identity must hold for any values of variables .
we observe the following .
convert this identity to a product of @xmath214 , @xmath215 .
the identity is equivalent to that his product equals 1 , so we can solve it for @xmath216 . thus @xmath216 is a product of @xmath217 , @xmath218 .
this identity must hold for any @xmath219 . by fixing the values of @xmath220 and choosing different values of @xmath132
, we obtain @xmath221 for all @xmath222 .
since @xmath223 , we obtain @xmath224 for all @xmath222 . under the assumption @xmath184
is written as @xmath227 for @xmath228 , where the exponents represents the number of repetitions .
then the identity @xmath182 is converted to the identity @xmath229 in @xmath38 .
_ length 4_. excluding trivial quandles from lemma [ lem : triv ] , we have the following cases . ( 1 ) @xmath238 , ( 2 ) @xmath239 , ( 3 ) @xmath240 .
lemma [ lem : type ] implies that ( 1 ) or ( 2 ) holds if and only if the quandle is a kei .
_ length 5_. excluding trivial quandles from lemma [ lem : triv ] , there are 10 identities @xmath182 ( all must have two distinct letters in @xmath184 ) , @xmath242 from lemma [ lem : type ] , the identities @xmath243 , @xmath244 , @xmath245 , @xmath246 , @xmath247 imply trivial quandle .
computer calculation shows that none of the remaining is satisfied by any of 790 rig quandles .
we conjecture that no connected quandle satisfies the remaining 5 identities of length @xmath248 .
the following words are satisfied by the same 202 rig quandles , which contain all 117 rig keis : @xmath250 , @xmath251 , @xmath252 .
the word @xmath253 is satisfied by 55 rig quandles , 4 of which are keis . _
length 7_. except for those words @xmath184 that give trivial quandles from lemmas [ lem : triv ] , [ lem : type ] and type 7 quandles ( 15 of them among 790 rig quandles ) , there are only two rig quandles that satisfy the identity @xmath182 with 2 letter words @xmath184 of length 7 , and they are : we consider alexander quandles @xmath262 , where @xmath263 is an automorphism of an abelian group @xmath5 , with @xmath264 .
inductively one computes @xmath265 setting @xmath266 and @xmath267 for @xmath268 , @xmath112 , we obtain @xmath269 let @xmath270 and @xmath271/ ( g_{m , n}(t ) ) $ ] . for all primes @xmath272 , @xmath273 , so that @xmath274 is invertible in @xmath5 and hence @xmath5 is connected , and @xmath275 in @xmath5 .
hence there is an infinite family of connected quandles that satisfy @xmath182 . for a group @xmath23 , the least possible integer @xmath4 such that @xmath278 for all @xmath279
is called the _ exponent _ of @xmath23 .
we note that if the exponent of @xmath280 is @xmath4 , then @xmath5 satisfies the inner identity @xmath281 for any word @xmath184 of length @xmath4 , where @xmath282 denotes @xmath283 repeated @xmath4 times with no parentheses .
we computed the exponent of @xmath280 by for all 790 rig quandles , and obtained the following data .
the notation @xmath284 $ ] below indicates that there are @xmath4 rig quandles @xmath5 such that the exponent of @xmath38 is @xmath285 .
the pairs are listed in order of increasing exponent .
@xmath286 & [ 10 , 4 ] & [ 12 , 59 ] & [ 14 , 3 ] & [ 15 , 1 ] & [ 18 , 47 ] & [ 20 , 15 ] \\ { } [ 21 , 2 ] & { } [ 22 , 1 ] & [ 24 , 38 ] & [ 26 , 1 ] & [ 30 , 22 ] & [ 34 , 1 ] & [ 36 , 31 ] \\ { } [ 38 , 1 ] & [ 39 , 6 ] & { } [ 40 , 6 ] & [ 42 , 22 ] & [ 46 , 1 ] & [ 48 , 4 ] & [ 50 , 5 ] \\ { } [ 52 , 2 ] & [ 54 , 9 ] & [ 55 , 4 ] & { } [ 57 , 2 ] & [ 58 , 1 ] & [ 60 , 44 ] & [ 62 , 7 ] \\ { } [ 66 , 4 ] & [ 68 , 2 ] & [ 70 , 3 ] & [ 72 , 13 ] & { } [ 74 , 1 ] & [ 78 , 13 ] & [ 82 , 1 ] \\ { } [ 84 , 24 ] & [ 86 , 1 ] & [ 90 , 9 ] & [ 93 , 2 ] & [ 94 , 1 ] & { } [ 100 , 10 ] & [ 110 , 4 ] \\ { } [ 111 , 2 ] & [ 114 , 2 ] & [ 116 , 2 ] & [ 120 , 27 ] & [ 129 , 2 ] & [ 136 , 4 ] & { } [ 140 , 6 ] \\ { } [ 148 , 2 ] & [ 155 , 4 ] & [ 156 , 10 ] & [ 164 , 2 ] & [ 168 , 4 ] & [ 171 , 6 ] & [ 180 , 12 ] \\ { } [ 186 , 2 ] & [ 203 , 6 ] & [ 205 , 4 ] & [ 210 , 4 ] & [ 222 , 2 ] & [ 240 , 3 ] & [ 253 , 10 ] \\ { }
[ 258 , 2 ] & { } [ 272 , 8 ] & [ 301 , 6 ] & [ 310 , 4 ] & [ 328 , 4 ] & [ 330 , 16 ] & [ 333 , 6 ] \\ { } [ 342 , 6 ] & [ 360 , 1 ] & { } [ 406 , 6 ] & [ 410 , 4 ] & [ 420 , 10 ] & [ 444 , 4 ] & [ 465 , 8 ] \\ { } [ 506 , 10 ] & [ 602 , 6 ] & [ 666 , 6 ] & { } [ 812 , 12 ] & [ 820 , 8 ] & [ 840 , 3 ] & [ 903 , 12 ] \\ { } [ 930 , 8 ] & [ 1081 , 22 ] & [ 1332 , 12 ] & [ 1640 , 16 ] & { } [ 1806 , 12 ] & [ 2162 , 22 ] & [ 2520 , 2 ] \end{array}\ ] ] we are grateful to jozef przytycki for valuable comments . m.s . was partially supported by the nih 1r01gm109459 .
the content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of nih .
vendramin , l. , _ rig a gap package for racks and quandles .
_ available at http://github.com / vendramin / rig/. vendramin , l. , _ on the classification of quandles of low order _ ,
j. knot theory ramifications , 21 ( 2012 ) , no .
9 , 1250088 .
zablow , j. , _ on relations and homology of the dehn quandle _ , algebraic & geometric topology , 8 ( 2008 ) 1951 .
|
quandle homology was defined from rack homology as the quotient by a subcomplex corresponding to the idempotency , for invariance under the type i reidemeister move .
similar subcomplexes have been considered for various identities of racks and moves on diagrams .
we observe common aspects of these identities and subcomplexes ; a quandle identity gives rise to a @xmath0-cycle , the abelian extension with a @xmath0-cocycle that vanishes on the @xmath0-cycle inherits the identity , and a subcomplex is constructed from the identity .
specific identities are examined among small connected quandles .
|
it is almost universally accepted that the majority of matter in the universe is non - baryonic .
existence of this `` dark matter '' is supported by several kinds of evidence : galaxy rotation curves , gravitational lensing of background objects by galaxy clusters , the small value of @xmath9 determined from big bang nucleosynthesis , and @xmath10 from the cosmic microwave background ( cmb ) anisotropy .
the gravitational effects of dark matter are observed , but in spite of numerous investigations with particle accelerators and direct detection experiments , no direct signature of dark matter interaction with ordinary matter has ever been found . supersymmetry theory ( susy )
provides a candidate dark matter particle : a linear combination of higgsino , z - ino , and photino states commonly known as the neutralino , @xmath11 , the lightest stable supersymmetric particle ( see jungman , kamionkowski , & griest 1996 for a review ) .
direct detection of neutralinos is difficult because of their very weak interactions with ordinary matter , but if the neutralino is a majorana fermion ( and therefore is its own anti - particle , @xmath12 ) , it self - annihilates and produces some combination of w and z bosons , mesons , @xmath1 pairs , and @xmath13-rays ( see gunn
_ et al . _
1978 for the basic scenario ) .
baltz & wai ( 2004 ) point out that a significant fraction of the power liberated in self - annihilation may go into ultra - relativistic @xmath1 pairs ( hereafter `` electrons '' ) , producing observable microwave synchrotron radiation .
these high energy electrons inverse compton scatter cmb and starlight photons , shifting them to soft and hard @xmath13-rays , respectively . to date , efforts have focused largely on @xmath13-ray observations of dark matter concentrations in dwarf spheroidal galaxies ( gondolo 1994 , blasi _ et al . _ 2003 , baltz & wai 2004 ) or the galactic center ( stoehr _ et al . _ 2003 ) , or synchrotron observations of satellite galaxies ( baltz & wai 2004 ) .
direct detection of high energy positrons near earth has also been discussed ( baltz & edsj 1998 , baltz _ et al . _
2002 ) , but is thought to be unlikely , given the foreground produced by other mechanisms and the high boost factors required to obtain an observable signal ( hooper _ et al . _
copious annihilation near the black hole in the galactic center would be detectable at radio frequencies for most neutralino models if there were a steep central cusp in the density profile ( gondolo 2000 ) but non - detection may indicate lack of a cusp .
detection of synchrotron emission near the black hole in the galactic center has also been considered ( bertone _ et al . _ 2002 ,
aloisio _ et al . _
2004 ) but detection of the synchrotron annihilation signal in the inner regions ( @xmath14 kpc ) of the galaxy was thought to be impractical because of the complexity of the ism microwave emission from other components ( thermal dust , spinning dust , free - free , and `` ordinary '' synchrotron from shock - accelerated electrons ) .
recent advances in our understanding of ism emission have made this problem tractable . the _ wilkinson microwave anisotropy probe _
( _ wmap _ ) has revealed the structure of the galactic ism with unprecedented precision at microwave ( @xmath15 ) frequencies , leading to a reassessment of the galactic foregrounds ( bennett _ et al . _ 2004 ,
finkbeiner 2004 , finkbeiner _
et al . _
one surprising result of the increasingly careful foreground analysis is an excess of emission in the inner galaxy , distributed with ( approximate ) radial symmetry within @xmath16 of the galactic center ( dubbed the `` haze , '' finkbeiner 2004 ) .
this emission is not correlated with any of the foreground templates representing the known microwave foregrounds .
although the spectrum of this excess is somewhat flatter than standard synchrotron emission , we argue that it is consistent with a population of ultra - relativistic electrons created in the inner galaxy and diffusing outward , and consider the possibility that these electrons are products of dark matter ( dm ) annihilation . to be concrete , we refer to the dm particle as the neutralino throughout , although _ any _ stable neutral particle with a mass of @xmath17 and the assumed @xmath18 would have the consequences described below . assuming the neutralino is the dominant form of dark matter
, the average relic density today is @xmath19 according to the _ wmap_cosmology ( spergel _ et al . _
2003 ) . by setting the annihilation rate equal to the expansion rate after freeze - out ,
the cross section is commonly refered to as the `` cross section '' even though that term would properly refer only to @xmath20 .
it would be more precise to call @xmath18 the `` annihilation rate coefficient . ''
] @xmath18 is determined by this relic abundance ( see [ sec_relic ] ) . using a fiducial mass of @xmath6 , a diffusion term ( @xmath21 ) derived from cosmic ray data , and an energy loss term ( @xmath22 ) including synchrotron losses and
inverse compton scattering ( ics ) off of starlight and cmb photons , we calculate the electron energy distribution as a function of galactocentric radius for various assumptions
. an analytic solution is presented for the simplest case , and numerical solutions for variable @xmath23 and @xmath21 terms . for the assumed dark matter distribution ,
all models produce a similar ( factor of @xmath24 ) power output , since the power input by annihilation is fixed _
a priori_. in all cases , the spectrum , morphology , and predicted synchrotron power output by the model is similar to the observed _ wmap _ excess .
this is tantalizing , because if one can determine the losses ( i.e. energy density of starlight and magnetic field ) in the inner galaxy , the fraction of power liberated as relativistic @xmath1 , and the dark matter density profile , then the total @xmath11 annihilation power can be computed .
given the dark matter distribution and the annihilation power in the inner galaxy , the approximate mass , @xmath25 , can be computed . with luck , an annihilation line or sharp cut off in @xmath13 ray continuum will be observed , fixing the mass more precisely . with refined galactic models and
improved microwave and @xmath13 ray data , it may actually be possible to determine the mass and cross section of a new particle astrophysically , making a specific prediction to be verified by accelerator experiments ( e.g. at the lhc , ellis _
et al . _
2004 ) .
the main difference between our model and previous models is that electron diffusion is solved for in a self - consistent way ( like baltz & wai 2004 ) and its effect on the synchrotron spectrum computed .
the consequent flattening of the spectrum , especially near the galactic center , has not been emphasized in the previous literature . in 2 , the excess microwave emission near the galactic center is described .
model parameters for the hot electron population and ism are justified in 3 , and the model is evaluated and compared to _ wmap _
data in 4 .
the _ wilkinson microwave anisotropy probe _ ( _ wmap _ ) has observed the sky with good sensitivity ( @xmath26 per pixel in the 1 yr data ) and measured the cosmological anisotropy power spectrum ( hinshaw _ et al . _
2003 ) and , with less significance , its polarization ( page _ et al . _
2003 ) , in order to determine fundamental cosmological parameters ( spergel _ et al . _
2003 ) . to the extent that galactic foreground signals from thermal dust emission , free - free ( thermal bremsstrahlung ) , `` ordinary '' synchrotron ( relativistic electrons accelerated by supernovae and spiraling in the galactic magnetic field ) and spinning dust ( see draine & lazarian 1998 ) are present in the data , they have been removed and do not interfere with the cosmological signal ( bennett _ et al . _
2003 ) . in the course of studying these foregrounds and exploring the possibility of spinning dust ( finkbeiner 2004 , finkbeiner
_ et al . _ 2004
) , evidence for an additional emission component has emerged .
this enigmatic emission is centered at the galactic center and extends @xmath27 away in all directions , falling rapidly ( @xmath28 ) with projected distance from the galactic center .
it is not likely to be an artifact of the foreground subtraction , because south of the gc it is similar in brightness to the other components , and the fit is sufficiently rigid ( only eight parameters for the whole sky to fit 5 bands @xmath29 pixels ) that it is robust to errors in the templates .
even north of the gc , the excess is approximately symmetrical about the gc in spite of much brighter foreground objects such as the large ( @xmath30 ) nearby nebula of @xmath31oph .
this emission component has been nicknamed `` the haze '' to avoid giving it an interpretive name prematurely ( see finkbeiner 2004 ) . the haze was first thought to be a free - free component from gas too hot to have recombination line emission ( @xmath32 ) and too cold to have significant x - ray emission in the rosat ( snowden _ et al . _
1997 ) 1.5 kev band ( @xmath33 ) . north of the gc there is sufficient gas and dust to hide the h@xmath34 recombination line and x - ray emission , but to the south , column densities are low enough that both would be seen . for @xmath35 gas
the x - ray emission would be soft enough that gas absorption could block it , but gas is thermally unstable at @xmath36 k because of efficient metal - line cooling ( e.g. , spitzer 1978 ) , and the amount of gas required is inconsistent with hii column densities determined from pulsar dispersion measures given by taylor & cordes ( 1993 ) .
furthermore , the free - free spectrum of @xmath36 k hii extends up to @xmath37 and the integrated spectral power suggests a substantial free - free power output of @xmath38 .
it is necessary to consider alternative sources of emission .
in this section , the model parameters are specified : annihilation cross section , mass , @xmath1 energy distribution , magnetic field strength , photon field available for ics , and dark matter mass profile in the milky way .
the parameters are listed in table 1 .
if the neutralinos present today are the cosmological relic of thermal production in the early universe , then the mean neutralino mass density as a fraction of the critical density , @xmath39 can be computed from its mass and cross section .
the _ wmap_cosmology ( spergel _ et al . _
2003 ) gives @xmath40 , where the subscripts `` dm '' , `` m '' and `` b '' denote `` dark matter '' , `` total matter '' and `` baryon '' , respectively , and @xmath41 is the hubble constant in units of @xmath42 .
the neutralino number density falls rapidly when the temperature of the universe ( @xmath43 ) falls much below the particle mass ( @xmath44 ) .
after this , neutralino annihilations cause the number density , @xmath45 , to drop rapidly [ @xmath46 to the level where the particle annihilation rate ( @xmath47 ) is comparable to the expansion rate , @xmath48 , after which the comoving density asymptotes to the `` freeze - out '' density . in the early , radiation - dominated , universe , @xmath49 ,
so the criterion is = _ avn ~_avn_0 a^-3 ~h ~a^-2 where @xmath50 is the co - moving number density . because temperature @xmath51 , n_0 t_f ~ where @xmath52 is the freeze - out temperature , proportional to the particle mass .
this leads to the remarkable conclusion that the co - moving relic mass density , @xmath53 , is inversely proportional to the cross section , so if one knows the relic mass density today ( or assumes it ) the cross section is fixed .
a more complete calculation ( jungman _ et al . _
1996 ) yields the same relation in terms of @xmath54 : _
h^2 = a proper calculation of the relic density takes account of the form of the density drop after freeze - out and introduces a weak mass dependence in the relic density , but the approximation given in eq .
( [ eq_relic ] ) is adequate for our purposes , and we adopt @xmath55 . we take the neutralino mass , @xmath44 to be @xmath6 , approximately at the weak scale , and lower than the bounds of @xmath56 given by ellis _ et al . _
( 2000 ) , or @xmath57 several tev ( baltz & gondolow 2004 ) .
this choice of mass is arbitrary , and is degenerate with the fraction of power going into energetic @xmath1 given in [ sec_result ] .
there is considerable uncertainty in both the injection energy distribution and the effects of diffusion on the resultant number density per energy , @xmath58 .
one possible assumption is that annihilating neutralinos go directly to quark anti - quark pairs which then go to pions : @xmath59 the pions then decay : @xmath60 @xmath61 followed by @xmath62 @xmath63 this results in an injection energy distribution that goes as @xmath64 in the low energy limit ( see blasi _ et al . _
2003 ) , with a significant fraction of the energy carried away by neutrinos
. however , if annihilation to @xmath65 leptons is important ( @xmath66 ) the resulting @xmath1 spectrum would be harder ( hooper _ et al . _
2004 ) .
the injection spectrum must then be evolved through the diffusion and energy loss equations to determine the steady state electron density .
the model in [ sec_model ] assumes spherical symmetry , but one might also consider cylindrical symmetry with boundaries a few kpc from the galactic midplane at which the electrons escape . if the electrons have time to diffuse to the boundaries before losing most of their energy , the electron spectrum can be considerably hardened . in the limit of rapid escape ,
the steady state spectrum is just the injection spectrum , so that sets a limit on the spectral hardness .
the starlight photon energy density in the inner galaxy is necessary to compute electron energy loss due to inverse compton scattering .
et al . _ ( 2000 ) have considered a model based on _ iras _ and dirbe infra - red data in addition to the usual stellar information , and find that the starlight energy density is approximately @xmath67 in the galactic plane , or about @xmath68 at @xmath69 .
this compares to the energy density of @xmath68 for a @xmath70 g magnetic field , and @xmath71 for the cmb .
the far ir emission is subdominant , but should be included in future models
. heiles ( 1996 ) presents measurements of different moments of the galactic magnetic field in the plane at galacto - centric distances of 4 and 8.5 kpc .
none of these measurements is exactly the quantity we need , @xmath72 , but they are consistent with @xmath70 g at the solar circle and increasing exponentially toward the galactic center . to demonstrate that our result is not strongly field dependent , we try two cases : @xmath70 g , and @xmath73 g .
this field is similar to the field radial dependence derived by han ( priv .
comm . ) with a scale length of @xmath74 .
because the @xmath75 field and starlight are so close to equipartition , we assume that they are exactly in equipartition for simplicity , even though there is no obvious mechanism to regulate this . also , the field is assumed to be tangled in an isotropic way , so that the particles move in a random walk .
a more detailed model would incorporate what is known about field strength variation with position in the galaxy , especially the fact that the field is stronger in the spiral arms than in inter - arm regions .
synchrotron emission is enhanced perpendicular to the field lines and suppressed along them ; we assume isotropic tangling of the field and therefore isotropic emission .
furthermore , the ratio of the tangled to ordered strength , and the mean direction of the ordered part may be important .
for example , han ( 2004 ) suggests that the bulge is dominated by a solenoidal component perpendicular to the galactic plane , which may facilitate rapid escape of electrons from the galactic center , whereas particles created in the arms may be trapped there for longer periods of time .
therefore we should be cautious about applying conclusions drawn from local cosmic ray propagation to particles in the inner galaxy .
this is a rich subject , and is well beyond the scope of this article .
we assume a truncated navarro , frenk , & white ( 1997 ; nfw ) profile . because we are only interested in the inner @xmath76 of the galaxy ,
the larger scale shape of the nfw profile is unimportant to this calculation , other than that it serves to tie the assumed milky way mass ( within 400 kpc ) of @xmath77 to a mass density at the center .
we assume an nfw profile with @xmath78 kpc , and arbitrarily truncate the core at 600 pc .
although recent computer simulations indicate that nfw is not valid at the core @xcite , there is theoretical support from the jeans equation for an index of -1 or steeper in the core ( hansen 2004 ) . in the case of a moore ( 1998 )
profile ( @xmath79 ) the annihilation power can be a factor of 5 times higher .
the details of the central truncation are unimportant for nfw profile ( where each concentric shell contributes equal annihilation power ) but becomes important for the moore profile .
at energies of tens of gev , electrons moving in @xmath80gauss fields take millions of years to lose half their energy , so it is a poor approximation to neglect electron diffusion .
we assume that electrons resulting from the annihilation are created at a rate proportional to @xmath11 number density squared and then diffuse in a random walk through the ( rather tangled ) galactic magnetic field .
the number density @xmath81 between energy @xmath82 and @xmath83 evolves according to n(e , * x * ) = ( k(e,*x*)n ) + + q(e,*x * ) where @xmath84 is the diffusion coefficient ( area per time ) , @xmath85 is the energy loss coefficient ( energy loss per time ) , and @xmath86 is the source term ( number density per time ) . neglecting the position dependence of @xmath21 and @xmath87 ,
the steady state equation is : -k(e ) ^2 n -= q(e,*x * ) . in the simple case where @xmath21 is energy independent and @xmath88 with @xmath23 constant ,
the solution for a mono - energetic delta - function source at the origin , @xmath89 , and zero density at @xmath90 , is simply a superposition of spherically symmetric gaussians losing energy and expanding with @xmath91 .
the number density of electrons produced between @xmath92 and @xmath93 in the past is n(e(t),r ) de / dt = q_0(-r^2/2(t)^2 ) with @xmath94 the creation rate ( number per second ) , and where electrons created with energy @xmath95 at @xmath96 lose energy as b(t)==-ae^2 so that e=. in general , the energy loss rate of each electron is b=_tc^2 ^ 2(u_b+u_ph ) where @xmath97 is the thomson cross section , @xmath98 is the speed of the electron , @xmath99 is the electron energy , @xmath100 is the magnetic field energy density and @xmath101 is the photon energy density . in the ultra - relativistic limit , a= , where @xmath102 . for electrons created time @xmath92 ago ,
^2=2kt= so that the number density as an explicit function of energy is : n(e , r ) = q_0 ^3/2 ( ) and the integral along the line of sight is n(e , r ) = ( ) where @xmath103 is the projected galacto - centric distance .
this green function solution must be convolved with the projected source density to obtain the solution for arbitrary @xmath104 , but just this simple expression provides some intuition .
( see appendix a for more details ) .
the energy distribution of the entire @xmath105 population is @xmath106 , as one would obtain for a population created at energy @xmath95 and confined to a box .
however , along any line of sight , the distribution is flatter ( @xmath107 ) because of diffusion .
the older electrons that have lost much of their energy have also had time to diffuse away from the source , leaving a flatter spectrum below some cut - off energy and a sharp drop above it .
this cut - off energy decreases with distance from the source .
another feature of the above solution is that the total projected radial density @xmath108 diverges in the absence of a lower energy bound .
this makes sense , because electrons have been created at rate @xmath94 forever and never disappear . in reality , the diffusion zone in the galaxy has finite size , and most electrons escape while still relativistic .
the effects of this escape will be seen in the numerical solutions in the next section . for less restrictive assumptions about @xmath21 , @xmath23 , and @xmath104 , analytic solutions to eq ( [ eq_pde ] ) are cumbersome or non - existent , and the pde
must be solved numerically .
being interested in the inner galaxy , where dark matter density is assumed to be highest , we assume spherical symmetry and use the fact that the laplacian of a radial function can be written @xmath109 it is convenient to solve for @xmath110 on a radius - energy @xmath111 grid and divide the result by @xmath112 . in most cases a grid with 30 logarithmic energy bins @xmath113 and 60 radial bins @xmath114 works well .
boundary conditions for @xmath115 are taken to be zero at @xmath116 ( as required ) and @xmath117 and an nfw profile squared as the input source distribution at maximum energy .
the density is truncated in the inner 600 pc to the value at 600 pc to avoid the central cusp .
results are not sensitive to this truncation .
the matrix @xmath118 describing the evolution of the system in state @xmath119 ( i.e. @xmath112 times density at discretely sampled positions and energies ) from time step @xmath120 to @xmath121 is constructed such that @xmath122 the steady state solution is the nullspace of @xmath118 ( i.e. @xmath123 ) , readily found with svd ( singular - value decomposition ) .
this is in general a multi - dimensional space providing exactly enough freedom to fit the boundary conditions stated above .
grid solutions have been computed for two cases ( in each case , @xmath124 and @xmath125 ) : ( 1 ) @xmath126 , and ( 2 ) @xmath127 ) , @xmath128 , and @xmath23 representing the energy loss from synchrotron and scattering of cmb and starlight photons assuming equipartition of starlight and the magnetic field , and an exponential radial dependence ( see eq .
[ eq_a ] ) .
the energy dependence of @xmath129 given here is proportional to but somewhat higher than that found in the study of cosmic ray propagation in the galaxy , including diffusion and convection , by webber , lee , & gupta ( 1992 ) .
although this study considered cosmic - ray nuclei and not leptons , at the high particle rigidity ( momentum per charge ) of interest , @xmath1 diffuse in a similar fashion . the electron energy distribution for constant @xmath23 and @xmath21 ( model 1 ) is shown in figure [ fig_simple ] , both as a function of energy , plotted for different radii , and as a function of radius , plotted for different energies
results for the analytic solution ( [ sec_analytic ] ) are also shown , to demonstrate that diffusion from the edges of the numerical grid is unimportant for energies @xmath130 . because these higher energies are the ones relevant for synchrotron emission at @xmath131 , the numerical solution is adequate for a comparison to _
the more realistic model 2 , with variable @xmath21 and @xmath23 , yields a somewhat steeper electron energy distribution ( fig .
[ fig_density ] ) , but results are qualitatively similar . the electron spectrum is integrated along each line of sight , and the synchrotron spectrum computed ( figure [ fig_synch ] ) .
the corresponding synchrotron spectrum for model 1 is very similar .
the power available from annihilation ( a total annihilation rate of @xmath132 for the galaxy ; see appendix a ) is the same in both models , so it is no surprise that the total power output in synchrotron is similar . the radial profile of synchrotron emission is shown in figure [ fig_synch]c for the _ wmap _ bands , but these assume that all annihilation power appears as ultra - relativistic @xmath1 .
comparison of the model with the observed _ wmap _ haze residual ( finkbeiner 2004 ) in figure [ fig_dm_haze ] indicates that a factor of 0.25 is appropriate , i.e. 25% of the annihilation power appears as synchrotron emission . for a different neutralino mass
, the annihilation power varies as @xmath133 so this result would appear to put a hard upper limit of a few hundred gev on the mass . for @xmath134 a smaller fraction of the power would be required , but most models with @xmath135 violate accelerator constraints .
of course , halo clumpiness of the kind invoked by de boer _ et al .
_ ( 2004 ) can increase the available power , and allow greater @xmath44 and/or more diffusion losses off the edge of the diffusion zone .
the hard @xmath13 rays produced by inverse compton scattering of starlight photons have an energy density similar to the microwave synchrotron emission ( [ sec_photons ] ) .
the softer @xmath13 rays from upscattered cmb photons are expected to be subdominant in the inner galaxy , but perhaps still detectable .
extensive searches for these signal have been carried out , and detections so far are suggestive . the egret data from the _ compton gamma ray observatory _
at @xmath136 have long been claimed to show an excess of emission in the inner galaxy over that expected from cosmic ray interaction with the ism ( hunter _ et al . _ 1997
; strong , moskalenko , & reimer 2000 ) .
the hard @xmath13 excess has remained enigmatic , though strong _
et al . _ ( 2000 ) were able to fit the observed spectrum by assuming a much flatter electron energy distribution above 1 gev than that observed in the cosmic ray population near earth .
recently , de boer _ et al . _ ( 2004 ) have considered dark matter annihilation as a source for these hot electrons and found reasonable agreement with @xmath13 ray data and other constraints .
some of the de boer _ et al . _
assumptions about dark matter structure in the galaxy may be controversial ( a high degree of clumpiness is assumed ) , but it is interesting that both groups conclude that a very hot electron spectrum is present .
note that the absence of clumpiness would imply a lower @xmath13 ray production , and would not violate the observations ; there could be other sources as well .
an excess at lower energies observed by comptel is at roughly the energy expected for scattered cmb photons ; however the observed amplitude is higher by one or two orders of magnitude .
the integral @xmath13 ray observatory has revealed numerous soft @xmath13 point sources in the inner galaxy that could account for this emission ( lebrun _ et al . _
the fact that these sources are resolved leaves open the possibility that the upscattered cmb may be detected at a lower level with more sensitive instruments in the future .
a detection of cmb scattering would be much easier to interpret than the putative starlight scattering , since the cmb photon density is isotropic and known .
starlight scattering calculations are significantly complicated by the fact that both energy and direction distributions must be known .
a great deal of work remains to tie the information provided by the observation of the _ wmap _ microwave haze to the important constraints already imposed by the @xmath13 ray data .
we have attempted to constrain models of self - annihilating dark matter by searching for synchrotron emission from the annihilation products , assumed to be ultra - relativistic @xmath1 pairs .
we refer to the dark matter particle as the supersymmetric neutralino , and the mass as @xmath44 , even though the results are of generic interest for any stable dark matter particle that annihilates into ultra - relativistic @xmath1 .
the galactic emission found by finkbeiner ( 2004 ) in the _ wmap_data in excess of the expected foreground galactic ism signal may be a signature of such dark matter annihilation .
it has the spectrum , morphology , and total power expected for the simple model given above , which uses reasonable values for the parameters of the dark matter halo density , particle mass , and cross section ; parameters that agree with other cosmological and accelerator constraints .
additional tests of this model are possible , given the x - rays and @xmath13-rays produced by inverse - compton scattered cmb and starlight photons , respectively .
current data are consistent with the model , but provides only loose constraints .
in addition to better x - ray and @xmath13-ray data , sensitive microwave observations of other edge - on spiral galaxies ( e.g. ngc 891 or ngc 4565 ) might reveal a similar signal , and verify these results .
tests of milky way satellite dwarf galaxies have also been suggested ( baltz & wai 2004 ) but the signal from these may be too weak for current technology . while the dark matter annihilation interpretation is exciting , it must be emphasized that other explanations are possible .
a hard electron energy spectrum could be produced by novel physics around the supermassive black hole at the galactic center .
although observations of other galaxies have not revealed such a flat spectrum of emission , they may have lacked the necessary resolution to separate the various emission components .
a more thorough investigation is required before declaring that the energetic population of electrons result from dark matter annihilation , and indeed , astrophysical observations alone are unlikely to make an entirely convincing case unless an annihilation line or sharp energy cut - off is found in @xmath13 rays .
complementary data from the lhc ( ellis _ et al . _
2004 ) and other particle accelerators will play an essential role in making the case for a new particle .
the vastly improved data expected over the next few years , both experimental and observational , may finally provide an observational window into one of the great mysteries of modern cosmology , dark matter .
i am indebted to jim gunn and bruce draine for help and encouragement ; david schlegel , robert lupton , and nikhil padmanabhan for inspiration , technical advice , and helpful skepticism ; carl heiles for advice on the galactic b - field and other matters ; marc davis for enlightening discussions ; and amber miller for reminding me what an important problem dark matter annihilation is .
gus and flora schultz s kind hospitality allowed me to work out the essentials of this problem while visiting their home in berkeley .
this research made use of the nasa astrophysics data system ( ads ) and the idl astronomy user s library at goddard .
i am supported by nasa ltsa grant nag5 - 12972 , the russell fellowship , and the cotsen fellowship of the society of fellows in the liberal arts at princeton university .
we assume the dm density in the galactic halo obeys the nfw density profile , _
nfw(r ) = with a core radius @xmath137 and @xmath138 , making the density in the solar neighborhood @xmath139 and @xmath140 .
the annihilation rate per volume is ( r ) = ( ) ^2 _ avfor majorana particles , the annihilation rate is @xmath141 , but in each annihilation , two particles are removed , which cancels the factor of 2 in the annihilation rate . note that the integral of @xmath142 over all space ( total number of annihilations per time ) is 4_0^ ( r ) r^2 dr = ^2(_0/m_)^2r_s^3 _
avwhich for the assumed parameters of @xmath143 , @xmath144 , @xmath138 , and @xmath78kpc , is @xmath145 for the whole galaxy . assuming that the @xmath1 pair creation rate is proportional to @xmath47
, we define the efficiency @xmath146 as the number of relativistic @xmath105 ( or @xmath147 ; hereafter simply called electrons ) created per neutralino annihilation , so that the creation rate is @xmath148 .
the electron density is obtained by convolving this density function with the green function obtained in eq ( [ eq_green ] ) . in general , such a convolution looks like h(e , r)=d dr^2 dr ( r ) g(|*x*-*x*| ) and since both functions depend only on radius , we can assume @xmath112 to be on the @xmath149 axis such that the angle between @xmath112 and @xmath150 is @xmath151 , then n(e , r)=d dr^2 dr ( r ) g ( ) inserting the functions gives n(e , r)= d dr^2 dr ( -(r^2+r^2 - 2rr)/2 ^ 2 ) where @xmath20 is the neutralino self - annihilation cross section , not to be confused with the gaussian width @xmath152 .
integrating over @xmath151 and @xmath153 we get n(e , r)= dr this quantity is number density per energy . integrating @xmath154 over the whole galaxy , we obtain a total annihilation rate of @xmath155
|
excess microwave emission observed in the inner galaxy ( inner @xmath0 kpc ) is consistent with synchrotron emission from highly relativistic @xmath1 pairs produced by dark matter particle annihilation .
more conventional sources for this emission , such as free - free ( thermal bremsstrahlung ) , thermal dust , spinning dust , and the softer galactic synchrotron traced by low - frequency surveys , have been ruled out . the total power observed in the range @xmath2 is between @xmath3 and @xmath4 , depending on the method of extrapolation to the galactic center , where bright foreground emission obscures the signal .
the inferred electron energy distribution is diffusion hardened , and is in qualitative agreement with the energy distribution required to explain the gamma ray excess in the inner galaxy at @xmath5 as inverse - compton scattered starlight .
we investigate the possibility that this population of electrons is produced by dark matter annihilation of @xmath6 particles , with cross section @xmath7 , and an @xmath8 dark matter mass profile truncated in the inner galaxy , and find this scenario to be consistent with current data .
|
two - dimensional ( 2d ) materials have been attracting tremendous interest since the discovery of graphene in 2004.@xcite many new 2d materials have been realized to date.@xcite in this family of 2d materials , graphene still stands out in that it is completely flat with only one - atom - thickness .
other monolayer materials such as silicene or phosphorene have buckled structures with atoms displaced perpendicular to the 2d plane ; while materials like transition metal dichalcogenides consist of multiple atomic layers . the single - atom - thickness endows graphene with superior properties.@xcite representing the ultimate limit of a material s thickness , it enables devices that are more compact in size ; and the atomic thickness also makes the material more susceptible to external control , e.g. , the carrier density in graphene can be tuned up to @xmath5/@xmath4 by gating.@xcite meanwhile , it is much desired to realize various symmetry - breaking phases in 2d materials .
2d materials with magnetic ordering are of particular interest , because such nanoscale magnetism is not only fascinating from a scientific point of view , it is also of great value for important applications such as magnetic information storage and spintronics devices.@xcite extensive efforts have been devoted in this direction , mostly by doping non - magnetic 2d materials with magnetic atoms or by creating certain structural defects.@xcite however , the induced magnetism is usually weak , and it is very difficult to control the distribution of dopants or defects in experiment . moreover , acting as extrinsically introduced scattering centers , these dopants or defects would strongly scatter the carriers and deteriorate the material s transport properties . it would become much better if we could generate magnetism without introducing extrinsic dopants or defects , which could avoid the above - mentioned severe disadvantages .
it was predicted that the 2d mon@xmath0 is an intrinsic magnetic material with its magnetic ordering tunable by strain.@xcite but monolayer mon@xmath0 has not been experimentally realized yet .
another interesting proposal is to achieve tunable ferromagnetism in monolayer gase via gate controlled hole - doping.@xcite the resulting spin - polarization can be large , however , the magnetic transition ( curie ) temperature is still low ( @xmath6 k ) , not high enough for practical applications . it hence remains a challenge to find a dopant - free 2d material with room - temperature magnetism .
in addition , these magnetic 2d crystals consist of multiple atomic layers , not being atomically thick . one may further ask : can we push the 2d magnetism down to one - atom - thickness ? in this work , we address the above question by proposing room - temperature magnetism in holey 2d c@xmath0n ( @xmath12d - c@xmath0n ) monolayer crystal with single - atom - thickness via purely electron doping that can be readily achieved by gating .
@xmath12d - c@xmath0n is a newly synthesized 2d material .
experimentally , monolayer @xmath12d - c@xmath0n has been successfully isolated and transferred onto insulating substrate.@xcite it shows a direct bandgap @xmath71.96 ev measured by optical spectroscopy , and a field - effect - transistor fabricated based on it exhibits a high on / off ratio of @xmath8.@xcite without charge doping , @xmath12d - c@xmath0n is non - magnetic . here , using first - principles calculations , we show that ferromagnetic ordering can be easily induced in @xmath12d - c@xmath0n monolayer by electron doping , and remarkably , in a wide doping range ( @xmath9/@xmath4 to @xmath10/@xmath4 ) the material is in fact a half - metal with fully spin - polarized carriers for conduction .
such doping levels can be readily achieved with currently available gating techniques , and gating provides a highly - efficient method to control magnetism in this material .
the curie temperature is estimated up to 320 k. furthermore , we find that the magnetism can also be effectively tuned by lattice strain : starting from a half - metal state , a uniaxial strain @xmath11 can completely turn off the magnetism . to our best knowledge
, this is the first time that room - temperature half - metallic ferromagnetism is ever proposed in a one - atom - thin 2d material via an intrinsic approach , without magnetic dopants or defects .
these important advantages the atomic thickness ,
the high curie temperature , the easy control via gating or strain , and the experimental availability of the material will make @xmath12d - c@xmath0n a highly promising platform to explore nanoscale magnetism and the many electronic and spintronic device applications .
our first - principles calculations are based on the density functional theory ( dft ) using the projector augmented wave ( paw ) method@xcite as implemented in the vienna @xmath12-@xmath13 simulation package ( vasp).@xcite the generalized gradient approximation with the perdew - burke - ernzerhof ( pbe ) realization@xcite is used for the exchange - correlation functional .
our main results are further checked by the hybrid functional ( hse06)@xcite approach .
the kinetic energy cutoff is set to be 550 ev . the lattice constants and the ionic positions are fully optimized until the force on each ion is less than 0.01 ev / .
the convergence criteria for energy is chosen as 10@xmath14 ev . and
a vacuum layer of 18 thickness is added to avoid artificial interactions between periodic images .
the effect of charge - doping is simulated by changing the total number of electrons in the unit cell , with a compensating jellium background of opposite charge to maintain the neutrality . in calculating magnetic properties ,
since magnetism has sensitive dependence on the states at fermi level @xmath15 , we use a refined @xmath16-mesh of size 27@xmath227@xmath21 to ensure the convergence of result .
the lattice structure of monolayer @xmath12d - c@xmath0n is shown in figure [ fig1 ] , which may be viewed as a honeycomb lattice of benzene rings connected through nitrogen atoms .
the lattice is completely flat , with only one atom thickness .
the primitive unit cell consists of 12 c and 6 n atoms ( see figure [ fig1](b ) ) , making a c@xmath0n stoichiometry .
the optimized lattice parameter that we obtain is 8.330 , which agrees well with the experiment measurement and previous theoretical studies.@xcite the structure involves three types of bonds , as indicated in figure [ fig1](b ) .
we find that c - n bond @xmath17 , c - c(1 ) bond @xmath18 , and c - c(2 ) bond @xmath19 .
2d - c@xmath0n monolayer .
the structure is flat with single - atom - thickness .
( b ) primitive cell of @xmath12d - c@xmath0n monolayer , in which the three different bond lengths are labeled . ( c )
schematic picture of the first brillouin zone .
, scaledwidth=100.0% ] the calculated electronic band structure with zero doping is shown in figure [ fig2](a ) .
it shows a semiconducting phase with a direct bandgap @xmath20 ev at the @xmath21 point , in good agreement with the previous dft results and the experimental value ( @xmath22 ev).@xcite the conduction band minimum ( cbm ) at the @xmath21 point involves three nearly degenerate states , and one band has almost flat dispersion along the @xmath21-m direction ( see figure [ fig2](b ) ) .
this leads to a prominent van hove singularity in the density of states ( dos ) at the cbm . in the right panel of figure [ fig2](a )
, we show the projected dos .
one observes that the states contributing to the van hove singularity at cbm are mainly from the @xmath23 orbitals of c and n atoms , i.e. , from the @xmath24 electrons .
nearly flat bands and van hove singularity also appear at the valence band maximum ( vbm ) in the pbe band structure , where the states are mainly consisting of @xmath25 and @xmath26 orbitals .
we have repeated the calculation by using the hybrid functional ( hse06 ) ( see supporting information ) .
the bandgap at hse06 level is larger ( @xmath272.45 ev ) , as one should expect .
interestingly , around vbm , we find that the hse result has a different band ordering , making the van hove singularity pushed below the vbm .
nevertheless , both pbe and hse functionals give almost the same band structure around cbm .
thus , to avoid any functional - dependent feature , we will focus on the cbm and consider the case of electron - doping , and our subsequent presentation will be based on the pbe result ( main results are also confirmed by hse calculations ) .
2d - c@xmath0n at zero doping along with the orbital projected density of states ( dos ) .
( b ) 3d plot of the band dispersion near the conduction band edge , exhibiting almost flat dispersion along the @xmath21-m directions which contributes to the pronounced van hove singularity as shown in ( a ) .
, scaledwidth=100.0% ] large dos associated with van hove singularities is usually connected with instabilities towards symmetry - breaking phases . at zero doping ( @xmath28 , where @xmath29 is the electron doping density )
, the @xmath12d - c@xmath0n monolayer is in the non - magnetic state .
however , due to the van hove singularity at cbm , we find that the system develops a spontaneous ferromagnetic ordering even with a slight electron doping . in figure [ fig3](a )
, we plot the magnetic moment and the spin polarization energy ( both are per doped electron carrier ) versus the doping density . here
the spin polarization energy is defined as the energy difference between the non - spin - polarized phase and the spin - polarized phase .
one observes that ferromagnetism already appears at a low doping with @xmath30/@xmath4 , and with increasing doping density , the magnetic moment per carrier has a sharp increase , reaching 0.96@xmath31/e at @xmath32/@xmath4 .
the moment is close to 1@xmath31 per carrier in a wide range from @xmath33/@xmath4 to @xmath10/@xmath4 , indicating that the carriers become fully spin - polarized in this doping range . above the range ,
the moment begins to decrease and has a quick drop above @xmath34/@xmath4 . in comparison , the spin polarization energy also increase first and then decrease as a function of @xmath29 , however , it does not show a plateau region like the magnetic moment .
it has a peak around @xmath35/@xmath4 with a value of @xmath36 mev per electron carrier .
this value is quite large , almost three times of that in monolayer gase,@xcite suggesting that the ferromagnetic ordering here would be more stable .
the spin polarization energy drops to zero above @xmath37/@xmath4 , indicating that the system returns to non - magnetic at higher doping .
( d ) spin - resolved density of states ( dos ) corresponding to the band structure in ( c ) . here
the fermi level is taken to be energy zero .
, scaledwidth=100.0% ] in figure [ fig3](b ) , we plot the spin polarization of the carriers @xmath38 , where @xmath39 ( @xmath40 ) is the population of spin - up ( spin - down ) carriers .
one clearly observes that the carriers are 100% spin - polarized in the doping range from @xmath33/@xmath4 to @xmath10/@xmath4 , where the system becomes half - metallic . as a representative ,
we take @xmath41/@xmath4 in this range , and plot the corresponding band structure in figure [ fig3](c ) and the spin - resolved dos in figure [ fig3](d ) .
compared with the undoped ( non - magnetic ) case , one observes that the shape of the band dispersion is more or less unchanged .
the main difference is that the bands are now spin - split , and the fermi level cuts through only one spin species .
the energy splitting between the two spin bands around cbm is about 0.1 ev , corresponding to an effective zeeman splitting from an external magnetic field of @xmath7860 t if one assumes a @xmath42 factor of 2 and ignores the orbital contribution to the magnetic moment .
such half - metal state with 100% spin polarization allows fully spin - polarized transport , hence it is very much desired for spintronics applications .
our result shows that half - metallic ferromagnetism can indeed be realized in a material with only one atom thickness without extrinsic dopant atoms or defects .
the emergence of ferromagnetism here can be explained in the stoner picture,@xcite which has been adopted to explain the magnetism in several @xmath43-orbital systems , particularly in so - called `` @xmath44 ferromagnetic semiconductors''.@xcite in this model , ferromagnetism would appear spontaneously if the stoner criterion @xmath45 is satisfied . here
@xmath46 is the strength of the magnetic exchange interaction , and @xmath47 is the dos at the fermi level of the non - spin - polarized band structure .
when the stoner parameter @xmath48 is greater than 1 , the relative gain in exchange energy would dominate over the loss in kinetic energy , and hence the system would favor a ferromagnetic ground state . indeed , in our dft calculation , we find that the stoner parameter is larger than 1 in a wide range from low doping @xmath49/@xmath4 up to @xmath50/@xmath4 ( its value is about 4 in the half - metallic range ) .
this analysis indicates that the huge dos associated with the van hove singularity plays a key role in the appearance of ferromagnetism .
as we mentioned , the electron doping level can be conveniently tuned by the available gating technique .
for example , the doping density on the order of @xmath5/@xmath4 has been demonstrated in graphene by ion liquid gating,@xcite and of @xmath51/@xmath4 in transition metal dichalcogenides by back - gate gating.@xcite owing to its atomic - thickness , we expect that doping range discussed in previous paragraphs can be readily achieved in @xmath12d - c@xmath0n monolayers . we have also estimated the magnetic transition ( curie ) temperature . at the mean field level , this can be done by evaluating the magnetic moment through minimizing the electronic free energy at a given temperature.@xcite we find that the estimated curie temperature can be up to 320 k at the doping density of @xmath10/@xmath4 , correlated with the large spin - polarization energy discussed in figure [ fig3](a ) .
thus , our calculation demonstrates that a room - temperature half - metallic ferromagnetism can be achieved in an existing atom - thin 2d material , and this magnetism can be readily turned on and off via gating .
n monolayer as the channel material .
( a ) without gate voltage , the channel is insulating .
no current can pass through .
( b ) by applying a gate voltage , the c@xmath0n monolayer is driven to the half - metallic phase .
hence a fully spin - polarized current can flow through.,scaledwidth=100.0% ] an experimental setup that can be used to probe the above predicted property is illustrated in figure [ fig4 ] .
it takes a standard field effect transistor configuration , with a @xmath12d - c@xmath0n monolayer as the channel material . without applied gate voltage ( figure [ fig4](a ) )
, the channel is a good semiconductor , and no current can pass through .
hence the transistor is in the off state .
when a gate voltage is applied ( figure [ fig4](b ) ) , the @xmath12d - c@xmath0n monolayer becomes half - metallic .
then current can flow through , and most importantly , the current is fully spin - polarized .
the spin - polarization can be measured by using magnetic electrode or by subsequent spin filtering .
in fact , the device proposed here can be well incorporated into spintronics circuits as a controlled spin generator .
it has the advantages of compactness , small operating voltage , and low energy consumption .
2d - c@xmath0n monolayer at zero doping under a uniaxial strain @xmath52 of @xmath53 .
( b ) magnetic moment per carrier versus strain @xmath52 for several different doping densities . here
the strain @xmath52 is applied along the armchair direction.,scaledwidth=100.0% ] besides gating , the magnetic property may also be tuned by lattice strain .
2d materials typically can sustain large strains , e.g. , graphene , mos@xmath0 , and phosphorene can sustain strains above @xmath54.@xcite a previous computational study has shown that the critical strain of @xmath12d - c@xmath0n monolayer is larger than @xmath55.@xcite in figure [ fig5](a ) , we plot the band structure of @xmath12d - c@xmath0n monolayer without doping under a uniaxial 8% strain along the armchair direction . compared with the unstrained band structure in figure [ fig2](a )
, one observes that the initially degenerate states at the cbm split due to the reduced symmetry , and a more dispersed band is pushed below the flat band , forming the new cbm . as a result , the original van hove singularity is higher in energy and the dos at the new cbm is much decreased .
according to the stoner model , this would reduce the tendency towards ferromagnetism under electron doping . indeed , as shown in figure [ fig5](b ) , starting from an initially half - metallic state , e.g. , with @xmath33/@xmath4 , by increasing uniaxial strain , the magnetic moment per carrier is decreased continuously , and is fully suppressed above 8% strain .
similar trend is observed at other doping densities and also for strains along the zigzag direction ( see supporting information ) .
experimentally , strain can be applied by stretching or bending a flexible substrate on which the 2d material is attached , or it can be controlled electrically by using a piezoelectric substrate.@xcite it provides a powerful alternative method to control the magnetism in @xmath12d - c@xmath0n monolayers .
we discuss several points before concluding .
first , in the above calculation , we take the spin - polarization direction to be perpendicular to the 2d plane . due to the very weak spin - orbit coupling ( soc ) strength of the light elements
c and n , the material exhibits very weak magneto - crystalline anisotropy .
we have performed the calculation including soc and found that the out - of - plane polarized state and the in - plane polarized state have negligible energy difference ( less than @xmath56 mev / e ) .
this shows that the material is a soft ferromagnet , with a low magnetic coercivity .
this could be an important advantage for its use as dynamic layers in nano - magnetic devices , which leads to low energy dissipation during magnetic reversal processes .
second , a recent theoretical study proposed to realize magnetism in @xmath12d - c@xmath0n by adsorption of magnetic atoms.@xcite from the analysis here , we can see that not only the localized magnetic moments of the adsorbed magnetic atoms contribute to the magnetism , the possible charge doping could also contribute .
in fact , the former factor is not a necessity for the induced magnetism in the current system .
we have checked that even by adsorption of alkali atoms , due to the charge doping , the @xmath12d - c@xmath0n monolayer still becomes magnetic ( see supporting information ) . in addition , as we mentioned in the introduction part , introducing extrinsic dopant atoms has several severe drawbacks .
the gate - controlled charge doping proposed here would be a more preferred approach .
finally , when @xmath12d - c@xmath0n monolayers are stacked along the perpendicular direction , they form multilayer structures bonded by weak van der waals interactions.@xcite hence one expects that the doping induced magnetism may also appear for @xmath12d - c@xmath0n multilayers .
it is also possible to combine @xmath12d - c@xmath0n with other 2d material into van der waals heterostructures,@xcite in which the interaction between magnetism and other possible orderings and carrier dynamics could lead to intriguing effects and device functionalities .
in conclusion , by using first - principles calculations , we show that tunable ferromagnetism and half - metallicity can be generated in the recently synthesized c@xmath0n-@xmath12d monolayers via charge doping controlled by electric gating .
the half - metallic state can be realized in a wide range of doping density from 4@xmath210@xmath3/@xmath4 to 8@xmath210@xmath3/@xmath4 , readily achievable by the existing gating technique .
the curie temperature is estimated up to 320 k. in addition , the magnetism can be effectively tuned using lattice strain .
this is the first time that room - temperature half - metallic magnetism is ever proposed in a 2d material with one - atom - thickness , which opens great opportunity to explore nanoscale magnetism and various electronic / spintronic device applications .
the authors thank n. li and d. l. deng for valuable discussions .
this work was supported by the most project of china ( nos .
2014cb920903 , 2016yfa0300603 ) , the national natural science foundation of china ( grant nos . 11574029 , and 11225418 ) , and singapore moe academic research fund tier 1 ( sutd - t1 - 2015004 ) .
butler s z , hollen s m , cao l , cui y , gupta j a , gutirrez h r , heinz t f , hong s s , huang j , ismach a f , johnston - halperin e , kuno m , plashnitsa v v , robinson r d , ruoff r s , salahuddin s , shan j , shi l , spencer m g , terrones m , windl w and goldberger j e 2013 _ acs nano _ * 7(4 )
* 2898 - 2926 http://dx.doi.org/10.1021/nn400280c stoner e c 1938 _ proc .
a _ * 165 * 372 - 414 http://dx.doi.org/ 10.1098/rspa.1938.0066[http://dx.doi.org/ 10.1098/rspa.1938.0066 ] ; 1939 _ proc .
a _ * 169 * 339 - 371 http://dx.doi.org/ 10.1098/rspa.1939.0003[http://dx.doi.org/ 10.1098/rspa.1939.0003 ]
|
exploring two - dimensional ( 2d ) materials with magnetic ordering is a focus of current research .
it remains a challenge to achieve tunable magnetism in a material of one - atom - thickness without introducing extrinsic magnetic atoms or defects . here ,
based on first - principles calculations , we propose that tunable ferromagnetism can be realized in the recently synthesized holey 2d c@xmath0n ( @xmath12d - c@xmath0n ) monolayer via purely electron doping that can be readily achieved by gating .
we show that owing to the prominent van hove singularity in the band structure , the material exhibits ferromagnetism even at a small doping level .
remarkably , over a wide doping range of 4@xmath210@xmath3/@xmath4 to 8@xmath210@xmath3/@xmath4 , the system becomes half - metallic , with carriers fully spin - polarized .
the estimated curie temperature can be up to 320 k. besides gating , we find that the magnetism can also be effectively tuned by lattice strain .
our result identifies @xmath12d - c@xmath0n as the first material with single - atom - thickness that can host gate - tunable room - temperature half - metallic magnetism , suggesting it as a promising platform to explore nanoscale magnetism and flexible spintronic devices .
|
in this paper , we propose a simple analytical approximation for backward stochastic differential equations ( bsdes ) .
these equations were introduced by bismut ( 1973 ) @xcite for the linear case and later by pardoux and peng ( 1990 ) @xcite for the general case , and have earned strong academic interests since then .
they are particularly relevant for the pricing of contingent claims in constrained or incomplete markets , and for the study of recursive utilities as presented by duffie and epstein ( 1992 ) @xcite .
for a recent comprehensive study with financial applications , one may consult yong and zhou ( 1999 ) @xcite , ma and yong ( 2000 ) @xcite and references therein .
the importance of bsdes , or more specifically non - linear fbsdes which have non - linear generators coupled with some state processes satisfying the forward sdes , has risen greatly in recent years also among practitioners .
the collapse of major financial institutions followed by the drastic reform of regulations make them well aware of the importance of counterparty risk management , credit value adjustments ( cva ) in particular . even in a very simple setup ,
if there exists asymmetry in the credit risk between the two parties , the relevant dynamics of portfolio value follows a non - linear fbsde as clearly shown by duffie and huang ( 1996 ) @xcite .
we have recently found that the asymmetric treatment of collateral between the two parties also leads to a non - linear fbsde @xcite .
furthermore , in may 2010 , regulators were forced to realize the importance of mutual interactions and feedback loops in the trading activities among financial firms , shocked by the astonishing flash crash of the dow jones index by almost @xmath0 points .
once we take the feedback effects from the behavior of major players into account , we naturally end up with complicated coupled fbsdes .
unfortunately , however , an explicit solution for a fbsde is only known for a simple linear example . in the last decade ,
several techniques have been introduced by researchers , but they tend to be quite complicated for practical applications .
they either require one to solve non - linear pdes , which are very difficult in general , or resort to quite time - consuming simulation .
although regression based monte carlo simulation has been rather popular among practitioners for the pricing of callable products , the appropriate choice of regressors and attaining numerical stability becomes a more subtle issue for a general fbsde .
in fact , in clear contrast to the pricing of callable products , one can not tell if the price goes up or down when one improves the regressors , which makes it particularly difficult to select the appropriate basis functions . in this paper
, we present a simple analytical approximation scheme for the non - linear fbsdes coupled with generic markovian state processes .
we have perturbatively expanded the non - linear terms around the linearized fbsde , where the expansion can be made recursively to an arbitrary higher order . in each order of approximation , the required calculations are equivalent to those for standard european contingent claims . in order to carry out the perturbation scheme , we need to express the backward components explicitly in terms of the forward components for each order of approximation . for that purpose , we propose to use the asymptotic expansion of volatility for the forward components , which is now widely adopted to price various european contingent claims and compute optimal portfolios ( see , for examples @xcite and references therein for the recent developments and review . ) . in the case
when the underlying processes have known distributions , of course , we can directly proceed to a higher order approximation without resorting to an asymptotic expansion .
we have also studied a perturbation scheme in the pde framework , or in the so - called four step scheme @xcite , for the generic fully - coupled non - linear fbsdes .
we have shown that our perturbation method renders the original non - linear pde into the series of classical linear parabolic pdes , which are easy to handle with standard techniques .
we then provided the corresponding probabilistic framework by using the equivalence between the two approaches .
we have shown that , also in this case , the required calculations in a given order are equivalent to those for the classical european contingent claims . as a by - product , by applying the asymptotic expansion method to the corresponding probabilistic model ,
it was actually found possible to derive an _ analytic expression _ for the solution of the non - linear pde up to a given order of perturbation .
therefore , our method can be interpreted as a practical implementation of the four step scheme in the perturbative approach .
the organization of of the paper is as follows : in section 2 , we will explain our new approximation scheme with perturbative expansion for generic decoupled fbsdes .
then , in section 3 , we shall apply it to the two concrete examples to demonstrate how it works and test its numerical performance .
one of them allows a direct numerical treatment by a simple pde and hence it is easy to compare the two methods . in the second example , we will consider a slightly more complicated model .
we compare our approximation result to the detailed numerical study recently carried out by gobet et al .
( 2005 ) @xcite using a regression - based monte carlo simulation . in section 4 ,
we explain how to use standard asymptotic expansion procedures to express the backward components explicitly when the forward components do not have known distributions . in section 5 , we will give an extension of our method to the fully coupled non - linear fbsdes under the pde framework , and then formulate the equivalent probabilistic approach in section 5 .
appendix contains slightly different scheme for coupled non - linear fbsdes which may be useful for the actual application .
let us briefly describe the basic setup .
the probability space is taken as @xmath1 and @xmath2 denotes some fixed time horizon .
@xmath3 , @xmath4 is @xmath5-valued brownian motion defined on @xmath6 , and @xmath7 stands for @xmath8-augmented natural filtration generated by the brownian motion .
we consider the following forward - backward stochastic differential equation ( fbsde ) dv_t&=&-f(x_t , v_t , z_t)dt+z_tdw_t + v_t&=&(x_t ) where @xmath9 takes the value in @xmath10 , and @xmath11 is assumed to follow a generic markovian forward sde dx_t=_0(x_t)dt+(x_t)dw_t . here
, we absorbed an explicit dependence on time to @xmath12 by allowing some of its components can be a time itself .
@xmath13 denotes the terminal payoff where @xmath14 is a deterministic function of @xmath15 .
the following approximation procedures can be applied in the same way also in the presence of coupon payments . @xmath16 and
@xmath17 take values in @xmath5 and @xmath18 respectively , and `` @xmath19 '' in front of the @xmath20 represents the summation for the components of @xmath21-dimensional brownian motion . throughout this paper
, we are going to assume that the appropriate regularity conditions are satisfied for the necessary treatments . in order to solve the pair of @xmath22 in terms of @xmath23
, we extract the linear term from the generator @xmath24 and treat the residual non - linear term as the perturbation to the linear fbsde .
we introduce the perturbation parameter @xmath25 , and then write the equation as [ mod_bsde1 ] dv_t^()&=&c(x_t)v_t^()dt - g(x_t , v_t^(),z_t^())dt+z_t^()dw_t + v_t^()&=&(x_t ) , [ mod_bsde2 ] where @xmath26 corresponds to the original model by as simply a parameter convenient to count the approximation order . the actual quantity that should be small for the approximation
is the residual part @xmath27 . ]
f(x_t , v_t , z_t)=-c(x_t)v_t+g(x_t , v_t , z_t ) .
usually , @xmath28 corresponds to the risk - free interest rate at time @xmath29 , but it is not a necessary condition .
one should choose the linear term in such a way that the residual non - linear term becomes as small as possible to achieve better convergence .
a possible linear term @xmath30 in the driver @xmath24 can be absorbed by the measure change and hence the simple reinterpretation of the drift term of the forward components @xmath31 results in the form ( [ mod_bsde1 ] ) .
see also the discussion in appendix .
now , we are going to expand the solution of bsde ( [ mod_bsde1 ] ) and ( [ mod_bsde2 ] ) in terms of @xmath25 : that is , suppose @xmath32 and @xmath33 are expanded as v_t^()&= & v_t^(0)+v_t^(1)+^2 v_t^(2)+ + z_t^()&= & z_t^(0)+z_t^(1)+^2 z_t^(2)+ .
once we obtain the solution up to the certain order , say @xmath34 for example , then by putting @xmath26 , _
t= _ i=0^k v_t^(i ) , _
t= _ i=0^k z_t^(i ) is expected to provide a reasonable approximation for the original model as long as the residual term is small enough to allow the perturbative treatment . as we will see , @xmath35 and @xmath36 , the corrections to each order can be calculated recursively using the results of the lower order approximations .
for the zero - th order of @xmath25 , one can easily see the following equation should be satisfied : dv_t^(0)&=&c(x_t)v_t^(0)dt+z_t^(0)dw_t v_t^(0)&=&(x_t ) .
it can be integrated as v_t^(0)=e[v0formula ] which is equivalent to the pricing of a standard european contingent claim .
since we have e^-_0^t c(x_s)ds(x_t)=v_0^(0)+_0^t e^-_0^u c(x_s)dsz_u^(0)dw_u it can be shown that , by applying malliavin derivative @xmath37 , _
t(e^-_0^t c(x_s)ds(x_t))=_t^t _ t ( e^-_0^u c(x_s)dsz_u^(0))dw_u + e^-_0^t c(x_s)dsz_t^(0 ) .
thus , by taking conditional expectation @xmath38 $ ] , we obtain z_t^(0)=e .
now , let us consider the process @xmath39 .
one can see that its dynamics is governed by [ 1stv ] d(v_t^()-v_t^(0))&=&c(x_t)(v_t^()-v_t^(0 ) ) -g(x_t , v_t^(),z_t^())dt+(z_t^()-z_t^(0))dw_t + v_t^()-v_t^(0)&=&0 .
now , by extracting the @xmath25-first order terms , we can once again recover the linear fbsde dv_t^(1)&=&c(x_t)v_t^(1)dt - g(x_t , v_t^(0),z_t^(0))dt+z_t^(1)dw_t + v_t^(1)&=&0 , which leads to v_t^(1)=e straightforwardly . by the same arguments in the zero - th order example
, we can express the volatility term as z_t^(1)=e . from these results
, we can see that the required calculation is nothing more difficult than the zero - th order case as long as we have explicit expression for @xmath40 and @xmath41 .
we can proceed the same way to the second order correction . by extracting the @xmath25-second order terms from @xmath42
, one can show that dv_t^(2)&=&c(x_t)v_t^(2)dt-(g(x_t , v_t^(0),z_t^(0))v_t^(1 ) + _ z g(x_t , v_t^(0),z_t^(0))z_t^(1))dt+z_t^(2)dw_t + v_t^(2)&=&0 is a relevant fbsde , which is once again linear in @xmath43 . as before
, it leads to the following expression straightforwardly : v_t^(2)&=&e + + z_t^(2)&=&e .
+ in the above calculation , we have assumed the driver function is differentiable .
if this is not the case , we need to approximate it using some smooth function or apply integration- by - parts technique for generalized wiener functionals ( e.g. a composite functional of dirac delta fucntion and a smooth wiener functional ) . in exactly the same way
, one can derive an arbitrarily higher order correction . due to the @xmath25 in front of the non - linear term @xmath27 ,
the system remains to be linear in the every order of approximation .
however , in order to carry out explicit evaluation , we need to give malliavin derivative explicitly in terms of the forward components . we will discuss this issue in the next .
firstly , let us introduce a @xmath44 matrix process @xmath45 , for @xmath46 $ ] , as the solution for the following forward sde : [ ysde1 ] d(y_t , u)^i_j&=&_k=1^d ( _ k ^i_0(x_u)(y_t , u)^k_j du+ _
k ^i(x_u)(y_t , u)^k_j dw_u ) + ( y_t , t)^i_j&=&^i_j , [ ysde2 ] where @xmath47 denotes the differential with respect to the @xmath34-th component of @xmath12 , and @xmath48 denotes kronecker delta .
now , for malliavin derivative , we want to express , for @xmath46 $ ] , ein terms of @xmath23 , where @xmath49 is a some deterministic function of @xmath12 , in general . thank to the known chain rule of malliavin derivative , we have _ t(e^-_t^uc(x_s)dsg(x_u))&=&_i=1^d\{e^-_t^u
c(x_s)ds_i g(x_u ) ( _ t x^i_u ) + & & -e^-_t^u c(x_s)dsg(x_u)(_t^u _
i c(x_s)(_t x^i_s)ds)}. thus , it is enough for our purpose to evaluate @xmath50 . since we have ( _ t x^i_u)=_k=1^d(_t^u _
k^i_0(x_s)(_t x^k_s)ds+ _ t^u _ k^i(x_s)(_t x^k_s)dw_s ) + ^i(x_t ) it can be shown that @xmath51 follows the next sde : d(_tx_u^i)&=&_k=1^d(_k^i_0(x_u)(_t x_u^k)du+_k^i(x_u)(_tx_u^k)dw_u ) + ( _ t x^i_t)&=&^i(x_t ) .
thus , comparing to eqs .
( [ ysde1 ] ) and ( [ ysde2 ] ) , we can conclude that ( _ t x_u)=y_t , u ( x_t ) .
[ mdy ] as a result , combining the sde for @xmath45 and the markovian property of @xmath12 , one can confirm that the conditional expectation eis actually given by a some function of @xmath23 .
therefore , in principle , both of the backward components can be expressed in terms of @xmath23 in each approximation order .
in fact , this is an easy task when the underlying process has a known distribution . in the next section ,
we present two such models , and demonstrate how our approximation scheme works .
we will also compare our approximate solution to the direct numerical results obtained from , such as pde and monte carlo simulation .
however , in more generic situations , we do not know the distribution of @xmath12 .
we will explain how to handle the problem in this case using the asymptotic expansion method for the forward components in sec.[asymptotic ] .
as the first example , we consider a toy model for a forward agreement on a stock with bilateral default risk of the contracting parties , the investor ( party-@xmath52 ) and its counterparty ( party-@xmath53 ) .
the terminal payoff of the contract from the view point of the party-@xmath52 is ( s_t)=s_t - k where @xmath54 is the maturity of the contract , and @xmath55 is a constant . we assume the underlying stock follows a simple geometric brownian motion : ds_t = rs_t dt+s_t dw_t where the risk - free interest rate @xmath21 and the volatility @xmath56 are assumed to be positive constants .
the default intensity of party-@xmath57 @xmath58 is specified as h_1= , h_2=+h where @xmath59 and @xmath60 are also positive constants . in this setup ,
the pre - default value of the contract at time @xmath29 , @xmath61 , follows dv_t&=&rv_t dt - h_1(-v_t,0)dt+h_2(v_t,0)dt+z_tdw_t + & = & ( r+)v_t dt+h(v_t,0)dt + z_t dw_t + v_t&=&(s_t ) .
now , following the previous arguments , let us introduce the expansion parameter @xmath25 , and consider the following fbsde : dv_t^()&=&v_t^()dt -g(v_t^())dt+z_t^()dw_t + v_t^()&=&(s_t ) + ds_t&=&s_t(rdt+dw_t ) , where we have defined @xmath62 and @xmath63 . in the zero - th order , we have dv_t^(0)&=&v_t^(0)dt+z_t^(0)dw_t + v_t^(0)&=&(s_t ) .
hence we simply obtain v_t^(0)&=&e + & = & e^-(t - t)(s_t e^r(t - t)-k ) and z_t^(0)=e^-(t - t)s_t . in the first order
, we have dv_t^(1)&=&v_t^(1)dt - g(v_t^(0))dt+z_t^(1)dw_t + v_t^(1)&=&0 .
thus , we obtain v_t^(1)&=&e + & = & -e^-(t - t)h_t^t edu + & = & -e^-(t - t)h_t^t c(u;t , s_t)du , where c(u;t , s_t)&=&s_te^r(t - t)n(d_1(u;t , s_t))-kn(d_2(u;t , s_t ) ) + d_1(2)(u;t , s_t)&=&(()^2 ( u - t ) ) , and @xmath64 denotes the cumulative distribution function for the standard normal distribution .
we can also derive z_t^(1)=-e^-(t - t)hs_t_t^t n(d_1(u;t , s_t))du .
finally , let us consider the second order value adjustment . in this case ,
the relevant dynamics is given by dv_t^(2)&=&v_t^(2)dt - g(v_t^(0))v_t^(1)dt+z_t^(2)dw_t + v_t^(2)&=&0 .
as a result , we have v_t^(2)&=&e + & = & e^-(t - t)h^2 _ t^t _ u^t eds du which can be evaluated as v_t^(2)=e^-(t - t)h^2 _ t^t _ u^t _ -d_2(u;t , s_t)^(z)c(s;u , s_u(z , s_t))dzdsdu , where we have defined s_u(z , s_t)=s_t e^(r-^2)(u - t)+z and ( z)=e^-z^2 . for this simple model
, we can directly evaluate the contract value @xmath61 by numerically solving the pde : v(t , s)+ ( rsv(t , s)+^2 s^2 v(t , s ) ) -v(t , s)=0 with the boundary conditions v(t , s)&=&s - k + v(t , m)&=&e^-(+h)(t - t)(me^r(t - t)-k ) , mk + v(t , m)&=&e^-(t - t)(me^r(t - t)-k ) , mk . in fig .
[ cva ] , we have plot the numerical results of the forward contract with bilateral default risk with various maturities with the direct solution from the pde .
we have used & & r=0.02,=0.01 , h=0.03 , + & & = 0.2,s_0=100 , where the strike @xmath55 is chosen to make @xmath65 for each maturity .
we have plot @xmath66 for the first order , and @xmath67 for the second order .
note that we have put @xmath26 to compare the original model .
one can observe how the higher order correction improves the accuracy of approximation . in this example , the counterparty is significantly riskier than the investor , and the underlying contract is quite volatile . even in this situation ,
the simple approximation to the second order works quite well up to the very long maturity . in this subsection
, we consider the valuation of self - financing portfolio under the situation where there exists a difference between the lending and borrowing interest rates . here
, we consider the problem under the physical measure .
the dynamics of the self - financing portfolio is governed by @xcite dv_t&=&rv_tdt-\{(r - r)(-v_t,0)-z_t}dt + z_t dw_t + v_t&=&(s_t ) + ds_t&=&s_t(dt+dw_t ) where @xmath21 and @xmath68 are the lending and the borrowing rate , respectively .
@xmath69 denotes the risk premium . for simplicity
, we assume all of the @xmath21 , @xmath68 , @xmath70 and @xmath56 are positive constants . here
, @xmath71 represents the amount invested in the risky asset , i.e. stock @xmath72 .
let us choose the terminal wealth function as ( s_t)=(s_t - k_1,0)-2(s_t - k_2,0 ) .
this spread introduces both of the lending and borrowing activities , which makes the problem more interesting .
the setup explained here is in fact exactly the same as that of adopted by gobet et al .
( 2005 ) @xcite .
they have carried our detailed numerical studies for the above problem and evaluate @xmath73 by regression - based monte carlo simulation . in the following , we will apply our perturbative approximation scheme to the same problem and test its accuracy . as usual , let us introduce the expansion parameter as dv_t^()&=&rv_t^()dt - g(v_t^(),z_t^())dt+z_t^()dw_t + v_t^()&=&(s_t ) , where we have defined the non - linear perturbation function as g(v , z)=(r - r)(-v,0)-z .
now , we are going to expand @xmath32 in terms of @xmath25 .
in the zero - th order , the bsde reduces to dv_t^(0)&=&rv_t^(0)dt+z_t^(0)dw_t + v_t^(0)&=&(s_t ) , which allows us to obtain v_t^(0)&=&e + & = & e^-r(t - t)(c(k_1,s_t)-2c(k_2,s_t ) ) , where we have defined c(k_i , s_t)&=&s_te^(t - t)n(d_1(k_i , s_t))-k_i n(d_2(k_i , s_t ) ) + d_1(2)(k_i , s_t)&=&\ { ( ) ^2 ( t - t ) } for @xmath74 .
the volatility term is given by z_t^(0)=e^(-r)(t - t)s_t(n(d_1(k_1,s_t))-2n(d_2(k_2,s_t ) ) ) .
now , in the first order , we have dv_t^(1)&=&rv_t^(1)dt - g(v_t^(0),z_t^(0))dt+z_t^(1)dw_t + v_t^(1)&=&0 . as before , we can easily integrate it to obtain v_t^(1)=e .
now , using the zero - th order results , one can show g(v_t^(0),z_t^(0))&=&e^-r(t - t)(r - r)\{k_1n(d_2(k_1,s_t))-2k_2n(d_2(k_2,s_t))}^+ + & -&e^-r(t - t)(-r)s_te^(t - t)\{n(d_1(k_1,s_t))-2n(d_1(k_2,s_t ) ) } , which leads to v_t^(1)&=&e^-r(t - t)_t^t du \
+ by setting s_u(s_t , z)=s_te^(u - t)(-^2 ( u - t)+z ) we can write the first order correction as & & v_t^(1)=e^-r(t - t)_t^t du_dz ( z ) + & & \{(r - r)^+
. + & & -.(-r)s_u(s_t , z)e^(t - u ) } .
+ the volatility term can also be derived easily as z_t^(1)&=&s_t v_t^(1)(s_t ) + & = & e^-r(t - t)_t^t du_dz(z ) , where we have defined ( u , z)&=&1 k_1n(d_2(k_1,s_u(s_t , z)))-2k_2n(d_2(k_2,s_u(s_t , z)))0 + & = & 0 . finally , in the second order , the relevant fbsde is given by dv_t^(2)&=&rv_t^(2)dt-(g(v_t^(0),z_t^(0))v_t^(1 ) + g(v_t^(0),z_t^(0))z_t^(1))dt+z_t^(2)dw_t + v_t^(2)&=&0 .
using the fact that g(v , z)&=&-(r - r)_\{-v0 } + g(v , z)&=&_\{-v0}-we obtain & & v_t^(2)=e^-r(t - t)_t^t du_u^t ds _
dz_1_dz_2 ( z_1)(z_2 ) . here , we have defined s_s(u,)&=&s_s(s_u(s_t , z_1),z_2 ) and also ( s , u,)&=&1 k_1n(d_2(k_1,s_s(u,)))-2k_2n(d_2(k_2,s_s(u,)))0 + & = & 0 .
if one needs , it is also straightforward to derive the volatility component .
gobet et al .
( 2005 ) @xcite have carried out the detailed numerical study for the above problem using the regression - based monte carlo simulation .
they have used & & = 0.05 , = 0.2,r=0.01 , r=0.06 + & & t=0.25 , s_0=100 , k_1=95,k_2=105 .
after trying various sets of basis functions , they have obtained the price as @xmath75 with standard deviation @xmath76 .
now , let us provide the results from our perturbative expansion .
we have obtained & & v_0^(0)=2.7863 + & & v_0^(1)=0.1814 + & & v_0^(2)=-0.0149 using the same model inputs .
thus , up to the first order , we have @xmath77 , which is already fairly close , and once we include the second order correction , we have @xmath78 , which is perfectly consistent with their result of monte carlo simulation . note
that , we have derived analytic formulas with explicit expressions both for the contract value and its volatility .
in this section , we consider the situation where the forward components @xmath79 consist of the generic markovian processes . in this case
, we can not express @xmath35 and @xmath36 in terms of @xmath79 exactly , which prohibits us from obtaining the higher order corrections in a simple fashion as we have done in the previous section .
however , notice the fact that what we have to do in each order of expansion is equivalent to the pricing of generic european contingent claims and hence we can borrow known techniques adopted there . in the following
, we will explain the use of asymptotic expansion method , now for the forward components .
although it is impossible to obtain the exact result , we can still obtain analytic expression for @xmath80 up to a certain order of the volatilities of @xmath79 .
for the details of asymptotic expansion for volatility , please consult with the works @xcite , for example .
let us introduce a new expansion parameter @xmath81 , which is now for the asymptotic expansion for the forward components .
we express the relevant sde of generic markovian process @xmath82 as dx_u^()=_0(x_u^(),)du+_a(x_u^(),)dw_u^a . here , we have used einstein notation which assumes the summation of all the paired indexes .
for example , in the above equation , the second term means _
a(x_u^(),)dw_u^a=_a=1^r _ a(x_u^(),)dw_u^a .
we assume _
a(x,0)=0 [ delassumption ] for @xmath83 .
intuitively speaking , it suggests that @xmath84 counts the order of volatility .
suppose that , in the @xmath85-th order of @xmath25 , we succeeded to express @xmath86 and @xmath87 in terms of @xmath88 .
then , in the next order , we can express the backward components as v_t^(i)&=&e + z_t^(i)&=&ewith some function @xmath49 . if there is no need to obtain @xmath89 , we can just run monte carlo simulation for @xmath90 to evaluate these quantities in a standard way .
however , if we want to obtain higher order corrections , we need somehow to express the @xmath80 in terms of @xmath88 .
what we are going to propose here is to expand the backward components around @xmath91 : v_t^(i)&=&v_t^(i,0)+v_t^(i,1)+^2 v_t^(i,2)+o(^2 ) + z_t^(i)&=&z_t^(i,0)+z_t^(i,1)+^2 z_t^(i,2)+o(^2 ) and express each @xmath92 and @xmath93 in terms of @xmath88 up to a certain order `` @xmath94 '' of @xmath84 . although we can proceed to arbitrarily higher order of @xmath84
, we will present explicit expressions up to the second order in this paper . for the interested readers ,
the work @xcite provides the systematic methods to obtain higher order corrections .
+ thank to the well - known chain rule for malliavin derivative , what we have to do is only expanding the two fundamental quantities , @xmath95 and @xmath51 for @xmath46 $ ] , in terms of @xmath81 .
firstly , let us introduce a simpler notation , dx_u^()&=&_0(x_u^(),)du+_a(x_u^(),)dw_u^a + & : = & _ ( x_u^(),)dw_u^ , where @xmath96 runs through @xmath97 to @xmath21 with the convention @xmath98 and @xmath99 for @xmath100 .
we set the time @xmath29-value of @xmath90 as @xmath15 .
thus our goal is to express @xmath101 and @xmath102 as functions of @xmath15 .
we first introduce a @xmath44 matrix process @xmath103 defined as d(y_t , u^())^i_j&=&_k^i_(x_u^(),)(y_t , u^())^k_jdw_u^ + ( y_t , t^())^i_j&=&^i_j , where @xmath47 denotes the differential with respect to the @xmath34-th component of @xmath12 .
since we have ( x_u^())^i = x^i+_t^u ^i_0(x_s^(),)ds+_t^u ^i_a(x_s^(),)dw_s^a applying a malliavin derivative @xmath104 with @xmath105 gives _
t,(x_u^())^i= _ t^u _ k_0^i(x_s^(),)_t,(x_s^())^kds + _ t^u_k^i_a(x_s^(),)_t,(x_s^())^kdw_s^a+ ^i_(x , ) . thus one can show that d(_t,(x_u^())^i)&=&_k^i_(x_u^ ( ) , )
_ t,(x_u^())^kdw_u^ + _ t,(x_t^())^i&=&^i_(x , ) . therefore , for @xmath106 , we conclude that _ t , a ( x_u^())^i=(y^()_t , u)^i_j^j_a(x , ) , [ yandm ] which implies that the asymptotic expansion of @xmath107 can be obtained from that of @xmath103 .
therefore , in the following , we first carry out the asymptotic expansion for @xmath12 and @xmath108 .
we are now going to expand for @xmath46 $ ] as x_u^()=x_u^(0)+d_t , u+^2 e_t , u+o(^2 ) , and y_t , u^()=y_t , u+h_t , u+o ( ) , where d_t , u=.|_=0 , e_t , u=.|_=0 , and y_t , u = y_t , u^(0 ) , h_t , u=.|_=0 .
since @xmath111 for @xmath100 , we have dx_u^(0)&=&_0(x_u^(0),0)du + d(y_t , u)^i_j&=&_k^i_0(x_u^(0),0)(y_t , u)^k_j du with the initial conditions @xmath112 and @xmath113 , which allows us to express @xmath114 and @xmath45 as deterministic functions of @xmath15 .
it is also convenient for later calculations to notice that @xmath115 is the solution of d(y^-1_t , u)^i_j&=&-(y_t , u^-1)^i_k_j_0^k(x_u^(0),0)du [ yinv ] with @xmath116 . by applying @xmath117 ,
we can easily obtain d(_x_u^())^i&=&(_(x_u^(),))^i_j_(x_u^())^jdw_u^+_^i_(x_u^(),)dw_u^ + d(_y^()_t , u)^i_j&=&\ { _ k ^i_(x_u^(),)(_y^()_t , u)^k_j + _ kl^i_(x_u^(),)(_x_u^())^l ( y_t , u^())^k_j + & & + _ k_^i_(x_u^(),)(y_t , u^())^k_j}dw_u^ .
+ putting @xmath91 , they leads to dd^i_t , u&=&_j^i_0(x_u^(0),0)d^j_t , udu+_^i_(x_u^(0),0)dw_u^ + d(h_t , u)^i_j&=&(_k^i_0(x_u^(0),0))(h_t , u)^k_j du + _ kl^i_0(x_u^(0),0))d^l_t , u(y_t , u)^k_j du + & & + ( _ k_^i_(x_u^(0),0))(y_t , u)^k_j dw_u^ .
now , by using eq.([yinv ] ) , one can show that & & d^i_t , u=(y_t , u)^i_j_t^u(y_t , s^-1)^j_k__^k(s)dw_s^ + & & ( h_t , u)^i_j=(y_t , u)^i_k_t^u(y^-1_t , s)^k_l \{(_mn^l_0(s))d^n_t , s(y_t , s)^m_jds+ ( _ m_^l_(s))(y_t , s)^m_jdw_s^ } + where we have defined the shorthand notation that ^i_j(s):=^i_j(x_s^(0),0 ) , which will be used in the following calculations , too .
applying @xmath118 to the sde of @xmath90 gives us d(_^2 x_u^())^i&=&\ { ( _ ( x_u^(),))^i_j_^2 ( x_u^())^j + _ jk^i_(x_u^(),)_(x_u^())^j_(x_u^())^k + & & + 2_j_^i_(x_u^(),)_(x_u^())^j+ _
^2 ^i_(x_u^(),)}dw_u^ .
thus , putting @xmath91 , we obtain de^i_t , u&=&(_0(x_u^(0),0))^i_je^j_t , udu+_jk^i_0(x_u^(0),0)d^j_t , ud^k_t , udu + & & + 2_j_^i_(x_u^(0),0)d^j_t , udw_u^+_^2^i_(x_u^(0),0)dw_u^ .
now we can integrate it as e^i_t , u=(y_t , u)^i_j_t^u(y^-1_t , s)^j_k\ { _ lm^k_0(s)d^l_t , sd^m_t , sds+(2_l_^k_(s)d^l_t , s + _ ^2^k_(s))dw_s^}.
+ we do not need the second order terms for @xmath103 . for convenience ,
let us define ( _ a^i)^()_t , u=(_t x_u^())^i_a and its expansion as ( _ a^i)^()_t , u=(_a^i)^(1)_t , u+^2 ( _ a^i)^(2)_t , u+o(^2 ) , where ( _ a^i)_t , u^(1)=. ( _ a^i)^()_t , u|_=0,(_a^i)_t , u^(2)=.(_a^i)^()_t , u|_=0 .
note that , the zero - th order term @xmath119 vanishes , due to the assumption ( [ delassumption ] ) . from ( [ yandm ] ) , we can easily show that ( _ a^i)^(1)_t , u&=&(y_t , u)^i_j(_^j_a(x,0 ) ) + ( _ a^i)^(2)_t , u&=&(y_t , u)^i_j(_^2 ^j_a(x,0))+ 2(h_t , u)^i_j(_^j_a(x,0 ) ) .
now , we try to express v_t^(i,)=_t^t edu as a function of @xmath121 using the previous results . for that purpose ,
we first need to carry out asymptotic expansion for r_t , u^():=e^-_t^u c(x_s^())dsg(x_u^ ( ) , ) to obtain r^()_t , u = r^(0)_t , u+r^(1)_t , u+^2 r^(2)_t , u+o(^2 ) , where r^(1)_t , u=.r^()_t , u|_=0 , r^(2)_t , u=.r^()_t , u|_=0 .
then , we can take the conditional expectation straightforwardly .
we have r^(0)_t , u = e^-_t^u c(x_s^(0))dsg(x_u^(0),0 ) which is a deterministic function of @xmath15 .
it is easy to obtain _
r^()_t , u&=&e^-_t^u c(x_s^())ds \{-g(x_u^(),)_t^u _
i c(x_s^())(_x_s^())^ids + & & + ( _ i g(x_u^(),))(_x_u^())^i+_g(x_u^ ( ) , ) } , which leads to r^(1)_t ,
u = e^-_t^u c(x_s^(0))ds\ { ( _ i g(x_u^(0),0))d_t , u^i+_g(x_u^(0),0)- g(x_u^(0),0)_t^u_i c(x_s^(0))d^i_t , sds}. + in the same way , we can show that r^(2)_t , u&=&e^-_t^u c(x_s^(0))ds\{(_ijg(x_u^(0),0))d^i_t , ud^j_t , u+2(_i_g(x_u^(0),0))d^i_t , u + & & + ( _ i g(x_u^(0),0))e^i_t , u+_^2 g(x_u^(0),0 ) + & & + ( _ t^u _ ic(x_s^(0))d^i_t , sds)^2g(x_u^(0),0 ) + & & -2(_t^u _
i c(x_s^(0))d^i_t , sds)(_j g(x_u^(0),0)d^j_t , u + _ g(x_u^(0),0 ) ) + & & -(_t^uds ) g(x_u^(0),0 ) } .
evaluation of the conditional expectation can be easily done by simply applying ito - isometry .
let us first define ^i_t , u&=&(y_t , u)^i_j_t^u ( y_t , s^-1)^j_k(_^k_0(s))ds + ^i_t , u&=&(y_t , u)^i_j_t^u ( y_t , s^-1)^j_k(_^k_a(s))dw^a_s and then we have @xmath123 .
since the first one is a deterministic function , we have , for @xmath124 , & : = & e + & = & ^i_t , u^j_t , s+ , where = ( y_t , u)^i_k(y_t , s)^j_l_t^us ( y^-1_t , v)^k_m(y^-1_t , v)^l_n ( _ ^m_a(v))(_^n_a(v))dv .
then , similarly , we can express & & _ t , u^i:=e + & & = ( y_t , u)^i_j_t^u ( y_t , s^-1)^j_k \{_lm^k_0(s ) + 2_l_^k_0(s)^l_t , s+_^2^k_0(s)}ds .
+ using these results , we have & & ^(0)_t , u:=e = e^-_t^u c(x_s^(0))dsg(x_u^(0),0 ) + & & ^(1)_t , u:=e + & & = e^-_t^u c(x_s^(0))ds\ { ( _ i g(x_u^(0),0))^i_t ,
u+_g(x_u^(0),0 ) -g(x_u^(0),0)_t^u _
i c(x_s^(0))^i_t , sds } , + and also & & ^(2)_t , u:=e = e^-_t^u c(x_s^(0))ds\ { + & & _ ijg(x_u^(0),0)+ 2(_i_g(x_u^(0),0))^i_t , u+_i g(x_u^(0),0)^i_t , u +
_ ^2 g(x_u^(0),0 ) + & & + g(x_u^(0),0)_t^u_t^u_i c(x_s^(0))_jc(x_v^(0))dsdv + & & -2_jg(x_u^(0),0)_t^u _ i c(x_s^(0))ds + & & -2_g(x_u^(0),0)_t^u _
i c(x_s^(0))^i_t , sds + & & -g(x_u^(0),0)_t^u ds}. now , we are able to express @xmath122 as a function of @xmath15 up to the second order of @xmath84 as desired : v_t^(i,)=_t^t \ { _ t , u^(0)+^(1)_t , u+ ^2 ^(2)_t , u}du + o(^2 ) . finally , we are going to express z_t^(i,)=_t^t edu as a function of @xmath121 .
let us introduce the two quantities : ( _ a^())_t , u&=&e^-_t^u c(x_s^())ds_i g(x_u^(),)(_t x_u^())^i_a + ( _ a^())_t , u&=&-e^-_t^uc(x_s^())dsg(x_u^ ( ) , ) ( _ t^u _ i c(x_s^())(_t x_s^())^i_a ds ) .
then , we have _
t(e^-_t^u c(x_s^())dsg(x_u^(),))=(_a^())_t , u + ( _ a^())_t , u . similarly to the previous section , we try to obtain the expressions as ( _ a^())_t , u&=&(_a^(1))_t , u+^2 ( ^(2)_a)_t , u+o(^2 ) + ( _ a^())_t , u&=&(_a^(1))_t , u+^2 ( ^(2)_a)_t , u+o(^2 ) where both of the zero - th order terms vanish .
we have & & _ ( ^()_a)_t , u = e^-_t^u c(x_s^())ds\ { -(_t^u _ i c(x_s^())(_x_s^())^ids ) _ jg(x_u^(),)(_t x_u^())^j_a + & & + ( _ ijg(x_u^(),))(_x_u^())^j(_t x_u^())^i_a + ( _ i_g(x_u^(),))(_t x_u^())_a^i+ ( _ i g(x_u^(),))(__t x_u^())^i_a } . + thus , we obtain ( _ a^(1))_t , u&:=&.(_a^())_t , u|_=0 + & = & e^-_t^u c(x_s^(0))ds_ig(x_u^(0),0)(_a^i)^(1)_t , u
. similarly we can show that ( ^(2)_a)_t , u&:=&.(^()_a)_t , u|_=0 + & = & -2(_t^u _ ic(x_s^(0))d_t , s^ids)(_a^(1))_t , u + & & + e^-_t^u c(x_s^(0))ds . + in the same way , for @xmath126 , we have & & _ ( ^_a)_t , u = e^-_t^u c(x_s^())dsg(x_u^ ( ) , ) ( _ t^u ( _ ic(x_s^()))(_x_s^())^ids ) ( _ t^u ( _ j
c(x_s^()))(_t x_s^())^j_a ds ) + & & -e^-_t^u c(x_s^())ds\{_g(x_u^(),)+(_i g(x_u^(),))(_x_u^())^i}(_t^u _ i c(x_s^())(_t x_s^())_a^i ds ) + & & -e^-_t^u c(x_s^())dsg(x_u^ ( ) , ) _
+ thus we can show that ( ^(1)_a)_t , u&:=&.(^_a)_t , u|_=0 + & = & -e^-_t^u c(x_s^0)dsg(x_u^(0),0)_t^u(_i c(x_s^(0)))(_a^i)^(1)_t , sds and similarly & & ( ^(2)_a)_t , u=. ( ^_a)_t , u|_=0 + & & = -2(_t^u ( _ i c(x_s^(0)))d_t , s^i ds)(^(1)_a)_t , u + & & -e^-_t^u c(x_s^(0))ds\ { 2(_g(x_u^(0),0)+(_i g(x_u^(0),0 ) ) d_t , u^i ) _ t^u ( _ j c(x_s^(0 ) ) ) ( ^j_a)^(1)_t , sds + & & + g(x_u^(0),0)_t^uds } . for the evaluation of the conditional expectation , let us define , for @xmath127 , & & ( _ t , u)^i_j:=e + & & = ( y_t , u)^i_k_t^u ( y^-1_t , s)^k_l\ { ( _ mn^l_0(s))^n_t , s+ ( _ m_^l_0(s))}(y_t , s)^m_jds + & & ( ) ^(1)_t , u:=e=(y_t , u)^i_j(_^j_a(x,0 ) ) + & & ( _ a^i)^(2)_t , u:=e=(y_t , u)^i_j(_^2 ^j_a(x,0))+2(_t , u)^i_j(_^j_a(x,0 ) ) . using these expressions ,
one show that ( ^(1)_a+^(1)_a)_t , u&:=&e + & & = e^-_t^u c(x_s^(0))ds\ { ( _ i g(x_u^(0),0 ) ) ( _ a^i)^(1)_t , u -g(x_u^(0),0)_t^u ( _ i c(x_s^(0 ) ) ) ( _ a^i)^(1)_t , sds } , + and in the same way that & & ( ^(2)_a + ^(2)_a)_t , u:=e + & & = -2(_t^u ( _ i c(x_s^(0)))_t , s^ids)(^(1)_a+^(1)_a)_t , u + & & + e^-_t^u c(x_s^(0))ds\ { ( _ i g(x_u^(0),0))(_a^i)^(2)_t , u+2(_ijg(x_u^(0),0))^j_t , u(_a^i)^(1)_t , u + 2(_i_g(x_u^(0),0))(_a^i)^(1)_t , u + & & -2_t^u ( _ j c(x_s^(0)))(_a^j)^(1)_t , sds + & & -g(x_u^(0),0)_t^uds } .
now that we are able to express @xmath128 as a function of @xmath121 as ( z_a^(i,))_t=_t^t \ { ( _ a^(1)+^(1)_a)_t , u + ^2 ( ^(2)_a+^(2)_a)_t , u}du+o(^2 ) .
+ this completes the goal of asymptotic expansion for @xmath120 and @xmath125 , which are now expressed as functions of @xmath15 as desired .
in this section , we will study the perturbation scheme under the pde ( partial differential equation ) framework following the so - called four step scheme @xcite .
we will see that our perturbative method makes the four step scheme tractable for the generic situations , which only requires standard techniques for the classical parabolic linear pde . in the next section
, we will explain the equivalent perturbation method in the probabilistic framework .
let us consider the following generic coupled non - linear fbsde : dv_t&=&-f(t , x_t , v_t , z_t)dt+z_tdw_t + v_t&=&(x_t ) + dx_t&=&_0(t , x_t , v_t , z_t)dt+(t , x_t , v_t , z_t)dw_t + x_0&=&x .
[ fbsde_org ] here , we made the dependence on @xmath29 explicitly to clearly distinguish it from the stochastic @xmath12 components .
as before , we assume that @xmath9 , @xmath16 , @xmath12 take value in @xmath10 , @xmath5 and @xmath129 respectively , and @xmath130 denotes a @xmath21 dimensional standard brownian motion . following the arguments of the four step scheme of ma and yong @xcite ,
let us postulate that @xmath61 is given by the function of @xmath29 and @xmath23 as v_t = v(t , x_t ) almost surely for @xmath131 $ ] .
then , applying it@xmath132 s formula , we obtain dv_t&= & _ t v(t , x_t ) dt + & & + \{_i v(t , x_t)_0^i(t , x_t , v(t , x_t),z_t)+_ijv(t , x_t)(^i^j ) ( t , x_t , v(t , x_t),z_t)}dt + & & + _
i v(t , x_t)^i ( t , x_t , v(t , x_t),z_t)dw_t .
thus , in order that @xmath133 is the right choice , it should satisfy [ t_pde ] & & v(t , x)=(x ) + [ pde_org ] & & _ t v(t , x)+\ { _ i v(t , x)^i_0(t , x , v(t , x),z(t , x))+_ijv(t , x ) ( ^i^j)(t , x , v(t , x),z(t , x ) ) } + & & + f(t , x , v(t , x),z(t , x))=0 + [ z_pde ] & & z(t , x)=_iv(t , x)^i(t , x , v(t , x),z(t , x ) ) , where the last equation arises to match the volatility term . + in the four step scheme , one first needs to find the solution @xmath134 satisfying the eq.([z_pde ] ) . and
secondly , one has to solve the pde ( [ pde_org ] ) to obtain @xmath135 , which then allows one to run @xmath12 as a standalone markovian process in the third step .
and then finally , one will obtain the backward components by setting @xmath136 and @xmath137 .
the crucial point in the above four step scheme is whether one can finish the step @xmath52 and @xmath53 successfully . even if one finds the solution for @xmath138 , the second step requires to solve the non - linear pde ( [ pde_org ] ) , which is very difficult in general . in the remainder of this section ,
let us study how our perturbation method works to achieve this goal .
+ we consider , as before , the original system eq ( [ fbsde_org ] ) as a linear decoupled fbsde with perturbations of non - linear generator and feedbacks of the order of @xmath25 . we write it as &
& dv_t^()=c(t , x^()_t)v_t^()dt - g(t , x_t^(),v_t^(),z_t^())dt+z_t^()dw_t + & & v_t^()=(x_t^ ( ) ) + & & dx_t^()=(r(t , x_t^())+(t , x_t^(),v_t^(),z_t^()))dt + & & + ( ( t , x_t^())+(t , x_t^(),v_t^(),z_t^()))dw_t + & & x_0^()=x [ p_fbsde ] and the corresponding pde : [ p_pde ] & & v^()(t , x)=(x ) + & & _ t v^()(t , x)+\ { _ i v^()(t , x)^i_0(t , x , v^(),z^())+_ijv^()(t , x ) ( ^i^j)(t , x , v^(),z^ ( ) ) } + & & + f(t , x , v^(),z^())=0 + & & z^()(t , x)=_i v^()(t , x)^i(t , x , v^()(t , x),z^()(t , x ) ) , where f(t , x , v^(),z^())&=&-c(t , x)v^()(t , x)+g(t , x , v^()(t , x),z^()(t , x ) ) + _ 0(t , x , v^(),z^())&=&r(t , x)+(t , x , v^()(t , x),z^()(t , x ) ) + ( t , x , v^(),z^())&=&(t , x)+(t , x , v^()(t , x),z^()(t , x ) ) .
we suppose that the solution of the above pde can be expanded perturbatively in such a way that v^()(t , x)&=&v^(0)(t , x)+v^(1)(t , x)+^2 v^(2)(t , x)+ + z^()(t , x)&=&z^(0)(t , x)+z^(1)(t , x)+^2 z^(2)(t , x)+ , and then try to solve @xmath139 order by order .
if the non - linear terms are small enough , we can expect to obtain a good approximation by putting @xmath26 in the above expansion to a certain order .
in the zero - th order , the pde ( [ p_pde ] ) reduces to & ( _ t + ( t , x))v^(0)(t , x)=0 & + & v^(0)(t , x)=(x ) & [ pde_v0 ] and z^(0)(t , x)=_i v^(0)(t , x)^i(t , x ) .
[ pde_z0 ] here , we have defined the operator @xmath140 as ( t , x)=r^i(t , x)_i+(^i^j)(t , x)_ij - c(t , x ) .
this is a standard parabolic pde and can be handled in the usual way .
one can easily check that @xmath141 and @xmath142 solves the fbsde ( [ p_fbsde ] ) when @xmath143 . by extracting @xmath25-first order terms from the pde ,
we obtain & ( _ t+(t , x))v^(1)(t , x)+g^(1)(t , x)=0 & + & v^(1)(t , x)=0 & [ pde_v1 ] and z^(1)(t , x)=_i v^(1)(t , x)^i(t , x)+_i v^(0)(t , x)^i(0)(t , x ) .
[ pde_z1 ] here , we have defined g^(1)(t , x)=_i v^(0)(t , x)^i(0)(t , x)+_ijv^(0)(t , x ) ( ^i^j(0))(t , x)+g^(0)(t , x ) and the following notations : ^(0)(t , x)&=&(t , x , v^(0)(t , x),z^(0)(t , x ) ) + ^(0)(t , x)&=&(t , x , v^(0)(t , x),z^(0)(t , x ) ) + g^(0)(t , x)&=&g(t , x , v^(0)(t , x),z^(0)(t , x ) ) . as a result
, we once again obtained a linear parabolic pde .
hence @xmath144 can also be solved , at least numerically , in a standard fashion . in the second order
, one can show @xmath145 and @xmath146 should satisfy & ( _ t + ( t , x))v^(2)(t , x)+g^(2)(t , x)=0 & + & v^(2)(t , x)=0 & [ pde_v2 ] and z^(2)(t , x)&=&_i v^(2)(t , x)^i(t , x)+_i v^(1)(t , x)^i(0)(t , x ) + & & + _
i v^(0)(t , x)(v^(1)(t , x)_v+z^(1)(t , x)_z)^i(0)(t , x ) .
[ pde_z2 ] here , @xmath147 is given by g^(2)(t , x)&= & _ i v^(1)(t , x)^i(0)(t , x)+_i v^(0)(t , x ) ( v^(1)(t , x)_v+z^(1)(t , x)_z)^i(0)(t , x ) + & & + _ ijv^(1)(t , x)(^i^j(0))(t , x)+_ijv^(0)(t , x ) ( ^i(0)^j(0))(t , x ) + & & + _ ijv^(0)(t , x)^i(t , x ) ( v^(1)(t , x)_v+z^(1)(t , x)_z)^j(0)(t , x ) + & & + ( v^(1)(t , x)_v+z^(1)(t , x)_z)g^(0)(t , x ) , where the partial differentials with respect to @xmath133 and @xmath138 are taken by considering @xmath148 and @xmath27 as functions of @xmath149 .
it is still a linear parabolic pde .
although we can proceed to higher orders in the same way and solve @xmath150 , there is another way with a clearer representation .
let us define v^[i](t , x)=_j=0^i ^j v^(j)(t , x ) , z^[i](t , x)=_j=0^i^j z^(j)(t , x ) . and the operator ^[k](t , x)&=&\{^l_0(t , x , v^[k](t , x),z^[k](t , x))_l+ ( ^l^m)(t , x , v^[k](t , x),z^[k](t , x))_lm } + then , one can easily check that the pde for @xmath151}$ ] with @xmath152 can be expressed as [ master - pde ] & ( _ t + ^[i-1](t , x))v^[i](t , x)+f(t , x , v^[i-1](t , x),z^[i-1](t , x))=0 & + & v^[i](t , x)=(x ) and [ master - z ] z^[i](t , x)=_lv^[i](t , x)^l(t , x , v^[i-1](t , x),z^[i-1](t , x ) ) .
it is straightforward to confirm the consistency with the summation of each @xmath153 for @xmath154 up to the error terms of @xmath155 , which is due to the additional @xmath25 in front of the non - linear terms .
note that , in an arbitrary order , the pde has a linear parabolic form .
+ the above formulation clearly shows that the perturbative treatment of non - linear effects of the original system allows us to obtain a series of linear parabolic pdes with the same structure . solving the pde for the zero - th order , and then recursively replacing the backward components by the solution of the previous expansion order
, we can obtain an arbitrary higher order of the approximation .
we have now seen the perturbation method in the pde framework can work even for the fully - coupled non - linear fbsdes . in this section , we will provide a corresponding perturbation scheme under the probabilistic framework . as we will see , it is nothing more difficult than the decoupled case studied in sec .
[ decoupled_approx ] , and reduces to the standard calculations for the european contingent claims . as a by - product , applying the asymptotic expansion method explained in sec .
[ asymptotic ] , we can also show that it is possible to obtain an analytic expression for the non - linear pde in the four step scheme up to the given order of expansion .
we try to solve the same fbsde ( [ fbsde_org ] ) treated in the pde framework .
suppose that we have somehow obtained a solution of @xmath156}(t , x ) , z^{[i-1]}(t , x))$ ] .
then , let us consider the following fbsde : dv^[i]_t&=&-f(t , x_t^[i],v^[i-1](t , x^[i]_t),z^[i-1](t , x_t^[i]))dt+z^[i]_tdw_t + v_t^[i]&=&(x^[i]_t ) + dx^[i]_t&=&_0(t , x^[i]_t , v^[i-1](t , x^[i]_t),z^[i-1](t , x_t^[i]))dt + & & + ( t , x^[i]_t , v^[i-1](t , x_t^[i]),z^[i-1](t , x_t^[i]))dw_t + x_0^[i]&=&x .
[ grading - fbsde ] here , one can immediately check that the solution of the above fbsde @xmath157 } , z_t^{[i]})$ ] , as a function of @xmath158})$ ] , actually satisfies the pde in the four step scheme given in ( [ master - pde ] ) and ( [ master - z ] ) by setting v^[i](t , x)=v^[i](t , x),z^[i](t , x)=z^[i](t , x ) .
hence the solution of the above fbsde can be interpreted as the @xmath159-th order approximation of the original fbsde in ( [ fbsde_org ] ) .
therefore , if we can solve the above fbsde in probabilistic way , we can proceed to an arbitrarily higher order of approximation by simply updating the backward components of the non - linear terms recursively .
we can also say that it is a probabilistic way to solve the non - linear pde ( [ fbsde_org ] ) order by order of @xmath25 .
one can check that the above fbsde is actually decoupled and linear by writing it explicitly as & & dv_t^[i]= c(t , x_t^[i])v_t^[i]dt - g(t , x^[i]_t , v^[i-1](t , x^[i]_t),z^[i-1](t , x^[i]_t))dt+z_t^[i]dw_t + & & v_t^[i ] = ( x_t^[i ] ) + & & dx_t^[i]= ( r(t , x_t^[i])+(t , x_t^[i],v^[i-1](t , x_t^[i ] ) , z^[i-1](t , x_t^[i])))dt + & & + ( ( t , x_t^[i])+(t , x_t^[i],v^[i-1](t , x_t^[i]),z^[i-1](t , x_t^[i])))dw_t + & & x_0^[i]=x , and hence , we can straightforwardly integrate it as [ fbsde - vi ] & & v_t^[i]=e + [ fbsde - zi ] & & z_t^[i]=e .
the result is equivalent to the pricing of standard european contingent claims , and also has the same form appeared in sec .
[ decoupled_approx ] .
thus , we can apply the asymptotic expansion method given in sec .
[ asymptotic ] to the forward components @xmath160}$ ] in the same way .
this will give us the analytical result of @xmath157},z_t^{[i]})$ ] as a function of @xmath158})$ ] , up to a given order of volatility parameter , say @xmath161 .
then we can set v^[i](t , x)=v_t^[i](t , x),z^[i](t , x)=z_t^[i](t , x ) .
up to the error terms of @xmath162 , and can move on to the higher order of approximations ( and also @xmath26 ) , the actual order of error terms are of @xmath163 in this example . ] . here
, let us summarize the procedures of our perturbation method .
firstly , in the zero - th order , the corresponding fbsde is given by dv_t^[0]&=&c(t , x_t^[0])v_t^[0]dt+z_t^[0]dw_t + v_t^[0]&=&(x_t^[0 ] ) + dx_t^[0]&=&r(t , x_t^[0])dt+(t , x_t^[0])dw_t + x_0^[0]&=&x .
this can be integrated as v_t^[0]&=&e + z_t^[0]&=&ewhich can be solved either exactly , or analytically up to the certain order of volatility by the asymptotic expansion method .
then we set v^[0](t , x)=v^[0](t , x ) , z^[0](t , x)=z^[0](t , x ) , and then put them back in the backward components of ( [ grading - fbsde ] ) with @xmath164 .
we then obtain ( [ fbsde - vi ] ) and ( [ fbsde - zi ] ) with @xmath164 .
we can express @xmath165}$ ] and @xmath166}$ ] in terms of @xmath29 and @xmath167}$ ] by using the asymptotic expansion method , and use them to define @xmath168}(t , x),z^{[1]}(t , x))$ ] in turn .
now , we can move to ( [ fbsde - vi ] ) and ( [ fbsde - zi ] ) with @xmath169 .
we repeat the same procedures to the desired order of approximation .
+ _ remark _ : although we have considered one - dimensional process for @xmath9 , it is straight forward to extend the method for higher dimensional cases . once we take the basis of @xmath12 in such a way that the linear drift term @xmath9 is diagonal , we can proceed without any difficulty .
the mixing from the other components of @xmath9 always appears in the lower order of @xmath25 , which keeps the diagonal form of drift term intact in an arbitrary order .
in this paper , we have presented a simple perturbation scheme for non - linear decoupled as well as coupled fbsdes . by considering the interested system as a decoupled linear fbsde with non - linear perturbation terms
, we succeeded to provide the analytic approximation method to an arbitrarily higher order of expansion .
we have shown that the required calculations in each order are equivalent to those for standard european contingent claims .
we have applied the method to the two simple models and compared them with the numerical results directly obtained from the pde and regression - based monte carlo simulation . both of the examples clearly demonstrated the strength of our method .
we have also shown that the use of the asymptotic expansion method for forward components allows us to proceed to the higher order of perturbation even if the forward components do not have known distributions . in the last part of the paper ,
we have studied the perturbative method in the pde framework based on the so - called four step scheme .
we have shown that our perturbative treatment renders the original non - linear fbsde into the series of linear parabolic pdes , which are straightforward to handle .
furthermore , by the equivalence of the two approaches , we were also able to provide the corresponding perturbative method in probabilistic framework which is explicitly consistent with the four step scheme up to a given order of expansion .
the perturbation theory presented in this paper may turn out to be crucial to investigate various interesting problems , such as those given in the introduction , which have been preventing analytical treatment so far .
the application of the new method to the important financial problems is one of our ongoing research topics .
finally , let us remark on the further extension to the cases including jumps .
although , in this work , we have only considered the dynamics driven by brownian motions , the same approximation scheme can also be applied to more generic cases .
although it will be more difficult to obtain explicit expressions in terms of forward components , if we choose the specific forward processes with appropriate analytical properties , we should be able to proceed in the similar way .
particularly , the separation of the original system into the decoupled linear fbsde and the non - linear perturbation terms can be done in a completely parallel fashion .
as we have briefly mentioned in section [ decoupled_approx ] , there are the situations where the driver contain the linear term in volatility @xmath16 .
although it is possible to absorb it by the measure change , there may be the situation where its direct treatment improves the numerical performance than inducing the drift modification in the forward components .
consider the situation where the backward component follows the following bsde : dv_t=(c(t , x_t)v_t+(t , x_t)z_t)dt - g(t , x_t , v_t , z_t)dt+z_tdw_t .
also in this case , the recursive procedures can be carried out in the same way by simply replacing the factor ( -_t^s c(u , x_u)du ) by ( t , s)=(-_t^s ( c(u , x_u)+||(u , x_u)||^2)du-_t^s ( u , x_u)dw_u ) .
for example , the results of section [ decoupled_zeroth ] can now be expressed as v_t^(0)&=&e + z_t^(0)&=&e .
higher order and coupled cases can be expressed similarly . in general , it is difficult to say which method performs better , since both of them have errors of the same order of @xmath170 at a given expansion .
in the appendix , we provide another method which is more closely related to that of sec.[decoupled_approx ] for generic coupled fbsdes .
we consider the same coupled fbsde as in sec .
[ four - step - prob ] : & & dv_t^()=c(x_t^())v_t^()dt - g(x_t^(),v_t^(),z_t^())dt+z_t^()dw_t + & & v_t^()=(x_t^ ( ) ) + & & dx_t^()=r(x_t^())dt+(x_t^(),v_t^(),z_t^())dt + ( ( x_t^())+(x_t^(),v_t^(),z_t^()))dw_t + & & x_0^()=x . here , we absorbed a possible explicit time dependency to @xmath12 .
as in sec.[decoupled_approx ] , we are going to expand the backward as well as forward components in terms of @xmath25 .
suppose that we have v_t^()&=&v_t^(0)+v_t^(1)+^2 v_t^(2)+ + z_t^()&=&z_t^(0)+z_t^(1)+^2 z_t^(2)+ + x_t^()&=&x_t^(0)+x_t^(1)+^2 x_t^(2)+ .
now , let us derive each term separately .
in the zero - th order , the original equation reduces to a linear and decoupled fbsde , which will going to serve as the center of expansion : dv_t^(0)&=&c(x_t^(0))v_t^(0)dt+z_t^(0)dw_t + v_t^(0)&=&(x_t^(0 ) ) + dx_t^(0)&=&r(x_t^(0))dt+(x_t^(0))dw_t + x_0^(0)&=&x .
@xmath171 is now the standard markovian process completely decoupled from the backward components .
we can easily solve the backward components as v_t^(0)&=&e + z_t^(0)&=&e . as explained in the previous section
, we can express @xmath172 and @xmath173 in terms of @xmath174 by the help of the asymptotic expansion for volatility even if the process of @xmath171 does not have known distribution .
now let us consider the dynamics of @xmath39 and @xmath175 .
following the same arguments in sec.[decoupled_approx ] , one can show straightforwardly that & & dv_t^(1)=c(x_t^(0))v_t^(1)dt - g^(1)(t)dt+z_t^(1)dw_t + & & v_t^(1)=^(1)(t ) + & & dx_t^(1)=_x r(x_t^(0))x_t^(1)dt+^(0)(t)dt + ^(1)(t)dw_t + & & x_0^(1)=0 , where we have used shorthand notations : ^(1)(t)&=&^(1)(x_t^(0),x_t^(1 ) ) = _ x ( x_t^(0))x_t^(1 ) + g^(0)(t)&=&g(x_t^(0),v_t^(0),z_t^(0 ) ) + g^(1)(t)&=&g^(1)(x_t^(0),x_t^(1),v_t^(0),z_t^(0 ) ) = g^(0)(t)-(_x c(x_t^(0))x_t^(1))v_t^(0 ) + ^(0)(t)&=&(x_t^(0),v_t^(0),z_t^(0 ) ) + ^(0)(t)&=&(x_t^(0),v_t^(0),z_t^(0 ) ) + ^(1)(t)&=&^(1)(x_t^(0),x_t^(1),v_t^(0),z_t^(0))=^(0)(t)+_x ( x_t^(0))x_t^(1 ) .
note that , since we have obtained @xmath172 and @xmath173 as the functions of @xmath174 , the pair @xmath176 consists of a markovian process , which is indeed decoupled from @xmath177 .
this means that we have ended up with the decoupled linear fbsde also for the first order correction .
hence , one can easily solve the backward components as [ v1_couple ] & & v_t^(1)=e + [ z1_couple ] & & z_t^(1)=eas a result , we have obtained @xmath178as well as @xmath179 as the functions of @xmath174 and @xmath180 .
we can continue to the higher order corrections in the same way . by considering @xmath181 and @xmath182 , and then extracting the @xmath25-second order terms
, one can show that & & dv_t^(2)=c(x_t^(0))v_t^(2)dt - g^(2)(t)dt+z_t^(2)dw_t + & & v_t^(2)=^(2)(t ) + & & dx_t^(2)=_x r(x_t^(0))x_t^(2)dt+^(2)(t)dt+^(2)(t)dw_t + & & x_0^(2)=0 .
here we have defined ^(2)(t)&=&^(2)(x_t^(0),x_t^(1),x_t^(2 ) ) + & = & _ x(x_t^(0))x_t^(2)+_ij(x_t^(0))x_t^i(1)x_t^j(1 ) + & & g^(2)(t)=g^(2)(x_t^(0),x_t^(1),x_t^(2),v_t^(0),v_t^(1),z_t^(0),z_t^(1 ) ) + & & = \{_x g^(0)(t)x_t^(1)+ g^(0)(t)v_t^(1)+_z g^(0)(t)z_t^(1 ) } + & & -(_x c(x_t^(0))x_t^(2)+_ijc(x_t^(0))x_t^i(1)x_t^j(1))v_t^(0 ) -(_x c(x_t^(0))x_t^(1))v_t^(1 ) . + and also for the process
@xmath183 , we have defined ^(2)(t)&=&^(2)(x_t^(0),x_t^(1),x_t^(2),v_t^(0),v_t^(1),z_t^(0),z_t^(1 ) ) + & = & \{_x ^(0)(t)x_t^(1)+ ^(0)(t ) v_t^(1)+_z ^(0)(t)z_t^(1 ) } + & & + _ ijr(x_t^(0))x_t^i(1)x_t^j(1 ) and ^(2)(t)&=&^(2)(x_t^(0),x_t^(1),x_t^(2),v_t^(0),v_t^(1),z_t^(0),z_t^(1 ) ) + & = & _ x ( x_t^(0))x_t^(2)+ _
ij(x_t^(0))x_t^i(1)x_t^j(1 ) + & + & \{_x^(0)(t)x_t^(1)+^(0)(t)v_t^(1)+ _
z^(0)(t)z_t^(1 ) } . one can check that the pair @xmath184 consists of a markovian process and it is decoupled from the backward components .
hence , once again , we have obtained the decoupled linear fbsde , which is solvable as before : v_t^(2)&=&e + z_t^(2)&=&e . in completely the same way
, we can proceed to an arbitrarily higher order correction . in each expansion order @xmath159
, the set @xmath185 follows markovian process decoupled from the backward components , and also the fbsde continues to be linear thank to the @xmath25 in front of the non - linear terms .
+ + _ remark _ : for simplicity of the presentation , we have used the common perturbation parameter @xmath25 both for the non - linearity in backward components as well as the feedback effects in the forward components . however , as one can easily expect , it is also possible to introduce multiple perturbation parameters . for completeness ,
let us check the consistency to the result of sec .
[ four - step - prob ] . in the zero - th order ,
the corresponding fbsdes are exactly equal . in the first order ,
we have obtained v_t^[1]&=&e[1st_order_n ] and @xmath166}$ ] as its malliavin derivative in sec .
[ four - step - prob ] .
now , in order to compare it to the current method , let us consider v_t^[1]-v_t^(0 ) and extract @xmath25 first order terms by expanding @xmath167}=x_t^{(0)}+\ep x_t^{(1)}$ ] .
since the second term of ( [ 1st_order_n ] ) has already @xmath25 in the front , we simply get a contribution from there as e. from the first term , we have to expand @xmath186}$ ] in the terminal payoff and also in the discount : ewhere we have used the gateaux derivative .
second term can be interpreted that there is a change in value by -(_x c(u , x_u^(0))x_u^(1))v_u^(0 ) at each point of time @xmath187 , which is summed and discounted back to the current time . by summing the above two terms
, one can easily confirm its equivalence to the result of the previous section .
applying malliavin derivative automatically tells the consistency of volatility terms . using the same arguments
, one can check the consistency between the current method and that of sec .
[ four - step - prob ] .
although we have solved @xmath188 separately , the sum @xmath189 can be shown equivalent to @xmath190}$ ] up to the error terms @xmath155 .
shiraya , k. , takahashi , a. , and toda , m , 2009 , `` pricing barrier and average options under stochastic volatility environment , '' carf working paper f-242 , available at http://www.carf.e.u-tokyo.ac.jp/workingpaper/ , forthcoming in journal of computational finance .
takahashi , a. , takehara , k. , and toda , m. , 2011 , `` a general computation scheme for a high - order asymptotic expansion method , '' carf working paper f-242 , available at http://www.carf.e.u - tokyo.ac.jp / workingpaper/.
|
in this work , we have presented a simple analytical approximation scheme for generic non - linear fbsdes . by treating the interested system as the linear decoupled fbsde perturbed with non - linear generator and feedback terms
, we have shown that it is possible to carry out a recursive approximation to an arbitrarily higher order , where the required calculations in each order are equivalent to those for standard european contingent claims .
we have also applied the perturbative method to the pde framework following the so - called four step scheme . the method is found to render the original non - linear pde into a series of standard parabolic linear pdes . due to the equivalence of the two approaches ,
it is also possible to derive approximate analytic solution for the non - linear pde by applying the asymptotic expansion to the corresponding probabilistic model .
two simple examples are provided to demonstrate how the perturbation works and show its accuracy relative to known numerical techniques .
the method presented in this paper may be useful for various important problems which have eluded analytical treatment so far .
*p * * keywords :* bsde , fbsde , four step scheme , asymptotic expansion , malliavin derivative , non - linear pde , cva
|
the scattering properties of dust provide a unique view into the physical properties of dust grains .
this view is different from other probes of dust grains ( eg
. , extinction , polarization due to differential extinction , abundances ) as it allows their absorption and scattering properties to be separated .
the scattering properties of dust grains are indicators of the size and composition of dust grains .
the scattering of photons by dust grains can be described by a single scattering albedo , @xmath0 , and a scattering phase function , @xmath1 , where @xmath2 is the scattering angle . in most studies ,
@xmath1 is approximated by the single parameter henyey - green ( 1941 ) phase function @xmath3 where @xmath4 the @xmath5 parameter is referred to as the scattering phase function asymmetry and varies from -1 ( complete back scattering ) to 0 ( isotropic scattering ) to 1 ( complete forward scattering ) .
this henyey - greenstein phase function is a good approximation for dust grains , except possible in the far - ultraviolet ( witt 1989 ) .
other analytic forms of the scattering phase function have been proposed ( e.g. , cornette & shanks 1992 ; draine 2003b ) , but have yet to be used in dust scattering studies .
there are three astrophysical objects which are suited for studies of milky way interstellar dust scattering properties .
these are reflection nebulae , dark clouds , and the diffuse galactic light ( dgl ) .
these objects consist of either high - density or diffuse dust illuminated by sufficiently strong sources to permit the detection of scattered light .
reflection nebulae usually consist of a single or small number of stars illuminating their natal cloud .
dark clouds are isolated dust clouds which are usually illuminated by the interstellar radiation field .
the dgl is the sum of the light of the stars in the milky way scattered by dust in the galaxy .
other objects which do show scattering light , but which are not suitable for such studies are circumstellar disks , regions , and external galaxies .
these objects have dust which is newly formed , processed , or non - milky way dust .
each of these three astrophysical objects have different strengths and weaknesses with respect to determining dust scattering properties .
the strengths of reflection nebulae are that they are bright , have simple , wavelength independent illuminator geometries , and are illuminated by a single or handful of stars .
their weaknesses are that they often have strong emissions in the red ( extended red emission [ ere ] ) and near - infrared ( non - equilibrium , thermal small particle emission ) . in addition
, they usually probe high density regions ( @xmath6 ) making the results harder to compare to most dust grain models .
the strengths of dark clouds are that they have simple , wavelength independent illuminator geometries and are externally illuminated by the interstellar radiation field and/or a small number of stars .
dark clouds share the same weakness as reflection nebulae as they also probe high density regions ( @xmath7 ) and , in addition , they are difficult to observe as they are faint .
the strength of the dgl is that it probes the diffuse interstellar medium ( @xmath8 ) .
the weaknesses of the dgl are that the illuminator geometry is wavelength dependent and the dust geometry is complex having a disk density distribution with embedded dust clouds .
in addition , dgl observations are challenging as observations of the faint dgl signal are needed over a fairly large region of the sky , especially a range of galactic latitudes , to allow for accurate measurements of @xmath0 and @xmath5 .
the combination of the results from all three astrophysical objects should give a good view of dust scattering properties without undue uncertainties resulting from each object type s weaknesses .
this review will concentrate on published studies of interstellar dust scattering properties in the ultraviolet , visible , and near - infrared .
the topic of x - ray scattering by dust is covered elsewhere in these proceedings . a great deal of work
has taken place in this area since the review of dust scattering properties at the last dust meeting ( witt 1989 ) . at that time ,
only a handful of studies existed which made quantitative determinations of dust scattering properties ( ie .
, actually determined values for @xmath0 and @xmath5 ) . since the last dust meeting
, a number of studies have specifically been addressed to this issue .
brief summaries of these studies , usually in the form of plots of @xmath0 and @xmath5 as a function of wavelength , have been presented in the literature ( eg .
, gordon et al . 1994 ; mathis 1996 ; li & greenberg 1997 ; witt & gordon 2000 ; draine 2003a , 2003b ) .
one drawback to these reviews has been the inclusion of all literature results without consideration of the quality of each study .
this has lead to a misplaced notion that our knowledge of dust scattering properties is quite uncertain in some wavelength regions .
this has been especially pronounced in studies of the dgl in the far - ultraviolet ( around 1500 ) . in a handful of these studies ,
preliminary analyses have been revised in later papers resulting in significantly higher determinations of both @xmath0 and @xmath5 .
thus , it is important to apply a uniform set of criteria when reviewing dust scattering studies to ensure the basic quality of the results .
this has been done in this review .
the literature was searched for determinations of dust scattering properties in objects possessing milky way interstellar dust .
astrophysical objects which satisfy this constraint are reflection nebulae , dark clouds , and the dgl . for inclusion in this review , each study had to satisfy four criteria .
these criteria are : 1 .
_ determine albedo & g _ : this requires that specific values be quoted for the two quantities .
studies which were only sensitive to one of the two quantities were included as long as a realistic range of the other quantity was probed . usually , the albedo ( @xmath0 ) was the quantity determined with calculations performed for @xmath5 values between zero and one .
while this criteria means that the large number of early studies which only put limits on @xmath0 and/or @xmath5 are neglected , their limits are usually consistent with the values determined from later studies .
2 . _ determine uncertainties _ : the importance of this criteria can not be understated .
the uncertainties presented here are either the uncertainties quoted in the original paper or reflect the range of @xmath0 & @xmath5 values allowed by models presented in each study .
this can mean that multiple models ( with different assumptions ) have been combined into a single measurement with larger uncertainties than a single model .
refereed publication _ :
this ensures that the full details of the work have be presented .
in addition to the usual journals ( eg .
, , , , , , etc . ) , theses were counted as refereed publications .
not be superseded by other work _ : a number of studies are superseded by newer work which usually used better radiative transfer models ( especially the case for dgl studies ) .
this usually results in significantly different determinations of @xmath0 & @xmath5 .
the studies satisfying the first three criteria are listed in table [ tab_litstudies ] , separated by astrophysical object studied .
these tables give the study reference , a brief description , a model code , and whether the study satisfied the fourth criteria and , as such , was included in figs .
[ fig_rn_vals]-[fig_all_vals ] .
each model code is explained in table [ tab_models ] where the method of calculation , sources , and dust geometry are summarized .
ideally , the model used to interpret the observations of a specific object would include realistic illuminators ( location and spectrum ) , a realistic dust distribution , and calculate the full multiple scattering of photons . this ideal is unlikely to be fully met by any specific model , but models which come closest have the best chance of producing good dust scattering properties .
the original study reference should be consulted for the full details of each study .
the @xmath0 and @xmath5 values for the studies which satisfy all four criteria are plotted separate in figs .
[ fig_rn_vals]-[fig_dgl_vals ] for reflection nebulae , dark clouds , and the dgl .
in addition , predictions for different dust grain models are included in these figures ( see [ sec_model ] for model details ) .
the results have been divided between object classes to highlight their differences .
it would not be surprising to find real differences in the @xmath0 and @xmath5 values between object classes .
for example , reflection nebulae and dark clouds possess , on average , larger dust grains than the dgl .
this is seen from the different @xmath9 values measured for these objects as @xmath9 is a rough measure of the average grain size .
examining the results separately also allows for the impact of the different strengths and weaknesses of each object on the derived scattering parameters to be examined .
the final reason to separate the results is to check if the assumptions in the modeling based on object type significantly affect the resulting scattering values .
all of the @xmath0 and @xmath5 values have been plotted together in fig .
[ fig_all_vals ] to check the consistency of determinations between the three astrophysical object types .
llcl reference & description & model & inc . + + witt et al .
1982 & ngc 7023 ; iue 12003000 & rn1 & yes@xmath10 + & & ground - based 34705515 & & + witt et al .
1992 & ngc 7023 ; uit 1400 & 2800 & rn1 & yes + witt et al .
1993 & ngc 7023 ; & rn1 & yes + & vogager 2 10001300 & & + gordon et al .
1994 & sco ob2 ; 1365 & 1769 & rn2 & yes + calzetti et al .
1995 & ic 435 ; iue 12003100 & & rn1 & yes + & ground - based b & v & & + burgh et al .
2002 & ngc 2023 ; fot 9001400 & rn1 & yes + gibson et al .
2003 & pleiades ; wisp 1650 & 2200 & rn3 & yes + + matilla 1970 & coalsack and libra dark & dc1 & yes + & cloud ; ubv & & + fitzgerald et al .
1976 & thumbprint nebula ; b & dc1 & yes + laureijs et al .
1987 & l1642 ; 35005500 & dc1 & yes@xmath11 + witt et al .
1990 & bok globule ; 46908560 & dc1 & yes + hurwitz 1994 & taurus molecular cloud ; & dc2 & yes + & berkeley uvx 1600 & & + haikala et al .
1995 & g251.2 + 73.3 cirrus cloud ; & dc1 & yes + & faust 14001800 & & + lehtinen & mattila 1996 & thumbprint nebula ; jhk & dc1 & yes@xmath11 + + witt 1968 & ground - based 3600 , 4350 , & dgl1 & no@xmath12 + & & 6100 & & + mathis 1973 & ground - based 3600 & & dgl2 & yes@xmath11 + & 4350 & & + witt & lillie 1973 & oao-2 15004200 & dgl3 & no@xmath13 + lillie & witt 1976 & oao-2 15004200 & dgl2 & yes + morgan et al .
1976 & td-1 2350 & 2740 & dgl3 & yes@xmath11 + toller 1981 & pioneer 10 4400 & dgl4 & yes + hurwitz et al . 1991 & berkeley uvx 1625 & dgl5 & no@xmath14 + murthy et al .
1993 & voyager 2 1050 & dgl6 & yes@xmath11 + murthy & henry 1995 & berkley uvx & others & dgl6 & yes@xmath11 + sasseen & deharveng 1996 & faust 1565 & dgl7 & no@xmath15 + petersohn 1997 & de 1 1565 & dgl8 & yes + witt et al . 1997 & faust 1564 & dgl8 & yes + schiminovich et al .
2001 & nuviews 1740 & dgl9 & yes + + @xmath10uv @xmath5 determinations not included , not enough radial data for unique @xmath5 solution + @xmath11no @xmath5 determination possible + @xmath12superseded by mathis 1973 + @xmath13superseded by lillie & witt 1976 + @xmath14superseded by murthy & henry 1995 + @xmath15superseded by witt et al .
1997 lccc name & method & sources & dust geometry + rn1 & monte carlo & single star & homogeneous sphere + rn2 & monte carlo & multiple stars & homogeneous sphere + rn3 & analytic & multiple stars & approximated + & single scattering & & clumpy slab + dc1 & monte carlo & mw isrf & & homogeneous sphere + & & specific stars & + dc2 & numerical int . &
mw isrf & 2d distribution + dgl1 & analytic & galaxy model & homogeneous slab + dgl2 & numerical int .
& mw isrf & infinite cylinder + & n=8 scatterings & boundary condition & + dgl3 & analytic & constant & infinite slab + dgl4 & numerical int .
& star counts & non - homogeneous + & n=2 scatterings & & + dgl5 & numerical int . & td-1 star catalog & clumped dust + dgl6 & numerical int . &
skymap catalog & dust based on + & & & hi survey + dgl7 & monte carlo & star catalogs & dust based on + & & & hi survey + dgl8 & monte carlo & td-1 star catalog & dust cloud spectrum + & & & based on hi survey + dgl9 & monte carlo & 3d td-1 star catalog & dust cloud spectrum + & & & based on hi survey +
the brightness and relative compactness of reflection nebulae have led to determinations of dust scattering properties with a finer wavelength sampling and smaller uncertainties than in either dark clouds or the dgl ( see fig . [ fig_rn_vals ] ) . due to the usual presence of strong emissions ( ere and non - equilibrium emission ) ,
reflection nebulae have only provided scattering information in the ultraviolet and blue - optical .
while dark clouds are faint , their relative compactness and usual lack of strong emissions has made them valuable as probes of dust scattering properties in the red - optical and near - infrared ( fig .
[ fig_dc_vals ] ) .
the one drawback to both reflection nebulae and dark cloud studies is that they probe higher density media than diffuse interstellar medium .
this leads to questions to the general applicability of reflection nebulae and dark clouds results .
the faintness and large extent of the dgl makes it the most challenging object to both observe and model .
the importance of measuring dust scattering properties in the diffuse interstellar medium and determinations of the extragalactic background has resulted in considerable effort being expended on the study of the dgl .
determinations of dgl dust scattering properties in the blue - optical and near - ultraviolet have been fairly uncontroversial , while such determinations in the far - ultraviolet have been quite controversial .
this controversy is highlighted by the back - to - back reviews on the ultraviolet background in which it was argued that the majority of the ultraviolet background is from dust scattered light ( bowyer 1991 ) or extragalactic light ( henry 1991 ) .
the difficulties in both observing and modeling the dgl in the far - ultraviolet has led to studies in the literature with conflicting results , either low @xmath0 and @xmath5 values or high @xmath0 and @xmath5 values .
this is the area which was most affected by the fourth criteria ( not superseded by a more recent study ) .
a number of initial analyses giving low @xmath0 and @xmath5 values were superseded by subsequent studies using more sophisticated models which found higher @xmath0 and @xmath5 values . as a result ,
the scatter in the plots of @xmath0 and @xmath5 for the dgl ( fig .
[ fig_dgl_vals ] ) is now within the quoted uncertainties .
the values of @xmath0 and @xmath5 derived from dgl measurements are similar to those derived from reflection nebulae and dark clouds . by examining figs .
[ fig_rn_vals]-[fig_dgl_vals ] , it can be seen that the @xmath0 and @xmath5 values derived for each object type are internally consistent .
for example , the measurements of @xmath0 and @xmath5 from different reflection nebulae all are consistent with a single wavelength dependence .
when examing these figures , it is important to remember that they include results from broad - band filters along with those derived from spectroscopy . comparing the results for all three object types in fig .
[ fig_all_vals ] , it can be seen that the same may be true in general and not just internal to an object type .
the only area where the data from different object types may disagree is in the far - ultraviolet region for @xmath0 measurements only .
the uncertainties are large enough to make a strong statement either way difficult .
the high level of agreement between object types implies that their different strengths and weaknesses as well as model assumptions are not severely biasing our view of dust scattering properties .
this result goes a long way in answering the concerns raised by mathis , whitney , & wood ( 2002 ) about possible biases in reflection nebulae measurements of the dust albedo caused by ignoring the known clumpiness of dust .
in fact , it is not unreasonable to say that the current data ( see fig . [ fig_all_vals ] ) indicates the wavelength dependence of the albedo is @xmath160.6 in the near - infrared / optical with a dip to @xmath160.4 for 2175 bump , a rise to @xmath160.8 around 1500 , and a drop to @xmath160.3 by 1000 . the wavelength dependence of @xmath5 is simpler with a monotonic rise from 0.6 to 0.8 from 10000 to 1000 .
the uncertainties do allow for significant real variation in the @xmath0 and @xmath5 values around these qualitative averages of at least 0.1 .
the study of dust scattering properties has shed light on the nature of the 2175 bump and the far - ultraviolet rise features of extinction curves .
early work on the 2175 bump indicated that it was likely an absorption feature with no scattered component ( eg . , lillie & witt 1976 ) .
subsequently , evidence for a scattered component in the 2175 bump was found in two reflection nebulae , ced 201 and ic 435 , by witt , bohlin , & stecher ( 1986 ) .
this results was found to definitely be spurious for ic 435 when much more sensitive observations were taken and analyzed by calzetti et al.(1995 ) . by inference ,
the observations of ced 201 are also likely to be spurious . as a result ,
all evidence currently supports a 2175 extinction bump which is only due to absorption .
similarly , examining the results of reflection nebulae ( see fig .
[ fig_rn_vals ] ) gives good evidence that a significant portion of the far - ultraviolet rise in extinction curves ( 1700 and shorter wavelengths ) is due to absorption .
one way to synthesize our knowledge of dust is to fit different observations of dust to a dust grain model .
the main observational constraints for dust grain models are usually the diffuse ism dust extinction curve and dust abundances ( weingartner & draine 2001 ; clayton et al .
in addition , some dust grain models also fit the dust polarization ( kim & martin 1994 ; li & greenberg 1997 ) and infrared emission spectrum ( zubko et al .
2003 ; li & draine 2001 ) .
some dust grain models also have been fit to the denser dust extinction curves ( kim , martin , & hendry 1994 ; weingartner & draine 2001 ) .
no dust grain models have used the measured dust scattering properties as fitting constraints , but most do use the scattering properties as a consistency check . in figs .
[ fig_rn_vals]-[fig_all_vals ] , predictions for @xmath0 and @xmath5 from recent dust grain models are plotted .
the zubko et al.(2003 ) model plotted is described as bare - nc - b .
all but one of the five model predictions plotted are for diffuse ism dust ( @xmath8 ) and are then only directly comparable to the dgl dust scattering results . as can be seen from fig . [ fig_dgl_vals ] , the dust grain models are in fairly good agreement with the results for @xmath5 , but most of the models underpredict @xmath0 for most wavelengths . all the models have a small dip in @xmath0 at the 2175 bump , but the dip is not as large as indicated by the dgl measurements .
the models all predict lower far - ultraviolet albedos than have been measured in the dgl . the one dust grain model for denser dust ( weingartner & draine ; @xmath17 ) is directly comparable to the dust scattering measurements in reflection nebulae ( fig .
[ fig_rn_vals ] ) and dark clouds ( fig .
[ fig_dc_vals ] ) as these objects have similar @xmath9 values .
similar to the dgl comparison , the @xmath5 values predicted by the model are in reasonable agreement for reflection nebulae and dark clouds , except possible in the optical for dark clouds .
the model predictions for @xmath0 agree with the measurements for dark clouds which mainly probe the optical and near - infrared .
the @xmath0 predictions for reflection nebulae are generally too high in the optical and too low in the far - ultraviolet .
the albedo dip at 2175 is much smaller in the model than the well measured values indicate .
basically , the recovery of the albedo at around 1500 is not seen in any of the dust grain models .
this review of dust scattering determinations has concentrated on giving a critical examination of results presented in the literature for reflection nebulae , dark clouds , and the dgl . unlike previous such reviews ,
each study was required to pass four simple criteria for inclusion in this review .
there were a total of 23 studies passing the four criteria with 7 on reflection nebulae , 7 on dark clouds , and 9 on the dgl .
a great deal of progress has been made since the last review of this area ( witt 1989 ) .
the uncertainty in far - ultraviolet albedo from dgl studies has been resolved with the help of better observations and more sophisticated modeling .
the 2175 bump has been shown to be an absorption feature ; earlier indications of scattering in the bump have been traced to bad data .
the measurements in reflection nebulae , dark clouds , and the dgl are roughly consistent alleviating earlier worries about the effects of the modeling assumptions specific to the object being studied ( eg .
reflection versus the dgl ) .
1 . does the assumption of the henyey - greenstein single parameter scattering phase function significantly bias the resulting dust scattering parameters ( @xmath0 and @xmath5 ) ?
this question is probably best answered with models of images of reflection nebulae and dark clouds which include either more complicated analytical phase functions ( eg .
, a double henyey - greenstein phase function ; eq . 4 of witt 1977 ) or scattering phase functions computed from dust grain models . 2 .
what are the @xmath0 and @xmath5 values in the near - infrared ?
currently , there is only @xmath0 measurements for a single dark cloud and no @xmath5 measurements .
the @xmath0 and @xmath5 values in the near - infrared probe the larger dust grains , information about which is difficult to determine from extinction curve measurements alone .
3 . what are the @xmath0 and @xmath5 values in the red - optical and near - infrared for reflection nebulae ? is it even possible to determine @xmath0 and @xmath5 in reflection nebulae for these wavelengths ?
there are reflection nebulae in which ere has not been detected ( witt & boroson 1990 ) .
are there nebulae without near - infrared , non - equilibrium emission ?
is it possible to measure dust scattering properties in the red - optical and near - infrared from dgl measurements ? 5 .
what does it mean to measure dust scattering properties in objects without ere or near - infrared , non - equilibrium emission if the diffuse ism has been shown to have ere ( gordon , witt , & friedmann 1998 ) and probably also has near - infrared , non - equilibrium emission .
is it possible to use the full multiwavelength appearance of reflection nebulae and dark clouds to alleviate lingering concerns about the geometrical assumptions inherit in modeling such objects ?
specifically , the inclusion of ultraviolet through far - infrared would provide direct measurements of the direct , scattered , and re - emitted light and thus require many fewer assumptions .
is it feasible to consider using the full multiwavelength appearance of the dgl to reduce number of assumption necessary to model the dgl ? or is this still too complex of a problem , especially in light of the need for accurate 3d star positions of all stars important for the dgl ( eg .
, hot stars in the ultraviolet and cooler stars at longer wavelengths ) ? 8 . what are the real differences between dust scattering properties in reflection nebulae , dark clouds , and the dgl ?
the current evidence indicates they are consistent with each other at the @xmath1620% level .
are the dust scattering properties in other galaxies different than those derived for milky way dust ?
the large and small magellanic clouds offer environments which have lower metallicities and high star formation rates than the milky way .
in addition , dust extinction curves in the magellanic clouds have been measured to be quite different from milky way extinction curves ( gordon et al . 2003 ) .
finally , the author would like to encourage readers who find studies which have not been included in this review to email them to the author ( currently [email protected] ) .
the author is committed to continuing to update the web - based version of this reviewkgordon / dust / scat_param / scat_data.html ] .
bowyer , s. 1991 , , 29 , 59 burgh , e. b. , mccandliss , s. r. , & feldman , p. d. 2002 , , 575 , 240 calzetti , d. , bohlin , r. c. , gordon , k. d. , witt , a. n. , & bianchi , l. 1995 , , 446 , l97 clayton , g. c. , wolff , m. j. , sofia , u. j. , gordon , k. d. , & misselt , k. a. 2003 , , 588 , 871 cornette , w. m. & shanks , j.g .
1992 , , 31 , 3152 draine , b. t. 2003a , , in press draine , b. t. 2003b , , in press fitzgerald , m. p. , stephens , t. c. , & witt , a. n. 1976 , , 208 , 709 gibson , s. j. & nordsieck , k. h. 2003 , , 589 , 362 gordon , k. d. , clayton , g. c. , misselt , k. a. , landolt , a. u. , & wolff , m. j. 2003 , , 594 , 279 gordon , k. d. , witt , a. n. , carruthers , g. r. , christensen , s. a. , & dohne , b. c. 1994 , , 432 , 641 gordon , k. d. , witt , a. n. , & friedmann , b. c. 1998 , , 498 , 522 haikala , l. k. , et al . 1995 , , 443 , l33 henry , r. c. 1991 , , 29 , 89 henyey , l. c. & greenstein , j. l. 1941 , , 93 , 70 hurwitz , m. 1994 , , 433 , 149 hurwitz , m. , bowyer , s. , & martin , c. 1991 , , 372 , 167 kim , s. & martin , p. g. 1994 , , 431 , 783 kim , s. , martin , p. g. , & hendry , p. d. 1994 , , 422 , 164 laureijs , r. j. , mattila , k. , & schnur , g. 1987 , , 184 , 269 lehtinen , k. & mattila , k. 1996 , , 309 , 570 li , a. & greenberg , j. m. 1997 , , 323 , 566 li , a. & draine , b. t. 2001 , , 554 , 778 lillie , c. f. & witt , a. n. 1976 , , 208 , 64 mathis , j. s. 1973 , , 186 , 815 mathis , j. s. 1996 , , 472 , 643 mathis , j. s. , whitney , b. a. , & wood , k. 2002 , , 574 , 812 matilla , k. 1970 , , 9 , 53 morgan , d. h. , nandy , k. , & thompson , g. i. 1976 , , 177 , 531 murthy , j. & henry , r. c. 1995 , , 448 , 848 murthy , j. , m. , henry , r. c. , & holberg , j. b. 1993 , , 419 , 739 petersohn , j. 1997 , ms thesis , univ . of toledo ,
ohio sasseen , t. p. & deharveng , j .-
1996 , , 469 , 691 schiminovich , d. , friedman , p. g. , martin , c. , & morrissey , p. f. 2001 , , 563 , l161 toller , g. n. 1981 , phd thesis , state univ . of new york ,
stony brook weingartner , j. c. & draine , b. t. 2001 , , 548 , 296 witt , a. n. 1968 , , 152 , 59 witt , a. n. 1977 , , 35 , 1 witt , a. n. 1989 , iau symp . 135 : interstellar dust , 135 , 87 witt , a. n. , bohlin , r. c. , & stecher , t. p. 1986
, , 305 , l23 witt , a. n. & boroson , t. a. 1990 , , 355 , 182 witt , a. n. , friedman , b. c. , & sasseen , t. p. 1997
, , 481 , 809 witt , a. n. & gordon , k. d. 2000 , , 528 , 799 witt , a. n. & lillie , c. f. 1973 , , 25 , 397 witt , a. n. , oliveri , m. v. , & schild , r. e. 1990 , , 99 , 888 witt , a. n. , petersohn , j. k. , bohlin , r. c. , oconnell , r. w. , roberts , m. s. , smith , a. m. , & stecher , t. p. 1992
, , 395 , l5 witt , a. n. , petersohn , j. k. , holberg , j. b. , murthy , j. , dring , a. , & henry , r. c. 1993 , , 410 , 714 witt , a. n. & stephens , t. c. 1974 , , 79 , 948 witt , a. n. , walker , g. a. h. , bohlin , r. c. , & stecher , t. p. 1982
, , 261 , 492 zubko , v. , dwek , e. , arendt , r. g. 2003 , , submitted
|
studies of dust scattering properties in astrophysical objects with milky way interstellar dust are reviewed .
such objects are reflection nebulae , dark clouds , and the diffuse galactic light ( dgl ) .
to ensure their basic quality , studies had to satisfy four basic criteria to be included in this review .
these four criteria significantly reduced the scatter in dust properties measurements , especially in the case of the dgl .
determinations of dust scattering properties were found to be internally consistent for each object type as well as consistent between object types .
the 2175 bump is seen as an absorption feature .
comparisons with dust grain models find general agreement with significant disagreements at particular wavelengths ( especially in the far - ultraviolet ) . finally , unanswered questions and future directions are enumerated .
|
sinspirant dun article de grimmett et marstrand sur certaines proprits de la percolation en dimension @xmath1 , bezuidenhout et grimmett ont dmontr dans un article clbre que le processus de contact steint au point critique .
leur technique de preuve a souvent t utilise pour tudier la croissance des modles de population .
le prsent texte se veut une introduction leurs ides , dans le cadre de lexemple simple du processus de galton - watson . on montrera
ici quun processus de galton - waltson non dgnr ne peut steindre que si sa fertilit dpasse strictement 1 .
la preuve classiquement enseigne
voir par exemple benam
el karoui @xcite est essentiellement analytique .
elle repose sur les fonctions gnratrices et des arguments de convexit .
la preuve prsente ici ne fait videmment pas usage des fonctions gnratrices et permet ainsi de retrouver une preuve plus conforme lintuition probabiliste .
soit @xmath2 deux lois sur @xmath3 .
@xmath4 est appele loi de reproduction et @xmath5 est la loi de la taille de la population initiale .
on appelle chane de galton - waltson de loi initiale @xmath5 et de loi de reproduction @xmath4 la chane de markov de loi initiale @xmath5 et de matrice de transition @xmath6 on peut fabriquer une telle chane comme suit : soient @xmath7 des variables alatoires de loi @xmath4 et @xmath8 une variable alatoire de loi @xmath5 indpendante des @xmath7 . on dfinit par rcurrence
la suite @xmath9 par @xmath10 alors @xmath11 est une chane de galton - watson de loi initiale @xmath5 et de loi de reproduction @xmath4 .
si je note @xmath12 et @xmath13 , on a classiquement@xmath14=m{\ensuremath{\mathbb{e}}}[y_n]\text { et } { \ensuremath{\mathbb{e}}}[y_n]=m^n{\ensuremath{\mathbb{e}}}[y_0]\ ] ] on pose @xmath15 .
[ mort ] si @xmath16 , @xmath17 . en particulier @xmath18 .
avec , on a @xmath19=m^n{\ensuremath{\mathbb{e}}}[y_0]$ ] .
[ galtonindep ] si @xmath20 et @xmath11 sont deux chanes de galton - waltson indpendantes de mme loi de reproduction @xmath4 , alors @xmath21 est encore une chane de galton - waltson de loi de reproduction @xmath4 .
comme @xmath20 et @xmath11 sont des chanes de markov indpendantes , @xmath22 est une chane de markov , de matrice de passage @xmath23 notons @xmath24 les lois des chanes de markov canoniquement associes .
il sagit maintenant de montrer que si lon pose @xmath25 , alors @xmath26 est une chane de markov .
pour cela , il suffit de montrer que si @xmath27 , alors @xmath28 ne dpend que @xmath29 et @xmath30 . or , sous @xmath24 , @xmath31 et @xmath32 sont deux variables alatoires indpendantes de lois respectives @xmath33 et @xmath34 , donc la loi de @xmath35 est @xmath36 .
ainsi , @xmath37 et @xmath21 est bien une chane de galton - waltson de loi de reproduction @xmath4 .
comme la loi initiale est @xmath38 , on a le rsultat voulu .
dans la suite on notera @xmath39 une probabilit sous laquelle une suite @xmath11 est une chane de galton - watson de loi initiale @xmath40 et de loi de reproduction @xmath4 .
soient @xmath41 . on a *
pour tout @xmath42 , @xmath43 * pour @xmath41 @xmath44 . * pour @xmath45 , on a @xmath46 grce au thorme [ galtonindep ] on a @xmath47 do par rcurrence @xmath48 .
cela donne le premier point .
avec la proprit de markov , on obtient alors le deuxime point .
le dernier point est vident .
[ souschaine ] soit @xmath49 . @xmath50 est une chane de galton - watson de loi de reproduction @xmath51 . comme @xmath52 est une chane de markov ,
il est bien connu que @xmath50 est une chane de markov .
il reste valuer les probabilits de transition .
soit @xmath53 .
en appliquant @xmath54 fois le thorme [ galtonindep],on voit que si @xmath55,@xmath56 @xmath57 sont des processus de galton - watson indpendants de lois initiales respectives @xmath58 et de loi de reproduction @xmath4 , alors @xmath59 est un processus de galton - watson de loi initiale @xmath60 et de loi de reproduction @xmath4 . en
consquence @xmath61 ce qui est le rsultat voulu .
avec lingalit de tchebitchef .
posons @xmath71 .
soit @xmath72-a}$ ] avec la proprit de markov , on a pour @xmath73 avec @xmath74 : @xmath75\\ & = { \ensuremath{\mathbb{e}}}[{1\hspace{-2.7mm}1}_a{\ensuremath{\mathbb{e}}}[{1\hspace{-2.7mm}1}_{y_{t+1}<an\}}|\mathcal{f}_t]]\\ & = { \ensuremath{\mathbb{e}}}[{1\hspace{-2.7mm}1}_a \phi(y_t , an)]\\ & \le { \ensuremath{\mathbb{e}}}[{1\hspace{-2.7mm}1}_a c / n]=c / n{\ensuremath{\mathbb{p}}}(a),\end{aligned}\ ] ] do @xmath76 .
+ on en dduit par rcurrence que pour @xmath77 , on a @xmath78 do @xmath79 * dans le sens direct , lide est de comparer la chane avec un processus de galton - watson surcritique , puis de conclure laide du thorme [ surcritique ] . * le sens rciproque est ici assez simple , puisquil sagit essentiellement de montrer que le nombre de particules explose ds quil y a survie .
toutefois , il faudra avoir lesprit que si lvnement local est plus compliqu , cette partie - l sera en ralit la plus difficile .
soient @xmath85 i.i.d .
de loi @xmath4 . on pose
@xmath86 , @xmath87 , puis @xmath88 avec @xmath89 . on montre par rcurrence que @xmath90 . en effet , si @xmath90 , on a @xmath91 @xmath92 est une chane de galton - walson de fertilit @xmath93=a{\ensuremath{\mathbb{p}}}^n(y_1\ge an)>1 $ ] , donc qui peut survivre daprs le thorme [ surcritique ] .
par comparaison , @xmath52 peut survivre . on a vu que @xmath103 . + comme @xmath104 , on peut crire @xmath105 or le thorme de convergence des martingales dit que @xmath106={\ensuremath{\mathbb{p}}}^n(\tau<+\infty|\mathcal{f}_t)\to { 1\hspace{-2.7mm}1}_{\{\tau<+\infty\}}\quad{\ensuremath{\mathbb{p}}}^n\text { p.s.}\ ] ] lorsque @xmath107 tend vers linfini .
+ en particulier , sur lvnement @xmath108 , @xmath109 tend presque srement vers @xmath110 et @xmath111 tend presque srement vers linfini . on a donc presque srement lingalit @xmath112 avec le lemme de fatou , il vient @xmath113= { \renewcommand{\arraystretch}{0.6 } \begin{array}{c } { \scriptstyle } \\ { \underline{\lim}\quad}\\ { \scriptstyle t\to + \infty } \end{array } \renewcommand{\arraystretch}{1}}{\ensuremath{\mathbb{p}}}^n(y_t\ge 2n)\ ] ] comme @xmath114 , il existe @xmath115 tel que @xmath116 . on va voir une deuxime preuve , un peu plus longue , mais aussi plus robuste .
cette mthode a notamment t utilise dans garet - marchand @xcite pour ltude des marches alatoires branchantes en milieu alatoire .
lide est alors de coupler la chane avec une chane sous - critique .
soient @xmath7 des variables alatoires de loi @xmath4 , @xmath122 une suite de variables de bernoulli de paramtre @xmath30 indpendantes des @xmath123 . on pose
@xmath87 , @xmath124 , puis par continuit squencielle croissante , @xmath126 donc il existe @xmath67 tel que @xmath127 . on a alors @xmath128 si on prend @xmath129 suffisamment grand , on a @xmath130 donc @xmath131 . or @xmath132 est un processus de galton - watson de loi de reproduction @xmath133 et de loi initiale @xmath134 , donc daprs le thorme [ equivalence ] ( sens direct ) , ce processus de galton - watson peut survivre
. mais @xmath135={\ensuremath{\mathbb{e}}}[b_1 ^ 1]{\ensuremath{\mathbb{e}}}[x_1 ^ 1]=pm = p<1,\ ] ] donc daprs le thorme [ mort ] , le processus ne peut survivre . contradiction .
|
les chanes de galton - watson , ou processus de branchements , font partie de la formation traditionnelle des tudiants en probabilit .
le thorme de base concerne ltude de la survie du modle en fonction de la fcondit : sauf cas dgnr , la survie nest possible que si le taux de fcondit dpasse @xmath0 .
la preuve classiquement enseigne est essentiellement analytique , reposant sur les fonctions gnratrices et des arguments de convexit . on se
propose ici , sinspirant des travaux de bezuidenhout et grimmett , de donner une preuve plus conforme lintuition probabiliste .
|
lattice gauge theory @xcite is a formulation of gauge theory on a euclidean space - time lattice @xmath0 .
@xmath0 consists of @xmath1 points @xmath2 ordered in some fashion and @xmath3 links @xmath4 where @xmath5 is the number of space - time dimensions .
this formulation of gauge theory is amenable to computer simulations .
lattice gauge theory simulations can perform non - perturbative computations for static observables in quantum chromodynamics ( qcd ) .
the expectation value of an observable @xmath6 can be obtained by evaluating a euclidean path integral as = d_i d|_i o e^-|_i ( d^(u)+ m_i)_i -s_g(u ) .
@xmath7 is a group valued matrix but other values of @xmath8 are also theoretically interesting . ]
defined on the link @xmath4 and plays the role of a parallel transporter ( gauge field ) .
@xmath9 represents a fermionic species with mass @xmath10 and is defined at each lattice point @xmath2 .
@xmath11 is a lattice dirac operator and @xmath12 is a discretized version of the continuum yang - mills lagrangian @xmath13 .
we first integrate out the fermionic fields to obtain = _ i ( d^(u)+ m_i ) oe^-s_g(u ) .
computing the determinant is an @xmath14 process for a @xmath15 matrix and for the 4-dimensional dirac oprator @xmath16 . even for moderate lattice sizes @xmath17 .
an explicit evaluation of this determinant is not feasible .
it is usually re - exponentiated by introducing additional bosonic fields @xmath18 called pseudofermions as = d^_i d_i o e^-^_i ( d^(u ) + m_i)^-1_i -s_g(u ) .
the path integral now involves evaluating the inverse of the matrix @xmath19 on the vector @xmath20 .
this can be done in at most @xmath21 steps using conjugate gradient ( cg ) type of algorithms .
though lattice gauge theory simulations are computationally expensive , they can be easily parallelized .
such simulations can therefore be efficiently carried out on a graphics processing units ( gpus ) which have about 500 compute cores .
since the first implementation of lattice gauge simulations on gpus in 2007 @xcite there has been numerous implementations of various formulations of lattice gauge theory on gpus .
a collection of gpu subroutines for several lattice actions has been put together in a package called quda @xcite .
implementation of pure gauge codes and spin models on gpus have been reported in @xcite and @xcite . in this article
we explore a new method of implementing lattice simulations on the gpu using register variables and avoid the slow global memory for storing run - time variables .
we consider one of the most popular formulations namely lattice gauge theory with two mass degenerate fermions .
this partition function can be written as = d^ de^-^ ( ) -s_g(u ) . where @xmath22 in section 2 we briefly describe the algorithm which we use .
section 3 describes the architecture of the gpu cards on which the program was tested and discusses the modifications we have to make in the usual cpu program for efficient running on the gpu .
finally in section 4 we report on the performance of the program and our conclusions .
currently the most efficient algorithm for generating gauge configurations with dynamical quarks on the lattice is known as hybrid monte carlo ( hmc ) @xcite . in this algorithm
the dynamics is generated by equations of motion corresponding to the hamiltonian ( u , p)=trp^2 + s_g(u)+^ ( ) . here
@xmath23 is an auxiliary momentum variable with @xmath24 and @xmath25 being conjugate to each other .
@xmath26 is drawn from a gaussian distribution and @xmath27 where @xmath28 is also drawn from a gaussian distribution .
both @xmath26 and @xmath28 are periodically refreshed .
the equations of motion ( molecular dynamics equations ) define a trajectory in the phase space . integrating these equations
numerically is the most time consuming part in the algorithm accounting for 80 to 90% of the computation time depending on simulation parameters .
this part of the algorithm involves repeated multiplication of a matrix @xmath29 on the vector @xmath18 .
parallelization of the matrix vector multiplication speeds up lattice gauge theory simulations to a great extent .
the hmc algorithm thus refreshes the momenta periodically to ensure ergodicity , uses the hamiltonian equations of motion to move quickly through the phase space and perofrms a monte carlo accept / reject step to correct for the finite time step used in the integration .
it is an exact algorithm .
our program was tested on two gpu cards the c2050 and x2090 ( different versions of the nvidia fermi card ) and we briefly discuss their architecture below .
both cards have intrinsic double precision support and are cuda compute version 2 . the computation on the cards
are carried out by multiprocessors each of which have 32 cores and a maximum of 64 kbytes of local storage ( shared memory + l1 cache ) .
the c2050 has 14 such multiprocessors giving it a total of 448 cores while the x2090 has 16 such multiprocessors giving it a total of 512 cores .
nvidia gpu cards have several different categories of memory and their access times vary widely .
the c2050 has a global memory of 3 gb while the x2090 has a global memory of 6 gb .
these are the largest memories but their access times are between 100 and 150 clock cycles .
every multiprocessor on the other hand has 32768 4-byte registers whose combined memory capacity is 128 kbytes and access to the register variables takes only 1 clock cycle . each thread can have a maximum of 63 registers .
fermi cards have the added feature that registers spill over to the shared memory which too can be accessed in 2 - 3 clock cycles .
the combined storage capacity of registers and shared memory is about 192 kbytes per multiprocessor or about 2.7 mb for the c2050 and 3 mb for the x2090 . in our program
we have attempted to use the registers as much as possible instead of the global memory .
the cuda compiler usually tries to place all scalar variables in the register @xcite while arrays are typically placed in the global memory . in hmc ,
the molecular dynamics part deals mainly with three large arrays : the force which is an array of dimension @xmath30 and the momenta and the pseudo - fermion fields , both of which are arrays of dimension @xmath31 our main challenge therefore is to make sure that the elements of these arrays are placed in the register and not in the global memory . the structure of the program is roughly the same for the cpu and the gpu .
there is an initialization step where the gauge field is initialized .
the gauge field is then evolved through a certain number of trajectories . at the end of each trajectory measurements
are carried out on the gauge configuration .
the evolution of the gauge fields consists of the following steps for each trajectory . 1 . setting up the pseudofermion fields and momenta .
2 . integrating the molecular dynamics equations of motion by the leap - frog method .
this sets up the proposed configuration .
3 . accepting or rejecting the proposed configuration depending on @xmath32 , where @xmath33 is the hamiltonian at the beginning of the trajectory and @xmath34 is the hamiltonian at the end of the trajectory .
communication between the cpu and the gpu is through a pci bus .
a major bottleneck if one frequently transfers data between them . to keep a check on this overhead
, we tried to strike a balance between the time taken for transferring the data and the computational time on the gpu . since all the input and output
is carried out by the cpu we had to transfer the data back to the cpu at the end of each trajectory . for the computation intensive parts we wrote cuda kernels ( enumerated below ) which carried out the following computations in parallel on the gpu . 1 . the result of operating the dirac operator and its hermitian conjugate on the pseudofermion field @xmath35 a matrix vector multiplication on a vector of size @xmath36 .
2 . the driving force for the molecular dynamics equation due to the gauge fields .
the driving force for the molecular dynamics equation due to the fermion fields and evolving the momenta .
evolution of the gauge fields .
we now briefly describe the essential modifications to the cpu subroutines for efficient running of the corresponding gpu kernels .
we first try to optimize the two matrix - vector multiplication routines which compute the action of the dirac operator and its hermitian conjugate on the pseudofermion field as a part of the conjugate gradient routine .
so we first try to optimize this routine .
the dirac operator requires four constant @xmath37 matrices ( @xmath38 ) for its definition .
only 8 elements of these matrices are non - zero .
we explicitly work out the result of the matrix - vector multiplication for the @xmath39 matrices retaining only the non - zero elements .
thus the loop over the dirac indices is fully unrolled .
this however is not special to the cuda kernels and is done also for the cpu subroutines .
generally we use the available cuda blas functions as much as possible .
however for the action of the dirac operator on the pseudofermion field , it is not possible to use the blas matrix - vector multiplication routines as the matrix representing the dirac operator is too large to be stored in either the cpu or the gpu memory .
so we store the source and the resultant vectors and compute the necessary matrix elements on the fly . for this kernel ,
the main optimization task was to break down the array representing the pseudofermion field into scalar variables .
since the pseudofermion field typically consists of @xmath40 elements , it is impossible to do this by hand .
we therefore wrote a code generator which runs over the necessary loops and yields the equations for the resultant vector in terms of independent scalar variables .
a naming convention we adopt is to name the scalar variables by concatenating the array indices with the vector name .
the second computationally intensive task is computing the driving force for the molecular dynamics equation due to the fermion fields . on a single thread
, they typically consume 15% of the trajectory time .
after parallelization of the matrix - vector multiplication routines on the gpu , this was consuming the largest amount of time . to optimize this routine
, we followed the same strategy as in the matrix - vector multiplication case .
unrolling the dirac loops in the force calculation is slightly more involved as they involve factors like @xmath41 .
however a bit of algebra again allows us to retain only the non - zero elements and with the help of a code generator we automatically wrote down the momenta evolution equations .
on the cpu this routine consumed significantly less time compared to the conjugate gradient routine .
so this optimization was not carried out on the corresponding cpu subroutines .
in addition to these optimizations , we also had a routine to compute the neighbour indices on the fly to avoid storing the neighbour array on the gpu .
the program is available from the authors on request .
the speed - up of the gpu program was primarily judged by its performance on a c2050 nvidia gpu card running at 1.15 ghz vis a vis its performance on an opteron cluster with qdr infiniband interconnect .
each node of the cluster consisted of two 6-core 2.2 ghz cpus .
we also have comparative results for one lattice size , viz .
@xmath42 , on a cray xe6 and a x2090 nvidia gpu card running at 1.3 ghz on a cray xk6 node . on the opteron cluster
the highest performance was obtained for one mpi process per node .
we therefore fixed the number of mpi processes to 8 ( number of available nodes ) and varied the number of openmp threads in each node .
for the sake of uniformity , the same test run pattern was followed even on the cray .
in addition to the total time taken , timings for each conjugate gradient call and fermionic force calculations ( the two most time consuming subroutines ) were written out . .timing ( in seconds ) of the program on a c2050 gpu , single cpu , smp ( 12 threads ) and mixed mpi and omp with 8 mpi processes each having 12 threads for different lattice volumes .
these runs were performed on an opteron cluster with qdr infiniband interconnect .
( @xmath43estimated ) [ cols="^,^,^,^,^",options="header " , ] in table [ time ] we record the time taken for the various lattice volumes using different compute modes of the opteron cluster and the c2050 gpu card mounted in the opteron cluster .
these volume scaling runs were performed at @xmath44 and @xmath45 with a time step of 0.2 and 50 integrations per trajectory .
the runs were for 200 trajectories each .
the timing for the single cpu run for the @xmath46 lattice is an estimate based on the @xmath47 scaling of the hmc algorithm , since the actual run was taking far too long . in table
[ speedup ] we see the speed - up of the different computing modes over a single cpu thread .
the cpu runs were performed using the intel fortran compiler with the intel mkl while for the gpu runs we used the nvidia cuda compiler ( nvcc ) . for these runs
we found that the performance of the c2050 card is roughly equivalent to the performance of 96 threads of the cluster ( for lattice size 24@xmath48 ) .
we now report the timing information we have on the cray hardware .
each cray xe6 node consisted of two 16 core opteron cpu at 2.1 ghz . on this machine
the cpu code was compiled by the cray fortran compiler and cray libraries were used .
the run parameters on the cray were the same as the opteron cluster .
however instead of 200 , only 2 trajectories were run starting from an ordered start .
we first tested the scaling of this program with the intel fortan compiler on the xe6 and found that the performance was roughly the same as the opteron cluster upto 64 threads . beyond 64 threads the performance gain continued on the xe6 , but not on the opteron .
the gpu run on the cray was done for a different cg convergence limit compared to the cpu runs ( @xmath49 instead of @xmath50 ) and so we could not directly compare the timing of the x2090 run with the other runs . nevertheless , since the cg inversion timing is linear with the number of cg iterations for both gpus ( see figure [ gpu - scale ] ) , it was possible to obtain a rough estimate for the timing for the x2090 for the same number of cg iterations as the other runs .
an additional information that we get from figure [ gpu - scale ] is that the overhead for the cg routine is about 0.5 secs for both the c2050 and x2090 . in figure [ inv_comp ]
we compare the inverter timings for the two gpus as well as the opteron cluster and the cray xe6 for a fixed number ( 123 ) of cg iterations .
the c2050 gives the performance of 8 cray threads or 32 cluster threads while the x2090 gives a performance of about 20 cray threads or 64 cluster threads .
figure [ time_comp ] shows the full timing of the runs for the cluster , the xe6 , the c2050 and the x2090 . from this plot
we conclude that the performance of the c2050 is roughly equivalent to about 19 threads of the xe6 , while that of the x2090 is that of about 26 threads . on the cray xe6 nodes , the force calculation time was about 30% of the inversion time with 64 threads , with 128 threads about 60% of the inversion time and the 256 thread inverter took 60% less time than the force calculation .
therefore , on hindsight , on a lattice of size @xmath42 it is worth unrolling the loops over the force calculations as with the inverter beyond 64 threads . on the opteron cluster on the other hand
the force calculation time was never more than 7% of the inversion time .
thus there not much is to be gained by unrolling the force loops .
our conclusion is that gpu programs can be speeded up to a great extent by using the register variables instead of the global memory . to effectively use the register variables we had to map array elements to scalar variables .
this was carried out automatically by a code generator . before doing this mapping
the performance gain was a modest factor of two . using the registers effectively pushed up the performance gain to a factor 10 compared to a single thread performance on the lattice volumes we investigated .
such automated techniques might prove useful in the design of accelerator compilers .
here we did not use optimization techniques like mixed precision solvers or reconstructing the gauge link on the fly .
we believe these optimizations can bring even more gains . at the moment
we have tested the performance only on a single gpu .
we hope to report our results on multiple gpus shortly .
the authors would like to acknowledge the dst ( india ) grant no .
sr / s2/hep/0035/2008 for funding the servers and gpu cards on which the programs were run .
the authors would also like to acknowledge cray supercomputers for testing the program on their hardware and sharing the results with the authors .
99 p. weisz and p. majumdar , scholarpedia , 7(4 ) ( 2012 ) 8615 .
g. i. egri , z. fodor , c. hoelbling , s. d. katz , d. nogradi , k. k. szabo comput .
comm . 177 ( 2007 ) 631 .
m. a. clark , r. babich , k. barros , r. brower , and c. rebbi , comput .
( 2010 ) 1571 .
n. cardoso , p. bicudo arxiv:1112.4533 [ hep - lat ] n. cardoso , p. bicudo j. comput .
( 2011 ) 3998 .
m. weigel j. comput .
( 2012 ) 3064 .
s. duane , a. d. kennedy , b. j. pendleton , d. roweth , phys .
b 195 ( 1987 ) 216 .
appendix b.2 , cuda c programming guide version 4.2
|
in this article we present our implementation of a hybrid monte carlo algorithm for lattice gauge theory using two degenerate flavours of wilson - dirac fermions on a fermi gpu .
we find that using registers instead of global memory speeds up the code by almost an order of magnitude . to map the array variables to scalars , so that the compiler puts them in the registers , we use code generators .
our final program is more than 10 times faster than a generic single cpu .
|
i zwicky 18 ( also mrk 116 or ugca 166 , hereafter izw18 ) is the most intriguing blue compact dwarf ( bcd ) galaxy in the local universe .
it became famous right after its discovery , when @xcite measured from its emission - line spectrum an oxygen abundance [ o / h]= 1.14 dex , only 7@xmath0 of the solar value , indicating an almost unprocessed gas content .
the first studies on its color and composition @xcite emphasized its almost primordial evolutionary status .
izw18 shows very blue colors , u@xmath1b@xmath2 and b@xmath1v@xmath3 @xcite , suggesting the presence of a very young stellar population , with a current star formation rate ( sfr ) much higher than the past mean value .
the total mass of izw18 from the rotation curve at a radius of 10@xmath412@xmath5 is estimated to be @xmath6 m@xmath7 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , where the neutral gas corresponds to @xmath870@xmath0 of the total mass , but only @xmath9 m@xmath7 of hi is associated with the optical part of the galaxy ( e.g. , * ? ? ?
* ; * ? ? ?
subsequent spectroscopic studies in izw18 @xcite have confirmed its extreme metal deficiency around 1/30@xmath101/50 of z@xmath7 .
nowadays izw18 remains the star forming galaxy with the second lowest metallicity and the lowest helium content known @xcite .
this makes the system a fundamental point in the derivation of the primordial helium abundance @xcite and in the study of the properties of chemically unevolved galaxies .
when discovered by zwicky ( 1966 ) , izw18 was described as `` two galaxies separated by [email protected] and interconnected by a narrow luminous bridge '' , surrounded by two `` very faint flares '' at 24@xmath5 northwest .
subsequent ccd ground based images @xcite revealed a more complex structure : the _ two galaxies _ turned out to be two star forming regions of the same galaxy ( usually indicated as nw and se components ) , while the _ two flares _ are just the most prominent of a few nebulosities surrounding izw18 .
these minor systems are roughly aligned toward the northwest and were initially considered at the same distance , but subsequent spectroscopic studies showed that only one component ( referred to as component c in * ? ? ?
* ) is at the same distance as izw18 and is physically associated with the main body @xcite .
the other diffuse objects have been recognized as background galaxies ( see , e.g. , * ? ? ?
* ) . in the following we will refer to izw18 and to component c respectively as main body ( mb ) and secondary body ( sb ) of izw18 .
both bodies have been resolved into single stars for the first time by hst with the wide field planetary camera 2 ( wfpc2 ) by @xcite and @xcite .
the observational features described above triggered the key question over the nature of izw18 : is it a young galaxy which is presently experiencing its first burst of star formation , or is it an old system which has already formed stars in the past ? hst studies by different authors have tried to characterize the evolutionary status of izw18 , with rather discordant results , but with a clear trend intriguingly reminiscent of oscar wilde s dorian gray s picture : the more one looks at it , the older it appears .
@xcite , using hst / wfpc2 data , were the first to detect asymptotic giant branch ( agb ) stars in the galaxy , and thus to demonstrate that izw18 is as old as at least several hundred myr , and possibly much older .
@xcite failed to detect red giant branch ( rgb ) stars with hst observations made by the advanced camera for surveys ( acs ) , and thus concluded that the galaxy is at most 500 myr old .
however , both @xcite and @xcite , from independent re - analyses of the same acs data set , suggested that izw18 should be older than at least 1 - 2 gyr , since it did appear to contain also red giant branch ( rgb ) stars .
our own time - series hst / acs photometry ( go 10586 , * ? ? ?
* hereafter al07 ) , has allowed us to shed light on the situation , thanks to the combined information from the color - magnitude diagram ( cmd ) and the lightcurves of the cepheids we were able to measure .
in fact , al07 not only characterized the rgb and identified its tip ( trgb ) at @xmath11 mag , but , more importantly , detected for the first time a few cepheids whose lightcurves allowed to independently and more reliably pin down the galaxy s distance to the couple mpc accuracy ( fiorentino et al .
2010 , hereafter f10 ; marconi et al .
the distance to izw18 was estimated from the cepheids using theoretical @xmath12 wesenheit relations specifically computed for the very low metal abundance of the galaxy , and by theoretically fitting the observed lightcurve of the most reliable classical cepheid with period 8.71 days .
the wesenheit relations provide an intrinsic distance modulus for izw18 of @xmath13 ( d=19.0@xmath14 mpc ) adopting a canonical mass - luminosity relation , and of 31.2 @xmath150.2 ( d=@xmath16 mpc ) when an overluminosity of 0.25 dex is assumed for each given mass ( non canonical scenario ) .
the theoretical modeling of the 8.71 days cepheid lightcurve provides 31.4 @xmath150.1 ( d=19.0@xmath17 mpc ) .
the trgb identified by al07 implies an instrinsic distance modulus of 31.30 @xmath15 0.17 , in excellent agreement with the cepheid - based estimates .
this agreement , in turn , provides independent support to our interpretation of the cmd faint red sequence as an rgb .
hence , izw18 is actually farther away than originally thought ( for instance * ? ? ?
* adopted from the literature a mean value of d=10 mpc ) and is in practice at slightly larger distance than the virgo cluster of galaxies .
this made it impossible in the past to resolve its older , hence fainter , stars , and quite challenging even with the outstanding performances of hst s current instrumentation .
today izw18 is one of the most distant system ever resolved into stars . in this work ,
we have used the photometric catalogues of our previous works , which have been focused on determining the age ( al07 ) and the distance through the variable stars ( f10 ) . in this paper
we describe the evolutionary properties and the spatial distribution of the resolved stellar populations and candidate star clusters of the main and secondary bodies of izw18 . in section 2
, we briefly summarize the main differences in the photometric procedures adopted by al07 and f10 .
the cmds and the spatial distribution of stars in different evolutionary phases are presented in section 3 and 4 , respectively .
the results are analysed in section 5 and summarized in section 6 .
the data acquisition and reduction have been extensively described by al07 and f10 .
here we only summarize their main aspects .
time series photometry of izw18 in the f606w ( broad @xmath18 ) and f814w ( @xmath19 ) filters was obtained with the wide field channel ( wfc ) of the acs in 13 different epochs properly spread over a time interval of 96 days ( program i d .
10586 , pi aloisi ) .
these proprietary data were complemented by 5 f555w ( @xmath18 ) and 3 f814w archival acs / wfc images of izw18 from hst ( program i d .
9400 , pi thuan ) .
izw18 is quite distant , plenty of bright stars hiding and/or contaminating the fainter ones , and plenty of gas , which complicates the photometry of individual stars .
in addition , the throughput of the f606w filter is significantly affected by the @xmath20 contribution of ionized gas ( quite abundant in the mb of izw18 ) and makes the star detection and measurement more complicated .
thus , depending on the goal , one can find more appropriate to restrict the analysis of the galaxy populations to safer but fewer objects ( al07 ) , or to more uncertain but numerous ones ( f10 ) . in al07 ,
to produce the galaxy cmd , we built four coadded images by stacking all single epoch images per filter for both archival and proprietary data , separately .
then we performed psf photometry of each of these four coadded images using the daophot / allstar package @xcite .
the calibration to the johnson - cousins photometric system was obtained using the transformations given by @xcite . since the goal of that paper was to demonstrate that faint red stars actually existed where none had previously been detected , the published cmd was produced following a strategy aimed at minimizing the number of false detections . as a result , it contained only the 2099 sources which were independently detected in all the four deep images ( the @xmath18s and @xmath19s of both datasets ) , with good match in the position ( within one pixel ) and magnitude ( within @xmath21 ) of each source in the different images , and with sharpness @xmath220.5 .
these are the most reliable objects measured in each of izw18 s bodies . in the case of f10 , to produce the lightcurves of the candidate variables , we made use of daophotii / allstar / allframe @xcite to get psf photometry on each single epoch ( 15 @xmath19 and 18 @xmath18 images , in total between proprietary and archival data ) . to study the variability we relaxed the al07 selection criteria , since variable stars are more easily distinguishable from spurious objects thanks to their brightness variation in the images at different epochs
moreover , to collect a good sample of candidate variables at all magnitudes , we examined objects close to the detection threshold .
this photometry resulted in a @xmath23 cmd of izw18 which contains 6859 sources with photometric errors @xmath24 mag and @xmath25 parameter in the range from @xmath26 to @xmath27 , and reaches @xmath28 30.0 , i.e. more than half a magnitude deeper than al07 s photometry .
the calibration to the johnson - cousins photometric system was obtained using the transformations given by @xcite , as for the other data set . as shown above
, there is a remarkable difference in the number of objects selected with the two approaches . to infer the properties of the stellar populations inhabiting izw18 minimizing the effects of uncertain photometry
, we will base our results on the hyper - selected al07 catalogue .
nonetheless , we consider it useful to show some features of the larger f10 catalogue both for completeness and for comparison .
since our aim is to analyse the evolutionary properties of izw18 s stellar populations , we have put special care in verifying whether the selection differences affect some cmd regions more than others .
we have hence divided the cmds of the two sets in 8 zones roughly representative of different evolutionary phases .
since the trgb was detected by al07 at @xmath29 , we have adopted this magnitude as the border between bright and faint stars : * zone 1 , stars in the extremely hot portion of the blue plume with @xmath30 and brighter than @xmath29 ; * zone 2 : stars in the extremely hot portion of the blue plume with @xmath30 and fainter than @xmath29 ; * zone 3 : stars in the standard " blue plume , with @xmath31 , and brighter than @xmath29 ; * zone 4 : stars in the standard " blue plume , with @xmath31 , and fainter than @xmath29 ; * zone 5 : objects with @xmath32 and any colour ; * zone 6 : stars with @xmath33 and @xmath34 ; * zone 7 : stars with @xmath35 and @xmath36 ; * zone 8 : all the remaining ones;. the number of objects measured in each zone is reported in table 1 for both bodies and both data sets .
lrrrrrrrr zone 1 & 76 & 6 & 649 & 13 & 0 & 0 & 0 & 0 + zone 2 & 84 & 7 & 917 & 20 & 43 & 5 & 108 & 5 + zone 3 & 277 & 23 & 820 & 17 & 93 & 9 & 98 & 4 + zone 4 & 186 & 16 & 707 & 15 & 262 & 29 & 584 & 28 + zone 5 & 34 & 3 & 51 & 1 & 2 & @xmath371 & 2 & @xmath371 + zone 6 & 105 & 8 & 210 & 4 & 60 & 6 & 110 & 5 + zone 7 & 126 & 11 & 349 & 8 & 120 & 13 & 375 & 18 + zone 8 & 305 & 26 & 1060 & 22 & 332 & 36 & 838 & 40 + total & 1187 & 100 & 4744 & 100 & 912 & 100 & 2115 & 100 + the zones are displayed on the al07 cmds in figure [ cmdmap07 ] .
figure [ image814 ] shows an f814w image of izw18 with the f10 measured objects colour - coded according to the 8 cmd zones of figure [ cmdmap07 ] .
we arbitrarily define as mb objects those with y coordinate in the image smaller than 1200 px and as sb objects those with y@xmath381200 px . in figure [ cmdmap07 ] , and all the following ones , the cmds of the two bodies are splitted following this definition .
we have found that the two photometries agree well with each other , once we take the different selection criteria and photometric depth into account .
the objects measured by both al07 and f10 fall consistently in the same cmd zone ( i.e. have the same colours and magnitudes ) .
the f10 objects not present in the al07 hyper - selected catalogue fall mostly in the fainter portions of the cmds , as expected since they can have larger photometric error .
an intriguing difference is that the mb zones 1 and 2 are much more populated in f10 than in al07 ( see table 1 ) , with a ratio n(f10)/n(al07)=8.5 in zone 1 and 10.9 in zone 2 , significantly higher than the total mb ratio 4744/1187=4.0 .
figure [ image814 ] reveals that most of the f10 objects of zones 1 , 2 and 4 fall on the gas - rich regions in and around the mb .
a possible cause for the discrepancy is therefore gas contamination in the f606w flux , even if in principle all photometric packages should overcome this kind of effects by subtracting the surrounding background flux to each source . as a matter of fact , many of the objects in zone 1 and 2 show a discrepancy in @xmath18-magnitude depending on the used filter , e.g. f606w or f555w .
if , we restrict the selection criteria on the f10 catalogue and keep only the objects with @xmath390.1 , most of those in the cmd zone 2 and several of those in zone 1 are rejected .
we thus consider as real stars on the gas filaments only the brighter ones with smaller photometric error .
finally , we have individually and visually inspected the brighter objects of f10 zone 1 using appropriate routines , and found that some of them do appear as real stars .
we consider the vast majority of the blue objects measured by f10 on the gas filaments and not present in the al07 catalogue as likely spurious detection due to gas peaks , but we do believe that some of the brighter zone 1 detections are real stellar objects . to be conservative , we will base all our results on the restricted al07 catalogue .
to let the reader easily figure out the involved stellar ages , the left panel of figure [ cmdisotracks ] shows the cmds of the hyper - selected al07 objects with overimposed theoretical isochrones @xcite .
the isochrones have metallicity z=0.0004 and are plotted adopting @xmath40=31.4 ( al07 and f10 ) and e(b - v)=0.032 @xcite .
the cmds corresponding to the less restrictively selected f10 photometry are displayed in the right panel of figure [ cmdisotracks ] , where stellar evolution tracks are also plotted to let the reader visualize what the involved star masses are .
the plotted tracks , again with z=0.0004 , are the classical padova models by @xcite .
we note that izw18 is located at high galactic latitude ( @xmath41 ) and therefore its cmds do not suffer from significant contamination from foreground stars belonging to our galaxy , meaning that is not necessary to correct for it our observed diagrams .
izw18 is a starforming dwarf and its cmds show all the features typical of these galaxies : a prominent blue plume with mean @xmath42 and a fainter and broad red plume with @xmath33 .
once compared with the stellar evolution tracks of figure [ cmdisotracks ] , the blue plume turns out to correspond to high- and intermediate - mass stars in the main sequence ( ms ) phase or at the blue edge of the blue loops covered during central he burning .
these two phases are resolved by hst photometry in nearby late - type galaxies , but at the distance of izw18 the photometric error inevitably is large enough to let them merge in just one broad sequence .
stars with colours intermediate between the blue and the red plume are either supergiants ( brighter than @xmath43 ) or intermediate - mass stars in the blue loop phase .
the brighter ones might also be unresolved star clusters .
the red plume is populated by red supergiants ( @xmath44 ) , agb ( @xmath45 ) and rgb stars ( @xmath46 ) .
our data allow to cover the brighter 1.5 magnitude portion of the rgb , where @xmath47 is @xmath150.1 mag in @xmath19 and @xmath150.2 in @xmath48 ( al07 ) and the photometric error , estimated from extensive artificial star tests @xcite , is smaller than @xmath150.3 mag in @xmath19 and @xmath150.4 in @xmath48 ( see figure [ cmdmap07 ] ) .
the colour extension of the rgb is larger than this error and reflects , as usual , both the spreads in age and in metallicity of the resolved stars .
however , given the extremely low metallicity of izw18 , in this case it is mostly the age that contributes to the rgb thickness . in other words ,
the reddest rgb stars are most likely low mass stars @xmath49@xmath50 gyr old .
the redder red plume stars with @xmath51 are presumably carbon stars , not too different from those measured in the magellanic clouds by @xcite .
the redder objects at @xmath52 are most probably stars of @xmath53@xmath54 m@xmath7 in their thermal pulsing ( tp ) phase at the end of the agb , since their location in the cmd is perfectly consistent with the predictions of marigo & girardi s ( 2007 ) tp - agb models .
these authors show in fact that at very low metallicities ( z @xmath55 ) in ( relatively ) massive intermediate - mass stars efficient nuclear burning at the base of the convective envelope may favour the formation of c - rich models and make their tracks describe pronounced excursions towards lower effective temperatures .
these excursions are recognizable in figure [ cmdisotracks ] in the tangled , messy portions of the isochrones above the trgb . 650 px , top panel ) from those in the se region ( y@xmath56650 px , bottom panel ) .
there are 478 objects in the nw region and 704 in the se one.,width=321 ] not only does the mb , by definition , host a larger number of stars ; it also contains a larger fraction of young , blue and bright stars than the sb .
both the luminosity and the number of the brighter stars in the sb are significantly lower than in the mb .
almost no supergiant is found in the sb and no cepheid . moreover , while the blue plume of the mb covers the entire range of blue massive stellar evolution tracks and youngest isochrones , that of the sb turns off the ms around @xmath57 , with apparently no bright star younger than 10 myr being detected there .
this is a significant difference , which reflects , as discussed already by @xcite , both a different rate and a different history of star formation , with the mb being more active than the other at recent epochs , and the sb more active earlier than about 100 myr ago , as suggested by the higher density of objects around the isochrones older than this age in the left - bottom panel of figure [ cmdisotracks ] . at the faintest magnitudes
, we should also consider that crowding affects the mb more severely , preventing the detection of its faintest stars .
this may be particularly the case in the nw part of the mb .
the variable stars in izw18 have been identified by f10 exploiting our acs time - series photometry .
the intensity - averaged mean magnitudes and colors of the confirmed cepheids are plotted as filled ( red ) triangles in figure [ cmdisotracks ] , while those of the confirmed long period variables ( lpvs ) are plotted as filled ( red ) squares .
their position ( see f10 for identifications and nomenclature ) in the cmd indicates that the two bright ultra - long period cepheids ( ulps ) v1 and v15 are supergiants , unless they are hosted by unresolved star clusters making them appear brighter .
the classical cepheid v6 with period p=8.71d is on the blue loop and is best fitted by a @xmath86 m@xmath7 star ( see also f10 and * ? ? ?
the brighter lpv v4 is a supergiant , while the other confirmed lpv v7 lies at the bright edge of the cmd region of the tp - agb stars and is presumably a mira . for a finer analysis of the evolutionary properties of izw18 components
, we have further divided the al07 objects according to their position in the mb .
figure [ mb_separate ] shows the cmds of the objects with y coordinate in the image larger ( top panel ) or lower ( bottom panel ) than 650 px ( see figure [ image814 ] ) .
this in practice separates objects in the nw ( top ) part of the mb from those in the se ( bottom ) .
the two cmds are fairly similar to each other , although the nw one contains more objects in zone 5 and is more dispersed than the other . in particular , the red plume of the se region is much better defined than that of the nw and allows to clearly distinguish the rgb and its tip from the agb .
the larger spread of the nw cmd is likely due to the presence of a larger number of bright objects and thick ionized gas filaments that make the measurements of fainter ( and older ) stars more uncertain .
it is plausible that the gas ionization is itself due to the massive luminous objects populating the nw region ; so , they may represent the only real difference between the northern and southern parts of the mb .
we will further discuss these objects in the next section . in spite of the very different number of objects selected in the two data sets , once the different depths of the resulting cmds ( figure [ cmdisotracks ] ) are taken into account , their stellar evolutionary phases turn out to be populated in similar proportions , both in the mb and in the sb ( see table 1 ) .
as discussed in the previous section , the only significant difference is that the mb blue plume is much more populated , bluer and thicker in the f10 cmd .
we have analysed the spatial location of the various types of stars populating izw18 and checked if stars of different ages are evenly or unevenly distributed . to this purpose
, we have used the cmd zones of figure [ cmdmap07 ] .
thanks to the comparison with the evolutionary tracks , we can assign to each zone the following evolutionary properties : * zone 1 corresponds to massive and intermediate mass stars on the ms , younger than @xmath8 1015 myr ; * zone 2 to fainter intermediate mass stars on the ms , younger than@xmath8 70 myr ; * zone 3 to massive and intermediate mass stars on the ms or at the blue edge of the blue loop phase , younger than @xmath8 70 myr ; * zone 4 to fainter intermediate mass stars on the ms or at the blue edge of the blue loop phase , younger than @xmath8 300 myr ; * zone 5 to supergiants , younger than @xmath8 10 myr and , possibly , to unresolved star clusters ; * zone 6 to agb stars , ages from @xmath8 35 to 300 myr ; * zone 7 to rgb stars , ages from @xmath8 1 to 13 gyr ; * zone 8 in most cases corresponds to intermediate - mass stars in the blue loops , but also to some supergiants , ages from @xmath8 15 to 300 myr .
the spatial distribution in the x - y map of izw18 of the al07 objects measured in the two bodies and falling in each of the 8 cmd zones is reported in figure [ map07 ] .
this map shows that the sb does not contain stars in zone 1 . in other words
the sb has had very little , if any , star formation activity in the last few myr .
zone 1 stars covering the whole range of magnitudes ( hence with masses from 1520 m@xmath7 up ) are numerous in the mb and show a sort of 8-shaped spatial distribution . if we look at the fainter portion of the extremely blue stars ( zone 2 ) , we find that both bodies contain several of these objects and that in the mb they seem to delineate the same 8-shape as their brighter counterpart .
the bright stars in zone 3 have masses above 6 m@xmath7 and are also present in both bodies .
their spatial distribution in figure [ map07 ] forms an irregular ellipsoidal clump of objects with significant central concentration ( the mb ) with a comma - shaped companion ( the secondary : zwicky s flare " ) .
this is the overall familiar appearance of izw18 , since these are the stars more easily visible , although not resolvable from ground .
once again , the sb stars of zone 3 do nt reach the luminosity of the brightest ones hosted in the mb ( see figure [ cmdmap07 ] ) : while the whole mass and luminosity ranges are covered in the latter , only stars less luminous than @xmath58 exist in the former .
taking into account that any initial mass function disfavours the most massive stars , we consider it more likely that the brighter blue stars in the sb are 1215 m@xmath7 stars at the blue edge of the blue loop rather than more massive stars . in this case
, the star formation activity in the secondary must have been quite low , or absent , in the last 10 myr .
fainter ( lower mass ) blue plume stars ( zone 4 ) can be either ms stars with masses between 620 m@xmath7 or stars at the blue edge of the blue loop with mass in the range 35 m@xmath7 ( see the tracks in right panel of figure [ cmdisotracks ] )
. they are present in both bodies , and are more abundant in the sb than in the mb , where they are not found in the central regions and have a spatial distribution complementary to that of their brighter counterpart of zone 3 .
this is clearly due to the severe crowding affecting the mb and , particularly , its inner regions , where the many bright stars completely hide fainter ones . in the sb
the recent star formation activity has been less intense , crowding is less severe , and faint stars are much more easily detectable . the lower activity of the sb at recent epochs is also confirmed by the objects of zone 5 . as apparent from figure [ map07 ] , we find only 2 such bright objects in the sb ( one of which is quite far away and may actually be a background system ) , and 31 in the mb .
the latter are fairly confined in the inner regions , and most often in the nw portion of the mb .
they significantly contribute to the severe crowding affecting the central area .
if interpreted in terms of individual stars , these objects are supergiants and confirm that the central region of the mb has experienced a strong star forming activity in the last 10 myr or so .
admittedly , some of these very bright objects might be not individual stars but unresolved star clusters .
we discuss this possibility in section [ cluster_candidates ] . the objects falling in the cmd zone 6
are ( see the tracks in figure [ cmdisotracks ] ) agb stars with mass lower than 9 m@xmath7 , and age ( see the isochrones in the same figure ) in principle between 35 myr and 13 gyr , but in practice peaked at , at least , several hundreds myr , once the stellar lifetimes and phase durations are taken into account .
they include the tp - agb and carbon stars with 4 - 5 m@xmath7 ( horizontal finger at @xmath5925.5 in the cmds ) predicted by @xcite for these low metallicities and the carbon stars at @xmath5926.4 discussed by al07 .
zone 6 objects are equally present and evenly distributed in both bodies , although missed in the most crowded portions of the mb .
the rgb stars are in zone 7 : objects with mass below 1.8 - 2.0 m@xmath7 and age between 1 and @xmath60 gyr .
the rgb is notoriously affected by the age - metallicity degeneracy , i.e. the ambiguity that stars with the same brightness and colour can be older and metal poorer or younger and metal richer .
hence , without independent spectroscopic metallicity measurements one can not assign the age to rgb stars simply from their position in the cmd . for this reason the photometric identification of the rgb normally puts only a lower limit of 1 - 2 gyr ( i.e. the minimum time required by low - mass stars to reach that phase ) to the age of the examined stellar population .
however , in the case of izw18 , its extremely low metallicity somewhat relaxes the effect of the age - metallicity degeneracy , since the redder rgbs can not be metal rich objects and therefore must be old . in these conditions , the colour extension of izw18 rgb , which is broader than the photometric error and covers the whole range of isochrone ages , can be taken as a signature of the presence of stars possibly as old as 13 gyr .
rgb stars are faint and therefore their detection is strongly affected by crowding and photometric performances .
this is why only with the acs has it been finally possible to resolve and measure them : not even hst was able to reach the rgb , with the less sensitive wfpc2 .
the red giants are evenly distributed in the sb and visible only in the outer regions of the mb , due to the excessive crowding of its inner parts .
it is interesting to notice that the mb rgbs are located along a semi - arc in the system periphery .
we ascribe this asymmetry to the presence of thick gas filaments in the regions where rgbs are not detected : overlapping filaments can clearly hide such faint objects .
we thus believe that rgb stars can actually be evenly distributed also in the mb .
finally , zone 8 collects all the remaining objects , with a large range of brightness and colour , and different ages and evolutionary phases .
those with @xmath6124.5 are stars in the 310 m@xmath7 mass range in the central helium burning phase and with ages between 30 and 300 myr .
the brighter ones are 1015 m@xmath7 mass stars in any post - ms evolutionary phase , with age of the order of 1530 myr .
these mixed objects are evenly distributed in both bodies , with the only exception of the two most crowded nw and se spots of the mb , where they are missing .
interestingly , like the objects of zones 2 , 4 and 7 , they are found in similar numbers in the two bodies . even considering the effect of the different crowding , this suggests that as far as low and intermediate mass stars are concerned , the sb is not significantly less populated than the mb . as for the variable stars ,
the two ulps v1 and v15 are located in the nw portion of the mb , while the classical cepheid v6 and the two lpvs v4 and v7 are in the se portion ( see f10 figure 5 ) .
interestingly , all the confirmed variables belong to the mb : of the 34 good candidate variables examined by f10 , almost half ( 15 ) were actually detected in , or around , the sb , but for none of them satisfactory lightcurves and periods were derived .
ground - based observations of the brighter variables are being performed to better characterize their pulsation and evolutionary properties .
in particular , further data on the ulps will allow us to verify if they are the extension to higher masses of classical cepheids and , in case , to calibrate them as extremely useful primary indicators of cosmological interest . finally , we emphasize that no star is found in the region between the mb and the sb , either in the al07 or the f10 catalogues .
that region looks completely transparent , with only background galaxies being detected there .
this shows that the two bodies are completely separated .
they are members of a two - body system , but have no star connecting each other
. if tidal interactions are at work , they do nt appreciably affect the stellar populations of the two components .
at 19 mpc , acs / wfc photometry can not recognize round regularly shaped clusters whose sizes fall within the point spread function . in this section we consider the possibility that zone 5 objects are actually not individual stars but unresolved star clusters , and discuss the implications on izw18 s evolutionary status . versus @xmath48 color - color diagram ( top ) and @xmath18 versus @xmath48 color - magnitude diagram ( bottom ) for the objects in zone 5 ( cluster candidates ) that have been cross - correlated with the wfpc2 @xmath62 photometry presented by @xcite .
full symbols denote objects located far away from the mb and sb which could be background galaxies .
overplotted are the padova simple stellar populations @xcite for a [email protected] metallicity , ages from 1 myr to @xmath813 gyr , and masses of @xmath64 , @xmath65 , and @xmath66 .
small dots indicate the log(age)=6.7 , 7 , 8 , 9 and 10 models .
the red arrow represents an @xmath67 extinction vector ( see text ) .
, width=302 ] with an intrinsic distance modulus of @xmath40=31.4 , the objects of zone 5 have an absolute bolometric magnitude brighter than -8.4 , certainly consistent with typical cluster luminosities . actually , the brightest objects , with @xmath68 , hence @xmath69 , would be populous " clusters , according to the definitions by @xcite and @xcite .
these bigger , more massive clusters are fairly common in starburst dwarfs , such as ngc 1705 @xcite , ngc 1569 @xcite and ngc 4449 @xcite , and tend to be centrally concentrated , just like these brightest objects of izw18 .
hence , also if taken as star clusters , the objects of zone 5 would confirm a higher star formation activity in the mb than in the sb . to infer the evolutionary properties of unresolved clusters , two photometric bands are insufficient .
fortunately izw18 was observed also in the b band with the wfpc2 by @xcite .
we cross - correlated the 31 + 2 objects in zone 5 with the wfpc2 ( f435w , f555w ) photometry presented by @xcite .
the cross - correlation provides 23 + 2 objects photometrized in f439w , f555w and f814w . a visual inspection in our images of the objects missed by the cross - correlation shows that they are all located in the most crowded mb region , and thus their position measurement is highly uncertain , also because of blending effects .
the wfpc2 magnitudes were transformed into johnson - cousins @xmath70 and @xmath19 magnitudes using the transformations by @xcite . in fig . [ clusters ]
we show the @xmath71 versus @xmath48 color - color diagram and the @xmath18 versus @xmath48 color - magnitude diagram for zone 5 objects .
the data are compared with the padova simple stellar populations ( ssps ) with metallicity [email protected] , ages from @xmath81 myr to @xmath813 gyr , and masses of @xmath64 , @xmath65 , and @xmath66 @xcite .
zone 5 objects have @xmath48 colors in the range @xmath72@xmath73 , and @xmath71 colors in the range @xmath72@xmath74 , and are compatible with ssp ages from a few myr to a hubble time .
the two objects with @xmath75 , @xmath760.8 , and @xmath77 , @xmath760.4 , are located in the most crowded region of the main body , and therefore their colors suffer larger uncertanties .
the objects concentrated around @xmath78 can be either @xmath79 myr old clusters or clusters in the red loop of the models at ages around @xmath80 myr .
we notice that there are some objects with @xmath81 which are redder than the oldest ssps .
we can exclude that they are background galaxies because they are all concentrated within izw18 s mb , except for two of them , which are more `` peripheral '' .
another possibility is that these objects are highly reddened young embedded clusters ; however , in this case they should have @xmath82 ( see reddening vector in figure [ clusters ] ) , and thus absolute magnitudes brighter than @xmath83 .
such high luminosities would imply a super star cluster ( ssc ) classification according to the definition by @xcite .
however , it is unlikely that izw18 host so many sscs , considering that ngc 1569 , the dwarf with the highest known current sfr , presents only 4 of them @xcite .
several studies have discussed stochastic effects on the integrated colors of clusters ( e.g. * ? ? ?
* ; * ? ? ?
stochasticity in the imf sampling of a low - mass cluster can heavily change the observed colors with respect to predictions , and in particular , red supergiants make appear the cluster colors redder .
@xcite investigated this issue and derived a luminosity limit under which stochastic effects are important . comparing the luminosities of our clusters redder than @xmath84 with their predictions for ages older than 100 myr , we found that our cluster luminosities are well above that limit .
therefore , the most likely explanation is that these objects are supergiants with ages younger than @xmath820 myr , implying that the mb has been forming stars at very recent epochs . in fig .
[ clusters ] have not been included .
, width=302 ] for the zone 5 objects with @xmath85 we have inferred ages and masses using the padova models . this was done minimizing the difference between model and data colors ( b@xmath1v , v@xmath1i , and b@xmath1i ) according to a @xmath86 criterion , varying the age between @xmath81 myr and @xmath813 gyr .
we accounted for a galactic reddening of e(b - v)@xmath630.032 , but did not solve for possible internal extinction , because of the small number of bands ( 3 ) and the relatively short wavelength range covered by our data .
following the prescriptions of @xcite , the 1 @xmath87 errors were derived as the limits of the 0.68 confidence level region defined by @xmath88 .
cluster masses were obtained from the mass - to - light ratios of the best - fitting models in the i band .
the results are presented in a mass versus age diagram in fig .
[ clusters_agemass ] .
cluster ages span the whole hubble time , even if we notice that the majority of clusters have ages younger than a few hundred myr .
cluster masses range from @xmath89 to @xmath90 .
the observed correlation between cluster ages and masses is an obvious effect of the magnitude limit adopted for zone 5 .
we notice that our analysis provides 5 `` old '' ( age@xmath911 gyr ) clusters , with masses between @xmath92 and @xmath90 , and totaling @xmath93 .
this implies that @xmath83% of the stellar mass in izw 18 is in the form of old clusters , which sounds unlikely given the fact that this ratio has been found to be considerably lower in other dwarf irregulars with evidence for significant sf at ages @xmath941 gyr ( e.g. , @xmath370.3% in ngc 4449 , annibali et al .
this suggests that the majority of the candidate old clusters in izw18 are instead resolved supergiants .
concluding , our analysis seems to suggest that the majority of the objects in zone 5 are indeed individual stars , even if we can not exclude that some of them are unresolved stars clusters .
nevertheless , even in this scenario , our analysis provides evidence for a particularly active sf over the past hundred myr .
the evolutionary properties derived above for the stars measured in izw18 are similar to those of other metal - poor starbursting dwarfs .
for instance , once the difference in distance modulus @xmath95 is taken into account , magnitudes and colours of the various types of stars on the red plume of izw18 are in perfect agreement with those of the corresponding stars in sbs1415 + 437 , the other very metal - poor bcd whose stellar populations have been resolved by hst / acs @xcite . indeed ,
the cmd of sbs1415 + 437 , whose intrinsic distance modulus is 30.66 , shows an unmistakable trgb at @xmath96 and the horizontal finger typical of carbon stars at @xmath97 @xcite .
sbs1415 + 437 has a red plume more populated than izw18 , as if it had been more active in previous epochs . in turn , the blue plume of izw18 s mb is more prominent than that of sbs1415 + 437 , suggesting that its current and recent star formation activity is proportionally higher .
the results described in the previous sections allow us to draw a qualitative picture of the evolution of the mb and sb .
both components contain old and young stars in all evolutionary phases , but it is apparent that the two stellar systems are distinct from each other and have had different histories .
the quantitative star formation history ( sfh ) of the two bodies of izw18 will be derived from the cmds using the synthetic cmd method ( see e.g. * ? ? ?
* and references therein ) and will be presented in a forthcoming paper @xcite .
the spatial distribution of stars in different evolutionary phases displayed in figures [ image814 ] and [ map07 ] suggests that the mb is indeed one single system and not two galaxies separated and interconnected by a narrow luminous bridge " as originally thought by @xcite .
we can however understand how he was misled by the 8-shaped distribution of the young blue stars .
young massive stars ( zones 1 and 3 of figure [ cmdmap07 ] ) cover the entire mb .
zone 3 stars are concentrated on the nw and se star forming regions , following the traditional scheme of starburst dwarfs , where massive young stars usually group together and are concentrated in the inner regions , although not necessarily at the very center of the galaxy .
the nw region contains a larger number of bright massive zone 3 objects than the se region .
this suggests a more recent / higher star formation activity in the nw clump .
zone 1 stars seem ( see figure [ map07 ] ) intriguingly to avoid the areas where zone 3 stars are located .
is this because they are different objects ?
since they are redder , zone 3 stars might be more evolved or more metal rich than zone 1 stars .
we rather believe that either zone 1 stars lie on gas filaments and are slightly blue - shifted by residual gas contamination increasing the f555w flux , or that differential reddening slightly affects zone 3 objects . throughout this paper
we have only considered the foreground galactic reddening , @xmath98 , but izw18 does contain a lot of gas and presumably dust and molecules , although in quantities limited by its low metallicity @xcite .
it is thus plausible that absorbing interlopers are more abundant in the innermost , star forming regions and that they make the stars of those regions appear redder . in this framework ,
zone 3 stars are the same kind of objects as zone 1 stars , only affected by internal extinction of the order of @xmath990.1
. lower mass young stars ( zones 2 and 4 ) could help in discriminating between the above interpretations , but they are faint and are much more severely affected by crowding than by any other effect .
indeed , although likely to exist also where massive young stars are found , they are only detectable where the latter do nt outshine them . as discussed in the previous sections , in f10 s catalogue many of these objects lie on the gas filaments protruding from the nw spur .
while the vast majority of these faint f10 detections on the filaments are likely to be gas peaks misinterpreted as stars , we do believe that some of the brighter ones , which appear as actual stellar sources when individually and visually inspected in the images , are indeed very young stars or unresolved star clusters ( see also * ? ? ?
this leads to speculate that star formation is occurring at least in the portions of the filaments closest to the mb .
unfortunately , we do nt know whether these filaments are only made of ionized gas or also of neutral one , since the current resolution of hi data ( f. fraternali private communication , lelli et al . in preparation ) does nt allow one to resolve them .
neither do we know whether the filaments are infalling , outflowing or comoving with respect to the mb . from van zee s et al .
( 1998 , their figure 6 ) comparison of the @xmath20 and hi distributions , the filaments appear to be a distinct feature of the ionized gas , but the hi resolution is definitely insufficient .
it is thus difficult to assess if star formation is physically plausible there ; certainly it is extremely unlikely that it occurs in hot , possibly outflowing bubbles , such as many of izw18 ionized filaments @xcite . also the massive objects of zone 5 are centrally concentrated , as usually found in starburst dwarfs .
whether they are resolved supergiants or unresolved star clusters , they are the signature of strong star formation activity .
the evidence that they are more abundant in the nw than in the se clump suggests that the activity in the former is higher than in the latter .
the nw clump is also known to contain wolf - rayet stars ( see e.g. * ? ? ?
* ; * ? ? ?
* ) . yet , from emission line spectroscopy , @xcite suggested it to be not as young as the se clump : 10 myr the former and 5 myr the latter ( and 20 myr the sb ) .
the cmds of figure [ mb_separate ] do not seem to confirm their suggestion .
rather , our cmds suggest that the star formation activity in the nw is more recent than in the se .
the circumstance that , of the variable stars reliably characterized by f10 , the two massive ulp cepheids v1 and v15 are located in the nw region , while the lower mass classical cepheid v6 and the fainter lpv v7 are located in the se region , favours the view that the nw on average contains younger stars than the se .
all the lower mass evolved stars are likely to be evenly distributed in the mb .
rgb stars seem to be located more externally than agbs , but this is most probably a selection effect .
we ascribe their apparent confinement in the outer regions to the severe central crowding which prevents their detections .
similarly , we interpret their asymmetrical horseshoe distribution as due to the extinction from the thick gas filaments of the northern areas ( see * ? ? ?
* for the spatial distribution of dust ) .
there is no doubt that faint objects are more easily detected in the se external portions of the mb .
the resulting overall scenario of the mb consists of a system with two active , distinct regions of current star formation , one ( the nw one ) stronger and younger than the other .
these two active clumps are embedded in a fairly homogeneous body of stars of all ages , from a few myr to several ( possibly 13 ) gyr .
the sb is in a rather different situation .
as mentioned above , the weaker emission lines of its spectrum already led @xcite to suggest that it is older than the mb emission regions .
wfpc2 photometry allowed @xcite to quantitatively assess that its recent sfr is much lower than in the mb .
the acs photometry confirms these results .
our cmds show that the sb does contain young blue plume stars , but they are more likely stars at least 10 myr old on the blue loop phase .
stars more massive than @xmath100 do not seem to be still alive .
the sb also appears to contain much less ionized gas than the mb : a further confirmation of its low recent activity . on the other hand ,
low and intermediate mass stars are at least as numerous in the sb as in the mb , with the objects of zone 4 ( low mass ms stars ) even outnumbering those in the mb ( see table 1 ) .
this is a strong indication that at epochs earlier than @xmath8100 myr ago , the sb must have been actively forming stars as , or more than , the mb .
if we compare the cmd of the mb se region ( figure [ mb_separate ] ) with that of the sb ( figure [ cmdisotracks ] ) , we see that the two red plumes are similar , both in the rgb and the agb sequences . the blue plume ,
although containing similar numbers of stars , clearly shows that those in the sb are more evolved than those in the se , that reach brighter magnitudes and bluer colours .
the agb and the rgb of the sb are well defined .
the colour extension of the latter , combined with the small metallicity range of the system , indicates that stars with the oldest possible ages ( 13 gyr ) are present .
we conclude that the sb has started its star formation activity up to @xmath60 gyr ago , and has remained active since then , at rates comparable with those in the mb until relatively recently .
however , in the most recent epochs ( i.e. later than 10 myr ago ) its sfr has dropped to values significantly lower than in the mb .
we have studied the evolutionary properties and spatial distribution of the stars resolved in izw18 by hst / acs proprietary and archival photometry . to this aim
we have analysed both the al07 and the f10 catalogues and found consistent results , independently of the different selection criteria adopted in the two approaches .
nonetheless , to be conservative , we have based our conclusions on the al07 most restrictively selected catalogue .
the comparison of the cmds resulting from the acs photometry with stellar evolution models ( tracks and isochrones ) indicates that stars of all ages are present in the two bodies . in spite of the impossibility of reaching the oldest ms turn - off in a galaxy 19 mpc away
, the extremely low metallicity of izw18 relaxes the age - metallicity degeneracy and makes the red edge of the rgb a signature of the presence of stars up to possibly 13 gyr old .
the oldest detectable stars are best visible in the sb and the se portion of the mb , where crowding is less severe , but are present also in the rest of the mb , although measured with larger uncertainties .
we have found that the stars are homogeneously distributed over the two bodies , with the younger ones more centrally concentrated , as always found in starburst dwarfs , and the older / fainter ones spread out to the system periphery , also because of selection effects in their detectability . from the maps of the stars spatial distribution ( figure [ map07 ] )
we conclude that old and intermediate age stars are actually distributed homogeneously over the two bodies .
* the distance independently estimated from the trgb ( al07 ) and the cepheids ( f10 ) is @xmath101 mpc . * as far as stars are concerned , the sb is completely separated from the mb : no star is detected between the two bodies . if tidal interactions are at work , they do nt appreciably affect the distribution of the stellar populations of the two components .
* izw18 contains stars in all the evolutionary phases visible at its distance : massive and intermediate mass stars on the ms , blue and red supergiants , agb and tp - agb stars , carbon stars , rgb stars .
hence , it hosts stars of all ages , from a few myr up to possibly 13 gyr . *
the mb hosts two bright star forming regions ( nw and se ) , embedded in a common and homogeneous environment of older stars .
some stars may have been recently formed on the gas filaments protruding from the nw portion of the mb . * while old and intermediate age stars are present in similar amounts in the mb and in the sb , the youngest stars are only found in the mb and mostly concentrated in its nw region .
the se region has also been active recently , but at a lower current rate ; the sb has had definitely less star formation than the mb in the last tens of myr .
* of the 34 candidate variable stars studied by f10 , 19 and 15 are in , or around , the mb and the sb respectively .
all the confirmed ones are however in the mb : the two ulp cepheids in the nw region ; the classical cepheid and the two lpvs in the se region . * at izw18 s distance ,
star clusters are not distinguishable from point - like sources .
it is thus possible that some of the brightest objects in the main body are actually clusters and not individual stars . comparing their colors with simple stellar population models
, we find that the candidate clusters would span ages from a few myr to a hubble time , with the majority of them being younger than @xmath8200 myr . however , the fraction of stellar mass in old ( age@xmath911 gyr ) clusters would be significantly higher than what found in other dwarfs , suggesting that the majority of the candidate old clusters are instead resolved stars .
the reddest ( @xmath102 ) brightest objects can not be unresolved clusters and must be supergiants with ages younger than @xmath820 myr .
these results show that izw18 is not unique in the sky , but shares the evolutionary properties typical of bcds and , more in general , of late - type dwarfs ( see * ? ? ?
* for a review ) .
it is not experiencing now its first burst of star formation , but has been forming stars over many gyrs , most likely over the whole hubble time .
this implies that its extremely low metallicity can not be explained by a low chemical enrichment due to extremely recent star formation .
at best , one may hope that the sfr has been low until relatively recently ( see e.g. * ? ? ?
* ) , although the conspicuous presence of stars of all ages does not make this likely .
the same problem obviously concerns all the other metal - poor bcds where stars many gyr old have been found , like sbs1415 + 437 @xcite .
we recall that no genuinely young galaxy has been found yet by anybody .
all the galaxies whose stellar content has been resolved so far contain stars as old as the lookback time allowed by the available instrumentation , whatever their morphological type and metallicity ( see e.g. * ? ? ?
* ; * ? ? ?
* and references therein ) .
hence , since stellar nucleosynthesis has had plenty of time to pollute them , the only ways to explain the low observed metallicity of these galaxies are either to dilute sufficiently their insterstellar medium by accretion of primordial or very metal poor gas , or to remove the stellar nucleosynthesis products through galactic winds triggered by supernova explosions , or both .
chemical evolution models of late - type dwarfs and bcds have shown long ago @xcite that infall of metal poor gas alone is not sufficient to reproduce the low metallicity of starburst dwarfs , and that galactic winds must be invoked too .
observational evidence for galactic winds has been found for a handful of starburst dwarfs @xcite , and izw18 is indeed one of those where the outflowing ( martin 1996 ) ionized gas associated with the most recent sf activity may indeed have a speed higher than the escape velocity and be able to remove metals from the system .
it would be crucial to get detailed information also on the cold gas , see how it is spatially distributed , and what its kinematics is . understanding the chemical and dynamical evolution of izw18 represents a key step to approach the evolution of all dwarf galaxies as well as that of more massive systems which may have formed from the assembly of small objects like it .
we thank filippo fraternali and renzo sancisi for interesting conversations on the hi gas properties .
financial support for this study was provided by asi through contracts cofis asi - inaf i/016/07/0 and asi - inaf i/009/10/0 .
support to the us co - authors was provided by nasa through grants associated with program go-10586 from the space telescope science institute ( stsci ) , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
aloisi , a. , tosi , m. , & greggio , l. 1999 , , 118 , 302 aloisi , a. , van der marel , r. p. , mack , j. , leitherer , c. , sirianni , m. , & tosi , m. 2005 , , 631 , l45 aloisi , a. , et al .
2007 , , 667 , l151 marel , r. p. , mack , j. , leitherer , c. , sirianni , m. , & tosi , m. 2005 , , 631 , l45 annibali , f. , tosi , m. , monelli , m. , sirianni , m. , montegriffo , p. , aloisi , a. , & greggio , l. 2009 , , 138 , 169 annibali , f. , tosi , m. , aloisi , a. & van der marel , r. p. , 2011
, , submitted avni , y. 1976 , , 210 , 642 billett , o. h. , hunter , d. a. , & elmegreen , b. g. 2002 , , 123 , 1454 bomans , d. j. , van eymeren , j. , dettmar , r .- j . , weis , k. , & hopp , u. 2007 , , 51 , 141 cannon , j. m. , skillman , e. d. , garnett , d. r. , & dufour , r. j. 2002 , , 565 , 931 cervio , m. , & luridiana , v. 2004 , , 413 , 145 chandar , r. , leitherer , c. , tremonti , c. a. , calzetti , d. , aloisi , a. , meurer , g. r. , & de mello , d. 2005 , , 628 , 210 cignoni , m. , & tosi , m. 2010 , advances in astronomy , 2010 , davidson , k. , & kinman , t. d. 1985 , , 58 , 321 davidson , k. , kinman , t. d. , & friedman , s. d. 1989 , , 97 , 1591 dufour , r. j. , garnett , d. r. , & shields , g. a. 1988 , , 332 , 752 dufour , r. j. , & hester , j. j. 1990 , , 350 , 149 dufour , r. j. , esteban , c. , & castaneda , h. o. 1996a , , 471 , l87 dufour , r. j. , garnett , d. r. , skillman , e. d. , & shields , g. a. 1996b , from stars to galaxies : the impact of stellar physics on galaxy evolution , 98 , 358 fagotto , f. , bressan , a. , bertelli , g. , & chiosi , c. 1994 , , 104 , 365 fiorentino , g. , et al .
2010 , , 711 , 808 garnett , d. r. , skillman , e. d. , dufour , r. j. , & shields , g. a. 1997 , , 481 , 174 363 , 142 girardi , l. , et al . 2010 , , 724 , 1030 grimes , j. p. , et al .
2009 , , 181 , 272 holtzman , j. a. , burrows , c. j. , casertano , s. , hester , j. j. , trauger , j. t. , watson , a. m. , & worthey , g. 1995 , , 107 , 1065 hunter , d. a. , & thronson , h. a. , jr .
1995 , , 452 , 238 hunter , d. a. , oconnell , r. w. , gallagher , j. s. , & smecker - hane , t. a. 2000 , , 120 , 2383 izotov , y. i. , thuan , t. x. , & lipovetsky , v. a. 1994 , , 435 , 647 izotov , y. i. , & thuan , t. x. 1998 , , 497 , 227 izotov , y. i. , & thuan , t. x. 2004 , , 616 , 768 izotov , y. i. , guseva , n. g. , fricke , k. j. , & papaderos , p. 2009
, , 503 , 61 kunth , d. , lequeux , j. , sargent , w. l. w. , & viallefond , f. 1994 , , 282 , 709 larsen , s. s. , & richtler , t. 2000 , , 354 , 836 legrand , f. 2000 , , 44 , 345 lequeux , j. , peimbert , m. , rayo , j. f. , serrano , a. , & torres - peimbert , s. 1979 , , 80 , 155 lequeux , j. , & viallefond , f. 1980 , , 91 , 269 maz apellniz , j. 2009 , , 699 , 1938 marconi , g. , matteucci , f. , & tosi , m. 1994 , , 270 , 35 marconi , m. , et al .
2010 , , 713 , 615 marigo , p. , & girardi , l. 2007 , , 469 , 239 martin , c. l. 1996 , , 465 , 680 matteucci , f. , & tosi , m. 1985 , , 217 , 391 momany , y. , et al .
2005 , , 439 , 111 olive , k. a. , steigman , g. , & skillman , e. d. 1997 , , 483 , 788 origlia , l. , leitherer , c. , aloisi , a. , greggio , l. , & tosi , m. 2001 , , 122 , 815 pagel , b. e. j. , simonson , e. a. , terlevich , r. j. , & edmunds , m. g. 1992 , , 255 , 325 petrosian , a. r. , boulesteix , j. , comte , g. , kunth , d. , & lecoarer , e. 1997 , , 318 , 390 pilyugin , l. s. 1993 , , 277 , 42 sargent , w. l. w. , & searle , l. 1970 , , 162 , l155 searle , l. , & sargent , w. l. w. 1972 , , 173 , 25 searle , l. , sargent , w. l. w. , & bagnuolo , w. g. 1973 , , 179 , 427 silva - villa , e. , & larsen , s. s. 2011 , , 529 , a25 sirianni , m. , et al .
2005 , , 117 , 1049 skillman , e. d. , & kennicutt , r. c. , jr .
1993 , , 411 , 655 2005 , , 117 , 1049 stasiska , g. , & leitherer , c. 1996 , , 107 , 661 stetson , p. b. 1987 , , 99 , 191 stetson , p. b. 1994 , , 106 , 250 tolstoy , e. , hill , v. , & tosi , m. 2009 , , 47 , 371 tosi , m. , aloisi , a. , mack , j. , & maio , m. 2007 , iau symposium , 235 , 65 tosi , m. 2009 , iau symposium , 258 , 61 van der marel , r. p. , & cioni , m .- r . l. 2001 , , 122 , 1807 van zee , l. , westpfahl , d. , haynes , m. p. , & salzer , j. j. 1998 , , 115 , 1000 zwicky , f. 1966 , , 143 , 192
|
the evolutionary properties and spatial distribution of i zwicky 18 stellar populations are analyzed by means of hst / acs deep and accurate photometry .
the comparison of the resulting colour - magnitude diagrams with stellar evolution models indicates that stars of all ages are present in all the system components , including objects possibly up to 13 gyr old , intermediate age stars and very young ones .
the colour - magnitude diagrams show evidence of thermally pulsing asymptotic giant branch and carbon stars .
classical and ultra - long period cepheids , as well as long period variables have been measured .
about 20 objects could be unresolved star clusters , and are mostly concentrated in the north - west ( nw ) portion of the main body ( mb ) . if interpreted with simple stellar population models , these objects indicate a particularly active star formation over the past hundred myr in izw 18 .
the stellar spatial distribution shows that the younger ones are more centrally concentrated , while old and intermediate age stars are distributed homogeneously over the two bodies , although more easily detectable at the system periphery .
the oldest stars are best visible in the secondary body ( sb ) and in the south east ( se ) portion of the mb , where crowding is less severe , but are present also in the rest of the mb , although measured with larger uncertainties .
the youngest stars are a few myr old , are located predominantly in the mb and mostly concentrated in its nw portion .
the se portion of the mb appears to be in a similar , but not as young evolutionary stage as the nw , while the sb stars are older than at least 10 myr .
there is then a sequence of decreasing age of the younger stars from the secondary body to the se portion of the mb to the nw portion .
all our results suggest that izw18 is not atypical compared to other bcds .
|
to discover and understand dusty galaxies at the highest redshifts , the emission of [ cii ] 158 is the single most important atomic line feature because it is the strongest far - infrared line in most sources @xcite and also provides the best opportunity for high redshift determinations and source diagnostics using submillimeter and millimeter spectroscopic observations .
already , for example , [ cii ] has been measured with the atacama large millimeter array ( alma ) in sources with 4 @xmath8 z @xmath6 6 @xcite , demonstrating that large scale star formation extends to high redshifts .
the [ cii ] line should be a diagnostic of star formation because it is primarily associated with the photodissociation region ( pdr ) surrounding starbursts @xcite . the objective of our studies is to compare characteristics of [ cii ] emission with classifications and luminosities of sources determined from mid - infrared spectra , so that [ cii ] can be used alone as a quantitative diagnostic when no other spectroscopy is available .
this is especially important for dusty sources at high redshifts , when the rest frame far - infrared can be observed in submillimeter and millimeter wavelengths , but other rest frame wavelengths are not observationally accessible .
our observations consist of a sample of 130 sources which include a wide range of active galactic nucleus ( agn ) through starburst ( sb ) classifications , using [ cii ] observations made with the photodetector array camera and spectrometer ( pacs ; poglitsch et al . 2010 ) on the @xmath0 space observatory @xcite .
the [ cii ] results are compared to mid - infrared spectroscopic diagnostics determined with the @xmath1 infrared spectrograph ( irs ; houck et al .
the sample is chosen to minimize effects of differing spatial resolutions when comparing @xmath0 far infrared observations with @xmath1 mid - infrared slit spectroscopy by choosing sources that are unresolved .
sample selection , classification , and results for [ cii ] fluxes for 112 sources are described in @xcite ; hereinafter paper 1 " . in that paper
, it was found that the [ cii ] line flux correlates closely with the flux of the polycyclic aromatic hydrocarbon ( pah ) 11.3 feature independently of agn / starburst classification , log [ f([cii ] 158 ) /f(11.3
pah ) ] = -0.22 @xmath9 0.25 . as a result
, we concluded that [ cii ] line flux measures the same photodissociation regions ( pdrs ) associated with starbursts as the pah feature @xcite and , therefore , measures the starburst component of any source . a calibration of star formation rate ( sfr ) was derived by comparing [ cii ] luminosity @xmath5)$ ] to total infrared luminosity @xmath10 using the precepts of @xcite , with the result that log sfr = log @xmath2)$ ] - 7.08 @xmath9 0.3 , for sfr in m@xmath3 yr@xmath4 and @xmath5)$ ] in l@xmath3 .
a comprehensive summary is given by de looze et al.(2014 ) of all available [ cii ] measurements ( including those in paper 1 ) and their use as a sfr indicator through calibration with continuum luminosities ( including both ultraviolet and infrared ) .
their emphasis is on determining uncertainties within the sfr calibration arising from cosmic variance and on discussing why differences can arise among sources . for starbursts ,
the summary results are very similar to ours in paper 1 , with the same calibration and similar scatter for the relation between @xmath5)$ ] and sfr .
the uncertainties arising from cosmic variance among sources do not include systematic uncertainties in the assumptions used to transform total radiated luminosities from a star forming region into sfrs .
it is important , therefore , to determine independent calibrations of @xmath2)$ ] compared to sfr which depend on different precepts .
the primary objective of the present paper is to obtain an independent calibration of @xmath2)$ ] as a sfr indicator by comparing [ cii ] luminosities with mid - infrared [ neii ] and [ neiii ] emission line luminosities , calibrated for sfr by @xcite , using new measurements of high resolution irs spectra ( section 3.2 ) .
in addition , we determine the [ cii ] line to continuum ratio at 158 and use this to calibrate sfr from the continuum luminosity at 158 ( section 3.3 ) ; this allows an independent comparison to a previous calibration of sfr from 160 luminosity @xcite . as part of these analyses
, we also provide new [ cii ] observations of an additional 18 sources and provide new redshifts for all sources determined using only the [ cii ] emission line .
as summarized in paper 1 , the sample selected for [ cii ] observations with @xmath0 pacs derives from sources having both @xmath1 irs spectra and full photometry with the infrared astronomical satellite ( iras ) , and for which f@xmath11(irs)/f@xmath11(iras ) comparisons at 25 indicate that sources are unresolved ( because comparable source fluxes are measured with the slit of the irs compared to the large beam of iras ) .
sources encompass a wide range of classification , from agn through composite to starburst , derived uniformly from the strength of the pah 6.2 feature in irs spectra . our agn / starburst classification , based on strength of the 6.2 pah feature ,
is described in paper 1 and in @xcite .
the classification criterion is similar to that used in many previous studies , although different authors use different pah features and different methods for measuring strength ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the primary motive for using the 6.2 feature instead of the stronger 11.3 feature is to compare with sources having sufficiently high redshifts that the 11.3 feature is not visible in irs spectra .
our measurement is simply the observed ew(pah 6.2 ) determined in a uniform fashion without any modeling assumptions regarding the mix of pah features and/or the true level of underlying dust continuum .
this measurement is made with the smart software for irs spectra @xcite , and the ew(pah 6.2 ) is determined using a gaussian fit to the pah feature and a linear fit to the continuum beneath the feature within the range 5.5 to 6.9 .
measurements from irs low resolution spectra and the empirical correlation of ew(pah 6.2 ) with source classification are given in @xcite .
all spectra are available in the cassis spectral atlas @xcite . 12.81 to [ oiv ] 25.89 emission line fluxes compared to equivalent width of pah 6.2 feature in .
dividing lines show the division into classifications based on ew(pah 6.2 ) used throughout this paper .
crosses are sources classified as pure agn , asterisks as pure starbursts , and circles are composites with contribution from both .
triangles are lower limits , in which the [ oiv ] feature is not measured . ] for purposes of the present paper , the most important use of these diagnostics is to select sources confidently classified as pure starbursts so these can be used to determine various relations among [ cii ] , infrared luminosity , pah luminosities , and emission line luminosities that apply to pure starburst sources .
these relations provide the calibration of sfr .
the composite and agn sources then illustrate changes that occur with increasing fractions of agn luminosity added to the starburst .
the classification utilized in the following discussion is summarized in figure 1 using sources discussed below having both pah and emission line measurements . the quantitative division among pure agn " and pure starbursts " , with the intermediate
composite " systems is illustrated .
the ew boundaries which are shown derive by combining optical classifications and mid - infrared template spectra to determine classification boundaries as summarized in @xcite and in paper 1 .
starbursts have ew(pah 6.2 ) @xmath12 0.4 , composites have 0.1 @xmath8 ew(pah 6.2 ) @xmath8 0.4 , and agn have ew(pah 6.2 ) @xmath8 0.1 .
pure starbursts are defined as systems with @xmath12 90% of the mid - infrared luminosity arising from a starburst , and pure agn as systems with @xmath12 90% of the mid - infrared luminosity arising from the agn ( paper 1 ) .
composites have mixtures of these components . for the full sample of 130 sources in this paper , these criteria give 60 agn , 36 composites , and 34 starbursts .
the summary of results in figure 1 comparing the ratio of [ neii ] 12.81 to [ oiv ] 25.89 emission line fluxes with the classification derived from pah ew illustrates the conclusions from many previous studies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) that pah strength correlates with level of ionization as seen in emission lines , which confirms the agn / starburst classification derived from pah strength . both the ew(pah 6.2 ) from the starburst pdr and the relative [ neii ] emission line strength from the starburst hii region decrease as the starburst component decreases , because the agn adds additional hot dust continuum in the mid - infrared that diminishes ew(pah 6.2 ) and also adds increasing strength of the higher ionization [ oiv ] ( the ionization potential required to produce neii is 21.6 ev compared to 55 ev to produce oiv ) .
the [ cii ] observations are described in paper 1 , where examples of observed line profiles are illustrated .
all [ cii ] observations were made using the @xmath0 pacs instrument @xcite for line spectroscopy in point source chop nod mode with medium throw .
[ cii ] line fluxes are obtained by summing fluxes in the nine central equivalent spatial pixels , or spaxels " , produced by the pacs image slicer . a correction communicated to us by the pacs calibration team
is applied for flux that would fall outside these spaxels for an unresolved source , correcting the 9 spaxel flux by a factor of 1.16(@xmath13/158 ) @xmath14 , with wavelength @xmath13 depending on redshift .
data reduction in paper 1 was done with version 8 of the @xmath0 interactive processing environment ( hipe ) , and the pacsman " software @xcite is used for fitting line profiles and continuum . an additional 18 sources were subsequently observed in program lsargsya - ot2 .
these [ cii ] fluxes were derived in a similar way and are given in table 1 .
results for the total sample of 130 sources are summarized in table 2 where [ cii ] fluxes for the 112 sources from paper 1 are also reproduced .
line fluxes are determined from the flux within a fitted gaussian profile except for 12 sources ( noted in table 2 ) for which the [ cii ] profile is asymmetric or shows component structure .
for these 12 , the total line flux is the integrated flux including all components rather than the flux within a single gaussian fit .
all results are determined with hipe version 8 .
the new sources were also extracted with hipe version 10 , which has different calibration from version 8 . by reducing new sources with both v8 and v10 , we find a systematic flux difference such that v8 fluxes are a factor of 1.07 brighter compared to v10 . this difference is noted in table 1 , but fluxes listed for the new sources arise from v8 to be consistent with results in paper 1 .
pacs observations of the [ cii ] line are made in a continuous scanning mode that produces a `` data cloud '' containing a large number of separate flux measurements within individual wavelength elements because the spectrometer is read out every 1/8 s ( poglitsch et al . 2010 and pacs observer s manual ) during the total integration time of 579 s for these observations .
the pacsman uncertainties at each wavelength are determined from the dispersion among the individual measurements within the data cloud .
the s / n of the overall profile fit is determined using the uncertainties within these individual wavelength elements . examples of profiles and fits are shown in paper 1 , including a noisy profile with plotted uncertainties .
uncertainties of individual [ cii ] emission line fluxes are determined by the signal to noise ( s / n ) of the profile fit by pacsman in the brightest spaxel , with uncertainties listed in paper 1 and table 1 individually by source .
s / n is defined as the ( line flux)/(uncertainty in line flux ) arising from the profile fitting . if s / n @xmath8 3 , fluxes were listed as upper limits in paper 1 and are shown as upper limits in plots within the present paper . no new sources in table 1 are limits .
s / n for the remaining sources indicates a median line flux uncertainty of @xmath9 15% .
the systematic uncertainty for [ cii ] fluxes depends on pacs flux calibration , estimated as @xmath9 12% in the pacs spectroscopy performance and calibration document picc - kl - tn-041 .
( the offset we find between hipe v8 and v10 reductions is within this uncertainty . )
combining these two sources of uncertainty yields a typical combined uncertainty of @xmath9 20% for [ cii ] fluxes . in paper 1 , redshifts were listed based on previously determined optical redshifts , verified by irs mid - infrared emission lines . in tables 1 and 2 of the present paper , we list new redshifts derived from the [ cii ] line only .
redshifts generally agree to @xmath15 0.002 , but the [ cii ] redshifts are given here to allow further study of any systematic differences between [ cii ] emitting regions and optical or mid - infrared regions .
uncertainties in measured [ cii ] line velocities are dominated by centering effects , as described in the pacs observer s manual .
depending on precisely where within a spaxel the source is centered , the observed wavelength of the emission line can change by @xmath6 @xmath9 40 km s@xmath4 . to estimate
empirically the uncertainties arising both from this effect and from fitting the line profile , we compare in figure 2 the [ cii ] centroid velocity within the brightest spaxel relative to the velocity within the second brightest spaxel ( only for sources in which both spaxels have s / n @xmath12 3 ) .
this plot does not include the 12 sources noted in table 2 with asymmetric or component structure to the [ cii ] line profile , because these sources may really have spatially different components on a sub - spaxel scale . from figure 2 ,
the brightest spaxel systematically gives the same velocity as other spaxels ; the median velocity offset between the brightest and second brightest is less than 1 km s@xmath4 .
the random errors from source position and profile fitting are determined as the one sigma dispersion among the velocity differences shown between the brightest spaxel and the second brightest , which gives a dispersion of @xmath9 50 km s@xmath4 .
this is our measurement of empirical uncertainty for the measured [ cii ] velocities .
this dispersion is much larger than the formal centroiding uncertainty of the line fit for the brightest spaxel , typically only a few km s@xmath4 in the pacsman fits , indicating that velocity uncertainties are dominated by positional effects of the source within the spaxel .
the final redshift measurement which is reported in tables 1 and 2 is that of only the brightest spaxel . in section 3.3 , the line to continuum ratio ( equivalent width , ew ) for the [ cii ] line is discussed .
this is important for understanding the relation between far infrared dust continuum reradiation and the origin of the [ cii ] emission line .
uncertainties in ew arise from both the [ cii ] emission as uncertainties in the line flux , and as uncertainties in the level of the underlying continuum . usually , the latter is a much larger source of uncertainty because source continua are faint compared to the background .
these two uncertainties as determined by pacsman are summed quadratically to give total uncertainty in the ew which is listed in table 2 .
comparisons of [ cii ] with pah fluxes in paper 1 and below are made using the 11.3 pah feature because this is the strongest pah feature in irs spectra .
the flux ratios for the new sources are given in table 1 , using measurements of the 11.3 feature taken from @xcite as determined with a gaussian fit to the pah feature and a linear fit to the continuum beneath the feature from 10.5 to 12 .
line profile velocities in km s@xmath4 between the profile centroid velocity in the brightest spaxel ( v1 ) compared to the velocity in the second brightest spaxel ( v2 ) , using only sources in which both spaxels have s / n @xmath12 3 for the measured line flux ; line flux is shown in units of w m@xmath16 .
the empirical uncertainty of final [ cii ] velocity measurements is taken as the one sigma dispersion among these velocity differences of @xmath9 50 km s@xmath4 , which includes uncertainties arising both from profile fitting and from location of the source within a spaxel .
this plot does not include the 12 sources with asymmetric or component structure for which we only use the velocities from the brightest spaxel , because these are sources which may have real spatially different components . ] in paper i , [ cii ] observations were compared with pah strength using irs low resolution spectra .
all objects in that paper also have irs high resolution spectra , which derive from archival observations in various @xmath1 observing programs ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . for many sources , line fluxes have been previously published . for the present paper , we remeasure emission lines in these high resolution spectra with the goal of a uniformly measured data set that uses the final irs flux calibrations and noise masks ; the new measurements also provide an empirical determination of line flux uncertainties that arise from the profile fitting process and noise removal .
line fluxes are measured using gaussian fits in the smart analysis program @xcite , beginning with the post - basic calibrated data products in the @xmath1 heritage archive .
the line fit is on top of the underlying continuum which is a combination of background and real source continuum . because of the short irs high resolution slit
, simultaneous background measurements can not be subtracted from a single on - source observation so we do not determine the real source continuum before fitting the line .
therefore , equivalent widths are not measured .
high resolution spectra are especially important for the [ neii ] 12.81 line fluxes because an adjacent pah feature makes accurate [ neii ] measures impossible on low resolution spectra .
we treat the wings of the pah as an underlying continuum when measuring the [ neii ] flux . even after removal
of known bad pixels , the irs high resolution spectra are contaminated by noise spikes from individual rogue pixels " .
these rarely interfere with individual emission lines at known wavelengths , although they are a serious contaminant when searching for weak , unknown emission features . in those few cases when a single pixel noise spike interferes with an otherwise smooth emission line profile ,
such spikes are removed by visual inspection .
this is done independently for each spectrum from the two nods of an observation , and line profiles are fit independently for each nod . to estimate the uncertainty of line fluxes arising from the noise removal and from the line fitting process ,
line fluxes from the two spectra arising from the two nods were compared . as expected , line uncertainty is greater for fainter lines .
the dispersion of the ratio of measured fluxes between the two nods is @xmath9 9% for all lines , and @xmath9 6% for bright lines with log flux @xmath12 -20.2 ( units of w cm@xmath16 ) . for quadratically combining irs line flux uncertainty with the [ cii ] uncertainty of @xmath9 20%
, we adopt the larger 9% dispersion for all lines , giving a final uncertainty for flux ratios [ cii]/irs lines = @xmath9 22% .
llcccccccccc 1 & ngc1808 & 050742.32 - 373045.7 & 0.00337 & 0.57 & 1.21e-14 & 256.1 & -0.28 & -2.14 & 7.87 & -2.88 & 1342270686 + 2 & irasf07260 + 3955 & 072930.29 + 394941.0 & 0.07888 & 0.03 & 4.70e-17 & 22.5 & 0.14 & -2.42 & 8.25 & -3.34 & 1342251036 + 3 & irasf08076 + 3658 & 081056.13 + 364945.0 & 0.02156 & 0.51 & 3.16e-16 & 101.1 & 0.00 & -1.83 & 7.92 & -2.57 & 1342251347 + 4 & irasf08273 + 5543 & 083109.29 + 553316.0 & 0.04475 & 0.30 & 2.13e-17 & 6.0 & -0.56 & -2.47 & 7.39 & -3.80 & 1342270666 + 5 & irasf08344 + 5105 & 083803.63 + 505509.0 & 0.09688 & 0.36 & 8.87e-17 & 29.7 & -0.26 & -2.36 & 8.71 & -3.30 & 1342270665 + 6 & irasf09168 + 3308 & 091954.54 + 325559.0 & 0.10376 & 0.48 & 4.37e-16 & 83.7 & -0.07 & -2.00 & 8.81 & -2.76 & 1342270660 + 7 & irasf09414 + 4843 & 094442.24 + 482916.0 & 0.05500 & 0.06 & 7.03e-17 & 21.5 & -0.19 & -2.41 & 8.10 & -3.11 & 1342270663 + 8 & irasf09471 + 3158 & 095004.02 + 314442.0 & 0.01640 & 0.55 & 1.70e-16 & 69.5 & -0.49 & -2.33 & 7.41 & -2.86 & 1342271043 + 9 & mrk25 & 100351.90 + 592610.0 & 0.01007 & 0.45 & 2.47e-16 & 88.7 & -0.23 & -2.05 & 7.14 & -2.73 & 1342270667 + 10 & irasf10332 + 6338 & 103636.12 + 632222.0 & 0.03812 & 0.51 & 9.89e-17 & 34.2 & -0.38 & -2.29 & 7.92 & -2.96 & 1342270668 + 11 & irasf10590 + 6515 & 110213.02 + 645924.0 & 0.07777 & 0.02 & 4.69e-17 & 16.0 & -0.46 & -2.58 & 8.23 & -3.23 & 1342270669 + 12 & mrk206 & 122417.00 + 672624.0 & 0.00442 & 0.50 & 2.63e-16 & 87.5 & -0.09 & -1.93 & 6.46 & -2.68 & 1342270670 + 13 & irasf12538 + 6352 & 125554.00 + 633644.0 & 0.00943 & 0.58 & 5.54e-16 & 119.6 & -0.08 & -2.22 & 7.42 & -2.73 & 1342270671 + 14 & irasf13007 + 6405 & 130239.21 + 634929.0 & 0.04165 & 0.45 & 3.24e-17 & 11.0 & -0.50 & -2.36 & 7.52 & -3.29 & 1342270672 + 15 & iras16487 + 5447 & 164946.88 + 544235.4 & 0.10376 & 0.29 & 1.18e-16 & 34.2 & -0.07 & -2.14 & 8.90 & -3.28 & 1342270675 + 16 & iras16569 + 8105 & 165236.82 + 810016.6 & 0.04927 & 0.59 & 4.58e-16 & 98.9 & -0.15 & -2.00 & 8.81 & -2.59 & 1342270674 + 17 & iras17028 + 5817 & 170341.91 + 581344.4 & 0.10608 & 0.38 & 1.52e-16 & 26.3 & -0.09 & -2.14 & 9.03 & -3.13 & 1342270676 + 18 & iras18580 + 6527 & 185813.90 + 653124.0 & 0.17630 & 0.35 & 7.54e-17 & 11.2 & -0.36 & -2.20 & 9.19 & -3.02 & 1342270679 + llcccccccc 1 & mrk0334 & 0.02214 & 6.70e-16 & 0.22 & 0.60 & 0.53 & 1.36 & 0.85 & 1.84 @xmath9 0.30 + 2 & mcg-02 - 01 - 051/2 & 0.02724 & 1.18e-15 & 0.11 & 0.85 & 0.55 & 1.64 & 1.73 & 1.71 @xmath9 0.16 + 3 & iras00199 - 7426 & 0.09609 & 1.98e-16 & 0.00 & 1.03 & 0.5 & @xmath121.76 & @xmath121.18 & 0.47 @xmath9 0.09 + 4 & e12-g21 & 0.03295 & 5.76e-16 & 0.68 & 1.08 & 1.13 & 1.48 & 0.58 & 1.93 @xmath9 0.41 + 5 & irasf00456 - 2904sw & 0.10994 & 1.26e-16 & 0.14 & 0.89 & 0.18 & @xmath121.93 & @xmath121.14 & 0.53 @xmath9 0.15 + 6 & mcg-03 - 04 - 014 & 0.03519 & 1.38e-15 & 0.31 & 1.18 & 0.73 & @xmath122.04 & 1.30 & 1.52 @xmath9 0.15 + 7 & ngc0454 & 0.01214 & 1.55e-16 & 0.54 & 0.39 & @xmath121.26 & 0.99 & 0.08 & 1.21 @xmath9 1.46 + 8 & eso244-g012 & 0.02253 & 1.14e-15 & 0.01 & 0.68 & 0.39 & 1.72 & 1.43 & 1.26 @xmath9 0.15 + 9 & eso353-g020 & 0.01607 & 1.56e-15 & 0.31 & 1.37 & 1.11 & 2.35 & @xmath121.91 & 0.96 @xmath9 0.04 + 10 & irasf01364 - 1042 & 0.04840 & 1.32e-16 & 0.22 & @xmath121.56 & 1.05 & @xmath121.74 & @xmath121.29 & 0.32 @xmath9 0.06 + 11 & ugc01385 & 0.01848 & 5.70e-16 & 0.10 & 1.03 & 0.28 & @xmath122.09 & @xmath121.29 & 0.86 @xmath9 0.07 + 12 & ngc0788 & 0.01353 & 8.43e-17 & 0.14 & -0.19 & 0.20 & 0.09 & -0.42 & 1.01 @xmath9 0.19 + 13 & iras02054 + 0835 & 0.34499 & @xmath82.00e-17 & @xmath80.25 & & & @xmath80.52 & & -0.03 @xmath9 0.29 + 14 & mrk0590 & 0.02659 & 2.06e-16 & 0.42 & 0.94 & @xmath121.31 & 1.29 & 0.81 & 1.03 @xmath9 0.86 + 15 & ugc01845 & 0.01570 & 1.42e-15 & 0.20 & 1.20 & 0.87 & @xmath122.26 & @xmath122.08 & 0.86 @xmath9 0.04 + 16 & ic1816 & 0.01693 & 3.42e-16 & 0.37 & 0.26 & 0.55 & 0.76 & 0.39 & 1.07 @xmath9 0.39 + 17 & ngc0973 & 0.01619 & 2.46e-16 & 0.53 & 0.46 & 0.67 & 1.04 & 0.30 & 0.67 @xmath9 0.23 + 18 & irasf02437 + 2122 & 0.02331 & 1.50e-16 & -0.09 & 0.98 & 0.76 & @xmath121.79 & @xmath121.18 & 0.24 @xmath9 0.03 + 19 & ugc02369 & 0.03161 & 8.21e-16 & 0.23 & 1.15 & 0.69 & 1.67 & @xmath122.21 & 0.97 @xmath9 0.12 + 20 & mrk1066 & 0.01216 & 9.39e-16 & -0.03 & 0.31 & 0.32 & 1.00 & 0.45 & 0.90 @xmath9 0.06 + 21 & irasf03217 + 4022 & 0.02350 & 5.96e-16 & 0.27 & 1.25 & 0.95 & @xmath122.03 & @xmath121.66 & 0.61 @xmath9 0.04 + 22 & mrk0609 & 0.03460 & 5.44e-16 & 0.47 & 0.90 & 0.85 & @xmath121.72 & @xmath121.66 & 1.46 @xmath9 0.20 + 23 & irasf03359 + 1523 & 0.03566 & 4.60e-16 & 0.17 & 0.68 & 0.54 & 1.45 & @xmath121.29 & 0.87 @xmath9 0.18 + 24 & irasf03450 + 0055 & 0.03079 & 2.69e-17 & 0.42 & 0.03 & @xmath120.33 & 0.13 & -0.04 & 0.78 @xmath9 1.57 + 25 & iras03538 - 6432 & 0.30069 & @xmath82.00e-17 & @xmath8 - 0.28 & & & & & 0.10 @xmath9 0.14 + 26 & iras04103 - 2838 & 0.11773 & 8.53e-17 & -0.07 & 0.05 & 0.41 & 0.78 & 0.25 & 0.77 @xmath9 0.51 + 27 & iras04114 - 5117 & 0.12512 & 6.32e-17 & & & @xmath120.66 & & @xmath120.99 & 0.46 @xmath9 0.58 + 28 & eso420-g013 & 0.01207 & 1.38e-15 & 0.07 & 0.73 & 0.50 & 1.27 & 0.43 & 0.63 @xmath9 0.02 + 29 & 3c120 & 0.03315 & 3.62e-16 & 0.68 & 0.12 & 0.57 & 0.19 & -0.48 & 1.78 @xmath9 0.58 + 30 & eso203-ig001 & 0.05293 & 5.14e-17 & 0.50 & & @xmath120.78 & @xmath120.98 & @xmath120.41 & 0.19 @xmath9 0.04 + 31 & mcg-05 - 12 - 006 & 0.01852 & 3.74e-16 & -0.11 & 0.96 & 0.29 & @xmath121.90 & @xmath121.24 & 0.43 @xmath9 0.03 + 32 & ngc1808 & 0.00337 & 1.21e-14 & & & & & & 0.96 @xmath9 0.02 + 33 & ark120 & 0.03285 & 1.34e-16 & 0.77 & 0.76 & @xmath120.89 & 0.93 & 0.58 & 0.78 @xmath9 0.48 + 34 & viizw31 & 0.05430 & 8.55e-16 & 0.22 & 1.06 & 0.67 & @xmath122.21 & & 1.02 @xmath9 0.07 + 35 & irasf05187 - 1017 & 0.02861 & 2.20e-16 & 0.30 & 1.06 & 1.02 & @xmath121.91 & @xmath121.44 & 0.29 @xmath9 0.03 + 36 & 2masxj05580206 - 3820043 & 0.03421 & 1.29e-17 & -0.27 & -0.61 & -0.60 & @xmath12 - 0.15 & -0.59 & 0.58 @xmath9 0.98 + 37 & irasf06076 - 2139 & 0.03763 & 1.11e-16 & -0.14 & @xmath121.28 & 0.66 & @xmath121.22 & @xmath120.86 & 0.19 @xmath9 0.02 + 38 & iras06301 - 7934 & 0.15632 & 3.10e-17 & 0.34 & & @xmath120.78 & @xmath120.88 & @xmath120.49 & 0.16 @xmath9 0.10 + 39 & iras06361 - 6217 & 0.15997 & 3.38e-17 & -0.03 & 0.55 & @xmath120.88 & 0.59 & 0.18 & 0.41 @xmath9 0.78 + 40 & ngc2273 & 0.00618 & 7.18e-16 & 0.25 & 0.58 & 0.70 & 1.20 & 0.71 & 1.05 @xmath9 0.10 + 41 & ugc03608 & 0.02163 & 1.46e-15 & 0.50 & 1.30 & 0.75 & 2.26 & @xmath122.22 & 1.15 @xmath9 0.11 + 42 & irasf06592 - 6313 & 0.02304 & 2.51e-16 & -0.12 & 0.83 & 0.47 & @xmath121.92 & @xmath121.60 & 0.41 @xmath9 0.04 + 43 & irasf07260 + 3955 & 0.07887 & 4.70e-17 & & & & & & 1.63 @xmath9 3.55 + 44 & mrk0009 & 0.04009 & 1.12e-16 & 0.59 & 0.65 & @xmath121.08 & 0.55 & 0.36 & 0.76 @xmath9 0.32 + 45 & iras07598 + 6508 & 0.14866 & 2.87e-17 & -0.06 & 0.12 & @xmath120.95 & @xmath121.34 & @xmath120.52 & 0.21 @xmath9 0.17 + 46 & mrk0622 & 0.02351 & 8.36e-17 & -0.06 & 0.22 & 0.33 & 0.66 & @xmath120.68 & 0.48 @xmath9 0.17 + 47 & irasf08076 + 3658 & 0.02155 & 3.16e-16 & & & & & & 2.26 @xmath9 1.10 + 48 & irasf08273 + 5543 & 0.04474 & 2.13e-17 & & & & & & 0.09 @xmath9 0.09 + 49 & irasf08344 + 5105 & 0.09688 & 8.87e-17 & & & & & & 0.59 @xmath9 0.25 + 50 & eso60-ig016 & 0.04488 & 4.99e-16 & 0.36 & 0.44 & 0.60 & & @xmath121.32 & 1.26 @xmath9 0.24 + 51 & mrk0018 & 0.01133 & 6.47e-16 & 0.43 & 0.88 & 0.86 & @xmath121.95 & @xmath121.63 & 2.64 @xmath9 0.80 + 52 & irasf09168 + 3308 & 0.10346 & 4.37e-16 & & & & & & 1.04 @xmath9 0.36 + 53 & mcg-01 - 24 - 012 & 0.01981 & 1.39e-16 & 0.36 & 0.38 & 0.36 & 0.81 & 0.16 & 0.80 @xmath9 0.43 + 54 & mrk0705 & 0.02879 & 1.34e-16 & 0.41 & 0.41 & 0.69 & 0.86 & 0.28 & 1.15 @xmath9 0.83 + 55 & irasf09414 + 4843 & 0.05500 & 7.03e-17 & & & & & & 1.62 @xmath9 4.43 + 56 & irasf09471 + 3158 & 0.01639 & 1.70e-16 & & & & & & 1.02 @xmath9 0.26 + 57 & mrk25 & 0.01007 & 2.47e-16 & & & & & & 1.88 @xmath9 0.65 + 58 & irasf10038 - 3338 & 0.03421 & 4.03e-16 & 0.36 & 0.89 & 0.62 & @xmath121.86 & @xmath121.48 & 0.74 @xmath9 0.15 + 59 & irasf10332 + 6338 & 0.03811 & 9.89e-17 & & & & & & 2.13 @xmath9 1.80 + 60 & ngc3393 & 0.01258 & 3.44e-16 & 0.30 & -0.25 & 0.01 & -0.19 & -0.77 & 1.88 @xmath9 0.85 + 61 & irasf10590 + 6515 & 0.07777 & 4.69e-17 & & & & & & 0.64 @xmath9 0.96 + 62 & iras11119 + 3257 & 0.18900 & @xmath82.00e-17 & @xmath8 - 0.10 & @xmath80.07 & @xmath80.40 & & & 0.23 @xmath9 0.44 + 63 & eso319-g022 & 0.01634 & 4.00e-16 & 0.22 & 1.35 & 0.85 & @xmath121.61 & @xmath121.26 & 0.29 @xmath9 0.02 + 64 & iras12018 + 1941 & 0.16809 & 2.41e-17 & -0.07 & 0.84 & @xmath120.46 & @xmath121.36 & @xmath120.19 & 0.41 @xmath9 0.44 + 65 & ugc07064 & 0.02507 & 3.94e-16 & 0.40 & 0.78 & 0.99 & 1.03 & 0.54 & 0.88 @xmath9 0.22 + 66 & mrk206 & 0.00441 & 2.63e-16 & & & & & & 2.36 @xmath9 0.93 + 67 & ngc4507 & 0.01187 & 5.62e-16 & 0.25 & 0.28 & 0.55 & 0.78 & 0.20 & 1.41 @xmath9 0.37 + 68 & pg1244 + 026 & 0.04862 & 1.65e-17 & 0.02 & 0.11 & @xmath120.25 & @xmath120.49 & @xmath120.08 & 1.10 @xmath9 1.59 + 69 & iras12514 + 1027 & 0.31820 & @xmath82.00e-17 & @xmath8 - 0.04 & @xmath8 - 0.18 & & @xmath80.33 & & 0.02 @xmath9 0.11 + 70 & irasf12538 + 6352 & 0.00942 & 5.54e-16 & & & & & & 1.13 @xmath9 0.17 + 71 & irasf13007 + 6405 & 0.04165 & 3.24e-17 & & & & & & 0.96 @xmath9 1.49 + 72 & eso507-g070 & 0.02154 & 9.51e-16 & 0.33 & 1.11 & 0.99 & @xmath122.34 & @xmath121.74 & 0.81 @xmath9 0.04 + 73 & ngc4941 & 0.00372 & 1.71e-16 & 0.11 & -0.13 & 0.35 & 0.23 & -0.2 & 0.85 @xmath9 0.25 + 74 & eso323-g077 & 0.01532 & 8.27e-16 & 0.33 & 0.69 & 0.71 & 1.02 & 0.52 & 1.02 @xmath9 0.08 + 75 & mcg-03 - 34 - 064 & 0.01682 & 1.90e-16 & -0.42 & -0.76 & -0.19 & -0.40 & -0.73 & 0.47 @xmath9 0.09 + 76 & irasf13279 + 3401 & 0.02364 & 2.58e-17 & -0.34 & 1.09 & 0.80 & @xmath121.28 & @xmath121.26 & 0.22 @xmath9 0.15 + 77 & m-6 - 30 - 15 & 0.00800 & 7.04e-17 & 0.31 & 0.06 & -0.03 & 0.00 & & 0.75 @xmath9 0.37 + 78 & iras13342 + 3932 & 0.17962 & 6.54e-17 & 0.10 & 0.15 & 0.39 & 0.52 & -0.17 & 1.08 @xmath9 3.53 + 79 & iras13352 + 6402 & 0.23660 & @xmath82.00e-17 & @xmath8 - 0.15 & & & & & 0.21 @xmath9 0.30 + 80 & irasf13349 + 2438 & 0.10840 & 2.54e-17 & 0.06 & -0.01 & @xmath120.94 & -0.26 & -0.41 & 0.41 @xmath9 0.44 + 81 & ngc5347 & 0.00813 & 1.15e-16 & 0.38 & 0.43 & 0.93 & 0.85 & 0.24 & 0.62 @xmath9 0.24 + 82 & iras14026 + 4341 & 0.32330 & @xmath82.00e-17 & & & & & & -0.01 @xmath9 0.07 + 83 & oq+208 & 0.07673 & 6.28e-17 & 0.20 & 0.21 & @xmath120.77 & @xmath121.15 & @xmath120.55 & 0.99 @xmath9 0.97 + 84 & ngc5548 & 0.01726 & 1.70e-16 & 0.30 & 0.26 & 0.53 & 0.62 & 0.14 & 0.89 @xmath9 0.31 + 85 & mrk1490 & 0.02595 & 3.18e-16 & -0.16 & 0.95 & 0.51 & @xmath121.75 & @xmath121.76 & 0.45 @xmath9 0.03 + 86 & pg1426 + 015 & 0.08633 & 2.30e-17 & 0.28 & 0.00 & 0.00 & 0.22 & -0.31 & 0.17 @xmath9 0.19 + 87 & pg1440 + 356 & 0.07769 & 5.77e-17 & 0.15 & 0.17 & 0.23 & 0.57 & 0.18 & 0.55 @xmath9 0.40 + 88 & ngc5728 & 0.00949 & 1.25e-15 & 0.57 & 0.36 & 0.72 & 0.59 & 0.04 & 1.10 @xmath9 0.06 + 89 & ngc5793 & 0.01164 & 8.98e-16 & 0.23 & 0.99 & 0.91 & @xmath122.14 & 1.07 & 0.89 @xmath9 0.05 + 90 & iras15001 + 1433 & 0.16236 & 6.40e-17 & 0.01 & 0.38 & 0.41 & 1.12 & @xmath120.52 & 0.43 @xmath9 0.38 + 91 & iras15225 + 2350 & 0.13875 & 3.23e-17 & -0.04 & 0.52 & & 0.14 & & 0.42 @xmath9 0.31 + 92 & mrk0876 & 0.12941 & 2.99e-17 & -0.09 & -0.00 & -0.15 & 0.35 & -0.28 & 0.28 @xmath9 0.43 + 93 & irasf16164 - 0746 & 0.02350 & 7.51e-16 & 0.19 & 0.74 & 1.07 & 1.84 & 1.02 & 0.65 @xmath9 0.03 + 94 & mrk0883 & 0.03827 & 1.84e-16 & 0.19 & 0.37 & 0.45 & 1.00 & 0.28 & 1.46 @xmath9 1.06 + 95 & cgcg052 - 037 & 0.02478 & 1.13e-15 & 0.19 & 1.32 & 0.75 & @xmath122.49 & @xmath122.28 & 0.98 @xmath9 0.06 + 96 & iras16334 + 4630 & 0.19085 & 3.62e-17 & -0.16 & 0.51 & 0.32 & 0.55 & @xmath120.81 & 0.49 @xmath9 0.73 + 97 & eso069-ig006 & 0.04644 & 1.08e-15 & 0.16 & 0.93 & 0.67 & @xmath121.97 & @xmath121.39 & 0.81 @xmath9 0.04 + 98 & irasf16399 - 0937n & 0.02728 & 1.49e-15 & 0.86 & 1.38 & 1.53 & @xmath122.49 & 1.47 & 1.27 @xmath9 0.65 + 99 & iras16487 + 5447 & 0.10361 & 11.7e-17 & & & & & & 0.67 @xmath9 0.46 + 100 & iras16569 + 8105 & 0.04926 & 4.58e-16 & & & & & & 2.06 @xmath9 0.68 + 101 &
2massj165939.77 + 183436.9 & 0.17069 & @xmath82.00e-17 & @xmath80.18 & @xmath8 - 0.17 & & @xmath8 - 0.04 & & 0.19 @xmath9 0.39 + 102 & pg1700 + 518 & 0.29200 & @xmath82.00e-17 & @xmath80.09 & & & & & -0.07 @xmath9 0.27 + 103 & iras17028 + 5817 & 0.10608 & 1.52e-16 & & & & & & 0.54 @xmath9 0.18 + 104 & iras17044 + 6720 & 0.13530 & 2.06e-17 & -0.43 & 0.16 & & 0.37 & & 0.39 @xmath9 0.35 + 105 & iras17068 + 4027 & 0.17933 & 3.37e-17 & -0.19 & -0.01 & 0.10 & @xmath121.11 & @xmath120.78 & 0.24 @xmath9 0.48 + 106 & irasf17132 + 5313 & 0.05110 & 6.09e-16 & 0.21 & 1.03 & 0.64 & @xmath122.19 & @xmath121.94 & 1.45 @xmath9 0.32 + 107 & eso138-g027 & 0.02093 & 1.44e-15 & 0.52 & 1.43 & 0.88 & @xmath122.59 & @xmath121.89 & 1.10 @xmath9 0.08 + 108 & cgcg141 - 034 & 0.02015 & 6.65e-16 & 0.25 & 0.99 & 0.75 & 2.07 & 1.18 & 0.79 @xmath9 0.07 + 109 & h1821 + 643 & 0.29700 & @xmath82.00e-17 & @xmath8 - 0.21 & @xmath8 - 0.7 & @xmath8 - 0.18 & @xmath8 - 0.46 & @xmath8 - 1.04 & 0.22 @xmath9 0.37 + 110 & ic4734 & 0.01548 & 1.36e-15 & 0.34 & 1.39 & 0.96 & 2.18 & @xmath122.15 & 0.43 @xmath9 0.02 + 111 & iras18443 + 7433 & 0.13417 & 2.31e-17 & -0.05 & 0.06 & @xmath120.83 & @xmath120.76 & @xmath120.91 & 0.32 @xmath9 0.45 + 112 & eso140-g043 & 0.01419 & 1.76e-16 & 0.28 & 0.111 & 0.31 & 0.33 & -0.12 & 1.04 @xmath9 0.36 + 113 & 1h1836 - 786 & 0.07478 & 2.49e-17 & 0.21 & 0.34 & @xmath120.211 & 0.41 & -0.32 & 0.27 @xmath9 0.18 + 114 & iras18580 + 6527 & 0.17630 & 7.54e-17 & & & & & & 0.67 @xmath9 0.31 + 115 & eso593-ig008 & 0.04923 & 9.51e-16 & 0.44 & 1.07 & 0.66 & 2.09 & @xmath121.71 & 0.77 @xmath9 0.06 + 116 & eso-141-g055 & 0.03725 & 1.32e-16 & 0.69 & 0.41 & @xmath120.99 & 0.35 & 0.09 & 0.87 @xmath9 0.40 + 117 & eso339-g011 & 0.01918 & 1.54e-15 & 0.56 & 0.56 & 0.85 & 1.33 & 0.54 & 1.08 @xmath9 0.09 + 118 & iras20037 - 1547 & 0.19219 & 4.63e-17 & -0.04 & @xmath120.62 & 0.38 & @xmath121.16 & @xmath120.96 & 0.30 @xmath9 0.32 + 119 & ngc6860 & 0.01479 & 3.69e-16 & 0.89 & 0.73 & 0.97 & 0.96 & 0.52 & 1.58 @xmath9 0.93 + 120 & eso286-g035 & 0.01748 & 1.63e-15 & 0.53 & 1.45 & 0.91 & @xmath122.65 & @xmath122.28 & 1.28 @xmath9 0.07 + 121 & ngc7213 & 0.00598 & 2.74e-16 & 0.06 & 0.33 & 0.94 & @xmath121.69 & @xmath121 & 0.87 @xmath9 0.26 + 122 & eso602-g025 & 0.02529 & 1.24e-15 & 0.41 & 1.22 & 1.04 & 2.15 & 1.37 & 1.39 @xmath9 0.10 + 123 & ugc12138 & 0.02516 & 1.43e-16 & 0.08 & 0.41 & @xmath121.23 & 0.79 & 0.16 & 1.50 @xmath9 1.67 + 124 & ugc12150 & 0.02171 & 8.96e-16 & 0.24 & 1.37 & 0.77 & @xmath122.31 & @xmath121.86 & 0.56 @xmath9 0.03 + 125 & eso239-ig002 & 0.04299 & 1.91e-16 & -0.14 & 0.76 & @xmath121.48 & @xmath121.71 & @xmath121.34 & 0.31 @xmath9 0.04 + 126 & zw453.062 & 0.02493 & 1.05e-15 & 0.61 & 1.20 & 1.24 & 2.27 & 1.27 & 0.90 @xmath9 0.07 + 127 & iras23060 + 0505 & 0.17300 & @xmath82.00e-17 & @xmath8 - 0.17 & @xmath8 - 0.14 & @xmath80.19 & @xmath8 - 0.02 & @xmath8 - 0.17 & 0.17 @xmath9 0.24 + 128 & ngc7603 & 0.02952 & 2.63e-16 & 0.36 & 0.73 & 0.69 & @xmath121.33 & 0.98 & 0.76 @xmath9 0.12 + 129 & mcg-83 - 1 & 0.02317 & 1.47e-15 & 0.44 & 1.17 & 0.87 & @xmath122.15 & @xmath122.33 & 1.71 @xmath9 0.22 + 130 & cgcg381 - 051 & 0.03091 & 2.34e-16 & 0.16 & 1.01 & 0.65 & @xmath121.49 & @xmath121.29 & 0.66 @xmath9 0.21 +
158 to pah 11.3 line fluxes , compared to source classification from ew(pah 6.2 ) measured in , for all sources .
crosses are agn from the ew(pah 6.2 ) classification , open circles are composite agn plus starburst , and asterisks are starbursts . sources with diamonds ( all agn ) are upper limits to [ cii ] line fluxes .
horizontal bars are medians within each category ; medians include limits because all limits fall below the median .
vertical error bar shows the observational line ratio uncertainty for individual points of @xmath9 22% derived in the text . ]
158 to [ neii ] 12.81 line fluxes , compared to source classification from ew(pah 6.2 ) measured in .
crosses are agn from the ew(pah 6.2 ) classification , open circles are composite agn plus starburst , and asterisks are starbursts . sources with diamonds ( all agn ) are upper limits to ratio because are limits to [ cii ] line fluxes but detections in [ neii ] .
horizontal bars are medians within each category , including limits . vertical error bar shows the observational line ratio uncertainty for individual points of @xmath9 22% derived in the text . ]
158 to [ neiii ] 15.55 line fluxes , compared to source classification from ew(pah 6.2 ) measured in .
crosses are agn from the ew(pah 6.2 ) classification , open circles are composite agn plus starburst , and asterisks are starbursts .
sources with diamonds ( all agn ) are upper limits to ratio because are limits to [ cii ] line fluxes but detections in [ neiii ] .
triangles are lower limits to the ratio because values arise from limits for [ neiii ] but detections in [ cii ] .
horizontal bars are medians within each category , including limits .
vertical error bar shows the observational line ratio uncertainty for individual points of @xmath9 22% derived in the text . ]
158 to [ neii ] 12.81 + [ neiii ] 15.55 line fluxes , compared to source classification from ew(pah 6.2 ) measured in .
crosses are agn from the ew(pah 6.2 ) classification , open circles are composite agn plus starburst , and asterisks are starbursts . sources with diamonds are upper limits to the ratio because values are limits to [ cii ] line fluxes but detections in [ neiii ] and/or [ neii ] .
horizontal bars are medians within each category , including limits . vertical error bar shows the observational line ratio uncertainty for individual points of @xmath9 22% derived in the text . ] in figures 3 - 6 , the [ cii ] flux is compared with mid - infrared features from irs spectra .
all scales are the same so ratios can be compared among the plots .
figure 3 repeats the comparison with pah 11.3 from paper 1 , adding the new sources in table 1 .
figure 4 compares [ cii ] to [ neii ] , figure 5 [ cii ] to [ neiii ] , and figure 6 [ cii ] to [ neii]+[neiii ] . the comparison with pah in figure 3 reconfirms the conclusion in paper 1 , that the scaling between [ cii ] and pah is independent of source classification , implying that the starburst component and sfr is measured equally well by pah emission or by [ cii ] emission in sources of all classifications . the correlation of [ cii ] with a sfr indicator can be checked in a completely independent way using figure 4 , which compares [ cii ] with [ neii ] , because [ neii ] is a luminosity indicator of star formation determined by the hii region instead of the surrounding pdr .
figure 4 shows a result very similar to the [ cii]/pah comparison .
the median ratios f([cii])/f([neii ] ) are independent of classification .
the dispersion about the median is also similar to the dispersion in [ cii]/pah .
because [ neii ] arises primarily from starbursts , this result is very important empirically because it confirms the previous conclusion from pah that [ cii ] scales with the starburst component within sources of all classifications .
figure 5 compares [ cii ] with [ neiii ] .
there is a greater difference of ratio with classification than for [ neii ] .
the increasing ratio f([cii])/f([neiii ] ) from agn to starbursts can be explained if [ neiii ] arises both from agn ionization and starburst ionization , but [ cii ] arises only from starbursts , so that agn contain an additional [ neiii ] component compared to [ cii ] .
the implications of these results are discussed in the next section .
the [ cii ] line is expected to arise primarily within the photodissociation region ( pdr ) surrounding star forming regions @xcite . as the primary cooling line for
low ionization regions , [ cii ] can also arise within more diffuse emission regions associated with the infrared cirrus " , for which the ionization and heating of cii ( and the accompanying far infrared dust continuum ) is not necessarily a measure of the young , ongoing starbursts .
this diffuse [ cii ] is observed in nearby , resolved galaxies @xcite .
the scaling of [ cii ] with pah found in paper 1 for luminous , unresolved sources indicates , however , that [ cii ] primarily measures the pdrs in spatially integrated observations that include entire galaxies rather than individual regions within galaxies . in paper 1 ,
sfr was determined by comparison of [ cii ] to total infrared luminosity , using the assumption @xcite that all of the radiation from ongoing star formation is absorbed and reemitted in the infrared spectrum by the dust .
a completely independent estimate of sfr can be made by ionization models of the hii region , where the emission lines are related only to the ultraviolet ionizing continuum of the starburst .
a great advantage of using mid - infrared emission lines for comparison to the models is that they are much less subject to uncertainties regarding extinction corrections compared to optical lines .
starburst models have been scaled to the luminosity of the infrared [ neii ] and [ neiii ] lines by @xcite and discussed in comparison with @xmath1 irs observations by @xcite .
the ho and keto result is sfr ( m@xmath3 yr@xmath4 ) = 4.34 x 10@xmath18 l([neii]+[neiii ] ) ( erg s@xmath4 ) , estimating about a factor of two uncertainty for individual sources depending on the fraction of ionizing photons which are absorbed by gas and the fractional ionizing state of neon .
these differences arising from cosmic variance among sources can explain the dispersion in the [ cii]/([neii]+[neiii ] ) ratio in figure 6 , for example . in what follows ,
we use only the pure starbursts in figure 6 for comparisons to avoid any uncertainty regarding neon luminosity from an agn component .
figure 6 shows the scaling with [ cii ] . for the pure starbursts in our sample , the result is log f[cii]/(f[neii]+ne[iii ] ) = 0.16 @xmath9 0.15 .
composites give a similar result , but , as expected from the individual plots for these lines in figures 4 and 5 , the value for agn is lower .
this scaling for pure starbursts applied to the ho and keto relation gives the result that sfr ( m@xmath3 yr@xmath4 ) = 3 x 10@xmath18 @xmath5)$ ] ( erg s@xmath4 ) , or log sfr = log @xmath5)$ ] - 6.93 @xmath9 0.15 for sfr in m@xmath3 yr@xmath4 and @xmath5)$ ] in l@xmath3 .
our result derived in paper 1 from calibrating sfr using the total infrared dust luminosity was log sfr = log @xmath2)$ ] - 7.08 @xmath9 0.3 , so the two independent results agree well within the cosmic variance among sources .
the systematic difference in sfr calibration is less than the variance within either calibration . from this , we conclude that the relation log sfr = log @xmath2)$ ] - 7.0 can measure sfr to a precision of @xmath9 @xmath15 50% for any individual source within the uncertainties arising from cosmic variance .
our conclusion that [ cii ] is a reliable sfr indicator differs from the conclusions of @xcite and @xcite .
this difference arises primarily from our calibration of sfr derived from either the bolometric luminosities or the neon line luminosities using only pure starbursts based on the pah classification criterion .
the spectral energy distributions of dusty sources show large variations , with changes in characteristic temperatures that may depend on luminosity or redshift ( e.g. * ? ? ?
* ; * ? ? ?
determining the relative agn / starburst contribution to dust luminosity at different infrared wavelengths is a complex problem to which various diagnostics can be applied @xcite . by using only a subset of sources having no indications of any agn contribution , contamination from agn luminosity can not affect the calibration of sfr , and it is not necessary to determine quantitatively the agn / starburst fraction within individual sources to determine a sfr calibration .
this issue and the differences from previous conclusions are considered further in the next section .
all of the results summarized above show consistent correlations between the [ cii ] luminosity and the luminosity of the starburst pdrs measured using the [ cii]/pah ratio and from the starburst hii regions measured with the neon lines .
these results imply that [ cii ] measures the same starbursts as are measured with the mid - infrared diagnostics .
are there any sources with indications of excess [ cii ] luminosity that may not be associated with the starburst pdr or hii region luminosities ?
additional [ cii ] could arise from lower ionization regions associated with an agn or from widespread low ionization from older starbursts and association with infrared cirrus @xcite .
there could also be excess [ cii ] compared to the mid - infrared diagnostics if starbursts are so dusty that the mid - infrared features suffer significant extinction compared to [ cii ] ( e.g. farrah et al .
2007 , 2013 ) .
the ratios among [ cii]/pah/[neii ] in figures 3 and 4 show median ratios that are the same for all classifications from agn through starburst .
the sources with the largest [ cii ] excess are among composites and agn , so those are the best candidates for sources with additional [ cii ] not associated with the starbursts . for this reason
, we attempt an estimate of the excess [ cii ] luminosity which may be present in composites and agn . to do this , we determine the [ cii ] luminosity remaining in the agn and composite sources , @xmath5,other)$ ] , after subtracting from the total @xmath5)$ ] the luminosity component from starbursts , @xmath5,sb)$ ] , assuming the median observed ratio log @xmath5,sb)$]/@xmath19 = 0.22 for starbursts in figure 3 .
for example , if log @xmath5,other)$]/@xmath5)$ ] = -0.3 , this means that 50% of the observed @xmath5)$ ] does not arise from the starburst component . in many cases ,
this leads to @xmath5,sb)$ ] @xmath12 @xmath5)$ ] . in these cases ,
we arbitrarily adopt a limit @xmath5,other)$ ] @xmath8 0.1 @xmath5)$ ] .
because of the scatter of @xmath6 0.2 dex in the @xmath5,sb)$]/@xmath19 ratio for individual starbursts ( figure 3 ) , only statistical estimates for the overall samples can be determined in this way .
of the 36 composite sources , only 4 have log @xmath5,other)$]/@xmath5)$ ] @xmath12 -0.3 .
this means that [ cii ] luminosity from alternative sources may exceed [ cii ] luminosity from conventional starbursts in only 11% of composite sources .
for agn , only 14 of 60 agn exceed log @xmath5,other)$]/@xmath5)$ ] = -0.3 , meaning that 23% of agn may have excess [ cii ] luminosity that exceeds the starburst [ cii ] luminosity . while these estimates are approximate and qualitative , they indicate that there is no strong evidence within our sample for sources in which [ cii ] luminosity arising from other mechanisms exceeds the [ cii ] luminosity from ongoing star formation the comparisons of [ cii ] with neon line fluxes in the preceding section confirm , as did the pah comparisons discussed in paper 1 , that the [ cii ] luminosity gives the same measure of sfr as is found using the mid - infrared diagnostics .
a related and important question is how sfr measured with [ cii ] compares with sfr measured using only the infrared continuum luminosity .
this is crucial because of the common use of the continuum luminosity alone , particularly the far infrared continuum , to measure the evolution of star formation in the universe @xcite .
the biggest question in the comparison of continuum luminosity and @xmath2)$ ] arises from previous observations that more luminous sources in @xmath20 have relatively weaker @xmath2)$ ] , termed the [ cii ] deficit " @xcite .
the @xmath20 is usually defined as the luminosity within 40 @xmath15 @xmath13 @xmath15 125 , measured quantitatively using the 60 and 100 photometry with iras according to the formulation in @xcite .
a deficit could be explained by starbursts with increasing ionization parameter and harder ionizing radiation in more luminous sources @xcite or by having dustier , more obscured starbursts @xcite .
if @xmath20 invariably measures sfr , then a deficit would mean that the sfr from [ cii ] is systematically underestimated for the most luminous sources . in paper 1
, we concluded that the deficit is not a consequence of differences among starbursts but instead arises because more luminous sources have increasing contributions from agn to the dust continuum luminosity .
this conclusion arose because luminous sources with an apparent deficit show the mid - infrared diagnostic of agn ( weak pah ) , indicating that most continuum luminosity is produced by agn dust heating .
this is illustrated in figure 9 of paper 1 , where our sources reach @xmath10 @xmath21 10@xmath22 l@xmath3 , but all sources except one having @xmath10 @xmath12 10@xmath23 l@xmath3 contain agn .
the one starburst does not show a deficit compared to lower luminosity starbursts , but the high luminosity agn do show a systematic deficit .
we illustrate this result again in figure 7 using the conventional parameters for the deficit , the ratio @xmath2)$]/@xmath20 compared to @xmath20 , determining @xmath20 from the iras fluxes tabulated in @xcite .
the plot is made only for pure starbursts , as defined by our classification criterion ew(pah 6.2 ) @xmath12 0.4 .
the distribution of results for our sample ( crosses ) confirms the conclusion of paper 1 - that no trend of ratio with luminosity is seen in the @xmath2)$]/@xmath20 ratio when only pure starbursts are used .
figure 7 also includes results ( circles ) for starbursts in the [ cii ] sample of @xcite defined by their stated criterion for pure starbursts , which is ew(pah 6.2 ) @xmath12 0.5 using pah measurements from @xcite .
these results show an overall median that agrees with ours . within the limited range of @xmath20 , any trend for @xmath2)$]/@xmath20 to decrease with @xmath20 in the diaz - santos sample
is less than the one sigma dispersion among starbursts at any given luminosity .
for example , from log @xmath20(l@xmath3 ) = 10.5 to log @xmath20(l@xmath3 ) = 11.5 , the median ratio decreases by a factor of two , but the dispersions at both luminosities exceed a factor of 2.5 .
their sources with 10 @xmath8 log @xmath20(l@xmath3 ) @xmath8 11 exceed our median ratio by only a factor of 1.5 , and their sources with 11 @xmath8 log @xmath20(l@xmath3 ) @xmath8 12 are below our median ratio by a factor of 0.8 .
these comparisons indicate consistency between the two samples of starbursts . to consider further the deficit issue , we can extend consideration of the [ cii ] line to continuum ratio to longer wavelengths ( representing emission from cooler dust ) than measured by @xmath20 .
this can be done because the continuum at rest frame 158 is also measurable for most of our sources having [ cii ] line observations . attributing a deficit to luminous starbursts invokes the assumption that the coolest dust in luminous sources is invariably heated by starbursts instead of agn .
even if the far - infrared continuum at @xmath15 100 can include contributions from agn heating , as we have concluded , it might be expected that cooler dust seen at longer wavelengths would be increasingly dominated by the starburst component .
if a deficit really exists among pure starbursts , therefore , the deficit should be as readily seen comparing @xmath2)$ ] to @xmath7(158 ) as when comparing to @xmath20 .
conversely , if the apparent deficit arises because @xmath20 is contaminated by hotter agn dust , any deficit arising because agn are included in the sample should diminish when measured with the cooler dust seen at 158 . )
/@xmath20 compared to @xmath20 for starbursts in l@xmath3 .
asterisks are starbursts from current sample and thick horizontal line is their median .
circles are starbursts from @xcite using their criterion for pure starbursts ; thin horizontal lines and error bars are medians and one sigma dispersions within this sample for 10 @xmath8 log @xmath20 @xmath8 11 , and 11 @xmath8 log @xmath20 @xmath8 12 . ]
( 158 ) in erg s@xmath4 , compared to equivalent width of [ cii ] emission line in for all sources in table 2 .
crosses are agn from the ew(pah 6.2 ) classification , open circles are composite agn plus starburst , and asterisks are starbursts .
error bar shows median uncertainty in luminosity determined by observational uncertainty of continuum flux measurement . ] the strength of [ cii ] compared to the continuum is defined by the equivalent width ( ew ) , which is a linear measurement of the line to continuum ratio .
ew relates to continuum luminosities by @xmath7(158 ) = @xmath24(158 ) = @xmath13@xmath5)$]/ ew([cii ] ) for @xmath13 = 158 .
measuring @xmath2)$ ] and @xmath7(158 ) with the same pacs observation has an additional advantage in that the spatial resolution of both measures is identical . for sources that may be resolved
, this removes one concern when comparing @xmath2)$ ] with @xmath20 determined from iras photometry because the iras spatial resolution is much poorer than the spaxels of pacs .
the [ cii ] line fluxes presented in section 2.2 are measured by summing over 9 spaxels .
the continuum flux densities are much weaker than the line and are often so weak in outlying spaxels that they are overwhelmed by noise . to minimize the effects of continuum noise , the continuum @xmath25(158 )
is determined by using the total @xmath26([cii ) ] ) from table 2 combined with ew([cii ] ) measured only in the brightest spaxel .
these ew are measured in all of our spectra and given in table 2 .
the uncertainties are often large , primarily because the continua are faint and have large uncertainties after backgrounds are subtracted .
the ew uncertainties that are given in table 2 include the combined uncertainty of the total line flux and of the continuum level in the brightest spaxel , but uncertainties are dominated by the uncertainty in the continuum .
the @xmath7(158 ) is compared to ew([cii ] ) in figure 8 for all of the [ cii ] detections in table 2 including agn , composites and starbursts ( the 10 non - detections are not used for this plot because their limits apply to both coordinates ) .
coordinates are in log units to make direct comparisons with the line to continuum ratio plots @xmath2)$]/@xmath20 used to discuss the deficit in previous references .
figure 8 shows similar ranges and dispersions of both values in all of the distributions for agn , composites , and starbursts .
the cosmic variance among individual sources is greater than any systematic differences between classes ( shown more quantitatively in the discussion below of figure 9 ) .
the overall conclusion from figure 8 is that the appearance of a
deficit " is much less conspicuous when using @xmath7(158 ) than when using @xmath20 .
for example , comparisons can be made with figure 1 of @xcite that compares @xmath2)$]/@xmath20 to @xmath20 for their combined sample of agn , composites , and starbursts .
the luminosity range is comparable to that in figure 8 , @xmath21 10@xmath27 , but the range of line to continuum ratio is much greater in the comparison with @xmath20 ( @xmath6 100 ) , compared to a range of @xmath6 10 in figure 8 .
another conspicuous difference is the clear trend for the upper envelope of points to show a deficit when @xmath5)$ ] is compared to @xmath20 , but this trend is not apparent in figure 8 .
some high luminosity sources show ew as large as some low luminosity sources .
these results indicate that the apparent deficit becomes less significant when continuum luminosity is measured at the longest wavelengths , consistent with the interpretation that the coolest dust arises in the same starbursts measured with the [ cii ] , pah , and neon features for starbursts of all luminosities .
this conclusion can be further illustrated by comparing ew([cii ] ) with source classification based on ew(pah 6.2 ) ; if @xmath7(158 ) and l([cii ) ] ) both measure only the starburst component of any source , the ew([cii ] ) should be independent of source classification . the results are shown in figure 9 , showing all ew([cii ] ) in table 2 .
the scatter increases for agn compared to starbursts primarily because of increased uncertainties caused by fainter continua . for starbursts , the typical observational uncertainty in ew(pah 6.2 )
given in table 2 is only @xmath6 10% because the lines and continua are strong , but this uncertainty rises to @xmath6 50% for the weak lines and continua of the agn . 158 emission line in , with uncertainties , compared to ew(pah 6.2 ) classification , in .
medians are shown as horizontal lines in each class .
error bars show median one sigma uncertainties of the individual ew measurements for different classifications using uncertainties for each source given in table 2 .
sources with upper limits in both ews are not plotted or included in medians , but sources with upper limits only in ew([cii ] ) are shown as small symbols ( all agn ) . ]
the crucial result in figure 9 is that the median ew([cii ] ) is only marginally greater for starbursts compared to composites and agn ; within the uncertainty , a median value of ew([cii ] ) = 0.9 falls within all error bars . for starbursts , the median ew([cii ] ) = 1.0 , decreasing to ew([cii ] ) = [email protected] for agn , and the difference is smaller than the dispersion in ew([cii ] ) among individual starbursts .
the ew([cii ] ) changes much less than ew(pah 6.2 ) .
the ew(pah ) ranges from @xmath8 0.01 to 0.6 , a range of more than 60 .
this happens because an agn has greatly increased mid - infrared continuum from warm dust compared to a starburst , thereby decreasing ew(pah 6.2 ) .
by contrast , the ew([cii ] ) ranges from @xmath6 0.2 to @xmath8 2 , a range of @xmath6 10 .
this result indicates quantitatively that the far infrared continuum at 158 scales much more closely with the starburst component of any source than does the continuum at shorter wavelengths .
this result illustrates why the @xmath7(158 ) could be a better estimate of sfr regardless of source classification than is @xmath20 , which includes shorter wavelength emission from hotter dust arising around agn .
the preceding section shows that there is no conspicuous deficit when @xmath2)$ ] is compared to @xmath7(158 ) for the full sample of agn , composites and starbursts over all luminosities . any systematic change in @xmath2)$]/@xmath7(158 ) with luminosity or with classification
is small compared to the the cosmic variance at a given luminosity .
this result encourages the conclusion that the far infrared luminosity at 158 scales similarly with @xmath2)$ ] in sources of all classifications and luminosity .
this would make @xmath7(158 ) a more reliable measure of sfr than the @xmath20 , if the latter sometimes includes an agn component of luminosity .
it was concluded in section 3.2 that log sfr = log @xmath5)$ ] - 6.93 @xmath9 0.15 for sfr in m@xmath3 yr@xmath4 and @xmath5)$ ] in l@xmath3 , calibrated by comparison with the [ neii ] and [ neiii ] luminosities .
the result derived independently in paper 1 by comparing [ cii ] luminosities with sfr derived from total infrared dust luminosity as in @xcite was log sfr = log @xmath2)$ ] - 7.08 @xmath9 0.3 .
both results agree within the uncertainties , so we take as an overall calibration combining these results that log sfr = log @xmath2)$ ] - 7.0@xmath9 0.2 . in both cases ,
the quoted uncertainties show only the cosmic variance among different starbursts but do not reflect any systematic uncertainties in the underlying calibrations .
the agreement of the two independent methods implies , however , that such systematic calibration uncertainties are smaller than the variance among sources .
results for the ew([cii ] ) in figure 9 allow a transformation from sfr as measured from @xmath2)$ ] to sfr measured from @xmath7(158 ) .
this leads to a comparison with another independent calibration . using the median ew([cii ] ) of 1.0 for starbursts alone to mimimize any residual contamination by agn in the continuum , the relations among ew , @xmath2)$ ] , sfr , and @xmath7(158 ) yield log sfr = log @xmath7(158 ) - 42.8@xmath9 0.2 for sfr in m@xmath3 yr@xmath4 and @xmath7(158 ) in erg s@xmath4 ( to match units in figure 8) . for the luminosities of our sample shown in figure 8 , the resulting sfrs range over 0.8 @xmath8 log sfr @xmath8 2.7 .
a previous effort with the same objective but a completely independent calibration was by @xcite , who used 160 photometry with @xmath1 for comparing continuum @xmath7(160 ) to the sfr calibrated by h@xmath28 luminosity .
their sample consisted of 189 nearby star forming galaxies with generally low infrared luminosities which are not obscured lirgs or ulirgs so that sfrs could be calibrated using h@xmath28 luminosities without any corrections for extinction .
they caution that lower luminosity systems may have significant 160 luminosity from dust heated by stellar populations not associated with the current star formation , but they conclude that the @xmath7(160 ) is dominated by ongoing star formation for @xmath7(160 ) @xmath12 2 x 10@xmath29 erg s@xmath4 ( which encompasses all of our sources ) . their resulting calibration is log sfr = log @xmath7(160 ) - 42.85 .
this is nearly identical to our result even though derived with completely independent parameters .
the agreement is somewhat fortuitous because calzetti et al .
use different relations between ionizing photons and sfr compared to those of kennicutt ( 1998 ) which were used for our calibration of sfr using bolometric luminosities .
calzetti et al .
note that with the kennicutt calibrations , their result would be log sfr = log @xmath7(160 ) - 42.68 .
although this gives a sfr measure from the 160 continuum that is formally 1.3 times larger than our result , it is within our uncertainties .
the combined results of the two samples cover a factor of 100 in sfrs and a wide range of dust obscuration , implying a generalized validity to use of the 160 continuum as a measure of sfr so long as there is confidence that the dust continuum is dominated by stellar heating . a calibration of sfr from @xmath30(158 )
was also determined by @xcite for a sample of 25 ulirgs ( dominated by agn ) for which they observe [ cii ] and continuum fluxes .
their calibration of sfr arises from an earlier comparison @xcite between pah(6.2 + 11.3 ) luminosities and sfrs determined from [ neii ] and [ neiii ] luminosities using the results of @xcite . by comparing sfrs determined from these pah luminosities with the observed @xmath30(158 ) for the 25 ulirgs , they determine an empirical relation between sfr and continuum luminosity density that log sfr = [email protected] + ( [email protected])log @xmath30(158 ) , for luminosity density in l@xmath3 per hz .
the non linear luminosity dependence implies that a greater sfr is measured from continuum luminosity compared to pah luminosity as luminosity increases , which is consistent with having a residual contamination of @xmath30(158 ) from agn luminosity for increasing luminosities . within the uncertainties of this fit and the luminosity dependence , our results for sfrs are similar .
for example , at log @xmath7(158 ) = 45 in erg s@xmath4 , or log @xmath30(158 ) = -0.86 in l@xmath3 per hz , our calibration from pure starbursts yields log sfr = 2.2@xmath9 0.2 compared to the farrah et al .
result of 1.7 @xmath8 log sfr @xmath8 2.6 .
the uncertainty within our results that is introduced when using @xmath7(158 ) to measure sfr compared to using @xmath2)$ ] in sources of different classification can be estimated from the differences among ew([cii ] ) for different classifications in figure 9 . if some of the @xmath7(158 ) for agn and composites arises from dust heating by an agn , then ew([cii ] )
should be less for these sources compared to starbursts . adopting the median ew([cii ] )
= 1.0 as the measure of line to continuum ratio for pure starbursts leads to an estimate of the sfr error that would derive from use only of the continuum for agn or composite sources .
the error in sfr measure is inversely proportional to the ew ; for example , an ew([cii ] ) = 0.5 would mean a continuum twice as strong as from starbursts , so the continuum alone would lead to an overestimate of sfr by a factor of two compared to the sfr measured from [ cii ] . for agn , the median ew([cii ] )
= 0.7 , which implies a systematic overestimate of sfr by a factor of 1.4 if using only the far infrared continuum as the sfr indicator in agn . for agn
, 21/56 have ew([cii ] ) @xmath8 0.5 , leading to an overestimate of sfr from the continuum alone by more than a factor of two in @xmath6 40% .
although these results lead to smaller discrepancies in sfr than would be measured using only @xmath20 , confidence in measuring sfr within a factor of two from continuum luminosities alone requires confidence that the source is dominated by a starburst without continuum contamination by an agn .
new [ cii ] 158 observations of 18 sources with the @xmath0 pacs instrument are presented , and a summary of our total sample of 130 [ cii ] sources is given that covers a wide range of agn to starburst classifications as derived from pah strength .
new redshifts derived from [ cii ] and line to continuum strengths ( equivalent width of [ cii ] ) are given for the full sample .
results for 112 sources are compared with emission line fluxes from high resolution @xmath1 irs spectra .
a new calibration of [ cii ] as a sfr indicator is determined by comparing [ cii ] fluxes with mid - infrared [ neii ] and [ neiii ] emission line fluxes .
this independently gives the same result as determining sfr using bolometric luminosities of reradiating dust from starbursts : log sfr = log @xmath2)$ ] - [email protected] , for sfr in m@xmath3 yr@xmath4 and @xmath5)$ ] in l@xmath3 .
this confirms that [ cii ] measures the same starburst component of sources as measured with mid - infrared pah and neon emission line diagnostics .
the line to continuum ratio measured at 158 , ew([cii ] ) , changes little with luminosity or with classification , indicating that the far infrared continuum at 158 arises primarily from the starburst component of any source . for pure starbursts
, the continuum alone gives log sfr = log @xmath7(158 ) - [email protected] for sfr in m@xmath3 yr@xmath4 and @xmath7(158 ) in erg s@xmath4 .
the change of ew([cii ] ) with classification ( median ew([cii ] ) = 1.0 for starbursts compared to 0.7 for agn ) implies a systematic overestimate of sfr in agn by a median factor of 1.4 if using only the far infrared continuum at 158 as a sfr indicator .
we thank those who developed the @xmath0 space observatory for the opportunity to observe with open time .
pacs has been developed by a consortium of institutes led by mpe ( germany ) and including uvie ( austria ) ; ku leuven , csl , imec ( belgium ) ; cea , lam ( france ) ; mpia ( germany ) , inaf - ifsi / oaa / oap / oat , lens , sissa ( italy ) ; and iac ( spain ) .
this development was supported by the funding agencies bmvit ( austria ) , esa - prodex ( belgium ) , cea / cnes ( france ) , dlr ( germany ) , asi / inaf ( italy ) , and cicyt / mcyt ( spain ) .
partial support for this work at cornell university was provided by nasa through contracts issued by the nasa @xmath0 science center .
we thank dieter engels and the hamburger sternwarte for hospitality during the initial phase of this effort .
abel , n. p. , dudley , c. , fischer , j. , satyapal , s. , and van hoof , p. a. m. 2009 , , 701 , 1147 brauher , j. r. , dale , d. a. , and helou , g. 2008 , , 178 , 280 calzetti , d. , wu , s .- y . ,
hong , s. , et al .
2010 , , 714 , 1256 carilli , c. l. , riechers , d. , walter , f. et al .
2013 , , 763 , 120 de looze , i. , cormier , d. , lebouteiller , v. , et al .
2014 , , submitted , arxiv1402.4075v1 desai , v. , armus , l. , spoon , h. w. w. , et al .
2007 , , 669 , 810 diaz - santos , t. , armus , l. , charmandaris , v. , et al .
2013 , , 774 , 68 elbaz , d. , dickinson , m. , hwang , h. s. , et al .
2011 , , 533 , 119 farrah , d. , bernard - salas , j , spoon , h. w. w. , et al .
2007 , , 667 , 149 farrah , d. , lebouteiller , v. , spoon , h. w. w. , et al .
2013 , , 776 , 38 gracia - carpio , j. , sturm , e. , hailey - dunsheath , s. , et al .
2011 , , 728 , 7 genzel , r. , lutz , d. , sturm , e. , et al .
1998 , , 498 , 579 higdon , s. j. u. , devost , d. , higdon , j. l. , et al .
2004 , , 116 , 975 helou , g. , khan , i. r. , malek , l. , and boehmer , l. 1988 , , 68 , 151 .
helou , g. , malhotra , s. , hollenbach , d. j. , dale , d. a. , and contursi , a. 2001 , , 548 , l73 ho , l. and keto , e. 2007 , , 658 , 314 houck , j. r. , roellig , t.l . , van cleve , j. , et al .
2004 , , 154 , 18 kennicutt , r. c. 1998 , , 36 , 18 kennicutt , r. c. , calzetti , d. , aniano , g. , et al .
2011 , , 123 , 1347 laurent , o. , mirabel , r. f. , charmandaris , v. , et al .
2000 , , 359 , 887 lebouteiller , v. , barry , d. j. , bernard - salas , h. , sloan , g. c. , houck , j. r. , and weedman , d. w. 2010 , , 196 , 8 lebouteiller , v. , cormier , d. , madden , s. c. , et al .
2012 , , 548 , a91 lo faro , b. , franceschini , a. , vaccari , m. , et al .
2013 , , 762 , 108 luhman , m. l. , satyapal , s. , fischer , j. , et al .
2003 , , 594 , 758 malhotra , s. , helou , g. , stacey , g. , et al .
1997 , apj , 491 , l27 malhotra , s. , kaufman , m. j. , hollenbach , d. , et al .
2001 , , 561 , 766 meijerink , r. , spaans , m. , and israel , f. p. 2007
, , 461 , 793 melbourne , j. , soifer , b. t. , desai , v. , et al .
2012 , , 143 , 125 nikola , t. , genzel , r. , herrmann , f. , et al .
1998 , , 504 , 749 peeters , e. , spoon , h. w. w. , and tielens , a. g. g. m. 2004 , , 613 , 986 petric , a. o. , armus , l. , howell , j. , et al .
2011 , , 730 , 2011 pilbratt , g. l. , riedinger , j. r. , passvogel , t. , et al .
2010 , , 518 , l1 poglitsch , a. , waelkens , c. , geis , n. , et al .
2010 , , 518 , l2 sanders , d. b. , and mirabel , i. f. 1996 , , 34 , 749 sargsyan , l. , weedman , d. , lebouteiller , v. , houck , j. , barry , d. , hovhannisyan , a. , and mickaelian , a. 2011 , , 730 , 19 sargsyan , l. , lebouteiller , v. , weedman , d. , et al .
2012 , , 755 , 171 stacey , g. , geis , n. , genzel , r. , et al .
1991 , , 373 , 423 stacey , g. , hailey - dunsheath , s. , ferkinhoff , c. , et al .
2010 , , 724 , 957 stierwalt , s. , armus , l. , surace , j. a. , et al .
2013 , , 206 , 1 swinbank , a. m. , karim , a. , smail , i. , et al .
2012 , , 427 , 1066 .
tielens , a. g. g. m. , and hollenbach , d. 1985 , , 291 , 722 sturm , e. , lutz , d. , verma , a. , et al .
2002 , , 393 , 821 tommasin , s. , spinoglio , l. , malkan , m. , and fazio , g. 2010 , , 709 , 1257 veilleux , s. , rupke , d. s. n. , kim , d .-
, et al . 2009 , , 182 , 628 verma , a. , lutz , d. , sturm , e. , sternberg , a. , genzel , r. , and vacca , w. 2003 , , 403 , 829 voit , g. m. 1992 , , 399 , 495 wang , r. , wagg , j. , carilli , c.l . , et al .
2013 , , 773 , 44 wright , e. l. 2006 , , 118 , 1711 wu , y. , helou , g. , armus , l. , et al .
2010 , , 723 , 895 zhao , y. , lu , n. , xu , c. k. , et al .
2013 , , 765 , l13
|
a summary is presented for 130 galaxies observed with the @xmath0 pacs instrument to measure fluxes for the [ cii ] 158 emission line .
sources cover a wide range of active galactic nucleus to starburst classifications , as derived from polycyclic aromatic hydrocarbon ( pah ) strength measured with the @xmath1 infrared spectrograph .
redshifts from [ cii ] and line to continuum strengths ( equivalent width of [ cii ] ) are given for the full sample , which includes 18 new [ cii ] flux measures .
calibration of @xmath2)$ ] as a star formation rate ( sfr ) indicator is determined by comparing [ cii ] luminosities with mid - infrared [ neii ] and [ neiii ] emission line luminosities ; this gives the same result as determining sfr using bolometric luminosities of reradiating dust from starbursts : log sfr = log @xmath2)$ ] - 7.0 , for sfr in m@xmath3 yr@xmath4 and @xmath5)$ ] in l@xmath3 .
we conclude that @xmath5)$ ] can be used to measure sfr in any source to a precision of @xmath6 50% , even if total source luminosities are dominated by an agn component . the line to continuum ratio at 158 , ew([cii ] ) , is not significantly greater for starbursts ( median ew([cii ] ) = 1.0 ) compared to composites and agn ( median ew([cii ] ) = 0.7 ) , showing that the far infrared continuum at 158 scales with [ cii ] regardless of classification .
this indicates that the continuum at 158 also arises primarily from the starburst component within any source , giving log sfr = log @xmath7(158 ) - 42.8 for sfr in m@xmath3 yr@xmath4 and @xmath7(158 ) in erg s@xmath4 .
|
in this paper i will review the sites of explosive nucleosynthesis relevant for gamma - ray astronomy .
two main types of explosion are responsible for the emission of gamma - rays in the galaxy : supernovae , both thermonuclear and gravitational ( core - collapse ) and classical novae .
thermonuclear supernovae , or supernovae of type ia , are exploding white dwarfs in close binary systems , which do not leave a remnant after the explosion .
core collapse supernovae ( supernovae of type ii , ib / c ) are exploding massive stars ( m@xmath0 ) , which leave as remnant either a black hole or a neutron star .
typical velocities of supernovae ejecta are some @xmath1 km s@xmath2 , energies involved @xmath3 erg and ejected masses some m@xmath4 .
classical novae are the result of the explosion of the external h - rich accreted shells of a white dwarf in a binary system .
these explosions are recurrent phenomena ( contrary to supernovae ) , since an explosion is expected every time the critical accreted mass on top of the white dwarf is reached .
typical velocities of novae ejecta are between some @xmath5 and @xmath6 km s@xmath2 , energies involved @xmath7 erg and ejected masses between @xmath8 and @xmath9 m@xmath4 .
the radioactive isotopes synthesized during explosive nucleosynthesis , either in novae or in supernovae , are summarized in table 1 .
three types of decay chains can occur : electron captures ( @xmath10ni @xmath11 @xmath10co , @xmath12ni @xmath11 @xmath12co @xmath11 @xmath12fe , @xmath13ti @xmath11 @xmath13sc and @xmath14be @xmath11 @xmath14li ) , @xmath15 decays ( @xmath10co @xmath11 @xmath10fe , @xmath13sc @xmath11 @xmath13ca , @xmath16al @xmath11 @xmath16 mg and @xmath17na @xmath11 @xmath17ne ) and @xmath18 decays ( @xmath19fe @xmath11 @xmath19co @xmath11 @xmath19ni ) .
the first six isotopes in the table ( @xmath10ni , @xmath10co , @xmath12ni , @xmath13ti , @xmath16al and @xmath19fe ) are produced in supernova explosions ( although not exclusively , at least in the case of @xmath16al ) , whereas the last two are synthesized in classical novae . in sections 2 and 3 below ,
i will discuss how the synthesis of these radioactive nuclei proceeds in these scenarios . in the case of core - collapse supernovae
, it is important to distinguish the nucleosynthesis during the pre - explosive stage of the massive star evolution from that in the explosive phases ; it is crucial as well to know which part of the star will finally be ejected , since this quantity will determine the final enrichment of the galaxy in radioactive ( and other ) elements .
i wo nt discuss the synthesis of radioactive isotopes in other ( non - explosive ) sites , like the agb stars ( see the contribution by mowlavi , these proceedings ) or the wolf - rayet stars ( which can eject radioactive nuclei by strong stellar winds , during their hydrostatic evolution ; see arnould and meynet 1997 and meynet and arnould , 1999 for recent reviews ) .
two types of isotopes can be distinguished , depending on their lifetime ( see diehl and timmes , 1998 , for a recent review ) .
short - lived isotopes , such as @xmath10ni , @xmath12ni ( and their daughters @xmath10co and @xmath12co ) , @xmath13ti and @xmath19co , have lifetimes short enough ( see table 1 ) to make them detectable in individual objects . @xmath10ni and @xmath12ni are produced in all types of supernovae ; @xmath13ti is mainly produced in core - collapse supernovae , but it can also be synthesized in thermonuclear supernovae of the sub - chandrasekhar type ( provided that they exist ; see discussion of sneia types below ) .
@xmath19co is produced directly and from @xmath19fe decay , with @xmath19fe belonging to the long - lived isotopes group .
long - lived radioactive isotopes , such as @xmath16al and @xmath19fe , have lifetimes long enough to make them undetectable in individual sources , because the nuclei can be quite far away from their source and mixed with those coming from other explosions ( since the lifetime is longer than the typical period between two succesive explosions in the galaxy ) . for these isotopes , only the accumulated emission in the galaxy can be observed and used as diagnostic of models and of the galactic distribution of the sources .
the same classification scheme applies to isotopes synthesized in novae ; in this case , @xmath14be belongs to the short - lived group , whereas @xmath17na belongs to both of them ( see section 3 below )
. a very recent and interesting compilation of papers about astronomy with radioactivities can be found in diehl and hartmann ( 1999 ) .
gamma - ray astronomy provides an unique opportunity to detect radioactive isotopes in individual objects , giving a proof of ongoing nucleosynthesis in them . in the case of supernovae ,
another tool for the determination of the amount of radioactive nuclei in the ejecta is the bolometric light curve ( uvoir , from _ ultraviolet , optical _ and _ infrared _ ) .
in addition to the well known fact that @xmath10co ( daughter of @xmath10ni ) powers the early evolution of the light curve ( @xmath10ni mass can be determined from luminosity at maximum ) , @xmath12co is responsible for powering the light curve from around day 1000 after maximum .
later on @xmath13ti will provide a floor to the bolometric light curve and @xmath17na and @xmath19co could also play a minor role , depending on the supernova type and the specific yields of these radioactivities ( see , e.g. , woosley , pinto and hartmann , 1989 , timmes et al .
1996 , diehl and timmes 1998 ) .
these two types of observational approaches to the radioactive content in the ejecta had been possible for only one object so far : the supernova 1987a , which exploded 13 years ago in the lmc , only 55 kpc away from us .
this was a type ii supernova , which is not the most favorable case to look for gamma - ray emission , since its is much more opaque and has a smaller content of the most relevant radioactivites than snia ( see below ) ; however , its very short distance allowed for detection of its gamma - ray emission and also for a follow - up until very late times of its light curve ( through photometry in the uvbrijhk bands ) .
gamma - ray lines from @xmath10co decay , at 847 and 1238 kev , were detected in sn 1987a with the grs instrument of the smm satellite ( matz et al .
1988 ) and confirmed by several ballon - borne instruments ( i.e. , teegarden et al .
1989 , mahoney et al .
1988 , sandie et al .
1988 , cook et al .
1988 , rester et al .
one surprising fact was the appearence of these lines only 200 days after the explosion ( matz et al . , 1988 ) , much earlier than expected .
this has been interpreted as a sign of some early extra mixing of @xmath10co in order to transport this isotope into regions of low gamma - ray optical depth ( see , e.g. , pinto and woosley , 1988 , leising 1988 , bussard , burrows and the , 1989 , leising and share , 1990 ) .
the line profiles observed with gris ( teegarden et al .
1989 , tueller et al .
1990 ) have also put constraints on theoretical models of supernova explosions , since sphericity and homogeneity of the ejecta were incompatible with the observed fluxes and widths of the @xmath10co lines .
another crucial gamma - ray observation of short - lived isotopes in sn 1987a was the detection of gamma - ray radiation from @xmath12co decay ( between 50 and 136 kev ) , with osse on the cgro ( kurfess et al .
the deduced @xmath12co content ( for models with low gamma - ray optical depth , see kurfess et al .
1992 ) was such that the original ratio @xmath12ni/@xmath10ni produced in the explosion should be about 1.5 times the solar @xmath12fe/@xmath10fe ratio .
observations up to now show the change of slope related to the sequence of @xmath10co-@xmath12co decays ( see figure 3 in timmes et al .
future uvoir observations would possibly be able to show the light curve powering from @xmath13ti .
the spi instrument onboard integral has some possibilities to detect the gamma - ray emission from this @xmath13ti , which would provide a unique proof of the nucleosynthesis in core collapse supernovae and an important link between the uvoir and the gamma - ray observations .
coming back to @xmath12co it was first thought that the sn 1987a @xmath12co content deduced from osse observations was not enough to power the available bolometric light curve , since 5 times solar @xmath12fe/@xmath10fe ratio was required ( suntzeff et al .
1992 ) and alternative mechanisms to power the bolometric light curve were suggested ( clayton et al .
1992 ) . however , more recent observations seem to require a smaller amount of @xmath12co - decay to power the light curve , in agreement with the @xmath12fe/@xmath10fe ratio deduced from osse observations ( see figure 9 in diehl and timmes 1998 ) .
the excitement induced by the above mentioned gamma - ray observations ( together with many other observations at other wavelength ranges ) has led the theorists to suggest different possibilities for mixing both during the explosion and the ejection phases ( to quote only a few of the early works , see e.g. , arnett , fryxell and mller 1989 , benz and thielemann 1990 , fryxell , arnett and mller 1991 , herant and benz 1992 , and also the general reviews of sn 1987a from arnett et al .
1989 and mccray 1993 and references therein ) .
another detection ( tentative ) of gamma - ray emission from a supernova , with comptel on cgro , was that of sn 1991 t ( which was an overluminous snia ) , in ngc 4527 at around 17 mpc distance . in that case , a marginal detection of the 847 kev line was reported ( morris et al .
1995 , 1997 ) , leading to a prediction of a quite large @xmath10ni mass , implying that all the white dwarf mass should have been incinerated to @xmath10ni ( in contradiction with current theoretical models ) .
more recently , upper limits to the fluxes of the 847 and 1238 kev lines from @xmath12co - decay in the type ia supernova 1998bu , in ngc 3368 at around 8 mpc distance , have been deduced from comptel observations ( georgii et al . 2000 ) .
although no detection has been obtained , these limits are restrictive enough to constrain some of the available models of sneia nucleosynthesis ( like a sub - chandrasekhar mass model from nomoto et al .
1997 ) .
there is another important observation of gamma - ray lines related to short and medium - lived radioactivities in supernovae : the discovery of @xmath13ti emission at 1157 kev in the cas a supernova remnant ( iyudin et al .
1994 ; see reviews from diehl and timmes 1998 and kndlseder , these proceedings ) .
again the observations in gamma - rays are in some way puzzling , because the amount of @xmath13ti deduced from observations implies a @xmath10ni content ( according to theoretical models of supernovae nucleosynthesis ) which should have originated a very bright supernova , in contrast with the absence of historical records ( timmes et al .
1996 ) . observations in gamma - rays push forward the theoretical models , in order to balance all the available possibilities and to consider new ones .
a different kind of information is obtained from the observations of long - lived radioactivities ( @xmath16al and @xmath19fe ) . in this case , what is seen is not the ongoing nucleosynthesis in a particular object , but the integrated nucleosynthesis in the galaxy . up to now
, this has been possible for the 1809 kev @xmath16al emission . the @xmath16al map obtained with the comptel instrument onboard the compton gamma - ray observatory cgro ( diehl et al .
1995 , 1997 , oberlack et al .
1996 , kndlseder 1997 , 1999 , plschke et al .
these proceedings ) has posed interesting questions about the origin of the galactic @xmath16al ( see , e.g. , review from prantzos and diehl , 1996 ) .
it provides a direct and unique insight on the integrated nucleosynthesis during the last @xmath20 years .
some regions of enhanced emission have been discovered ( cygnus , carina , vela ) , indicating the presumable link between @xmath16al emission and massive star formation , as well as the relationship with spiral structure of the galaxy ( diehl et al .
1996 , kndleseder et al .
1996a , b , diehl et al . 1999 ,
kndlseder , these proceedings ) . for @xmath19fe ,
a similar map should be observed by integral , because the sources of this isotope are the same as those of @xmath16al , being the yields smaller by some factor ( timmes et al .
1995 , diehl et al . , 1997 ) .
it is worth mentioning that the integrated nucleosynthesis of @xmath13ti and @xmath17na may also be seen , if instruments are sensitive enough .
the future @xmath13ti and @xmath17na maps will provide a precious information about their sources , i.e. , supernovae ( mainly core collapse ones ) and novae , respectively .
the defining characteristic of sneia is the lack of hydrogen in their spectra , as well as the presence of a p cygni feature related to siii , @xmath21 , at maximum light ( wheeler and harkness 1990 ) ; in general , intermediate - mass elements ( o , mg , si , s , ar , ca ) appear in the spectrum near maximum light with high velocities ( 8000 - 30000 km s@xmath2 ) .
sneia are quite homogeneous from the observational point of view ( i.e , @xmath2290% of all sneia have similar spectra , light curves and peak absolute magnitudes ) , although some differences exist ( i.e. , subluminous explosions , like sn1991bg and sn1992k , and overluminous ones , like sn1991 t ) .
sneia appear in both elliptical and spiral galaxies and , therefore , their progenitors should be long - lived .
these facts all together suggest that the thermonuclear disruption of mass - accreting carbon - oxygen ( co ) white dwarfs is responsible for these explosions .
already in the sixties , hoyle and fowler ( 1960 ) suggested that thermonuclear burning in an electron - degenerate stellar core might be responsible for type i supernova ( there was no subclassification at the epoch ) explosions , with the explosion energy coming from the thermonucler burning of co into higher mass elements ( see also , e.g. , the pioneering works by arnett 1969 , hansen and wheeler 1969 ) .
it was also suggested at the epoch that the early supernova luminosity might have its origin on the radioactive decay of @xmath10ni ( colgate and mckee 1969 ) , which was already known to be a product of supernova nucleosynthesis , and that gamma - ray lines should be emitted from those explosions ( clayton , colgate and fishman , 1969 ) .
but the particular _ scenario _ where the explosion occurs ( see , e.g. , livio 1999 ) and the _ physics of the flame _ itself ( see , e.g. , hillebrandt and niemeyer 2000 ) are far from being understood . two types of progenitors have been suggested so far , concerning the mass of the exploding co white dwarf : chandrasekhar and sub - chandrasekhar mass models . in the chandrasekhar mass models , a co white dwarf explodes when reaching that mass , with central carbon ignition propagating outwards being responsible for the explosion .
the main problems related to this model are the uncertainties concerning burning propagation ( deflagration , detonation , delayed detonation , see below ) , but also the scenario is unclear . either a double degenerate ( merging of two co white dwarfs ) or a single degenerate scenario is possible . in all cases ,
the growth to the chandrasekhar mass is problematic , because both mass loss ( through nova episodes , for instance ) and accretion induced collapse ( if the initial mass is high enough and/or the white dwarf is made of oxygen and neon ) should be avoided ( see , e.g. , canal , isern and labay 1990 , 1992 , canal et al .
1990 , isern , canal and labay 1991 , nomoto and kondo 1991 , bravo and garca - senz 1999 ) .
in addition , for the double degenerate scenario there is a problem of statistics : there are not enough double white dwarf systems with sufficiently short period and total mass in excess of the chandrasekhar mass able to explode in less than the hubble time and to explain the galactic sneia rate .
in fact there was nt any observed system fulfilling these conditions until the very recent discovery of kpd 1939 + 2752 ( maxted , marsh and north , 2000 ) , which is the first snia progenitor candidate observed . in the sub - chandrasekhar mass models , a co white dwarf of low - mass ( 0.6 - 0.8 m@xmath4 ) accretes helium ( @xmath23 m@xmath4 ) , reaching a final mass smaller than the chandrasekhar mass .
provided the accretion rate is moderate ( around @xmath24m@xmath25 ) , there is he ignition on the top of the co core .
this ignition causes an outward propagating he - detonation wave ( basically transforming he into ni at high velocity ) and an inward propagating pressure wave .
the last one finally provoques a carbon ignition ( central or off - center ) , which leads to an outward carbon - detonation incinerating all the white dwarf , and synthesizing intermediate - mass elements , in addition to ni ( see , e.g. , livne 1990 , livne and glasner 1991 , woosley and weaver 1994 ) .
therefore , in this model ( called `` indirect double detonation '' , idd , or `` edge lit detonation '' , eld ) there is an outer layer of high - velocity ni and he above the intermediate - mass elements , which does not exist in the chandrasekhar - mass models .
sub - chandrasekar mass models are not considered as good sneia progenitors nowadays , because of both observational and theoretical problems ; observational : the high velocity ni above intermediate mass elements is not seen in the spectra ; theoretical : the he - driven carbon detonation is very model dependent ( see for instance the 3d models from garca - senz , bravo and woosley 1999 ) .
but it is still a possibility that sub - chandrasekar mass models explain some subluminous sneia , like sn1991bg ( see , e.g. , ruiz - lapuente , canal and burkert 1997 ) . in summary ,
the bulk of normal sneia are assumed to be exploding chandrasekhar - mass co white dwarfs , but there is still room for the sub - chandrasekhar mass models to explain some peculiar objects . therefore , whether sneia come from single or double - degenerate scenarios and whether they come from carbon or helium plus carbon ignition are not closed issues ( see , e.g. , the recent paper from branch 2000 ) .
the main problems still remaining on the modeling of sneia affect the ignition process and the flame propagation .
different possibilities exist : deflagration ( subsonic flame speed ) , detonation ( supersonic ) and a combination of both ( delayed detonation ) .
a detonation with densities larger than @xmath26 g @xmath27 is not a viable mechanism , since all the star would be incinerated to ni , without synthesis of intermediate - mass elements . on the contrary , if the density is lower , intermediate - mass elements are synthesized , in agreement with the observations . concerning deflagrations , they produce nucleosynthesis at velocities in general agreement with the observed spectra , but some neutronized isotopes ( such as @xmath28fe , @xmath28cr and @xmath29ni ) are overproduced in amounts incompatible with the chemical evolution of the galaxy . to overcome this problem , delayed detonations
were suggested ( khokhlov 1991a ) .
there are two situations in which a deflagration to detonation transition ( ddt ) could occur in supernovae ( see , e.g. , khokhlov , oran and wheeler , 1997 ) : ddt could occur directly or as a result of a previous expansion . for instance , in the pulsation delayed detonation ,
a first slow deflagration is quenched because of the expansion of the white dwarf , which subsequently pulses and recontracts , causing a detonation upon recollapse ( khokhlov 1991b ) .
the propagation of the detonation wave through the pre - expanded star produces the required intermediate mass elements in the outer layers at densities lower than @xmath26 g @xmath27 ( which are not synthesized in detonations at larger densities ) . in these models , the problem of overproduction of highly - neutronized nuclei is alleviated but not solved ( khokhlov 1991a , b ) .
models of delayed detonations in 2d , both of the first deflagration phase and of the subsequent detonation phase , have been performed by arnett and livne ( 1994a , b ) ; they show that the first slow deflagration is insufficient to unbind the star , that a pulsation of large amplitude is generated and that reignition occurs after the first contraction phase . in summary
, there is a general consensus about the fact that , in order to explain spectroscopic observations , burning should proceed subsonically ( deflagration ) in the inner core ( where densities are large , i.e. , @xmath30 g @xmath27 ) , whereas burning becomes supersonic ( detonation ) in the outer lower density zones ( see examples of models in bravo et al .
1993 , hflich and khokhlov 1996 , bravo et al .
1996 , woosley 1997 ) .
but the way in which the deflagration to detonation transition ( ddt ) occurs is not yet clear , ( see , e.g. , discussions in recent papers by niemeyer and woosley 1997 , niemeyer 1999 , lisewski , hillebrandt and woosley 2000 , and in the review by hillebrandt and niemeyer 2000 ) .
there is also ample debate about the way in which the initial burning occurs : flame instabilities , flame - turbulence interactions ( see review about turbulence and thermonuclear burning by hillebrandt and niemeyer 1997 , and references therein ) .
all 1d models ( which were the unique ones available up to the nineties and still are the only ones to include _ complete nucleosynthesis _ ) rely on prescriptions based on some parametrization of the flame speed and , in the case of delayed detonations , of the deflagration - detonation transition -ddt- densities .
different groups work in models of thermonuclear snia and their nucleosynthesis , including the radioactivities .
it is out of the scope of this paper to mention even a small fraction of them , but a small sample can be useful to show the main results and the main caveats still remaining ( see the recent books _ thermonuclear supernovae _ , edited by ruiz - lapuente , canal and isern , 1997 , and _ type ia supernovae : theory and cosmology _ , edited by niemeyer and truran , 2000 ) .
nomoto and coworkers have computed detailed nucleosynthesis in carbon deflagration supernovae ( nomoto , thielemann and yokoi , 1984 , thielemann , nomoto and yokoi , 1986 ) , and also in other types of explosive carbon burning ( such as delayed detonations , with parametrized ignition densities and deflagration - detonation transition -ddt- densities , see iwamoto et al .
the yields of radioactive isotopes are mainly affected by the ddt density ( i.e. , synthesized mass of @xmath10ni ranges from 0.55 to 0.77 m@xmath4 , and @xmath12ni from 9.6x@xmath8 to 1.98x@xmath31 m@xmath4 , in iwamoto et al.s models ) .
these yields are larger than those from core collapse supernovae ( see below ) and distributed in less opaque zones , since there is nt much mass above them .
this makes type ia supernovae better targets for integral than sneii ( but see section 2.1 for observational results ) . in the context of gamma - ray astronomy , it is important to stress that sub - chandrasekhar mass models synthesize larger amounts of @xmath13ti than chandrasekhar mass ones ( see , e.g. , woosley and weaver , 1994 ) .
gamma - ray spectra of sneia for the different models provide important signatures of the explosion mechanism , although unfortunately there is nt much observational data to compare with ( see previous section ) .
prospects for sneia explosion mechanism identification with gamma - rays have been analyzed recently by gmez - gomar et al .
( 1998a ) , with a special emphasis on detectability with the instruments that will be onboard integral ( see also burrows and the 1990 , hflich , khokhlov and mller 1994 , kumagai and nomoto 1997 , hflich , wheeler and khokhlov 1998 ) .
lines from @xmath10ni - decay ( 158 , 750 , 812 kev ) are prominent during the first days after the explosion , but they disappear very fast , because of the short @xmath10ni - lifetime . lines from @xmath10co ( 847 , 1238 kev ) and @xmath12co ( 122 , 136 kev ) appear later and have longer durations .
the most intense lines are those at 847 , 1238 , 812 and 158 kev , in addition to the annihilation line at 511 kev .
the strongest line is always the 847 kev one ( detectable up to 11 - 16 mpc with spi on integral ) , whereas the 158 kev line ( from @xmath10ni - decay ) is the most interesting to discriminate between models .
the 158 kev line is narrower and , therefore , detectable at longer distances with spi , than another @xmath10ni - line ( at 812 kev ) , despite being fainter .
it is almost undetectable in pure deflagration models , whereas it is even stronger than the 1238 line in detonation models .
another interesting signature of the models is the ratio between the 847 kev and the 158 kev line fluxes ( 200 days after maximum and at maximum , respectively ) , because it provides information about the ratio between total @xmath10ni in the ejecta and @xmath10ni in the external layers : the late emission at 847 kev comes from @xmath10co - decay ( coming from @xmath10ni - decay ) , while only the @xmath10ni present in the outermost shells is responsible for the 158 kev line flux ( see gmez - gomar et al .
1998a for details ) .
finally , line profiles will also provide important information allowing for discrimination between the models , for explosions at distances short enough ( see again gmez - gomar et al .
1998a ) .
all supernova types except type ia s ( i.e. , type ii , ib / c ) are explained by the explosion of massive stars .
stars with initial masses ( m@xmath32 m@xmath4 ) do nt end their lives as white dwarfs .
succesive phases of thermonuclear burning ( c , ne , o , si ) give as a result a star with an `` onion - skin '' structure , where a central iron core is surrounded by shells made of elements of progressively lower atomic mass .
the chemical composition along the star is the following ( see , e.g. , figure 10.8 , corresponding to a 25 m@xmath4 star , in arnett 1996 ) : fe core , _ si - burning _ zone ( made mainly of elements from si to ni , without o ) , _ o - burning _ zone ( o , si - ca ) , _ ne - burning zone _ ( ne , mg and o , no c ) , _ c - burning _ zone ( c and o , ne and mg ) , radiative he - burning zone ( he , c and o ) , convective he zone , inert part of old he core ( interior to h - burning shell ) , material above the h - burning shell ( plus some inert zones associated with the si , o , ne and c - burning zones and just outside them ) .
once the fe - core reaches the chandrasekhar mass , it becomes unstable and collapses to form a neutron star .
the gravitational energy released ( @xmath33 erg ) during core collapse is responsible for the ensuing supernova explosion , but it is not yet completely understood how the conversion of this potential energy into kinetic energy proceeds ( only 0.1% of the available potential energy is needed ) .
baade and zwicky ( 1934 ) were the first to suggest that the gravitational energy released during the formation of a neutron star could produce a supernova explosion .
colgate and white ( 1966 ) built a supernova model , considering that the transfer of energy takes place by the emission and deposition of neutrinos ; wilson ( 1971 ) showed that the electron capture neutrino burst was not strong enough to eject material .
the weinberg - salam model of electroweak interactions opened new possibilities of neutrino interactions with matter ( neutral currents ) . in 1974 , freedman noticed the importance of neutral currents in the physics of core collapse supernovae ; as a result of the increased cross section of core material to neutrinos , these particles are trapped during the collapse .
it was shown by bethe et al . in 1979
that one of the consequences of neutrino trapping is that the entropy of the core changes little during collapse ( it remains low ) , leaving the collapse continue up to nuclear densities .
further compression is prevented by the repulsive component of the strong interaction ( stiffness of nuclear matter ) , leading to the core bounce .
a shock wave is generated at its boundary and propagates outwards .
but it has been shown that the energy of this shock is mainly invested in the photodisintegration of heavy nuclei and in neutrino losses ; therefore , the shock stalls and the explosion via the so called `` prompt mechanism '' is unsuccessful . in the `` delayed mechanism '' , there is a revival of the stalled shock because of neutrino heating behind the shock ( bethe and wilson 1985 ) .
however , the explosion energy does not reach easily the necessary @xmath3 erg .
further works introduced the effect of convective instabilities , caused by a negative entropy gradient , in order to deliver energy to the shock ( see , e.g. , bethe , 1990 , herant , benz and colgate 1992 , herant et al .
1994 , bethe 1995 , janka and mller , 1995 , burrows , hayes and fryxell , 1995 , to quote only a few of the papers dealing with this topic ) .
convection aids the explosion because it increases the efficiency at which neutrino energy is deposited ( material that rises cools and converts energy from neutrino deposition into kinetic energy , instead of re - radiating it as neutrinos ) and also reduces the energy required to launch the explosion ( by reducing the pressure at the accretion shock ) ( see recent reviews by fryer , 2000 , burrows 2000 , and references therein ) .
the handling of this process is very model dependent : treatement of neutrino transport , multidimensional aspects .
also the structure of the stellar core before collapse ( i.e. , the presupernova model ) are important for the final outcome of the explosion .
fortunately , nucleosynthesis during core collapse supernova explosions can be computed without a complete knowledge of the explosion mechanism itself . as in the case of thermonuclear supernovae
, all the details of the physics involved in the explosion are not required to have an approximate , but quite good when compared with the observations , idea of which are the main nucleosynthetic yields of core collapse supernovae .
two steps are needed to compute snii ( and ib / c ) yields : nucleosynthesis during the massive star evolution ( i.e. , pre - supernova phase ) and explosive burning when a shock wave crosses the mantle surrounding the collapsing core .
there are different ways to simulate the explosion artificially .
one is by means of a `` thermal bomb '' , i.e. , injecting thermal energy inside the fe core , in a way such that the ejecta attains the desired kinetic energy , @xmath34erg ( see , e.g. , thielemann , nomoto and hashimoto , 1996 ) .
another alternative is the injection of momentum , through a piston , inward - moving during the infall previous to the explosion , and outward - moving during the explosion , with a velocity such that the desired kinetic energy of the ejecta is obtained ( see , e.g. , woosley and weaver , 1995 ) .
the _ mass cut _ between the collapsing core and the ejecta determines the amount of mass ejected ( and that of @xmath10ni and other radioactive isotopes , in particular ) . in the `` thermal bomb ''
method , they adjust it taking into account the relationship between supernova progenitor masses and @xmath10ni masses ejected deduced from some observations . in the piston approach
, the mass cut is obtained from the choice of piston position and energy ; a mass cut located outside the piston is often obtained ( for a discussion of the differences between both models , including an analysis of the influence of the nuclear reaction rates , see hoffman et al . , 1999 ) . in summary ,
both groups have performed calculations of detailed nucleosynthesis by inducing the core - collapse supernova explosion on massive stars ( previously evolved following all the nucleosynthesis phases ) .
other groups have performed studies of massive star evolution , but there is no room in this short review to mention all of them .
the masses studied by thielemann et al .
( 1996 ) range between 13 and 25 m@xmath4 , with initial metallicities , z , equal to solar ( see nakamura et al .
1999 for the effect of low z ) .
woosley and weaver ( 1995 ) studied the range 11 - 40 m@xmath4 , for z=0 and z between @xmath35 and z@xmath4 .
si , o , ne and c explosive burning occur when the shock wave crosses the corresponding zones in the pre - supernova ( see above for the description of its structure ) .
a brief description of the results concerning the synthesis of radioactive isotopes follows .
@xmath10ni and @xmath12ni are produced when either oxygen or silicon - rich layers with low neutron excess ( y@xmath36 ) are heated to temperatures above 4x@xmath37 k ( explosive o- and si - burning ) .
they are produced whether the material ejected is alpha - rich or not , although @xmath12ni synthesis is favored in alpha - rich freeze - out ; this happens when material , initially in nuclear statistical equilibrium ( nse ) at relatively low density , is cooled so rapidly that the free alpha particles do not have time to merge via the 3@xmath38 reaction and , therefore , matter cools down in the presence of a large concentration of @xmath38-particles , which modify the final composition ( with respect to the normal freeze - out ) .
@xmath13ti is also produced during @xmath38-rich freeze - out from nse in the hottest and deepest layers ejected during the explosion .
therefore , the yields of these radioactive isotopes are very sensitive to the mass - cut location ( woosley and hoffman 1991 , hoffman et al .
1995 , woosley and weaver 1995 , timmes et al .
for example , stars with masses larger than 30 m@xmath4 do nt eject any @xmath10ni ( nor @xmath12ni and @xmath13ti ) if the kinetic energy ( at infinity ) is around 1.2x@xmath3 erg .
if this energy is enhanced , the mass - cut is lowered and some @xmath10ni ( and @xmath12ni and @xmath13ti ) are ejected . ejected masses of @xmath10ni are around 0.1 m@xmath4 and those of @xmath13ti between @xmath39 and @xmath35 m@xmath4 .
similar results are obtained by thielemann et al .
( 1996 ) , except for the larger amounts of @xmath13ti , probably because of the different way of simulating the explosion , which possibly injects larger entropy in the inner shells and favors a larger ejected mass and an enhanced @xmath38-rich freeze - out ( see , e.g. , aufderheide , baron and thielemann 1991 and hoffman et al .
1999 ) .
@xmath16al is another important radioactive isotope which is produced in core collapse supernovae ( and in other scenarios ) through the @xmath40mg(p,@xmath41 ) reaction .
@xmath16al yields depend on pre - supernova evolution ( h- and o - ne burning shells ) and on the explosion .
two factors enhance @xmath16al production during the explosion : explosive burning in o - ne shells and @xmath42-spallation reactions on @xmath43ne , @xmath44o , @xmath45na , @xmath46 mg , which liberate protons that are captured by @xmath40 mg .
it is important to stress that another important long - lived radioisotope , @xmath19fe , is coproduced with @xmath16al in the same regions within snii ( this isotope is synthesized by neutron captures on @xmath47fe in the o - ne burning shell and in the base of the he - burning shell , both pre - explosively and explosively ) .
therefore , these nuclei should have similar spatial distributions in the ejecta ( timmes et al .
the @xmath16al/@xmath19fe ratio depends on the mass of the pre supernova : @xmath16al/@xmath19fe is larger than 1 for m larger than 25 m@xmath4 and similar to 1 for smaller masses .
the typical yields of @xmath16al are @xmath35 m@xmath4 and those of @xmath19fe 4x@xmath9 m@xmath4 .
the final yields depend on three aspects : presupernova evolution , explosion energy and details of the explosion mechanism ( see diehl and timmes , 1998 and thielemann 1999 for recent analyses ) . the main issues concerning presupernova evolution are those affecting general stellar evolution of low - mass stars , plus some specific ones relative to massive stars . for instance , the treatement of convection affects nucleosynthesis ; in particular , convective burning in the o - shell of massive stars . models with m=20 m@xmath4 have deserved a particular attention for the theorists , since they are crucial to understand the mixing of radioactive ( and other ) isotopes , like @xmath10ni , in supernova ejecta , which has been deduced from the observations of sn1987a ( see section 2.1 above ) .
2d models of o - burning ( bazan and arnett 1994 ) obtain significant mixing beyond the boundaries defined by mixing - length convection .
what they obtain are perturbations in density in the oxygen shell that are sufficiently large to `` seed '' hydrodynamic instabilities , which will mix the `` onion - skin '' composition of the presupernova ( bazan and arnett 1998 ) .
this occurs in precisely the region in which @xmath10ni is explosively produced by oxygen burning behind the explosion shock .
this result poses some problems to the models of explosive nucleosynthesis based on 1d presupernova evolution .
rotation can also have some effect ( see , e.g. , works by meynet and maeder 1997 , heger , langer and woosley 2000 ) , as well as mass - loss during the presupernova evolution , in the final yields of radioactive elements . concerning the energy of the explosion and the details of the explosion mechanism , one of the main problems is the location of the mass - cut , which determines how much mass falls back into the collapsing core ( and therefore whether it will be a neutron star or a black hole ) and how much mass is ejected and with wich composition ( the profile of some isotopes is steep around the mass - cut location and , therefore , the yield is affected by it ) .
therefore , the mass - cut determines crucially the final yields of radioactive elements , specially for those produced in the inner regions of the supernova ( @xmath10ni , @xmath12ni and @xmath13ti ) .
as mentioned above , the explosion energy , and the corresponding entropy in the inner shells , crucially affect the degree of @xmath38-rich freeze - out and , therefore , the yields of the fe - group nuclei and of @xmath13ti .
classical novae explosions are the most common explosions in the galaxy .
the cause of the explosion is a thermonuclear runaway ( tnr ) on top of a white dwarf , ensuing the degenerate burning of the accreted hydrogen ( starrfield 1989 , hernanz & jos 2000 ) .
the synthesis of radioactive isotopes in classical novae is important for two reasons .
first , some of the isotopes produced are crucial for the explosion mechanism itself ( i.e. , @xmath48o , @xmath49o , @xmath50f with lifetimes 102 , 176 and 93s , respectively ) .
the reason is that these isotopes are transported by convection to the outer layers of the envelope , during the runaway , where they subsequently decay ( @xmath51 ) and cause the expansion of the envelope and the increase in visual luminosity .
second , the decay of the unstable nuclei originates gamma - ray emission , because of either direct emission of gamma - ray photons or positrons ( for @xmath52-unstable nuclei ) , which annihilate with electrons .
the photons emitted ( 511 kev , positronium continuum , 478 and 1275 kev , see table 1 ) experience comptonization in the nova expanding envelope . therefore , the emission from novae consists of lines plus a continuum ( see gmez - gomar et al .
1998b , hernanz et al
. 1999 , hernanz et al .
these proceedings and references therein ) .
the potential role of classical novae as gamma - ray emitters had been alredy pointed out many years ago ( clayton and hoyle 1974 , clayton 1981 , leising and clayton 1987 ) .
the first available hydrodynamic models of nova explosions ( starrfield et al . 1978 and prialnik et al . 1978 ) realized that there was a need of an initial enrichment in cno isotopes both to power the explosion and to explain some observed abundances .
two and three dimensional simulations of the thermonuclear runaway of a co white dwarf , valid when the accreted envelope has been already built up , are the only available up to now ( glasner et al . , 1997 , kercek et al
1998 , 1999 ) .
they predict that enrichment proceeds too slowly if the accreted gas has nearly solar cno abundances at the onset of the thermonuclear runaway , and conclude that fast nova outbursts require huge enrichments of c and o. the mechanism which leads to such enhancements must operate prior to the outburst and has not been modeled up to now .
therefore , it is known that some mixing with core material ( either co or one ) during the accretion phase prior to the runaway should occur , but this process has not been modeled yet in a self - consistent way .
another approach to the problem of initial enrichment comes from the multicycle 1d models ( of co novae only , up to now ) , from prialnik and kovetz ( prialnik and kovetz 1995 , kovetz and prialnik , 1997 ) ; diffusion is responsible for the enrichment , which becomes larger after a number of flashes .
however , the large metallicities and neon abundances observed in some novae are not well reproduced .
another approach is based on 1d models with an initial ( parametrized ) enrichment , such that the general properties of observed novae ( mainly abundances ) are well modeled ( see for instance starrfield et al .
1998 , jos and hernanz 1998 ) .
classical novae synthesize many radioactive isotopes , which vary depending on the nova type ( which in turn depends on the type -co or one- of the underlying white dwarf ) .
co novae produce mainly @xmath14be , whereas one produce @xmath17na and @xmath16al .
other radioactivities with shorter lifetimes , such as @xmath53n and @xmath54f ( @xmath55s and 158min , respectively ) are produced in similar amounts in co and one novae ( see jos and hernanz , 1998 , jos , coc and hernanz , 1999 , hernanz et al . 1999 and hernanz et al .
, these proceedings for details ) .
typically , @xmath56m@xmath4 of @xmath53n , @xmath57m@xmath4 of @xmath54f are ejected in both co and one explosions .
co novae also eject @xmath58m@xmath4 of @xmath14be , and one novae @xmath57m@xmath4 of @xmath17na and @xmath24m@xmath4 of @xmath16al .
the reason of the different explosive nucleosynthesis results in co and one nova types is that some seed nuclei ( such as @xmath43ne , @xmath17ne , @xmath59 mg ) are necessary to synthesize @xmath17na and @xmath16al . that s because temperatures attained at the peak of the nova outburst are not high enough to break the cno cycle towards nena - mgal cycles . the two short - lived isotopes @xmath53n and @xmath54f are crucial for the prompt gamma - ray emission of novae , which is the most intense emission ( @xmath8 phot @xmath60 s@xmath2 ) , but of very short duration ( a few days ) and appearing before optical detection . the medium - lived isotopes ( @xmath14be and @xmath17na )
produce fluxes of around @xmath61 and @xmath9 phot @xmath60 s@xmath2 , for distances of 1 kpc .
the prospects for detectability with the spi instrument onboard integral are analyzed in hernanz et al .
( these proceedings , and references therein ) .
it is important to remind that , in addition to the @xmath14be and @xmath17na emission from individual novae , the cumulative emission from all the galactic novae can give important information about the distribution of the sources , specially if there is only one dominant source for that particular isotope . for @xmath17na ,
novae are the main individual contributors .
therefore , the detection of galactic @xmath17na emission , and the corresponding 1275 kev emission map ( see jean et al .
, 1999 , 2000 ) , would be a very valuable tool to study the distribution of novae in the galaxy , which is very poorly known from optical - uv and ir observations because of interstellar extinction .
finally , @xmath16al is produced in one novae in such an amount that makes it quite improbable that novae contribute largely to the @xmath16al content of the galaxy , as observed through its emission at 1809 kev .
research partially supported by the cicyt - p.n.i.e .
( esp98 - 1348 ) , by the dges ( pb98 - 1183-c03 - 02 and pb98 - 1183-c03 - 03 ) and by the aihf1999 - 0140 aufderheide m.b . , baron e. , thielemann f .- k . , 1991 ,
apj 370 , 630 arnett d. , 1969 , astrophys .
space sci . 5 , 180 arnett d. , 1996 , supernovae and nucleosynthesis , princeton university press arnett d. , livne e. , 1994a , apj 427 , 315 arnett d. , livne e. , 1994b , apj 427 , 330 arnett d. , bahcall j.n . , kirshner r.p . ,
woosley s.e . , 1989 ,
27 , 629 arnett d. , fryxell b. , mller e. , 1989 , apj 341 , l63 arnett d. , truran j.w . ,
woosley s.e . , 1971 , apj 165 , 87 arnould m. , meynet g. , 1997 , in 2@xmath62 integral workshop `` the transparent universe '' .
esa sp-382 , noordwijk , p. 33
baade w. , zwicky f. , 1934 , phys .
45 , 138 bazan g. , arnett d. , 1994 , apj 433 , l41 bazan g. , arnett d. , 1998 , apj 496 , 316 benz w. , thielemann f .- k . , 1990 , apj 348 , l17 bethe h.a . , 1990 , rev .
62 , 801 bethe h.a . , 1995 ,
apj 449 , 714 bethe h.a . , wilson j.r . , 1985 ,
apj 295 , 14 bethe h.a .
, brown g.e . , applegate j. , lattimer j. , 1979 , nucl .
a324 , 487 branch d. , 2000 , pasp , in press ( astro - ph/0009337 ) bravo e. , garca - senz d. , 1999 , mnras 307 , 984 bravo e. , tornamb a.,domnguez i. , isern j. , 1996 , a&a 306 , 811 bravo e. , domnguez i. , isern j. , canal r. , hflich p. , labay j. , 1993 , a&a 269 , 187 burrows a. , 2000 , nature 403 , 727 burrows a. , the l.s . , 1990 ,
apj 360 , 626 burrows a. , hayes j. , fryxell b.a . , 1995 ,
apj 450 , 830 bussard r.w . , burrows a. , the l.s . , 1989 , apj 341 , 401 canal r. , isern j. , labay j. , 1990 , ann .
astrophys . , 28 , 183 canal r. , isern j. , labay j. , 1992 , apj 398 , l49 canal r. , garca d. , isern j. , labay j. , 1990 , apj 356 , l51 clayton d.d .
, 1981 , apj 244 , l97 clayton d.d . , hoyle f. , 1974 , apj 187 , l101 clayton d.d . , colgate s.a . ,
fishman g.j . , 1969 , apj 155 , 75 clayton d.d . , leising m.d . , the l.s : , johnson w.n . ,
kurfess , j.d . , 1992 ,
apj 399 , l141 colgate s.a .
, mckee c. , 1969 , apj 157 , 623 colgate s.a .
, white r.h . , 1966 , apj 143 , 626 cook w. r. , palmer d. m. , prince t. a. , schindler s. m. , starr c. h. , stone e.c .
, 1988 , apj 334 , l87 diehl r. , hartmann d. , eds . , 1999 ,
astronomy with radioactivities , mpe report 274 diehl r. , timmes f.x .
, 1998 , pasp 110 , 637 diehl et al .
1996 , a&a suppl .
120 , 321 diehl et al . 1995 , a&a 298 , 445 diehl et al .
1997 , 4th compton symp . ,
diehl et al .
1999 , astroph .
lett . & commun .
, 38 , 357 freedman d.z .
, 1974 , phys .
d 9 , 1389 fryer c.l .
, in in cosmic explosions , eds .
s.s.holt and w.w .
zhang , aip conf .
522 ( new york ) , p. 113
fryxell b. , arnett d. , mller e. , 1991 , apj 367 , 619 garca - senz d. , bravo e. , woosley s.e . , 1999 ,
a&a 349 , 177 georgii r. , et al . 2000 , in 5@xmath63compton symposium , eds .
mcconnell m.l . &
ryan j.m . , aip 510 , new york , p. 49 glasner s.a .
, livne e. , truran j.w . , 1997 , apj 475 , 754 gmez - gomar j. , isern j. , jean p. , 1998a
, mnras 295 , 1 gmez - gomar j. , hernanz m. , jos j. , isern j. , 1998b , mnras 296 , 913 hansen c.j , wheeler j.c . , 1969 ,
space sci . 3 , 464 heger a. , langer n. , woosley s.e .
, 2000 , apj 528 , 368 herant m. , benz w. , 1992 , apj 387 , 294 herant m. , benz w. , colgate s. , 1992 , apj 395 , 642 herant m. , benz w. , hix w.r .
, fryer c.l . ,
colgate s.a . , 1992 , apj 435 , 339 hernanz m. , jos j. , coc a. , gmez
gomar j. , isern j. , 1999 , apj 526 , l97 hernanz m. , jos j. , 2000 , in cosmic explosions , eds .
s.s . holt and w.w .
zhang , aip conf .
522 ( new york ) , p.339 hillebrandt w. , niemeyer j.c , 1997 , in thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) , p. 337 hillebrandt w. , niemeyer j.c , 2000 , ann .
, in press ( astro - ph/0006305 ) hflich p. , khokhlov a.m. , 1996 , apj 457 , 500 hflich p. , khokhlov a.m. , mller e. , 1994 , apjs 92 , 501 hflich p. , wheeler j.c . , khokhlov a.m. , 1998 , apj 492 , 228 hoffman r.d . ,
woosley s.e .
, weaver t.a .
, rauscher t. , thielemann , f .- k .
, 1999 , apj 521 , 735 hoffman r.d . , woosley s.e . ,
weaver t.a .
, timmes f.x . , eastman r. , hartmann d.h . , 1995 , in the gamma - ray sky with compton gro and sigma , eds .
m. signore , p. salati and g. vedrenne , kluwer ( dordretch ) , p. 267
hoyle , fowler , 1960 apj 132 , 565 isern j. , canal r. , labay j. , 1991 , apj 372 , l83 iwamoto k. , brachwitz f , nomoto k. , kishimoto n. , umeda h. , hix w.r . ,
thielemann , f .- k . , 1999 ,
apj 125 , 439 iyudin a. , et al . , 1994 ,
a&a 284 , l1 jean p. , hernanz m. , gmez - gomar j. , jos j. , 1999 , astrophys .
lett . & comm . , 38 , 433 jean p. , hernanz m. , gmez - gomar j. , jos j. , 2000 , mnras , in press janka h .-
, mller e. , 1995 , apj 448 , l109 jos j. , hernanz m. , 1998 , apj 494 , 680 jos j. , coc a. , hernanz m. , 1999 , apj 520 , 347 kercek a. , hillebrandt w. , truran j.w . , 1998 , a&a 337 , 339 kercek a. , hillebrandt w. , truran j.w . , 1999 , a&a 345 , 831 khokhlov a.m. , 1991a , a&a 245 , 114 khokhlov a.m. , 1991b , a&a 245 , l25 khokhlov a.m. , oran e.s .
, wheeler j.c . , 1997 , apj 478 , 678 kndlseder j. et al .
, 1996a , a&a suppl .
120 , 327 kndlseder j. et al . , 1996b , a&a suppl .
120 , 335 kndlseder j. 1997 , phd thesis kndlseder j. 1999 , a&a 345 , 813 kovetz a. , prialnik d. , 1997 , apj 477 , 356 kumagai s. , nomoto k. , 1997 , in thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) , p. 515
kurfess j.d . , et al . , 1992 ,
apj 399 , l140 leising m.d . , 1988 ,
nature 332 , 516 leising m.d . ,
clayton d.d . , 1987 ,
apj 323 , 157 leising m.d .
, share g.h . , 1990 ,
apj 357 , 638 lisewski a.m. , hillebrandt w. , woosley s.e .
, 2000 , apj 538 , 831 livio m. , 1999 , in type ia supernovae : theory and cosmology , cambride univ . press ( astro - ph/9903264 ) livne e. , 1990 , apj 354 , l53 livne e. , glasner a.s . , 1991 ,
apj 370 , 272 mahoney w.a . , varnell l.s .
, jacobson a.s . , ling j.c .
, radocinski r.g . ,
wheaton w.a . , 1988 , apj 334 , l81 matz s.m . , share g.h .
, leising m.d . ,
chupp e.l .
, vestrand w.t .
, 1988 , nature 331 , 416 maxted p.f.l . ,
marsh t.l . ,
north r.c . , 2000 , mnras , in press ( astro - ph/0007257 ) mccray r. , 1993 , ann .
31 , 175 meynet g. , arnould m. , 1999 , astronomy with radioactivities , mpe report 274 , p. 85
meynet g. , maeder a. , 1997 , a&a 321 , 465 morris d.j .
, et al . , 1995 , in 17@xmath63 texas symposium , eds . bhringer et al . , new york , p. 397
morris d.j . , et al . , 1997 , in 4@xmath63compton symposium , eds . dermer c.d . , strickman m.s . and kurfess j.d .
, aip 410 , new york , p. 1084
nakamura t. , umeda h. , nomoto k. , thielemann f .- k . , burrows a. , 1999 , apj 517 , 193 niemeyer j.c . , 1999 , apj 523 , l57 niemeyer j.c . ,
truran , j.w .
, 2000 , type ia supernovae : theory and cosmology , cambride univ .
press niemeyer j.c . , woosley s.e . , 1997 ,
apj 475 , 740 nomoto k. , kondo y. , 1991 , apj 367 , l19 nomoto k. , thielemann f .- k . , yokoi k. , 1984 , apj 286 , 644 nomoto k. et al . , 1997 , in thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) , p. 349
oberlack et al .
1996 , a&a suppl .
120 , 311 pinto p.a . ,
woosley s.e . , 1988 ,
nature 333 , 534 prantzos n. , diehl r. , 1996 , phys . rep .
267 , 1 prialnik d. , shara m.m . , shaviv g. , 1978 , a&a 62 , 339 prialnik d. , kovetz a. , 1995 , apj 445 , 789 rester a. c. , coldwell r.l .
, dunnam f.e . , eichhorn g. , trombka j.i . , starr r. , lasche g.p . , 1989 , apj 342 , l71 ruiz - lapuente p. , canal r. , burkert a. , 1997 , in thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) , p. 205
ruiz - lapuente p. , canal r. , isern j. , eds .
, 1997 , thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) sandie w. g. , nakano g. h. , chase l. f. jr .
, fishman g.j .
, meegan c. a. , wilson r. b. , paciesas w. s. , lasche g. p. , 1988 ,
apj 334 , l91 starrfield s. , 1989 , in classical novae , eds .
bode and a. evans , john wiley ( chichester ) , p. 39 starrfield s. , truran j.w . , sparks w.m . , 1978 ,
apj 226 , 186 .
starrfield s. , truran j.w .
, wiescher m. , sparks w.m . , 1998 ,
mnras 296 , 502 suntzeff n.b . ,
phillips m.m . , elias j.h . ,
depoy d.l .
, walker a.r . , 1992 ,
apj 384 , l33 teegarden b.j . ,
barthelmy s.d . ,
gehrels n. , tueller j. , leventhal m. , 1989 , nature 339 , 122 thielemann f .- k . , 1999 , in astronomy with radioactivities , mpe report 274 , p. 123
thielemann f .- k .
, nomoto k. , hashimoto m. , 1996 , apj 460 , 408 thielemann f .- k . , nomoto k. , yokoi k. , 1986 , a&a 158 , 17 timmes f.x . , woosley s.e . , hartmann d.h . ,
hoffman r.d . ,
weaver t.a . , matteucci f. , 1995 , apj 449 , 204 timmes f.x . , woosley s.e . ,
hartmann d.h . ,
hoffman r.d . , 1996 , apj 464 , 332 tueller j. , barthelmy s.d . , gehrels n. , teegarden b.j . , leventhal m. , maccallum c.j . , 1990 ,
apj 351 , l41 wheeler j.c , harkness r.p . , 1990 , rep .
phys . , 53 , 1467 wilson j.r . , 1971 , apj 163 , 209 woosley s.e . , 1997 , in thermonuclear supernovae , nato asi series , kluwer ( dordrecht ) , p. 313
woosley s.e . ,
hoffman r.d . , 1991 , apj 368 , l31 woosley s.e . , pinto p.a . , hartmann d. , 1989 , apj 346 , 395 woosley s.e . , weaver t.a . , 1994 ,
apj 423 , 371 woosley s.e .
, weaver t.a . , 1994 ,
apjs 101 , 181
|
many radioactive nuclei relevant for gamma - ray astrophysics are synthesized during explosive events , such as classical novae and supernovae .
a review of recent results of explosive nucleosynthesis in these scenarios will be presented , with a special emphasis on the ensuing gamma - ray emission from individual nova and supernova explosions .
the influence of the dynamic properties of the ejecta on the gamma - ray emission features , as well as the still remaining uncertainties in nova and supernova modelling will also be reviewed .
|
the chirality of molecules in living organisms must have been fixed at an early stage in the development of life .
all life that we know is based on rna and dna molecules with dextrarotatory sugars .
there is growing evidence that the rna world ( woese , 1967 ; crick , 1968 ; orgel , 1968 ; see also wattis & coveney 1999 ) must have been preceded by a simpler pre - rna world made up of achiral constituents ( bada , 1995 , nelson _
et al . , _ 2000 ) .
an alternative carrier of genetic code are peptide nucleic acids or pna ( nielsen , 1993 )
. these can be rather simple and are currently discussed in connection with the idea to build artificial life ( rasmussen _ et al . , _
2003 ) . furthermore
, although pna can still be chiral ( tedeschi _ et al . , _ 2002 ) , there are also forms of pna that are achiral ( pooga _ et al . , _ 2001 ) , suggesting that chirality may have developed later when the first rna molecules formed . in current proposals to build artificial life
, chirality does not seem to be crucial .
the pna molecules is proposed to act primarily as charge carrier , i.e. a very primitive functionality compared to the genetic code in contemporary cells ( rasmussen _ et al . , _
2003 ) . at this stage ,
homochirality may have been introduced by chance .
this is also supported by the fact that chiral polymers of the same chirality tend to have a more stable structure ( pogodina _ et al . , _ 2001 ) and would therefore be genetically preferred . since the introduction of chiral molecules
is assumed to take place at a stage when there is already growth and self - replication , it is also plausible to assume that the existence of chiral molecules has an autocatalytic effect in producing new chiral molecules of the same chirality ( kondepudi _ et al . , _
this is the basis of the recently proposed polymerization model of sandars ( 2003 ) ; see also wattis & coveney ( 2004 ) .
the purpose of the present paper is to reconsider this model ( or a slightly modified version of it ) and to analyze its stability behavior and conservation properties .
we also discuss and illustrate some of the salient features of the model in more detail .
the model is then compared with earlier models of homochirality where the detailed polymerization process is ignored and the dynamics of single variables representing left and right handed polymers are modeled instead ( frank , 1953 ; kondepudi and nelson , 1984 ; goldanskii and kuzmin 1989 ; avetisov and goldanskii , 1993 ; saito and hyuga , 2004 ) . in order to appreciate the nature of the many terms in the model of sandars we begin by discussing first the basic principle of the model in connection with homochiral polymer growth and then turn to the full set of reactions that are included in the model .
in this section we discuss the growth of polymers by adding monomers of the same chirality , i.e. we ignore reactions with monomers of the opposite chirality .
this is conceptually the simplest case , but its equilibrium solution also corresponds to a solution of the full system discussed below .
we write down the equations for left - handed polymers , but the same applies also to right - handed polymers . a left - handed polymer of length @xmath0
is assumed to react with a left - handed monomer via the reaction @xmath1 the reaction rate is @xmath2 , but since @xmath3 can bind to @xmath4 on either side , the total reaction rate is @xmath5 and proportional to the product of the concentrations of the two constituents .
we denote the concentration of @xmath6 chains by @xmath7 $ ] , so in a volume @xmath8 the number of @xmath6 chains is @xmath9v$ ] . for @xmath10
the number of possible pairs of @xmath11 and @xmath4 is @xmath12 .
a special situation arises for @xmath13 , because then @xmath4 is interacting with another @xmath4 , and the number of possible pairs is only @xmath14 .
[ this problem is familiar from the physics of nuclear reactions ; see , e.g. , kippenhahn and weigert ( 1990 ) . ]
we therefore introduce the factor @xmath15 defined by @xmath16 ( later we shall use this factor also with @xmath17 instead of @xmath18 . ) the corresponding contribution to the evolution of the concentration of @xmath6 is therefore @xmath19\over{{\rm d } { } } t}= ... +2k_s\sigma_n^{(1/2)}[l_{n}][l_1],\ ] ] where the dots denote the presence of other terms that will be discussed later .
obviously , the concentrations of @xmath3 and @xmath4 have to decrease at the same rate by the same amount , so @xmath20\over{{\rm d } { } } t}= ... -2k_s\sigma_n^{(1/2)}[l_{n}][l_1 ] , \label{lna}\ ] ] @xmath21\over{{\rm d } { } } t}= ... -2k_s\sigma_n^{(1/2)}[l_{n-1}][l_1 ] .
\label{l1a}\ ] ] in the following we regard @xmath0 as a general index with @xmath22 , the evolution of @xmath7 $ ] is governed by the difference of two terms ( gain from @xmath11 chains and loss in favor of producing @xmath23 chains ) .
the production of each @xmath6 contributes to a loss of @xmath4 monomers , so the right hand side of eq .
( [ l1a ] ) becomes a sum over all @xmath0 .
the full set of equations is then @xmath21\over{{\rm d } { } } t}=q_l-\lambda_l[l_1 ] , \quad\mbox{where}\quad\lambda_l=2k_s\sum_{n=1}^{n-1}[l_{n } ] , \label{l1b}\ ] ] @xmath20\over{{\rm d } { } } t}=2k_s[l_1]\big(\sigma_n^{(1/2)}[l_{n-1}]-[l_{n}]\big ) , \label{lnb}\ ] ] where @xmath24 denotes the production of new @xmath4 monomers ( see below ) .
a corresponding set of equations applies also to right - handed polymers , i.e. @xmath25 and @xmath26 . note that eqs .
( [ l1b ] ) and ( [ lnb ] ) obey the conservation law @xmath27[l_n ] , \label{deldt}\ ] ] where @xmath28 \label{eldef}\ ] ] is the total number of left - handed building blocks .
this number reaches an equilibrium if the supply of new left - handed monomers , @xmath24 , balances the loss associated with reactions involving the longest polymers possible for a given value of @xmath29 .
equation ( [ lnb ] ) shows that in the steady state we have @xmath7=\half[l_1]$ ] for all @xmath30 .
using eq .
( [ l1b ] ) , we find @xmath31 $ ] , and therefore @xmath32=[l_1]=\sqrt{q_l / k_sn}\quad\mbox{(steady state ) } \label{homochiralequilib}\ ] ] is a possible equilibrium solution .
new left and right handed monomers are assumed to be continuously reproduced from an achiral ( racemic ) substrate .
the rates of regeneration , @xmath24 and @xmath33 , depend on the concentration of the substrate , @xmath34 $ ] , and in some fashion on the relative concentrations of right and left handed polymers .
so , in general , we write @xmath35\big\{\half(1+f)c_l+\half(1-f)c_r+c_{0l}\big\ } , \label{qldef}\ ] ] @xmath36\big\{\half(1+f)c_r+\half(1-f)c_l+c_{0r}\big\ } , \label{qrdef}\ ] ] where @xmath37 and @xmath38 are some measures of the catalytic effect of the already existing right and left handed polymers , and the terms @xmath39 and @xmath40 allow for the possibility of non - catalytic production of left and right handed monomers possibly at different rates .
( unless noted otherwise , we keep @xmath41 . )
the concentration of the substrate is assumed to be maintained by a source @xmath42 , so we have @xmath43\over{{\rm d } { } } t}=q-\left(q_l+q_r\right ) , \label{dsdt}\ ] ] where @xmath44(c_l+c_r+c_{0l}+c_{0r})$ ] ; see eqs .
( [ qldef ] ) and ( [ qrdef ] ) . in general
, we expect @xmath37 and @xmath38 to be some function of @xmath6 and @xmath26 , respectively .
sandars ( 2003 ) assumed @xmath45 $ ] and @xmath46 $ ] , i.e. the catalytic effect depends on the concentrations of the longest possible chains of left and right handed polymers .
this assumption imposes a dependence on the cutoff value @xmath29 , a dependence that should preferably be avoided in numerical or other technical considerations .
the model should for example be stable and consistent in the limit when @xmath29 is infinite .
another option would be to assume @xmath47 $ ] and @xmath48 $ ] , where @xmath49 is a fixed value that is independent of the maximum chain length .
both alternatives have the disadvantage that @xmath50 $ ] and @xmath51 $ ] can never grow unless @xmath52 $ ] or @xmath53 $ ] are initially also finite . while it is plausible that long chains carry more catalytic weight than shorter ones , the dependence of the results on the particular choice of @xmath54 seems artificial .
( the allowance of finite values of @xmath39 and @xmath40 would remove this problem , although in practice both of these values should still be quite small . ) on may expect that the catalytic properties of the existing left and right handed polymers depend on the length of the polymer . the exact functional expression for this dependence is not known .
it is therefore important that a model that explains homochirality is not sensitive to the details of the catalytic properties and hence the functional form of @xmath55 and @xmath56 .
it turns out that the qualitative behavior of the model of sandars is indeed robust in this respect , e.g. a pitchfork bifurcation exists in both sandars original and in our model . to avoid artificial dependence on the maximal chain length @xmath29 , we chose to let the catalytic functions have the following form @xmath57 where @xmath58 is given by eq .
( [ eldef ] ) , and @xmath59 is defined analogously .
this is similar to the choice of wattis & coveney ( 2004 ) who assumed , independently of us , @xmath60 $ ] and @xmath61 $ ] .
we now comment on another aspect of the model of sandars .
he assumed that in the evolution of @xmath62 $ ] the loss is not @xmath63[l_n]$ ] , as it would be if eq .
( [ lnb ] ) were applied to @xmath64 , but he introduced an explicit linear damping term instead .
this implies that the model behaves discontinuously at the end of the chain .
we feel that an `` extrapolating '' ( continuous ) behavior is more reasonable , so we choose to apply eq .
( [ lnb ] ) also at @xmath64 . .
the time difference between the different curves is @xmath65 .
we have shown the first and last times as solid and dashed lines , and all other times as dotted lines .
the parameters are @xmath66 and @xmath67 . ]
it is interesting to note that in the continuous limit , eq . ( [ lnb ] ) becomes @xmath68 { \partial\over\partial n}\right)[l_n]=0,\ ] ] which describes waves traveling toward larger @xmath0 .
this is shown in figure [ pwave ] , where we have perturbed the equilibrium solution ( [ homochiralequilib ] ) by a gaussian and have solved eqs .
( [ l1b])([lnb ] ) numerically .
the wave is damped and has a speed that is proportional to @xmath69 , because for the steady state background solution @xmath50\sim ( k_s / q)^{1/2}$ ] .
note that the extrapolating boundary condition at @xmath64 allows the wave to escape freely .
in this paper we do not adopt the nondimensionalization of sandars .
instead , we note that there are only two physical dimensions in this problem : time and volume .
characteristic quantities with the dimensions of time and volume are @xmath70 and @xmath71 , respectively .
we therefore present all results by explicitly quoting these dimensions . in practice
this means that from now on we use @xmath72 as numerical values , but we keep the symbols in the equations for clarity . throughout this paper
we also assume @xmath67 ; calculations with different values do not seem to affect our results in any important way .
the fact that the equilibrium solution is constant for all @xmath30 implies that this value will decrease for longer choices of @xmath29 . in that sense
the solution is never converged .
this situation changes when we allow the ends of the left - handed polymers to be spoiled by right - handed monomers , as done by sandars ( 2003 ) .
this will be discussed in the next section .
already 20 years ago , joyce _ et al . , _ ( 1984 ) showed in an important paper describing experiments with template - directed polymerization that , once a monomer of the opposite chirality is bound to one end of the chain , the polymerization terminates on that end of the chain .
sandars ( 2003 ) incorporated this effect in his model and showed that this can lead to a bifurcation into two possible solutions of opposite chirality and hence to homochirality .
the full set of reactions included in his model is ( for @xmath73 ) @xmath74 and for all four equations we have the complementary reactions obtained by exchanging @xmath75 .
following sandars ( 2003 ) , we have introduced the new parameter @xmath76 , which quantifies that rate of enantiomeric cross - inhibition .
the special case @xmath77 corresponds to the case discussed in the previous section .
the most important effect of enantiomeric cross - inhibition is that a certain fraction of chains becomes spoiled by producing @xmath78 and @xmath79 polymers
. equation ( [ lnb ] ) and its complementary equation for right - handed polymers suffer therefore a loss proportional to @xmath80 , so we have instead @xmath20\over{{\rm d } { } } t}=2k_s[l_1]\big(\sigma_n^{(1/2)}[l_{n-1}]-[l_{n}]\big ) -2k_i[l_{n}][r_1 ] , \label{ln_new2}\ ] ] @xmath81\over{{\rm d } { } } t}=2k_s[r_1]\big(\sigma_n^{(1/2)}[r_{n-1}]-[r_{n}]\big ) -2k_i[r_{n}][l_1 ] . \label{rn_new2}\ ] ] these equations allow us to see what happens in the racemic case with @xmath82=[l_n]$ ] . in a steady state
we have ( for @xmath73 ) @xmath83=\half a^{-(n-1)}[l_1 ] \quad\mbox{(racemic solution)},\ ] ] where we have defined @xmath84 .
in particular , if @xmath85 , then @xmath7=2^{-n}[l_1]$ ] , i.e. @xmath7 $ ] drops by a factor of 2 from one @xmath0 to the next , except for @xmath86 to 2 , where it drops by a factor of 4 .
we should note , however , that this solution can be unstable ( see section [ sstability ] ) .
so far , we have not yet considered the evolution equations for the concentrations of the mixed terms , @xmath87 and @xmath88 . following sandars ( 2003 ) ,
we abbreviate the corresponding concentrations by @xmath89 $ ] and @xmath90 $ ] , respectively , i.e. without the subscript 1 on the terminating end of the chain .
the effect of generating these terms was already manifested in eqs .
( [ ln_new2 ] ) and ( [ rn_new2 ] ) through the appearance of the last term proportional to @xmath80 . nevertheless
, we do need to solve for @xmath89 $ ] and @xmath90 $ ] explicitly , because the reactions ( [ react3 ] ) and ( [ react4 ] ) consume @xmath4 and @xmath25 monomers , respectively .
the evolution equations for @xmath50 $ ] and @xmath51 $ ] are therefore given by @xmath21\over{{\rm d } { } } t}=q_l-\lambda_l[l_1],\quad { { { \rm d } { } } [ r_1]\over{{\rm d } { } } t}=q_r-\lambda_r[r_1 ] , \label{l1r1}\ ] ] where @xmath91 + 2k_i\sum_{n=1}^{n}[r_n ] + k_s\sum_{n=2}^{n-1}[l_nr ] + k_i\sum_{n=2}^{n}[r_nl ] , \label{lambdal}\ ] ] @xmath92 + 2k_i\sum_{n=1}^{n}[l_n ] + k_s\sum_{n=2}^{n-1}[r_nl ] + k_i\sum_{n=2}^{n}[l_nr ] , \label{lambdar}\ ] ] are the decay rates that quantify the losses associated with the reactions ( [ react1])([react4 ] ) , respectively . in eqs .
( [ lambdal ] ) and ( [ lambdar ] ) the concentrations @xmath89 $ ] and @xmath90 $ ] enter , so we have to solve their corresponding evolution equations ( for @xmath30 ) @xmath93\over{{\rm d } { } } t}= k_s[l_1]\big(\sigma_n^{(0)}[l_{n-1}r]-[l_{n}r]\big ) + k_i[r_1]\big(2[l_{n}]-[l_nr]\big ) , \label{lrevol}\ ] ] @xmath94\over{{\rm d } { } } t}= k_s[r_1]\big(\sigma_n^{(0)}[r_{n-1}l]-[r_{n}l]\big ) + k_i[l_1]\big(2[r_{n}]-[r_nl]\big ) , \label{rlevol}\ ] ] where the @xmath95 factor turns off the first term for @xmath13 ; see eq . ( [ sigman ] ) . in eqs .
( [ lrevol ] ) and ( [ rlevol ] ) the first two terms proportional to @xmath2 correspond to the homochiral growth on the unspoiled end , i.e. reaction ( [ react3 ] ) .
the third term comes from reaction ( [ react2 ] ) and the fourth term comes from reaction ( [ react4 ] ) and enters here and also in eqs .
( [ lambdal ] ) and ( [ lambdar ] ) as a loss term . for completeness
, we note that the corresponding gain enters in the evolution equations @xmath96\over{{\rm d } { } } t}=k_i[r_1][l_nr],\quad { { { \rm d } { } } [ lr_{n}l]\over{{\rm d } { } } t}=k_i[l_1][r_nl ] , \label{rlnr+lrnl}\ ] ] which are not explicitly required for constructing a solution , because these polymers no longer react with the monomers . nevertheless , solving eq .
( [ rlnr+lrnl ] ) simultaneously with eqs .
( [ ln_new2 ] ) and ( [ rn_new2 ] ) and eqs .
( [ l1r1])([rlevol ] ) can be useful for monitoring the evolution of the net chirality ; see section [ conservation ] .
note that , in contrast to the equations given by sandars ( 2003 ) , the truncation levels for the terms @xmath7 $ ] , @xmath97 $ ] , and @xmath98 $ ] are here the same , i.e. @xmath99 for both terms , whereas in the model of sandars the longest @xmath78 chain has @xmath100 , and the longest @xmath101 chain has only @xmath102 .
the reason we need to keep the same truncation levels for all three types of polymers is that we want to ensure that the behavior near the end of the chain ( @xmath64 ) does not deviate from the behavior elsewhere ( @xmath103 ) ; see the discussion in section [ homochiralgrowth ] .
for example , to ensure continuous behavior of @xmath7 $ ] at @xmath64 we need to keep the term @xmath104[r_1]$ ] in eq .
( [ ln_new2 ] ) .
this term , however , is the loss resulting from the gain of @xmath105 $ ] , so we have to keep the evolution equation for this term as well .
furthermore , the evolution equation for this term involves , in turn , the term @xmath106[l_nr]$ ] , which is the loss corresponding to the gain of @xmath107 $ ] .
if one regards the truncation level @xmath29 as an unrealistic feature of the model , as we do , then all three polymer types should be truncated at the same level .
$ ] ( left ) and @xmath97 $ ] ( right ) of equilibrium solutions for different values of @xmath108 .
for @xmath109 we have @xmath97=0 $ ] , which can not be seen in the logarithmic representation . ] in figure [ pspectra ] we show @xmath7 $ ] and @xmath97 $ ] for a number of equilibrium solutions for different values of @xmath108 and @xmath110 .
the corresponding values of @xmath82 $ ] and @xmath111 $ ] are small and not shown , except when @xmath112 in which case the solution is fully racemic with @xmath82=[l_n]$ ] and @xmath111=[l_nr]$ ] and is simply @xmath113=(n-1)[l_n]=(n-1)2^{-n}[l_1 ] .
\label{racemicsolution}\ ] ] for @xmath109 the solution is given by eq .
( [ homochiralequilib ] ) .
for @xmath114 the results are similar to those for @xmath110 provided @xmath115 . for @xmath116 , however , the solution is fully racemic and therefore the curves are independent of @xmath108 .
this racemic solution is similar to the case @xmath110 and @xmath117 that is shown in figure [ pspectra ] .
a realistic model of homochirality must also have an achiral ( racemic ) equilibrium solution .
it is generally anticipated that this racemic solution can be destabilized in the presence of catalytic reactions ( frank , 1953 ; avetisov and goldanskii , 1993 ) . if the probabilities of adding left and right handed monomers to a homochiral polymer are equal , i.e. if @xmath85 , the racemic solution given by eq .
( [ racemicsolution ] ) is also a possible solution for other values of the fidelity than @xmath112 , but it may of course be unstable .
we have carried out a numerical stability analysis by adding a small ( @xmath118 ) relative perturbation to the value of @xmath50 $ ] of the racemic solution .
it turns out that for certain values of the fidelity @xmath108 the departure of @xmath50 $ ] from the racemic equilibrium solution , @xmath119 $ ] , growth exponentially in time like @xmath120 . in figs
[ bifurc ] and [ bifurc01 ] we plot @xmath121 obtained from the slope of the graph of @xmath122(t)$ ] during the exponential growth phase , i.e. before a new nonlinear equilibrium is attained . in figs
[ bifurc ] and [ bifurc01 ] we also plot the corresponding chiral polarization parameter , @xmath123 , as a function of @xmath108 . here
we have chosen to define @xmath123 as @xmath124 it turns out that for @xmath110 the racemic solution is unstable when @xmath125 , and for @xmath114 it is unstable when @xmath126 .
the transition from an achiral to a chiral solution is a typical example of a pitchfork bifurcation ; see figs [ bifurc ] and [ bifurc01 ] .
this result is in qualitative agreement with sandars ( 2003 ) who found that for @xmath110 the critical value of @xmath108 is around 0.21 .
the differences in the numerical values are explained by differences in the model ( e.g. , the coupling to the substrate and the length of the maximum polymer length ) . , and @xmath66
. the dashed line indicates an unstable solution . ] , but for @xmath114 . ]
the growth rate of the instability is important for determining the time it takes for an almost racemic solution to become homochiral ( or at least non - racemic for @xmath127 ) . when @xmath110 , the growth rate @xmath121 is around 0.5 , but it becomes significantly smaller when the value of @xmath76 is reduced .
this shows explicitly that homochirality emerges as being due to enantiomeric cross - inhibition .
for homochiral growth the relevant conservation law is given by eq .
( [ deldt ] ) for @xmath58 , and similarly for @xmath59 . in general , however , because of the interaction with left and right handed monomers , there are no longer separate conservation laws for @xmath58 and @xmath59 . instead
, the complete set of equations , eqs .
( [ ln_new2 ] ) and ( [ rn_new2 ] ) together with eqs .
( [ l1r1])([lambdar ] ) , satisfies @xmath128 where @xmath129 and @xmath130 are the net input and output rates of chirality , respectively , and @xmath131 is the total chirality , where @xmath132+\sum_{n=2}^n(n-1)[r_nl ] + \sum_{n=3}^n(n-2)[lr_nl ] , \label{ertilde}\ ] ] @xmath133+\sum_{n=2}^n(n-1)[l_nr ] + \sum_{n=3}^n(n-2)[rl_nr ] , \label{eltilde}\ ] ] denote the total numbers of right and left handed building blocks ( or enantiomers ) , where opposite enantiomers are counted such that they annihilate enantiomers of the opposite chirality .
the loss terms resulting from the finite truncation level , @xmath29 , are denoted by @xmath134[r_n]+k_s(n-1)[r_1][r_nl],\ ] ] @xmath135[l_n]+k_s(n-1)[l_1][l_nr].\ ] ] in order to evaluate the quantities @xmath136 and @xmath137 we have to integrate the evolution equations ( [ rlnr+lrnl ] ) for the production of terminally spoiled polymers even though they undergo no further evolution . in a sense
the integration of the terminally spoiled polymers acts only as counters that keeps track of the number of polymers that are lost during the polymerization process .
the expressions for @xmath136 and @xmath137 involve sums over @xmath138 $ ] and @xmath98 $ ] , but since these quantities do not occur on the right hand sides of the governing evolution equations , their values are not constrained by the dynamics and depend on the initial conditions and continue to evolve in time even though the system may have reached an equilibrium .
the so defined net chirality can therefore not be used to characterize a particular solution , and we have to restrict ourselves either to @xmath59 and @xmath58 , or to @xmath139 and @xmath140 , which are defined by taking only the first two sums in eqs .
( [ ertilde ] ) and ( [ eltilde ] ) , i.e.@xmath141+\sum_{n=2}^n(n-1)[r_nl ] , \label{ertilde}\ ] ] @xmath142+\sum_{n=2}^n(n-1)[l_nr ] .
\label{eltilde}\ ] ] in table [ tres ] we list the resulting values of @xmath123 , defined in eq . ( [ eta ] ) , an analogously defined @xmath143 , @xmath144 , and @xmath145 .
we also give the mean polymer lengths , @xmath146/\sum[r_n]$ ] and @xmath147/\sum[l_n]$ ] , of right and left handed polymers .
the polymerization model of sandars ( 2003 ) is significantly different from all the previously proposed models of homochirality that ignore the detailed polymerization process by only describing some scalar quantities , say @xmath148 and @xmath149 , that are representative of the number of left and right handed polymers . in the papers by saito and hyuga ( 2004 )
it was shown that neither linear nor nonlinear autocatalytic behavior suffice to produce homochirality , and that a backreaction term is needed .
their model equations are @xmath150 where @xmath151 and @xmath152 is the feedback parameter .
for @xmath153 there is a continuous range of solutions along the line @xmath154 , i.e. homochirality does not emerge unless the initial condition is already homochiral . for finite ( but small ) values of @xmath152
there are two nontrivial stable fixed points .
( the trivial solution , @xmath155 , is always a stable fixed point in this model . )
the model of saito and hyuga ( 2004 ) , hereafter the sh model , does capture the expected behavior , but it remains unsatisfactory in that its functional form has been introduced ad hoc .
it is therefore desirable to derive simple model equations based on the polymerization equations of sandars ( 2003 ) .
it turns out that , without changing the basic properties of the model , a minimal version is still meaningful for @xmath156 , and that the equations for the semi - spoiled polymers , @xmath157 $ ] and @xmath158 $ ] , can be ignored ( as already done by sandars ) .
thus , we only solve eqs .
( [ ln_new2 ] ) and ( [ rn_new2 ] ) together with eqs .
( [ l1r1])([lambdar ] ) .
following sandars ( 2003 ) , we also assume that @xmath159 $ ] and @xmath160 $ ] ( instead of @xmath161 and @xmath162 , which would yield more complicated expressions ) .
a further simplification can be made by regarding @xmath163 $ ] as a rapidly adjusting variable that is enslaved to @xmath50 $ ] ( and similarly for @xmath164 $ ] ) .
this technique is also known as the adiabatic elimination of rapidly adjusting variables ( e.g. , haken , 1983 ) .
equation ( [ ln_new2 ] ) becomes @xmath165 ^ 2 - 2[l_2]\big(k_s[l_1]+k_i[r_1]\big),\ ] ] which is solved for @xmath163 $ ] ( and similarly for @xmath164 $ ] ) , which in turn couples back to the equations for @xmath50 $ ] and @xmath51 $ ] via @xmath24 and @xmath33 .
finally , we also treat the substrate @xmath34 $ ] as a rapidly adjusting variable , i.e. we have @xmath166=q/([l_2]+[r_2])$ ] .
we emphasize that the adiabatic elimination does not affect the accuracy of steady solutions .
it is convenient to introduce new dimensionless variables , @xmath167(2k_s / q)^{1/2},\quad y=[l_1](2k_s / q)^{1/2},\quad \tau = t(qk_s/2)^{1/2}.\ ] ] in order to compare first with the sh model we restrict ourselves to the special case @xmath168 , which leads to the revised model equations @xmath169 where dots denote derivatives with respect to @xmath170 and @xmath151 and @xmath171 has been introduced for brevity .
equations ( [ modeleqn ] ) resemble the equations of the sh model in that both have a quadratic term proportional to @xmath172 ( or @xmath173 ) , which is quenched either by a @xmath174 factor ( in the sh model ) or by a @xmath175 factor in our model .
furthermore , both models have a backreaction term proportional to @xmath176 ( or @xmath177 ) , but the coefficient in front of this term ( @xmath152 in the sh model ) is not constant but equal to @xmath178 . ) for two different values of @xmath108 .
the starting points of each trajectory are marked by small dots and stable fixed points are marked by big dots . ] in the general case with @xmath179 , @xmath127 , as well as finite values of @xmath180 , and @xmath181 , the equations read @xmath182 where we have introduced the abbreviations @xmath183 , @xmath184 , @xmath185 , @xmath186 , @xmath187 , @xmath188 , and @xmath189 . in figure [ pruns051 ]
we show trajectories of solutions of eq .
( [ modeleqnf ] ) for two different values of @xmath108 in an @xmath190 phase diagram .
note that all equilibrium solutions lie on the line @xmath191 .
this property allows us to calculate equilibrium solutions for general values of @xmath108 .
inserting @xmath192 yields a cubic equation of which one solution is always @xmath193 .
this reduces the problem to a quadratic equation with the solution @xmath194 linearizing the equations around the racemic solution , @xmath195 , yields the growth rate @xmath196 in agreement with our numerical results for large values of @xmath29 , this equation gives a linear dependence of the growth rate on the fidelity .
this result also shows that for @xmath197 perturbations decay exponentially . ) for @xmath198 and @xmath199 using the newton - raphson method . ] in the presence of a biased , non - catalytic generation of monomers ( finite @xmath200 or @xmath201 with @xmath202 ) there is no longer a perfectly racemic equilibrium solution .
the sign of @xmath123 for the solution for @xmath112 depends on the sign of @xmath203 . along this solution branch @xmath123
goes further away from zero in a continuous fashion until @xmath109 . at some value of @xmath108
a pair of new solutions emerges , one is stable and the other one unstable , but both have the opposite sign of @xmath123 ; see figure [ pbranches ] . among these new branches ,
the stable one can only be reached via a finite amplitude perturbation .
this behavior is called an imperfect bifurcation and has long been anticipated in this context ( kondepudi and nelson , 1983 ; kondepudi _ et al . , _ 1986 ; goldanskii and kuzmin , 1989 ) . the steady solutions shown in figure [ pbranches ]
have been obtained by solving eq .
( [ modeleqnf ] ) using the newton - raphson method .
this method allows us to find both stable and unstable solutions .
near the bifurcation point the diagram is extremely sensitive to the addition of a bias parameter .
it is remarkable that for a value as small as @xmath204 a relatively large gap has been produced in the bifurcation diagram .
finite values of @xmath200 and @xmath201 could result from physical influences , for example polarized synchrotron radiation from neutron stars ( but see bonner , 1999 ) , uv radiation in star - forming regions ( bailey , 2001 ) , or the parity violation of the electroweak force ( e.g. , hegstrom , 1984 ) .
in all these cases the expected effect is however very small ( bada , 1995 ) .
we emphasize , however , that the main reason for homochirality is the instability of the racemic ( or nearly racemic ) solution , which is hardly modified by a finiteness of @xmath200 or @xmath201 .
the origin of homochirality has long been thought to be the result of a bifurcation process that can vastly amplify a very small random enantiomeric excess which can then prevail forever .
generic model equations reproducing the expected bifurcation behavior have so far mostly been proposed on an ad hoc basis .
it was therefore difficult to establish a connection between model and reality . according to the work of saito and hyuga ( 2004 ) one expects two effects to be important : nonlinearity and backreaction .
however , the functional form of these terms remained open .
furthermore , the meaning of non - perfect catalytic fidelity and enantiomeric cross - inhibition within the framework of the model were not clear . in the present paper
we have established a direct connection between the more detailed polymerization model of sandars ( 2003 ) and the simpler model equation approach with only two ordinary differential equations .
in particular , the present work has confirmed that the relevant nonlinearity is indeed quadratic ( as in the sh model ) , but it is not quenched like @xmath174 , but rather like @xmath175 , where @xmath178 and @xmath205 are measures of the total concentrations of monomers ( both right and left handed ) .
furthermore , the feedback coefficient is not a small constant , as in the sh model , but it is itself proportional to @xmath178 . more importantly ,
imperfect fidelity and enantiomeric cross - inhibition , as well as the effects of a weakly biased non - catalytic production of new monomers , have a quantitative meaning within the framework of the reduced model . for a more quantitative comparisons of the polymerization process with experiments
the full set of equations of sandars ( with the revisions discussed above ) is to be preferred . a number of features that can only be captured by the full model .
an example is the wave - like propagation in the distribution of homochiral polymers .
an experimental confirmation would help to quantify the growth coefficient @xmath2 characterizing the probability that a polymer grows by a monomer of the same chirality . on the other hand ,
the growth coefficient for enantiomeric cross - inhibition , @xmath76 , determines primarily the minimum fidelity parameter , @xmath108 , above which bifurcation and hence homochiral growth is at all possible .
it is indeed quite remarkable , that the main reason homochiral growth occurs is that binding with a wrong enantiomer spoils further polymerization on the corresponding end of the chain .
this leads to competition which is always a key feature of natural selection processes such as these .
homochirality in living organisms is a singular phenomenon .
non - living chemical systems do in general not have a preferred chirality . in the models presented in this paper this
is reflected in figs [ bifurc ] and [ pbranches ] .
the region of the phase diagram displaying homochirality is characterized by high fidelity , i.e. high auto - catalytic accuracy .
the fidelity is expected to be significantly higher in living systems .
when an organism dies the auto - catalytic polymerization stops and as a consequence the fidelity is sharply decreased .
the characteristic behavior of the polymerization changes from the chiral to the racemic region of the phase diagram .
the relaxation of the system from the homochiral to the racemic state is often very slow .
it was in fact suggested by hare and mitterer ( 1967 ) and later by bada _
et al . , _ ( 1970 ) that racemization of amino acids in fossil material could be used as a dating method .
unfortunately it has turned out that the rate of racemization is strongly temperature dependent , which tends to make this dating method unreliable .
pogodina , n. v. , lavrenko , v. p. , srinivas , s. , and winter , h. h. : 2001 , rheology and structure of isotactic polypropylene near the gel point : quiescent and shear - induced crystallization , _ polymer _ * 42 * , 90319043 .
tedeschi , t. , corradini , r. , marchelli , r. , pushl , a. , and nielsen , p. e. : 2002 , racemization of chiral pnas during solid - phase synthesis : effect of the coupling conditions on enantiomeric purity , _ tetrahedron : asymmetry _ * 13 * , 16291636 .
|
the stability and conservation properties of a recently proposed polymerization model are studied . the achiral
( racemic ) solution is linearly unstable once the relevant control parameter ( here the fidelity of the catalyst ) exceeds a critical value .
the growth rate is calculated for different fidelity parameters and cross - inhibition rates .
a chirality parameter is defined and shown to be conserved by the nonlinear terms of the model .
finally , a truncated version of the model is used to derive a set of two ordinary differential equations and it is argued that these equations are more realistic than those used in earlier models of that form .
|
the pions @xmath11 are the goldstone bosons of spontaneous chiral symmetry breaking in qcd : @xmath12 .
their low - energy dynamics can therefore be calculated systematically ( and accurately ) with chiral perturbation theory in form of a loop - expansion based on an effective chiral lagrangian .
the very accurate two - loop prediction @xcite for the isospin - zero s - wave @xmath8-scattering length @xmath13 has recently been confirmed in the e865 @xcite and na48/2 @xcite experiments by analyzing the @xmath14 invariant mass distribution of the rare kaon decay mode @xmath15 .
one particular implication of that good agreement between theory and experiment is that the quark condensate @xmath16 constitutes the leading order parameter @xcite of spontaneous chiral symmetry breaking ( considering the two - flavor sector of qcd ) . as a consequence
, one can assert that more than @xmath17 of the ( squared ) pion mass , @xmath18mev/@xmath19 , must stem from the term linear in the light quark mass multiplied with the quark condensate @xmath20 @xcite .
furthermore , the dirac experiment @xcite at cern has been proposed to determine the difference of the isospin - zero and isospin - two s - wave @xmath8-scattering lengths @xmath21 by measuring the life time ( @xmath22fs ) of pionium ( i.e. @xmath14 bound electromagnetically and decaying into @xmath23 ) . in the meantime
the na48/2 collaboration @xcite at cern has accumulated very high statistics for the charged kaon decay modes @xmath24 , which allowed them to extract the value @xmath25 for the @xmath8-scattering length difference from the cusp effect in the @xmath23 mass spectrum at the @xmath14 threshold .
this experimental result is again in very good agreement with the two - loop prediction @xmath26 of chiral perturbation theory @xcite . for a discussion of isospin breaking corrections which have to be included in a meaningful comparison between theory and experiment , see also ref.@xcite . clearly , these remarkable confirmations give confidence that chiral perturbation theory is the correct framework to calculate reliably and accurately the strong interaction dynamics of the pions at low energies .
electromagnetic processes offer further possibilities to probe the internal structure of the pion .
whereas the charge ( or vector ) form factor of the pion , @xmath27 , is largely dominated by the low - lying @xmath28-resonance , pion compton scattering @xmath1 allows one to extract the electric and magnetic polarizabilities of the ( charged ) pion . in a classical picture
these polarizabilities characterize the deformation response ( i.e. induced dipole moments ) of a composite system in external electric and magnetic fields . in the proper quantum field
theoretical formulation the electric and magnetic polarizabilities , @xmath29 and @xmath30 , are defined as expansion coefficients of the compton scattering amplitudes at threshold .
since pion targets are not directly available , real pion compton scattering has been approached using different artifices , such as high - energy pion - nucleus bremsstrahlung @xmath31 @xcite , radiative pion production off the proton @xmath32 @xcite , and the crossed channel two - photon reaction @xmath33 as embedded in the @xmath34 process @xcite .
the corresponding results for the electric and magnetic pion polarizabilities scatter substantially and they have also large uncertainties . on the other hand , chiral perturbation theory at two - loop order gives the firm prediction @xmath35@xmath36 @xcite for the polarizability difference .
it is however in conflict with the existing experimental results from serpukhov @xmath37@xmath36 @xcite and mami @xmath38@xmath36 @xcite which amount to values more than twice as large .
the result @xmath39@xmath36 @xcite extracted from the mark ii data @xcite is consistent with the chiral prediction , but the corresponding low - energy cross sections for @xmath40 are ( within their errorbars ) rather insensitive to even sizeable changes of the pion polarizabilities @xcite .
we also note that radiative pion photoproduction @xmath32 has recently been considered in the framework of heavy baryon chiral perturbation theory @xcite .
it has been argued in ref.@xcite that additional contributions from @xmath41 vertices in the chiral lagrangian @xmath42 may have effects on the cross section which are comparable to those of the pion polarizabilities .
in that situation , it is very promising that the ongoing compass experiment @xcite at cern aims at measuring the pion polarizabilities , @xmath29 and @xmath43 , with high statistics using the primakoff effect .
the scattering of high - energy negative pions in the coulomb field of a heavy nucleus ( of charge @xmath44 ) gives access to cross sections for @xmath45 reactions through the equivalent photon method : @xmath46 here , @xmath47 is the @xmath45 center - of - mass energy and @xmath48 the squared invariant momentum transfer between the initial and final state pion .
@xmath49 denotes the momentum transferred by the virtual photon to the heavy nucleus , and one aims at isolating the coulomb peak @xmath50 from the strong interaction background . due to the small range of @xmath51
the nuclear charge form factor can be approximated by @xmath52 . a first pilot run performed at compass
has already accumulated statistics for primakoff events comparable to previous experiments of this type @xcite .
for a recent discussion of possible problems ( nuclear scattering , kinematical limitations etc . ) associated with isolating the coulomb peak , see refs.@xcite .
the purpose of the present paper is to review the cross sections for various low - energy @xmath45 reactions in the framework of chiral perturbation theory .
we treat pion compton scattering , @xmath1 , up to one - loop order , where the pion s internal structure appears for the first time in form of the difference of the electric and magnetic polarizability , @xmath53 . in the view
the expected accuracy of the compass experiment , and the knowledge that two - loop corrections @xcite are relatively small , this approximation should be sufficient .
then , we consider ( at one - loop order ) the reaction @xmath54 which is also under study in the compass experiment .
it serves as a test of the qcd chiral anomaly via measuring the @xmath4-coupling constant @xmath5 .
as the pertinent one - loop amplitudes are available from the literature , we restrict here ourselves to a concise but explicit presentation of the relevant cross section formulas .
actually new is our calculation of total cross sections for neutral and charged pion - pair production , @xmath55 and @xmath56 , which interestingly are determined by the chiral @xmath8-interaction . with the larger data sample planned in future compass
runs sufficient statistics will become available for these reactions . finally , we investigate the radiative correction process @xmath9 and calculate the corresponding two - photon mass spectrum .
this information will be helpful for analyzing the @xmath10 primakoff scattering events with photons in the final state .
we start with the process of compton scattering off a negatively charged pion : @xmath57 .
the corresponding t - matrix in the center - of - mass frame has ( in coulomb gauge @xmath58 ) the form : @xmath59 \big\}\,,\ ] ] with @xmath60 the fine - structure constant , and @xmath61 and @xmath62 the independent mandelstam variables .
performing the sums over transversal polarizations and applying flux and appropriate phase space factors , the resulting differential cross section reads : @xmath63 with @xmath64 where @xmath65 is the cosine of the cms scattering angle . the decomposition in eq.(2 ) into two invariant amplitudes @xmath66 and @xmath67 has been done with hindsight to an expression as simple as possible for the ( unpolarized ) differential cross section .
the amplitudes at tree level coincide with the ones from scalar quantum electrodynamics : @xmath68 note that the contribution of the @xmath69-channel pole diagram vanishes in coulomb gauge since the coupling of the initial state photon becomes zero in that gauge , @xmath70 .
the one - pion loop diagrams of chiral perturbation theory generate , after renormalization of the pion mass @xmath71 , the following ( finite ) contribution to the compton amplitude @xmath66 @xcite : @xmath72 with @xmath73mev the pion decay constant .
heuristically , it can be interpreted as the ( leading ) correction arising from photon scattering off the pion cloud around the pion .
the internal structure of the pion enters through its electric and magnetic polarizabilities , which obey at the one - loop order the constraint @xmath74 @xcite .
the pertinent @xmath75 contact vertices from the chiral lagrangian @xmath76 give rise to the contribution @xcite : @xmath77 to the compton amplitude @xmath66 .
the relevant combination of low - energy constants @xmath78 can be extracted from the axialvector - to - vector form factor ratio @xmath79 measured in the pibeta experiment @xcite at psi via the radiative pion decay @xmath80 .
the two - loop analysis of ref.@xcite yields the value @xmath81 implying the pion polarizability difference : @xmath82@xmath36 .
it should be stressed that the current - algebra relation @xmath83 constitutes a low - energy theorem @xcite which must hold ( to leading order ) in any chiral invariant theory . from this point of view , the parametrization of @xmath2 in terms of a phenomenological @xmath84-meson exchange , as chosen in ref.@xcite , is problematic since it ignores that profound theoretical constraint . moreover , as stressed in ref.@xcite the outcome @xmath85 of the chiral invariant linear @xmath84-model is completely ruled out by experiment , because it leads to an axialvector - to - vector form factor ratio @xmath86 of the wrong sign . as a matter of fact ,
the actual physics behind the low - energy constant @xmath78 is the excitation of vector and axialvector meson resonances @xcite . in the same way the existing experimental determinations of @xmath87@xmath36 @xcite give reason to doubts since they violate the chiral low - energy theorem notably by a factor @xmath88 . in order to clarify the situation on the theoretical side ,
a complete two - loop calculation of pion polarizabilities has been performed in ref.@xcite , grinding out the analytical result : @xmath89 expressed in terms of the chiral low - energy constants @xmath90 , @xmath91 , @xmath92 , @xmath93 .
the last term proportional to @xmath94 stems from the nonfactorizable acnode diagrams .
the additional counterterm @xmath95 has been estimated via resonance saturation as @xmath96 @xcite and we have set the scale in the chiral logarithm @xmath97 equal to the @xmath98-meson mass @xmath99mev/@xmath19 .
taking into account various theoretical uncertainties , the two - loop prediction amounts to : @xmath100@xmath36 @xcite .
as this value is fully compatible with the current - algebra result , it is now assured that there are no significant corrections to the chiral low - energy theorem for the pion polarizability difference @xmath101 . for a more detailed discussion of these issues , see ref.@xcite .
the non - vanishing two - loop prediction for the pion polarizability sum , @xmath102@xmath36 @xcite , is in fact consistent with results from dispersion sum rules @xcite but presumably too small to cause an observable effect in pion compton scattering .
moreover , as indicated e.g. by fig.6 in ref.@xcite the one - loop approximation should be sufficient for low energies @xmath103 . )
contribute at the level of @xmath104 . ]
the dashed curves represent the born terms in eq.(4 ) . ]
fig.1 shows the calculated total cross section @xmath105 for @xmath10 compton scattering in the low - energy region @xmath106 . at higher energies
the excitation of the broad @xmath28-resonance becomes prominent ( @xmath107 ) .
one observes that the one - loop result is almost indistinguishable from the cross section for a point - like pion : @xmath108\,.\ ] ] the effects from the pion s low - energy structure ( pion - loops and electric minus magnetic polarizability ) contribute at the level of only @xmath104 .
this finding is of course not entirely new @xcite .
the angular dependence of the differential cross section @xmath109 is shown in fig.2 for three selected center - of - mass energies , @xmath110 .
the full curves are calculated with ( assumed ) pion polarizabilities @xmath111@xmath36 and the dashed lines correspond to the case of a structureless pion : @xmath112 \over s[s(1+z)+m_\pi^2(1-z)]^2}\,,\ ] ] where @xmath113 .
one gets instructed that the effects of the pion s low - energy structure on observables are rather small . for @xmath114 , they amount to at most an @xmath115 percent reduction of the differential cross section in backward directions , @xmath116 .
this gives some impression of the experimental challenge posed for measuring with good accuracy the pion polarizabilities .
first , low - energy @xmath45 cross sections ( e.g. @xmath117 ) need to be extracted from the primakoff events in the coulomb peak . secondly , these must come with an accuracy such that the deviations from the point - like cross section become statistically significant .
we also note that the ( quadratic ) pion - loop contribution @xmath118 in eq.(5 ) works against the ( linear ) polarizability term @xmath119 .
therefore , leaving it out in an analysis of pion compton scattering data would lead to an underestimation of the pion polarizabilities @xmath120 . in the optimal case their fitted value
should also be independent of the upper limit put on the center - of - mass energy @xmath121 . in order to visualize more directly the effect of the pion polarizabilities , we show in fig.3 the ratio @xmath122 between the differential cross sections ( at one - loop order ) in backward direction , @xmath123 , evaluated with finite polarizabilities @xmath124@xmath36 , and with polarizabilities set equal to zero . at @xmath125
the effect becomes now quite sizeable , where it amounts to almost a @xmath126 reduction of the ( backward ) differential cross section .
fig.3 reemphasizes the necessity to include ( at least ) the pion - loop correction eq.(5 ) in the analysis of primakoff data . of course
, it would also be interesting to see how the ratio displayed in fig.3 gets affected by the two - loop corrections of ref.@xcite , in particular because of the presence of a branch point at @xmath127 . between the differential cross sections in backward direction ( @xmath123 ) calculated with finite , @xmath128@xmath36 , and with zero pion polarizabilities .
next , we come to the neutral pion production reaction @xmath129 . the corresponding t - matrix :
@xmath130 involves ( as any process which violates natural parity ) the totally antisymmetric @xmath131-tensor and @xmath132 is a dimensionless invariant function equal to @xmath133 in the soft - pion limit . the prefactor @xmath134gev@xmath135 is fixed by the chiral anomaly of qcd ( i.e. the anomalous @xmath136 rectangle quark diagram ) . in the effective chiral lagrangian ,
the @xmath137-vertex is provided by the famous wess - zumino - witten term ( together with the @xmath138 decay vertex ) .
neutral pion production in primakoff reactions serves therefore as test of the qcd chiral anomaly .
the low - statistics experiment performed some time ago at serpukhov has obtained the somewhat high value @xmath139gev@xmath135 @xcite .
an improved measurement of @xmath5 with much higher statistics has been proposed by the compass collaboration @xcite and preliminary data are presently already being analyzed @xcite . here , it should be noted that electromagnetic corrections @xcite as well as two - loop corrections @xcite have been calculated for the process @xmath140 .
if these corrections are included in the analysis of the serpukhov data the extracted value of @xmath5 gets somewhat reduced to @xmath141gev@xmath135 @xcite .
the total cross section for @xmath140 following from the t - matrix in eq.(10 ) has the form : @xmath142 after substituting @xmath143 with @xmath144 the cosine of the @xmath10 cms scattering angle .
the prefactor @xmath145 signals that the final state pions are produced in a relative p - wave .
the one - loop corrections for the process @xmath140 have been calculated in chiral perturbation theory some time ago in ref.@xcite .
together with the tree - level contribution the invariant amplitude @xmath132 reads : @xmath146 -{1\over 2 } -\ln{m_\pi \over m_\rho } \bigg\ } + { 3m_\pi^2 \over 2m_\rho^2}\,,\ ] ] where @xmath147 and we have introduced the loop function : @xmath148 the correction terms to @xmath133 in eq.(12 ) represent one - loop pion - rescattering in the @xmath69- , @xmath149- , and @xmath150-channel .
the last term in eq.(12 ) corresponds to the counterterm contribution estimated via vector meson exchange and using simplifying symmetry relations for the coupling constants @xcite . as a function of the center - of - mass energy @xmath121 . ]
the three lower curves in fig.4 show the total cross section for @xmath151 in the region @xmath152 .
one sees that the pion - loop and the counterterm corrections each enhance the total cross section by about @xmath153 in comparison to the tree - level approximation . above center - of - mass energies of @xmath154 the effects of the prominent @xmath28-resonance
can no more be represented by a local counterterm .
this is clearly demonstrated by the upper dashed - dotted curve in fig.4 which stems from a @xmath98-meson exchange model : @xmath155 including in the @xmath69-channel exchange term an energy - dependent @xmath98-meson decay width : @xmath156 here , @xmath157 is determined from the empirical decay width @xmath158mev @xcite , and the coupling constant @xmath159 can be inferred from the empirical branching ratio br@xmath160 @xcite . at the @xmath98-meson
peak the resonant cross section exceeds the one due to the chiral low - energy terms by about a factor 5 .
potentially , this spectral shape offers a strategy to avoid absolute cross section measurements , namely by covering in the compass experiment the whole region from the @xmath98-peak down to threshold . yet
, some model dependence will remain in such a procedure of extracting @xmath5 . clearly , a careful analysis of the near - threshold data for @xmath161 should the include radiative and higher loop corrections of refs.@xcite .
next , we turn to double neutral pion production @xmath162 .
this reaction has not been considered so far , mainly because of the lack of any experimental data . with the expected high - statistics of the compass experiment @xcite this
may change in the near future . .
in the case of charged pion - pair production , @xmath163 , the incoming photon couples also to the other two outgoing pions . ] the tree level diagrams for @xmath164 are shown in fig.5 . in the convenient parametrization of the special - unitary matrix field
@xmath165 a @xmath166 contact - vertex does not exist .
( in other parametrizations its contribution gets canceled by the off - shell part of the @xmath167-vertex . )
moreover , since the left diagram vanishes in coulomb gauge @xmath168 we need to evaluate only one single diagram . performing the polarization sum and three - body phase space integration
, we end up with the following expression for the total cross section for @xmath169 : @xmath170 here , the integration variable @xmath171 is @xmath172 times the cms energy of the outgoing @xmath10 , and the factor @xmath173 originates from the chiral @xmath174 interaction .
the lower dashed curve in fig.6 shows this total cross section as a function of the center - of - mass energy in the region @xmath175 .
near threshold it grows cubically with the excess energy , @xmath176 $ ] . with values up to one microbarn , this cross section is still quite sizeable .
note that a resonant @xmath98-meson contribution is now not possible , since @xmath177 is forbidden by g - parity . for the sake of completeness
we treat also the charged pion - pair production process @xmath178 .
two additional diagrams where the photon couples to the other outgoing pions , need then to be included . putting all pieces together
, we end up with the following expression for total cross section for @xmath179 : @xmath180 ^ 2 + \big[k_0 \sqrt{s}-p_0\omega_1-k_0 q_1 x\big]^2 \bigg\ } \nonumber \\ & & + { 2q_1q_2(z - x y ) \over ( \omega_1 -q_1 x)(\omega_2-q_2 y)}\bigg\{\big[p_0 ( \sqrt{s}-\omega_2)-\sqrt{s}\omega_1-k_0 q_2 y\big ] \nonumber \\ & & \times \big[p_0(\sqrt{s}-\omega_1)-\sqrt{s}\omega_2-k_0 q_1 x\big]+ 2\big[k_0\sqrt{s } -p_0\omega_2-k_0 q_2 y\big ] \nonumber \\ & & \times \big[p_0(\omega_1+\omega_2)- \sqrt{s}\omega_1+k_0 ( q_1 x+q_2 y)\big ] \bigg\ } \bigg\ } \ , , \end{aligned}\ ] ] with @xmath181 , @xmath182 , @xmath183 and @xmath184 and @xmath185 as a function of the total center - of - mass energy @xmath121 . ] in eq.(17 ) we have exploited the permutational symmetry of the three - pion phase space in order reduce the number of independent interference terms in the integrand .
this cross section grows in the threshold region also cubically with the excess energy , @xmath186 $ ] .
after numerical evaluation of the four - dimensional integral we obtain the total cross sections shown by the upper full curve in fig.6 .
they come out about a factor @xmath187 larger than the ones for @xmath188 , reaching @xmath189b at @xmath190 .
this enhancement can be traced back to the larger number of contributing diagrams . at their upper
ends the curves in fig.6 should be considered only as indicative since there higher orders may become significant .
finally , we consider the radiative correction process @xmath192 to pion compton scattering . as a smooth background process it interferes also with the neutral pion production @xmath193 . therefore
a knowledge of its order of magnitude will be helpful in analyzing the primakoff data . .
the coupling of the incoming photon at the bottom of the pion line vanishes in coulomb gauge . diagrams with the two outgoing photons interchanged are not shown . ] a representative set of tree diagrams is shown in fig.7 .
these need to be supplemented by diagrams with the two outgoing ( right ) photons interchanged .
we exploit the fact that the coupling of the incoming ( left ) photon at the bottom of the pion line vanishes in coulomb gauge and thus are left with 9 non - vanishing diagrams . after performing the triple sums over transversal photon polarizations for the 45 interference terms and applying flux and phase space factors , we end up with a double differential cross section of the form : @xmath194 here , @xmath195 stands for a rational function of its six arguments ( which is much too long to be reproduced here ) . @xmath196 and @xmath197 denote the cms energies of the final state photons and @xmath198 as a function of the variable @xmath199 . ] the two equations @xmath200 and @xmath123 determine the phase space boundaries as a straight line and a hyperbola in the @xmath201-plane .
it is well - known that soft photons with @xmath202 cause a logarithmic infrared divergence of the integrated ( total ) cross section .
therefore , we introduce as a suitable lorentz - invariant variable the @xmath203 invariant mass squared : @xmath204 related to the sum of the photon energies @xmath205 and integrate over their half difference @xmath206 . for the purpose of a uniform representation
we express the two - photon mass spectrum @xmath207 in terms of the dimensionless variable @xmath208 which ranges for all @xmath121 in the unit interval , @xmath209 .
fig.8 shows the calculated two - photon mass spectrum @xmath210 for the process @xmath211 at four values of @xmath212 .
for @xmath213 the curves diverge as @xmath214 .
it is astonishing that the @xmath203-spectra possess only a very weak dependence on the total center - of - mass energy @xmath121 .
a possible reason herefore could be the similar weak variation of the ratio @xmath215 three - body phase space over flux factor : @xmath216 .
we have also considered the interference term of the bremsstrahlung diagrams in fig.7 with the @xmath217-production from @xmath218-decay .
the amplitude for the latter process involves the large scale factor @xmath219mev to the fourth power in the denominator . as a consequence of that suppression factor , the integrated interference cross sections ( taking e.g. @xmath220 turn out to be extremely small . , @xmath221 , and @xmath222 .
the latter ones have been obtained as infrared - regularized integrals @xmath223 with @xmath224 and @xmath225 . ]
moreover , we show in fig.9 ( on a logarithmic scale ) the total cross sections for the three final states @xmath0 ( full curve ) , @xmath221 ( dashed curve ) , and @xmath226 ( dashed - dotted curves ) , altogether . in the latter case these
have been obtained ( in infrared - regularized form ) by integrating the two - photon mass spectrum @xmath210 over the interval @xmath227 and @xmath228 , respectively . with the exception of the region @xmath229
very near to the @xmath218-production threshold , the non - resonant two - photon radiation amounts to at most a few percent correction .
as already mentioned , the two - photon mass spectrum @xmath230 of the reaction @xmath9 possesses an infrared singularity of the form : @xmath231 in the limit @xmath232 . in practice
the leading @xmath233 term provides a good description of the two - photon mass spectrum up to @xmath234 . after an elaborate calculation we obtain the following analytical expression for the energy dependent function @xmath235 associated with the infrared singular part : @xmath236\big[x ( \hat s -1)^2 + 2 \hat s\big ] \over ( \hat s+x - x \hat s)^2 \sqrt{x(\hat s-1)^2 + 4\hat s } } \nonumber \\ & & \qquad \qquad \qquad \times \ln { \sqrt{x}(\hat s -1)+\sqrt{x ( \hat s -1)^2 + 4\hat s } \over 2 \sqrt{\hat s } } \ , , \end{aligned}\ ] ] where @xmath237 .
the corresponding behavior near threshold is , @xmath238 , and it comes exclusively from the first term in eq.(23 ) . in order to arrive at this result
we have first established in a careful numerical study the 20 non - vanishing contributions to @xmath235 from the interference terms of the diagrams in fig.7 as well as the identities holding between them . as a result , only six independent interference terms needed to be considered and for these the three - dimensional phase space integral over @xmath239 )
could then be solved ( almost ) analytically . in order to bring the integral - term in eq.(23 ) into its compact form we substituted @xmath240 at the end .
fig.10 shows the cross section @xmath235 as a function of the total center - of - mass energy @xmath121 .
quite interestingly , it reaches a maximum of @xmath241nb at @xmath242 and beyond that it decreases slowly as @xmath243 .
this behavior is in accordance with the weak @xmath121-dependence of the two - photon mass spectra @xmath210 observed in fig.8 . of the infrared singular part of the two - photon mass spectrum .
] clearly , in order to establish the complete infrared finiteness of the qed radiative corrections to pion compton scattering ( la bloch - nordsieck ) , one has to evaluate and include the corresponding one - photon loop diagrams .
work along this line is in progress @xcite .
we thank s. paul for triggering this work and for informative discussions .
we thank j. gasser for critical reading of the manuscript and for suggesting improvements .
99 g. colangelo , j. gasser , and h. leutwyler , _ nucl .
phys . _ * b603 * , 125 ( 2001 ) ; and refs . therein .
s. pislak et al . , _ phys .
rev . _ * d67 * , 072004 ( 2003 ) .
batley et al . , _ eur .
phys . j. _ *
c54 * , 411 ( 2008 ) . g. colangelo , j. gasser , and h. leutwyler , _ phys .
lett . _ * 86 * , 5008 ( 2001 ) .
b. adeva et al . , _ j. phys . _
* g30 * , 1929 ( 2004 ) .
batley et al . , _ phys .
lett . _ * b633 * , 173 ( 2006 ) .
j. gasser , hep - ph/0710.3048 .
antipov et al . , _ phys .
* b121 * , 445 ( 1983 ) ; _ z. phys . _ * c26 * , 495 ( 1985 ) .
j. ahrens et al .
phys . j. _ * a23 * , 113 ( 2005 ) .
donoghue and b.r .
holstein , _ phys .
* d48 * , 137 ( 1993 ) . j. boyer et al . , _ phys .
rev . _ * d42 * , 1350 ( 1990 ) .
j. gasser , m.a .
ivanov , and m.e .
sainio , _ nucl .
phys . _ * b745 * , 84 ( 2006 ) ; and refs . therein .
filkov and v.l .
kashevarov , hep - ph/0802.0965 .
kao , b.e .
norum , and k. wang , hep - ph/0409081 ; _ few body systems _ * 41 * , 31 ( 2007 ) .
compass collaboration : p. abbon et al .
, hep - ex/0703049 .
antipov et al . , _ phys .
rev . _ * d36 * , 21 ( 1987 ) .
g. fldt , _ phys .
rev . _ * c76 * , 014608 ( 2007 ) .
g. fldt and u. tengblad , _ phys .
_ * c76 * , 064607 ( 2007 ) .
u. brgi , _ nucl .
phys . _ * b479 * , 392 ( 1996 ) ; and refs . therein .
donoghue and b.r .
holstein , _ phys .
rev . _ * d40 * , 2378 ( 1989 ) .
e. frlez et al . , _ phys .
lett . _ * 93 * , 181804 ( 2004 ) ; hep - ex/0606023 .
j. bijnens and p. talavera , _ nucl .
phys . _ * b489 * , 387 ( 1997 ) ; + c.q .
geng , i.l . ho , and t.h .
wu , _ nucl .
_ * b684 * , 2815 ( 2004 ) .
terentev , _ sov .
* 16 * , 87 ( 1973 ) .
t. nagel , tu mnchen , e18 , private communications . l. ametller , m. knecht , and p. talavera , _ phys . rev . _ * d64 * , 094009 ( 2001 )
. t. hannah , _ nucl . phys . _ * b593 * , 577 ( 2001 ) .
j. bijnens , a. bramon , and f. cornet , _ phys .
lett . _ * b237 * , 488 ( 1990 ) ; + j. bijnens , _ int
phys . _ * a8 * , 3045 ( 1993 ) .
particle data group : w.m .
yao et al . , _ j. phys . _
* g 33 * , 1 ( 2006 ) .
n. kaiser and j.m .
friedrich , work in progress .
|
we review the cross sections for low - energy @xmath0 reactions in the framework of chiral perturbation theory .
charged pion compton scattering , @xmath1 , is considered up to one - loop order where the pion s internal structure enters through the difference of the electric and magnetic pion polarizability , @xmath2 .
the ongoing compass experiment aims at measuring this important structure constant with high statistics using the primakoff effect . in the same way
, the reaction @xmath3 serves as a test of the qcd chiral anomaly ( via the @xmath4-coupling constant @xmath5 ) .
furthermore , we calculate the total cross sections for neutral and charged pion - pair production , @xmath6 and @xmath7 , which are governed by the chiral @xmath8-interaction . finally , we investigate the radiative ( correction ) process @xmath9 and calculate the corresponding two - photon mass spectrum .
this information will be useful for analyzing the @xmath10 primakoff scattering events with photons in the final state .
25.cm 17.cm -2.2 cm
-0.6 cm -0.6 cm # 1#2 # 2 cm n. kaiser and j.m .
friedrich + pacs : 12.20.ds , 12.39.fe , 13.60.fz , 13.75.lb
|
collectivity , isospin symmetry and shell structure are generic features of the nuclear many - body quantum system .
collective nuclear valence - shell excitations are a key to understand how these features coexist , interplay and compete . in vibrational nuclei ,
the development of predominantly proton - neutron symmetric collective nuclear structures at low excitation energies is governed by the strong residual proton - neutron interaction .
their existence implies - due to quantum - mechanical orthogonality - the formation of collective states with , at least partial antisymmetry with respect to the contribution of proton and neutron valence - space components to their wave functions .
such excited states are said to have mixed symmetry @xcite .
the investigation of mixed - symmetry states ( mss ) is an important source of information on the effective proton - neutron interaction in the valence shell of heavy atomic nuclei @xcite .
mss have been defined in the framework of the proton - neutron interacting boson model ( ibm-2 ) @xcite . in analogy to the isospin symmetry of nucleons , the symmetry of a multi - boson wave function formed by @xmath14 proton bosons and @xmath15 neutron bosons is quantified by the so - called @xmath16-spin .
states with @xmath17 have wave functions that contain at least one pair of proton and neutron bosons antisymmetric under the exchange of proton and neutron labels . the signatures of @xmath0 mss are ( @xmath18 ) strong @xmath19 transitions to fully symmetric states ( fss ) with matrix elements of about 1@xmath20 and ( @xmath21 ) weakly collective @xmath22 transitions to fss .
the prediction of the ibm-2 with respect to a multi - phonon structure of mss in vibrational nuclei was confirmed about ten years ago by the observation of large @xmath19 transition strengths between low - energy states of @xmath23mo @xcite .
the @xmath24 states were also investigated with electron scattering experiments at the superconducting electron accelerator s - dalinac and with proton scattering at ithemba labs @xcite .
the combined analysis supported a one - phonon structure of the @xmath25 states of @xmath23mo . in the neighboring even - even isotone @xmath1zr with two neutrons
outside the @xmath26 closed shell and with the @xmath27 sub - shell closure , a stronger configurational isospin polarization of the one - phonon states than in @xmath23mo is expected @xcite .
recent work showed that the collectivity of the low - lying symmetric and mixed - symmetric quadrupole excitation in vibrational nuclei originates from the coupling of the giant quadrupole resonance to the dominant valence - space configurations @xcite .
experimental evidence in @xmath1zr and @xmath23mo stems from the observation of a difference of the respective matter - transition radii ( deduced from proton scattering ) while charge - transition radii ( deduced from electron scattering ) were found to be about equal .
the difference results from a sign change of the dominant valence neutron amplitude in mss with respect to the fss .
the present work provides an in - depth study of the electron scattering results on @xmath1zr .
in particular , we discuss a new method for a model - independent determination of the ratio of the @xmath22 transition strengths of fully symmetric and mixed - symmetric one - phonon excitations in heavy vibrational nuclei , which at the same time provides an estimate of the sensitivity to the transition - radius difference between these two states .
the results are furthermore interpreted in the framework of the qpm ( for an introduction to the model see @xcite ) .
the experiment has been carried out at the darmstadt superconducting electron linear accelerator s - dalinac .
the lintott spectrometer was used with a focal - plane detector system based on four single - sided silicon detectors , each providing 96 strips with thickness of 500 @xmath28 m and a pitch of 650 @xmath28 m @xcite .
electrons with an incident beam energy @xmath29 mev and beam currents ranging from 0.5 to 1 @xmath28a impinged on a @xmath1zr target with an isotopic enrichment of 94.57% and an areal thickness of 9.75 mg/@xmath30 .
data were taken at five different scattering angles @xmath31 , @xmath32 , @xmath33 , @xmath34 , and @xmath35 covering roughly the first maximum of an @xmath22 form factor .
examples of electron - scattering spectra are shown in fig .
[ zr92_2spe ] .
the energy resolution was about 55 kev ( full width at half maximum , fwhm ) .
the observed peaks correspond to the elastic line , the collective one - phonon @xmath36 ( @xmath37 mev ) and @xmath38 ( @xmath39 mev ) states , and the one - phonon mss ( @xmath40 , @xmath41 mev ) .
zr(@xmath4 ) reaction at incident electron energy @xmath42 mev and electron scattering angles @xmath43 , @xmath33 and @xmath34.,width=302 ] peak areas @xmath44 of the transitions were obtained from a spectrum decomposition using the line shape described in ref .
absolute differential cross sections were determined from normalization to the elastic scattering peak .
theoretical elastic scattering cross sections were obtained from calculations with the code phashi @xcite using charge density distributions from ref .
the resulting inelastic cross sections with statistical uncertainties are given in tab .
[ tab : ff ] normalized to the mott cross section .
the overall systematic uncertainties of the normalization due to the model description of the charge density and experimental kinematic parameters ( electron energy , scattering angle ) were estimated to 5% , which were added in quadrature .
.cross sections of electroexcitation of the @xmath11 and @xmath45 states in @xmath1zr normalized to the mott cross section in units @xmath46 and the ratio @xmath47 of the @xmath11 kinematical functions defined in eq .
( [ eqn : ratiodr ] ) .
only statistical errors are given .
[ cols="^,^,^,^,^",options="header " , ]
the transition strengths derived from the @xmath48 data depend of course on the applied nuclear structure model . in light nuclei it has been shown that transition strengths can be extracted in a nearly model - independent plane - wave born approximation ( pwba ) analysis @xcite . it assumes that coulomb distortions of the electron wave function can be approximated by an overall correction factor determined from the g.s .
charge distribution of the nucleus . for kinematics where transverse contributions can be neglected ,
the differential cross sections for electric transitions are then given by @xmath49 where @xmath50^{-2}$ ] , @xmath51 , and @xmath52 .
the symbol @xmath53 denotes the fine structure constant , @xmath54 is the transition multipolarity , and @xmath55 is a kinematic function given e.g. in ref .
the quantities b@xmath56 and the reduced transition probabilities b@xmath57 from real - photon experiments can be related by siegert s theorem @xcite @xmath58 the coulomb correction factor @xmath59_{\rm theo}\ ] ] is determined from the ratio of dwba and pwba calculations employing the qpm transition densities .
the reduced transition probabilities can thus be related to the experimental differential cross section by @xmath60^{-1 } \left(\frac{d\sigma}{d\omega}\right)_{e\lambda}\\ \nonumber & \equiv & \left[f_{\rm kin}f_{\rm c}\left(q_{\rm x } , e_{0},e_{x}\right)\right]^{-1 } \left(\frac{d\sigma}{d\omega}\right)_{e\lambda}.\end{aligned}\ ] ] for low momentum transfers , @xmath61 can be expanded in a power series of the momentum transfer @xmath62 here , @xmath63 is defined by @xmath64 with @xmath65 desribing the transition density of multipolarity @xmath54 . in general ,
the pwba approximation is not valid for a heavy nucleus like @xmath1zr .
however , it may hold for the ratio of cross sections populating the @xmath0 fss and mss for the following reasons : ( i ) the kinematics for both transitions are almost identical , and ( ii ) transition densities of collective transitions of a given multipolarity are similar ( see , e.g. , fig . 2 in ref .
@xcite for the cases studied here ) .
figure [ coulombratio ] shows the coulomb - correction factors calculated with the qpm for the transitions to @xmath36 ( middle ) and @xmath40 ( top ) states in @xmath1zr together with their ratio ( bottom ) as a function of @xmath66 .
states in @xmath1zr for an incident electron energy @xmath29 mev.,width=302 ] the ratio is unity to better than 1% over the range of the momentum transfer included in our experiments . consequently , the effects from coulomb distortion can be neglected in a relative analysis , and the extraction of the @xmath10 ratio can be achieved with improved accuracy , since systematic errors in the determination of absolute cross sections cancel .
employing eqs .
( [ eqn : pwba ] ) and ( [ eqn : expansion ] ) and defining the transition radius @xmath67 , the ratio of reduced transition strengths can be approximated by @xmath68 where the indices @xmath69 indicate the transitions to the @xmath70 and @xmath2 state , respectively .
@xmath47 denotes the ratio of kinematic functions @xmath71 , given in tab .
[ tab : ff ] , and @xmath72 the difference of the corresponding charge - transition radii .
the experimental ratio depends on the square root of the ratio of the peak areas @xmath73 only . for the approximation on the r.h.s . of eq .
( [ eqn : ratiodr ] ) use is made of siegert s theorem and of the tassie model @xcite which provides a good description of the surface behaviour of transition densities for collective states .
we have checked that the approximation @xmath74 employed in eq .
( 6 ) within this model , yields very accurate results , and thus it is used in our analysis below .
this approximation may be questionable for the results at the highest @xmath66 in tab .
[ tab : ff ] which also have poor statistics .
the data point is thus omitted in the further analysis .
mss and fss ( solid squares ) of @xmath1zr as a function of the squared elasic momentum transfer @xmath75 . an additonal data point ( full circle ) at @xmath76 stems from the ratio of @xmath10 strengths obtained from @xmath77-decay lifetime measurements @xcite .
the solid line is a fit of eq .
( [ eqn : ratiodr ] ) with @xmath78 error bars given by the dashed lines . , width=302 ] figure [ fitdata ] shows a plot of @xmath79 as a function of the squared elastic momentum transfer .
a fit of eq .
( [ eqn : ratiodr ] ) to the data has 3 parameters , viz . the ratio of @xmath10 strengths , @xmath80 and @xmath81 . in a first step ,
@xmath82 fm is fixed using the qpm results .
a @xmath83-minimization of eq.([eqn : ratiodr ] ) to the data then determines @xmath84 with the @xmath85 value from tab .
[ tab : b ] , we obtain @xmath86 w.u . , in agreement with ref .
@xcite and with the value obtained above from the dwba analysis .
while the ratio can be determined precisely with an uncertainty of about 1% , the accuracy of the absolute value is presently limited by the error of the @xmath87 value in the literature .
the second parameter @xmath81 in eq .
( [ eqn : ratiodr ] ) provides information about the change of the proton transition radii between both @xmath24 states .
this is particular interesting in view of the recent results of walz _
et al_. @xcite providing evidence for a significant difference of the neutron transition radii for these two states , while their proton transition radii are expected to be very close based on qpm calculations .
this may serve as a new experimental signature of mss in vibrational nuclei with a specific shell structure .
an experimental confirmation of this conjecture is of considerable interest .
the fit of eq .
( [ eqn : ratiodr ] ) leads to @xmath88 fm , where the uncertainty is dominated by the limited number of data points with small enough error bars at sufficiently low momentum transfers .
one way to improve the fit is the inclusion of the results of ref .
@xcite providing an additional data point at @xmath89 .
the resulting fit ( red curves ) with @xmath78 error bars ( green curves ) shown in fig .
[ fitdata ] corresponds to @xmath12 fm consistent with equal proton transition radii to about half a fm . and @xmath2 states in @xmath1zr obtained from eq .
( [ eqn : ratiodr ] ) as a function of the transition radius @xmath80 .
the arrow indicates the prediction of the qpm calculation.,width=302 ] finally , we briefly comment on a possible dependence of the result on a variation of the absolute size of the charge transition radius @xmath80 .
we have repeated the analysis for 9 different values of @xmath80 between 4 and 7 fm , thereby overexhausting the range of possible values expected from model calculations and from the phenomenological finding that the transition radii of collective excitations differ not too much from the radius of the nuclear ground state @xcite . as demonstrated in fig .
[ r9value ] , the deduced difference of the charge transition radii is independent of the choice of @xmath80 .
to summarize , an investigation of the nature of one - phonon symmetric and mixed - symmetric @xmath24 states in @xmath1zr has been performed using inelastic electron scattering at low momentum transfers .
a comparison of the measured form factors with qpm calculations confirms the dominant one - phonon structure of the transitions to the @xmath36 and @xmath40 states .
it is shown that a pwba analysis of the form factors , which usually fails for heavy nuclei , can nevertheless be applied to extract the ratio of the g.s .
@xmath90 ) transition strengths in a relative analysis . this is a new promising approach to determine the g.s .
transition strength of the @xmath0 mss in vibrational nuclei with a precision limited only by the experimental information about the @xmath91 strength .
the pwba approach furthermore provides information about differences of the proton transition radii of the respective states , containing independent information about the mixed - symmetry character of @xmath0 states and the sign change of leading valence shell components between fss and mss @xcite . for @xmath1zr ,
the proton transition radii agree within about 0.5 fm , consistent with predictions that the sign change arises in this case from the neutron component .
further analysis of the data indicates that an improved precision for the proton transition radii difference can be achieved by additional data , in particular in momentum transfer ranges presently not covered well ( e.g. , @xmath92
@xmath93 ) .
elementary to the present approach is an approximate cancellation of coulomb corrections of the fss and mss .
this may be questioned when moving away from shell closures , where the collectivity of the mss ground - state decay decreases .
systematic investigations are necessary to establish the range of applicability of this new promising method .
work along these lines ( e.g. studies of @xmath94zr and mo isotopes ) is underway .
|
background : : mixed - symmetry @xmath0 states in vibrational nuclei are characterized by a sign change between dominant proton and neutron valence - shell components with respect to the fully symmetric @xmath0 state .
the sign can be measured by a decomposition of proton and neutron transition radii with a combination of inelastic electron and hadron scattering [ c. walz _ et al .
_ , phys .
rev .
lett . * 106 * , 062501 ( 2011 ) ] . for the case of @xmath1zr ,
a difference could be experimentally established for the neutron components , while about equal proton transition radii were indicated by the data .
purpose : : determination of the ground - state ( g.s . ) transition strength of the mixed - symmetry @xmath2 state and verification of the expected vanishing of the proton transition radii difference between the one - phonon @xmath0 states in @xmath1zr .
method : : differential cross sections for the excitation of one - phonon @xmath0 and @xmath3 states in @xmath1zr have been measured with the ( @xmath4 ) reaction at the s - dalinac in a momentum transfer range @xmath5 @xmath6 .
results : : transition strengths @xmath7 , @xmath8 and @xmath9 weisskopf units are determined from a comparison of the experimental cross sections to quasiparticle - phonon model ( qpm ) calculations . it is shown that a model - independent plane wave born approximation ( pwba ) analysis can fix the ratio of @xmath10 transition strengths to the @xmath11 states with a precision of about 1% .
the method furthermore allows to extract their proton transition radii difference . with the present data @xmath12 fm is obtained .
conclusions : : electron scattering at low momentum transfers can provide information on transition radii differences of one - phonon @xmath0 states even in heavy nuclei .
proton transition radii for the @xmath11 states in @xmath1zr are found to be identical within uncertainties .
the g.s .
transition probability for the mixed - symmetry state can be determined with high precision limited only by the available experimental information on the @xmath13 ) value .
|
active systems are fundamental elements in high precision measurements in particular measurement on quantum systems .
amplifiers are used either for amplifying the signal to a macroscopic level or to make the system work around its optimal working point with the help of feedback loops .
the analysis of sensitivity limits in these devices rises many questions related to fundamental processes as well as experimental constraints .
how does the active feedback act on the quantum system ?
how does the coupling with the environment influence the sensitivity of the measurement ? how are these process related to the fluctuation dissipation theorem ?
how do the experimental constraints interplay with the fundamental limitations of the sensitivity ?
the aim of the present paper is to address the question of measurement on quantum systems with real measurement devices .
we address this problem with quantum network theory .
this approach provides a rigorous thermodynamical framework able to withstand the constraints of a quantum analysis of the measurement . in the same time , it makes possible a realistic description of real measurement devices .
thermodynamic and quantum fluctuations are treated in the same footing .
the measurement process is described as a scattering process allowing for a modular analysis of real quantum systems .
active systems such as the linear amplifier or the ideal operational amplifier are described in this framework . here , the approach will be illustrated by analyzing the sensitivity of a cold damped capacitive accelerometer developed for fundamental physics applications in space @xcite .
we first present the analysis of a quantum measurement with a passive systems in term of quantum networks .
then , we use this approach to present the quantum analysis of measurements with an operational amplifier working in the ideal limit of infinite gain , infinite input impedance and null output impedance and we illustrate the theoretical framework with the example of a cold damped accelerometer .
we conclude with an analysis of chained measurements
the measurement of a physical quantity or signal corresponds to the obtention of a result or readout from a measureing device . in general , this is obtained after a few stages .
many times , the signal is transduced into an electrical signal which is amplified .
after further processing , the resulting voltage drives a displaying or a storing device .
the final result is thus obtained after a succession of stages . as far as measurement at the quantum level is concerned
, the principle is the same : a chain of processes leads from the quantum system to the readout of the meter . as a first insight into a quantum analysis of measurement , we will consider the detection of an electromagnetic wave with an antenna and first restrict our attention to the transduction of the wave into an electrical signal .
as sketched on figure [ antenne ] , a detecting antenna is coupled with an electrical resonator a coaxial line is used to extract an electrical signal from this system .
the further steps as for instance amplification will be considered in the next section .
the the antenna is coupled to the electromagnetic field though one mode obtained as a linear combination of plane waves or spherical waves and corresponding to the radiation diagram of the antenna . in this mode
two counterpropagating fields can be separated an incoming field @xmath0 and an outgoing field @xmath1 .
the incoming electromagnetic wave @xmath2 puts into motion the electrons in the antenna and produces a current flow in the circuit . as a consequence of this current
, a signal @xmath3 with informations on the detected radiation is sent into the coaxial line .
this electrical signal may then be used to feed a further element , for instance an amplifier .
in addition , any incoming signal @xmath4 arriving in the coaxial line also creates a current in the detector and can perturb the measurement .
for example , if the line is at thermal equilibrium , it corresponds to the johnson nyquist noise .
finally , energy and information can be lost because of the emission of radiation by the antenna .
these phenomena , present in every measurement can be synthetized in the following way . in the frequency domain the output fields
can be written as linear combinations of input fields . throughout the paper
, we will consider that a function @xmath5 is defined in the time domain ( notation @xmath6 ) or in the frequency domain ( kubo s notation @xmath7 $ ] ) and that these two representations are related trough the fourier transform with the convention of quantum mechanics @xmath8
e^{-i\omega t}\ ] ] the electronics convention may be recovered by substituting @xmath9 to @xmath10 .
the _ detection and measurement _ is then described as a scattering process : @xmath11 the measurement of the field @xmath12 with the current @xmath13 can be described with an estimator@xmath14@xmath15 the incoming current fluctuations correspond to a noise in the measurement .
it is also possible to express the output field @xmath1@xmath16 @xmath17 coreponds to the back action noise on the measured device .
this kind of description of a measurement can be generalized to more complex systems in the framework of quantum network theory . in this approach ,
the role of the antenna and the line are to couple the measurement device with the outside world for inputs as well as for outputs .
this coupling is associated to _ dissipation and fluctuations _ which are thus inavoidable in measurement processes_. _ dissipation corresponds to the loss of energy by the resonator as emitted radiation by the antenna or as an electrical signal by the line .
fluctuations corresponds to the detection of the thermal and quantum fluctuations of the input fields .
since einstein and its study of the viscous damping of mechanical systems @xcite the link between fluctuations and dissipation has been widely studied .
for example with johnson - nyquist noise in resistive electrical elements @xcite .
in our situation , the energy loss of the oscillator has to be counterbalanced by the detection of thermal radiation by the antenna and the thermal current of the line . in the high temperature limit
, it leads to the usual thermodynamic @xmath18 per degree of freedom , with @xmath19 being boltzmann constant and @xmath20 the radiation temperature . note that as a consequence
, thermodynamics imposes constraints on the scatering matrix resulting from the two principles .
we will analyse precisely these constraints in the general framework of quantum networks .
the classical result on the link between fluctuation and dissiaption was extended to take into account the quantum statistical properties of fluctuations@xcite .
a general approach of these relations was widely studied in the framework of linear response theory @xcite .
in particular , in the zero temperature limit , the detected field corresponds to the vacuum fluctuations of the electromagnetic field and the induced energy of the oscillator is the zero point energy @xmath21 , with @xmath22 the resonance frequency of the oscillator .
the elementary systems described up to now as well as more complex devices to be studied later in this paper may be described by using a systematic approach which may be termed as `` quantum network theory '' .
initially designed as a quantum extension of the classical theory of electrical networks @xcite , this theory was mainly developed through applications to optical systems @xcite .
it has also been viewed as a generalized quantum extension of the linear response theory which is of interest for electrical systems as well @xcite .
it is fruitful for analyzing non - ideal quantum measurements containing active elements @xcite . in this quantum network approach ,
the various fluctuations entering the system , either by dissipative or by active elements , are described as input fields in a number of lines as depicted on [ network ] .
for example , a resistance @xmath23 is modeled as a semi - infinite coaxial line @xmath24 with characteristic impedance @xmath23 . in this line ,
current and voltage may be treated as quantum fields propagating in a two dimensional space - time .
free field operators @xmath25 and @xmath26 can be defined as the fourier components of @xmath27and @xmath13 and related to the current and voltage at the end of the line : @xmath28 they are normalized so that they obey the standard commutation relations @xmath29 , a^{{\rm in}}\left [ \omega ^{\prime } \right ] \right ] & = & \left [ a^{{\rm out}}\left [ \omega \right ] , a^{{\rm out}}% \left [ \omega ^{\prime } \right ] \right ] \nonumber \\ & = & 2\pi \ \delta \left ( \omega + \omega ^{\prime } \right ) \
\varepsilon \left ( \omega \right ) \label{commutfree}\end{aligned}\ ] ] where @xmath30 denotes the sign of the frequency @xmath31 .
this relation just means that the positive and negative frequency components correspond respectively to the annihilation @xmath32 and creation @xmath33 operators of quantum field theory @xmath34 = a_{\omega } \theta \left ( \omega \right ) + a_{-\omega } ^{\dagger } \theta \left ( -\omega \right)\ ] ] @xmath35 denotes the heavyside function .
similarly , mechanical fluctuations as well as electromagnetical fluctuations can be expressed in term of these normalized fields . to characterize the fluctuations of these noncommuting operators , we use the correlation function defined as the average value of the symmetrized product . with stationary noise ,
the correlation function depends only on the time difference @xmath36 \cdot a^{{\rm in}}\left [ \omega ^{\prime } \right ] \right\rangle & = & 2\pi \
\delta \left ( \omega + \omega ^{\prime } \right ) \
\sigma _ { aa}^{{\rm in}}\left [ \omega \right]\end{aligned}\ ] ] the dot symbol denotes a symmetrized product for quantum operators . in the case of a thermal bath ,
the noise spectrum is @xmath37 = \frac{1}{\exp \frac{\hbar \left| \omega \right| } { k_{b}t_{a}}-1}+\frac{1}{2}=\frac{1}{2}\coth \frac{% \hbar \left| \omega \right| } { 2k_{b}t_{a}}\ ] ] one recognizes the black body spectrum or the number of bosons per mode for a field at temperature @xmath38 and a term @xmath39 corresponding to the quantum fluctuations . the energy per mode will be denoted in the following as an effective temperature @xmath40 @xmath41 = \frac{\hbar \left| \omega \right| } { 2}\coth \frac{\hbar \left| \omega \right| } { 2k_{b}t_{a}}\ ] ] in the high temperature limit the classical energy for an harmonic field of @xmath42 per mode is recovered . in the low temperature limit , the energy @xmath43 corresponding to the ground state of a quantum harmonic oscillator is obtained .
note that the term @xmath39 corresponding to the zero point quantum fluctuations was added by planck so that the difference with the classical result @xmath44 tends to zero in the high temperature limit @xcite .
these results are easily translated to obtain the expression of the johnson nyquist noise power @xmath45 = 2r\hbar \left| \omega \right| \sigma _ { aa}^{{\rm in}}\left [ \omega \right ] = 2rk_{b}\theta _ { a } \label{thermal}\ ] ] our symmetric definition of the noise power spectrum leads to a factor @xmath46 difference with the electronic convention where only positive frequencies are considered .
the fields entering and leaving the network can be represented with column vectors @xmath47 with components @xmath48and @xmath49 input fields corresponding to different lines commute with each other . for simplicity
, we also consider that the fields entering through the various ports are uncorrelated with each other . the whole network is then associated with a scattering @xmath50 matrix , also called repartition matrix @xcite , describing the transformation from the input fields to the output ones @xmath51 the output fields @xmath26 are also free fields which obey the same commutation relations [ commutfree ] as the input ones .
in other words , @xmath52 matrix is unitary .
more generally , the unitarity of the @xmath50 matrix is required to ensure the quantum consistency of the description . in the following section
, we will make use of this property to deduce general properties of amplifiers . for passive systems
the unitarity of @xmath50 matrix ensures as well the verification of the two principles of thermodynamics . however , this unitarity is also valid in the presence of amplificating device .
for the analysis of the measurement , two fields can be isolated , the input signal @xmath53 and the output detection field @xmath54 the detection corresponds to figure [ mesure ] .
the detection is performed with the output detection signal @xmath55 .
it is a linear combination of the incoming signal @xmath53 and of input fields in the various noise lines .
we normalize this expression so that the coefficient of proportionality appearing in front of the incoming signal @xmath53 is reduced to unity . with this normalization
, we obtain an estimator @xmath56 which is just the sum of the true signal @xmath53 to be measured and of an equivalent input noise : @xmath57 where @xmath58 denote the various input fields corresponding to the active and passive elements in the detection device . the added noise spectrum @xmath59 is obtained as @xmath60 within this approach , the succession of the different sages of the measurement are studied by using the output field of one stage as the input field of the next one .
we consider in this section the amplifying stage of the measurement . quantum noise associated with linear amplifiers has been the subject of numerous works . in the line of thought initiated by early works on fluctuation - dissipation relations ,
active systems have been studied in the optical domain when maser and laser amplifiers were developed @xcite .
general thermodynamical constraints impose the existence of fluctuations for amplification as well as dissipation processes .
the added noise determines the ultimate performance of linear amplifiers @xcite and plays a key role in the question of optimal information transfer in optical communication systems @xcite .
most practical applications of amplifiers in measurements involve ideal operational amplifiers operating in the limits of infinite gain , infinite input impedance and null output impedance . in order to deal with the pathologies that could arise in such a system , we consider that it operates with a feedback loop which fixes its effective gain and effective impedances @xcite . in the framework of quantum networks
, the noise sources of the amplifier can be modelled with two lines @xmath24 and @xmath61 . as depicted on figure [ ampliop ] , by
coupling two coaxial lines denoted @xmath62 and @xmath63 respectively on the left port and the right port of the amplifier , one realizes a model of amplifying stage .
the left line comes from a transducer so that the inward field @xmath53 plays the role of the signal to be measured .
meanwhile , the right line goes to an electrical meter _ _ so that the outward field @xmath55 plays the role of the meter readout . in connection with the discussions of quantum non demolition measurements
@xcite , @xmath64 appears as the back - action field sent back to the monitored system and @xmath65 represents the fluctuations coming from the readout line .
a reactive impedance @xmath66 acts as feedback for the amplifier .
we now present the electrical equations associated with this measurement device .
we first write the characteristic relations between the voltages and currents @xmath67 here , @xmath68 and @xmath69 are the voltage and current at the port @xmath70 , i.e. at the end of the line @xmath71 or @xmath63 , while @xmath72 and @xmath73 are the voltage and current noise generators associated with the operational amplifier itself ( see fig.1 ) .
@xmath66 is the impedance feedback .
all equations are implicitly written in the frequency representation and the impedances are functions of frequency . equations ( [ amplifier ] ) take a simple form because of the limits of infinite gain , infinite input impedance and null output impedance assumed for the ideal operational amplifier .
we also suppose that the fields incoming through the various ports are uncorrelated with each other as well as with amplifier noises . to obtain compact equations as well as for physical interpretations , we introduce here a voltage noise @xmath72 and a current noise @xmath73 as linear combination of the fields @xmath25 and @xmath74 .
@xmath75 & = & \sqrt{2\hbar \left|
\omega \right| r_{a}}\left ( a^{% { \rm in}}\left [ \omega \right ] -a^{\prime { \rm in}}\left [ -\omega \right ] \right ) \nonumber \\ i\left [ \omega \right ] & = & \sqrt{\frac{2\hbar \left| \omega \right| } { r_{a}}}% \left ( a^{{\rm in}}\left [ \omega \right ] + a^{\prime { \rm in}}\left [ -\omega % \right ] \right ) \label{deflignes}\end{aligned}\ ] ] note the presence of a conjugated field @xmath76 = a^{\prime { \rm in}}\left [ \omega \right ] ^{\dagger } $ ] associated to the amplification process . in the following
, the conjugation of @xmath61 will be assumed . the parameter @xmath77 is determined by the ratio between voltage and current noise spectra of the amplifier @xmath78 we can deduce from ( [ deflignes ] ) that the voltage and current fluctuations @xmath72 and @xmath73 obey the following commutation relations @xmath79 , u\left [ \omega ^{\prime } \right ] \right ] & = & % \left [ i\left [ \omega \right ] , i\left [ \omega ^{\prime } \right ] \right ] = 0 \nonumber \\ \left [ u\left [ \omega \right ] , i\left [ \omega ^{\prime } \right ] \right ] & = & 2\pi \
\hbar \omega \ \delta \left ( \omega + \omega ^{\prime }
\right ) \label{uu}\end{aligned}\ ] ] hence , voltage and current fluctuations verify heisenberg inequalities which determine the ultimate performance of the ideal operational amplifier used as a measurement device @xcite .
the fields @xmath25 and @xmath74 are described by temperatures @xmath38 and @xmath80 .
we have assumed that these fluctuations are the same for all field quadratures , i.e. that the amplifier noises are phase - insensitive .
although this assumption is not mandatory for the forthcoming analysis , we also consider for simplicity that the specific impedance @xmath77 is constant over the spectral domain of interest .
we use the characteristic equations ( [ amplifier],[defnyquist],[deflignes ] ) associated with the amplifier and the lines to write the output fields @xmath64 and @xmath55 in terms of input fields @xmath81 , @xmath65 and of amplifier noise sources @xmath72 and @xmath73 @xmath82 the unitarity of the input output relations for @xmath64 and @xmath83 is ensured by the non commutation of the current and voltage noises .
in order to characterize the performance of the measurement device in terms of added noise , we introduce the estimator @xmath84 of the signal @xmath53 as it may be deduced from the knowledge of the meter readout @xmath55 @xmath85 the estimator @xmath84 would be identical to the measured signal @xmath53 in the absence of added fluctuations .
hence the noise added by the measurement device is described by the supplementary terms assigned respectively to the nyquist noise @xmath65 in the readout as well as nyquist noises @xmath25 and @xmath74 in the two lines representing amplification noises .
notice that proper fluctuations of @xmath86 are included in the signal and not in the added noise .
the whole added noise is characterized by a spectrum obtained as a sum of the uncorrelated noise spectra associated with these nyquist noises @xmath87 the nyquist spectra are given by thermal equilibrium relations ( [ thermal ] ) with temperatures @xmath88 , @xmath89 , @xmath40 .
the optimal sensitivity is obtained by matching the amplifier noise impedance @xmath90 with the impedance of the left line@xmath91 .
we rewrite ( [ estim],[sigma ] ) under the simple forms @xmath92
@xmath93 corresponds to the gain of the amplification for the normalized fields . in the limit of very lare
gain the added noise reads : @xmath94 finally this added noise is still decreased by going to a temperature @xmath95 as low as possible . at the limit of a null temperature , we recover the optimum of @xmath96 added noise which is the same as for phase - insensitive linear amplifiers @xcite . it is also possible analyse the succession of two amplifications
@xmath55 is itself a quantum field it can be send at the iput of tan other amplifier as a field @xmath97and can be measured in a similar manner with a quantum operational amplifier leading to an estimator @xmath98 where the quantitities @xmath99 and @xmath100 correspond to the second detection . as far as @xmath101 is concerned , this is equivalet to a single detection . however ,
if we consider the estimator @xmath84 for the two successive amplification of the signal @xmath53 one obtains : is concerned , the added noise is negligeable thanks to the large amplification of the first detection . as a consequence ,
as soon as only the information on the first system @xmath53 is concerned , the second detection can be considered as noiseless . in these condition
the result @xmath55 of the first measurement can be considered as a classical quantity in the further stage of the signal processing .
@xmath102 as aconsequence , with a large gain of the first amplification the only important noise source is @xmath103
we come to the discussion of the ultimate performance of the cold damped capacitive accelerometer designed for fundamental physics experiments in space @xcite .
the central element of the capacitive accelerometer is a parallelepipedic proof mass placed inside a box .
the walls of these box are electrodes distant from the mass off a hundred micrometers .
the proof mass is kept at the center of the cage by an electrostatic suspension . since
a three dimensional electrostatic suspension is instable , it is necessary to use an active suspension . in the cage reference frame ,
an acceleration is transformed in an inertial force acting on the proof mass .
the force necessary to compensate this inertial force is measured .
in fact , as in most ultrasensitive measurements , the detected signal is the error signal used to compensate the effect of the measured phenomenon .
the essential elements of the accelerometer are presented in figure [ accelero ] .
the proof mass and the cage form two condensators .
any mass motion unbalances the differential detection bridge and provides the error signal . in order to avoid low frequency electrical noise ,
the electrical circuit is polarized with an ac voltage with a frequency of a hundred kilohertz .
after demodulation , this signal is used for detection and as an error signal for a servo control loop which allows to keep the mass centered in its cage .
furthermore , the derivative of this signal provides a force proportional to the mass velocity and simulates a friction force .
this active friction is called cold damping since it may be noiseless .
more precisely , the effective temperature of the fluctuations of this active friction is much lower than the physical temperature of the device .
the detection is performed with the output detection signal @xmath104 .
it is a linear combination of the external force @xmath105 and of input fields in the various noise lines .
we normalize this expression so that the coefficient of proportionality appearing in front of the external force @xmath106 is reduced to unity . with this normalization
, we obtain a force estimator @xmath107 which is just the sum of the true force @xmath106 to be measured and of an equivalent input force noise . in the absence of feedback ,
the force estimator reads @xcite : @xmath108 where @xmath58 denote the various input fields corresponding to the active and passive elements in the accelerometer .
when the feedback is active , the servo loop efficiently maintains the mass at its equilibrium position and the velocity is no longer affected by the external force @xmath105 .
the residual motion is interpreted as the difference between the real velocity of the mass and the velocity measured by the sensor .
this means that the servo loop efficiently corrects the motion of the mass except for the sensing error .
however the sensitivity to external force is still present in the correction signal .
quite remarkably , in the limit of an infinite loop gain and with the same approximations as above , the expression of the force estimator @xmath107 is the same as in the free case @xcite . the added noise spectrum @xmath109 is obtained as @xmath110 we have evaluated the whole noise spectrum @xmath109 for the specific case of the instrument proposed for the @xmath111scope space mission devoted to the test of the equivalence principle .
some of the main parameters of this system are listed below @xmath112 @xmath113 is the mass of the proof mass , @xmath114 is the residual mechanical damping force , @xmath115 is the frequency of the measured mechanical motion , @xmath116 is the operating frequency of the electrical detection circuit . @xmath77 and @xmath40 are the characteristic impedance and temperature of the amplifier . in these conditions , the added noise spectrum is dominated by the mechanical langevin forces @xmath117 this corresponds to a sensitivity in acceleration @xmath118 taking into account the integration time of the experiment , this leads to the expected instrument performance corresponding to a test accuracy of @xmath119 . in the present state - of - the - art instrument
, the sensitivity is thus limited by the residual mechanical langevin forces .
the latter are due to the damping processes in the gold wire used to keep the proof mass at zero voltage @xcite .
with such a configuration , the detection noise is not a limiting factor .
this is a remarkable result in a situation where the effective damping induced through the servo loop is much more efficient than the passive mechanical damping .
this confirms the considerable interest of the cold damping technique for high sensitivity measurement devices .
future fundamental physics missions in space will require even better sensitivities .
to this aim , the wire will be removed and the charge of the test mass will be controlled by other means , for example uv photoemission .
the mechanical langevin noise will no longer be a limitation so that the analysis of the ultimate detection noise will become crucial for the optimization of the instrument performance .
this also means that the electromechanical design configuration will have to be reoptimized taking into account the various noise sources associated with detection @xcite .
in this analysis of chained quantum measurements , the preamplification settles the von neumann problem of determining where the measurement result can be considered as classical . notice that in our analysis of the cold damped accelerometer , this fact is important to treat properly the feedback .
the result of the measurement is used to act on the quantum system .
this action is provided by a quantum signal , however , we have shown that the noise added in the active feedback loop is negligeable compared to the noise of the first stage .
|
we analyse non ideal quantum measurements described as scattering processes providing an estimator of the measured quantity .
the sensitivity is expressed as an equivalent input noise .
we address the von neumann problem of chained measurements and show the crucial role of preamplification .
|
the electronic transport properties in nano - size ferromagnetic structures has been drawing growing interest recently . of most interest would be the magnetoresistance effect , namely , the change of the resistance by application of a magnetic field .
the magnetoresistance in nano - structures is due to the rearrangement of the magnetization , namely the creation and motion of domain walls , driven by the external field .
recently high - sensitivity measurements on magnetic nano - structures such as wires and contacts has become possible and intensive experimental studies of the magnetoresistance due to a single domain wall have begun@xcite @xcite .
these recent experiments have evoked a renewal of interest in theories of the scattering of the electron by a domain wall@xcite @xcite , although the first theoretical study of this problem goes back to more than 25 years ago@xcite .
the scattering mechanism is due to the non - adiabatic deviation of the electron spin from the local magnetization as the electron goes through the wall . in the case of a wall with thickness @xmath0 ( @xmath1 being the fermi wave length )
, the electron spin can adiabatically follow the spatial rotation of the magnetization , resulting in a negligiblly small reflection .
this is the reason the wall contributes very little in the magnetoresistance in the 3@xmath2 metals in the bulk@xcite .
the situation changes in nano - magnets , where the profile of the wall can be different from that in the bulk magnets .
the extreme case would be an atomic size contact .
there the wall is believed to be constrained in a small contact region of an atomic scale due to a rapid change of the contact shape@xcite , and thus the adiabaticity does not hold resulting in a strong scattering@xcite .
indeed an experiment on ni contact revealed a magnetoresistance of about 300% at room temperature@xcite , which is significantly large compared with other giant magnetoresistive materials such as multilayers .
the result has been interpreted as due to a strong electron scattering by a domain wall , which is trapped in the narrow contact region in the absence of the magnetic field , and which is expelled by applying a magnetic field@xcite . in this paper the domain wall resistance in nano - wires or contacts is calculated by use of landauer s formula .
the advantage of this formalism is that we can see how the density of states of the sample and the lead appears in the resistance , which is not possible in he linear response calculation as carried out in ref .
the system is treated as in one dimension and the calculation is at zero temperature .
the transmission coefficients are calculated in section 2 , and the resistance is calculated in the case of four- and two - terminals in section 3 . the result is shown to agree with the result of the mori formula of the linear response theory in the case of the equilibrium flow in a four - terminal case . throughout the paper we consider only channels which contribute to the transport and so do not discuss the quantized behavior of the conductance .
the phase of the transmission coefficient contains important information on the stability of the electron system
. this has been considered elsewhere@xcite .
we consider the exchange interaction between the electron spin and the local spin .
the hamiltonian is @xmath3 where @xmath4 ( @xmath5 being the fermi energy ) .
the spin index is denoted by @xmath6 and @xmath7 is the pauli matrix ( spin indices are suppressed in the second term ) .
the local spin @xmath8 has a spatial dependence of a domain wall . in terms of a polar coordinate @xmath9
the wall is represented as @xmath10 for @xmath11 and @xmath12 for @xmath13 , @xmath14 being the thickness of the wall .
we consider two typical profiles of the wall ; @xmath15 where the first case corresponds to the wall determined by the uniaxial anisotropy energy , which would be the case of bulk magnets , and the second linear profile would be realized in small contacts .
to proceed we carry out a local gauge transformation in the electron spin space so that the @xmath16-axis is chosen to be along the local direction of spin @xmath17@xcite .
the transformation is written as @xmath18 , where @xmath19 is a position - dependent matrix defined as @xmath20 and the electron in the new frame is denoted by @xmath21 .
the hamiltonian is transformed to be @xmath22 where @xmath23 and @xmath24is the energy of uniformly polarised electron with the exchange splitting @xmath25 .
the term @xmath26 describes the interaction between the electron and the wall and is given by @xcite @xmath27 , \ ] ] where @xmath28 is a domain - wall form factor .
note that @xmath29 for @xmath30 and so the interaction is localized near the wall .
the spin - conserving process in the rotated frame as the electron goes through the wall corresponds to the adiabatic process , where the electron spin follows the rotation of the magnetization , and the spin flip corresponds to the deviation from the adiabaticity .
the interaction @xmath26 leads to a scattering of the electron and thus contributes to a resistivity due to the wall as has been discussed before based on a linear response theory@xcite .
here we take a different approach of calculating the @xmath31-matrix perturbatively with respect to @xmath26 .
the solution of the schrdinger equation with the energy of @xmath32 , which we denote by @xmath33 ( two component vector corresponding to two spin state ) , is given as @xmath34 where @xmath35 is the incoming wave , @xmath36 is the free green function with energy @xmath32 and @xmath31-operator is defined as @xmath37 .
the small imaginary part in @xmath38 is to choose the correct boundary condition of outgoing wave .
we consider the incoming wave with energy @xmath32 and spin @xmath39 .
they are explicitly written as @xmath40 where @xmath41 is the wave vector of the spin @xmath39 component .
the spin @xmath42 component of the full wave function for the incident wave @xmath43 is then obtained from eq .
( [ psidef ] ) as @xmath44 where @xmath45 and @xmath46 .
the transmission and reflection coefficients are defined from the asymptotic behavior as @xmath47 the result is @xmath48 the diagonal ( in spin ) component of the @xmath31-matrix for the incoming wave vector of @xmath49 is calculated as @xmath50+o(a^4 ) .
\label{tdiag}\end{aligned}\ ] ] off - diagonal components are given by @xmath51 the result of the transmission and reflection coefficients are @xmath52 the scattering probabilities are defined as @xmath53 and @xmath54 .
these probabilities for the electron at fermi level ( i.e. , @xmath55 ) are obtained to the second order in @xmath56 as @xmath57 and @xmath58 , where @xmath59 and @xmath60 describes a backward and forward scattering with spin flip , respectively ( @xmath61 and @xmath62 ) .
the conservation of the current , @xmath63 is satisfied .
to relate these quantities to resistance , we consider a four - terminal case where the system is coupled with two leads , @xmath64 and @xmath65 and two terminals , @xmath66 and @xmath67 ( fig .
[ figterminal ] ) .
( the gauge transformation does not affect the resistance , since the shift of the current due to the transformation vanishes far away from the wall . )
@xmath68 and @xmath65 are connected with a reserver ( current source ) and @xmath66 and @xmath67 are to measure the voltage difference across the sample .
the argument goes similarly to the non - magnetic case described in refs . @xcite but taking account of spin dependence .
the chemical potential of @xmath68 and @xmath65 are fixed to be @xmath69 and @xmath70 ( @xmath71 is the applied voltage ) , respectively .
the chemical potential of @xmath66 and @xmath67 , @xmath72 and @xmath73 are determined below by the condition of vanishing current flow through each terminal .
the true voltage difference across the sample is different from @xmath71 and is given as @xmath74 .
the resistance of the system is written as @xmath75 , @xmath76 being the current through the sample .
we first consider the case where the leads are ferromagnets with density of states for the spin state @xmath39 given by @xmath77 .
the case of non - magnetic leads is discussed at the end of this section .
the number of the electron with spin @xmath39 injected from @xmath68 is @xmath78 , and thus the electric current is written by use of transmission probability as @xmath79 the voltage @xmath72 and @xmath73 are determined by the balance between the terminal and lead . in @xmath68 , the density ( per energy ) of spin @xmath39 electron is written as @xmath80 here a factor of @xmath81 comes from choosing right moving electron .
the density at the other lead is given by @xmath82 the condition of vanishing flow between @xmath68 and @xmath66 and @xmath83 and @xmath67 are given as @xmath84 respectively .
this leads to the voltage difference across the sample of @xmath85 by use of eq .
( [ trresults ] ) the resistance to the lowest order in domain wall scattering is obtained as @xmath86 , \label{r1}\ ] ] where @xmath87 .
naively it appears correct to choose @xmath88 , equal to the one - dimensional density of state .
this is wrong if the equilibrium situation with a steady current flow is considered .
in fact , the electron with down spin , which has a smaller velocity of @xmath89 , can not leave the scattering region quickly after the scattering and thus the density becomes higher by a factor proportional to @xmath90 ( see fig .
[ figdensity ] ) . therefore under a steady flow ,
the effective density of states becomes proportional to @xmath91 , namely , @xmath92 , where a prefactor is chosen to conserve the total density , i.e. , @xmath93 .
this point has not been discussed in the multi - channel case carried out so far@xcite . then @xmath94 and we obtain @xmath95 .
\label{req}\ ] ] this result is the same as that obtained based on a linear response theory@xcite , where the equilibrium is assumed . in the case of two - terminals ,
the system is out of equilibrium and the resistance is modified .
the conductance in this case is given as the transmitted current , eq .
( [ ieq ] ) with @xmath96 , divided by the applied voltage , @xmath97 .
thus @xmath98 the wall resistance is written as @xmath99 , where @xmath100 is the conductance without the wall .
thus to the lowest order in the wall scattering , @xmath101 this resistance contains only backward scattering @xmath102 .
this result is obtained from the four terminal case , eq .
( [ r1 ] ) , by putting @xmath96 .
two resistances , ( [ req ] ) and ( [ rne ] ) ( dashed and solid lines , respectively ) , are plotted in fig .
[ figr ] for the two profiles , ( [ bulkwall ] ) and ( [ dwprofiles ] ) . in the bulk case , ( [ bulkwall ] )
, @xmath103 is given by @xmath104 and so resistance decays rapidly for @xmath105 . in the case of a linear wall , ( [ dwprofiles ] ) ,
decay is slower ( fig .
[ figr ] ) ; @xmath106 this difference in the form factor leads to a slightly different resistance , but strong suppression of the resistance for @xmath107 is universal .
the difference between the two- and four - terminal cases is small if the zeeman splitting is large but becomes manifest when @xmath108 is small .
let us mention the case where the leads connected to the sample are non - magnetic metals . if we write the hopping probability between the lead and the sample as @xmath109 , the balance of the density of states requires @xmath110 , where @xmath111 is the density of states in the lead .
thus the voltage across the sample is determined by the same eqs . as ( [ n1])-([muab ] ) , and so the resistance is equal to that for the ferromagnetic lead discussed above .
the scattering of the electron by a domain wall is studied theoretically . by use of a local gauge transformation ,
the interaction between the electron and the wall is treated perturbatively to the lowest order .
the transmission coefficients are obtained and the resistance was calculated using the laundauer s formula with four - terminals .
the comparison with the result of linear response theory was made .
the author thanks h. matsukawa and y. tokura for stimulating discussion .
he also thanks max planck institut fur mikrostrukturphysik for its hospitality during his stay at the early stage of this work and alexander von humboldt stiftung for financial support .
this work is supported by a grand - in - aid for scientific research from the ministry of education , science , sports and culture .
99 k. hong and n. giordano : euro .
* 36 * ( 1996 ) 147 .
y. otani , k. fukamichi , o. kitakami , y. shimada , b. pannetier , j. p. nozieres , t. matsuda and a. tonomura : proc .
mrs spring meeting , san francisco , 1997 .
u. ruediger , j. y. s. zhang , a. d. kent and s. s. p. parkin : phys .
lett . * 80 * ( 1998 ) 5639 .
u. ruediger , j. yu , l. thomas , s. s. p. parkin , a. d. kent : phys . rev .
* b59 * ( 1999 ) 11914 .
n. garcia , m. munoz and y. w. zhao : phys .
* 82 * ( 1999 ) 2923 .
t. ono , y. ooka , h. miyajima and y. otani : appl .
* 75 * ( 1999 ) 1622 .
g. tatara , y .- w .
zhao , m. muoz and n. garca : phys .
* 83 * ( 1999 ) 2030 .
f. komori and k. nakatsuji : j. phys .
68 * ( 1999 ) 3786 .
y. w. zhao , m. munoz , g. tatara and n. garcia : preprint ( 1999 ) .
m. yamanaka and n. nagaosa : j. phys .
* 65 * ( 1996 ) 3088 .
g. tatara and h. fukuyama : phys .
* 78 * ( 1997 ) 3773 .
p. m. levy and s. zhang , phys .
* 79 * ( 1997 ) 5110 .
y. l .- geller , i. l. aleiner and p. m. goldbart : phys .
* 81 * ( 1998 ) 3215 .
j. b. a. n. van hoof , k. m. schep , a. brataas , g. bauer and p. j. kelly : phys . rev .
* b59 * ( 1999 ) 138 .
a. brataas , g. tatara and g. bauer : phys
* b60 * ( 1999 ) 3406 .
h. imamura , n. kobayashi , s. takahashi and s. maekawa : phys .
* 84 * ( 2000 ) 1003 .
k. nakanishi and y. nakamura : phys . rev .
* b61 * ( 2000 ) 11278 .
g. g. cabrera and l. m. falicov : phys .
( b)*61 * ( 1974 ) 539 .
bruno : phys .
* 83 * ( 1999 ) 2425 .
g. tatara and y. tokura , preprint . g. tatara and h. fukuyama : phys . rev
* 72 * ( 1994 ) 772 ; j. phys .
soc . jpn .
* 63 * ( 1994 ) 2538 .
m. buttiker , y. imry , .
landauer and s. pinhas : phys . rev .
* b31 * ( 1985 ) 6207 . y. imry , `` introduction to mesoscopic physics '' , oxford univ . press ( 1997 ) chapter 5 .
|
the scattering of the electron by a domain wall in a nano - wire is calculated perturbatively to the lowest order .
the resistance is calculated by use of landauer s formula .
the result is shown to agree with the result of the linear response theory if the equilibrium is assumed in the four - terminal case .
|
theories defined on non - anti - commutative superspace have been studied extensively during last ten years @xcite .
superspace in such non - anti - commutative theories is a superspace whose fermionic supercoordinates are not anticommutative .
one could construct a field theory in non - anti - commutative superspace in terms of superfields with the star - product where lagrangian is deformed from the original theory by the non - anti - commutative parameters .
recently some renormalisability aspects of the non - anti - commutative field theories have been studies .
it has been shown non - anti - commutative field theories are not power - counting renormalisable ; however it has been discussed that they could be renormalisable if some additional terms have been added to the lagrangian in order to divergences to all orders @xcite-@xcite .
the renormalisability of non - anti - commutative versions of the wess - zumino model has been discussed @xcite , with explicit computations up to two loops @xcite .
the renormalisability of non - anti - commutative gauge theory with @xmath8 supersymmetry has been studied in @xcite .
the authors in @xcite show that the theory is renormalisable to all order of perturbation theory .
the conditions of the renormalisability of non - anticommutative ( nac ) field theories have been studied with explicit computations up one and two loops @xcite-@xcite .
the renormalisability of supersymmetric gauge field theories has been discussed in wz gauge @xcite .
the explicit one loop corrections in component formalism have been done in @xcite-@xcite .
the authors in @xcite have claimed the precise form of the lagrangian is not preserved by renormalisation .
they have shown by explicit calculation that there are problems with assumption of gauge invariance which is required to rule out some classes of divergent structure in non - anti - commutative theory.from their calculation , one can see even at one loop divergent non - gauge - invariant terms are generated . in order to remove the non - gauge - invariant terms and restore gauge invariance at one loops they introduce a one loop divergent field redefinition in the case of pure @xmath8 supersymmetry ( i.e. no chiral matter ) .
on the other hand , the authors in @xcite have started from superspace formalism and discussed renormalisability and supergauge invariance .
they proved that the field redefinition is not necessary and the original effective action is not only gauge but also supergauge invariant up one loop corrections .
the disagreement between two approaches put a big question mark which approaches we should relay on in @xmath0 supersymmetric gauge theory . in this paper
we investigate the renormalisability of @xmath9 supersymmetric pure gauge theory at one - loop perturbative corrections in component formalism .
we shall show @xmath8 supersymmetric gauge theory is renormalisable in a usual manner without any needs for field redefinition ( there is not theoretical justification or interpretation for the field redefinition as mentioned by authors @xcite ) which leads to the lagrangian change .
therefore we shall prove two approaches lead to the same conclusion .
the paper is organized as follows : first we briefly review nac supersymmetric gauge theories and their lagrangian
. then an explicit one - loop calculation of the three and four - point functions of the theory in the c - deformed sector is carried out to calculate the corrections .
we show some anomaly terms appears in the 1pi functions which spoil the renormalisability of theory .
next we introduce extra terms to the original lagrangian in order to renormalise nac pure gauge supersymmetric theory and calculate corrections which come from these new terms .
finally we discus the source of the extra lagrangian , and show that these new terms have hidden because of the component @xmath3 redefinition @xcite ) , so in order to reproduce them one should reverse gaugino field redefinition .
the original non - anticommutative theory defined in superfields appears to require a u(n ) gauge group . here , at
first we would like to consider u(n ) gauge theory for non-(anti)commutative ( nac ) superspace .
the action for an n = 1/2 supersymmetric u(n ) pure gauge theory is given by : @xmath10,\end{aligned}\ ] ] where @xmath11,\nonumber\\ d_\mu\lambda&=&\partial_\mu\lambda+ig[a_\mu , \lambda],\end{aligned}\ ] ] and @xmath12 corresponding to any index @xmath13 for su(n ) we introduce the index @xmath14 , so that @xmath15 runs from 0 to @xmath16 . with @xmath17 being the group matrices for u(n ) in the fundamental representation .
these satisfy @xmath18&=&if^{abc}r^c , \
\{r^a , r^b\}=d^{abc}r^c,\end{aligned}\ ] ] where @xmath19 is completely antisymmetric , @xmath20 is the same as for su(n ) and @xmath21 , while @xmath22 is completely symmetric ; @xmath23 is the same as for su(n ) , @xmath24 and @xmath25 .
in particular , @xmath26 .
we have also @xmath27 the following identities hold in u(n ) group and will be extensively used below @xmath28 where @xmath29 , @xmath30 . upon substituting the above relations in eq .
( 1 ) , we obtain the action in the u(n ) case in the form : @xmath31\nonumber\\,\end{aligned}\ ] ] with gauge coupling @xmath32 , gauge field @xmath33 and gaugino @xmath3 . beside , definition for @xmath34 and @xmath35
are given by : @xmath36 @xmath37 is related to the non - anti - commutativity parameter @xmath38 by : @xmath39 also , we have : @xmath40 where @xmath41 the useful identity is : @xmath42 there are some properties of c in app a. in above eqs . , @xmath38 is the non - anti - commutativity parameter , and our conventions are consistent with ref @xcite .
the action for pure @xmath43 supersymmetric gauge theory ( eq . 11 ) is invariant under the standard @xmath44 gauge transformations and the @xmath45 supersymmetry transformations .
the standard @xmath44 version of the nac gauge theory is not renormalisable @xcite .
therefore ; we would like to present a @xmath1 lagrangian which has @xmath0 supersymmetric properties , so we introduce the following action : @xmath46\nonumber\\,\end{aligned}\ ] ] one beauty of the above equation is one could easily switch between the original @xmath44 theory and @xmath47 theory . in our work
we define @xmath48 .
indeed we give them in terms of @xmath32 and @xmath49 .
they are given by : @xmath50 where @xmath51 .
the above action is similar to the @xmath1 action in ref @xcite .
the n=1/2 supersymmetry transformation is : @xmath52,\ \delta { \bar\lambda_{\dot\alpha}}^a=0~,\\ & & \delta d^a=-\epsilon\sigma^\mu d_\mu { \bar\lambda}^a.\end{aligned}\ ] ]
in our calculation , we use standard gauge fixing term @xmath53 and consider the feynman rules in the super - fermi - feynman gauge@xmath54 . in this section
we first review the one - loop perturbative corrections to the undeformed @xmath55 part of the theory .
it has been shown that the quantum corrections of @xmath55 part of the theory are not affected by @xmath2-deformation @xcite . therefore ;
gauge field and gaugino anomalous dimensions and gauge @xmath56-functions are the same as those in the ordinary @xmath55 case .
the @xmath2-independent part of the bare action can be written as : @xmath57\end{aligned}\ ] ] the @xmath2-independent part of the action is renormalisable if one introduce bare fields and couplings according to : @xmath58 that @xmath59 and @xmath60 are known as a renormalisation constants .
also one can define : @xmath61 finally , one should add the following counter terms to the lagrangian of theory in order to renormalise theory : @xmath62 where , @xmath63 and @xmath64 is given by : @xmath65 here @xmath66 is the regulator .
( we have given here the renormalisation constants corresponding to the @xmath67 sector of the @xmath44 theory ; those for the @xmath68 sector , namely @xmath69 are given by ommiting the terms in @xmath70 and replacing @xmath32 by @xmath49 . )
in this part we will present on - loop graphs contributing to the new terms arising from @xmath2-deformed part of the action .
the one - loop one - particle - irreducible ( 1pi ) graphs of the @xmath2-deform @xmath71 three point functions are depicted in figs(2 ) .
using feynman rules one could compute the divergent contributions from the graphs . as an example we calculate the one loop corrections to fig(2-a ) .
it is given by : the diagram ( 2-a).,width=302 ] @xmath72 using feynman parameter in app .
b : @xmath73 ^ 3},\end{aligned}\ ] ] we can simplify denominator of eq . (
31 ) @xmath74+p_1 ^ 2x(1-y)+p_2 ^ 2(1-x)]^3}.\end{aligned}\ ] ] by changing variables to @xmath75 the denominator of integral in eq .
( 33 ) is given by : @xmath76 ^ 3}\ ] ] where @xmath77 ^ 2\ ] ] so , the integral of eq .
( 32 ) is given by : @xmath78 ^ 3}\end{aligned}\ ] ] then we arrive : @xmath79 ^ 3 } = \frac{-ig_{k\lambda}}{32\pi^2\varepsilon}\end{aligned}\ ] ] we finally have for eq .
( 31 ) : @xmath80 moreover , as it be seen in fig .
( 2-a ) , we have momentum - energy conserving in the loop : @xmath81 the divergent contributions up to one loop correction to diagrams in fig .
2 are given by : @xmath82\nonumber\\&&+one\ permutation\nonumber\\ \gamma_{2-c}^{(1)\mu\dot\alpha\dot\beta}&=&\frac{1}{4}inl\gamma^{abc}d^{abc}c^ad^bc^c[\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}(4p_1 + 5p_2)_\nu -\frac{2}{3}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}(2p_1 + 7p_2)_\nu]\nonumber\\&&+one\ permutation\nonumber\\ \gamma_{2-d}^{(1)\mu\dot\alpha\dot\beta}&=&-3inl\gamma^{abc}d^{abc}c^a\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}q_\nu\nonumber\\ \gamma_{2-e}^{(1)\mu\dot\alpha\dot\beta}&=&-\frac{1}{2}inl\gamma^{abc}d^{abc}c^ad^bc^c[\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu } -2{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]p_{2\nu}\nonumber\\&&+one\ permutation\end{aligned}\ ] ] @xmath83is been defined , @xmath84 where tensor @xmath83 is symmetric respect to both lorentz and spinor indices and tensor @xmath37 is anti - symmetric . in our computation permutations has taken into account by changing the position of @xmath2 as well as symmetry factors . adding the different divergent contributions from the diagrams of fig 2 corresponding to different @xmath68 and @xmath67 parts
, we have : @xmath85 let us now continue with the relevant diagrams containing only @xmath2-deformed vertex and contributing to the four point functions ( fig . 3 and fig . 4 ) .
using the feynman rules , and considering all permutations between the same fields , the final result for 1pi graphs of fig .
3 are given by : @xmath86d^ac^bc^cc^d\nonumber\\&&[\frac{1}{4}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}-\frac{1}{6}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]+three\ permutations\nonumber\\ \gamma_{3-b}^{(1)\mu\nu\dot\alpha\dot\beta}&=&nlg[\gamma^{eac}d^{ace}f^{bde}+\gamma^{bde}f^{ace}d^{bde } ] d^ac^bc^cc^d\nonumber\\&&[\frac{1}{8}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}-\frac{1}{4}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]+one\ permutations\nonumber\\ \gamma_{3-c}^{(1)\mu\nu\dot\alpha\dot\beta}&=&2nlg\gamma^{eab}d^{abe}f^{cde}c^ac^b\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}\nonumber\\ \gamma_{3-d}^{(1)\mu\nu\dot\alpha\dot\beta}&=&nlg[\gamma^{ead}d^{ade}f^{bce}+\gamma^{bce}f^{ade}d^{bce } ] d^ac^bc^cc^d\nonumber\\&&[\frac{1}{8}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}-\frac{5}{12}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]+one\ permutations\nonumber\\ \gamma_{3-e}^{(1)\mu\nu\dot\alpha\dot\beta}&=&-nlg[\gamma^{eab}d^{abe}f^{cde}-\gamma^{cde}f^{abe}d^{cde } ] d^ac^bc^cc^d\nonumber\\&&[\frac{1}{4}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}+\frac{7}{12}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]+three\ permutations\nonumber\\ \gamma_{3-f}^{(1)\mu\nu\dot\alpha\dot\beta}&=&nlg[\gamma^{eab}d^{abe}f^{cde}+\gamma^{cde}f^{abe}d^{cde } ] d^ac^bc^cc^d\nonumber\\&&[\frac{3}{16}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}-\frac{3}{8}{(y^{\mu\nu})}^{\dot\alpha\dot\beta}]+one\ permutations\nonumber\\ \gamma_{3-g}^{(1)\mu\nu\dot\alpha\dot\beta}&=&nlg[\frac{3}{8}\gamma^{eab}d^{abe}f^{cde}d^ac^b\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}+\frac{1}{4}\gamma^{cde}f^{abe}d^{cde}c^cc^d{(y^{\mu\nu})}^{\dot\alpha\dot\beta } ] \nonumber\\&&+one\ permutations\nonumber\\ \gamma_{3-h}^{(1)\mu\nu\dot\alpha\dot\beta}&=&-\frac{3}{2}nlg\gamma^{eab}d^{abe}f^{cde}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}\nonumber\\ \gamma_{3-i}^{(1)\mu\nu\dot\alpha\dot\beta}&=&\frac{3}{4}nlg\gamma^{eab}d^{abe}f^{cde}\epsilon^{\dot\alpha\dot\beta}c^{\mu\nu}\end{aligned}\ ] ] considering all diagrams the final result for fig .
3 is given by : @xmath87 finally , the final divergence contributions ( fig .
4 ) which come from the last term containing @xmath88 are given by : @xmath89\nonumber\\ \gamma_{4-b}^{(1)\dot\alpha\dot\beta\dot\delta\dot\gamma}&=&il\epsilon^{\dot\alpha\dot\beta}\epsilon^{\dot\delta\dot\gamma } \mid c\mid^2[\frac{n}{2}\gamma_0d^{abe}d^{cde}-2\gamma_2]\nonumber\\ \gamma_{4-c}^{(1)\dot\alpha\dot\beta\dot\delta\dot\gamma}&=&il\epsilon^{\dot\alpha\dot\beta}\epsilon^{\dot\delta\dot\gamma } \mid c\mid^2[-\frac{n}{2}\gamma_0d^{abe}d^{cde}+2\gamma_2]\nonumber\\ \gamma_{4-d}^{(1)\dot\alpha\dot\beta\dot\delta\dot\gamma}&=&il\epsilon^{\dot\alpha\dot\beta}\epsilon^{\dot\delta\dot\gamma } \mid c\mid^2[-2\gamma_0 d^{abcd}+\frac{1}{3}(\tilde d^{abcd}-\tilde d^{acdb})+3\gamma_2]\end{aligned}\ ] ] note that we have considered all permutations between the same fields and changing the position of @xmath2 and @xmath90 , and adds all divergences come from fig .
4 . the final result for 1pi graphs of fig .
4 is given by : @xmath91 in order to renormalise the theory and remove the divergences one should rescale and redefine coupling constants .
this procedure is equivalent to introduce of some counter - terms to the lagrangian .
therefore , we introduce bare couplings according to : @xmath92 however in the language of counter - terms , a theory is renormalisable if the counter - terms are of the same form as those appearing in the original lagrangian ( these counter terms are required to cancel the divergences ) . if we look at three and four point functions we see that the anomaly term @xmath93 called @xmath94 term is problematic because we can not add some kinds of counter - terms which cancel @xmath94 term , or one can say these terms spoil renormalisation . in the next section we add some extra lagrangian to the theory and prove that the nac pure gauge theory is renormalisable .
in this section we shall renormalise the theory and remove the divergences . in order to renormalise the theory
we should add extra term ( @xmath95 ) to the original lagrangian .
the extra lagrangian is considered as follow : @xmath96 the two last terms in eq .
( 46 ) help us to make renormalisable nac @xmath44 gauge theory , the first two terms are needed because of gauge transformation rules .
these terms are absent from the original lagrangian because of @xmath3 redefinition in ref @xcite as we shall explain in next part . adding the above terms to original lagrangian
, the total lagrangian is given by : @xmath97 since we add some terms to original lagrangian , we have to modify the gauge transformation and susy transformation .
it is easy to show that @xmath98 is preserved under following gauge transformation in @xmath44 group : @xmath99 where @xmath100 is considered as arbitrary coupling which depends on @xmath101 values .
the gauge transformation is not canonical because the transformation of @xmath3 depends on the non- anti- commutative parameter @xmath2 .
however ; the susy transformation does not change except @xmath102\ ] ] so yet we can refer to eqs .
( 21 - 23 ) . in our work in order to renormalise the nac @xmath1 gauge theory , we choose @xmath103 here @xmath104 is considered as a coefficient .
then the extra lagrangian is given by : @xmath105 where @xmath106 in according to eq.(46 ) , the @xmath94 terms leads to new interactions hence , we have to consider new vertices in 1pi graphs .
it means we display these interactions that have been hidden .
so , we should calculate 1pi diagrams considering new vertices or we should modify vertices .
finally , we find that theory is renormalisable . in this case
the new action is given by : @xmath107.\end{aligned}\ ] ] we have to calculate new ( 1pi ) diagrams contributing to the new terms(those containing both parameter @xmath2 and @xmath94 ) which are depicted in figs . 5 and 6 . the results for the new graphs contributing to the new interaction terms in eq .
( 53 ) are the same as the @xmath2 terms so we shall not give detailed results .
for example in order to calculate fig .
5-a , we should only change the nac vertex @xmath108 in eq .
( 31 ) to @xmath109 .
the result for the graphs in fig .
5-a is given by : @xmath110 beside , the total divergent contribution for new graphs in fig .
5 is given by : @xmath111 in order to obtain divergent contribution for 1pi graphs with both @xmath2 and @xmath94 parameters , one should add results of eq .
( 39 ) and eq .
( 56 ) @xmath112 there are new graphs(fig .
6 ) for one loop corrections of the four point function .
the total contributions as different @xmath1 parts corresponds to fig .
5 is given by : @xmath113 then , the total four point 1pi divergent contribution is given by : @xmath114 fortunately , new terms does not effect on four point function is containing @xmath88 . in order to compute counter terms
we should decompose the lagrangian to the @xmath115 parts because some interaction terms such as term which correspond to @xmath116 receive no quantum corrections .
it is given by : @xmath117 the @xmath2 and @xmath94-dependent part of the action are renormalisable if one introduces bare fields and couplings according to eqs .
( 26 ) and eqs .
( 44 , 45 ) as well as : @xmath118 then in order to find some counter terms we have to introduce the following identities : @xmath119 adding the following counter - term terms to the part of the total action and comparing the expression with the bare action , the theory should be renormalisable .
the full @xmath2 dependent part of the action can be written as : @xmath120.\end{aligned}\ ] ] where in order to renormalise the @xmath2-dependent part of the action we have to obtain the values of @xmath121 by solving the following equations : @xmath122 where @xmath123s come from counter terms ( appendix e ) . then , using @xmath124 for @xmath67 sector , and @xmath125 for @xmath68 sector , we obtain the renormalisation constants as : @xmath126 as it is understood because @xmath104 is renormalised , so @xmath100 should be renormalised
. moreover we obtain @xmath127 it is found that our result is compatible with ref @xcite .
of course a natural expectation would be that @xmath128 , because of eq .
( 36 ) we know that @xmath129 .
we demonstrate that the theory is renormalisable and
the n = 1/2 supersymmetric as well as nac @xmath1 pure gauge theory is preserved .
this point also has been concluded in ref @xcite and suggested in ref @xcite , where was supposed to be correct to all orders .
we also arrive at the conclusion that it is not needed to renormalise the non - anticommutativity parameter @xmath2 . beside our full lagrangian
is the same form as derived from non - anticommutative superspace , however @xmath130 and @xmath131 depends on whether @xmath132 are @xmath133 indices or @xmath68 indices .
it seems to imply that the renormalised theory is not @xmath134 non - anticommutative theory any more . because the @xmath134 structure is broken by renormalisation . in order to clarify @xmath95 we
redefine the component @xmath3 as @xmath135 then , put it in eq .
( 18 ) , and obtain eq .
( 55)(in other words @xmath98 is result of @xmath3 redefinition in @xmath136).the @xmath3 redefinition just affects the gaugino kinetic term .
our redefinition is opposite to refs @xcite .
they have redefined @xmath3 in order to make gauge transformations be canonical ; however it causes theory unrenormalisable because in that case some terms are been hidden in the lagrangian . in order to reverse process we should add hidden terms by hand or come back to original definition of @xmath3 eq .
( 71 ) ; however we lose gauge canonical transformation . beside , because @xmath100 is obtained renormalised , the redefinition of @xmath137 is called @xmath138 should be renormalised .
finally divergent field redefinition in ref @xcite is reinterpreted as renormalised @xmath138 .
our results show it is not needed to deform the classical action if one do not use the field redefinition of ref @xcite .
it is worthwhile to investigate our results in the case of limit @xmath139 . indeed in this case
, 1pi graphs from new terms are vanished .
we would like to present this claim as follow .
instance we know that @xmath140 comes from extra lagrangian
. then total divergent contribution for three point function would be finite if we add some counter terms as follow : @xmath141 we expect to obtain @xmath142 ( it means the nac structure would be preserved under the procedure of renormalisation ) .
then , the above equation leads to : @xmath143 note that second terms in eqs .
( 73 ) and ( 78 ) are related to divergent contribution from new 1pi graphs . in order to find @xmath144 we have to benefit from eqs .
hence @xmath144 up to one order is given by : @xmath145 now , in the case of @xmath139 eqs .
( 73 - 79 ) result to @xmath146 in fact we have neglected the divergent contributions from new graphs , however ; @xmath144 has not been changed .
this event could be checked for four point function as well . in this case
we obtain the results of ref @xcite , but because @xmath147 it easy to show it is not necessary to define a nonlinear divergence field redefinition .
if we consider @xmath148 then the variation of @xmath137 is written by @xmath149 consequently , after the procedure of renormalization , the nonlinear divergent field redefinition in ref @xcite automatically is generated : @xmath150
we have compute 1pi corrections for the pure @xmath0 supersymmetric @xmath151 gauge theory at one loop order .
we have proved the theory is renormalisable up one loop order using a standard way of renormalisation if one adds some new terms to the original lagrangian .
we have shown it is possible to interrupt these are hidden terms because of @xmath3 redefinition .
it is worth to investigate if it is possible to show that the problems which arise in renormalisation of @xmath0 supersymmetric theories comes from the redefind vector superfield .
we have shown there is not need to define divergent redefinition of @xmath3 .
moreover we suggest all works which have been done based on divergent field redefinition should be reviewed .
we have used the @xmath0 @xmath44 gauge group action because as discussed in @xcite , just non - anticommutative theory with @xmath44 gauge group is well - defined .
moreover it is worth to investigate the renormalization of theory at higher loops or including chiral matter in the standard form of renormalisation method .
we guess the problem of renormalisation of non - anticommutative theory at component formalism is because of @xmath3 redefinition in @xcite .
it is a pleasure to thank f. yagi for helpful discussions and n. ghasempour for checking calculations .
we have found the new properties for c and y parameters that we have made frequent use in our calculations .
@xmath37 is related to the non - anti - commutativity parameter @xmath38 by : @xmath39 also , we have : @xmath40 where @xmath152 we have used the following notations : @xmath153 where in last equation the symbol of t is used for transposed .
also for y parameter we have : @xmath154\mid c\mid^2 ~,\\ & & tr[~\mid y\mid^2~]=-\frac{3}{2}\mid c\mid^2~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\theta}\sigma_{\nu\gamma\dot\gamma}=\frac{1}{2}[-2c^\alpha ~_\theta{\bar\sigma}^{\mu\dot\alpha\theta}\epsilon_{\alpha\gamma}\epsilon_{\dot\theta\dot\gamma}+c_\gamma~ ^\theta{\sigma^r}_{\theta\dot\gamma}\delta^{\dot\alpha}_{\dot\theta}]~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\theta}{\bar\sigma}_\nu ~^{\dot\gamma\gamma}=\frac{1}{2}c^\gamma ~_\theta[-2{\bar\sigma}^{\mu\dot\alpha\theta}\delta^{\dot\gamma}_{\dot\theta}+{\bar\sigma}^{\mu\dot\gamma\theta}\delta^{\dot\alpha}_{\dot\theta}]~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\theta}{\bar\sigma}_\mu ~^{\dot\gamma\gamma}\sigma_{\nu\delta\dot\delta}=c^\gamma ~_\delta[2\epsilon^{\dot\alpha\dot\gamma}\epsilon_{\dot\theta\dot\delta}+\delta^{\dot\gamma}_{\dot\delta}\delta^{\dot\alpha}_{\dot\theta}]~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\theta}{({\bar\sigma}_{\nu\mu})}^{\dot\beta}_{\dot\delta}=0 ~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\theta}g_{\mu\nu}=0 ~,\\ & & { ( y^{\mu\nu})}^{\dot\alpha}_{\dot\alpha}=0,~ ( y\ is\ traceless).\end{aligned}\ ] ]
@xmath155 ^ 2}\\ \frac{1}{a^nb}&=&n\int_0 ^ 1dx\frac{x^{n-1}}{[(1-x)b+xa]^{n+1}}\\ \frac{1}{abc}&=&2\int_0 ^ 1dx\int_0 ^ 1dy\frac{x}{[axy+bx(1-y)+c(1-x)]^3}\nonumber\\ & = & 2\int_0 ^ 1du\int_0^{1-u}dw\frac{1}{[aw+b(1-u - w)+cu]^3}\\ \frac{1}{a_1a_2 ... \
a_n}&=&(n-1)!\int_0 ^ 1dx_ndx_{n-1 } ...
\ dx_2\nonumber\\ & & \times\frac{x_n^{n-2}x_{n-1}^{n-3} ... \ x_3 ^ 1x_2 ^ 0}{[(1-x_n)a_n + x_n[(1-x_{n-1})a_{n-1}+x_{n-1}[ ... +x_3[(1-x_2)a_2+x_2a_1]] ... ]^n}\nonumber\\\end{aligned}\ ] ]
@xmath156^n}=\frac{(-1)^ni}{(4\pi)^{\frac{d}{2}}}\frac{\gamma(n-\frac{d}{2})}{\gamma(n)}(\frac{1}{\delta})^{n-\frac{d}{2}}\\ & & \int\frac{d^dl}{(2\pi)^d}\frac{l^2}{[l^2-\delta]^n}=\frac{(-1)^{n-1}i}{(4\pi)^{\frac{d}{2}}}\frac{d}{2}\frac{\gamma(n-\frac{d}{2}-1)}{\gamma(n)}(\frac{1}{\delta})^{n-\frac{d}{2}-1}\\ & & \int\frac{d^dl}{(2\pi)^d}\frac{l^\mu l^\nu}{[l^2-\delta]^n}=\frac{(-1)^{n-1}i}{(4\pi)^{\frac{d}{2}}}\frac{g^{\mu\nu}}{2}\frac{\gamma(n-\frac{d}{2}-1)}{\gamma(n)}(\frac{1}{\delta})^{n-\frac{d}{2}-1}\\ & & \int\frac{d^dl}{(2\pi)^d}\frac{(l^2)^2}{[l^2-\delta]^n}=\frac{(-1)^ni}{(4\pi)^{\frac{d}{2}}}\frac{d(d+2)}{4}\frac{\gamma(n-\frac{d}{2}-1)}{\gamma(n)}(\frac{1}{\delta})^{n-\frac{d}{2}-2}\\ & & \int\frac{d^dl}{(2\pi)^d}\frac{l^\mu l^\nu l^\rho l^\sigma}{[l^2-\delta]^n}=\frac{(-1)^ni}{(4\pi)^{\frac{d}{2}}}\frac{\gamma(n-\frac{d}{2}-2)}{\gamma(n)}(\frac{1}{\delta})^{n-\frac{d}{2}-2}\nonumber\\ & & \times\frac{1}{4}(g^{\mu\nu}g^{\rho\sigma}+g^{\mu\rho}g^{\nu\sigma}+g^{\mu\sigma}g^{\nu\rho}).\end{aligned}\ ] ]
here , we collect feynman rules would be used in order to calculate 1pi digrams for current theory .
propagators for each field could be as follow : gauge field @xmath157 : @xmath158 gaugino field @xmath159 : @xmath160 auxiliary boson field @xmath161 : this field could not be propagated .
scalar field @xmath162 : this field could not be propagated .
spinor field @xmath163 : @xmath160 scalar field @xmath164 :
@xmath165 vertices comes from each interaction in the theory so , we have some vertices as follow : three- gauge coupling @xmath166 with momentum @xmath167 respectively : @xmath168,\ ] ] four- gauge coupling @xmath169 : @xmath170@xmath171@xmath172\ ] ] gaugion @xmath173 -gaugino @xmath174 -gauge @xmath175 vertex : @xmath176 nac in gauge @xmath157 with momentum @xmath177 -gaugino @xmath178
-gaugino @xmath179 vertex : @xmath180 nac in gauge @xmath181 -gauge @xmath182 -gaugino @xmath173 -gaugino
@xmath183 vertex : @xmath184 nac in four gauginos @xmath185 vertex:@xmath186 for total lagrangian we can write @xmath187 so , feynman rule would be : * @xmath157 gauge , @xmath178 gaugino , @xmath179 gaugino in nac parameter @xmath2 : @xmath188 * @xmath181 gauge , @xmath189 gauge , @xmath173 gaugino , @xmath183 gaugino in nac parameter @xmath2 : @xmath190
we have made use from the follow counter terms in the procedure of the renormalization : @xmath191
@xmath192d^a c^b c^c c^d,\nonumber\\ & & d^{abcd}=tr[f^af^bd^cd^d]=c^ac^b[\frac{1}{2}c^cc^d(\delta_{ac}\delta_{bd}-\delta_{ab}\delta_{cd}+\delta_{ad}\delta_{bc})\nonumber\\&&+\frac{n}{8}d^cd^d(-f^{abe}f^{cde}-f^{ace}f^{bde}-d^{abe}d^{cde}-d^{ace}d^{bde})],\nonumber\\ & & \tilde d^{abcd}=tr[f^ad^c f^b d^d],\end{aligned}\ ] ] where @xmath29 , @xmath30 .
the gauge index of u(n ) runs a = 0 , 1 , ... ,
@xmath193 where a = 0 corresponds to overall u(1 ) while a = a = 1 , ...
, @xmath193 corresponds to su(n ) . for the simplicity of the calculation
, we introduce the matrix @xmath162 and @xmath161 , whose component is given by @xmath19 and @xmath22 as@xmath194 taking into account that @xmath19 is totally antisymmetric tensor and @xmath22 is totally symmetric tensor .
@xmath195 99 n. seiberg , `` noncommutative superspace , @xmath0 supersymmetry , field theory and string theory '' .
jhep 0306 ( 2003 ) 010 .
t. araki , k. ito and a. ohtsuka , `` supersymmetric gauge theory on noncommutative superspace '' .
lett . b573 ( 2003 ) 209 .
r. britto and b. feng and soo - jong rey , `` non(anti)commutative superspace , @xmath196 mixing and open wilsons lines '' , phys .
91(2003 ) 201610 .
a. romagnoni , `` renormalizability of @xmath0 wess - zumino model in superspace '' , jhep 0310 ( 2003 ) 061 . m. t. grisaru , s. penati , a. romagnoni,``two - loop renormalization for nonanticommutative @xmath197 wess - zumino model , '' jhep 0308,(2003 ) 003 o. lunion and s. - j. rey,``renormalizability of non(anti)commutative gauge theories with n=1/2 supersymmetry '' , jhep 0309 ( 2003 ) 045 .
d. berenstein and s. j. re , phys .
d68 ( 2003 ) 121701 .
s. terashima and j. - t. yee , `` comments on noncommutative superspace '' , jhep 0312 ( 2003 ) 053 .
m. alishahiha , a. ghodsi , n. sadooghi , `` one loop perterbative corrections to non(anti ) commutative parameter of @xmath0 supersymmetric u(n ) gauge theory '' nucl .
b691 ( 2004 ) 111 .
i. jack , d. r. t. jones and l. a. worthy , `` one loop renormalisation of @xmath0 supersymmetry gauge theory '' .
b611 ( 2005 ) 199 .
i. jack , d.r.t .
jones and l.a .
worthy , phys .
d72 ( 2005 ) 065002
. s. penati and a. romagnoni , jhep 0502 ( 2005 ) 064 .
m. t. grisaru , s. penati , a. romagnoni , `` non(anti ) commutative sym theory : renormalization in superspace '' , jhep 0602 : 043 ( 2006 ) .
i. jack , d.r.t .
jones and l.a .
worthy , phys .
d75 ( 2007 ) 045014 .
i. jack , d.r.t .
jones and l.a .
worthy jhep 1101 ( 2011):052 .
michael e. peskin , daniel v. schroeder , `` an introduction to quantum field theory '' , addison - wesley publishing company , 1995 .
marija dimitrijevic and voja radovanovic , jhep04(2009)108 .
jian - wei cui , yue - liang wu , `` one - loop renormalization of non - abelian gauge theory and @xmath56 function based on loop regularization method '' , int.j.mod.phys.a23 : 2861 - 2913 , ( 2008 ) , arxiv : hep - ph/ 0801.2199v2 .
m. buric , v. radovanovic , j. terampetic , `` the one - loop renormalization of the gauge sector in the noncommutative standard model '' , jhep 0703 : 030 , 2007 , arxiv : hep - th/ 0609073v1 . t. araki , k. ito and a. ohtsuka , `` supersymmetric gauge theories on noncommutative superspace , '' phys .
b 573 , 209 ( 2003 ) ; a. ohtsuka `` symmetries in non(anti)commutative supersymmetric gauge theories '' phd thises 2005 .
s. terashima and j. - t. yee , `` comments on noncommutative superspace '' , jhep 0312 ( 2003 ) 053 .
|
we study one loop corrections to @xmath0 supersymmetric @xmath1 pure gauge theory .
we calculate divergent contributions of the 1pi graphs contain the non - anti - commutative parameter @xmath2 up to one loop corrections .
we find the disagreement between component formalism and superspace formalism is because of the field redefinition in component case .
we modify gaugino field redefinition and lagrangian .
we show extra terms of lagrangian have been generated by @xmath3 redefinition and are necessary for the renormalisation of the theory .
finally we prove @xmath4 supersymmetric gauge theory is renormalisable up one loop corrections using standard method of renormalisation # 1@xmath5 .3 in * * .3 in * a. f. kord @xmath6 , m. haddadi moghaddam @xmath7 * .3 in
|
recent achievements in manipulating quantum gases of ultracold atoms have opened up exciting possibilities for the experimental study of many - body quantum effects in low - dimensional systems @xcite .
experimental observation of superfluidity and phase separation in imbalanced fermi atomic gases @xcite have stimulated great interest in exploring exotic quantum phases of matter with two mismatched fermi surfaces .
the pairing of fermionic atoms with mismatched fermi surfaces may lead to a breached pairing phase @xcite and a nonzero momentum pairing phase of fulde - ferrell - larkin - ovchinnikov ( fflo ) states @xcite . in general the nature of pairing and superfluidity in strongly interacting systems
is both subtle and intriguing @xcite .
pairing is well known to be a momentum space phenomenon , in which two fermions with opposite spin states form a bound pair which behaves like a boson .
the bound pairs form a superfluid , while the unpaired fermions remain as a separated gas phase in momentum space .
such superfluid states with gapless excitations in ultracold atomic gases provide an exciting insight into the superfluid regime in quantum many - body physics .
fermi gases of ultracold atoms with population imbalance have been predicted to exhibit a quantum phase transition between the normal and superfluid states @xcite .
mismatched fermi surfaces can appear in different quantum systems , such as type ii superconductors in an external magnetic field @xcite , a mixture of two species of fermionic atoms with different densities or masses @xcite and charge neutral quark matter @xcite .
these exotic phases have attracted newfound interest in the one - dimensional ( 1d ) integrable two - component fermi gas @xcite which was used to study bcs - bec crossover @xcite and quantum phase separation in a trapping potential @xcite .
the 1d fermi gases can be experimentally realized by applying strongly transverse confinement to the fermi atomic clouds @xcite . in the 1d interacting fermi gas ,
the fermi surface is reduced to the fermi points .
the lowest excitation destroys a bound pair close to the fermi surface .
charge and spin propagate with different velocities due to the pair - wise interaction .
the external magnetic field triggers energy level crossing such that the fermi surfaces of paired fermions and unpaired fermions vary smoothly with respect to the external field .
as we shall see in this paper , the presence of the external field at zero temperature has an important bearing on the nature of quantum phase transitions in 1d interacting fermions . in general the exact bethe ansatz ( ba ) solution of any model provides reliable physics beyond mean field theory @xcite .
the thermodynamic bethe ansatz ( tba ) @xcite provides a way to obtain the ground state signature and finite temperature behaviour of integrable 1d quantum many - body systems . at zero temperature
, the tba equations naturally reduce to dressed energy equations in which the external field is explicitly involved .
thus the band fillings are subsequently varied with respect to the external field .
this gives an elegant way to analyze quantum phase transitions in the presence of an external field by means of the dressed energy formalism .
our aim here is to obtain new exact results from this formalism for characteristics of pairing phases and quantum phase transitions in the 1d two - component strongly attractive fermi gas of cold atoms .
we present a systematic way to obtain the critical fields and magnetic properties at zero temperature for strongly interacting fermions .
we find that the bound pairs of fermions with opposite spin states form a singlet ground state when the external field @xmath0 .
a completely ferromagnetic phase without pairing occurs when the external field @xmath1 . in the region
@xmath2 we observe a mixed phase of matter in which paired and unpaired atoms coexist . however , in the absence of the external magnetic field , we show that the bound pairs of fermions behave like hard - core diatoms obeying nonmutual generalized exclusion statistics ( ges ) at temperatures much less than the binding energy .
this paper is set out as follows . in section [ model ] , we present the ba solution of the 1d two - component interacting fermi gas .
the ground state properties are also analysed . in section [ tba ] , we introduce the tba in order to set up the dressed energy formalism .
the quantum phase transitions and magnetic properties are studied by means of the dressed energy formalism in section [ qp ] .
we discuss the distribution profiles and the thermodynamics of the 1d strongly attractive fermi gas of atoms at low temperatures in section [ ges ] , along with the connection to ges .
section [ conclusion ] is devoted to concluding remarks .
the model we consider has interacting atoms in two hyperfine levels @xmath3 and @xmath4 , which are coherently coupled with laser or rabi frequency fields . under strong transverse confinement ,
the system is effectively described along the axial direction by the 1d hamiltonian @xmath5 . the first term @xmath6 contains the kinetic energy and the trapping potential @xmath7 .
the second term @xmath8 describes @xmath9-wave interaction and @xmath10 is the coupling term . here
@xmath11 and @xmath12 are the atomic field creation operators , @xmath13 is the 1d interaction strength and @xmath14 is the rabi frequency of coupling fields . defining @xmath15 and @xmath16 ,
the hamiltonian becomes @xmath17 the new field operators @xmath18 and @xmath19 describe the atoms in the states @xmath20 and @xmath21 .
this hamiltonian also describes the 1d @xmath22-interacting spin-@xmath23 fermi gas with an external magnetic field @xmath24 . here
we consider the homogeneous case @xmath25 with periodic boundary conditions for a line of length @xmath26 @xcite .
unless specifically indicated , we use units of @xmath27 .
we define @xmath28 and a dimensionless interaction strength @xmath29 for the physical analysis , with linear density @xmath30 , where @xmath31 is the number of fermions .
the inter - component interaction can be tuned from strongly attractive ( @xmath32 ) to strongly repulsive ( @xmath33 ) via feshbach resonances .
the model was solved by nested ba @xcite for the energy eigenspectrum @xmath34 in terms of the @xmath31 ba wave numbers @xmath35 , which are the quasimomenta of the fermions .
they satisfy the ba equations @xcite @xmath36 here @xmath37 and @xmath38 , with @xmath39 the number of spin - down fermions .
the additional parameters @xmath40 are the rapidities for the internal spin degrees of freedom .
the distribution of the quasimomenta in the complex plane was studied recently @xcite . for weakly attractive interaction
, the system describes weakly bound cooper pairs where the quasimomenta are distributed in a bcs - like manner @xcite ( figure [ fig : cartoon](a ) ) . in this limit
, the ground state energy per unit length is given by @xmath41 with the polarization @xmath42 .
the bound state has a small binding energy @xmath43 and is therefore unstable against thermal fluctuations . for strongly attractive interaction , the bound pairs form hard - core bosons ( figure [ fig : cartoon](b ) ) .
the energy per unit length derived directly from eqs ( [ be ] ) is @xcite @xmath44 with binding energy @xmath45 . generally , the total momentum for bound pairs and that for unpaired fermions are both zero . thus the ba roots for the model with population imbalance do not show sufficient evidence for the existence of a fflo state which might exist in the asymmetric bcs pairing models @xcite . in fflo
states the unpaired fermions have an asymmetric distribution at the fermi surface resulting in a net total momentum for bound pairs @xcite . in the complex plane .
( a ) for weakly attractive interaction , the quasimomenta of unpaired fermions sit in the outer wings of the distribution .
( b ) for strongly attractive interaction , they can penetrate into the central region.,title="fig : " ]
the thermodynamic bethe ansatz ( tba ) has been well established in quantum integrable systems @xcite . for the sake of completeness , we sketch the main idea and results of the tba for the fermion model in this section . at zero temperature ,
all quasimomenta @xmath46 of @xmath31 atoms form two - body bound states , i.e. , @xmath47 , accompanied by the real spin parameter @xmath48 . here
. however , at finite temperature , spin quasimomenta form complex strings @xmath50 with @xmath51 @xcite . here the number of strings @xmath52 .
@xmath53 is the position of the center for the length-@xmath54 string on the real axis .
the number of @xmath54-strings @xmath55 satisfies the relation @xmath56 .
there are @xmath57 real @xmath48 and there are @xmath58 real @xmath46 for unpaired fermions . in the presence of a magnetic field ,
the ground state consists of two fermi seas : one contains bound pairs , another contains unpaired fermions . in the thermodynamic limit ,
i.e. , @xmath59 with @xmath60 finite , it is assumed that the distributions of bethe roots is sufficiently dense along the real axis . after introducing the root distribution functions @xmath61
, @xmath62 and @xmath63 for paired fermions , unpaired fermions and the spin @xmath54-string as well as their hole densities @xmath64 , @xmath65 and @xmath66 , the ba equations ( [ be ] ) can be transformed into the form @xcite @xmath67 here @xmath68 denotes the convolution integral @xmath69 and @xmath70 the function @xmath71 can be found in takahashi s book @xcite .
the equilibrium states at finite temperature @xmath72 are described by the equilibrium quasiparticle and hole densities .
the partition function @xmath73 is defined as @xmath74 where the densities satisfy the ba equations ( [ be - string ] ) , and @xmath75 is the number of states corresponding to the given densities . by introducing the combinatorial entropy @xmath76 the grand partition function
can be presented as @xmath77 , where the gibbs free energy @xmath78 . here
@xmath79 is the chemical potential .
the entropy and gibbs free energy are given in terms of the ba root distribution functions of particles and holes for bound pairs and unpaired fermions as well as spin degrees of freedom .
the energy per unit length is defined by @xmath80 here @xmath81 is the external magnetic field and @xmath82 denotes the atomic magnetic momentum per unit length ( where the bohr magneton @xmath83 and the lande factor are absorbed into the magnetic field @xmath81 ) .
the entropy per unit length is given by @xcite @xmath84 the equilibrium states are determined by the minimization condition of the gibbs free energy , which gives rise to a set of coupled nonlinear integral equations
the tba equations @xcite . in terms of the dressed energies
@xmath85 and @xmath86 for paired and unpaired fermions these are @xmath87 the function @xmath88 is the ratio of the string densities .
the gibbs free energy per unit length is given by @xmath89 the tba equations provide a clear picture of band fillings with respect to the field @xmath81 and the chemical potential @xmath90 at arbitrary temperatures .
however , it is a challenging problem to obtain analytic results for the thermodynamics at low temperatures from the tba ( [ tba - full ] ) .
we focus on quantum phase transitions in the 1d strongly attractive fermi gas at @xmath91 by analyzing the dressed energy equations @xmath92 which are obtained from the tba equations ( [ tba - full ] ) in the limit @xmath93 .
the dressed energy @xmath94 ( @xmath95 ) for @xmath96 ( @xmath97 ) correspond to the occupied states .
the positive part of @xmath98 ( @xmath99 ) corresponds to the unoccupied states .
the integration boundaries @xmath100 and @xmath101 characterize the fermi surfaces for bound pairs and unpaired fermions , respectively .
the gibbs free energy per unit length at zero temperature is given by @xmath102 the magnetization @xmath103 per unit length is determined by @xmath81 , @xmath13 and @xmath54 through the relations @xmath104
the ground state is antiferromagnetic , i.e. , the number of the fermionic atoms with up - spin states and the number of the fermionic atoms with down - spin states are equal . in this case the integral limit for the unpaired fermi sea @xmath105 and @xmath106 . for strong coupling ,
i.e. @xmath107 the dressed energy equations ( [ tba - f ] ) reduce to the form @xmath108 for convenience of notation we denote @xmath109 as the pressure for bound pairs and unpaired fermions . substituting equation ( [ tba - g ] ) into @xmath110 , we have @xmath111 furthermore , from the fermi points @xmath112 , we have @xmath113 from the relation ( [ deffield ] ) together with the equations ( [ tba - g1 ] ) and ( [ tba - g2 ] ) , we obtain the pressure and the ground state energy per unit length as @xmath114 for the strongly attractive 1d fermi gas , the low - energy excitations split into collective excitations carrying charge and collective excitations carrying spin .
this leads to the phenomenon of spin - charge separation .
the spin excitation is gapped with a divergent spin velocity @xmath115 therefore the spin sector can not be described by a conformal field theory
. however , the charge sector is still critical @xcite with central charge @xmath116 and the charge velocity @xmath117 for the fully paired ground state , where the bound pairs behave like hard - core bosons . in the above equation @xmath118 .
the bound pairs can be broken by a strong enough external field or thermal fluctuations . in the strong interaction limit , it was demonstrated in a recent experiment that the nature of the pairing is likely to be molecular and the mismatched fermi surfaces do not prevent pairing but indeed quench the superfluidity @xcite .
the state with polarization can be viewed as an ideal mixture of bosonic pairs and fermionic quasiparticles . with polarization @xmath119 , from the equations ( [ tba - f ] ) we obtain @xmath120 from the fermi points @xmath121 and @xmath122 we have
@xmath123 it follows that @xmath124 with the help of the relation ( [ deffield ] ) and by lengthy iteration we find the effective chemical potentials for pairs @xmath125 and unpaired fermions @xmath126 are given by @xmath127 and @xmath128 these results can give explicit chemical potentials for the two different species : @xmath129 in addition we have the total chemical potential @xmath130 . here
the energy per unit length with polarization ( [ e - p ] ) follows from the relation @xmath131 .
indeed , the energy obtained from the tba formalism is in agreement with the result ( [ e - p ] ) derived from the ba @xcite .
the integration boundaries @xmath132 are the largest quasimomentum for bound pairs and unpaired fermions .
analysis of the dressed energy equations ( [ tba - f ] ) shows that the fully paired ground state with @xmath133 is stable when the field @xmath134 , where @xmath135 this critical field makes the excitation gapless . if the external field @xmath136 , the _ pairing gap _ , defined by @xmath137 , is completely diminished by the external field . slightly above the critical point @xmath138 ,
the system has a linear field - dependent magnetization @xmath139 with a finite susceptibility @xmath140 this behaviour differs from the pokrovsky - talapov - type phase transition occurring in a gapped spin liquid .
we note that this smooth phase transition in the attractive fermi gas is reminiscent of the transition from the meissner phase to the mixed phase in type ii superconductors @xcite . on the other hand , if the external field @xmath141 , where @xmath142 all bound pairs are broken and the ground state becomes a normal ferromagnetic state of fully polarized atoms . slightly below @xmath143
the phase transition is determined by the linear field - dependent relation @xmath144 with a finite susceptibility @xmath145 a mixed phase occurs in the region @xmath146 , with coexistence of spin singlet bound pairs and unpaired fermions with ferromagnetic order .
the external field - magnetization relation @xmath147 follows by solving equations ( [ mu - b ] ) and ( [ mu - u ] ) .
it indicates the energy transfer relation among the kinetic energy variation @xmath148 , the binding energy @xmath149 and the zeeman energy @xmath150 : @xmath151 which qualitatively agrees with the relation identified in experiment @xcite . in the above equation
we take the bohr magneton @xmath152 and @xmath153 . and
susceptibility @xmath154 vs the external field @xmath81 in the units @xmath155 , according to eq . (
[ h]).,title="fig : " ] figure [ fig : mz](a ) shows the magnetization and the susceptibility for @xmath156 and @xmath157 .
the magnetization gradually increases from @xmath133 to @xmath158 as the field increases from @xmath138 to @xmath143 .
it is important to note that the points of intersection at @xmath159 indicate where the fermi surface of unpaired fermions exceeds the one for the bound pairs .
this point separates the mismatched pairing phase into different breached pairing phases @xcite .
the susceptibility shows discontinuities at the critical points , with @xmath160 for @xmath134 and @xmath141 .
however , @xmath154 is finite and quickly decreases in the vicinity of @xmath138 . for larger densities ( figure [ fig : mz](c ) illustrates the case @xmath161 )
, @xmath154 slowly increases as @xmath162 .
we note that the coexistence of pairing and magnetization in the 1d attractive fermi gas is similar to the shubnikov phase of superconductivity and magnetization in type ii superconductors @xcite .
according to eqs .
( [ hc1 ] ) , ( [ hc2 ] ) and ( [ h]).,title="fig : " ] figure [ fig : pd ] shows the phase diagram in the @xmath163 plane for the particular value @xmath164 . as @xmath165 ,
the two critical fields approach the binding energy @xmath149 .
the two critical fields have opposite monotonicity : @xmath138 decreases with increasing @xmath54 whereas @xmath143 increases with @xmath54 . for the 1d fermi gas in a harmonic trapping potential @xcite ,
the density is position - dependent and decreases away from the trapping centre .
thus , for sufficiently large centre - density , the system has subtle segments ; the mixed phase lies in the centre and the fully paired phase ( or the fully unpaired phase ) sits in the two outer wings for @xmath166 ( or @xmath167 ) .
nevertheless , for sufficiently low centre - density , the cloud is either a fully paired phase or a fully unpaired phase for @xmath166 or @xmath167 , respectively . for the 1d fermi gas
, the local pair correlation is defined by @xmath168 for weakly attractive interaction , @xmath169 indicating a two - component free fermi gas phase . for strongly attractive interaction the local pair correlation is given by @xmath170.\ ] ] this has maximum and minimum values corresponding to the fully paired phase for @xmath134 and the fully unpaired phase for @xmath141 , respectively .
depairing weakens the pair correlation in the region @xmath171 .
the phase transitions in the vicinities of the critical points are of second order .
strong thermal fluctuations can destroy the magnetically ordered phases , delineated by the two critical fields , above a critical temperature @xmath172 , which in principle can be also calculated from the tba equations . on the other hand , at temperatures much lower than the degeneracy temperature @xmath173 , the bound pairs are stable against weak thermal fluctuations . however , the individual pair wavefunctions do not overlap coherently , i.e. , the existence of bound pairs does not lead to long range order at finite temperatures .
this can be seen from the finite temperature tba equations ( [ tba - full ] ) in which the unpaired band is empty due to a large negative chemical potential .
this behaviour is similar to that of 3d attractive fermions @xcite . in 1d
the dynamical interaction and the statistical interaction in the pairing scattering process are inextricably related @xcite .
this means that one bound pair excitation may cause a fractional number of holes below the fermi surface due to the collective signature .
this is the key point in understanding ges for 1d interacting many - body systems .
we shall show that the bound pairs can be viewed as ideal particles obeying ges .
ges has recently been applied to the 3d unitary fermi gas @xcite . in the absence of the magnetic field and at low temperatures ,
the unpaired dressed energy is positive due to a large negative chemical potential .
it follows that the ba and tba equations can be written as @xmath174 after neglecting exponentially small terms in ( [ p - b ] ) , the pressure for the bound pairs is given by @xmath175 with the function @xmath176 .
furthermore , using sommerfeld expansion and iterating the pressure @xmath110 with equation ( [ tba - f2 ] ) , we obtain the cut - off energy @xmath177 in the form @xmath178 , \label{tba - mu}\ ] ] where @xmath179 here @xmath180 is the degenerate temperature . the pair distribution function @xmath181 is given by @xmath182 where @xmath183 .
the chemical potential follows as @xmath184 \nonumber\\ & & \,-\frac12{\epsilon_b}.\end{aligned}\ ] ] here @xmath185 which is consistent with ( [ mu - b ] ) .
the total energy per unit length and the free energy per unit length in the strong coupling regime are @xmath186 \nonumber\\ & & \,-\,\frac12 { n \epsilon_{\rm b } } , \nonumber\\ f&\approx & e_0\left[1-\frac{16\tau^2}{\pi^2 } \left(1-\frac{2}{|\gamma| } \right ) -\frac{1024\tau^4}{15\pi^4 } \left(1-\frac{4}{|\gamma|}\right)\right ] \nonumber\\ & & \,-\ , \frac12 { n\epsilon_{\rm b}},\label{tba - ef}\end{aligned}\ ] ] respectively .
here @xmath187 is consistent with ( [ e0 ] ) obtained from ( [ tba - f ] ) .
we see that in the strongly attractive regime and in the absence of the magnetic field the bound pairs behave like hard - core bosons at low temperatures and have massless excitations , i.e. @xmath188 here the central charge @xmath116 and @xmath189 is given by ( [ vc ] ) .
the specific heat is given by @xmath190 on the other hand , the statistical signature of the fully paired state can be described by ges @xcite . in this formalism
the pair distribution function is given by @xmath191 where @xmath192 satisfies the ges relation @xmath193 here @xmath194 denotes the energy of pairs . for the strongly attractive fermi gas
, we find the ges parameter @xmath195 now following isakov _
@xcite , at low temperatures , i.e. , for @xmath196 , we find the cut - off energy @xmath197,\nonumber\end{aligned}\ ] ] which agrees with ( [ tba - mu ] ) to leading order and next leading order in the strong coupling regime ( for higher order terms , the reader is referred to ref .
figure [ fig : n ] shows the close agreement between the tba distribution function ( [ tba - n ] ) and the ges most probable distribution of fermion pairs ( [ ges - n ] ) for different values of interacting strength at low temperatures .
we see clearly that the dynamical interaction @xmath198 continuously varies the ges , with the most probable distribution of fermion pairs approaching that of hard - core bosonic molecues with an effective statistics parameter @xmath199 as the interaction increases . in this sense
the dynamical attractive interaction makes the fermions more exclusive . in the ges formalism the total energy per unit length and the free energy per unit length follow as @xmath200 -\frac12\ , { n \epsilon_{\rm b}},\nonumber\\ f&\approx & e_0\left[1-\frac{16\tau^2}{\pi^2\alpha^3}+o(\tau^4)\right ] -\frac12\ , { n \epsilon_{\rm b}},\label{ges - ef}\end{aligned}\ ] ] which again agree well with the tba results ( [ tba - ef ] ) for strong coupling , see figure [ fig : ef ] . for the values
@xmath201 and @xmath202 at different values of the degeneracy temperature @xmath203
. at zero temperature @xmath204 leads to a fermi surface at @xmath205 .
fermi - dirac statistics with ges parameter @xmath199 appear for @xmath206 .
attractive interaction thus results in a more exclusive state than for pure fermi - dirac statistics .
the solid and dashed lines are obtained from the tba distribution function ( [ tba - n ] ) .
the symbols show the most probable distribution evaluated from the ges result ( [ ges - n ] ) .
the results from both approaches are seen to coincide well . ] and total energy per unit length @xmath207 in units of @xmath208 _ vs _ the degenerate temperature @xmath203 for different coupling strengths @xmath209 .
the solid and dashed lines are evaluated from the tba results ( [ tba - ef ] ) .
the symbols follow from the ges results ( [ ges - ef ] ) . for @xmath210 , a small discrepancy between the tba and ges results
is observed due to the approximation condition @xmath211 . in the strong coupling regime
, the universality ( [ f - t ] ) of low temperature behaviour suggests that strongly attractive fermions can be viewed as ideal particles obeying nonmutual ges with @xmath212 for temperatures @xmath213 . ]
in conclusion , we have studied pairing and quantum phase transitions in the strongly attractive 1d fermi gas with an external magnetic - like field .
analytic results have been obtained for the critical fields @xmath138 and @xmath143 , magnetization , critical behaviour and local pair correlation .
the pairing induced by an interior gap in the system differs from conventional bcs pairing and gapped spin liquids . the smooth pair breaking phase transitions seen in the attractive fermi gas are reminiscent of the superconductivity breaking phase transitions in type ii superconductors @xcite . at low temperatures
we predict that the hard - core bound pairs of fermionic atoms obey ges .
the thermodynamics of the hard - core pairs obey universal temperature dependent scaling .
we emphasize here that in the presence of an external magnetic field , pair breaking in the 1d two - component strongly attractive fermi gas sheds light in understanding the pairing signature of the 3d strongly interacting fermi gas of ultracold atoms in which superfluid and normal phases can coexist .
although there is no long range order in 1d quantum many - body physics , the mismatched fermi surfaces do not prevent pairing .
this pairing signature has also been observed in the 3d two - component atomic gas with high spin population imbalances @xcite .
in addition , for the 1d fermi gas , the magnetic field triggers spin imbalances when the external field is greater than the first critical field @xmath138 ( [ hc1 ] ) and less than the second critical field @xmath143 ( [ hc2 ] ) . the energy transfer relation ( [ h - e - relat ] ) found for this model is also consistent with experimental observations in the 1d fermi gas @xcite .
the phase diagram presented in figure [ fig : pd ] clearly shows the phase separation and the pairing signature with changing magnetic field . it may be possible to experimentally test our theoretical predictions for the quantum phase transitions and critical fields in the 1d two - component strongly interacting fermi gas via the experimental advances in trapping ultracold atoms .
m. w. zwierlein , a. schirotzek , c. h. schunck , and w. ketterle , science * 311 * , 492 ( 2006 ) ; m. w. zwierlein , c. h. schunck , a. schirotzek and w. ketterle , nature * 442 * , 54 ( 2006 ) ; y. shin , m. w. zwierlein , c. h. schunck , a. schirotzek and w. ketterle , phys .
lett . * 97 * , 030401 ( 2006 ) . g. b. partridge , w. h. li , r. i. kamar and r. g. hulet , science * 311 * , 503 ( 2006 ) ; g. b. partridge , w. h. li , y. h. liao , r. g. hulet , m. haque and h. t. c. stoof , phys .
lett . * 97 * , 190407 ( 2006 ) .
w. v. liu and f. wilczek , phys .
90 * , 047002 ( 2003 ) .
m. m. forbes , e. gubankova , w. v. liu and f. wilczek , phys .
* 94 * , 017001 ( 2005 ) .
m. a. cazalilla , a. f. ho and t. giamarchi , phys .
95 * , 226402 ( 2006 ) .
j. n. fuchs , a. recati and w. zwerger , phys .
lett . * 93 * , 090408 ( 2004 ) ; i. v. tokatly , phys .
lett . * 93 * 090405 ( 2004 ) ; g. e. astrakharchik , d. blume , s. giorgini and l. p. pitaevskii , phys . rev . lett . *
93 * 050402 ( 2004 ) .
|
we investigate pairing and quantum phase transitions in the one - dimensional two - component fermi atomic gas in an external field .
the phase diagram , critical fields , magnetization and local pairing correlation are obtained analytically via the exact thermodynamic bethe ansatz solution . at zero temperature , bound pairs of fermions with opposite spin states form a singlet ground state when the external field @xmath0 .
a completely ferromagnetic phase without pairing occurs when the external field @xmath1 . in the region
@xmath2 we observe a mixed phase of matter in which paired and unpaired atoms coexist .
the phase diagram is reminiscent of that of type ii superconductors . for temperatures below the degenerate temperature and in the absence of an external field , the bound pairs of fermions form hard - core bosons obeying generalized exclusion statistics .
|
amongst the various techniques used to discretize partial differential equations , spectral methods , introduced by d. gottlieb and s. orszag in the seventies @xcite , offer high accuracy at a low computational cost .
their principle is to approximate solutions of pdes by truncated fourier series or high degree polynomials .
these methods have an infinite order of convergence , because the error between the exact and discrete solutions is only limited by the regularity of the exact solution .
spectral methods resort to fourier series in the case of periodic boundary conditions ; whereas polynomial approximations use orthogonal polynomials ( such as chebyshev or legendre polynomials ) .
these methods also employ quadrature formulas to compute integrals in variationnal formulations of pdes . in the present work ,
we explore a new family of orthogonal polynomials : the @xmath0 jacobi polynomials . after a brief summary about spectral methods in sec .
[ s : spectral ] , we detail several general properties of the jacobi polynomials , and present the necessary tools to build a spectral method based on the @xmath0 jacobi polynomials in sec . [ s : jacobi ] . finally , we show some applications to poisson problems including numerical tests in sec . [
s : test ] .
concluding remarks are given in sec .
[ s : concl ] , and important technical proofs can be found in the appendix ( sec .
[ s : annex ] ) .
both numerical analysis and implementation of spectral methods are based on orthogonal polynomials , whose major properties are hereby recalled .
we shall then emphasize the importance of sturm - liouville problems in the case of legendre polynomials , and mention results about polynomial approximation and interpolation .
several textbooks cover this domain , for instance @xcite , @xcite , or @xcite , and provide comprehensive proofs of the following results .
let @xmath2 be the interval @xmath3 and @xmath4 denote a weight on @xmath2 , i.e , @xmath5 and @xmath6 on @xmath2 .
we define @xmath7 which is a hilbert space for the scalar product @xmath8 one can construct a basis of monic orthogonal polynomials by using a gram - schmidt orthogonalisation process on the basis @xmath9 , @xmath10 : @xmath11 is fixed to @xmath12 , then , assuming the @xmath13 , @xmath14 are known , @xmath15 is chosen by @xmath16 we recall the following property of orthogonal polynomials : for all positive integer @xmath17 , the roots of @xmath15 are real , distinct and strictly bounded by @xmath18 and @xmath12 . in particular , those polynomials do not vanish at @xmath12 , we can thus define any family of orthogonal polynomials by imposing their value at @xmath12 . besides
, the monic orthogonal polynomials satisfy the following induction formula . for all integer @xmath19 , @xmath20 finally , a wide class of orthogonal polynomials belongs to singular sturm - liouville solutions , on which spectral discretizations rely .
we present the legendre case ( @xmath21 and @xmath22 ) , since a general description would be too long . for all integer @xmath10 , @xmath23 satisfies the following differential equation @xmath24 the @xmath25 operator , defined by @xmath26 is self - adjoint in @xmath27 by integration by parts . besides , it is positive and of singular sturm - liouville type .
the equation ( [ eq - dif - legendre ] ) means that legendre polynomials are eigenvectors of this operator , which is the origin of the spectral " adjective of the numerical methods hereby discussed .
a consequence of ( [ eq - dif - legendre ] ) ( through integration by parts ) is that , for all positive integers @xmath28 and @xmath17 : @xmath29 this means that the @xmath30 , @xmath31 form a basis of orthogonal polynomials in @xmath32 .
this section illustrates why spectral methods are so accurate .
the sturm - liouville operator @xmath25 enables us to put some upper bounds on the distance between functions with a given regularity and a polynomial space , still in the case of legendre polynomials .
this distance is computed , by mean of orthogonal projectors .
it has been first estimated in @xcite and @xcite .
let @xmath33 be a positive integer , @xmath34 stands for the space of polynomials with degree less than @xmath33 , and the orthogonal projector from @xmath27 to @xmath34 is denoted by @xmath35 . for any integer @xmath36 , @xmath37 stands for the sobolev space of order @xmath28 on @xmath2 .
[ t : polapprox ] for all integer @xmath36 , there exists a positive constant @xmath38 , only dependant of @xmath28 such that , for all @xmath39 in @xmath37 , @xmath40 the key of the proof of this theorem is the following one .
we denote @xmath41 the @xmath17-th coefficient of @xmath39 in its expansion onto the legendre basis
. then , @xmath42 since @xmath25 is self - adjoint in @xmath27 , @xmath43 by iterating @xmath44 times this result , one can deduce that @xmath45 since @xmath46 is the @xmath17-th coefficient of @xmath47 in its decomposition over the legendre basis , a continuity result on the operator @xmath25 will allow to conclude . thus we have seen that any function , given its regularity , can be approximated by polynomials , with an error decreasing as a power law of the degree of the polynomial .
the power depends only on the chosen norm , and the regularity of the function . however , this result has few direct numerical applications .
indeed , one has to compute integrals to obtain the polynomial approximating the function , which is highly time - expensive .
thus , one shall approximate these integrals with quadrature formulas , and replace the orthogonal projectors by interpolation operators @xmath48 in the nodes of these quadrature formulas .
first , we recall the definition of these operators . [
p : definterp ] given @xmath49 distinct points @xmath50 in @xmath2 , and a continuous function @xmath39 on @xmath2 , there exists a unique polynomial @xmath51 in @xmath52 such that @xmath53 this greatly simplifies the set up of the method , and does nt reduce the precision according to the following theorem , first demonstrated in @xcite .
[ t : polinterp ] for all integer @xmath54 , there exists a positive constant @xmath38 , depending only in @xmath28 , such that , for all function @xmath39 in @xmath37 , we have @xmath55 the latter result shows spectral methods have an infinite order of accuracy .
a @xmath56 function is interpolated faster than any power of the discretization parameter .
moreover , it can be shown that an exponential convergence of the interpolant is achieved if the function is analytic .
in this section , several results about jacobi polynomials are detailed .
first , we present how jacobi polynomials arise naturally from sturm liouville singular problems . then , after presenting some basic properties , we prove a new result about jacobi - gauss - lobatto quadratures , enabling us to build a discrete jacobi transform . finally , we show some differentiation and integration results about @xmath0 jacobi polynomials .
we saw previously that the liouville operator was crucial to obtain an efficient polynomial approximation .
more generally , a sturm - liouville problem consists in looking for solutions @xmath57 of @xmath58 the coefficients @xmath59 , @xmath60 and @xmath4 are three given , real - valued functions such that @xmath59 is continuously differentiable , strictly positive in @xmath3 and continuous at @xmath61 ; @xmath60 is continuous , non - negative and bounded in @xmath3 ; the weight @xmath4 is continuous and non - negative in @xmath3 .
one must notice that every sturm - liouville problem does not ensure the convergence of the associated spectral method .
let us consider the following example ; @xmath62 its eigenvalues are @xmath63 , with eigenfuctions @xmath64 .
thus , a function will be correctly approximated by the cosine series , iff all its even derivatives vanish in @xmath18 and in @xmath12 .
this is due to the non - vanishing of @xmath59 in the extremities of the interval.this sturm - liouville problem is _
regular_. inversely , if @xmath59 vanishes in @xmath18 and in @xmath12 , the problem is said to be _
singular_. in the case of singular problems , efficient polynomial approximation is ensured .
we send back to @xcite for a more complete discussion .
thus , we are led to consider polynomial solutions to singular sturm - liouville problems in order to build an efficient spectral method .
let us note @xmath65 a family of polynomial solutions of degree @xmath44 of a singular sturm - liouville problem with eigenvalues @xmath66 , then one can see that @xmath67 is a polynomial of degree zero , @xmath68 has degree @xmath69 , and @xmath70 has degree @xmath12 , according to the following identity : @xmath71 since @xmath59 vanishes in @xmath18 and in @xmath12 , @xmath72 .
let us note @xmath73 , then @xmath74=a_1/(1-x)+b_1/(1+x)$ ] .
after integration , one has @xmath75 with @xmath76 , @xmath77 and @xmath78 constants . as a result , @xmath79 .
an integrability condition on @xmath4 imposes @xmath80 and @xmath81 .
we then define the jacobi polynomials @xmath82 of index @xmath83 as the orthogonal polynomials for the weight @xmath84 , normalized by @xmath85 where @xmath86 is the euler gamma function .
these polynomials are solutions of the previous singular sturm - liouville problem .
indeed , @xmath87 is a polynomial with degree @xmath88 , and verifies that for all polynomial @xmath39 with degree @xmath89 @xmath90 therefore there exists @xmath91 such that @xmath92 jacobi polynomials are thus the eigenvectors of the previous sturm - liouville problem . identifying the highest degree coefficients , one obtains @xmath93 .
thus , this configuration enables to generalize theorem [ t : polapprox ] .
standard orthogonal polynomials are special cases of jacobi polynomials .
legendre polynomials are jacobi polynomials with index @xmath94 .
if we denote @xmath95 the @xmath17-th chebyshev polynomial , one can verify that @xmath96 finally , @xmath97 jacobi polynomials have been studied in @xcite .
we present here basic results about jacobi polynomials .
it can be proven they are special cases of hypergeometric functions ( see @xcite ) , which provides the following results : * analytical expression @xmath98 * induction formula @xmath99j_{k}^{(\alpha,\beta)}(x ) \\ - 2(k+\alpha)(k+\beta)(2k+\alpha+\beta+2 ) j_{k-1}^{(\alpha,\beta)}(x).\end{gathered}\ ] ] * highest degree coefficient + @xmath100 * norm + @xmath101 * links between the various jacobi polynomials families . @xmath102 the latter formulae show , for example , that @xmath0 jacobi polynomials are linked to the legendre polynomials via the following equation : @xmath103 finally , as we saw it previously for legendre polynomials , derivatives of jacobi polynomials are orthogonal for the weight @xmath104 , which can also be shown from the identity @xmath105 we show in a very illustrative manner the plot of @xmath106 , @xmath107 , in fig.1 . plot of @xmath106 , @xmath107 . notice that @xmath108 the extrema of orthogonal polynomials are useful in order to construct high precision numerical quadrature formulas , i.e , which are exact within a space of high degree polynomials : the so - called gauss - lobatto formulas . in this section
, we recall the definition of these quadrature formulas in proposition [ p : defquad ] , then compute some useful formulas to build numerical spectral methods .
[ p : defquad ] let @xmath33 be a positive integer .
let us denote @xmath109 and @xmath110 .
there exists a unique set of @xmath111 points @xmath112 in @xmath2 , @xmath113 , and a unique set of @xmath49 real @xmath114 , @xmath115 , such that we have the following identity for all @xmath116 in @xmath117 @xmath118 the nodes @xmath112 , @xmath113 , are the zeros of @xmath119 .
moreover , the @xmath120 , @xmath121 are positive . the quadrature formula @xmath122 is called _ jacobi type gauss - lobatto quadrature formula _ with @xmath49 points .
the expression of the coefficients ( also called weights ) @xmath120 , @xmath121 is given by the following proposition .
[ p : jglw ] @xmath123 @xmath124 , \ : \rho_i = \frac{2^{\alpha+\beta+1}}{n(n+\alpha+\beta+1 ) } \frac{\gamma(n+1+\alpha)\gamma(n+1+\beta)}{\gamma(n+1)\gamma(n+1+\alpha+\beta ) } \frac{1}{(j_n^{(\alpha,\beta)}(x_i))^2}.\ ] ] @xmath125 some commentaries arise naturally at this point . as far as the authors know
, this expression of the jacobi gauss - lobatto weights has not been given previously in the literature .
szego , in @xcite , gives a very similar expression , but only for gauss quadrature ( i.e , the nodes of the quadrature are the zeros of @xmath126 ) .
one can also check the validity of this formula for known values of @xmath77 and @xmath78 . for legendre polynomials ( i.e , @xmath94 )
, one can check that @xmath127 \
, \rho_i = 2/[n(n+1)l_n^2(x_i)]$ ] .
as far as @xmath97 jacobi polynomials ( denoted @xmath128 hereafter ) are concerned , our formulas coincide with those found in @xcite , namely @xmath129 $ ] and @xmath130 \ , \rho_i = 4/[n(n+2)m_n^2(x_i ) ] $ ] . for chebyshev polynomials ( @xmath131 ) , this expression is also valid .
indeed , @xmath124 \ : \ : \ : \ : \rho_i = \frac{1}{n^2}\frac{\gamma(n+1/2)^2}{\gamma(n+1)\gamma(n)}\frac{1 } { ( j_n^{(-1/2,-1/2)}(x_i ) ) ^2 } .\ ] ] then equation ( [ e : tcheb ] ) allows us to write @xmath132 $ ] .
however , in the chebyshev case , @xmath133 , so that @xmath134 . finally , @xmath135 \ , \rho_i = \pi /n $ ] , and @xmath136 . for the numerical applications we have in mind , @xmath137 and @xmath138 , so that @xmath139 $ ] and @xmath140 \ , \rho_i = 8/[n(n+3)(j_n^{(0,2)}(x_i))^2]$ ] .
now , it is necessary to know how to compute the nodes of the gauss - lobatto formulas .
looking through the zeros of @xmath141 via a secant method is simple , but zeros tend to accumulate near the boundaries of the interval . with no preliminary knowledge of the distribution of those knots , it is preferable to use a eigenvalue location method which is enabled by the following proposition .
[ p : matrixnodes ] the @xmath142 , @xmath143 , are the eigenvalues of the tridiagonal symmetric matrix @xmath144 with @xmath145 @xmath146 the form of the matrix allows for a fast and robust computation of the eigenvalues , e.g , by a givens - householder algorithm . in the case of @xmath147 and @xmath148 , @xmath149 finally , these quadrature formulas , enable us to build a discrete jacobi transform , i.e , if one knows the values of a function on the gauss - lobatto nodes , on can compute the coefficients of its polynomial interpolant in the gauss - lobatto nodes .
[ p : djt ] let us note @xmath150 the interpolant of f on the gauss - lobatto nodes , then @xmath151 @xmath152 various cases of these formulae can be checked . for legendre polynomials
, they reduce to @xmath153 in our case ( @xmath154 and @xmath138 ) , @xmath155 this discrete jacobi transform is an essential ingredient in order to build a tau method @xcite . indeed
, any function @xmath156 is represented by the coefficients @xmath157 from the latter proposition .
differential operators on @xmath156 can be translated as linear operators on the @xmath158 , boundary conditions as linear constraints , etc .
since we have detailed the properties of jacobi - gauss - lobatto quadrature formulas , we may investigate the polynomial interpolation error .
theorem [ t : polinterp ] has been proven for legendre in @xcite , chebyshev in @xcite , @xmath97 jacobi in @xcite , and gegenbauer polynomials ( @xmath159 ) in @xcite .
it still remains to be computed in the @xmath0 case . as an illustration of this section , we show the plot of @xmath160 and its gauss - lobatto nodes in fig.2 , we also include the location of chebyshev gauss - lobatto nodes of the same order for comparison .
plot of @xmath160 , location of jacobi and chebyshev gauss - lobatto nodes . ] the following result is enclosed here for completeness .
it may not be of interest for our numerical purposes , but may be useful in order to devise some _ a priori _ inequalities in non - linear problems .
see @xcite for more details .
[ t : invineq ] there exists a real @xmath38 such that the following inequality is satisfied for all positive integer @xmath33 and all polynomial @xmath161 in @xmath34 .
@xmath162 this inequality is optimal . indeed , there exists a constant @xmath163 such that @xmath164 let us have a closer look at the orthogonality property of @xmath0 jacobi polynomials , which we shall denote @xmath165 hereafter .
@xmath166 by mean of the change of variable @xmath167 , we obtain @xmath168 thus appears the crucial @xmath1 weight in spherical coordinates .
the utility of such a family of polynomial appears when computing global quantities .
let us take the example of a magnetic field in a spherically symetric geometry .
its total energy is proportionnal to @xmath169 . denoting @xmath170 its coefficients in a decomposition on a jacobi polynomial basis
, its total energy will then be proportionnal to @xmath171 .
using chebyshev polynomials would have required to employ desialasing techniques to compute this integral ( see @xcite ) . besides , when dealing with non - linear terms , one can use , for example , truncation techniques , and in our case , truncating jacobi coefficients does nt increase the energy of the field . for instance , if the previous magnetic field obeys an induction equation with dissipation , exact solutions have decreasing energy , and we ensure the desaliasing techniques do not increase the energy of the numerical solution .
various approaches can be examined when building a spectral method .
a c++ library called lorene @xcite has been built and mostly uses the lanczos - tau method exposed in @xcite .
therefore , we need operator matrices such as derivation , integration , multiplication and division by @xmath172 , whose expressions in the @xmath0 case are given in the following proposition [ p : opmatrices ] for all positive integer @xmath17 , the following identities are true : * derivation : + @xmath173 j_j .\ ] ] * integration : + @xmath174 * division by @xmath175 + @xmath176 j_j .\ ] ] * multiplication by @xmath175 + @xmath177
numerical tests have been performed using the c++ library lorene @xcite which provides useful tools for numerical relativity , especially elliptic solvers .
poisson equations are of outmost importance when looking for solutions of einstein equations .
indeed , various formulations of these equations lead to elliptic equations which can be solved through iterative poisson resolutions .
lorene performs 3d multi - domain resolutions with spectral accuracy thanks to chebyshev polynomials in the radial direction , and spherical harmonics or trigonometric polynomials in the @xmath178 and @xmath39 directions .
moreover , it uses several spherical domains , a nucleus centered around @xmath179 , some shells , and an external compactified domain ( using @xmath180 as a variable ) .
some applications can be found in @xcite , @xcite , and @xcite .
we implemented the use of the @xmath0 jacobi polynomials in the nucleus , where their orthogonality properties are the most interesting , and developped a poisson solver thanks to these polynomials .
decay of the errors ( maximum difference with theoretical solution on collocation points ) in each domain .
settings are @xmath181 . ]
the poisson equation reads @xmath182 where @xmath183 is the laplace operator . our procedure is the following one : we perform a decomposition of functions on spherical harmonics , through @xmath184 where @xmath185 denote the spherical harmonics . since @xmath186
, we reduce equation ( [ e : poisson ] ) to a set of 1d equations @xmath187 then , assuming @xmath188 , one can assemble the matrix of the operator appearing in ( [ e : poisson1d ] ) , and adding some boundary conditions , invert it to find the coefficients @xmath189 .
more precisely , we decompose @xmath190 in three domains , a sphere of radius @xmath12 , a shell between radii @xmath12 and @xmath69 , and an external zone @xmath191 .
we map those three domains to @xmath192 $ ] via the following equations : @xmath193 then , for each @xmath194 , we assemble a three - block matrix containing the matrices of the following operator in the @xmath0 jacobi basis for the nucleus thanks to ( [ e : derjacobi ] ) and ( [ e : divjacobi ] ) , and in the chebyshev basis for the two other domains decay of the errors ( maximum difference with theoretical solution on collocation points ) in the nucleus for jacobi and chebyshev methods .
settings are @xmath181 . ]
@xmath195 - \frac{l(l+1)}{(1+x)^2}[f - f(-1 ) - ( 1+x)f'(-1)].\ ] ] in the nucleus , it corresponds to the operator in ( [ e : poisson1d ] ) , under the assumption that for @xmath196 , @xmath197 , for @xmath198 , @xmath199 , and @xmath200 , @xmath201 . in order to enforce those conditions when solving the problem
, we replace the last lines of the first block - matrix by the values of @xmath202 or @xmath203 , and those of the source by @xmath204 , depending of the value of @xmath194 .
finally , we include @xmath205 matching conditions between the domains in the 3-block matrix .
several numerical tests have been passed .
first , we used the following smooth source in eq.([e : poisson ] ) @xmath206 which admits the following solution @xmath207 we compute the error between the analytical and numerical solution by looking at the maximum difference on the collocation points in each domain , which gives an estimation of the corresponding @xmath208 norm of the difference .
results are displayed in fig .
3 and show as expected an exponential decay of the error as a function of the number of radial collocations points .
one can compare the efficiency of chebyshev and jacobi polynomials for the solution ( [ e : smoothsource])-([e : smoothsolution ] ) .
when lorene uses chebyshev base in the nucleus , the mapping is @xmath209 , where @xmath210 is the radius of the nucleus , and we use the parity of @xmath211 with respect to @xmath194 to use odd or even chebyshev bases .
notice this is nt possible in the jacobi case because jacobi polynomials with @xmath212 do not have a definite parity .
results are displayed in fig .
one can see no loss or gain of precision is acquired with this method .
in fact , jacobi polynomials are expected to be more useful in dynamical evolutions to ensure @xmath213 stability ( see sec .
[ motivation ] ) .
finally , we examined a non - smooth solution , namely @xmath214 where @xmath210 is the outer radius of the shell .
decay of the errors ( maximum difference with theoretical solution on collocation points ) in the nucleus for jacobi and chebyshev methods .
settings are @xmath181 . ]
results are displayed in fig .
5 , and one can see as expected a geometric decay of the error as a function of the number of radial coefficients .
we found that the rate of convergence in the chebyshev case is approximately @xmath215 , and @xmath216 in the jacobi case .
indeed , @xmath217 , and @xmath218 .
we have described a new numerical spectral method and showed some applications to poisson problems in a sphere . in order to build this method ,
we have first recalled basic results about orthogonal polynomials and generalized quadrature formulas and discrete polynomial tranforms . besides , we computed various differential and algebraic operators . through numerical tests of poisson equations
, we have shown that the method is convergent and accurate .
in particular , for smooth solutions , we find as expected a error decaying exponentially with the number of radial coefficients .
this method has also been integrated to a 3d multi - domain spectral solver .
however , the interpolation polynomial error on the gauss - lobatto nodes remains to be computed .
studying the nodes of the gauss lobatto quadrature formulas shows evidence that their distribution is very similar to the legendre case , or the @xmath97 case , and therefore , that a similar theorem could be proved .
other applications for @xmath0 jacobi polynomials could be foreseen .
for instance , hyperbolic equations in a sphere , with applications to fluid dynamics in a star .
the authors wish to thank jrme novak for his help on numerical issues , eric gourgoulhon and christine bernardi for their critical reading of the manuscript .
this work was supported by anr grant 06 - 2 - 134423 _ mthodes mathmatiques pour la relativit gnrale_.
in this section we provide proofs for the major new results listed in this paper . to simplify the expressions , @xmath219 will be denoted @xmath82 in this proof , except when the index is explicitly written .
the proof resorts to a already known @xcite preliminary lemma .
we shall note @xmath220 the induction formula of the @xmath221 , since they are orthogonal polynomials with respect to the weight @xmath104 .
let us fix @xmath222 in @xmath223 $ ] , et define @xmath224 in @xmath34 .
the exactness of the quadrature formula ( [ quadformula ] ) then gives @xmath225 the computation of the integral may be done through the study of the following quantity @xmath226 the induction formula of the @xmath221 allows us to write @xmath227 we here recall that @xmath228 , according to the induction formula of orthogonal polynomials .
then @xmath229 since @xmath230 , one can see that @xmath231 this is already known christoffel - darboux formula ( see @xcite ) for orthogonal polynomials .
let us take that equality in @xmath232 , @xmath233 , @xmath234 , multiply by @xmath235 , then integrate with respect to @xmath4 .
left member is given by @xmath236 in the right member , by orthogonality , every term in the sum vanishes , except for @xmath237 , to give @xmath238 since @xmath239 is a constant .
then , one deduces that @xmath240 * computation of @xmath241 + after integration by parts , and using the differential equation ( [ eqdifjacobi ] ) , one has + @xmath242 + i.e , @xmath243 * computation of @xmath244 + the identification of highest degree coefficients leads to @xmath245 * computation of @xmath246 + developing the differential equation ( [ eqdifjacobi ] ) , and evaluating in @xmath112 results in + @xmath247 * computation of @xmath248 + induction formula of @xmath221 evaluated in @xmath112 gives @xmath249 + besides , deriving the induction formula ( [ inducjacobi ] ) of the @xmath219 , and evaluating in @xmath112 gives + @xmath250 + but @xmath251 + after substitution , one has + @xmath252 + so + @xmath253 let us now compute @xmath255 .
we apply the quadrature formula ( [ quadformula ] ) to @xmath256 .
we then obtain @xmath257 but @xmath258 , according to ( [ linkjacobi ] ) . multiplying this by @xmath259 and integrating leads to @xmath260 hence @xmath261 moreover
, @xmath262 finally , @xmath263 then , @xmath264 since @xmath265 , one has @xmath125
first , let us recall that @xmath267 , then we will look for the zeros of @xmath268 .
but the @xmath269 satisfy the following induction formula @xmath270 j_n^{(\alpha+1,\beta+1 ) } \\ - 2 ( n+\alpha+1 ) ( n+\beta+1 ) ( 2n+\alpha+\beta+4 ) j_{n-1}^{(\alpha+1,\beta+1 ) } \end{gathered}\ ] ] let us then define @xmath271 then , the reccurence formula of the @xmath272 can be written as such @xmath273 or similarly @xmath274 which can be written in a matrix form @xmath275 otherly said , the @xmath142 , @xmath143 , zeros of @xmath276 , are the eigenvalues of the ( [ e : matrixnodes ] ) matrix . for @xmath279 , since @xmath280 ,
only @xmath281 is left in the right - hand side . then @xmath282 .
eq.([e : jacobinorm ] ) and the expressions of @xmath114 from proposition [ p : jglw ] lead to the desired result . for @xmath283
, the exactness of the quadrature formula allows us to suppress the @xmath111 first terms of the sum , but the last one is not equal to the norm of @xmath126 . precisely , the right side member is equal to @xmath284 . using the expressions of the weights ,
one finds one can deduce that @xmath295^{1/2 } \\
\leq \| \varphi \|_{l^2_w(\lambda ) } \left [ n^4 \frac{1}{4}\left ( \frac{1}{\alpha+1}+\frac{1}{\beta+1}\right ) \text{sup}_{0\leq n \leq n } \left(\frac{n}{n}\frac{n+\alpha+\beta+1}{n}\frac{2n+\alpha+\beta+1}{n}\right ) \right]^{1/2}.\end{gathered}\ ] ] therefore , we have @xmath162 * derivation : + we recall that @xmath301 from these equalities , we obtain @xmath302 which allows us to write @xmath303 the term in between brackets can be computed using a simple element decomposition and leads to @xmath304 one can deduce that @xmath305 which gives the desired result , thanks to @xmath306 .
* integration : + let us recall that @xmath307 then , @xmath308 or @xmath309 by taking @xmath310 , we obtain the primitive given by the propostion [ p : opmatrices ] , besides , since @xmath311 , this primitive vanishes at @xmath12 . *
division by @xmath175 : + we will use the following formula : @xmath312 from which @xmath313 but we saw previously that @xmath314 so @xmath315 the term in between brackets can be computed using the following simple element decomposition : @xmath316 the sum presents lots of cancellations and is equal to @xmath317 finally , @xmath318 notice that this proof can also be achieved using a christoffel - darboux formula .
* multiplication by @xmath175 : + we recall that @xmath319 then , @xmath320 or @xmath321
|
a new spectral method is built resorting to @xmath0 jacobi polynomials .
we describe the origin and the properties of these polynomials .
this choice of polynomials is motivated by their orthogonality properties with the respect to the weight @xmath1 used in spherical geometry .
new results about jacobi - gauss - lobatto quadratures are proven , leading to a discrete jacobi transform .
numerical tests for poisson problems in a sphere are presented using the c++ library lorene .
spectral methods , jacobi polynomials , numerical relativity
|
we consider a cellular mobile network in which each base station ( bs ) is equipped with a large - scale antenna array whose size @xmath0 is much greater than the number of single - antenna mobile users it serves .
such a large - scale ( distributed ) multiple - input , multiple - output ( mimo ) system or a massive mimo system for short has the potentiality of achieving transmission rate much higher than those offered by current cellular systems with enhanced reliability and drastically improved power efficiency .
it takes advantage of the so - called channel - hardening effect @xcite which implies that the channel vectors seen by different users tend to be mutually orthogonal and frequency - independent @xcite . as a result ,
linear receivers are near - optimal in the uplink and simple multiuser ( mu ) precoders are sufficient to guarantee satisfactory downlink performance .
we consider a single cell within such a network whose bs array size @xmath1 , @xmath2 being the number of single antenna mobile users within its coverage range , and refer to this system as a distributed massive mimo system . to estimate a massive mimo channel
, we can employ the conventional pilot - assisted channel estimators ( ces ) such as least - squares ( ls ) @xcite or minimum mean square error ( mmse ) estimators @xcite .
taking advantage of the massive mimo channels spatial sparseness , ces based on compressive sensing ( cs ) have been proposed @xcite .
however , the complexity of cs recovery is high and cs - based methods often call for the use of randomized pilots to produce a sensing matrix which satisfies the restricted isometry property ( rip ) .
they also assume that the channel has limited paths with separable angles of arrivals ( aoas ) .
in fact , there are not many articles consider mimo channels with a cluster of stochastic aoas around a mean angle .
to circumvent such a limitation , a complicated bayesian learning approach with a gaussian - mixture channel model , was developed in @xcite .
assuming the channel can be represented by a nonnegative combination of limited number of elements from the so - called atomic set , @xcite presented a denoising - based ce . by exploiting the transform domain sparseness , @xcite presented transform domain ces that use only significant components in the transform domain to reduce the noise effect and computing complexity .
while @xcite and @xcite explored various transforms , @xcite and the references therein focused on the discrete fourier transform ( dft ) domain . in this paper , we extend and analyze the performance of the transform - domain approach that exploits the spatial correlation - induced channel rank reduction .
we show that with a judicial choice of the unitary transform and a proper selection of the dominant subspace , the resulting ce gives excellent performance as it involves only the critical channel dimensions which is much smaller than that required by the conventional methods @xcite .
in other words , the efficiency of a transform - domain ce depends not only on the unitary basis ( transform ) used but on how one selects a proper subset of the basis vectors which span the dominant channel subspace .
furthermore , we verify that improved performance is attainable if an additional angle alignment operation is in place so that the receive array points to the mean aoa .
we discuss the effect of the basis ( transform ) used has on the ce s mse performance and explain the near - optimality of the type-2 discrete cosine transform ( dct-2 ) basis .
our analysis enables us to find the relations between some channel parameters such as channel rank , array size , signal - to - noise ratio ( snr ) , and the mse performance for a given basis .
it also leads to a robust rank determination algorithm .
the proposed ce , as will be seen , can provide additional information about the mean aoa as well .
this information is very useful in designing a downlink precoder . due to space limitation
, we focus on the scenario where each mobile station ( ms ) served use a single omnidirectional antenna and there is only one dominant uplink arrival cluster for every ms - bs link . the rest of this paper is organized as follows . in section [ section :
model ] , a brief review on the system and channel models is given and the advantage of using one of the models is explained .
section [ sec : ssfc ] presents the proposed ce and its mse analysis , discusses possible implementation complexity reduction , and provides a different perspective on our approach . in section [ sec : basis ]
we study the basis selection issue from the perspective of transform coding and analyze various system and channel parameters impacts on the estimator s mse performance in section [ sec : rr_mo ] .
we present a rank and the corresponding subspace indices determination algorithm in section [ subsec : mod ] .
simulation results are provided in section [ section : simulation ] to validate the superiority of our ce and rank determination algorithm .
we summarize our main contributions in section [ section : conclusion ] . _
notation : _
@xmath3 represents the operator that stacks columns of the enclosed matrix , whereas @xmath4 , @xmath5 , @xmath6 , and @xmath7 denote respectively the expectation , vector @xmath8-norm , matrix spectral norm , and frobenius norm of the enclosed item .
we denote by @xmath9 , @xmath10 , and @xmath11 , the @xmath12 identity matrix and @xmath13-dimensional all - one and all - zero column vectors .
@xmath14 translates the enclosed vector or elements into a diagonal matrix with the diagonal terms being the enclosed items .
throughout this paper , we consider a single - cell massive mu - mimo system having an @xmath0-antenna bs and @xmath2 single - antenna mss , where @xmath15 .
we assume a narrowband ( nb ) communication environment in which a transmitted signal suffers from both large- and small - scale fading .
we denote @xmath2 ms - bs link ranges by @xmath16 and assume that each uplink packet place its pilot of length @xmath17 at the same time location .
given the initial frequency and time synchronization has been achieved , we express the corresponding @xmath18 noise variance normalized received samples at the bs , arranged in matrix and frame - synchronized form , @xmath19 $ ] as rcl = _ k=1^k_k_k^h+ = _ ^+ , [ eq : sys_mod ] where @xmath20\in \mathbb{c}^{m\times k}$ ] and @xmath21 contain respectively the small - scale fading coefficient ( ssfc ) vectors @xmath22 and large - scale fading coefficient ( lsfc ) vector @xmath23^t$ ] that characterize the @xmath2 uplink vector channels and @xmath24 $ ] is the white gaussian noise matrix with independent identically distributed ( i.i.d . )
elements , @xmath25 .
each element of the lsfc vector , @xmath26 , is the product of random variable @xmath27 representing the shadowing effect and the pathloss @xmath28 , @xmath29 .
assume that @xmath27 s are i.i.d .
lognormal random variables , i.e. , @xmath30 .
the @xmath31 matrix @xmath32^h$ ] , where @xmath33 , consists of orthogonal uplink pilot vectors @xmath34 .
the optimality of using orthogonal pilots was verified in @xcite .
a normalized nb uplink ssfc vector can be expressed as @xmath35 where @xmath36 , @xmath37 denotes the number of incoming clusters , @xmath38 the number of subpaths in a cluster , @xmath39 the steering vector associated with aoa @xmath40 , and @xmath41 with @xmath42 being the mean aoa of the @xmath43th cluster and @xmath44 the corresponding as .
we consider only the single - cluster case , i.e. , @xmath45 , the extension to multi - cluster channels is touched upon in section [ section : simulation ] where an example is given to validate the extendibility of the proposed approach .
it is reasonable to assume that the mobile users are relatively far apart ( with respect to the signal wavelength ) so that the @xmath46th uplink ssfc vector is independent of the @xmath43th vector , @xmath47 . with mutually independent @xmath48 s ,
each of them can be represented by rcl _
k=_k^_k , where @xmath49 is the spatial correlation matrix at the bs side with respect to the @xmath46th user and @xmath50 .
we assume that the ssfc matrix @xmath51 remains constant during a pilot sequence period , i.e. , the channel s coherence time is much greater than @xmath17 , while the lsfc @xmath52 varies even slower . due to the use of orthogonal pilots and
the fact that the channel estimation procedure is identical for all users , we shall omit the user index @xmath46 henceforth . in @xcite , two analytic correlated mimo channel models were proposed .
these models generalize and encompass as special cases , among others , the kronecker @xcite , virtual representation @xcite , and weichselberger @xcite models .
they often admit flexible reduced - rank ( rr ) representations . for a single - input , multiple - output ( simo ) channel the analytic mimo models of @xcite degenerates to the following obvious result .
[ lemma : reducerank ] the channel vector seen by the @xmath46th user can be approximated by @xmath53 or alternately by the phase - modulated ( angle - aligned ) model @xmath54 where @xmath55 \in \mathbb{c}^{m\times m}$ ] is a predetermined ( unitary ) basis matrix , @xmath56^t,{\mathbf{c}}=[c_1,c_2 , \cdots , c_m]^t
\in \mathbb{c}^{m\times 1}$ ] are the transformed channel vectors for the @xmath46th ms - bs link , and @xmath57 is diagonal with unit magnitude entries whose phases depend on a single phase parameter @xmath58 .
the approximations become exact when @xmath59 .
the model ( [ c_only ] ) seems to be trivial and obvious , after all @xmath60 is just a transform - coded version of @xmath61 .
the not - so - obvious part is that it induces the decomposition of the spatial correlation matrix @xmath62 where @xmath63 .
similarly , with @xmath64 and @xmath65 , the phase - modulated ( pm ) channel vector , ( [ rrapp ] ) implies rcl ^h()()&=&_k__k_k_^h = \{rr^h}.[eq : c_mtx ] both ( [ c_only_1 ] ) and ( [ eq : c_mtx ] ) provide an alternate and perhaps better conceptual interpretation on our modeling , i.e. , we attempt to approximate the spatial correlation matrix by a two - dimensional ( 2d ) separable transform of a lower - rank matrix . as ( [ c_only_1 ] ) can be rewritten as @xmath66 , we are in fact performing a 2d transform coding on @xmath67 to obtain a compressed representation .
the interpretation of ( [ eq : c_mtx ] ) , on the other hand , is given in the following remark .
further elaboration from the transform coding perspective is given in section [ sec : basis ] where we discuss the choice of the basis matrix @xmath68 .
[ remark : lpm_merits ] we claim that the rank - reduction capability of the pm model ( [ rrapp ] ) can potentially be better than that of ( [ c_only ] ) .
we argue as following .
let @xmath69 be the pdf of the aoa and @xmath70^t$ ] be the cartesian coordinates of the @xmath71th antenna element .
without loss of generality , we assume that antenna elements @xmath71 and @xmath72 lie on the @xmath73-axis and the impinging waveform spreads over @xmath74 $ ] , where @xmath58 is the mean aoa and @xmath75 the as . following @xcite , we can show that , for antenna spacing @xmath76 and small @xmath75 , rcl [ ] _
ij&=&\{h_i h_j^*}=_-^p(+ ) e^-j(+ ) + & & e^-j
_ -^p(+)e^-j + & & e^-j_ij [ eq : phi_ij ] where @xmath77_{ij}$ ] is real if @xmath69 is symmetric about @xmath58 .
this then implies that , for a uniform linear array ( ula ) with antenna spacing @xmath78 and incoming signal wavelength @xmath79 , if we set @xmath80_{ii}=\exp\left[-j2\pi \frac{(i-1)\xi } { \lambda } \sin\phi\right]$ ] , @xmath81 is linearly pm ( lpm ) and @xmath82 is approximately real . hence , its _ real _ dimension is effectively only _ half _ that of @xmath67 .
for other array geometries , if the diagonal elements of @xmath57 form the steering vector associated with the mean aoa , the same conclusion holds .
a direct implication of the this remark is that the mean aoa of each user link is extractable if the associated as is not large which , as reported in a recent measurement campaign @xcite , is likely to be the case at the bs .
the mse of a channel estimator based on ( [ rrapp ] ) , as to be proved later , is also smaller as a result of a smaller required modeling order .
we begin with the channel model ( [ rrapp ] ) and denote by @xmath83 the modeling error .
let @xmath84 and assume that an @xmath0-element ula with antenna spacing @xmath78 is used . then
, when lsfcs are perfectly known , the post - detection pilot - matched filter output vector is given by rcl [ mf output ] & = & ^2 + + & & ( ( ) _ m+ ) + . if the modeling error @xmath85 is negligible , the ml joint estimator for @xmath86 is the solution of rcl [ prob:28 ] _ , & & -()_m^2 + & & ( ) = ( _ 1 ( ) , , _ m ( ) ) , + & & _
i()=(-j2 ) and is of a least - squares ( ls ) form similar to ( 17 ) of @xcite . with @xmath87 and @xmath88 , the optimal solution to ( [ prob:28 ] )
is rcl & = & ^h ^h_m ( ) _ m^ ( ) + & = & ( ( ) _ m ) ^h^2 , [ eqn : w_hati ] + & = & _ m^ ( ) = _
m^h ^h ( ) .
[ c_hat ] when the true lsfc is not available we use its estimate , @xmath89 , and obtain the ssfc estimate @xmath90 where the subscript @xmath91 for @xmath92 emphasizes that the ssfc vector estimate is obtained by using a rank-@xmath91 basis matrix @xmath93 . both ( [ eqn : w_hati ] ) and ( [ c_hat ] ) require no matrix inversion while @xmath94 can be found by a line search . because of @xmath95
, we shall refer to ( [ h_hat ] ) as the lpm - aided ( rr ) channel estimator .
the one based on ( [ c_only ] ) , called the regular rr channel estimator , is simply @xmath96 equation ( [ c_hat ] ) can be further rewritten as @xmath97 , where @xmath98 is an @xmath99 matrix .
if @xmath100 contains time - domain samples and a unit - modulus basis is used , i.e. , @xmath68 consists of unit - modulus column vectors and so does @xmath101 , then @xmath102 can be realized with @xmath91 instead of @xmath0 rf chains and the remaining operations can be performed digitally at the baseband , making possible a hybrid receive array structure .
similarly , ( [ eqn : w_hati ] ) also allows such a hybrid implementation in searching for the mean aoa .
the implementation complexity is reduced significantly if @xmath103 . when there is a limitation on the number of rf chains , our rr channel estimator is particularly desirable .
if @xmath91 exceeds the available rf chain number , say @xmath104 , a serial realization is possible if the channel s coherent time is greater than @xmath105 periods and an @xmath37-period pilot sequence is sent .
the receiver performs similar operations at each period with a basis matrix that consists of non - overlapping columns of @xmath101 .
as @xmath106 in a noiseless environment , the operation @xmath107 gives us yet another interpretation of our approach : spatial filtering the channel vector @xmath61 by a filter bank , where each column of @xmath101 represents an @xmath0-element spatial filter ( receive beamforming vector ) .
we shall address the basis ( matrix ) selection issue in section [ sec : basis ] . from the spatial search viewpoint , ( [ eqn : w_hati ] ) tries to search for the mean aoa by finding which spatial filter bank has the maximum squared sum output magnitude .
this seems not to be a very plausible approach after all , the equivalent filter bank is pointing at multiple directions simultaneously .
using the sum output without looking at individual filter ( beam ) outputs is equivalent to determining the aoa by the sum spatial response which may yield poor spatial resolution .
however , reflecting on ( [ eqn : w_hati ] ) and the description of ( [ eq : chnl_mod ] ) should make it clear that , unlike some known methods which assume a single or multiple separable , deterministic incoming plane wave , each with a fixed aoa @xcite , we are searching for the _ mean _ aoa of a group of plane waves with random incident angles around the mean aoa .
unless the as is much smaller than the beamwidth of a single filter ( beam ) , the channel vector will induce responses ( leakage ) in multiple filters .
the sum output is resulted from a cluster of steering directions centered around a mean aoa with an as determined by @xmath91 , the modeling order .
[ rem : sp_filter ]
let @xmath108 , where @xmath109 is the unitary group of degree @xmath0 , i.e. , the set of all @xmath110 unitary matrices . performing unitary transform @xmath111 on the lpm channel vector @xmath112 , we obtain @xmath113 which is equal to @xmath114 in ( [ rrapp ] ) when @xmath59 .
for a reason to become clear later we refer to the correlation matrix of @xmath115 rcl \{^h^h } = ^h [ cu ] as the _ bias matrix _ , where @xmath116 $ ] .
equation ( [ cu ] ) indicates that the bias matrix is the ( separable ) 2d unitary transform of the correlation matrix @xmath117 .
let @xmath118_{\ell\ell}$ ] , @xmath119 , be the diagonal entries of @xmath120 , i.e. , the energy spectrum of the transformed channel vector . invoking the identity rcl [ ] _
= _ ^ * ( ) [ ] _ _ ( ) = _ = 1 , we obtain rcl _ = 1^m_^2&= & ( ) = ( ^h)=_=1^m [ ] _ = m , an alternate expression to re - confirm that a unitary transform is energy - preserving , and @xmath121 .
we note that the 2d unitary transform in effect redistributes the energy ( or variance ) of @xmath61 ( and @xmath112 ) .
the lpm operator @xmath122 , besides helping extract the mean aoa information , reshapes the channel s energy spectrum . assuming perfect lsfc information and compensation , we substitute ( [ c_hat ] ) into ( [ h_hat ] ) to obtain rcl _ m&= & ( ) _ m_m^h ^h()[eq : rr_ce_yp ] + & = & ( ) _ m_m^h ^h()(+ ) + & & \{_m}+ , [ eqn : h_biased ] where @xmath123 . denoting by @xmath124 , the mse of the enclosed channel estimator we have the decomposition rcl ( _ m)&= & \{_m- ^2 } + & = & _ \{_m}+ _ b(_m ) + [ eqn:23 ] where @xmath125 and @xmath126 represent respectively the variance and bias of the estimator @xmath127 .
we prove in [ app : pf_mse ] that [ thm : mse ] for the channel estimator based on ( [ rrapp ] ) @xmath127 , with @xmath68 comprising @xmath91 distinct columns of @xmath111 , we have rcl \{_m}&= & [ eqn : var_term ] and rcl b(_m ) & = & ( _ m)=_^c_m [ ] _ , [ eqn : bias_term ] where @xmath128 consists of the indices of @xmath111 s column vectors that are included in @xmath68 , @xmath129 , @xmath130 is the _ bias _ matrix defined by ( [ cu ] ) with @xmath58 substituted by @xmath94 , and @xmath131 is diagonal with @xmath132_{\ell\ell}=\left\ { \begin{array}{ll } 1 , & \hbox{$\ell\in\mathcal{f}_m^c$ ; } \\ 0 , & \hbox{otherwise . }
\end{array } \right.\ ] ] using an argument similar to [ app : pf_mse ] , we verify that [ thm : mse2 ] for the regular ssfc estimator @xmath133 , rcl \{_m } & = & , [ eqn : var_term2 ] + b(_m ) & = & ( _ m ) [ eqn : bias_term2 ] where @xmath134 .
note that @xmath135 is the post - detection snr per antenna port ( i.e. , an element of the vector ( [ mf output ] ) ) and so @xmath136 can be regarded as the effective pre - detection snr per dimension . equation ( [ eqn : var_term ] ) thus says that the estimator s variance is inverse proportional to this pre - detection sample snr and is due entirely to noise @xmath137 but independent of the basis ( matrix ) used . the estimator s bias , as shown in ( [ eqn : bias_term ] ) or ( [ eqn : bias_term2 ] ) , is equal to the sum of the @xmath138 diagonal terms , indexed by @xmath139 , of the positive semidefinite matrix @xmath130 or @xmath140 and is caused by the modeling error @xmath85 .
the bias can be minimized by choosing the @xmath91 columns that span the @xmath91-dimensional subspace containing the maximum energy of @xmath61 .
when the full - rank model , i.e. , @xmath59 , is used , we have @xmath141 and @xmath142 .
both estimators become unbiased and equivalent to the conventional ls estimator @xcite rcl _
m=_m= , [ eqn : convetional_ls ] which is independent of the basis used , and give the same performance .
the mse performance for channel estimator @xmath127 based on ( [ rrapp ] ) is given by ( [ eqn:23])([eqn : bias_term ] ) .
we note that rcl b(_m)=\{-^(m)^2 } where @xmath143 is the truncated version of @xmath144 obtained by nulling its @xmath138 elements associated with @xmath139 . as @xmath145 is independent of the basis matrix , to find a basis matrix that minimizes the mean squared modeling error for a given modeling order @xmath91 is equivalent to find a bias - minimizing basis matrix @xmath146 for which the transformed channel vector would have the least average power in the truncated @xmath138 coordinates , i.e. , rcl
_ = \{^h -(^h)^(m)^2}. [ prob : minrank ] the above problem is related to the notion of _ energy compactness _ of a unitary transform in digital waveform coding @xcite .
a unitary transform @xmath111 on the vector @xmath112 redistributes the vector s energy @xmath147 as @xmath148 .
the resulting energy compactness is measured by the ( subband ) coding gain ( * ? ? ?
1 ) rcl ( ) = .
[ eq : tau ] it is known that , under certain regularity conditions , the maximizer of ( [ eq : tau ] ) is the karhunen - love transform ( klt ) @xmath149 which diagonalizes the spatial correlation matrix @xmath117 defined by ( [ eq : c_mtx ] ) .
klt is also the minimizer of the truncation error rcl _
m\{-^(m)^2 } , = .
hence we have the best dimension reduction of @xmath61 is achieved by setting @xmath68 as the one that consists of the eigenvectors associated with the largest @xmath91 eigenvalues of @xmath117 .
the use of klt basis guarantees rcl ( _ ) |_m = m_1>(_)|_m = m_2 [ mono ] for two bs antenna arrays of sizes @xmath150 .
[ cor:1 ] the asymptotic coding gain ( @xmath151 ) of a large - scale mimo system can easily be derived by the transform coding analogy as well .
_ lemma [ cor:1 ] _ reveals the strictly increasing nature of @xmath152 , therefore , the asymptotic coding gain is also a coding gain upper bound which increases as the spectrum roughens , i.e. , as the degree of spatial correlation increases .
we conclude that increasing @xmath0 and/or spatial correlation improves the rr channel modeling efficiency . despite of its modeling efficiency ,
klt basis is nonflexible in that it is channel - dependent and computationally intensive .
prior to the channel estimation , eigen - decomposition and eigenvalue ordering must be performed on the spatial correlation matrix @xmath117 , which varies from user to user and can be accurately estimated only if sufficient observations are collected . as a result ,
we resort to signal - independent predetermined bases .
the use of predetermined unitary matrices @xmath68 in ( [ c_only ] ) or ( [ rrapp ] ) avoids the needs of eigen - decomposing @xmath67 or @xmath117 and ordering the eigenvectors to obtain the associated klt basis .
two candidate bases are of special interest to us for their analogousness to the klt basis . for a correlated environment having a smooth spatial correlation function ,
polynomial basis vectors are adequate to model the spatial variation .
an orthonormal discrete polynomial basis @xmath153 $ ] is constructed via qr decomposition @xmath154 , where @xmath155_{ij}=(i-1)^{j-1}$ ] , @xmath156 . as @xmath157 s
are arranged in the ascending order of degrees , the rr basis @xmath68 is able to track more rapid variation parts of the channel vector by including more higher - degree columns .
dct , especially type-2 dct ( dct-2 or simply dct ) , is widely used for image coding for its excellent energy compaction capability @xcite . for a finite - length real sequence ,
its dct is often energy - concentrated in lower - indexed coefficients as an @xmath158-point dct is equivalent to a @xmath159-point dft of a sequence obtained by cascading the original @xmath158-point sequence with its mirrored version . as mentioned in _ remark [ remark : lpm_merits ] _ , the correlation matrix @xmath117 of the lpm channel vector is approximately real for nb channels with small as although @xmath160 is complex - valued in general .
hence , ( [ h_hat ] ) would require a modeling order @xmath91 smaller than that needed by @xmath133 if @xmath58 is known or can be reliably estimated .
we want to remark that there are fast algorithms for computing dct that require @xmath161 operations as fft does .
it is also known that the energy compaction efficiency of dct is perhaps the closest to that of klt @xcite in the high correlation regime among popular unitary transforms such as walsh - hadamard , slant , haar , and discrete legendre transforms .
the last one is a polynomial - based transform .
we compare the efficiencies of the dct and polynomial bases in representing nb mimo channels in section [ section : simulation ] .
we now examine the relations of @xmath91 with other system and channel parameters for a given basis .
recall that ( [ eqn : bias_term ] ) implies that the estimation bias is minimized by the semi - unitary basis matrix @xmath162 , which consists of the @xmath91 eigenvectors associated with the largest @xmath91 eigenvalues of @xmath117
. therefore , denote by @xmath163 the set of all @xmath99 semi - unitary matrices , we have rcl ( _ m ) = b(_m ) = ( _ ) _ m , where the first equality is because @xmath164 is independent of the basis used . for two bs antenna arrays of sizes @xmath165 and a fixed @xmath91 ,
the biases associated with the channel estimator ( [ h_hat ] ) satisfy rcl b(_m)|_m = m_1 & < &
b(_m ) |_m = m_2 , if @xmath166 and @xmath167 . the normalized mse ( nmse ) also has the monotonicity property : rcl ( _ m)|_m = m_1 & & ( _ m)|_m = m_1 + & < & ( _ m)|_m = m_2 . [ remark:11 ] to analyze the effect of the modeling order on the mse performance of the estimator , we define the _ optimal _ modeling order @xmath168 is the one that minimizes @xmath169 when @xmath167 or @xmath170 .
[ def : opt_mod_ord ] we denote by @xmath171 the average pre - detection pilot snr per spatial sample at the bs and rewrite the estimator mse as rcl ( _ m ) = + ( _ m ) = m - m(1- ) , + [ eqn:27 ] where the first equality is resulted from the zero spatial correlation assumption ( i.e. , @xmath172 ) .
the above equation predicts that , for low snr ( @xmath173 ) , a larger modeling order does not help as one just tries to fit more parameters to noise - dominant observations .
however , if @xmath174 , the mse improves with increasing modeling order @xmath91 .
we conclude that the optimal modeling order for uncorrelated channels is @xmath0 if @xmath175 .
as indicated by ( [ eqn:23])([eqn : bias_term ] ) and section [ sec : basis ] , to minimize the estimation mse , the energy compactness of @xmath120 has to be considered . for a given basis ,
the diagonal terms become more and more nonuniform with increasing spatial correlation . by _ definition [ def : opt_mod_ord ] _ the optimal modeling order @xmath176 is given by @xmath177_{\ell\ell}\right).\ ] ] the above optimal modeling order is difficult to determine especially when @xmath178 . however , the channel spectrum @xmath179_{\ell\ell}$ ] or @xmath180_{\ell\ell}$ ] , as shown in fig .
[ fig : sketch_d ] , is usually concentrated in a relative narrow band ( set of consecutive frequency indices ) .
furthermore , the lpm channel spectrum ( solid curve ) , as predicted in _ remark [ remark : lpm_merits ] _ and section [ sec : dct ] ) , is a lowpass one and whose bandwidth is only about half that of the regular channel spectrum without lpm , i.e. , this figure also verifies the advantage of the model ( [ rrapp ] ) over ( [ c_only ] ) . for both models ,
the _ passband _ boundaries are located at the bottom of the ` waterfall ' regions and most of the channel energy lies within the subspace spanned by those columns of @xmath111 whose indices belong to the corresponding _ passband_. to simplify the task of determining the optimal modeling order , we henceforth restrict @xmath128 to be of the form @xmath181 for some @xmath182 and define the bias - reduction modeling order @xmath183 and the associated subspace index set @xmath184 of efficiency @xmath185 , @xmath186 , are rcl ( m _ , _ ) = \{m : > } [ eq : m_c ] where each candidate @xmath187 consists of @xmath91 consecutive integers in @xmath188 . for the convenience of subsequent discussion
, we refer to the subspace @xmath189 as the _ dominant subspace _ and @xmath187 the _ dominant support_. [ def : m_c ] note that @xmath183 and @xmath184 do not require the snr information and results in negligible energy loss if @xmath185 is chosen to be close to @xmath190 , say @xmath191 .
it is is also much easier to find @xmath183 than @xmath176 .
this definition can be extended to model ( [ c_only ] ) directly by replacing @xmath120 with @xmath140 .
both modeling orders reflect the energy compactness of the transformed channel spectrum ; they decrease when the spatial correlation increases or equivalently , the dominant rank of @xmath67 decreases .
furthermore , _ definition [ def : m_c ] _ and ( [ eqn : bias_term ] ) imply that , for any @xmath192 , rcl < 1- , m m _ , i.e. , using a dominant support @xmath187 containing @xmath184 ensures that @xmath193 .
[ rm : mse_obs ] when @xmath194 , increasing @xmath91 beyond @xmath183 can not improve and , in the variance - dominant region , @xmath195 , even degrades the mse performance .
this is because in this region when @xmath196 , the mse grows approximate linearly with @xmath91 , i.e. , rcl ( _ m)\{_m } = .
therefore , in the low snr ( @xmath197 ) regime , for @xmath198 , rcl ( _ m_1)<(_m_2 ) . [ obs:1 ] on the other hand , in the bias - dominant region , mse increases with decreasing @xmath91 for @xmath199 . this can be easily seen by observing that when @xmath200 rcl ( _ m_1)-(_m_2 ) & & _ ^c_m_1^c_m_2[]_>0 .
[ obs:2 ] this remark will be verified in section [ section : simulation ] via simulation with some typical correlated channels .
we now investigate the snr influence on the optimal modeling order .
_ lemma [ thm : mse ] _ says that , for a given modeling order and @xmath187 pair , @xmath201 depends on snr only , therefore , we have for a fixed modelling order @xmath202 , @xmath203 if and only if @xmath204 . on the other hand , it can be shown that if @xmath205 , then @xmath206 , where rcl ( m_j,_m_j)&=&_(m,_m)(_m)|_=_j , [ eqn : m1 ] and so for two average pre - detection snrs , @xmath207 , the corresponding mse - minimizing solutions rcl ( m_j,_m_j)&=&_(m,_m)(_m)|_=_j , j=1,2 , [ eqn : m2 ] satisfy @xmath208 .
[ lemma : mse_snr ] since @xmath209 and @xmath210 are respectively the mse - minimizing orders corresponding to @xmath211 and @xmath212 and @xmath213 , rcl & & _ ^c_m_1 [ ] _ -_^c_m_2 [ ] _ + & = & \ { rl 0 , & m_1=m_2 ; + _ ^c_m_1^c_m_2[]_,&m_1<m_2 ; + -_^c_m_2^c_m_1[]_,&m_1>m_2 . . for @xmath212 ,
@xmath214 gives rcl & & \
{ rl 0 , & m_1=m_2 ; + _ ^c_m_1^c_m_2[]_,&m_1<m_2 ; + -_^c_m_2^c_m_1[]_,&m_1>m_2 . .
thus , it is possible that @xmath215 and the other two possibilities lead to rcl \ { rl _ 1 _ 2,&m_1<m_2 ; + _ 2 _ 1,&m_1>m_2 . .
the case @xmath216 results in a contradiction to the assumption that @xmath207 , we thus conclude that @xmath217 .
* initialization : * set @xmath218 and termination criterion @xmath219 . * ( updating bias matrix ) calculate + rcl = ^h^h ( ) ( ) . *
( finding dominant subspace ) solve ( [ eq : m_c ] ) with @xmath220 to obtain an estimate of @xmath221 and construct @xmath68 . * ( updating mean aoa ) with @xmath222 , find @xmath94 via ( [ eqn : w_hati ] ) . *
( recursion ) go to step * 1 * ; or terminate and output @xmath223 , @xmath94 , and @xmath68 if @xmath224 is satisfied .
* initialization : * set step size @xmath225 , adjustment ratio @xmath226 , threshold @xmath227 , window size @xmath228 , dct basis @xmath111 , and termination criterion @xmath219 * ( coarse estimation ) randomly draw a @xmath94 from @xmath229 $ ] . repeat until @xmath94 satisfies @xmath230 . *
( fine adjustment ) let @xmath231 and @xmath232 . find @xmath233 _
= m - w+1^m q_^h^h()()q_. * ( recursion ) let @xmath234 and go to step * 2 * ; or terminate and output @xmath94 , if @xmath219 is satisfied . in discussing the optimal or near - optimal modeling order
, we assume that the true bias matrix @xmath235 or at least its diagonal entries ( the channel s energy spectrum ) is known .
as @xmath58 is not known a priori , we propose a joint rank and mean aoa ( @xmath58 ) estimation method called iterative modeling order determination ( imod ) algorithm by assuming known @xmath67 whose estimation is addressed later in this section . as shown in the ensuing section , @xmath94 and @xmath236 do not need accurate estimate of @xmath67 , hence neither does @xmath237 .
the imod algorithm is summarized as * algorithm * [ alg : imod ] . if we are interested in finding the mean aoa only , an alternate search method we refer to as * algorithm * 2 , which computes the channel spectrum sequentially , is more efficient .
this algorithm is based on the fact mentioned in section [ ssec : ana_corrch ] that the @xmath58-aligned channel spectrum with respect to the dct basis is a lowpass one ( cf .
[ fig : sketch_d ] ) and the dominant support is the set @xmath238 .
once the modeling order and mean aoa are determined , we then proceed to estimate the channel vector according to ( [ h_hat ] ) or ( [ h_hat ] ) , which needs to know the dominant rank and support only instead of the complete information about @xmath67 . compared with other rank estimation methods
@xcite , the imod algorithm finds @xmath239 that gives near - minimum mse with much less complexity and faster convergence .
in contrast to the singular value decomposition ( svd)-based methods @xcite which are applicable only if the klt basis is used , the imod method is basis - independent and do not have to perform svd . in many cases , the spatial correlation rcl & = & ( \ { ( ) ^h}-^2_m ) + & & ( -^2_m)[eqn : est_scm ] is not available and has to be estimated prior to applying the imod algorithm .
an estimate can be derived by first noticing that the ml estimate for @xmath240 is rcl = _ i=1^j_i_i ( _ i_i)^h[eq : psi ] where @xmath241 and @xmath242 denote respectively the samples received and pilot transmitted at the @xmath71th training period , and @xmath243 is the number of pilot periods used . substituting ( [ eq : psi ] ) into ( [ eqn : est_scm ] ) , we then obtain an estimate of @xmath67 .
unlike the klt - based estimate which needs accurate spatial correlation knowledge , for a predetermined basis ce , all one has to know is the index range where most of the channel energy @xmath244_{\ell\ell}$ ] lies , the detail shape within the range is of not much concern ; see ( [ eq : m_c ] ) .
since the spatial correlation and lsfcs often vary slowly in time , we can collect enough pilot samples from multiple training blocks to obtain their reliable estimates .
simulation results reveal that @xmath245 or 3 is sufficient in most cases .
the complete rr channel estimation process is described in fig .
[ fig : rank_det_bl_diag ] .
[ rm : chcorr_aoa ] it is important at this point to mention two known properties governing the mean aoa , as , spatial correlation , and channel rank .
the first property is that the spatial correlation decreases as ula element spacing and/or the as @xmath75 increases @xcite .
the second property further incorporates the effect of mean aoa @xmath58 , stating that the asymptotic normalized channel rank @xmath246 is given by ( * ? ? ?
2 ) @xmath247 where @xmath248 .
the above equation indicates that for a fixed @xmath58 , @xmath249 increases as @xmath75 becomes larger and ; for a fixed @xmath75 , a @xmath58 that is closer to @xmath250 leads to a larger @xmath249 . in the finite @xmath0 regime such behaviors
are also verified in and in the following section .
in this section , we investigate the effects of key system design parameters such as the basis used , modeling order , and channel parameters as snr ( [ eq : snr ] ) , channel correlation , as , and mean aoa on the rr channel estimator performance . unless otherwise specified , we assume a @xmath251-element ula with antenna spacing @xmath252 and use the standardized scm channel model @xcite with a single cluster that contains @xmath253 subpaths . in section [ sec : dct ]
, we have mentioned that the estimator @xmath127 and @xmath133 differ in that the former attempts to redistribute the average energy of an lpm channel vector having an approximately real - valued correlation matrix in a reduced - dimension subspace while the latter has to deal with one that has complex - valued spatial correlations .
we therefore predicted that the lpm based estimator @xmath127 performs better than @xmath133 , the one without the lpm operation , in that the former offers more reduction on rank .
[ fig : sketch_d ] confirms such a prediction ; it shows that one needs to find only one index @xmath239 instead of two required by @xmath133 which also has a inferior rank reduction capability .
mean aoa estimate is a by - product of @xmath127 .
, @xmath254 @xmath255 , known mean aoa @xmath256.,width=360 ] _ { \ell\ell}$ ] ) seen by different bases .
the left and right columns of sub - figures assume @xmath255 and @xmath257 , respectively .
the top , middle , and bottom rows are the spectra with respect to klt , dct , and polynomial bases.,width=360 ] fig .
[ fig : as72aoa60_as15aoa0_klt_dct_poly_bdiag ] shows the energy spectra of highly ( @xmath255 ) and moderately ( @xmath258 ) correlated channels with respect to the klt , dct , and polynomial bases after lpm operation .
for all three bases considered , the channel rank is an increasing function of the as and a decreasing function of the mean aoa @xmath58 .
this trend has been predicted in _ remark [ rm : chcorr_aoa]_. as expected , the klt basis is indeed the most efficient in that it requires the least modeling order @xmath91 , among the three bases , to characterize the @xmath0-dimensional channel .
the dct and polynomial bases give similar performance when as is small but the polynomial basis degrades significantly at @xmath258 . with known lsfcs ; @xmath255 , @xmath256 . both theoretical mse performance with known @xmath58
( solid curves ) and simulated counterpart ( dashed curves ) are shown .
also shown is the theoretical performance as a function of @xmath58 when the modeling order is @xmath259.,width=345 ] s simulated ( dashed ) and theoretical ( solid curves ) mse performance with known lsfcs ; @xmath258 , @xmath260 .
the impact of @xmath58 on a fixed-@xmath91 estimator is also shown.,width=345 ] the effect of the modeling order on the dct- and polynomial - based channel estimators mse performance is examined in figs .
[ fig : model_order_72_poly ] and [ fig : model_order_15_dct ] for @xmath261 and @xmath262 , respectively .
when the as is small ( @xmath263 ) , the spatial correlation is relatively high ( cf . _
remark [ rm : chcorr_aoa ] _ ) so we need a small modeling order
[ fig : model_order_72_poly ] shows that over - modeling the channel ( as @xmath91 increases from @xmath253 to @xmath251 ) in fact degrades the mse performance .
but when @xmath264 increases to @xmath257 , the spatial correlation decreases and the spatial channel roughens , hence a larger @xmath91 is called for , especially at higher snr ; see fig .
[ fig : model_order_15_dct ] . in both figures
we also show the mse performance ( solid curves without markers ) for some selected mean aoas .
we find that , for a fixed as and snr , the optimal modeling order is a decreasing function of @xmath265 , which is consistent with what _ remark [ rm : chcorr_aoa ] _ and fig . [
fig : as72aoa60_as15aoa0_klt_dct_poly_bdiag ] have predicted : the channel correlation increases with decreasing @xmath265 . ) of the rr channel estimator versus modeling order and @xmath197 ; @xmath255 , known mean aoa @xmath256.,width=360 ] ) behaviors of the rr channel estimator versus @xmath91 and @xmath197 ; @xmath258 , known mean aoa @xmath260.,width=360 ] in figs .
[ fig : wc_vs_c_compare_as72 ] and [ fig : wc_vs_c_compare_as15 ] , we plot the bias , variance , and mse with respect to modeling order @xmath91 of the estimator ( [ h_hat ] ) in a high and moderate spatial correlated channels .
the mse - minimizing modeling order @xmath176 , for each scenario corresponding to different bases and snrs can be found in these two figures .
_ lemma [ lemma : mse_snr ] _ is numerically verified and the associated @xmath176 and @xmath239 are listed in tables [ tab : imod_72 ] and [ tab : imod_15 ] . when snr becomes larger , mse becomes more bias - dominant ( cf .
_ remark [ rm : mse_obs ] _ ) and so @xmath185 should be chosen to be closer to @xmath190 .
when @xmath185 is so adaptive to snr , @xmath239 is close to @xmath266 and we have near - optimal mse performance . as expected , the klt - based estimator requires the smallest modeling order and achieves the best mse performance .
0.05 in [ tab : imod_72 ] @xmath197&@xmath250 db & @xmath267 db&@xmath253 db + @xmath266 & @xmath268&@xmath267&@xmath267 + @xmath239 ( @xmath58 , @xmath67 known ) & 8(@xmath269)&9(@xmath270)&10 ( @xmath271 ) + @xmath272 ( @xmath67 known ) & @xmath273(@xmath269)&@xmath274 ( @xmath270)&@xmath275 ( @xmath271 ) + @xmath272 ( @xmath276 , @xmath277 ) & @xmath278(@xmath279)&@xmath280 ( @xmath269)&@xmath281 ( @xmath270 ) + + @xmath197&@xmath250 db & @xmath267 db&@xmath253 db + @xmath266 & @xmath282&@xmath283&@xmath284 + @xmath239 ( @xmath58 , @xmath67 known ) & 11(@xmath269)&12 ( @xmath270)&14 ( @xmath271 ) + @xmath272 ( @xmath67 known ) & @xmath285(@xmath269)&@xmath286 ( @xmath270)&@xmath287 ( @xmath271 ) + @xmath272 & @xmath274(@xmath279)&@xmath288 ( @xmath269 ) & @xmath289 ( @xmath270 ) + 0.05 in [ tab : imod_15 ] @xmath197&@xmath250 db & @xmath267 db&@xmath253 db + @xmath266 & @xmath290&@xmath291&@xmath292 + @xmath239 ( @xmath58 , @xmath67 known ) & 27(@xmath269)&28 ( @xmath270)&30 ( @xmath271 ) + @xmath272 ( @xmath67 known ) & @xmath293(@xmath269)&@xmath294 ( @xmath270)&@xmath295 ( @xmath271 ) + @xmath272 & @xmath296(@xmath279)&@xmath297 ( @xmath269 ) & @xmath298 ( @xmath270 ) + + @xmath197&@xmath250 db & @xmath267 db&@xmath253 db + @xmath266 & @xmath291&@xmath299&@xmath300 + @xmath239 ( @xmath58 , @xmath67 known ) & 27(@xmath269)&36 ( @xmath270)&56 ( @xmath271 ) + @xmath272 ( @xmath67 known ) & @xmath301 ( @xmath269)&@xmath302 ( @xmath270)&@xmath303 ( @xmath271 ) + @xmath272 ( @xmath276 , @xmath277 ) & @xmath304(@xmath279)&@xmath305 ( @xmath269)&@xmath306 ( @xmath270 ) + db ; @xmath255 , and @xmath256 ( known ) .
estimated lpm spectra in the left column of sub - figures are obtained with @xmath307 or @xmath308 samples , whereas @xmath267 or @xmath309 in the right .
the sub - figures in the top , middle , and bottom rows are the spectra with respect to the klt , dct , and polynomial bases .
the blue solid curves correspond to the results of the true @xmath67.,width=336 ] db ; @xmath258 , and @xmath260 ( known ) .
estimated spectra in the left sub - figures are obtained with @xmath307 or @xmath308 samples , whereas @xmath267 or @xmath309 samples are used elsewhere .
the sub - figures in the top , middle , and bottom rows are the spectra with respect to the klt , dct , and polynomial bases .
the blue solid curves are associated with the true @xmath67.,width=336 ] also shown in table [ tab : imod_72 ] and [ tab : imod_15 ] are the average @xmath272 s obtained via the imod algorithm , which usually converges in just a few ( no more than five ) iterations . a more realistic scenario when @xmath67 is estimated by substituting @xmath240 in ( [ eqn : est_scm ] ) with ( [ eq : psi ] ) using @xmath277 periods samples
is considered there as well . figs .
[ fig : as72_sub_50 ] and [ fig : as15_sub_50 ] show the corresponding estimated channel spectra . as @xmath310 is obtained by averaging @xmath243 outer product matrices and has a rank less than or equal to @xmath243 , we can not find a proper estimate of @xmath239 for klt basis when the true dominant rank exceeds the number of sample periods .
this is a shortcoming of the klt approach when the associated correlation matrix has to be estimated by averaging small time - domain samples .
this sample - deficient problem exists for other similar rank determination methods @xcite but is of much less concern for the predetermined basis approach we have adopted .
db . the left and right sub - figures assume @xmath261 and @xmath262 and those in the top , middle , and bottom rows are generated by the true channel vector , mmse @xcite , and rr estimate ( [ h_hat ] ) , respectively . red and blue curves represents respectively the results with perfect @xmath67 or its estimated counterpart using @xmath277 samples.,width=336 ] we now verify the effectiveness of our rr channel estimator from the receiver s viewpoint which is similar to but slightly different from what was discussed in _ remark [ rem : sp_filter]_. we notice that a maximum - ratio combining ( mrc ) receiver would first multiply a received vector @xmath311 by @xmath312 , the pseudoinverse of the channel estimate @xmath92 , before making hard or soft symbol decisions .
this operation is equivalent to receive beamforming .
a receive beamformer which acts as a spatial filter should conceivably match the incoming signal beam pattern : it should point toward the mean aoa with a ( main ) beamwidth approximately equal to the 2-sided as so that the dominant part of the signal carried by the channel vector is filtered with minimum distortion while noise or interference from other directions are suppressed .
this prediction is confirmed by fig .
[ fig : ch_bp_as72_as15 ] ; see ( [ eq : rr_ce_yp ] ) . although the mmse channel estimator @xcite @xmath313 yields performance similar to the rr estimators when @xmath314 is perfectly known , it is sensitive to the correlation matrix estimation error .
in contrast , our estimator is more robust as it does not need complete and accurate information about @xmath314 , all it needs is the dominant rank of the channel spectrum and the associated dominant support . finally , to demonstrate that our channel estimator is applicable to multi - cluster channels , we show two examples in fig . [ fig:2path_dft_bdiag_i ] . in the upper sub - figure ,
the two clusters are separable both spatially , i.e. , @xmath315=[52.8^\circ,67.2^\circ]$ ] and @xmath316= [ 0^\circ,30^\circ]$ ] , and in the dft domain . as a result ,
the dominant support consists of two disjoint sets , @xmath317\cup[{m_\eta}_2,{m_\eta}_3]$ ] , so the multi - cluster channel can be estimated via ( [ c_only ] ) with @xmath318_{\ell\in\mathcal{f}_m}$ ] . in the lower sub - figure we consider a channel with two overlapped clusters which are inseparable . by treating this channel as a single - cluster channel
, we still can apply our scheme with dominant support @xmath319 $ ] to obtain a ce .
-cluster channels with respect to the dft basis .
( upper part ) @xmath320 and @xmath321 ; ( bottom ) @xmath322 and @xmath323.,width=360 ]
in this paper , we extend a model - based rr channel estimator for use in spatially - correlated massive mimo systems .
the two proposed estimators rank - reduction nature and looser spatial correlation information requirement result in performance improvement over conventional ls and mmse estimators and lower complexity in post - channel - estimation signal processing .
one of the estimator the lpm - aided one obtain a mean aoa estimate as a by - product .
the lpm operation is shown to offer another benefit of further rank reduction when the dct basis is employed , resulting in near - optimal energy compaction and performance .
we analyze the impacts of and interrelations among the critical design issues such as basis selection and rank determination and other system / channel parameters ( e.g. , snr , spatial correlation , mean aoa and as ) on the estimators mse performance .
the analytic relation between the channel s energy spectrum and the estimator s mse enable us to develop efficient rank and/or mean aoa estimation methods .
viewing the channel estimation problem from different perspectives helps casting new insights into the problem of concern .
although we focus our discussion on single - cluster channels , we briefly demonstrate the feasibility of extending our approach to more general multi - cluster channels .
in the following , we derive the variance and bias terms of the mse of @xmath127 ( [ rrapp ] ) .
those of @xmath133 for the regular model ( [ c_only ] ) can be similarly obtained .
we start with 25 f. rusek , d. persson , b. k. lau , e. g. larsson , t. l. marzetta , o. edfors , and f. tufvesson , scaling up mimo : opportunities and challenges with very large arrays , " _ ieee signal proces . mag .
_ , vol . 30 , no . 1 , pp .
4060 , jan .
2013 .
m. biguesh and a. b. gershman , training - based mimo channel estimation : a study of estimator tradeoffs and optimal training signals , " _ ieee trans .
signal process .
3 , pp . 884893 , mar .
h. yin , d. gesbert , m. filippou , and y. liu , a coordinated approach to channel estimation in large - scale multiple - antenna systems , " _ ieee j. sel .
areas commun .
2 , pp . 264273 , feb . 2013 .
wen , s. jin , k .- k .
wong , j .- c .
chen , and p. ting ,
channel estimation for massive mimo using gaussian - mixture bayesian learning , " _ ieee trans .
wireless commun .
3 , pp . 13561368 , mar .
2015 .
q. zhang , x. zhu , y. yang , p. shang , and j. liu , efficient transform - domain channel estimation technique for large - scale multiple antenna systems , " in _ proc .
ieee pimrc 2014 _ , pp .
448452 , washington , d.c .
2014 .
v. kristem , s. sangodoyin , c. u. bas , m. kske , j. lee , c. schneider , g. sommerkorn , j. zhang , r. thom , and a. f. molisch ,
3d mimo outdoor to indoor macro / micro - cellular channel measurements and modeling , " in _ proc .
globecom 2015 _ , san diego , ca , dec .
shiu , g. j. foschini , m. j. gans , and j. m. kahn , fading correlation and its effect on the capacity of multielementantenna systems , " _ ieee trans .
3 , pp . 502513 , mar .
a. m. sayeed , deconstructing multiantenna fading channels , " _ ieee trans .
signal process .
10 , pp . 25632579 , oct .
w. weichselberger , m. herdin , h. zcelik , and e. bonek , a stochastic mimo channel model with joint correlation of both link ends , " _ ieee trans .
wireless commun .
_ , vol . 5 , no .
1 , pp . 90100 , jan . 2006 .
p. r. haddad and a. n. akansu , _ multiresolution signal decomposition , second edition : transforms , subbands , and wavelets _ , academic press , 2000 .
n. ahmed , t. natarajan , and k. r. rao , discrete cosine transform , " _ ieee trans .
c-23 , no .
1 , pp . 9093 , jan . 1974 .
j. a. tsai , r. m. buehrer , and b. d. woerner , the impact of aoa energy distribution on the spatial fading correlation of linear antenna array , " in _ proc .
ieee vtc 2002 _ , vol .
2 , pp . 933937 , birmingham , al , may 2002 .
|
we present two reduced - rank channel estimators for large - scale multiple - input , multiple - output ( mimo ) systems and analyze their mean square error ( mse ) performance . taking advantage of the channel s
transform domain sparseness , the estimators yield outstanding performance and may offer additional mean angle - of - arrival ( aoa ) information . it is shown that , for the estimators to be effective , one has to select a proper predetermined unitary basis ( transform ) and be able to determine the dominant channel rank and the associated subspace .
further mse analysis reveals the relations among the array size , channel rank , signal - to - noise ratio ( snr ) , and the estimators performance .
it provides rationales for the proposed rank determination and mean aoa estimation algorithms as well . an angle alignment operation included in one of our channel models is proved to be effective in further reducing the required rank , shifting the dominant basis vectors range ( channel spectrum ) and improving the estimator s performance when a suitable basis is used .
we also draw insightful analogies among mimo channel modeling , transform coding , parallel spatial search , and receive beamforming .
computer experiment results are provided to examine the numerical effects of various estimator parameters and the advantages of the proposed channel estimators and rank determination method .
channel estimation , channel rank , channel spatial correlation , massive mimo , transform - domain approach .
|
the magnetorotational instability ( mri ) is probably the main cause of turbulence and accretion in sufficiently ionized astrophysical disks @xcite , which has inspired searches for mri in taylor - couette flow .
experiments on magnetized couette flow aiming to observe mri have been performed @xcite .
those experiments have demonstrated mri - like modes , but not yet for background flows that approximate keplerian disks .
@xcite observed nonaxisymmetric modes and enhanced transport in a spherical experiment , but with a hydrodynamically linearly unstable background state .
@xcite demonstrated unsteady behavior in a very low - rm experiment with a strong toroidal as well as axial field ; the relationship between the experimental results and linear instabilities is still controversial @xcite , but if these are indeed the linear modes , then the corresponding ratio of toroidal to axial field in a thin ( thickness @xmath0 ) disk would have to be @xmath1 , where @xmath2 is the radius of the accretion disk .
some other experiments have been proposed or are still under construction @xcite .
standard mri modes will not grow unless both the rotation period and the alfvn crossing time are shorter than the timescale for magnetic diffusion .
this requires that both the magnetic reynolds number @xmath3 ( see definitions of @xmath4 , @xmath5 , @xmath6 and @xmath7 in fig . [ princeton_full_scheme ] and their values in table .
[ ta_parameter ] ) and the lundquist number @xmath8 be @xmath9 , where @xmath10 is the alfvn speed , @xmath11 is the density of liquid gallium and @xmath12 is the background magnetic field parallel to the angular velocity . because the magnetic prandtl number @xmath13 in liquid metals , where @xmath14 is the kinematic viscosity of liquid gallium , @xmath15 and fields of several kilogauss for standard mri must be achieved in typical experimental geometries .
recent linear analyses have shown that mri - like mode could grow with much reduced magnetic reynolds number and lundquist number in the presence of a helical background magnetic field @xcite , at least for cylinders of infinite or periodic axial extent ( liu et al 2006b ) .
the simulations of this paper , however , are limited to axial background fields the experiment is complicated by the large ( @xmath16 ) reynolds number and by ekman circulation and stewartson layers @xcite , even though the princeton mri experimental apparatus has been constructed to minimize the circulation by the use of independently controlled split endcaps @xcite .
it is known that ekman circulation is significantly modified when the elsasser number @xmath17 exceeds unity : @xmath18 where @xmath19 is the characteristic rotation frequency @xcite . at 100% of the maximum designed rotation rate of our experiment ( table.[ta_parameter ] ) and with @xmath20 , where @xmath21 is the external coil current ( fig .
[ princeton_full_scheme ] ) , the elsasser number @xmath22 .
there have been few published studies of the nonlinear saturation of mri in an experimental taylor - couette geometry apart from @xcite , and even fewer with realistic experimental boundary conditions . as noted in @xcite ,
the nature of saturation of mri in a couette flow is essentially different from accretion disks , in which mri is believed to saturate by turbulent reconnection @xcite . in accretion disks ,
differential rotation arises from a balance between the gravitational attraction of the accreting body and centrifugal force .
thermal and turbulent energies are probably small compared to orbital ones , so saturation is not likely to occur by modification of the mean flow profile . in experiments , however , differential rotation is imposed by viscous or other weak forces , so that mildly nonlinear mri might well change the mean rotation profile . this paper is the second of a series , following @xcite .
several idealizations made in that paper , notably the vertically periodic boundary conditions , are dispensed with here .
we discuss the linear phase and nonlinear saturation of mri in finite cylinders with realistic fluid and magnetic boundary conditions .
all simulations reported here were performed with the zeus - mp 2.0 code @xcite , which is a time - explicit , compressible , astrophysical ideal mhd parallel 3d code , to which we have added viscosity , resistivity ( with subcycling to reduce the cost of the induction equation ) for axisymmetric flows in cylindrical coordinates @xmath23 @xcite . the implementation of fully insulating and partially conducting boundary conditions are discussed in @xcite .
the computational domain is shown in fig .
[ princeton_full_scheme ] and the parameters are summarized in table [ ta_parameter ] .
six coils with dimensions as shown were used , with @xmath24 turns in the two coils nearest the midplane and @xmath25 in the rest .
they are split into two sets of three in parallel , with the upper three in series and the bottom three also in series .
currents @xmath26 were adjusted according to the experimental values . note that in the simulations the magnetic diffusivity @xmath5 is fixed to the experimental value @xmath27 ( table.[ta_parameter ] ) , however the kinematic viscosity is varied for the purpose of extrapolation from numerically tractable to experimentally realistic reynolds numbers .
the measured current waveform is displayed in fig .
[ experiment_coil ] for target currents of @xmath28 and @xmath29 .
the waveform displayed has an overshoot at the early stage , a linear decline and then a linear ramp where the controller tries to adjust the output voltage to reach the programmed set point .
this behavior is peculiar to the high current runs ( left panel of fig .
[ experiment_coil ] ) .
lower currents have a much flatter waveform ( right panel of fig . [ experiment_coil ] ) . in the simulation
a wave form like that in fig .
[ coil ] is used to approximate the experimental coil currents with ramp time @xmath30 .
since the container is made of stainless steel , not a perfectly insulator , the radial magnetic angular momentum flux ( @xmath31 ) at the cylinders need not vanish . in principle , the magnetic coupling of the fluid to the cylinders might modify the growth of the instability@xcite . fortunately , at the frequencies relevant to princeton mri experiment ( 100% run , @xmath32 ) the skin depth of stainless steel @xmath33 ( @xmath34 is the conductivity of stainless steel , see the value in table .
[ ta_parameter ] ) , which is much larger than the thickness of the steel vessel surrounding the gallium in the experiment , @xmath35 , so that the magnetic field diffuses rather easily into the boundary .
if one considers axial currents , the gallium and the steel wall act as resistors in parallel ; taking into account their conductivities and radial thicknesses , one finds that the resistances of steel walls are much larger than the resistance of the liquid gallium [ @xmath36 ; see fig .
[ princeton_full_scheme ] for the subscripts ] .
therefore , the currents carried by the steel walls can be neglected for the toroidal field , so that the @xmath37 stress at the boundary is expected to be unimportant and an insulating boundary condition suffices .
the linear growth rate and saturated final state based on partially conducting boundary condition differs slightly from the results based on perfectly insulating boundary condition with partially conducting walls , which verifies our argument .
hereafter we ignore the conductivity of the stainless steel walls , but regard them as insulating materials , that is : @xmath38 .
please note that in this whole paper the rotation profile used is @xmath39=[4000,533,1820,650]\;{\rm rpm}$ ] ( 100% run , @xmath40 , @xmath4 , @xmath41 , @xmath42 , and @xmath43 are the rotation speed of the inner cylincer , outer cylinder , inner ring and outer ring , respectively ) ( table.[ta_parameter ] ) , under which the simulations predict that the bulk flow is almost ideal couette state @xcite . and
this is the desired basic state for mri to grow .
however , the experimental scans find that only if we choose @xmath39=[400,53,146,40]\;{\rm rpm}$ ] ( 10% run , @xmath44 ) , the bulk flow profile is found to be very close to the ideal couette state with the deviation ( the boundary layer ) highly confined near both endcaps @xcite .
@xcite also shows that non - magnetic quasi - keplerian flows at reynolds numbers up to millions are essentially steady , if the boundary effects have been properly considered .
the stewartson layer at the junction of the rings may be smoothed by localized circulation and/or turbulence from these instabilities @xcite .
the rotation profile to be used in the magnetized runs would be @xmath39=[2400,320,876,240]\;{\rm rpm}$ ] ( 60% run , @xmath44 ) .
the difference of the rotation profile at the end caps between in the experiment and in the simulation could possibly be explained by the wobbling of the inner cylinder in the experimental runs due to the difficulties of aligning the inner cylinder perfectly and the gaps between the rings and cylinders @xcite . and these effects
are speculated to be more important with higher reynolds number @xmath45 as in the experiment .
unfortunately , the modern computers and codes can not afford a simulation with reynolds number as high as several million .
although this is not experimentally realizable , the low - reynolds - number results reported in this paper without wobbling , gaps and starting from an initially ideal couette state with fully insulating magnetic boundary conditions still gives us some useful hints about what is going on in the experiment .
the result maybe is even closer to what is happening in the experiment based on the purely hydrodynamic experiment ( steady flow , mostly ideal couette state in the bulk ) , although this argument needs to be proven .
we are addressing the problem step by step , adding the complexity into the code one by one . to include all subtle points at one time would perplex the problem and make it difficult for us to catch the essence of the problem .
the detailed study of the influence of the wobbling and gaps between the rings is on the way to investigate the discrepancies between the experimental results and simulations .
also we want to point out : as demonstrated by @xcite , the viscosity of liquid metals is so small as to be almost irrelevant to mri , at least in the linear regime .
thus a low - reynolds - number run would be fine , at least in this sense .
most of results presented in this paper are the results of 100% run except stated explicitly .
the simulation results have proved very valuable in design , operation and understanding of the experiment .
the outline of this paper is as follows .
we present the results of the linear mri in sec .
[ linear ] . and
the results of the nonlinear saturation of mri are discussed in sec .
[ nonlinear ] .
[ summary ] summarizes the results and presents the final conclusions .
the first convincing evidence of the existence of mri is its linear growth rate ( fig .
[ princeton_ring2_br_outside ] ) . we find this linear growth rate is reduced from @xmath46 with ideal couette state at both endcaps ( left panel of fig .
[ princeton_ring2_br_outside ] , the end effect is removed completely by enforcing ideal couette state at both endcaps , thus no ekman circulation is present ) to @xmath47 with two rings just as in the real experiment ( right panel of fig .
[ princeton_ring2_br_outside ] ) .
this is due to the residual magnetic ekman circulation , which modifies the background flow .
also as pointed out in @xcite , the linear growth rate of an absolute instability in a bounded geometry is reduced by the absorping " end plate , with the reduction extent proportional to @xmath48 , where the aspect ratio is @xmath49 and @xmath50 is the height of the cylinders ( see fig .
[ princeton_full_scheme ] ) .
interestingly the linear growth rate with ideal couette state at both endcaps ( right panel of fig .
[ princeton_ring2_br_outside ] ) matches the growth rate of the fastest growing mode , which naturally is the dominant mode , got from a linear code developed by @xcite , in which vertical periodicity is assumed , but the radial equations are solved directly by finite differences with perfectly insulating boundary conditions ( table [ princeton_compare_with_global ] ) .
the agreement suggests that viscous effects are slight , since the no - slip conditions on the horizontal velocities at the endcaps differ from the periodicity imposed by the linear code .
mri essentially is a weak field instability .
it is characteristic of mri for a strong magnetic field to suppress this instability .
our experimental facility can only allow @xmath51 .
thus we need to try carefully to find one set of proper parameters under which one growing mode is present with low @xmath21 while absent with larger @xmath21 .
[ suppress ] demonstrates this property for simulations in which the boundary rotation rates @xmath52 are scaled to 45% of their designed values ( table.[ta_parameter ] ) .
for @xmath53 , the final state is not steady .
typical time averaged flow and field patterns are shown in fig .
[ princeton_ring2_final_pattern ] .
the poloidal flux and stream functions are defined so that @xmath54 which imply @xmath55 and @xmath56 .
the divergence of the velocity field in these compressible but subsonic simulations is nonzero but small .
we note that the induced toroidal field is around @xmath57 of the initial axially imposed magnetic field at this magnetic reynolds number @xmath58 : @xmath59 .
the most striking feature is the inflowing `` jet '' centered near @xmath60 in fig .
[ princeton_ring2_final_pattern ] ( see also fig . [ ring2_100_1000_vr ] ) , which is opposite to the usual ekman circulation @xcite .
it seems that the rapid outflowing jet " found in @xcite with vertically periodic boundary condition , where the position of the jet " is arbitrary , is shifted to the boundary layer , near the end caps in the experiment .
the inward jet " near the inner cylinder part is a direct consequence of mri . for mri - stable regimes ,
@xcite discuss the influence of independently controlled split endcaps upon the poloidal circulation and the influence of the axial magnetic field and reynolds number upon the stewartson and ekman layers .
in such regimes , an _ outflowing _ jet occurs near the inner cylinder , in our apparatus , as it does in purely hydrodynamic ekman circulation ( i.e. , @xmath61 ) .
however in the mri - unstable regime for the same elsasser number @xmath17 , the middle clock - wise cell enlarges both horizontally and vertically and eventually dominates the other cells , which retreat to the corners if the instability is strong enough .
this leads to one inflowing jet " near the inner cylinder ( fig .
[ ring2_100_1000_vr ] ) . further evidence that mri causes
the inflowing jet was found in simulations with the ideal couette state imposed at the end caps , to remove ekman circulation ( fig .
[ princeton_couette_stream ] ) .
[ princeton_couette_stream ] shows a clear inflowing jet near the inner cylinder in this case ( see also fig .
[ ideal_100_1000_vr ] ) .
this suggests that the poloidal circulation seen in the final state in the split - endcap cases ( fig .
[ princeton_ring2_final_pattern ] ( b ) ) is caused solely by saturation of mri rather than ( magnetically modified ) ekman circulation .
astrophysicists are interested in the angular momentum transport due to mri since mri is supposed to be the most probable mechanism to explain the fast accretion in the astrophysical disks .
[ princeton_ring2_integral_torque_r ] displays the time - averaged @xmath2-profiles of the radial advective , viscous , and magnetic torques , i.e. the angular momentum fluxes integrated over cylinders coaxial with the boundaries : @xmath62\\[1ex ] \gamma_{\rm total , r}(r)&= & \gamma_{\rm advective , r}(r)+\gamma_{\rm magnetic , r}(r)+\gamma_{\rm viscous , r}(r)\ , , \end{aligned}\ ] ] where @xmath63 and @xmath64 are the radial and vertical cell sizes respectively .
in contrast to the final state for vertically periodic boundary conditions @xcite , the total radial torque is not constant with radius . since our numerical scheme conserves angular momentum
exactly , we can infer a vertical flux arising from exchange of angular momentum with the endcaps ( see also fig .
[ princeton_ring2_integral_torque_z ] ) . in the statistical steady state
( nonlinearly saturated final state ) the sum of the angular momentum flux on all boundaries should be zero , at least on time average . in this case it is fluctuating since
the final state is not steady given high reynolds number . from the gradients of the radial torque , we identify four ekman circulation cells : where @xmath65 ( @xmath66 ) , the fluid is losing ( gaining ) angular momentum at the endcaps and the boundary - layer flow is therefore radially inward ( outward ) .
this is consistent with the discussion in @xcite of the poloidal circulation driven by two split rings .
the radial magnetic and advective torques vanish at @xmath6 and @xmath7 because of the boundary conditions but are important at intermediate radii , especially the advective .
all components of the radial torque are positive , which means that the angular momentum is transported radially outwards .
[ princeton_ring2_integral_torque_z ] displays the time - averaged @xmath2-distributions of the vertical total angular momentum flux at both endcaps @xmath67 : @xmath68\,.\ ] ] since at both endcaps , both the advective and magnetic angular momentum fluxes are zero due to the boundary conditions , the total angular momentum at both endcaps @xmath69 are simply the viscous angular momentum flux @xmath70 .
from these two figures , we can see that the global angular momentum is entering from the inner cylinder then most of it is flowing out from the outer cylinder while the rest of it is flowing out at the two end caps .
_ angular momentum is transported radially outward .
_ it is very interesting to derive the dependence of the total torque at the inner cylinder , which is mainly responsible for driving the rotation , and of the sum of the vertical torques at both endcaps , on reynolds number ( @xmath71 ) . from fig .
[ 100_ratio ] ( a ) , we infer the following scalings ( @xmath72 run , @xmath73 ) : \(1 ) in the absence of magnetic field ( 100% run , @xmath74 ) , @xmath75 these hydrodynamic states ( rather than ideal couette flow ) will be used as the initial conditions for the mri experiments .
the maximum relative error ratio , defined as the ratio of the absolute difference of the value from the simulation and the fitted value ( eq .
[ torque_scaling1 ] ) over the former one , is @xmath76 .
\(2 ) for @xmath20 and @xmath22 ( 100% run , @xmath74 ) . @xmath77
this corresponds to the final states of the mri experiment .
the maximum relative error ratio is @xmath78 .
from these two scaling laws , _ the mri indeed enhances the angular momentum transport at saturation , though slightly _
( see discussion below ) .
the enhancement is far beyond the arrange of the error bar ( see the error bars on panel ( a ) of fig . [ 100_ratio ] ) .
it is reasonable for @xmath79 in both magnetized and unmagnetized cases to decrease with increasing @xmath71 since the viscous coupling to the walls scales as @xmath80 in standard unmagnetized ekman layers ( _ i.e. _ the endcaps are corotating solidly with the outer cylinder ) . in this more complicated case ( with split rings ) , we also expect a similar relationship , though the magnetized ekman and stewartson layers complicate the problem @xcite . if eq .
[ torque_scaling1 ] and eq .
[ torque_scaling2 ] also work at larger reynolds numbers , so that they may safely be extrapolated to the experimental reynolds number ( @xmath81 , 100% run ) , then the total radial torque of the initial and final state at the inner cylinder may be as large as @xmath82}}\nonumber\ ] ] and @xmath83}}\nonumber\ ] ] respectively .
thus , the ratio of the increase of the torque over the initial torque is : @xmath84 , which is quite measurable and indicates that at the experimental reynolds number mri would dominate the residual magnetic ekman circulation in the point of view of transporting the angular momentum .
there are , however , reasons for caution in accepting this estimate . for example , the experimental flow may be three - dimensional and turbulent , which might result in an even higher torque in the final state , and the absolute values of both the exponents seem to decrease at larger reynolds number and the difference of these two exponents is small .
these concerns all make the extrapolation of eq .
[ torque_scaling1 ] and eq .
[ torque_scaling2 ] to the experimental reynolds number a bit risky .
nevertheless , we expect a noticeable torque enhancement in the mri - unstable regime . from fig .
[ 100_ratio ] ( b ) , we can see that : ( 1 ) at larger reynolds number , @xmath85 is increased , which means that larger part of the total angular momentum is transported vertically .
( 2 ) in the mri stable regime ( @xmath86 ) , the magnetic field enhances the vertical transport of the angular momentum .
this is also reasonable since the magnetic field would align the flow , thus having the cells elongating and penetrating deeper into the bulk .
the size of the middle cells is increased vertically by the residual magnetic ekman circulation .
these are consistent with the conclusions deduced in @xcite .
( 3 ) in the mri unstable regime ( @xmath87 ) , the onset of the mri results in more angular momentum transported radially outwards and less vertically .
the mri would increase the scale of the middle cell horizontally .
therefore it transports more angular momentum radially outwards .
in conclusion of this paper we have simulated the nonlinear development of magnetorotational instability in a nonideal magnetohydrodynamic taylor - couette flow .
the simulations mimic an on - going experiment except that the conductivity of the stainless steel walls is neglected and the simulation is started from an ideal couette state rather than an actual hydrodynamic statistical steady state driven by split end caps .
we have also restricted our study to smaller fluid reynolds number ( @xmath71 ) than in the experiment , however we have used exactly the same magnetic reynolds number ( @xmath58 ) .
mri grows from small amplitudes at rates in good agreement with linear analyses without the end cap effects . 1 .
a clear linear phase is observed ; the linear mri growth rate is reduced by the residual magnetized ekman circulation .
2 . strong magnetic field suppresses mri .
3 . in the final state
one inflowing jet " opposite to the usual ekman circulation jet " @xcite is found near the inner cylinder , a direct consequence of mri rather than the residual magnetic ekman circulation ( 100% run ) .
4 . the mri enhances the angular momentum transport at saturation .
( 100% run ) .
the final state contains horizontal fields about @xmath57 as large as the initial vertical field for @xmath88 ( 100% run ) .
we emphasize that these conclusions are based on axisymmetric simulations restricted to the range @xmath89 with the idealizations mentioned above .
the simulation results considering the conductivity of the steel container , starting from an actual hydrodynamical equilibrium and the comparison with the experimental results would be given in a future paper .
the author would like to sincerely thank jeremy goodman and hantao ji for their very inspiring discussion and constructive comments .
the author would also like to thank james stone for the advice on the zeus code and thank stephen jardin for the advice of implementing full insulating boundary condition .
this work was supported by the us department of energy , nasa under grants atp03 - 0084 - 0106 and apra04 - 0000 - 0152 and also by the national science foundation under grant ast-0205903 .
|
in preparation for an experimental study of magnetorotational instability ( mri ) in liquid metal , we present non - ideal axisymmetric magnetohydrodynamic simulations of the nonlinear evolution of mri in the experimental geometry .
the simulations adopt fully insulating boundary conditions .
no - slip conditions are imposed at all boundaries .
a clear linear phase is observed with reduced linear growth rate .
mri results in an inflowing jet " near the midplane and enhances the angular momentum transport at saturation .
|
although the bulges of disk - dominated galaxies can appear to be the remnants of long - ago mergers , their kinematics may point to close ties with ongoing formation processes in the disk ( for reviews , see * ? ? ?
* ; * ? ? ?
* and references therein ) .
the strong correlation of bulge and disk color seen in local galaxies @xcite suggests a shared history , with similar ages and metallicities in their stellar populations .
so it is important to probe bulge and disk colors at significant look - back times .
of particular interest might be the link between bulge and central black hole mass , for example , as that correlation plausibly develops at an epoch when bulges are first undergoing collapse or mergers , providing fuel for the black hole @xcite .
separating the bulge from disk in images of high - redshift field galaxies is a difficult task though , because bulges are small ; typically less than a few kpc across in the local universe @xcite . at redshift
@xmath4 each kpc corresponds to only 016 in a concordance cosmology ( @xmath5 km @xmath6 mpc , @xmath7 , @xmath8 ) which we adopt throughout .
the _ hubble space telescope _ ( hst ) has been used to decompose many @xmath9 field galaxies into bulges and disks in the optical @xcite .
but equally high resolution near - infrared ( nir ) color maps would also be valuable , as they should lessen contamination by dust and probe older stellar populations , especially for galaxies with @xmath10 where the @xmath11 filter desirably samples rest - frame @xmath12 , which is below the 4000 break .
our initial efforts to do high - resolution nir photometry of @xmath0 galaxies with natural - guide - star ( ngs ) ao were hampered by poor control over the point - spread function ( psf ) @xcite .
better results are now possible with the new generation of laser guide - star ( lgs ) ao systems .
the motivation for lgs ao is to increase sky coverage by providing a pointable artificial beacon on which to guide .
lgs ao will also limit anisoplanatism , and thus improve image quality , by allowing the wavefront sensor and deformable mirror to be aligned with the target , instead of a bright nearby star .
one complication is that the laser spot can only be used to sense high - order aberrations , not phase tip and tilt .
thus , a nearby star , which can be several magnitudes fainter than that used in full ngs mode , is still needed to account for image motion . a separate sensor guides on this tip - tilt star " and drives a fast steering mirror . despite this added complexity , tests with the lick lgs
ao system show that the resulting point - spread function ( psf ) is no harder to constrain than with a conventional ngs system @xcite .
here we discuss our ongoing effort to employ the latest ao technology to obtain images of high - redshift field galaxies .
this work is part of the center for adaptive optics ( cfao ) treasury survey ( cats ) @xcite .
we have employed keck lgs ao in the extended groth strip ( egs ) , an hst field extensively imaged in both the optical and near infrared .
the data are presented in section [ data ] , including a description of the improved psf provided by the laser and the comparison of ao and hst images .
details of photometry and image decomposition follow for two @xmath3 galaxies in the field in section [ photometry ] , and color gradients between the bulge and disk are discussed further in section [ discussion ] .
the ao field was chosen from the egs where hst data exist from several programs . besides wfpc2 data in f606w and f814w ( go-5090 : pi - groth , go-8698 : pi - mould ) , acs wide field camera ( wfc ) data in the equivalent filters and nicmos nic3 in f160w have been taken ( go-10134 : pi - davis ) .
we make use of the acs data in our analysis , as these are deeper and have finer pixel sampling than the previous wfpc2 data .
exposures were 2260 s in acs f606w , 2100 s in acs f814w , and 640 s in nicmos f160w , enough to provide excellent @xmath13 ( @xmath14 ) in the peak pixel ( 005 for acs and 020 for nicmos ) for galaxies brighter than @xmath15 mag .
all hst images were processed with the standard pipelines .
the reduced acs data were kindly provided to us by j. lotz , and the reduced nicmos data by s. kassin .
fluxes were converted to vega magnitudes using the zeropoints in the data handbooks for acs @xcite and nicmos @xcite .
we will henceforth refer to the hst filters f606w , f814w , and f160w as @xmath11 , @xmath16 , and @xmath17 , respectively .
the egs acs data were taken in a rectangular mosaic designed to avoid bright stars .
thus , no @xmath18 mag star is available to serve as a natural guide star ( ngs ) for the keck ao system .
several stars in egs are , however , bright enough ( @xmath19 ) for tip - tilt guiding in the lgs mode , and one of the stars happens to lie within one of the parallel nicmos pointings . using this @xmath20 star to guide tip - tilt correction
, we obtained keck ii lgs - mode ao observations with the nirc2 camera with the @xmath2 filter on 2 march 2005 .
see @xcite and @xcite for more information on the keck ii lgs ao system and its operation .
individual exposures were 120 s , obtained in a nonrepeating circular dither pattern of 4 radius , for a total integration of 5640 s. for these small dithers no re - acquisition is necessary with each laser move , and so overheads are no more than for ngs mode . the center of the dither pattern was aligned with the center of the nicmos field , which required an 18 offset between the laser spot and tip - tilt star .
exposures were obtained with nirc2 with 004 pixel wide - field mode ( 40 @xmath21 40 ) which slightly undersamples the keck diffraction limit in the nir ( @xmath22 at 1.6@xmath1 m ) .
this is a good match to nicmos nic3 field ( 50 @xmath21 50 ) although the 020 nic3 pixels significantly undersample the hst psf ( @xmath23 at 1.6@xmath1 m ) .
data reductions closely followed those described in our earlier ngs work @xcite .
a photometric zeropoint of @xmath24 ( from the keck lgs ao nirc2 webpages ) was used to convert to standard johnson @xmath2 .
apart from the tip - tilt guide star , we detect seven objects with keck ao in the region of overlapping hst coverage .
one is a star , bright but unsaturated , which is suitable as a psf estimator for hst .
three are galaxies of known redshift , listed in table [ table_targets ] .
two of these were measured as part of the deep2 redshift survey @xcite .
a third has a photometric redshift obtained from the canada - france - hawaii telescope legacy survey @xcite .
there is another galaxy in the acs images of unknown , although presumably low , redshift .
it is only marginally detected at @xmath2 .
no redshifts are published for the remaining two objects either , and they are too faint in all bands for our analysis . the hst psf is well characterized .
the field distortions of acs and nicmos are modest and can be modeled in detail with the tiny tim software @xcite .
the psf strehl ratio - the peak intensity relative to that of the ideal diffraction pattern - is high , approaching @xmath25 for nicmos nic3 @xmath17 at optimal focus @xcite .
so to first order the fwhm will describe the psf , and differences in strehl ratio can indicate psf variation over the field .
the hst psf fwhm was measured on the bright but unsaturated star in the field using a gaussian fit with the iraf task imexam .
the acs wfc has a psf fwhm of 011 in @xmath11 and 010 in @xmath16 .
the nicmos nic3 @xmath17 psf has @xmath26 .
variation in psf fwhm for the acs data was estimated by comparing the psf star in the field with another suitable star 74 to the southwest , outside the overlapping nicmos coverage .
this was found to differ in fwhm by less than 001 ( @xmath27% in @xmath11 , 6% in @xmath16 ) .
we can not measure psf variation in the nicmos field as there is only one unsaturated star available , but variation in psf fwhm over the nic3 field is also known to be slight , less than 002 ( @xmath28% ; barker et al . , 2007 ) .
variation in the psf strehl ratio for our acs data can be estimated from the relative flux - normalized maxima in the imexam fits .
this is a difference of @xmath29 in @xmath11 and @xmath30 in @xmath16 over a range of 74 .
the largest separation of psf and target is 37 , so variation in strehl ratio over our field will be less . the case should be similar for nicmos @xmath17 , as its psf fwhm variation is comparable to acs .
variations in lgs ao performance , both in time and angular separation from the guide source , are similar to those experienced in ngs mode .
see @xcite and @xcite for a discussion of keck ii lgs ao performance characterization .
changes in seeing with time result in fluctuations in ao correction over the entire field of view .
increasing angular separation @xmath31 between the target and laser spot results in a drop in strehl ratio proportional to @xmath32 .
it also increases radial elongation of the psf towards the laser spot .
although we did not record the seeing at @xmath11 , previous experience with keck ao indicates that for the correction achieved it was @xmath3307 , and fairly stable during the night .
two steps were taken to track changes in ao performance and characterize the data .
first , we interleaved the scientific observations with short , unsaturated exposures of the tip - tilt star .
this establishes the psf at the tip - tilt guide location immediately prior to each scientific exposure .
second , we followed the scientific observations by duplicating the dither pattern used for science observations in a dense star field . the globular cluster m5 was observed , which was at comparable airmass to egs and is familiar from our previous ao calibration work @xcite .
we used a @xmath34 star for tip - tilt guiding . that this was brighter than the star used for
egs should not impact the results , as @xmath20 is already sufficient for excellent tip - tilt correction .
we applied the same offset as the science observations and duplicated the circular dither pattern .
we then shut off the laser , and continued the observations in m5 using this tip - tilt star as the ngs - mode guide for comparison with our lgs observations .
these calibration data were combined to give the frame - by - frame performance of the ao system . the strehl ratio of the unsaturated images of the tip - tilt star and another faint star visible in the scientific exposures were used to interpolate the on - axis " strehl ratio - the value at the center of each frame , coincident with the laser spot - via an @xmath32 fit .
this was repeated in each of the m5 fields . in the latter case , the fit was to all strehl ratios of isolated unsaturated stars .
measured strehl ratios are plotted in figure [ plot_strehl_ratio ] , along with the elongation of the psf at the tip - tilt guide star position . image quality improved during the observations and peaked at a strehl ratio of @xmath35 just before we moved the telescope to m5 .
these last science images also correspond to the least elongated images , and the strong anti - correlation with strehl ratio is mirrored in the calibration data .
strehl ratio and elongation data are plotted again in the lower panel of figure [ plot_strehl_ratio ] as a function of separation between the tip - tilt guide star and the laser spot .
note the expected trend towards increasing psf elongation with increasing offset .
however , one can see that strehl - ratio performance is fairly stable as a function of dither position , and so it is probably reasonable to neglect the effect of anisoplanatism due to separation of the tip - tilt star and the laser spot .
the final images of both the target and the m5 field were combined after pruning poor data . image quality and per - pixel
@xmath13 were found to be optimal by setting a cutoff for on - axis strehl ratio of @xmath36 .
this eliminated 10 of the egs frames or about a fifth of the data ; and similarly for m5 .
the strehl ratio at the center of the combined m5 frames is @xmath37 , which is consistent with the interpolated mean on - axis strehl ratio for the science frames ( @xmath38 , see figure [ plot_strehl_ratio ] ) .
it is worth emphasizing that this is based on a model fit to the on - axis strehl ratio in each frame , not a measured value there .
although the true psf field variation may differ from the model , this method allows us to uniformly compare all of the m5 and egs frames based on the best correction expected in each .
the results are shown in figure [ figure_compare ] .
the variation in image quality over the combined images can be determined from the m5 field .
although lgs observations still suffer from anisoplanatism , the image quality across the field should be improved over our previous ngs observations .
this is due to being able to move the laser spot around the field , and thus _ spread out the best correction over the combined frames_. to help characterize the improved performance of lgs over ngs we employed an analytic model of ao performance discussed in @xcite .
a synthetic starfield was generated by producing a spatially varying psf for the location of every star in each m5 frame .
this is a simple psf model comprising a diffraction - limited gaussian core and a gaussian halo into which core light is scattered as the strehl ratio diminishes .
anisoplanatic degradation of the strehl ratio is determined by the fit to the calibration data .
these fake frames were then combined and their strehl ratio determined for each isolated star in the same manner as the original data .
the results are shown along the left hand column in figure [ figure_contours ] .
these are contour plots of equal strehl ratio .
notice how the region of best correction in the lgs observation is predicted to be larger than for ngs and closer to the center of the field .
this is due to dithering the laser spot in a broad circle around the center of the field , which is indicated by the dashed curve .
the observed results are shown in the right - hand panels .
the advantage of this lgs geometry is demonstrated by inferring the strehl ratio at the positions of the three targets in the field .
the greatest disparity in strehl ratio between any two target positions is roughly @xmath39 for ngs .
for lgs , correction is both better and more uniform , with a maximum disparity of @xmath40 .
there is clearly some spatial dependence not accounted for in the model , but this could be due to , for example , static aberrations in the nirc2 camera .
overall , the model seems to give a qualitatively correct picture of lgs performance , which indicates that visiting many locations in the field with the laser is the right approach to reduce psf spatial variation . as a final reduction step
, the combined nirc2 science and m5 calibration frames were resampled to match the acs pixellation .
no smoothing was done prior to resampling .
figure [ figure_psf_plot ] shows a radial plot of the psf star in the science frames .
this has a fwhm of 016 , measured with a gaussian using the iraf task imexam .
although this is a radial average , the elongation of the psf is slight , less than 0.2 . the psf fwhm is smaller at the positions of the three targets : for gss 044_6719 it is 011 ; gss 044_6712 , 014 ; and 141811.0 + 523149 , 015 .
the acs and nicmos profiles are also plotted for comparison .
the three targets are shown in figure [ figure_panels ] along with the hst data .
each panel is 6@xmath216 .
north is up , and east is left .
the psf star is shown along the bottom row in all bands with the same scaling .
there is a good match between this psf in @xmath2 and that obtained from m5 , both having @xmath41 016 .
galaxy 141811.0 + 523149 ( @xmath42 ) seems somewhat elongated in acs @xmath11 and @xmath16 , but it is only marginally resolved in our nirc2 @xmath2 image and no disk is evident .
we therefore choose to exclude it from further analysis and instead focus on the two @xmath3 targets , both of which are resolved in all bands .
the acs @xmath11 and @xmath16 images reveal this @xmath43 galaxy to be nearly face on , with the disk easily seen in both images . also apparent
are well defined spiral arms with bright star - formation regions .
the central bulge - like core is very compact ; only a few pixels across in these images , and appears more pronounced in the @xmath16 image . with 020 pixels and a psf fwhm of 035 , the nicmos @xmath17 image does not resolve the central core .
in contrast , our @xmath2 galaxy image has a psf more comparable to that of the acs and shows a marginally resolved core with clear signs of the disk .
figure [ figure_color ] shows color images produced by combining the acs @xmath11 and @xmath16 with either nicmos @xmath17 or nirc2 @xmath2 . in each case the acs images have been degraded with a gaussian filter to match the resolution of either the nicmos or nirc2 psf .
notice that the improved resolution of nirc2 over nicmos nic3 helps reveal the separate bulge and disk components of the galaxy .
the acs @xmath11 and @xmath16 images of this @xmath44 galaxy are dramatic , clearly showing a smooth , inclined disk and an elongated core , more bulge - like in @xmath16 .
there is a broad dust lane towards the center of the disk in the @xmath16 image .
the nicmos @xmath17 image also shows the galaxy disk and a core which is elongated in the same direction as the disk .
the clump of white pixels just north of center in this image is an artifact due to a poorly removed cosmic ray .
the disk is also faintly visible in our nirc2 @xmath2 image .
the @xmath2 core is asymmetric , elongated in the same sense as in acs @xmath16 .
synthetic aperture photometry was performed on the two @xmath3 galaxies .
a circular 3 aperture was used for gss 044_6719 and a 5 aperture for the larger gss 044_6712 .
the zeropoints are those discussed in section [ data ] . the results are presented in table [ galaxy_aperture_photometry ] .
our total @xmath2 magnitude for gss 044_6719 of @xmath45 is comparable to seeing - limited photometry from palomar giving @xmath46 @xcite .
note the similar colors of the two galaxies , especially in @xmath47 . in order to probe the colors of the cores
, we repeated the photometry using a smaller circular aperture centered on the brightest pixel in each galaxy in @xmath2 .
the size of 04 was chosen to be slightly larger than the nic3 psf fhwm , and for @xmath3 corresponds to about 3 kpc in the rest frame of the galaxies .
also , for gss 044_6719 the edge of this aperture nicely falls in trough " between the bright core and the spiral arms in acs @xmath11 and @xmath16 .
we find that the cores of both galaxies are redder than their total light , especially gss 044_6712 in @xmath47 . as an improvement over our small aperture we employed gim2d ( marleau & simard 1998 , simard et al .
2002 ) to obtain separated bulge and disk photometry .
the gim2d code models a galaxy s total flux , @xmath48 , bulge and disk sizes and orientations , and galaxy pixel center from an input image , or simultaneously from multiple images .
sersic bulge plus exponential disk models are convolved with an input psf and directly compared to data during an optimization with the metropolis algorithm @xcite .
once the algorithm converges , 99% confidence intervals are determined via monte - carlo sampling of parameter space ( @xmath49 ) . a method similar to that described in @xcite
was applied , adopting a de vaucouleurs bulge ( sersic index @xmath50 ) plus exponential disk ( @xmath51 ) model and using the psfs discussed in section [ data ] .
first , we determined the center of the galaxy in the reddest band ( nirc2 @xmath2 ) , as that fit was least likely to be affected by dust or by star - forming hotspots .
then , because gim2d does not allow simultaneous fits in more than two bands , we fit the galaxy structural parameters from the @xmath16-band acs image , which has both high resolution and high signal - to - noise .
structural parameters fit were bulge and disk position angles ( pas ) and sizes ( disk @xmath52 and bulge @xmath53 ) , disk inclination @xmath54 , and bulge ellipticity @xmath55 . in all bands we kept the center fixed to
that found in @xmath2 , except for our @xmath17-band fit , in which we allowed the center to wander 005 in both the x andy y directions . because the intrinsic pixel size of this image is much larger than for the images in other bands , the @xmath17-band image was difficult to register against the other bands at the sub - pixel level . with the galaxy center and structural parameters fixed ,
the following parameters were then allowed to float in all bands : total flux , bulge - to - total flux ratio ( @xmath48 ) , and sky background .
the results are shown for gss 044_6719 in figure [ figure_models_e ] . in the left column of panels in this figure ,
the acs @xmath11 and @xmath16 , nicmos @xmath17 , and nirc2 @xmath2 data are shown .
best - fit models for each passband are displayed in the center column , and the residual images are to the right .
note how the positive residuals from modeling the acs @xmath11 and @xmath16 images highlight clumpy star formation regions that are not accounted for in our ( smooth ) models .
negative residuals in @xmath11 and @xmath16 may indicate significant dust in the disk .
for gss 044_6712 , a dust lane is prominent , and is visible even in @xmath2 . in fact , the asymmetric core hindered a robust centering of the model , preventing meaningful constraints on what could be a small bulge .
worse results were obtained by centroiding instead on the @xmath16 image , which is better resolved but less symmetric than in @xmath2 ( see figure [ figure_models_c ] ) . masking out the dust lane in the @xmath16 image
did not significantly improve the fit .
the resulting structural parameters for gss 044_6719 are shown in table [ table_parameters ] , quoted with 99% confidence limits .
the photometry of the disk and bulge components for gss 044_6719 was derived from the total fluxes and the @xmath48 .
bulge magnitudes correspond to limits that have been propagated from the total magnitude measurements and their uncertainties .
total , bulge , and disk colors are presented in table [ galaxy_photometry ] ; errors again represent 99% confidence limits .
the redness of the bulge relative to the disk is apparent in both @xmath56 and @xmath47 , but the former is likely a more reliable measurement due to the better match between nirc2 and acs psfs .
two tests were carried out on the photometry of gss 044_6719 as a check on the colors obtained from gim2d fitting .
the first test was to repeat the gim2d analysis but instead as a simultaneous fit in acs @xmath11 and @xmath16 only and using psfs derived from tinytim .
the results of the fit and the correponding photometry are shown in tables [ table_parameters ] and [ galaxy_photometry ] . that the acs - only and the combined acs , nicmos , and nirc2 datset agree on an upper limit of @xmath57
03 is reassuring , as both their psfs and gim2d fitting methods are slightly different .
a second test was to carry out aperture photometry in @xmath11 , @xmath16 , and @xmath2 in concentric annuli .
first , the acs @xmath16 image was degraded slightly with a gaussian filter to match the resolution of the nirc2 @xmath2 image .
then , the iraf task ellipse was used to measure photometry in annular apertures with outer radii ranging from 01 to 08 ( in 01 increments ) in all three bands .
near - circular apertures were used , with the center , position angle , and ellipticity of each aperture fixed to the best - fit values from gim2d disk component fits to the acs @xmath16 band image of the galaxy .
the resulting aperture photometry is given in table [ table_aperture_photometry ] .
note that only for apertures of outer radius smaller than 03 are the @xmath58 and @xmath47 colors redder than 1.35 and 2.55 magnitudes respectively , the total galaxy colors derived from gim2d fitting .
this confirms that most of the red light is confined to what gim2d finds as the bulge effective radius .
the integrated acs @xmath16 magnitude of gss 044_6719 is 21.1 , which gives an absolute rest - frame @xmath59 magnitude of -21.0 using the @xmath60-corrections of @xcite . for gss 044_6712 ( @xmath61 ) this is @xmath62 .
these @xmath63 , along with a disk scale length @xmath64 kpc found by gim2d fitting are similar to those of the three deep galaxies we observed in keck ngs mode reported in @xcite .
these are seemingly normal spirals , near the peak of the luminosity function at @xmath65 as determined in the deep2 survey @xcite .
the keck deimos spectrum from that survey also provides velocity dispersion for gss 044_6719 of @xmath66 km @xmath67 , consistent with a disk - dominated galaxy . from our gim2d fitting , we can see that while the bulge is weak in the optical ( restframe uv ) , it may comprise 20% of the @xmath2 ( restframe @xmath68 ) light .
interestingly this galaxy is detected at @xmath69 ( restframe 14 @xmath1 m ) with spitzer @xcite , indicating significant dust may be present . in an attempt to understand the very red optical to nir colors of these galaxies , and especially the bulge of gss 044_6719
, we used the galaxev code of @xcite to compare population synthesis models with the data .
galaxev constructs spectra for arbitrary star formation histories by adding together the single - stellar - population spectra from various ages , weighted appropriately .
we produced @xmath58 versus @xmath56 colors from these spectra by scaling them according to the @xcite dust attenuation curve , shifting them to @xmath3 , and then convolving this with the appropriate filter bandpasses .
we constructed two models , both with solar metalicity ( @xmath70 ) .
the first has an exponential ( @xmath71 gyr ) star formation rate , and the other passively evolves - that is , there is no active star formation - after a brief ( 50 myr ) burst .
these were also run assuming a large amount of dust ( @xmath72 ) , for a total of four models .
the results are shown in figure [ figure_bulge_colors ] , for ages of 10 myr , 100 myr , 1 gyr , 2 gyr , 4 gyr , and 7 gyr ( the age of the universe at @xmath3 for our chosen cosmology ) after the onset of star formation .
the total colors of gss 044_6719 from aperture photometry are well fit by an old @xmath73 model with no dust ; a value of @xmath74 gyr will just shift the track blueward .
the core ( 04 aperture ) is redder than the total light , and for a uniform age throughout the galaxy more closely matched by the single - burst+passive model . by adding
dust ( and keeping @xmath71 ) the same models can also hold for the other galaxy , gss 044_6712 .
however , the bulge of gss 044_6719 can only be fit by either a @xmath73 model or a single - burst+passive model after invoking significant dust .
one way to reconcile this bulge color with the total light , which seems essentially unreddened , would be a clumpy distribution of dust in the galaxy .
we could be seeing an old bulge through a dust clump in a thick disk .
the bulge is very small , and so any obscuration would affect its color much more dramatically than the disk .
the same might be true for an agn which , highly obscured , could contribute to these red colors , as well as produce significant mid - ir emission .
the combination of the high resolution and long wavelength of our nirc2 ao imaging , along with existing hst imaging , give us an opportunity to measure and interpret the optical to nir color gradients of the bulge and disk of two galaxies at @xmath3 .
aperture photometry shows that the centers of the galaxies are significantly redder than the outer regions . while we have only explored a few stellar population models in our interpretation of this , we find that the color gradient can be explained by two different star - formation histories for bulge and disk , with similar ages . however , aperture photometry is much more difficult to interpret in terms of galaxy subcomponents , as the light in any given aperture is a combination of light from the galaxy bulge and disk .
furthermore , our gim2d analysis shows that the bulge of gss 044_6719 is extremely small , so only the central aperture we used is likely to contain a significant amount of bulge light .
a much larger sample of galaxies is needed , but it seems we have established a path towards obtaining it . employing the keck laser in our nirc2 ao observations
has resolved the psf issues that troubled our ngs work and made it particularly difficult to measure and analyze galaxy subcomponent colors . we have demonstrated a method of dithering the laser guide star that stabilizes the psf across the field of view during science observations .
furthermore , our method of interleaving science and psf calibration frames , and of imaging a crowded stellar field for further calibration , allows us to much better characterize the psf in our lgs science images than in the ngs images we analyzed in our previous work .
this is important because lgs ao is quickly becoming a standard feature of large telescopes , opening up the sky to deep nir imaging surveys in archival hst faint galaxy fields .
this is the path that the cfao treasury survey is pursuing .
we gratefully acknowledge matthew barczys and shelley wright for their contributions to cats .
our thanks go to jennifer lotz and susan kassin for providing the hst images and kai noeske for providing us the palomar @xmath60 magnitude of gss 044_6719 .
we appreciate the hard - working keck staff , especially observing assistant steven magee and keck ao team members randy campbell , antonin bouchez , david le mignant ( currently at cfao ) , and peter wizinowich , who made the huge complexity of laser operations seem ( to us ) like routine observations .
we acknowledge the great cultural significance of mauna kea to native hawaiians , and express gratitude for permission to observe from its summit .
data presented herein were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california , and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w. m. keck foundation .
this work was supported by the national science foundation science and technology center for adaptive optics , managed by the university of california at santa cruz under cooperative agreement no .
ajm appreciates support from the national science foundation from grant ast-0302153 through the nsf astronomy and astrophysics postdoctoral fellows program .
andredakis , y.c .
, peletier , r.f . , & balcells , m. 1995 , , 275 , 874 baggett , s. , et al .
2002 , in hst wfpc2 data handbook , v. 4.0 , ed .
b. mobasher , baltimore , stsci barker , e. , dahlen , t. , et al .
2007 , in nicmos instrument handbook , v. 10.0 , baltimore , stsci bender , r. , burstein , d. & faber , s.m .
1992 , , 399 , 462 boehm , a. & ziegler , b.l .
2007 , asp conf .
, 379 , 278 bruzual , g. , charlot , s. 2003 , , 344 , 1000 calzetti , d. 1997 , , 113 , 162 davis , m. , et al .
2007 , , 660 , l1 de jong , r.s .
1996 , a&a .
, 313 , 45 farrarese , l. & merritt , d. 2000 , , 539 , l9 gebhardt , k. , bender , r. , bower , g. , dressler , a. , faber , s.m . ,
filippenko , a.v .
, green , r. , grillmair , c. , ho , l.c . ,
kormendy , j. , lauer , t.r .
2000 , , 539 , l13 gebhardt , k. , faber , s.m . ,
koo , d.c .
, i m , m. , simard , l. , illingworth , g.d . , phillips , a.c . , sarajedini , v.l . ,
vogt , n.p . , weiner , b. , & willmer , c.n.a .
2003 , , 597 , 239 ilbert , o. , arnouts , s. , mccracken , h.j . , bolzonella , m. , bertin , e. , le fevre , o. , mellier , y. , zamorani , g. , pello , r. , iovino , a. , tresse , l. , bottini , d. , garilli , b. , le brun , v. , maccagni , d. , picat , j.p . , scaramella , r. , scodeggio , m. , vettolani , g. , zanichelli , a. , adami , c. , bardelli , s. , cappi , a. , charlot , s. , ciliegi , p. , contini , t. , cucciati , o. , foucaud , s. , franzetti , p. , gavignaud , i. , guzzo , l. , marano , b. , marinoni , c. , mazure , a. , meneux , b. , merighi , r. , paltani , s. , pollo , a. , pozzetti , l. , radovich , m. , zucca , e. , bondi , m. , bongiorno , a. , busarello , g. , de la torre , s. , gregorini , l. , lamareille , f. , mathez , g. , merluzzi , p. , ripepi , v. , rizzo , d. , vergani , d. , astro - ph/0603217 koo , d.c . ,
simard , l. , willmer , c.n.a .
, gebhardt , k. , bouwens , r.j . ,
kauffmann , g. , crosby , t. , faber , s.m .
, harker , j. , sarajedini , v.l . , vogt , n.p . , weiner , b.j . , phillips , a.j . , i m , m. , & wu , k.l .
2005 , , 157 , 175 koo , d.c . , ammons , s.m . ,
barczys , m. , larkin , j.e . ,
max , c.e . ,
melbourne , j. , metevier , a.j .
, steinbring , e. , wright , s.a . &
the cats team 2006 , galaxy evolution across the hubble time , iau symp .
235 , held 14 - 17 august , 2006 in prague , czech republic , s235 , 233 kormendy , j. & kennicutt , r.c .
2004 , , 42 , 603 krist , j. in asp conf .
77 , astronomical data analysis software and systems iv , ed .
r. a. shaw , h. e. payne , & j. j. e. hayes ( san francisco : asp ) , 349 le mignant , d. , van dam , m.a . ,
bouchez , a.h . ,
chin , j.c.y . ,
chock , e. , campbell , r.d . , conrad , a. , doyle , s. , goodrich , r.w . ,
johansson , e.m .
, kwok , s.h . ,
lafon , r.e . ,
lyke , j.e . ,
melcher , c. , mouser , r.p .
, summers , d.m . , stomski , p.j .
jr . , wilburn , c. , wizinowich , p.l .
2006 , advances in adaptive optics ii .
ellerbroek , brent l. , & bonaccini calia , d. proc .
spie , volume 6272 , 1 maggorian , j. , tremaine . , s. , richstone , d. , bender , r. , bower , g. , dressler , a. , faber , s.m . ,
gebhardt , k. , green , r. , grillmair , c. , kormendy , j. , & lauer , t. 1998 , , 115 , 2285 marleau , f.r . ,
& simard , l. 1998 , , 507 , 585 melbourne , j. , wright , s.a . , barczys , m. , bouchez , a.h . ,
chin , j. , van dam , m.a . ,
hartman , s. , johansson , e. , koo , d.c . ,
lafon , r. , larkin , j. , le mignant , d. , lotz , j. , max , c.e . ,
pennington , d.m .
, stomski , p.j . ,
summers , d. , & wizinowich , p.l .
, , 625 , l27 metropolis , n. , rosenbluth , a. , rosenbluth , m. , teller , a. , & teller , e. 1953 , j. chem .
phys . , 21 , 1087 mobasher , b. , et al .
2006 , nicmos data handbook " , version 6.0 , ( baltimore : stsci ) noeske , k. g. 2006 , private communication pavlovsky , c. , et al .
2006 , acs data handbook " , version 5.0 , ( baltimore : stsci ) pierce , m.j . &
tully , r.b .
1992 , , 387 , 47 peletier , r.f . & balcells , m. 1996 , , 111 , 2238 salpeter 1955 , , 121 , 161 schade , d. , lilly , s.j . ,
crampton , d. , hammer , f. , le fevre , o. & tresse , l. 1995 , , 451 , l1 simard , l. , willmer , c.n.a .
, vogt , n.p . ,
sarajedini , v.l . ,
phillips , a.c . , weiner , b.j . ,
koo , d.c .
, i m , m. , illingworth , g.d . &
faber , s.m .
2002 , , 142 , 1 steinbring , e. , faber , s.m . ,
macintosh , b.a . , gavel , d. , gates , e.l .
, 2005 , , 117 , 847 steinbring , e. , metevier , a.j . ,
norton , s.a .
, raschke , l.m . ,
koo , d.c . ,
faber , s.m . ,
willmer , c.n.a . ,
larkin , j.e .
, glassman , t.m . , 2004 , , 155 , 15 steinbring , e. , faber , s.m .
, hinkley , s. , macintosh , b.a . , gavel , d. , gates , e.l . ,
christou , j.c .
, le louarn , m. , raschke , l.m . ,
severson , s.a . ,
rigaut , f. , crampton , d. , lloyd , j.p . ,
graham , j.r .
2002 , , 114 , 1267 van dam , m.a . ,
bouchez , a.h .
, le mignant , d. , johansson , e.m . , wizinowich , p.l . ,
campbell , r.d . ,
chin , j.c.y . ,
hartman , s.k . ,
lafon , r.e . ,
stomski , p.j .
, & summers , d.m . 2006 , , 118 , 310 van dam , m.a . ,
sasiela , r.j . ,
bouchez , a.h .
, le mignant , d. , campbell , r.d . ,
chin , j.c.y . ,
hartman , s.k . ,
johansson , e.m .
, lafon , r.e . , stomski , p.j .
, summers , d.m . , & wizinowich , p.l . 2006 , , in advances in adaptive optics ii
ellerbroek & d. bonaccini calia , 6272 , 95 vogt , n.p .
, koo , d.c . , phillips , a.c . , wu , k. , faber , s.m . ,
willmer , c.n.a . ,
simard , l. , weiner , b.j . ,
illingworth , g.d .
, gebhardt , k. , gronwall , c. , guzmn , r. , i m , m. , sarajedini , v. , groth , e.j . , rhodes , j. , brunner , r. , connolly , a. , szalay , a. , kron , r. , & blandford , r. 2005 , , 159 , 41 weiner , b.j . , phillips , a.c . ,
faber , s.m . ,
willmer , c.n.a . ,
vogt , n.p . ,
simard , l. , gebhardt , k. , i m , m. , koo , d.c . ,
sarajedini , v.l .
, wu , k.l . , forbes , d.a . ,
gronwall , c. , groth , e.j . ,
illingworth , g.d . ,
kron , r.g . ,
rhodes , j. , szalay , a.s . , & takamiya , m. , 620 , 595 willmer , c.n.a . ,
faber , s.m . ,
koo , d.c . ,
weiner , b.j .
, newman , j.a . ,
coil , a.l .
, connolly , a.j . , conroy , c. , cooper , m.c . ,
davis , m. , finkbeiner , d.p .
, gerke , b.f . , guhathakurta , p. , harker , j. , kaiser , n. , kassin , s. , konidaris , n.p . ,
lin , l. , luppino , g. , madgwick , d.s . ,
noeske , k.g . ,
phillips , a.c . , yan , r. 2006 , , 647 , 853 wizinowich , p.l . , le mignant , d. , bouchez , a.h . , campbell , r.d . ,
chin , j.c.y . , contos , a.r . , van dam , m.a . ,
hartman , s.k . ,
johansson , e.m .
, lafon , r.e . ,
lewis , h. , stomski , p.j . , summers , d.m . , brown , c.g . ,
danforth , p.m. , max , c.e .
, & pennington , d.m .
2006 , , 118 , 297 wyse , r.f.g . , gilmore , g. & franx , m. 1997 , , 35 , 637 ziegler , b.l .
, bohm a. , fricke , k.j .
, jager , k. , nicklas , h. , bender , r. , drory , n. gabash , a. , saglia , p. , seitz , s. , heidt , j. , mehlert , d. , mollenhoff , c. , noll , s. & sutorius , e. 2002 , , 564 , l69 lcccccccc gss 044_6719 & 30 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & @xmath82 + & 04 & @xmath83 & @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath88 & @xmath89 + gss 044_6712 & 50 & @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 + & 04 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 lcccccccccc acs @xmath11 and @xmath16 , nicmos @xmath17 , & nirc2 @xmath2 & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath113 + acs @xmath11 and @xmath16 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath110 & @xmath120 & & lccccccc total & @xmath121 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + disk & @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 + bulge & @xmath135 & @xmath136 & @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 + total & @xmath142 & @xmath143 & & & @xmath144 & & + disk & @xmath145 & @xmath146 & & & @xmath147 & & + bulge & @xmath148 & @xmath149 & & & @xmath150 & & cccccc 00 to 01 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 + 01 to 02 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 + 02 to 03 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 + 03 to 04 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 + 04 to 05 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 + 05 to 06 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 + 06 to 07 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + 07 to 08 & @xmath186 & @xmath187 & @xmath188 & @xmath189 & @xmath190
|
we have employed laser guide star ( lgs ) adaptive optics ( ao ) on the keck ii telescope to obtain near - infrared ( nir ) images in the extended groth strip ( egs ) deep galaxy survey field .
this is a continuation of our center for adaptive optics treasury survey ( cats ) program of targeting @xmath0 galaxies where existing images with the _ hubble space telescope _ ( hst ) are already in hand .
our ao field has already been imaged by the advanced camera for surveys ( acs ) and the near infared camera and multiobject spectrograph ( nicmos ) .
our ao images at 2.2@xmath1 m ( @xmath2 ) are comparable in depth to those from hst , have strehl ratios up to 0.4 , and fwhm resolutions superior to that from nicmos . by sampling the field with the lgs at different positions ,
we obtain better quality ao images than with an immovable natural guide star . as examples of the power of adding lgs ao to hst data we study the optical to nir colors and color gradients of the bulge and disk of two galaxies in the field with @xmath3 .
|
over the past decade , direct imaging surveys have discovered a small but significant number of faint , apparently planetary - mass ( @xmath220 @xmath3 ) companions that orbit their primary star hosts ( @xmath40.21.5@xmath5 ) at ultrawide separations ( @xmath6100au , extending to thousands of au ) @xcite .
these planetary mass companions ( pmcs ) represent intriguing analogs to the recent discoveries of smaller separation planets , like hr 8799 bcde @xcite , beta pic b @xcite , hd 95086 b @xcite , gj 504 b @xcite , and lkca15 b @xcite .
the large orbital separations of the pmcs are markedly different from the planets in our own solar system and the vast population of exoplanets detected with the radial velocity and transit methods , so it is not clear whether pmcs formed via similar processes .
planets at orbital radii of @xmath2100 au can feasibly be formed via traditional methods like core accretion @xcite and class ii disk instability @xcite . at wider radii ,
the most plausible process is likely disk fragmentation at the class 0/i stage(e.g .
nonetheless , pmcs are a potential boon for exoplanet studies ; their large separations make them relatively easy to observe , so they could serve as the fully characterized templates against which the more difficult measurements of traditional " planets are compared .
in particular , pmcs offer an unique window into the process of giant planet assembly and the associated formation of moon systems .
most pmcs have been found in star - forming regions with ages @xmath7myr , where gas - rich protoplanetary disks are common .
several pmcs exhibit emission lines or possible mid - infrared excesses @xcite that are commonly associated with disks and outflows .
hubble space telescope _
( _ hst _ ) data confirm that some pmcs have large optical / ultraviolet excess emission , indicative of shocks due to the accretion of disk material @xcite .
such observations demonstrate that pmcs can host their own circum(sub)stellar disks , composed of material left over from their own formation or accreted from the disk / envelope of their ( much more massive ) host .
measurements of the masses , structures , and lifetimes of these disks provide constraints on the assembly timescale for pmcs and the duration of their satellite formation epoch which could be compared with measurements of the complementary free - floating substellar population ; ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
furthermore , hydrodynamic models of giant planet formation ( e.g. , * ? ? ?
* ) make predictions of the disk radii , scale heights , and mass distributions that could be tested with spatially resolved observations , and wide companions also provide context for observations of close - in planets with ring systems , such as the recently discovered companion to 1swasp j140747.93 - 394542.6 @xcite .
however , these disks also pose challenges : if they are oriented at large viewing angles ( edge - on ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , the disk material can obscure the pmc and complicate the determination of its mass .
fw tau is a member of the taurus - auriga association , a nearby ( @xmath8 pc ) , young ( @xmath9 myr ) region of ongoing star formation .
the well - studied primary consists of a close pair ( @xmath1075mas , or 11au , separation ) of m5 stars ( fw tau ab ) with a total mass of 0.20.3@xmath5 @xcite .
fw tau ab exhibits no spectroscopic evidence of ongoing accretion @xcite , nor any clear signature of an infrared excess shortward of 24@xmath11 m @xcite , indicating that it does not host a significant optically - thick disk at orbital radii @xmath1250au .
intriguingly , @xcite noted a 4@xmath13 detection of the system at 850@xmath11 m ( @xmath14mjy ) , suggesting that a small amount of cool dust is present within a @xmath1015diameter region around the system .
recent herschel observations by @xcite also find a far - infrared excess within a similar ( or larger ) beam around fw tau ( @xmath15 , @xmath16 , and @xmath17mjy at 70 , 100 , and 160@xmath11 m , respectively ) .
@xcite first reported an extremely faint candidate ( tertiary ) companion near fw tau ab in _ hst _ optical images , with anomalously red @xmath18 photometry and strong ( narrowband ) h@xmath19 emission .
@xcite subsequently confirmed that this companion ( named either fw tau c or fw tau b , depending on its poorly - understood nature and whether it is best considered as a binary companion or planetary - type companion ) and fw tau ab were co - moving using ground - based adaptive optics imaging , and began a multi - pronged effort to better characterize its properties .
if the companion s flux can be attributed to unobscured photospheric emission , the near - infrared flux suggests a companion mass as low as 10@xmath3 , making it another example of a pmc .
however , the true mass could be significantly higher or lower if there is an excess from accretion or disk emission ( as for many disk - hosting stars and brown dwarfs ; @xcite ) and/or obscuration from an envelope or edge - on disk ( as for a number of other stellar or substellar companions ; @xcite ) .
@xcite recently obtained a near - infrared spectrum of fw tau c that confirms the previous hints of accretion signatures , but found that veiling obscures any photospheric features except broad water absorption bands , which are present for late - m or early - l dwarfs .
the spectra therefore remain consistent with either a pmc or a brown dwarf obscured by an edge - on disk .
the tentative detection of submillimeter emission by @xcite from the unresolved fw tau system hinted that a disk could be present around at least one component , and the high h@xmath19 line flux observed from the faint companion by @xcite strongly suggested that material was accreting onto the companion . in this letter , we present sensitive , component - resolved 1.3mm - wavelength observations from the atacama large millimeter array ( alma ) designed to confirm and localize the putative long - wavelength continuum emission ( to fw tau ab or to fw tau c ) and to more accurately measure the mass of the dust responsible for it .
fw tau was briefly observed with alma in cycle 1 on 2013 december 2 , using 27 available 12 m antennas in an intermediate configuration ( baseline lengths of 17460 m ) and the band 6 receivers under excellent conditions ( 0.6 mm of precipitable water vapor ) .
the correlator was configured to process four spectral windows in dual polarization , centered at 215 , 217 , 230.5 , and 233 ghz ( a mean frequency of 224 ghz , or @xmath20 mm ) , each with 128 coarse channels ( 15.625 mhz resolution ) to maximize continuum sensitivity .
observations cycled between fw tau and the nearby quasar j051002 + 180041 on @xmath107 minute intervals , with additional visits to j0522 - 3627 and ganymede for calibration purposes .
the total on - source time for the fw tau field was 22 minutes .
the raw alma visibilities were calibrated and imaged with the casa software package .
after phase correction using the water vapor radiometers , a system temperature correction , and initial flagging ( which included the rejection of data from one antenna ) , the bandpass structure in each spectral window was corrected using observations of j0522 - 3627 .
the absolute flux scaling was bootstrapped from observations of ganymede .
gain variations due to intrinsic changes in the array and atmosphere were determined from the monitoring of j051002 + 180041 and corrected .
the spectrally - averaged calibrated visibilities were fourier inverted ( assuming natural weighting ) , deconvolved with the clean algorithm , and restored with a @xmath21 ( p.a .
= 131 ) synthesized beam after a single iteration of phase - only self - calibration .
the rms noise level in the resulting image is 28 @xmath11jy beam@xmath22 .
figure [ fig : image ] shows contours of the alma 1.3 mm continuum emission overlaid on the keck / nirc2 @xmath23 image of the fw tau components ( cf .
continuum emission at 1.3 mm is firmly detected ( peak @xmath24 @xmath25 ) and centered on the faint companion , located 228 @xmath26 005 or 330 @xmath26 30 au ( p.a .
= 296@xmath262 ) from the fw tau ab photocenter ( @xmath19 = 04@xmath2729@xmath282971 , @xmath29 = + 26165282 , based on the allwise astrometry and the proper motion estimated by * ? ? ?
. a gaussian fit to the emission indicates that it is unresolved in the @xmath10200@xmath30100au beam and has an integrated @xmath31 mjy ( with an additional @xmath1010% systematic uncertainty in the flux scale ) as estimated from a gaussian model of the visibilities .
this mm - wave continuum emission is faint enough to be entirely optically thin for any reasonable dust opacity ( e.g. , * ? ? ?
* ) , so the flux measurement can be simply converted into a dust mass estimate . assuming a characteristic dust temperature of @xmath101020 k and a standard opacity of 2.3 @xmath32 g@xmath22 at 1.3 mm , we infer @xmath332 m@xmath34 .
this measurement is entirely consistent with the unresolved , marginal detection of 850 @xmath11 m emission ( @xmath35 mjy ) from the fw tau system by @xcite ; the corresponding spectrum scales like @xmath36 , in excellent agreement with the mean value for disks in this wavelength range .
the herschel fluxes from @xcite are also consistent with the far - ir sed that fits our alma flux , but do not provide a sufficient constraint on the sed shape to further constrain the dust temperature .
the upper limit to the dust emission around the fw tau ab binary , @xmath37 @xmath11jy ( @xmath38 ) , is @xmath1030@xmath30 lower than typical constraints , and corresponds to a dust mass limit of @xmath390.1 @xmath1 .
the co @xmath40=3@xmath412 transition was not detected in the very wide channels ( chosen to optimize continuum sensitivity ) .
the 3@xmath13 upper limit on the peak flux is @xmath103 mjy beam@xmath22 in a single 20 km s@xmath22-wide channel .
the alma data clearly show that all of the disk material in the system is associated solely with the faint companion , fw tau c. given the dust mass estimated from the observed 1.3 mm flux ( @xmath1013 @xmath1 ) and its complementary signatures of disk accretion @xcite , we have confirmed that the faint companion to fw tau hosts one of the least massive primordial disks known to date . if that disk proceeds to form its own system of planetary " or
satellite " companions , they will not achieve sufficient mass to become standard gas or ice giants .
however , the mass is well matched to the total mass of compact systems of sub - earth planets seen around field ultracool dwarfs ( e.g. , kepler-42 bcd ; @xcite .
it is also interesting to note that the dust mass reservoir still exceeds the total sum of the galilean satellites ( 0.066@xmath1 ; @xcite ) by just over an order of magnitude .
expectations for the disk mass distribution for very low - mass primaries remain uncertain throughout the substellar and planetary mass regime , as there are only a handful of detections and many non - detections ( e.g. , * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
however , a comparison of the infrared and millimeter fluxes to a population of free - floating young counterparts could still provide context as to the nature of the disk host as either a brown dwarf tertiary companion ( fw tau c ) or a pmc ( fw tau b ) . in figure 2 , we show
how the 3.6@xmath11 m magnitude ( converted from the @xmath42 measurement reported by * ? ? ?
* ) and the new 1.3 mm flux for the companion compare to the flux versus spectral type relations for disk - bearing stars and brown dwarfs in the taurus - auriga region @xcite . a shaded band is used for fw tau c to denote the range of possible spectral types , limited at the upper end ( @xmath10m5 ) by the presence of deep water absorption bands in the near - infrared spectrum @xcite . at 3.6@xmath11 m ,
fw tau c is fainter than nearly all known substellar companions in the region , clearly in the regime of objects with spectral types later than m9 ( @xmath43@xmath3 ) , though some edge - on disks also sit well below the median relation .
however , comparisons of its millimeter flux with the low - mass taurus - auriga population are more ambiguous .
most mm - wave surveys for brown dwarf disks were conducted with facilities that are substantially less sensitive than alma .
the best available upper limits of those surveys approach the 1.3 mm flux detected here for fw tau c , so there are at least some brown dwarf disks with masses that could be quite a bit lower .
the disk mass measured here is consistent with the range of possible values for isolated taurus - auriga brown dwarfs with spectral types later than @xmath10m5 , assuming it has a similar age and formed in a similar process .
a more appropriate comparison would consider only companions , rather than the full population of isolated objects .
even a very low - mass companion might be able to retain a substantial disk via bondi - hoyle accretion from the disk or envelope of the primary , but free - floating objects do not have such a reservoir to build or maintain a comparable disk mass .
unfortunately , the number of substellar companions in nearby star - forming regions is quite low , and the number with component - resolved millimeter measurements is even lower , so such a comparison is not yet feasible
. trends observed for more massive stars might also apply , but it is unclear which processes are dominant in this specific case .
a clear trend is seen in older regions for less massive free - floating objects to retain their disks longer @xcite , perhaps even for several tens of myr among brown dwarfs @xcite .
observations of ( stellar ) binaries in taurus suggest that primaries usually dominate the disk mass budget @xcite , though perhaps not out of proportion to the observed scaling of disk mass with primary mass @xcite , and the disks ultimately seem to have similar lifetimes @xcite .
furthermore , close binaries ( such as the fw tau ab pair , with @xmath44 au ) appear to have sharply reduced disk frequencies even at early ages @xcite , so it might not be surprising that the primary has cleared its disk quickly .
the relative rate of disk dispersal is therefore an ambiguous feature in the classification of this system . in the meantime ,
the most unambiguous way to determine the mass and geometry of fw tau c will be to spectroscopically detect photospheric absorption features , either at higher spectral resolution or in the @xmath101@xmath11 m regime where veiling is minimized .
disk mass determinations for more securely identified pmcs also could shed light on the distinction between companions formed via binary processes or planetary processes . in any case , the alma data presented here signals that we are entering a new regime of sensitivity that will ultimately enable the characterization of fundamental circum(sub-)stellar and circumplanetary disk properties .
the authors thank gaspard duchne for an insightful discussion of the nature of the fw tau companion , and the referee for a helpful critique of this paper .
this paper makes use of the following alma data : ads / jao.alma#2012.1.00989.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj .
the nrao is a facility of the nsf operated under cooperative agreement by associated universities , inc .
|
we present alma observations of the fw tau system , a close binary pair of m5 stars with a wide - orbit ( 300au projected separation ) substellar companion .
the companion is extremely faint and red in the optical and near - infrared , but boasts a weak far - infrared excess and optical / near - infrared emission lines indicative of a primordial accretion disk of gas and dust .
the component - resolved 1.3 mm continuum emission is found to be associated only with the companion , with a flux ( @xmath0mjy ) that indicates a dust mass of 12@xmath1 . while this mass reservoir is insufficient to form a giant planet , it is more than sufficient to produce an analog of the kepler-42 exoplanetary system or the galilean satellites .
the mass and geometry of the disk - bearing fw tau companion remains unclear .
near - infrared spectroscopy shows deep water bands that indicate a spectral type later than m5 , but substantial veiling prevents a more accurate determination of the effective temperature ( and hence mass ) .
both a disk - bearing `` planetary - mass '' companion seen in direct light or a brown dwarf tertiary viewed in light scattered by an edge - on disk or envelope remain possibilities .
|
the outer solar system beyond the orbit of neptune is a relatively unfamiliar world to the human being , and yet it is believed to contain debris of the primordial solar disk and therefore carries much information about the early solar system . although the existence of objects in the outer solar system was proposed long time ago @xcite , except for pluto discovered by c. w. tombaugh in 1930 , the first of such objects was not found until 1992 @xcite . since then
several surveys have been performed and about 1000 tnos have been discovered ( http://cfa-www.harvard.edu/cfa/ps/lists/tnos.html ) . the differential size distribution of the observed tnos , which are all larger than about 100 km in diameter , follows a power law like @xmath1 with @xmath2 ( e.g. @xcite ) .
the size distribution , in addition to representing a major population property of tnos , provides important constraints on the theory of how our planetary system was formed . in coagulation models of planet formation with collisional disruption cascade ,
for example , the power - law size distribution does not simply extend towards smaller size but turns flatter ( a smaller @xmath3 ) below a certain break radius ( e.g. @xcite ) .
the magnitude of the break radius depends on the initial mass in the trans - neptunian region , the epoch of neptune s formation , and the tensile strength of these small bodies .
extending our knowledge of the size distribution to smaller tnos is clearly desirable .
tnos smaller than 100 km are too dim to detect . a hubble space telescope survey reported the detections of 3 tnos within 0.02 square degrees of the sky , whose estimated diameters are about 30 km @xcite .
these detections were achieved by blindly integrating images repeatedly over all possible tno orbits for a total exposure of 22 ks .
these 3 detections are a factor of 25 less than that expected from the extrapolation of the size distrbution of tnos larger than 100 km . for even smaller tnos , occultation of background stars as a way to study their properties was proposed thirty years ago @xcite .
searches for such occultation events in optical bands have been conducted by many groups but without any definite detection so far .
three possible detections were just reported recently @xcite . on the other hand , occultation events in x - rays may stand a better chance of detection , compared with the optical band .
the photon - counting nature of x - ray detectors like proportional counters provides much faster photometry and the much shorter wave length of x - rays means much less diffraction .
these two features make it more feasible to detect events caused by smaller bodies , and henceforth can provide a larger number of possible events .
this technique requires a bright background x - ray source .
sco x-1 , the brightest x - ray source in the sky , serves this purpose well .
its location , only about 5.5 degrees north of the ecliptic , allows good sampling of the tno population . the proportional counter array ( pca ) on board rossi x - ray timing explorer ( rxte ) @xcite , a nasa satellite , has the largest effective area in the kev x - ray band .
the typical rxte / pca count rate of sco x-1 is @xmath4 counts per second .
it is therefore possible to search for occultation events at millisecond time scales .
millisecond dip events in the x - ray light curve of sco x-1 , in total 58 in 322-ks data , were reported recently @xcite , which were attributed to occultation caused by small tnos .
later , signatures of possible instrumental effects related to high energy cosmic rays for many of these dip events were found @xcite and their nature became uncertain . in this paper
, we report further analysis of 107 dip events ( including the original 58 ) , which were found in 564-ks rxte archival data spanning over 7 years from 1996 to 2002 .
the procedure to identify these events and their properties are described in section 2 .
signatures of possible instrumental effects are discussed in section 3 . in section 4
we propose some criteria to tentatively separate true astronomical events from those likely due to the instrumental dead - time effect .
we then further argue that those dip events without indication of instrumental effects are occultation caused by small tnos and derive the total number of tnos in the corresponding size range in section 5 .
to search for possible occultation events , we examined the light curves of sco x-1 binned in different bin sizes from all the available rxte / pca archival data and computed their deviation distributions .
the definition of the ` deviation ' employed here can be found in @xcite .
the deviation distribution of all the time bins in the light curve should be of a poisson nature modified by the dead - time and coincidence - event effects if all the fluctuations are random and there is no occultation event .
one example of such a deviation distribution is shown in fig . 1 , in which the search was conducted with 2-ms time bins .
the excess at the negative deviation end is obvious .
we performed the search with different time - bin sizes ranging from 0.5 ms to 20 ms . in each search , those time bins with a negative deviation of random probability lower than @xmath5 were selected .
it corresponds to a negative deviation of about @xmath6 in the deviation distribution , depending on the total number of time bins in that search . in total , 107 dip events were identified in 564-ks data , many of which appear more than once in searches with different bin sizes .
the epochs of all the 107 events are listed in table 1 for later referring and for readers convenience to find them in the rxte data .
[ eventlist ] [ cols="<,<,<,<,<,<,<,<,<,<,<",options="header " , ] some example light curves of these dip events are shown here in fig .
2 . these light curves show a variety of characteristics .
about one half of all the events have a duration of 2 ms , and most of the rest are 1.5-ms and 2.5-ms events .
the longest one is 7 ms .
the flux - drop fraction of these events , @xmath7 , defined as @xmath8 with @xmath9 being the ratio of the average counts per bin within the event to that outside the event in a 100-ms window , is distributed between 0.3 and 0.8 .
3 shows the distribution of the 107 events in duration and in the flux - drop fraction .
the event duration is defined as the length of the time interval in which the photon count number in a 0.5-ms bin is smaller than one standard deviation below the average counts per 0.5-ms bin in a 100-ms window .
this definition , instead of considering all the bins with photon counts just below the average , avoids a probably inadequate designation of a longer duration to cases like event 31 in fig . 2 and event 5 in fig .
however , a more complicated dip light curve may simply be missed with our definition , such as event 9 and 101 in fig .
2 . adopting the average counts as the threshold to define
the duration will include the trailing dip of event 9 but not event 101 .
to have an operational definition without arbitrary , subjective selection judgment , we adopt our current definition .
this definition tends to somewhat underestimate the duration and overestimate the flux - drop fraction .
we note that our detection method as described above is biased against events of a small flux - drop fraction . in the case of a search with time bins of 2
ms , for example , those time bins falling between @xmath10 and @xmath11 contain a mixture of random and good events and are not selected ( see fig .
this procedure is also biased against events much shorter than about 1 ms because of the small count number in one bin .
many of the events have an asymmetric light curve with fast drop and slow recovery , like event 21 and 84 shown in fig . 2 . to characterize the degree of asymmetry
, we define the skewness , @xmath12 , of each event as @xmath13 where @xmath14 is the total number of ` missing photons ' within the event duration relative to the average count level of a 100-ms window , @xmath15 is time and runs for each time bin , @xmath16 is the number of missing photons in time bin @xmath15 , and @xmath17 is the standard deviation of this missing - photon distribution in time ( @xmath18 and @xmath19 ) .
with such a definition , @xmath14 and @xmath16 are in general not integers .
the skewness is then normalized to @xmath20 , which is roughly equal to the standard deviation in skewness randomly drawn from a gaussian distribution of @xmath14 photons @xcite .
the skewness is zero for symmetric light curves and positive for light curves with fast drop and slow recovery .
the skewness distribution of the 107 events is shown in fig .
it appears that there seems to be two populations of events in terms of skewness .
it was pointed out in @xcite that many of these dip events may be related to the so - called ` very large event ' ( vle ) , in the rxte terminology .
unfortunately , the vle counts were recorded with a 125-ms resolution . besides , not all the individual vle bins covering the dip epochs have an unusual high count number .
the association of vles and dip events is so far not conclusive .
prompted by the report in @xcite , we examined some more particle - related properties of these dips , in the hope to properly assess the instrumental signatures and to identify non - instrumental dip events , if the current data allows .
a vle is an event that deposits more than about 100 kev energy in any one of the six active xenon - layer anodes or the propane - layer anode in each proportional counter unit ( pcu ) of the pca on board rxte .
technical details of rxte instrumentation can be found in the rxte web site and also in @xcite .
a vle saturates the preamplifier and causes ringing in the signal chain when the amplified signal is restored to the baseline .
a larger dead - time window is set for each vle . for most of the sco x-1 observations
it was chosen to be about 70 @xmath21s @xcite .
the actual duration in which the signal chains are affected depends on the pulse height of the saturating event .
the vle counts are recorded with a 125-ms time resolution in the ` standard-1 ' data mode for every pca observations .
the standard-1 data mode also records , with the same 125-ms resolution , the propane - layer counts , which are events that trigger only the propane layer , and the ` remaining ' counts , which are all the events except for ` good ' events , vles , and propane events .
good events are those trigger only one xenon anode . the propane layer is designed to distinguish events caused by soft electrons from those by x - rays .
the ` remaining ' counts contain many ` multiple - anode ' events , which are usually treated as particle background . because sco x-1 is very bright , most of the propane - layer events and the ` remaining ' events are in fact caused by photons from sco x-1 . in most of the sco x-1 observations , the data mode recording events triggering both the ` left ' and ` right ' xenon anodes of the first layer ( l1&r1 ) with a 250-@xmath21s time resolution is also employed .
almost all of these l1&r1 coincidence events are caused by photons from sco x-1 , instead of particles .
all the above information is useful for investigating the possible instrumental effects for the dip events .
unfortunately the 125-ms resolution is much larger than the dip event duration .
more detailed information is recorded in the ` standard-2 ' data mode but with an even coarser resolution of 16 s. in fig .
4 , example light curves of good counts , l1&r1 coincidence events , propane - layer events , very large events , and ` remaining ' events are shown .
the time zero of all the 5 light curves is referred to the same epoch . for all the 107 events , the l1&r1 coincidence counts all drop consistently with the dip , although sometimes not so obviously because of the small count number .
the propane - layer events in fact do not show any anomaly .
many of the 107 events are associated with higher vle counts , like event 79 in fig .
4 . their vle counts are not unusually high .
only when all the events are studied together , the tendency of having a higher vle count number becomes clear .
we note that , however , a higher vle count does not always come with a dip event . in fig .
4 , there are several vle bins with a higher count number similar to that at the dip epoch of event 79
. there are no dip events in those time bins .
a clearer indication of possible instrumental effects appears in the ` remaining ' counts .
almost all the 107 events are associated with a ` remaining ' count number larger than the average in a local window . to better present all these count excesses , we define , for each dip event , the propane deviation , vle deviation , and ` remaining ' deviation as the difference between the corresponding count number and the average in a certain window encompassing the bin in question .
the deviation is then further expressed in units of the square root of that average .
these deviation distributions are plotted in fig .
5 . the average ` remaining ' count rate sometimes varies noticeably at time scales about 2 s or so .
we therefore used a window containing 9 bins ( about 1 s ) for the computation of these deviations .
a comparioson deviation distribution is also shown together in fig .
the comparison distribution is the average of four distributions drawn from the bin at -4s , -2s , + 2s and + 4s away from the bin covering the dip epoch respectively .
this comparison distribution is a representation of the distribution without contamination from the suspected particle events .
one can see that there is nothing unusual in the propane - count deviation .
the vle deviation distribution exhibits a positive shift but still contains a considerable population of negative deviation .
the ` remaining'-count deviation distribution shows a very significant positive shift . to further investigate the possible particle association , we computed the time difference between the dip epoch and the nearest saa peak passage .
it was done with the data obtained with the two particle monitors of the instrument hexte @xcite .
some examples are shown in fig .
the event distribution in the time since saa is shown in fig .
7 , in which all the dip events within 45 minutes before and after the saa peak passage are included .
one can see that there is no indication of any correlation between the dip epoch and the saa passage .
the suspected particle events are not related to the saa .
we have also examined other characteristics of these dip events and found no obvious correlations among the duration , flux - drop fraction , skewness , vle deviation , ` remaining ' deviation and time since saa either . that the dip skewness is not related to the vle or ` remaining ' deviations is somewhat puzzling .
the skewness versus those deviations and the time since saa is plotted in fig .
68 dip events out of the 107 are in observations with a data mode carrying two - channel spectral information .
the energy bands for these two channels are about 2 - 5 kev and 5 - 18 kev .
there seems to be a trait that the flux drop fraction is higher in the high energy band than in the low energy one .
some examples are shown in fig .
9 . the distribution of the ratio of the flux - drop fraction in the high energy band to that in the low energy one for each dip events is plotted in fig . 10 .
occultation events , because of diffraction , will have a smaller flux drop in a lower energy band , while the instrumental dead - time effect does not have that energy dependence .
although random fluctuation may render individual identification unreliable , fig .
10 suggests that , statistically speaking , some dip events may not be caused by instrumental effects .
it was proposed in @xcite that ` the dips are caused by electronic dead time in response to some type of charged particle shower in the spacecraft . ' indeed , the positive deviation in the vle and ` remaining ' counts associated with dip events within 125-ms uncertainty ( fig .
5 ) indicates that , at least statistically speaking , many of the dip events are likely due to the dead - time effect caused by particles . to have such a large flux drop for such a long period ,
however , several pcus must be affected at the same time and a very large amount of charged particles is needed .
the pca has 5 almost identical pcus .
the usual signal dead time is about 10@xmath21s , and that for a vle signal is about 70@xmath21s .
that the ` remaining ' excess is much more significant than the vle one indicates that the dead time caused by the hypothesized lower - energy secondary particles or by the vle ringing events may be the major , direct cause for these dips .
although the crab nebula is about 8 times dimmer than sco x-1 and therefore individual dips can not be identified , an excess at the negative end of the deviation distribution may be expected if the data amount is enough . however , because the flux - drop fraction of these dips is not close to unity , but distributed between 0.3 and 0.8 , they , at a rate of about one per 5 ks , are not noticeable in the deviation distribution drawn from the 380-ks rxte / pca archival data of the crab nebula . on the other hand , the indication so far is not strong enough to argue that all the dips are caused by the cosmic - ray induced dead - time effect .
there are dip events whose associated ` remaining ' and vle deviations are negative or very small .
there are time bins ( 125 ms ) with large vle or ` remaining ' excesses but without any dips .
furthermore , the flux - drop fraction of the dip events is generally larger in the high - energy band than in the low - energy one , a signature of diffraction .
it is likely that these dips are heavily contaminated by particle - related events , but not all of them are caused by the instrumental dead - time effect .
limited by the 125-ms time resolution of the standard-1 data , it is impossible to definitely identify individual dip events as caused by the instrumental dead - time effect or as of a probable astronomical origin .
nevertheless , as discussed in the previous section , it is likely that these dip events are a mixture of the two kinds . since the light - curve skewness and the ` time since saa '
are not correlated with the vle or ` remaining ' deviations , we propose to select those probable non - instrumental dip events solely based on the two deviations . from fig . 5
, one can see that the comparison population has a half width at half maximum of about 1.25 in both deviation distributions . if a non - instrumental population exists , the half width in the deviation distribution is about the same . in this regard ,
a somewhat conservative criterion to identify probable non - instrumental events can be set as the vle and ` remaining ' deviations being both smaller than 1.25 . from the ` remaining ' deviation distribution in fig .
5 , the total number of the probable non - instrumental events can be estimated by assuming all the events with a negative ` remaining ' deviation are the probable non - instrumental events and they have a more or less symmetric deviation distribution centered at zero deviation .
such a possible distribution is plotted in dashed lines in fig .
the total number estimated this way is about 16 .
this number is of course somewhat arbitrary and only suggestive .
the sum of the two deviations can be taken as a measure and the second criterion can be set , to be conservative again , as that sum being smaller than 1.0 so that the total number is 12 .
these two criteria are also shown in fig . 11 . the light curves of the 12 probable non - instrumental events are shown in fig . 12 .
the second criterion is somewhat arbitrary .
we have conducted such a selection procedure with different window size for computing the ` remaining ' deviation .
8 dip events among the 12 are always selected even when the window is as large as 10 s. the 4 left out ( event 41 , 52 , 59 , and 74 ) are all those with the ` remaining ' count level varying at 2-s time scales .
we think the 1-s window , which we employed to compute the ` remaining ' deviation distribution shown in fig .
5 , is a more meaningful one and therefore adopt the 12 dip events selected with the procedure mentioned above as the probable non - instrumental events . among the 12 probable events ,
7 ( event 19 , 41 , 52 , 59 , 61 , 66 , and 74 ) have two - channel spectral information .
their flux - drop - fraction ratios of high- to low - energy bands are all larger than unity . only event 4 and event 66 show obvious asymmetry in their light curve , that is , with fast drop and slow recovery in the flux .
we suggest that , in the context of the above discussion , the 12 events , whose light curves are plotted in fig .
12 , are probable non - instrumental events .
it is about 10% of all the identified dip events .
hereafter we assume the 12 probable events proposed above are really non - instrumental .
these events of abrupt and significant flux drop at millisecond time scales are not random fluctuations , as evidently shown in the deviation distribution ( e.g. fig .
1 ) . based on spectral analysis , x - ray emission from sco x-1
is believed to originate from the neutron star surface thermal emission and corona comptonization near the inner accretion disk ( e.g. @xcite ) .
a sudden quench of x - ray emission from sco x-1 in such a short time scale seems very unlikely .
we have also examined the x - ray color for those events with two spectral channels .
no detectable color change in the x - ray emission before and after any event was found .
occultation by objects in the line of sight is the most viable cause for these events .
the flux - drop fraction , @xmath7 , of these events ranges from 0.36 to 0.72 . to have such a significant flux drop ,
the occulting body must be a few times larger than the corresponding fresnel scale , which is @xmath22 , where @xmath23 is the wavelength and @xmath24 is the distance .
it is about 30 m for @xmath25 au and @xmath26 nm ( 4 kev ; most photons in the data are at this energy ) .
some x - ray diffraction patterns for objects of different sizes at different distances are computed and plotted in fig .
13 , with the pca - detected photon spectrum from sco x-1 as the input radiation for the computation .
13 shows the case for a spherical blocking object and a crossing impact parameter equal to 0.33 .
the crossing impact parameter is the distance to the shadow center in units of the object size .
computations with different crossing impact parameters , such as 0 and 0.67 , do not yield too different a flux - drop fraction , although the diffraction patterns are different , particularly regarding the central brightening in the shadow . from the diffraction pattern computation
, one can see that to have @xmath27 , the size of the occulting body at 40 au away must be larger than 30 m , the fresnel scale of 4-kev photons .
we note that , given a certain ratio of the object size to the square root of the distance , the diffraction patterns are the same , up to a scaling factor in length proportional to the object size or the square root of the distance ( e.g. @xcite ) .
the flux - drop fraction @xmath7 is therefore also the same for these patterns . at the distance of sco x-1 , 2.8 kpc @xcite ,
the fresnel scale is about 100 km ( hereafter all the fresnel scales are referred to that for 4-kev photons ) .
it is clear if these events are occultation by objects moving around sco x-1 , those objects must be larger than about 100 km and also larger than the x - ray emitting region , which ranges from a few hundred kilometers to about 50,000 km , depending on different models ( e.g. @xcite ) . to produce a 2-ms event , the speed of those objects needs to be larger than @xmath28 km / sec , a very unlikely speed .
if we consider , more realistically , the event duration as the shadow size instead of the object size , an even higher speed is required . the same argument can be applied to other objects in the interstellar space . for objects at the distance of the oort cloud ,
typically 10,000 au , the fresnel scale is about 500 m. again , to produce the events that we found , the oort cloud objects need to be larger than that scale .
it is unlikely that a kilometer - size object in the oort cloud can produce millisecond events , since its speed relative to the earth is only about 30 km / sec . on the other hand , other solar system small bodies at a shorter distance
can satisfy the duration and flux - drop constraints at the same time .
a major source candidate is the main - belt asteroids . their expected event rate for occulting sco x-1 , however , is too low ( @xmath29 sec@xmath30 , down to 10-m size ; see @xcite ) .
we therefore propose that the 12 events , if really not instrumental , are occultation caused by objects in the trans - neptunian region , that is , by tnos .
to pin down the distance and the size of each occulting body more accurately for all the events is an important issue but requires further analysis on the dead - time - corrected light curves , diffraction patterns and different orbital inclinations and eccentricities .
that study will be reported in another forthcoming paper . for an approximate estimation of the occulting - object sizes at this stage , we contend ourselves by considering a typical relative speed of 30 km / s between the occulting tnos and rxte and a random crossing through the shadow . the average crossing length of a random crossing is @xmath31 times that of a central crossing , assuming a circular shadow .
we therefore set the size range of these 12 objects to be from 60 m to 100 m for durations from 1.5 ms to 2.5 ms .
we note that the orbital speed of the earth is about 30 km / s , rxte s speed relative to the earth is about 7 km / s , a tno at 43 au has an orbital speed about 5 km / s for a low eccentricity orbit , the shadow size is larger than the object , and our definition of the event duration tends to somewhat underestimate .
we next proceed to estimate the total number of tnos of these sizes . for a background point source ,
the event rate is @xmath32 where @xmath14 is the number of detected events ( assuming a 100% detection efficiency ) , @xmath33 the total exposure time , @xmath34 the differential size distribution , @xmath35 the typical sky - projection relative speed , @xmath24 the typical distance to the tnos , and @xmath36 the total solid angle of the sky distributed with tnos .
to derive @xmath34 from the event rate , a functional form of the distribution needs to be assumed . on the other hand ,
if the integration is only over a small range of the size , we may derive an average value at that size .
noting that @xmath37 , we have @xmath38 and @xmath39 we can estimate @xmath36 with the inclination distribution obtained from the cfht survey @xcite , which reported a 20@xmath40 half - angle of an assumed gaussian distribution .
this half - angle in the inclination distribution translates to about 12.8@xmath40 for a corresponding half - angle in the apparent ecliptic latitude distribution , assuming circular orbits .
sco x-1 is 5.5@xmath40 north of the ecliptic . to use the detection rate in the direction toward sco x-1 to represent the whole tno population ,
the equivalent sky area is a zone occupying @xmath41 in latitude .
therefore , @xmath42 . assuming a typical distance @xmath43 au , a typical relative sky - projection speed @xmath44 km / s , and @xmath45 ks , we have in the size range from 60 m to 100 m @xmath46 the total number of tnos in that size range is about @xmath47 .
there are considerable uncertainties in the above estimate .
first of all , the suggestion that 12 out of the 107 dip events are non - instrumental is inconclusive .
the detection efficiency of the procedure that we employed needs to be investigated in more details .
the size of these possible occulting tnos is only very roughly estimated .
with all these cautions in mind , we tentatively discuss the implication of the discovery of tno x - ray occultations .
the estimated total number , about one quadrillion , for tnos of size between 60 m and 100 m is several orders of magnitude larger than usual model predictions , but is close to , within the 1-@xmath17 level , the direct extrapolation of the distribution obtained from the cfht survey @xcite , which gives a best fit of @xmath1 with @xmath48 ( 1 @xmath17 ) and the total number for tnos larger than 100 km equal to @xmath49 ( fig .
our result seems to indicate that the power - law size distribution of tnos may extend down to the order of 100 m without any breaking or flattening . on the other hand ,
if the assumed albedo is appropriate when converting the luminosity function into a size distribution from the hst observation @xcite , and if that size distribution inferred from 3 detections in 0.02 square degrees of the sky is reliable , a break will occur at about 50 km and there must exist a second component at smaller sizes . this second component may require a new mechanism in the theory of planet formation .
we report 107 dip events of millisecond time scales in 564.3-ks rxte / pca archival data of sco x-1 .
these dip events are apparently contaminated by those due to instrumetal dead - time effects induced by high - energy cosmic rays .
we performed detailed analysis and propose that about 10% of these dip events are probable tno - occultation events .
the inferred size distribution of tnos at size about 100 m is several orders of magnitude larger than current theoretical predictions . confirming or disproving this result is important for our understanding of planet formation processes in the early solar sytem .
x - ray diffraction patterns in the occultation light curves may also be exploited to study the background x - ray source , potentially in a very high angular resolution .
it is unfortunately impossible to definitely separate true occultation events from instrumental ones with currently available data .
observations of rxte / pca with an adequately designed data mode or of other new instruments in the future may resolve this issue .
we thank r. rothschild , f. roques , and g. georgevits for helpful discussion .
the examination of the event rate versus the time - since - saa with the saa passage determined from hexte / pm data were first done by tommy thompson and richard rothschild .
this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center , and of the jpl horizons on - line solar system data and ephemeris computation service .
this work was supported by the national science council of the republic of china under grant nsc 95 - 2112-m-007 - 050 and by academia sinica in taipei .
99 alcock c. , dave r. , giammarco j. , goldader j. , lehner m. , king s .- k . , lee t. , wang a. , wang s .- y . , wen c .- y . , et al .
, 2003 , earth , moon , and planet , 92 , 459 bailey m.e . , 1976 ,
nature , 259 , 290 barnard r. , church m.j . , balucinska - church m. , 2003 , a&a , 405 , 237 bernstein g.m . , trilling d.e . ,
allen r.l . ,
brown m.e . , holman m. , malhotra r. , 2004 , aj , 128 , 1364 bradshaw c.f .
, fomalont e.b . ,
geldzahler b.j . , 1999 ,
apj , 512 , l121 bradshaw c.f .
, geldzahler b.j . ,
fomalont e.b . , 2003 ,
apj , 592 , 486 bradt h.v .
, rothschild r.e . ,
swank j.h . , 1993 ,
a&as , 97 , 355 brown m.j.i . ,
webster r.l . , 1997 , mnras , 289 , 783 chang h .- k . , king s .- k . , liang j .- s . ,
wu p .- s .
, lin l.c .- c .
, chiu j .- l . , 2006 ,
nature , 442 , 660 cooray a. , farmer a.j .
, 2003 , apj , 587 , l125 edgeworth k.e . ,
1943 , j. br . astron .
soc . , 53 , 181 edgeworth k.e
. , 1949 , mnras , 109 , 600 farinella p. , davis d.r . , stern s.a . ,
2000 , in protostars and planets iv , eds mannings v. , boss a.p .
, russel s.s .
( the university of arizona press , tucson ) , pp 1255 - 1282 jahoda k. , et al . , 2006 , apjs , 163 , 401 jewitt d.c . , luu j.x . , 1993 ,
nature , 362 , 730 jones t.a . , levine a.m. , morgan e.h . ,
rappaport s. , 2006 , astro - ph/0612129 kenyon s.j . , 2002 , pasp , 114 , 265 kenyon s.j . ,
bromley b.c . , 2004 ,
aj , 128 , 1916 kuiper g.p . , 1951 , in astrophysics , ed hynek j.a .
( mcgraw - hill , new york ) , pp 357 - 424 leonard f.c . , 1930 , leaflet astron . soc . pacific , 80 , 121 luu j.x .
, jewitt d.c . , 2002 ,
ara&a , 40 , 63 oort j.h . , 1950 , bull .
. inst .
netherlands , 11 , 91 press w.h . , teukolsky s.a . , vetterling w.t . ,
flannery b.p . , 1992 ,
numerical recipes in c ( cambridge university press , cambridge ) , pp 610 - 615 .
roques f. , doressoundiram a. , dhillon v. , marsh t. , bickerton s. , kavelaars j.j . , moncuquet m. , auvergne m. , belskaya i. , chevreton m. , et al . , 2006 ,
aj , 132 , 819 roques f. , moncuquet m. , lavillonire n. , auvergne m. , chevreton m. , colas f. , lecacheux j. , 2003 , apj , 594 , l63 roques f. , moncuquet m. , sicardy b. , 1987 , aj , 93 , 1549 rothschild r.e . , et al . , 1998 , apj , 496 , 538 trujillo c.a . , jewitt d.c .
, luu j.x . , 2001 , aj , 122 , 457
|
millisecond dips in the rxte / pca light curve of sco x-1 were reported recently ( chang et al .
2006 ) , which were interpreted as the occultation of x - rays from sco x-1 caused by trans - neptunian objects ( tno ) of hundred - meter size .
inconclusive signatures of possible instrumental effects in many of these dip events related to high - energy cosmic rays were later found ( jones et al .
2006 ) and the tno interpretation became shaky .
here we report more detailed analysis aiming at distinguishing true occultation events from those related to cosmic rays . based on some indicative criteria derived from housekeeping data and two - channel spectral information , we suggest that about 10% of the dips are probable events of occultation .
the total number of tnos of size from 60 m to 100 m is estiamted to be about @xmath0 accordingly .
limited by the coarser time resolution of standard data modes of rxte / pca , however , definite results can not be obtained .
adequately configured observations with rxte or other new instruments in the future are very much desired .
[ firstpage ] occultations kuiper belt solar system : formation stars : neutron x - rays : binaries .
|
the dense neutron star(ns ) matter shows us an interesting subject to study the properties of nucleon matter with the density higher than nuclear saturation density @xmath3 .
there , it has pointed that the various new degrees of freedom , such as hyperons , quarks and their mixed phases , realize according to the density@xcite .
it is well known that ns matter has the properties of the strong degeneracy and exists the attractive interaction between two baryons , which are the conditions for the occurrence of superfluid states in fermi systems .
thus , ns is already considered as the key laboratories of various superfluidity in nuclear matter@xcite . in recent years
, many studies have been focussing on @xmath0 protonic superfluidity .
since the protonic superfluid states as well as their pairing strength can greatly suppress the neutrino processes involving nucleons about @xmath4 years of ns cooling phases , affect the properties of rotating dynamics , post - glitch timing observations and possible vertex pinning of nss@xcite .
the possibility of protonic superfluid states in ns matter is first suggested by migdal in 1960@xcite .
the interaction between two protons is the combination of strong repulsive short - range interaction and weaker attractive long - range interaction . in proton matter ,
when the interparticle distance is much larger than the range of the repulsive interaction , protons will condense into superfluid states due to the attractive interaction .
for the size of @xmath0 protonic pairing energy gap , the calculations based on the microscopic theory have already carried out a great amount of work .
all the numerical calculations yield qualitatively similar ranges for the appearance of @xmath0 protonic pairing .
however , obtaining the exact numerical results for @xmath0 protonic pairing energy gap and evaluating its quantitative influence on ns have been proved to be a hard problem , which is due to that there are many uncertainties about the proton - proton(pp ) interactions in ns matter , methods of approximation , paucity of experimental data in extreme conditions and so on . during this period , lots of researches of ns
have been performed adopting various frameworks .
currently , many relativistic models call attention in researches on ns since they are well suited to describe ns in accord with the special relativity .
the most commonly used among them is the relativistic mean field(rmf ) model , extremely successful in nuclear matter studies@xcite .
this article mainly does the following work .
we study ns matter for two cases : ( i)ns is made up of neutron , proton , electron and muon only ( npe@xmath5 ) , ( ii)ns is composed of nucleons , hyperons(@xmath6 and @xmath7 ) , electron and muon ( nphe@xmath5 ) .
our model excludes @xmath8 hyperons on account of the remaining uncertainty of the form of @xmath8 potential in nuclear matter at the saturation density@xcite .
we use the rmf theory to describe the properties of ns .
the @xmath0 protonic pairing energy gap is calculated by the reid soft core(rsc ) potential@xcite .
we mainly focus on the influence of hyperon degrees of freedom on @xmath0 protonic pairing energy gap in ns matter .
in the rmf theory , baryonic interactions are described by the exchanged mesons including isoscalar scalar and vector mesons @xmath9 and @xmath10 , an isovector vector meson @xmath11 , two additional strange mesons @xmath12 and @xmath13 . the total lagrangian density of ns matter is@xcite , @xmath14\psi_b + \frac{1}{2}(\partial_\mu\sigma\partial^\mu\sigma - m_\sigma^2\sigma^2)-\frac{1}{3}a \sigma ^{3}\\ -\frac{1}{4}b \sigma^4 + \frac{1}{2}m_\omega^2 \omega_\mu\omega^\mu+\frac{1}{4}c_3(\omega_\mu\omega^\mu)^2 + \frac{1}{2}m_\rho^2{\boldsymbol{\rho}}_\mu{\boldsymbol{\rho}}^\mu -\frac{1}{4}f^{\mu v}f_{\mu v}\\-\frac{1}{4}g^{\mu v}g_{\mu v}+ + \frac{1}{2}(\partial_v\sigma^*\partial^v\sigma^*-m^2_{\sigma^*}\sigma^{*2 } ) -\frac{1}{4}s^{\mu v}s_{\mu v}\\+\frac{1}{2}m^2_{\phi}\phi_\mu\phi^\mu+\sum_l\overline{\psi}_l[i\gamma_\mu\partial^\mu - m_l]\psi_l.\end{gathered}\ ] ] here the field tensors of the vector mesons , @xmath9 and @xmath10 , are denoted by @xmath15 , and @xmath16 .
the meson fields are seen as classical fields and field operators are instead of their expectation values in the rmf approximation .
the field equations derived from the lagrange function are @xmath17 here @xmath18 denotes baryonic isospin projection . @xmath19 and @xmath20 are baryonic scalar and vector densities , respectively .
they have the following form , @xmath21 here @xmath22 is baryonic spin projection , @xmath23 is baryonic fermi momentum , @xmath24 is baryonic effective mass . for ns matter consisting of n , p , @xmath6 , @xmath25 , @xmath26 , e , @xmath5 , the charge neutrality condition is given by @xmath27 the @xmath28 equilibrium conditions are expressed as @xmath29 at zero temperature the chemical potential of baryon and lepton are written by @xmath30 the @xmath0 protonic pairing energy gap @xmath31 can be obtained by solving the bcs gap equation , @xmath32 where @xmath33 , @xmath34 is protonic single particle energy . for the pp interaction ,
we adopt the rsc potential which is well suited for applying in ns matter .
@xmath35 is defined the matrix element of @xmath0 component of rsc potential in momentum space , @xmath36 here @xmath37 is @xmath0 pp interaction potential in coordinate space . it is expressed in the five - range gaussian and depends on @xmath20 , asymmetry parameter @xmath38 and two - nucleon state @xmath28 and @xmath39 as discussed in refs@xcite , @xmath40 the critical temperature @xmath41 of @xmath0 protonic superfluidity is given by its energy gap @xmath31 at zero temperature approximation@xcite , @xmath42 combining eqs.(1)-(6 ) with the charge neutrality and @xmath28 equilibrium conditions , eqs.(8 ) and ( 9 ) , we can solve the system with a fixed @xmath20 .
thus , @xmath0 protonic pairing energy gap @xmath31 and critical temperature @xmath41 can be obtained from eqs.(11)-(14 ) .
as mentioned above , the onset of @xmath0 protonic superfluidity is determined by the energy gap function @xmath31 . theoretical calculation of @xmath31 is sensitively dependent on the model of pp interaction and many - body theory adopted . in this paper
, we use the rsc potential as an example to analyze @xmath0 protonic superfluidity in npe@xmath5 and nphe@xmath5 matter , respectively .
we mainly focus on the influence of hyperon degrees of freedom on @xmath0 protonic superfluidity . in order to make the results more clearly , we employ two successful parameter sets of tm1 and tma to calculate separately @xmath0 protonic pairing energy gap in ns matter .
the parameter sets are listed in table i and ii . for the vector couplings of hyperons ,
we take the relations derived from su(6 ) quark model(see ref@xcite for details ) , @xmath43 .[tab : table i]the coupling constants of the meson - nucleon of tm1 and tma sets , the masses are unit of mev@xcite . [ cols="<,<,<,<,<,<,<,<",options="header " , ] up to now , we do not known @xmath0 protonic pairing energy gap increase or decrease , if hyperons appear in ns matter . as the uncertainty of the pp interaction ,
we calculate @xmath31 on the basis of the rsc potential .
we concentrate on the influence of hyperon degrees of freedom on @xmath0 protonic pairing energy gap in ns matter .
the peak values of @xmath0 protonic pairing energy gap @xmath44 and the corresponding critical temperature @xmath45 , baryonic density @xmath20 in npe@xmath5 and nphe@xmath5 matter are listed in table iii , for the tm1 and tma parameter sets .
as shown in table iii , the appearance of hyperons makes @xmath44 and @xmath45 decrease which will inevitably cause the cooling rate of ns changing . in fig.4
, we show @xmath0 protonic pairing energy gap @xmath31 as a function of baryonic density @xmath46 in npe@xmath5 and nphe@xmath5 matter , for parameter sets tm1 and tma . as seen from fig.4 , the protonic pairing energy gap @xmath31 as the function of @xmath46 is a typical bell - shaped curve from zero to a maximum value to zero again .
the change of @xmath31 in behavior is due to the reduction of mean interparticle distance . from fig.4 , we also can see that @xmath0 protonic superfluidity appears within the baryonic density range of @xmath20 = 0.0 0.649 ( 0.685 ) @xmath47 and @xmath20 = 0.0 0.953 ( 0.853 ) @xmath47 for parameter sets tm1 ( or tma ) in npe@xmath5 and nphe@xmath5 matter , respectively .
it is clear from these data that the appearance of hyperons makes baryonic density range of @xmath0 protonic superfluidity widen .
such baryonic density ranges can cover or partially cover the cores of nss and are highly relevant to the direct urca processes involving nucleons which play a leading role in ns cooling .
in particular , from fig.4 , the appearance of hyperons also leads to @xmath0 protonic pairing energy gap in nphe@xmath5 matter obviously larger than the corresponding values in npe@xmath5 matter when baryonic density @xmath48 0.377 ( 0.409 ) @xmath47 for tm1(or tma ) parameter set ( see the two dots shown in fig.4 ) . according to eq.(14 ) , the critical temperature of @xmath0 protonic superfluidity positively increases which could further suppress the cooling rate of ns .
we study the influence of hyperon degrees of freedom on the size and baryonic density range of @xmath0 protonic paring energy gap by adopting the rmf and bcs theories in ns matter .
it is shown that the appearance of hyperons makes the peak values of @xmath0 protonic pairing energy gap and critical temperature decrease . and baryonic density range of existing @xmath0 protonic superfluidity is widen from 0.0 0.649 ( 0.685 ) @xmath47 in npe@xmath5 matter to 0.0 0.953(0.853)@xmath47 for tm1 ( or tma ) parameter sets in nphe@xmath5 matter .
in addition , when the baryonic density @xmath48 0.377 ( 0.409 ) @xmath47 for tm1(or tma ) parameter set , the appearance of hyperons leads to @xmath0 protonic pairing energy gap obviously larger than the corresponding values in npe@xmath5 matter which could further suppress the cooling rate of nss . in our results , the appearance of hyperons in ns matter not only changes the eos and bulk properties but also changes the properties of @xmath0 protonic superfluidity .
0 meng j , toki h , zhou s g , et al .
2006 prog . part .
phys * 57 * 470 bednarek i , manka r. 2005 j. phys .
g : nucl . part .
phys * 31 * 1009 yang f , shen h. 2008 phys .
c * 77 * 025801 amundsen l , stgaard e. 1985 nucl .
a * 437 * 487 chen j m c , clark j w , dav r d et al .
1993 nucl .
a * 555 * 59 zhao e g , wang f. 2001 chin .
bull * 56 * 3797 zuo w , li z h , lu g c , et al .
2004 phys .
b * 595 * 44 zuo w and lombardo u. 2010 aip conf .
proc * 1235 * 235 xu f r. 2012 chin .
bull * 57 * 4689 tanigawa t , matsuzaki m , chiba s. 2004 phys . rev .
c * 70 * 065801 .
shen c , lombardo u , schuck p. 2005 phys .
c * 71 * 054301 .
cao l g , lombardo u , schuck p. 2006 phys .
c * 74 * 064301 .
xu y , liu g z , liu c z , et al .
2013 chin . phys .
lett * 30 * 062101 .
gnedin o y , yakovlev d g. 1993 astron .
lett * 19 * 104 chen w , lam y y , wen d h , et al .
2006 chin .
lett * 23 * 271 yakovlev d g , kaminker a d , gnedin o y , et al .
2001 phys .
rep * 354 * 1 xu y , liu g z , wu y r , et al .
2012 plasma sci .
technol * 14 * 375 xu y , liu g z , liu c z , et al . 2013 chin .
lett * 30 * 129501 xu y , liu g z , liu c z , et al . 2014
bull * 59 * 273 migdal a b. 1960 soviet phys .
jetp * 10 * 176 nishizaki s , takatsuka t , yahagi n , et al .
1991 prog .
phys * 86 * 853 wambach j , ainsworth t l , pines d. 1993 nucl .
a * 555 * 128 takatsuka t , tamagaki r 2004 nucl .
a. * 738 * 387 xu y , liu g z , wang h y , et al .
lett * 29 * 059701
|
the possible @xmath0 protonic superfluidity is investigated in neutron star matter , and the corresponding energy gap as a function of baryonic density is calculated on the basis of bcs gap equation .
we have discussed particularly the influence of hyperon degrees of freedom on @xmath0 protonic superfluidity .
it is found that the appearance of hyperons leads to a slight decrease of @xmath0 protonic pairing energy gap in most density range of existing @xmath0 protonic superfluidity .
however , when the baryonic density @xmath10.377 ( or 0.409 ) @xmath2 for tm1 ( or tma ) parameter set , @xmath0 protonic pairing energy gap is significantly larger than the corresponding values without hyperons . and the baryonic density range of existing @xmath0 protonic superfluidity is widen due to the appearance of hyperons . in our results , the hyperons not only change the eos and bulk properties but also change the size and baryon density range of @xmath0 protonic superfluidity in neutron star matter .
|
standard models for studying collective complex behavior in natural systems consist typically of ensembles of interacting dynamical elements @xcite .
such kind of models has proven to be extremely versatile in the mathematical description , both analytical and numerical , of a wide variety of phenomena within the scopes of physics , biology , and other branches of science @xcite . according to the range of the involved interactions
, these models can be divided into two well distinct classes .
local interactions which are paradigmatically represented in reaction - diffusion systems @xcite give rise to macroscopic evolution in which space variables play a relevant role , such as spatial structures and propagation phenomena . on the other hand , with global interactions where the coupling range is of the order of , or larger than , the system size space becomes irrelevant and collective behavior is observed to develop in time , typically , in the form of synchronization @xcite .
an essential model of globally coupled elements is given by a set of @xmath0 identical oscillators described , in the so - called phase approximation , by phase variables @xmath1 ( @xmath2 ) .
their evolution is governed by the equations @xmath3 it is known that , for any value of the coupling intensity @xmath4 , all the elements converge to a single orbit whose frequency @xmath5 coincides with that of an individual oscillator @xcite . in this case , @xmath6 measures the time required to reach such synchronized state . in this note
we present results on a generalization of the above model , where time delays are introduced .
the effect of time delays in synchronization phenomena has already been considered for two - oscillator systems , both periodic @xcite and chaotic @xcite .
ensembles with local interactions @xcite and globally interacting inhomogeneous systems have also been studied @xcite .
none of these contributions make however explicit reference to the relevant case where interactions are global but their propagation occurs at a finite velocity @xmath7 .
this situation , which naturally introduces time delays , provides a realistic description of highly connected systems where the time scales associated with individual evolution and with mutual signal transmission are comparable . instances of such systems are neural and informatic networks @xcite , and biological populations with relatively slow communication media such as sound @xcite .
our main result is that , since a finite signal velocity makes spatial variables relevant even when interactions are global , globally coupled ensembles with time delays exhibit typical features of systems driven by local interactions , in particular , structure formation and propagation .
we consider an ensemble of @xmath0 identical oscillators in the phase approximation , governed by the equations @xmath8,\ ] ] where @xmath9 is the time required for the signal to travel from element @xmath10 to element @xmath11 at velocity @xmath7 , and @xmath12 is the distance between @xmath11 and @xmath10 .
note that coupling is still global , since its intensity @xmath4 does not depend on the distance between elements .
however , the relative position of the oscillators becomes now relevant through time delays .
the full specification of our system requires to fix the topology and the metric properties of the ensemble , by fixing the values @xmath12 for all @xmath13 .
moreover , initial conditions for @xmath14 must be provided . in the case of delay equations like ( [ ens ] ) , it is necessary to specify the evolution of @xmath14 at times prior to @xmath15 up to a time @xmath16 @xcite . in the following
we shall assume that for @xmath17 the oscillators evolve independently from each other at their proper frequency @xmath5 and with random relative phases .
namely , for @xmath17 we have @xmath18 , where @xmath19 is chosen at random from a uniform distribution in @xmath20 . at @xmath15 coupling in
switched on , so that we formally have a time - dependent coupling intensity @xmath21 , where @xmath22 is the heaviside step function . through extensive numerical calculations for a variety of topologies , ranging from one - dimensional arrays to tree ( ultrametric ) structures
, we have found that the system evolves to a state where all the oscillators have the same frequency . on the other hand ,
in contrast with the case without time delays @xcite , their phases can be different .
this asymptotic state corresponds thus to a situation of frequency synchronization .
the long - time evolution of each oscillator can then be written as @xmath23 , where in general @xmath24 for @xmath25 .
the fact that these phases are different could have been expected for topologies where not all the elements are equivalent for instance , when boundaries are present . as we show later , however , such states are also found in homogeneous topologies . in this case , they are associated with propagating structures . in general , the synchronization frequency is different from the proper frequency of each oscillator , @xmath26 . according to ( [ ens ] ) , the synchronization frequency satisfies @xmath27 note that the sums @xmath28 are in general different for each @xmath11 .
however , their numerical value must coincide if the synchronization frequency is to be well defined . for a given value of @xmath29
, this constraint provides @xmath30 independent equations for the phases @xmath31 : @xmath32 since phases are defined up to an additive constant we can choose for instance @xmath33 , and solve these equations for @xmath34 .
then , @xmath29 can be found self - consistently from ( [ omega ] ) .
for large values of @xmath0 , and due to the involved nonlinearities , this results to be a quite hard numerical problem .
an approximate solution can however been found in the case of short delays , i.e. close to the situation where the system is also synchronized in phase , @xmath35 for all @xmath36 .
assuming that @xmath37 , we can write @xmath38 where @xmath39 is the average of the time delays associated with element @xmath11 . taking into account eq .
( [ s = s ] ) , we get @xmath40 where @xmath41 is a constant , independent of @xmath11 , that can be chosen arbitrarily .
this result indicates that , in this short - delay limit , oscillators with small average delays are relatively ahead in the evolution , as their phases are larger .
this is plausibly due to the fact that , in average , they receive the information on the system state before other elements with larger values of @xmath42 , which are thus relatively retarded .
note moreover that in a homogeneous topology all the elements are equivalent , so that @xmath43 is the same for all oscillators . in this case , the system is also synchronized in phase . within the approximation of short delays ,
the synchronization frequency is given by @xmath44 where @xmath45 is the overall average delay .
it therefore results that @xmath29 is smaller than the proper frequency of each oscillator .
in this note , we specifically focus the attention on the case of one - dimensional arrays .
two different topologies are considered , namely , with periodic boundary conditions where all the elements are equivalent and with free boundaries where the neighborhood of each element depends on its distance to the center of the array . for periodic boundary conditions , which we consider first ,
the distance between two elements is not univoquely defined , since it can be measured around the ring in both directions .
we fix @xmath12 by taking the minimum of these values , namely , @xmath46 ) , the proper frequency @xmath5 can be used to define time units so that , without loss of generality , we fix @xmath47 . moreover , our numerical simulations are restricted to the case @xmath48 . as a matter of fact , we have found that other coupling intensities do not produce qualitatively different results .
note that this would not be the case if the oscillators had chaotic individual dynamics .
in such situation , the value of @xmath4 is expected to control the existence of synchronized states .
we have solved numerically equations ( [ ens ] ) for ensembles of @xmath49 to @xmath50 oscillators with a standard finite - difference scheme . for small values of @xmath51
we find that the above described random - phase initial conditions evolve to a state of synchronized frequency where the phases of all oscillators coincide , @xmath35 for all @xmath11 , @xmath10 .
this fully synchronized state is completely analogous to that of globally coupled identical oscillators without time delays , and corresponds to the approximate solution ( [ appr ] ) for the present homogeneous topology . in this case ,
( [ omega ] ) becomes an autonomous equation for @xmath29 .
the sum in the right - hand side can in fact be explicitly evaluated though its expression depends on @xmath0 being even or odd and the synchronization frequency can be found numerically by standard methods . in general , this equation admits several solutions . for the values of @xmath51 where the state of phase synchronization is encountered , however , there is only one possible value of @xmath29 . at @xmath52 a qualitative change occurs . above this critical value ,
the asymptotic synchronized state is not characterized by a homogeneous phase anymore .
instead , the phase varies linearly along the system , in such a way that a phase difference @xmath53 accumulates in a whole turn around .
the sign of @xmath54 is defined by the initial condition .
symmetry considerations , in fact , indicate that both signs will be found with equal probability over the set of initial conditions that lead to this kind of asymptotic state .
the individual phases are given by @xmath55 , with @xmath56 and @xmath57 an arbitrary constant .
due to the time evolution of @xmath58 a structure propagates around the system at velocity @xmath59 .
similar qualitative changes are found at larger values of @xmath51 . for @xmath60 ,
the asymptotic states modify their phase structure in such a way that the phase difference around the whole system , @xmath61 with @xmath62 , increases progressively .
the corresponding individual evolution is @xmath63 , which defines a propagation velocity @xmath64 .
the synchronization frequency is given by @xmath65.\ ] ] for @xmath66 this reduces to the case of full synchronization found for small @xmath51 .
figure [ f1 ] shows the solutions of equation ( [ omegam ] ) for various values of @xmath67 , and @xmath68 .
bolder curves indicate the intervals where each mode has been observed in the numerical calculations with random - phase initial conditions .
note the zones where more than one solution exist for @xmath66 and @xmath69 .
are the transitions observed at the above quoted values of @xmath51 actual bifurcations , associated with changes in the stability of the asymptotic states ?
in view of the difficulty of dealing with the linear stability problem for a many - dimensional system with time delays such as ( [ ens ] ) @xcite , we choose to answer this question by numerical means . for a given value of @xmath51
we calculate the frequency @xmath29 of a given mode @xmath67 from equation ( [ omegam ] ) and generate an initial condition which corresponds to that mode added with a certain typically random small perturbation .
then , we run the evolution and study the asymptotic behavior .
this has been carried out for @xmath70 at several values of @xmath51 in ( @xmath71 ) , for a 100-oscillator ensemble . in almost all cases
, it has been found that for sufficiently small perturbations the considered states are stable for any value of @xmath51 .
the only exceptions seem to be the states whose frequencies are multiple solutions of equation ( [ omegam ] ) , since in this case the only stable state correspond to the smallest frequency .
the observed transitions are therefore not related to stability changes in the propagation modes .
rather , several modes coexist and the system is multistable . the specific asymptotic state is thus selected by the initial condition .
the fact that from the random - phase initial conditions considered previously the system evolves to a well defined synchronous mode , whose order @xmath67 grows with @xmath51 , suggests that the attraction basins of the various solutions could considerably vary in size as @xmath51 changes .
indeed , from a probabilistic viewpoint , most initial conditions are of the random - phase type .
initial conditions that , for a given value of @xmath51 , do not evolve to the mode marked with a bold line in fig .
[ f1 ] should be considered probabilistically rare .
we consider now the case of a one - dimensional array with free boundaries . here
, the distance between elements can be defined in the standard form , @xmath72 , so that the time delays are @xmath73 . in this
topology sites are not equivalent .
delays for elements near the center of the array are in average lower than for elements towards the boundaries . as a consequence
, no homogeneous stable states are expected for the coupled ensemble .
numerical results show that , in fact , in the asymptotic evolution all the oscillators have the same frequency , given by latexmath:[\[\label{omega2 } \omega = \omega-{\epsilon\over n}\sum_j \sin ( \omega \tau_0 @xmath25 for any nearest - neighbor time delay @xmath51 .
unexpectedly , however , the associated spatial structures not always preserve the topological symmetry of the system , as shown in the following .
our numerical calculations for the case of free boundaries correspond to a 100-oscillator ensemble with the random - phase initial conditions described above . for small values of @xmath51
we find a symmetric asymptotic pattern , @xmath75 , where the central elements have larger phases than near the boundaries ( fig .
[ f2 ] for @xmath76 ) .
this structure corresponds to the approximate solution ( [ appr ] ) which , for this topology , predicts a parabolic phase profile with a maximum at the center of the array . beyond a critical value @xmath77
random - phase initial conditions are instead attracted towards an asymmetric structure , where the phase varies in @xmath78 from one end to the other , and attains a maximum in between .
figure [ f2 ] shows such structure for @xmath79 . in average
, of course , half of the realizations produce the symmetric counterpart of this asymptotic state .
the situation changes again at @xmath80 . beyond this point , stationary structures
are again symmetric , as shown in fig .
[ f2 ] for @xmath81 .
they result however to be more complicated than for small @xmath51 , with inflection points at @xmath82 and a much flatter maximum .
a new critical point occurs at @xmath83 , where phase structures become asymmetric once more ( see fig .
[ f2 ] , for @xmath84 ) .
the phase variation between the ends is similar to that observed for smaller @xmath51 but the intermediate geometry is considerably more complex . an analytical or semi - analytical study of these structures including their existence and stability properties requires considering the consistency problem discussed in connection with eq .
( [ omega ] ) .
fixing @xmath33 , the @xmath30 equations for @xmath31 ( @xmath85 ) read here @xmath86.\ ] ] this problem will be discussed in detail in a separate publication @xcite .
let us stress for the moment that , though the appearance of spatial structures was to be expected in an inhomogeneous system as the present one - dimensional array with free boundaries , these patterns are found to exhibit an unexpected richness upon variation of @xmath51 including , in particular , symmetry breaking .
we have found that an ensemble of identical globally coupled oscillators with finite interaction velocity , which gives origin to time delays , evolves to an asymptotic state where all the oscillators have the same frequency but different phases .
generally , the synchronization frequency differs from the proper frequency of individual oscillators , so that the dynamics of each element in the collective asymptotic motion does not coincide with its individual ( uncoupled ) dynamics .
phases , in turn , are distributed according to spatial patterns with nontrivial topological and dynamical properties .
specifically , in a one - dimensional periodic array several asymptotic states coexist , corresponding to propagation modes with different velocities . in a bounded one - dimensional array
we have observed stationary phase structures whose symmetry properties depend on the size of time delays .
these features , which are reminiscent of the behavior of reaction - diffusion systems with local interactions , point out sharp differences with the collective motion of coupled oscillators without time delays .
it is natural to ask whether any structure similar to those described above is observed in other , more complex topologies . to advance an answer to this question
, we have performed a preliminary analysis of a two - dimensional array of @xmath87 elements with periodic boundary conditions . in this case , each element can be labeled by two indices , @xmath88 and @xmath89 , according to its cartesian coordinates in the lattice . for algorithmic convenience
we have defined the distance between elements as @xmath90 , with @xmath91 in our case . as above , the delay time is @xmath92 . in complete agreement with the one - dimensional analog
, we have here found that for small @xmath51 the system synchronizes both in frequency and phase . beyond a critical value @xmath93 , instead , propagating phase patterns are observed .
figure [ f3 ] shows the simplest of these patterns , corresponding to the propagation mode with @xmath94 .
work in progress is being devoted to the detailed characterization of the phase structures described in this note , as well as those that could arise in other topologies .
the next step will be to study the effects of the present kind of time delays in ensembles formed by chaotic oscillators , where coupling competes as a stabilizing mechanism against the inherently unstable dynamics of individual elements .
this work was partially carried out at the abdus salam international centre for theoretical physics .
the author wishes to thank the centre for hospitality .
g. m. shepherd , _ neurobiology _ ( oxford university press , new york , 1983 ) ; j. foss , a. longtin , b. mensour , and j. milton , phys . rev
* 76 * ( 1996 ) 708 ; l. kleinrock , _ communication nets : stochastic message flow and delay _
( dover , new york , 1964 ) .
|
we consider an ensemble of globally coupled phase oscillators whose interaction is transmitted at finite speed .
this introduces time delays , which make the spatial coordinates relevant in spite of the infinite range of the interaction .
we show that one - dimensional arrays synchronize in an asymptotic state where all the oscillators have the same frequency , whereas their phases are distributed in spatial structures that in the case of periodic boundaries can propagate , much as in coupled systems with local interactions .
|
occuring in a broad class of natural phenomena metastability has been intensively studied in different research fields ranging from condensed matter to chemical physics , from particle and nuclear physics to cosmology . while the mathematics of the metastable phase in statistical physics was elaborated by langer @xcite forty years ago , the thorough extension of the formalism to the quantum field theory came in the seventies of the last century @xcite .
as the semiclassical approximation is known to be appropriate @xcite to deal with quantum tunneling , the path integral method in the euclidean version @xcite has emerged as the fundamental tool for quantitative theories of the decay rate @xmath1 of a metastable state .
the latter is given in semiclassical theory by @xmath2 \label{eq:100}\end{aligned}\ ] ] where @xmath3 and @xmath4 depend on the specific shape of the potential .
in the path integral approach , @xmath4 is the euclidean action calculated at the classical path around which the action varies the least while @xmath3 is obtained by summing over the quantum fluctuations paths . in quadratic approximation
the quantum fluctuations are decoupled and their contribution can be evaluated by the theory of the functional determinants which implements the path integral method @xcite .
( [ eq:100 ] ) is formally identical to the tunneling energy in the bistable @xmath5 potential @xcite which admits an instanton @xcite as the classical path interpolating between the two vacua .
the fundamental difference lies however in the fact that the instanton minimizes the euclidean action whereas the classical path in the metastable potential , the bounce , is a saddle point for the action .
hence the bounce time derivative , which is eigenfunction of the quantum fluctuations sturm - liouville operator with zero eigenvalue ( due to time translational invariance ) , is not the ground state .
there must be a negative eigenvalue @xcite in the fluctuation spectrum causing an imaginary contribution to the total partition function .
the decay rate of the false vacuum arises precisely from this negative eigenvalue which , moreover , is unique as the bounce path has one node along the time axis .
while the general formalism of metastability is settled at least in one dimension and in the semiclassical method @xcite , the evaluations of @xmath1 for specific systems are generally based on the langer - coleman approach @xcite which assumes a system with _
infinite size _ @xmath6 and therefore , strictly speaking , holds in the zero temperature limit of the quantum field theory once @xmath6 is taken equal to @xmath7 .
the same formal approach can be used in classical systems to calculate the noise - induced transition rate between locally stable states , for example the domain reversal rate in micromagnets @xcite and the instabilities in metallic nanowires @xcite . in such systems
however the character of the transition between activation regimes crucially depends on the finite spatial extent @xcite , thus quantitative estimates of the kramer escape rate require an extension of the langer - coleman method to the more complex
_ finite size _ case . in quantum tunneling systems ,
the counterpart of the latter is represented by the _
finite t _ quantum regime below the crossover temperature @xmath8 at which thermal activated processes set in . below @xmath8 , the decay rate stems from purely quantum fluctuations effects and the @xmath1 dependence on @xmath0 is governed by the softening of the fluctuation spectrum . to determine this phenomenon with accuracy one has to face the non trivial task of computing fluctuation functional determinants at low but finite @xmath0 @xcite which also includes a knowledge of the negative eigenvalue in the spectrum .
this paper focuses on this issues developing the finite temperature theory of metastability below @xmath8 for the case of a model cubic potential .
in particular , the softening of the fluctuations is analysed quantitatively by means of the jacobi elliptic functions formalism @xcite which has the advantage to monitor the evolution of the system at any @xmath0 , free from approximations regarding the form of the bounce and its fluctuation spectrum .
the explicit equations which determine such softening are provided .
activation effects @xcite are not discussed in this work which assumes a closed system .
metastability in the presence of dissipation has been studied in ref.@xcite . as in the space - time euclidean path integral
the time is an inverse temperature , the zero @xmath0 theory assumes that the bounce classical motion lasts an _ infinite time _ whereas the finite @xmath0 theory realistically admits that the excursion time for the bounce may be finite .
this consideration permits to derive the family of energy dependent paths which solve the classical equation of motion as shown in section 2 for the cubic potential . in section 3 , using the theory of the functional determinants , i determine analytically and compute the quantum fluctuations contribution to the path integral .
the solution of the lam equation which governs the fluctuation spectrum is presented together with the evaluation of the temperature effects on the lowest lying eigenmodes and their eigenvalues .
section 4 contains the results for the decay rate of the metastable cubic potential and a comparison with the case of the metastable quartic potential .
some final remarks are made in section 5 .
consider a particle of mass @xmath9 in the one dimensional cubic potential : @xmath10 plotted in fig .
[ fig:1](a ) for several energies @xmath11 .
say @xmath12 the position of the top of barrier whose height is @xmath13 with @xmath14 .
i take throughout the paper , @xmath15 and @xmath16 with @xmath17 being the electron mass . at @xmath18
the particle is in a local minimum , see fig .
[ fig:1](a ) , from which it can not escape classically and , in the real time formalism , only @xmath19 solves the classical equation of motion .
such local minimum is however metastable as quantum fluctuations allow the particle to explore the abyss at @xmath20 .
let s see how .
a non trivial classical solution can be found in the euclidean space .
in fact , performing a wick rotation from the real to the imaginary time , @xmath21 , the equation of motion reads : @xmath22 where @xmath23 means derivative with respect to @xmath24 .
the particle path @xmath25 has been split in the sum of a classical and a quantum component , @xmath26 .
the wick rotation is equivalent to turn the potential upside down with respect to the real time as shown in fig .
[ fig:1](b ) for @xmath27 . in the reversed potential
the classical motion can take place . integrating eq .
( [ eq:1 ] ) , one gets : @xmath28 with the constant @xmath29 representing the classical energy . defining : @xmath30 eq .
( [ eq:2 ] ) is easily integrated to yield : @xmath31 where @xmath32 is the center of motion between the turning points at which the path velocity vanishes .
thus the physical picture set by the boundary conditions is that , see fig .
[ fig:1](b ) , of a particle starting from @xmath33 at the time @xmath34 , reaching @xmath35 at @xmath36 and returning to the initial position at @xmath37 .
then , eq . ( [ eq:57 ] ) has a time reversal invariant solution whose period @xmath6 is finite and dependent on @xmath29 .
phase transitions phenomena in spatially extended systems are analysed by formally similar models with @xmath6 being the size of the system @xcite .
looking at fig .
[ fig:1](b ) , one captures qualitatively that the amplitude @xmath38 attains the largest value for the @xmath39 motion while @xmath33 and @xmath40 coincide .
@xmath40 becomes negative for @xmath41 motions .
the turning points @xmath35 , @xmath33 , @xmath40 are given by the zeros of the equation @xmath42 ( @xmath43 ) which admits three real solutions for @xmath44 $ ] , that is for @xmath45 $ ] .
after some algebra i find : @xmath46 at the bounds of the energy range , eq .
( [ eq:58 ] ) yields : @xmath47 thus , at the sphaleron energy @xmath48 @xcite , the amplitude of the finite time solution has to shrink into a point .
let s find the general solution of eq .
( [ eq:57 ] ) by pinning the center of motion at @xmath49 and using the result @xcite : @xmath50 where @xmath51 is the elliptic integral of the first kind with amplitude @xmath52 and modulus @xmath53 . then , through eqs .
( [ eq:56 ] ) , ( [ eq:57 ] ) , ( [ eq:61 ] ) , i derive the bounce solution of the _ finite time _
theory : @xmath54 \ , \nonumber \\ & & \varpi=\ , \sqrt{{\chi_1 - \chi_3}}\,{\omega \over 2}(\tau - \tau_0)\ , \nonumber \\ \label{eq:62}\end{aligned}\ ] ] @xmath55 and @xmath56 are the _ sine- _ and _ cosine- _ amplitudes respectively @xcite .
the modulus @xmath53 incorporates the classical mechanics of the problem through the second of eq .
( [ eq:56 ] ) and eq .
( [ eq:58 ] ) . at e = 0 , @xmath57 , the bounce of the _ infinite time _
theory is recovered : @xmath58 at the sphaleron energy , @xmath59 , the bounce solution is in fact a point - like object set at the bottom of the valley in the reversed potential : @xmath60 .
thus , eq . ( [ eq:62 ] ) defines the transition state which is a saddle for the action below the sphaleron . representing the jacobi elliptic functions in eq .
( [ eq:62 ] ) through trigonometric series
@xcite suitable to computation , i plot the classical path in fig .
[ fig:2 ] which makes evident how the bounce amplitude contracts by increasing the _ energy over potential height _ ratio ( in absolute value ) .
while the bounce can be physically interpreted as a pair of domain walls , it is clear that the extent of the @xmath61-region over which the domain walls separate the turning points continuously shrinks by approaching the crossover .
the center of motion @xmath32 has been set equal to zero with no loss of generality being the system invariant for time translation . as the bounce is a combination of squared jacobi elliptic functions ,
its period is @xmath62 with @xmath63 being the complete elliptic integral of the first kind @xcite .
hence , from eq .
( [ eq:62 ] ) , i get : @xmath64 which establishes the relation between the oscillation period and the classical energy embedded in the turning points . as in the matsubara formalism
, one can map the imaginary time onto the temperature axis , @xmath65 , where @xmath66 is the temperature at which the particle makes the journey to and from the edge of the abyss for a given @xmath29 . then , only periodic bounces whose period equals the inverse temperature determine the decay rate and the _ finite time _ theory can be viewed as a finite @xmath66 theory . from eq .
( [ eq:64 ] ) i get : @xmath67 the @xmath6 and @xmath66 values given in fig .
[ fig:2 ] are computed on the base of eqs .
( [ eq:64 ] , ( [ eq:64a ] ) respectively . at the sphaleron , eq .
( [ eq:64a ] ) yields @xmath68 that marks the transition temperature between quantum and activated regimes .
such value represents the upper bound for the validity of the model and precisely sets the goldanskii criterion @xcite for the cubic potential .
taking @xmath69 , the following calculations are carried out in the low temperature range up to @xmath70 .
the classical action @xmath71 $ ] for the bounce in the finite temperature theory can be computed either in terms of the path velocity @xmath72 : @xmath73=\ , 2 m \int_{-l/2}^{\tau_0}d\tau [ \dot{x}_{cl}(\tau)]^2 - e\cdot l(e ) \ , \nonumber \\ & & \dot{x}_{cl}(\tau)=\,{{3a } \over 2}\mathcal{f}\cdot sn(\varpi , p)cn(\varpi , p)dn(\varpi , p)\ , \nonumber \\ & & \mathcal{f}=\,-\omega ( \chi_1 - \chi_2 ) \sqrt{{\chi_1 - \chi_3 } } \label{eq:65}\end{aligned}\ ] ] ( @xmath74 is the _
delta- _ amplitude @xcite ) or , in terms of the potential @xmath75 : @xmath76}}=\ , { { 27v(a ) } \over { \omega } } \int_{\chi_2}^{\chi_1}d\chi \sqrt{{\kappa } + \chi^2 - \chi^3 } - { e\cdot l(e ) } \ , \nonumber \\ \label{eq:65a}\end{aligned}\ ] ] both ways require computation of @xmath77 through eqs .
( [ eq:56 ] ) , ( [ eq:58 ] ) , ( [ eq:64 ] ) . in the @xmath78 limit , i get the result @xmath79 } \over \hbar } \rightarrow \ , { { 6 m^3 \omega^5 } \over { 5\hbar \gamma^2 } } \label{eq:66}\end{aligned}\ ] ] which serves as testbench for the computational method .
the inverse dependence on the anharmonic force constant @xmath80 reflects the well known fact that metastable systems are intrinsecally non perturbative and provides the fundamental motivation for the semiclassical treatment .
( [ eq:66 ] ) permits to set the potential parameters such as the condition @xmath81 \gg \hbar}$ ] holds and the semiclassical method is thus justified . as @xmath9 and @xmath12 have been taken constant , @xmath82 } \over \hbar}\propto \omega$ ] in the @xmath78 limit . the bounce velocity and the classical action
are displayed in fig .
[ fig:3 ] and fig .
[ fig:4 ] respectively .
note that @xmath81}$ ] decreases smoothly with increasing temperature and absolute value of energy ( inset ( a ) in fig .
[ fig:4 ] ) suggesting that the transition to the activated regime above @xmath83 is of second order @xcite .
although i m dealing here with a closed system , it is worth emphasizing that the inclusion of a dissipative environment @xcite would decrease @xmath83 but would not change the character of the crossover . in general , both first and second order transitions have been found in tunneling systems depending on the shape of the potential @xcite .
the criterion used to establish the order of the transitions in periodic tunneling systems has been formulated by chudnovsky @xcite on the base of the behavior of @xmath77 : a monotonic dependence of the period below the sphaleron points to a smooth crossover whereas a nonmonotonic @xmath77 indicates a sharp transition . the inset ( b ) in fig .
[ fig:4 ] clearly shows that the former case applies to the cubic potential in eq .
( [ eq:55 ] ) .
@xmath77 decreases monotonically and , consistently , the action versus @xmath66 is convex upwards . at the sphaleron
, i find numerically : @xmath84 .
note that such value corresponds to @xmath85 @xcite and this is not by accident .
in fact at the sphaleron the bounce is a point , that is a static solution of eq .
( [ eq:2 ] ) but , near the sphaleron , the periodic path is the sum of the sphaleron and an oscillation with negative eigenvalue @xmath86 whose period tends to @xmath87 @xcite .
then one infers that , for @xmath88 , @xmath89 .
i ll come back to this point towards the end of the next section .
let s expand the action around the classical path up to second order in the fluctuations @xmath90\sim \,a[x_{cl } ] + { \delta } a[\eta ] + { 1 \over { 2 ! } } { \delta}^2 a[\eta]\ , \nonumber \\ & & { \delta } a[\eta]=\ , \int_{-l/2}^{l/2 } d\tau \biggl({m } \dot{x}_{cl}(\tau ) \dot{\eta}(\tau ) + v'(x_{cl}(\tau)){\eta}(\tau )
\biggr ) \ , \nonumber \\ & & { \delta}^2 a[\eta]\equiv \,2 a_f[\eta ] \ ,
\nonumber \\ & & a_f[\eta]=\,\int_{-l/2}^{l/2 } d\tau \biggl({m \over 2 } \dot{\eta}^2(\tau ) + { 1 \over 2}v''(x_{cl}(\tau))\eta^2(\tau ) \biggr ) \ , \nonumber \\ \label{eq:66+}\end{aligned}\ ] ] being the classical solution an extremum for the action , the first functional derivative vanishes and , after partial integration in the first term of @xmath91 , one gets the euler - lagrange eq . ( [ eq:1 ] ) .
further , differentiating eq .
( [ eq:1 ] ) with respect to @xmath61 , one observes that @xmath92 solves the homogeneous equation associated to the second order schrdinger - like differential operator in @xmath93 $ ] @xmath94 where the @xmath95 are the fluctuations eigenvalues while the coefficients @xmath96 of the series expansion in ortonormal components @xmath97 define the measure of the fluctuation paths integration .
then , as a general consequence of the @xmath61-translational invariance , @xmath92 is proportional to the eigenmode @xmath98 with @xmath99
. the latter can not be the ground state as , see fig .
[ fig:3 ] , @xmath92 has one node hence , the quantum fluctuation ground state has negative eigenvalue . from eq .
( [ eq:65 ] ) , note that for any two points @xmath100 such that @xmath101 , @xmath102 .
the important consequence is that the fluctuation eigenmodes obey periodic boundary conditions ( pbc ) . after these observations we are ready to calculate the space - time euclidean path integral between the positions @xmath103 and @xmath104 connected in the time @xmath6 . in the semiclassical model
, it is given by @xmath105 } \over { \hbar } } \biggr ] \cdot \int d\eta \exp\biggl[- { { a_f[\eta ] } \over { \hbar } } \biggr ] \
, \nonumber \\ & & \int d\eta=\,\aleph \prod_{n=\,-1}^{\infty } \int_{-\infty}^{\infty } { { d\varsigma_n}\over { \sqrt{2\pi\hbar / m } } } \label{eq:66+++}\end{aligned}\ ] ] @xmath106 depends only on the functional integral measure . @xmath103 and @xmath104 coincide for the periodic bounce , thus eq .
( [ eq:66+++ ] ) represents the single bounce contribution @xmath107 to the total partition function @xmath108 .
the latter also contains the effects of all multiple ( non interacting ) excursions to and from the abyss which is equivalent to sum @xcite over an infinite number of single bounce contributions like @xmath107 .
moreover , also the static solution of eq .
( [ eq:1 ] ) , @xmath19 contributes with the harmonic partition function @xmath109 which can be easily determined using the measure in eq .
( [ eq:66+++ ] ) . summing up
, @xmath108 is given by @xmath110|^{-1/2 } \label{eq:66a+++}\end{aligned}\ ] ] @xmath111 $ ] being the harmonic fluctuation determinant .
being the decay rate @xmath1 proportional to the imaginary exponential argument through the feynman - kac formula @xcite , it follows that there is no need to determine @xmath106 as it cancels out in the ratio @xmath112 .
let s proceed to evaluate the fluctuation term in @xmath107 which , formally , is worked out by carrying out gaussian path integrals yielding : @xmath113 } \over { \hbar } } \biggr]= \ , \aleph \cdot det\bigl[\hat{o}\bigr]^{-1/2 } \
, \nonumber \\ & & det[\hat{o}]\equiv \ , \prod_{n=\,-1}^{\infty}\varepsilon_n \label{eq:66++++}\end{aligned}\ ] ] eq . ( [ eq:66++++ ] )
is however divergent due to the goldstone mode which reflects the fact that @xmath32 , the center of the bounce , can be located arbitrarily inside @xmath6 .
however such mode can be treated separately @xcite : the divergent integral over the coordinate @xmath114 associated to the zero mode in the measure @xmath115 is transformed into a @xmath116 integral . accordingly the eigenvalue @xmath99 is extracted from @xmath117 $ ] and its contribution to eq .
( [ eq:66++++ ] ) is replaced as follows @xmath118 to be rigorous , this replacement holds in the approximation of quadratic fluctuations @xcite while higher order terms may be significant around the crossover .
it is also worth noticing that eq .
( [ eq:66+++++ ] ) is often encountered in the form @xmath119}\over { 2\pi\hbar}}}l$ ] .
however , this is correct only in the zero @xmath0 limit where @xmath71 $ ] equals @xmath120 ( see eq .
( [ eq:65 ] ) ) while , approaching @xmath121 , the difference between the two objects gets large .
exactly at @xmath121 , the norm of the bounce velocity vanishes while the action is finite .
this fact is crucial in establishing the behavior of the decay rate at the crossover as shown below .
now i tackle the problem of the evaluation of the regularized determinant @xmath122 $ ] defined by @xmath117=\,\varepsilon_0 \cdot det^r[\hat{o}]$ ] .
the calculation of @xmath122 $ ] requires knowledge only of the classical paths .
this is a fundamental feature of the theory of functional determinants of second order differential operators first developed for dirichlet boundary conditions @xcite and then extended to general operators and boundary conditions in several ways @xcite .
as shown above the path velocity obeys pbc for any two points @xmath100 separated by the period @xmath62 .
the latter corresponds to the oscillation period @xmath6 along the @xmath61-axis .
it can be easily checked that also the path acceleration fulfills the pbc .
then , the regularized determinant is given by @xmath123=\,{{<f_0 |f_0 > \bigl(f_1(\varpi_2 ) - f_1(\varpi_1)\bigr ) } \over { { f_0(\varpi_1 ) w(f_0 , f_1 ) } } } \label{eq:67}\end{aligned}\ ] ] where @xmath124 are two independent solutions of the homogeneous equation : @xmath125 .
@xmath126 is their wronskian and @xmath127 is the squared norm .
@xmath128 is obviously @xmath129 while @xmath130 can be taken as @xmath131 using eq .
( [ eq:62 ] ) , derivatives and properties of the elliptic functions , i derive for @xmath130 : @xmath132\ , \nonumber \\ { { \partial \chi_1 } \over { \partial q}}&=&\ , { { 1 - p^2 } \over { 2(1 - p^2 + p^4)^{3/2}}}\ , \nonumber \\ { { \partial \chi_2 } \over { \partial q}}&=&\ , { { -1 } \over { 2(1 - p^2 + p^4)^{3/2}}}\ , \nonumber \\ { { \partial f(\lambda , p)}\over { \partial q}}&=&\,{1 \over { 2\bar{p}^2}}\biggl [ { { e(\lambda , p ) - \bar{p}^2 f(\lambda , p ) } \over { p^2 } } - { { \sin\lambda \cos\lambda } \over { \sqrt{1 - p^2\sin^2\lambda } } } \biggr]\ , \nonumber \\ & & \bar{p}^2=\ , 1 - p^2 \label{eq:68}\end{aligned}\ ] ] where @xmath133 is the elliptic integral of the second kind .
the wronskian , being constant along @xmath61 , can be calculated in any convenient point .
let s take @xmath32 as @xmath134 .
then : @xmath135 and @xmath136 given in eq .
( [ eq:66+++++ ] ) .
thus , all ingredients are known to calculate @xmath137}$ ] : evaluating eq .
( [ eq:68 ] ) on the @xmath138 range boundaries , eq .
( [ eq:67 ] ) transforms into @xmath139 } = \,{{2 } \over { \omega \sqrt{\chi_1 - \chi_3 } \bar{p}^2 } } \biggl [ { { e(\pi/2,p ) - \bar{p}^2k(p ) } \over { p^2 } } \biggr ] \cdot { { < f_0 |f_0 > } \over { w(f_0 , f_1 ) } } \label{eq:70}\end{aligned}\ ] ] which can be computed using eqs .
( [ eq:66+++++ ] ) , ( [ eq:69 ] ) .
it is however known in the theory of functional determinants @xcite that only ratios of determinants are meaningful in value and sign , such ratios arising naturally in the path integral method as it has been pointed out above .
in fact , @xmath122 $ ] would diverge in the @xmath140 limit due to the fact that the determinant is the product over an infinite number of eigenvalues with magnitude greater than one . consistently with eq .
( [ eq:66a+++ ] ) , @xmath122 $ ] has to be normalized over @xmath111 $ ] ( @xmath141 ) which , in the case of pbc , is : @xmath142}=\,-4\sinh^2(\omega l/2)$ ] .
the normalization cancels the exponential divergence and makes the ratio finite .
then , observing that for @xmath143 ( @xmath140 ) : @xmath144 from eq .
( [ eq:70 ] ) , i finally get the finite ratio @xmath145 } \over { det[\hat{h } ] } } \rightarrow -{{1 } \over { 60\omega^2 } } \label{eq:71+}\end{aligned}\ ] ] the dimensionality @xmath146 $ ] correctly accounts for the fact that one eigenvalue has been extracted from @xmath122 $ ] .
the computation of eq .
( [ eq:70 ] ) becomes rather time consuming in the @xmath140 limit where elliptic integrals need to be evaluated with high accuracy in order to get the correct exponential divergence .
the plot of @xmath122 $ ] versus temperature is shown in fig .
[ fig:5 ] up to @xmath147 while the inset displays the computed inverse ratio normalized over @xmath148 : the @xmath140 limit is in fact an excellent estimate up to @xmath149 whereas a strong deviation is found at larger @xmath66 up to @xmath150 .
exactly at @xmath8 , @xmath122 $ ] vanishes thus causing the divergence of the inverse ratio .
this divergence may look surprising since the zero mode had been extracted from the determinant : as it will be unravelled in the next subsection , there must be a quantum fluctuation mode which softens by increasing @xmath66 and ultimately vanishes at the sphaleron .
i want to emphasize the key role played by the bounce velocity norm in getting @xmath122 \sim 0 $ ] at @xmath151 : close to the sphaleron , @xmath152 and @xmath153 thus the squared norm drops to zero faster than any other term in eq .
( [ eq:70 ] ) .
this feature is made evident in fig .
[ fig:6 ] where the plots of @xmath154}/ { < f_0 | f_0>}}$ ] and the squared norm are given separately against temperature up to @xmath121 .
eventually , note that the determinants ratio is negative and this sign has physical meaning as it is due precisely to the negative ground state of the fluctuation spectrum . as @xmath112 contributes to the partition function by the square root of the fluctuation determinants ( inverse ) ratio it follows that such contribution is purely imaginary . to weigh the effect of the negative eigenvalue @xmath86
, its analytical expression has to be determined at finite @xmath66 .
this is done in the next subsection .
let s take eq .
( [ eq:66++ ] ) with the second derivative of the potential given by @xmath155 then , using eq .
( [ eq:62 ] ) and working out the algebra , i get the stability equation which governs the fluctuation spectrum @xmath156 \eta_n(\tau ) \ , \nonumber \\ & & \mathcal{a}_n=\,{{4(1 - 3\chi_1 ) } \over { \chi_1 - \chi_3 } } - { { 4\varepsilon_n } \over { \omega^2(\chi_1 - \chi_3 ) } } \ , \nonumber \\ & & 12 \equiv l(l + 1 ) \label{eq:73}\end{aligned}\ ] ] this is the lam equation in the jacobian form for the case @xmath157 @xcite . for a given @xmath158 and @xmath53 , eq .
( [ eq:73 ] ) yields periodic solutions ( which can be expanded in infinite series ) for an infinite sequence of characteristic @xmath159 values .
the continuum of the fluctuation spectrum stems from this sequence .
however , being @xmath158 positive and integer , the first @xmath160 solutions of eq .
( [ eq:73 ] ) are not infinite series but polynomials in the jacobi elliptic functions with real period @xmath62 or @xmath161 .
being the period of the potential @xcite , @xmath62 plays the role of a lattice constant .
then , eq . ( [ eq:73 ] ) admits 7 polynomial solutions with eigenvalues @xmath162 $ ] , from which the corresponding @xmath95 are derived @xcite : @xmath163 \bigr ) \ ,
\nonumber \\ & & \varepsilon_{-3}=\,\omega^2\bigl(\alpha_1 + \alpha_2\bigl[5(1 + p^2 ) - 2\sqrt{4p^4 - 7p^2 + 4}\bigr ] \bigr ) \ ,
\nonumber \\ & & \varepsilon_{2}=\,\omega^2\bigl(\alpha_1 + \alpha_2\bigl[5 + 2p^2 + 2\sqrt{p^4 - p^2 + 4}\bigr ] \bigr ) \ , \nonumber \\ & & \varepsilon_{-2}=\,\omega^2\bigl(\alpha_1 + \alpha_2\bigl[5 + 2p^2 - 2\sqrt{p^4 - p^2 + 4}\bigr ] \bigr ) \ , \nonumber \\ & & \varepsilon_1=\ , \omega^2\bigl(\alpha_1 + \alpha_2\bigl[2 + 5p^2 + 2\sqrt{4p^4 - p^2 + 1}\bigr ] \bigr ) \nonumber \\ & & \varepsilon_{-1}=\ , \omega^2\bigl(\alpha_1 + \alpha_2 \bigl[2 + 5p^2 - 2\sqrt{4p^4 - p^2 + 1}\bigr ] \bigr)\ , \nonumber \\ & & \varepsilon_0=\,0 \ , \nonumber \\ & & \alpha_1\equiv\ , 1 - 3\chi_1 \ ,
\nonumber \\ & & \alpha_2\equiv\ , { { \chi_1 - \chi_3 } \over 4 } \label{eq:75}\end{aligned}\ ] ] the physically relevant feature is that the @xmath95 depend , through @xmath53 , on the system size @xmath6 ( or on @xmath66 ) .
however not all the @xmath95 are good fluctuation eigenvalues .
in fact the first four values have to be discarded as their eigenfunctions do not fulfill the pbc required for the fluctuation components : @xmath164 .
thus , three good eigenvalues in polynomial form are found : _
i ) _ @xmath99 is the zero mode eigenvalue which is consistently found also through the stability equation . _ ii ) _
@xmath165 lies in the continuum and vanishes at @xmath59 : this is the soft mode responsible for the sharp drop observed in @xmath122 $ ] as the sphaleron is approached . at @xmath83 , @xmath165 and @xmath166
merge consistently with the double degeneracy of the corresponding eigenmodes above the crossover .
@xmath86 is the negative eigenvalue responsible for the decay of the metastable state . in fig .
[ fig:7 ] , the ratio @xmath122/\varepsilon_{-1}$ ] and @xmath86 are plotted separately to emphasize the role of the latter .
the dependence on @xmath66 is obtained as usual by means of eq .
( [ eq:64a ] ) . while the overall shape of @xmath122 $ ] is not essentially modified by extracting @xmath86 ( compare with fig .
[ fig:5 ] ) , the interesting feature is represented by the softening of @xmath86 with respect to the value @xmath167 found at @xmath168 .
the substantial reduction ( in absolute value ) starts up at @xmath169 , that is in the same range at which the classical properties deviate from the predictions of the zero temperature theory . finally , at the sphaleron , from eq .
( [ eq:75 ] ) i get @xmath170 consistently with the observation made at the end of the previous section .
this completes the analysis of the quantum fluctuation spectrum .
the decay rate in eq .
( [ eq:100 ] ) for the finite temperature theory can be now computed as the coefficients @xmath3 and @xmath4 , identified with @xmath171/ det[\hat{o}]\bigr|}/l$ ] and @xmath71 $ ] respectively , have been analysed in detail . the general expression for @xmath172 is : @xmath173 } \over { det^{rr } [ \hat{o}]{\bigl|\varepsilon_{-1}\bigr| } } } \exp\biggl[-{{a[x_{cl } ] } \over { \hbar } } \biggr ] \ , \nonumber \\
\label{eq:76}\end{aligned}\ ] ] with @xmath174\equiv \,det^r[\hat{o}]/\varepsilon_{-1}$ ] .
essentially , the two bound state eigenvalues @xmath175 and @xmath86 have been extracted from @xmath117 $ ] and treated separately .
it is now clear that , in order to calculate the decay rate , there would be no need to compute @xmath176 due to the zero mode @xcite as it cancels out with the same term embedded in @xmath137}$ ] and discussed above .
thus , it is the piece of determinant in fig .
[ fig:6](a ) which enters the computation of @xmath1 . eq . ( [ eq:76 ] ) is plotted in fig .
[ fig:8 ] against temperature up to @xmath177 together with the constant decay rate of the _ zero temperature _ theory .
there is a substantial increase in @xmath1 above @xmath149 where the combined effects of quantum fluctuations and classical action softening become evident . approaching the crossover ( at @xmath178 )
, @xmath1 reaches a peak value which is larger by a factor @xmath179 with respect to the value of the zero @xmath0 theory although the linewidth remains much smaller than the oscillator energy , @xmath180 .
certainly , such ratio strongly depends on the particle mass through the exponential gamow factor as @xmath81 } \propto m$ ] . beyond @xmath181 and _ before _
@xmath121 , the quantum decay rate smoothly merges with the classical arrhenius factor as @xmath82}/ { \hbar } } \rightarrow v(a)/k_bt^*_c$ ] . in principle
this should occur exactly at @xmath121 once the nonuniform bounce leaves room to the constant solution @xmath60 .
but close to @xmath121 the decay rate drops to zero as @xmath182 signalling that the quantum tunneling effects is ending .
the power law dependence is driven by @xmath183 } \propto p$ ] .
for this reason the arrhenius behavior in the decay rate , which is a continuous function of @xmath66 , takes over slightly before @xmath121 and precisely at @xmath184 corresponding to the symbol in fig .
[ fig:8 ] .
the inset shows the plot of the quantum statistical decay rate @xmath185 @xcite which i have computed , on the base of the wkb turning point formula @xcite , by summing over the particle energies in the cubic potential case . in the quantum regime , the calculation converges by taking about 1000 energy values ( below the sphaleron @xmath186 ) in the boltzmann average . in the activated regime ( with energies above the sphaleron ) , @xmath185 continuously grows with exponential behavior while , the matching between the quantum and activated @xmath185 plots occurs at around @xmath187 , with states of energy @xmath188 yielding a significant contribution to @xmath185 in the classical regime . the deviation with respect to the @xmath189 result found in the quantum decay rate @xmath172 ( and culminating in the peak value ) is interesting also in view of a comparison with classically activated systems in which spatio - temporal noise induces transitions between locally stable states of a nonlinear potential . the changes in radius and the stability conditions of metastable metallic nanowires are an example of current interest @xcite . in these systems
, the finite size shapes the activation behavior and a power - law divergence in the escape rate with critical exponent @xmath190 is predicted once a critical lenghtscale is approached . in the decay rate of eq .
( [ eq:76 ] ) the small parameter is @xmath191 which , unlike the noise in classical systems , can not be varied as a function of @xmath0 ( or @xmath6 ) .
accordingly the quantum tunneling decay rate is always small .
similar conclusions regarding the transition can be drawn by carrying out the study of @xmath1 for the metastable quartic potential @xmath192 according to the same semiclassical method described so far .
the parameters @xmath9 , @xmath193 and @xmath12 have been taken as in the cubic case while @xmath194 , in units @xmath195 , is given by @xmath196 .
there are in fact some fundamental differences which may be summarized as follows : _ 1 ) _ for the quartic potential , the bounce solution is linear in the jacobi elliptic @xmath197-amplitude and the crossover temperature @xmath83 is larger by a factor @xmath198 with respect to the value given in eq .
( [ eq:64b ] ) .
_ 2 ) _ the classical action is slightly larger than that computed in fig .
[ fig:4 ] and , in the @xmath78 limit , @xmath71\rightarrow \ , { { 4 m^2 \omega^3 } / { ( 3\hbar\delta)}}$ ] . _
3 ) _ the _ harmonic oscillator over regularized fluctuations determinants _ ratio is smaller and , in the @xmath140 limit : @xmath199 } / { det^r[\hat{o } ] } } \rightarrow - { 12\omega^2}$ ] . _
4 ) _ the lam equation , which has to be solved for the case @xmath200 , has the negative eigenvalue scaling from @xmath201 at @xmath168 to @xmath202 at @xmath203 .
all these features work together to decrease the decay rate of the metastable quartic potential with respect to the cubic one .
the quantitative results are reported on in fig .
[ fig:9 ] by plotting @xmath1 for the two potentials .
it appears that , at a given @xmath66 within the quantum regime , the life of a particle set in a metastable quartic potential is longer ( by an order of magnitude ) than in a cubic potential with the same structural parameters .
close to @xmath121 , the decay rate of the quartic potential shows the same behavior found in the cubic case as a consequence of the softening of the polynomial fluctuation eigenmode with lowest lying positive eigenvalue .
the semiclassical path integral method has been applied to compute the decay rate of a metastable cubic potential at finite temperature .
i have considered a non dissipative system where the inverse temperature , introduced as the imaginary time of the euclidean path integral , represents a measure of the system _ size_. solving the euler - lagrange equation for the cubic potential i have determined the classical background , the bounce , in terms of jacobi elliptic functions whose energy dependent modulus keeps track of the classical mechanics .
the periodicity property of the elliptic solution sets the fundamental relation between bounce oscillation period and classical energy . as the former varies monotonically with the latter ,
the transition from the quantum to the activated regime is in fact a temperature driven smooth crossover occuring at @xmath204 .
then the quantum regime may cover a broader temperature range in systems with characteristic frequencies larger than the value taken here . in the latter cases
however the quadratic approximation for the quantum fluctuations may become questionable . in general , the character of this transition in tunneling systems may depend on the form of the potential . for the metastable potential the classical background represented by the bounce has zero topological charge , fulfills periodic boundary conditions and the amplitude of its motion progressively shrinks by increasing @xmath66 . at @xmath8 ,
the bounce degenerates into a point , the constant solution which determines the classical arrhenius factor .
then the nonuniform bounce samples a @xmath66 dependent portion of the ( reversed ) potential valley and continuosly adapts its shape to join the turning points .
that s why the crossover is smooth . as in the semiclassical method
classical and quantum degrees of freedom are intertwinned the quantum fluctuation spectrum can be evaluated , by the theory of the functional determinants , in terms of the classical bounce .
more precisely , in terms of its time and modulus derivatives .
i have obtained a compact expression for the fluctuations contribution to the path integral and computed the renormalization of the fluctuation spectrum due to temperature effects inside the quantum regime .
the overall softening is remarkable : it shows up at about @xmath205 whereas at lower temperatures the predictions of the zero @xmath0 ( infinite time ) theory are accurate .
the polynomial eigenmodes of the spectrum have been analytically determined by solving the stability equation associated to the sturm - liouville differential operator , the lam equation .
the focus on the ground state is justified by the fact that its eigenvalue is negative hence , it is responsible for the decay rate @xmath1 of the metastable potential .
the negative eigenvalue softens considerably by approaching the sphaleron where , in turn , it determines the oscillation period of the classical bounce . also the positive polynomial eigenvalue plays a key role as it even vanishes at the sphaleron and ultimately induces the rapid increase in the inverse determinants ratio .
the explicitly derived equations governing the temperature dependence of the fluctuations may be adapted to specific situations by tuning the potential parameters . for the specific choice of parameters here presented ,
@xmath16 and @xmath27 , the computed @xmath1 starts to increase over the zero @xmath0 decay rate at about @xmath205 and , close to @xmath8 , the peak value is larger by a factor @xmath179 , a sizeable effect . at the crossover the quantum decay process ends and
the arrhenius classical behavior sets in .
finally , a comparison between cubic and quartic metastable potentials having the same structural parameters shows that the latter is _ less metastable _ in the sense that the lifetime is found to be larger by a factor ten at any temperature in the quantum regime .
j.s.langer , ann . phys . * 41 * , 108 ( 1967 ) .
m.b.voloshin , i.y.kobzarev , l.b.okun , sov . j. nucl
. phys . * 20 * , 644 ( 1975 ) .
t.banks , c.bender , t.t.wu , phys .
d * 8 * , 3346 ( 1973 ) .
s.coleman , phys .
d * 15 * , 2929 ( 1977 ) ; c.g.callan , s.coleman , phys .
d * 16 * , 1762 ( 1977 ) .
l.d.landau , e.m.lifshitz , _ quantum mechanics _
@xmath206 edn .
( butterworth - heinemann , oxford , 1977 ) .
r.p.feynman , rev .
* 20 * , 367 ( 1948 ) . l.s.schulman _ techniques and applications of path integration _ ( wiley&sons , new york , 1981 ) .
i.m.gelfand , a.m.yaglom , j.math.phys . * 1 * , 48 ( 1960 ) .
m.zoli , phys .
b , * 72 * , 214302 ( 2005 ) .
r.rajaraman , _ solitons and instantons _
( north holland , amsterdam , 1982 ) .
e.brzin , g.parisi , j.zinn-justin , phys .
d * 16 * , 408 ( 1977 ) .
i.affleck , phys . rev .
* 46 * , 388 ( 1981 ) .
u.weiss , w.haeffner , phys .
d * 27 * , 2916 ( 1983 ) .
m.b.voloshin , phys
. lett .
b * 599 * , 129 ( 2004 ) .
h.kleinert , _ path integrals in quantum mechanics , statistics , polymer physycs and financial markets _ ( world scientific publishing , singapore , 2004 ) .
b.braun , phys .
. lett . * 71 * , 3557 ( 1993 ) .
j.brki , c.a.stafford , d.l.stein , phys .
lett . * 95 * , 090601 ( 2005 ) .
w.g.faris , g.jona-lasinio , j. phys . a * 15 * , 3025 ( 1982 ) .
the concept of temperature needs a clarification . in theories of phase transitions with symmetry breaking
the effective potential is a function of the temperature . by cooling the system
, a minimum ( corresponding to a non - zero higgs field @xmath207 ) appears in the potential .
thus the metastable state placed at @xmath208 decays via quantum tunneling to the @xmath209 energetically favourable state . here
i m taking a specific potential which does not change its shape with @xmath0 : it has a metastable state at any @xmath0 within the considered quantum regime .
instead , it is the particle moving in such potential which feels the temperature effects .
the bounce plus its quantum fluctuations depend on @xmath0 which is here an inverse _
time_. z.x.wang , d.r.guo , _ special functions _
( world scientific , singapore , 1989 ) .
p.hnggi , p.talkner , m.borkovec , rev .
phys . * 62 * , 251 ( 1990 ) .
n.antunes , f.c.lombardo , d.monteoliva , p.i.villar , phys .
e * 73 * , 066105 ( 2006 ) .
p.s.riseborough , p.hnggi , e.freidkin , phys .
a * 32 * , 489 ( 1985 ) .
d.l.stein , brazilian j. phys . * 35 * , 242 ( 2005 ) .
n.s.manton , phys .
d * 28 * , 2019 ( 1983 ) .
i.s.gradshteyn , i.m.ryzhik , _ tables of integrals , series and products _ ( academic press , new york , 1965 ) .
v.i.goldanskii , sov.phys.dokl . * 4 * , 74 ( 1959 ) .
a.i.larkin , y.n.ovchinnikov , sov.phys.jetp * 37 * , 382 ( 1983 ) ; ibid . * 59 * , 420 ( 1984 ) .
h.grabert , u.weiss , phys .
rev . lett . * 53 * , 1787 ( 1984 ) .
d.a.gorokhov , g.blatter , phys .
b * 56 * , 3130 ( 1997 ) .
a.n.kuznetsov , p.g.tinyakov , phys .
b * 406 * , 76 ( 1997 ) .
q.liang , h.j.w.mller-kirsten , d.k.park , f.zimmerschied , phys .
lett . * 81 * , 216 ( 1998 ) .
e.m.chudnovski , phys
. rev .
a * 46 * , 8011 ( 1992 ) .
this can be directly deduced from eq .
( [ eq:64 ] ) as @xmath210 .
s.y.lee , h.kim , d.k.park , j.k.kim , phys .
b * 60 * , 10086 ( 1999 ) .
r.forman , invent.math . * 88 * , 447 ( 1987 ) ; ibid .
. math.phys . * 147 * , 485 ( 1992 ) .
k.kirsten , a.j.mckane , j.phys a : math . gen . * 37 * , 4649 ( 2004 ) .
r.s.ward , j.phys.a : math.gen . *
20 * , 2679 ( 1987 ) .
the way to order eigenvalues and eigenstates in terms of @xmath211 is purely conventional .
either @xmath212 $ ] or @xmath213 $ ] can be used .
the former is assumed here consistently with having labelled the unstable eigenvalue by @xmath214 .
a.j.mckane , m.b.tarlie , j.math.phys .
* 28 * , 6931 ( 1995 ) .
a.i.yanson , i.k.yanson , j.m .
van ruitenbeck , nature * 400 * , 144 ( 1999 ) .
h.zhang , f.kassubek , c.a.stafford , phys .
b * 68 * , 165414 ( 2003 ) .
|
the decay rate for a particle in a metastable cubic potential is investigated in the quantum regime by the euclidean path integral method in semiclassical approximation .
the imaginary time formalism allows one to monitor the system as a function of temperature .
the family of classical paths , saddle points for the action , is derived in terms of jacobian elliptic functions whose periodicity sets the energy - temperature correspondence .
the period of the classical oscillations varies monotonically with the energy up to the sphaleron , pointing to a smooth crossover from the quantum to the activated regime .
the softening of the quantum fluctuation spectrum is evaluated analytically by the theory of the functional determinants and computed at low @xmath0 up to the crossover . in particular , the negative eigenvalue , causing an imaginary contribution to the partition function , is studied in detail by solving the lam equation which governs the fluctuation spectrum . for a heavvy particle mass , the decay rate shows a remarkable temperature dependence mainly ascribable to a low lying soft mode and , approaching the crossover , it increases by a factor five over the predictions of the zero temperature theory . just beyond the peak value ,
the classical arrhenius behavior takes over .
a similar trend is found studying the quartic metastable potential but the lifetime of the latter is longer by a factor ten than in a cubic potential with same parameters .
some formal analogies with noise - induced transitions in classically activated metastable systems are discussed .
|
electron density ( n@xmath1 ) is a fundamental physical parameter of a plasma , and one of the main ways for determining this quantity in remote astrophysical sources is via emission line intensity ratios which are sensitive to variations in n@xmath1 . in the case of the high temperature solar transition region and corona ,
many of the emission lines lie at extreme - ultraviolet ( euv ) wavelengths ( @xmath01001200 ) , and for more than half a century numerous researchers have worked on the calculation of theoretical line ratio n@xmath1-diagnostics involving euv transitions ( see , for example , @xcite ; @xcite ; @xcite ) .
such diagnostics have generally been used in conjunction with medium to high spectral resolution solar observations obtained with rocket - borne or satellite - based instrumentation , required to adequately resolve and hence reliably measure the relevant emission lines .
examples include the s082a instrument on board the skylab space station , which had a resolution of @xmath00.1 @xcite , the solar euv rocket telescope and spectrograph ( serts ) at 0.050.08resolution @xcite , the coronal diagnostic spectrometer ( cds ) on the solar and heliospheric observatory ( soho ) satellite at @xmath00.4resolution @xcite and the euv imaging spectrometer ( eis ) on the hinode satellite at @xmath00.07resolution @xcite .
the solar dynamics observatory ( sdo ) , launched on 2010 february 11 , contains a number of instruments including the extreme ultraviolet variability experiment ( eve ) , which obtains a full - disk euv spectrum every 10s using its megs - a component , albeit at a relatively low resolution of @xmath01 @xcite .
although the eve temporal resolution is unprecedented for an euv spectrometer , allowing detailed studies of time - dependent phenomena such as flares , one would expect the spectral resolution to be too low to provide reliable electron density diagnostics .
however , @xcite found that a few fexxi emission line ratios in the 90160 portion of the megs - a flare spectra provide useful electron density diagnostics for values of n@xmath1 in the range @xmath010@xmath310@xmath4@xmath5 , hence allowing the study of the temporal evolution of relatively low flare density .
the short - wavelength spectrometer on the extreme ultraviolet explorer ( euve ) satellite obtained spectra of astrophysical sources in the 70190 region at a resolution of @xmath00.5 @xcite .
it observed many highly - active late - type stars ( see , for example , @xcite ) , and due to the comparable spectral coverage and resolution with eve flare data , the euve and eve observations can appear remarkably similar in terms of the presence of highly ionised fe lines , as illustrated in figure 1 .
hence the euve spectra may be employed as a ` testbed ' for fe ion line ratio diagnostics in eve observations , and in particular allow us to assess the potential usefulness of such diagnostics for high n@xmath1 flare plasmas , as active late - type stars can contain high density ( up to @xmath010@xmath6@xmath5 or greater ) coronal material . in the present paper
we significantly expand upon the work of @xcite to establish if other high temperature ( @xmath710mk ) fe lines apart from those of fexxi can be used to derived plasma densities under flare conditions .
specifically , we undertake a comparison of eve and euve observations , plus published results for well - diagnosed high density tokamak plasmas , to test the validity of additional highly ionized fe line ratios present in eve spectra between 90160 as reasonable density diagnostics for flaring plasmas .
more importantly , we also assess if they can be employed to derive densities in the high n@xmath1 ( @xmath210@xmath410@xmath6@xmath5 ) regime .
the solar observations considered in the present paper consist of eve spectra for two x - class solar flares obtained at event peak , namely that of 2011 february 15 at 01:55:32 ut ( x2.2 class ) and 2011 august 9 at 08:05:07 ut ( x6.9 class ) .
these flares were selected to maximise both the intensities of the highly ionized fe lines of interest ( and hence minimise the effect of blending from low temperature transitions ) , and the electron density of the emitting plasma ( ideally reaching densities close to 10@xmath6@xmath5 ) .
details of the eve instrument and data reduction procedures may be found in @xcite and @xcite .
briefly , the multiple euv grating spectrograph ( megs)-a of eve obtains full - disk spectra of the sun in the 65370wavelength range , at a resolution of @xmath01.0 , every 10s .
we have employed the most recent release of the eve data ( level 2 , version 5 ) , based on new rocket calibration flights , which improves the accuracy in the irradiance values presented in earlier releases ( @xcite ) .
our paper is focused on the @xmath090160 region of megs - a , which contained numerous emission lines arising from transitions in fexviii fexxiii .
@xcite have previously undertaken an analysis of fexviii
fexxiv features in eve spectra , to investigate which of these may be relatively free of blending in solar flares and hence have the potential to be employed as diagnostics .
however , these authors did not consider electron density diagnostics in detail , in particular the consistency of densities derived from different line ratios .
furthermore , they focused on plasmas with n@xmath1 in the range @xmath010@xmath310@xmath4@xmath5 , where many of the fe line diagnostics are not particularly useful .
for example , the fexxi ( 142.14 + 142.28)/128.75 ratio recommended by @xcite only varies by @xmath035% between 10@xmath3 and 10@xmath4@xmath5 .
the fexxi 145.73/128.75 ratio recommended by both @xcite and @xcite varies by a much greater amount over this density interval ( a factor of 4.8 ) , but with the problem that the 145.73 line is very weak at such values of n@xmath1 , being predicted to be typically only 5% of the intensity of 128.75 .
by contrast , our work is primarily concerned with the identification and assessment of diagnostics for much higher density flare plasmas , ideally up to n@xmath1 @xmath2 10@xmath6@xmath5 or even greater . in table 1
we summarise the lines judged by @xcite to be blend free and which may provide electron density sensitive intensity ratios .
however , in addition we include the fexxi 123.83 and fexxii 156.02 transitions .
the former was not considered by @xcite , while in the case of fexxii 156.02 these authors noted that the line appeared to be blended as it implied a very high value of electron density . to investigate the potential blending of these features ,
we have calculated synthetic spectra with the latest version ( 8.0.1 ) of the chianti collisional - radiative modelling packages , appropriate for plasmas with densities up to @xmath010@xmath8@xmath5 @xcite .
spectra have been generated for electron densities in the range 10@xmath310@xmath6@xmath5 , and for the standard chianti flare differential emission measure ( dem ) distribution .
although each solar flare will have its own characteristics ( including dem ) , we note that the highly ionized fe lines are formed over a narrow range of temperatures of maximum fractional abundance , log t@xmath9 = 7.07.1 @xcite .
the dems of different flares do not vary significantly over this temperature interval , as shown by for example figure 5 of @xcite .
our chianti synthetic spectra indicate that blending species ( primarily crxvii 122.97 , mnxxi 124.08 and crxx 156.02 ) should make at most @xmath025% contributions to the total measured line fluxes of both the fexxi 123.83 and fexxii 156.02 lines .
.highly ionized fe lines considered in the present analysis [ cols="<,^,^",options="header " , ] @xmath10ratios are in photon units .
+ @xmath11total exposure time of the summed euve short - wavelength spectrum .
+ @xmath12from @xcite .
+ @xmath13from @xcite .
+ @xmath14from @xcite .
+ @xmath15l indicates that the observed ratio is greater than the theoretical low density limit .
+ @xmath16stellar electron densities derived from a range of high temperature ions in the euve spectra .
the tokamak values are those measured by @xcite or @xcite .
the final sets of euv spectra considered here are from the frascati tokamak upgrade ( ftu ) and princeton large torus ( plt ) tokamaks , obtained by @xcite and @xcite , respectively .
all three datasets have a spectral resolution of @xmath00.7 , similar to those of the eve and euve observations .
the tokamak spectra will not show the same level of blending as the astrophysical data , due to the latter sampling a larger number of elements .
however , they should still provide a useful comparison with the solar / stellar observations , as most of the significant emission lines in the euv region of interest arise from other high ionization fe transitions .
this is shown by , for example , a comparison of the fe spectrum of the ftu tokamak between 85140 in figure 2 of @xcite with the eve and euve data in figure 1 , which appear very similar .
hence the tokamak spectra should allow the reliability and consistency of highly ionized fe line diagnostics to be assessed .
tokamaks are also particularly useful to the present study as their plasmas have very large values of n@xmath1 , similar to or greater than those of high electron density flares .
line fluxes for the fe ion lines in both the eve and euve spectra were determined by fitting multiple gaussian profiles to the observations , with some example fits shown in figures 2 and 3 .
the resultant line intensity ratios ( in photon units ) , along with the @xmath171@xmath18 errors , are listed in table 2 .
also included in the table are the observed line ratios for the ftu and plt tokamak spectra , taken directly from @xcite and @xcite .
we have derived values of electron density from the observed eve , euve and tokamak line ratios using theoretical results from the latest version ( 8.0.1 ) of chianti @xcite .
all of the line ratio calculations were performed at the value of t@xmath9 of the relevant fe ion @xcite , although we note that the results are not sensitive to the adopted temperature .
for example , changing the electron temperature for the fexxi ( 142.14 + 142.28)/128.75 line ratio calculations from t@xmath9 = 1.1@xmath1910@xmath20k to 2@xmath1910@xmath20k leads to a @xmath210.1 dex variation in the derived value of n@xmath1 .
similarly , for fexxii 114.41/135.79 , adopting t@xmath1 = 2.5@xmath1910@xmath20k rather than t@xmath9 = 1.3@xmath1910@xmath20k also results in a @xmath210.1 dex change in the density estimate .
hence even if the plasma is not in ionization equilibrium and the temperature is significantly different from t@xmath9 , as may be the case during a flare ( see , for example , @xcite ) , this should not affect the derived values of n@xmath1 .
furthermore , the timescale to achieve ionization equilibrium scales with electron density , and for the high n@xmath1 events considered here the ionization state of fe should be close to equilibrium @xcite .
also listed in table 2 are the electron densities , denoted n@xmath1(other ) , determined from a range of high temperature ions in the euve spectra , in particular those with lines known to be blended in the eve spectra @xcite but which should be relatively unblended in the ( somewhat ) greater resolution euve observations .
these include density diagnostics such as fexx 121.85/110.63 , fexxi 121.21/128.75 and fexxii 116.27/117.15 .
for the tokamak observations , n@xmath1(other ) is the experimental plasma density , determined from interferometry measurements to an accuracy of approximately @xmath1715% @xcite .
the first point to note from an inspection of table 2 is that the fexxi 123.83 line is clearly blended to a very significant degree in the eve observations , as the measured values of fexxi 123.83/(142.14 + 142.28 ) are both far greater ( by factors of 1.7 and 3.1 ) than the theoretical low density limit of 0.29 .
as noted in section 2 , this line was not considered by @xcite in their assessment of eve flare data , although it is not predicted to be severely blended . however
, blending species must make a much larger contribution to the total intensity than evisaged , including perhaps unidentified transitions .
for the other fe ion lines in the eve spectra , the resultant intensity ratios imply electron densities that are reasonably consistent , with discrepancies that average only 0.3 dex from the mean value of log n@xmath1 = 12.4 for both flares .
similarly , for the euve datasets , the line ratios lead to electron density estimates for each star that are consistent within the errors , and indicate average densities which show no significant discrepancies with those derived from a wider range of high temperature n@xmath1-diagnostic lines .
differences between the mean densities and the values of n@xmath1(other ) average only 0.1 dex , and do not exceed 0.3 dex .
we note that as the euve spectra analysed here are the sum of several sw datasets , the derived densities will not represent specific flare values , but rather time - averaged estimates over the duration of the observations . however , this is not an issue as we are only concerned with the consistency of densities derived from different line ratios , and hence a comparison of time - averaged values is appropriate .
in addition , it is worth pointing out that the ratio values for individual flares are actually very similar to those from the summed spectra .
for example , for au mic the measured fexxii 114.41/135.79 and 114.41/117.15 ratios in table 2 are 0.26 and 0.22 , respectively , while those for a specific flare ( interval b of @xcite ) are 0.30 and 0.23 .
the similar ratio values reflect the fact that the spectral emission is dominated , as might be expected , by high density flaring plasma .
however , we employ summed euve spectra in our work as these provide better signal - to - noise and hence more reliable detections , especially for weak features . in the case of the tokamak spectra ,
the derived average values of n@xmath1 are consistent with the experimental data in three out of four instances , with discrepancies of less than 0.2 dex .
the exception is the plt data of @xcite at an experimental density of log n@xmath1 = 12.7 , where the diagnostics indicate log n@xmath1 = [email protected] .
this is due to the fexxii 114.41/135.79 ratio being much larger than expected ; the measured value is 0.60 , while for an experimental density of log n@xmath1 = 12.7 the theoretical ratio is only 0.28 .
@xcite also noted this discrepancy , and it remains unexplained .
however , the problem must lie with the 135.79 line measurement in the spectrum , as the fexxii 114.41/117.15 ratio shows no significant differences between theory ( 0.27 at log n@xmath1 = 12.7 ) and observation ( 0.33 ) , and hence the 114.41 experimental intensity should be secure .
we note that there is no issue with the 135.79 line in the ftu spectrum of @xcite , with good agreement between the density derived from fexxii 156.02/135.79 ( log n@xmath1 = [email protected] ) and the measured value ( log n@xmath1 = 13.9 )
. furthermore , there is consistency between theory and observation for fexxii 114.41/135.79 in the plt spectrum for the log n@xmath1 = 13.5 plasma , with the measured value indicating log n@xmath1 = 13.7 , providing support for the reliability of the chianti line ratio calculations .
the consistency of the electron densities determined from the eve observations , combined with that also found from the euve and tokamak data of similar spectral resolution , indicates that the line ratios in table 2 with the exception of fexxi 123.83/(142.14 + 142.28 ) should provide reasonable estimates of n@xmath1 for the high temperature ( @xmath010mk ) plasma in solar flares .
more importantly , the euve and tokamak results provide support for the line ratio diagnostics being reliable up to very high values of n@xmath1 ( @xmath210@xmath6@xmath5 ) , and confirm the large densities derived for the solar events .
@xcite found that two of the fexxi ratios in table 2 145.73/128.75 and ( 142.14 + 142.28)/128.75 provided good estimates of density at n@xmath1 @xmath2 10@xmath4@xmath5 .
( we note in passing that these authors also considered fexxi 121.21/128.75 , which we exclude as @xcite point out that the 121.21 feature is very weak and in the wing of the strong fexx 121.84 line ) .
however , the present work significantly extends the @xcite analysis to show that four additional ratios of fexx and fexxii may be confidently used as diagnostics , even at the low spectral resolution of eve , and furthermore allow a high electron density regime to be reliably investigated . as eve obtains full - disk spectra at a very high cadence of 10s ( as opposed to its low spectral resolution ) , this in turn indicates that the observations may be employed for reliable temporally - resolved studies of flare densities in the high n@xmath1 regime ( @xmath010@xmath410@xmath6@xmath5 ) , and in particular the investigation of the maximum density that can be achieved during flare peak .
the eve instrument has observed hundreds of flares during its 4 years of operation , providing a very large dataset for such studies .
measurement of as many as possible of the line ratios in table 2 would be preferred , as the average densities are in better agreement with n@xmath1(other ) than the individual values , which show a large amount of scatter . however , the best diagnostics based on their density sensitivity particularly at high n@xmath1 , and how well the derived densities agree with n@xmath1(other ) are probably fexx 113.35/121.85 , fexxi ( 142.14 + 142.28)/128.75 and fexxii 114.41/135.79 .
now that we have confirmed the usefulness of megs - a data for determining high flare densities , the next step is to extend our analyses to the megs - b spectral region ( 3501050 ) , which samples lower temperature emission lines .
the overall aim of this work will be to identify diagnostics which allow time profiles of flare densities simultaneously across a broad temperature range ( @xmath00.110mk ) .
fpk and mm are grateful to the science and technology facilities council for financial support .
the research leading to these results has received funding from the european community s seventh framework programme ( fp7/2007 - 2013 ) under grant agreement no .
606862 ( f - chroma ) .
we are also grateful to the leverhulme trust for financial support via grant f/00203/x .
rom acknowledges support from nasa lws / sdo data analysis grant nnx14ae07 g .
chianti is a collaborative project involving george mason university , the university of michigan ( usa ) and the university of cambridge ( uk ) .
the euve data presented in this paper were obtained from the mikulski archive for space telescopes ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
support for mast for non - hst data is provided by the nasa office of space science via grant nnx09af08 g and by other grants and contracts .
|
the extreme ultraviolet variability experiment ( eve ) on the solar dynamics observatory obtains extreme - ultraviolet ( euv ) spectra of the full - disk sun at a spectral resolution of @xmath01 and cadence of 10s .
such a spectral resolution would normally be considered to be too low for the reliable determination of electron density ( n@xmath1 ) sensitive emission line intensity ratios , due to blending .
however , previous work has shown that a limited number of fexxi features in the 90160 wavelength region of eve do provide useful n@xmath1-diagnostics at relatively low flare densities ( n@xmath1 @xmath2 10@xmath310@xmath4@xmath5 ) . here
we investigate if additional highly ionised fe line ratios in the eve 90160 range may be reliably employed as n@xmath1-diagnostics .
in particular , the potential for such diagnostics to provide density estimates for high n@xmath1 ( @xmath010@xmath6@xmath5 ) flare plasmas is assessed .
our study employs eve spectra for x - class flares , combined with observations of highly active late - type stars from the extreme ultraviolet explorer ( euve ) satellite plus experimental data for well - diagnosed tokamak plasmas , both of which are similar in wavelength coverage and spectral resolution to those from eve .
several ratios are identified in eve data which yield consistent values of electron density , including fexx 113.35/121.85 and fexxii 114.41/135.79 , with confidence in their reliability as n@xmath1-diagnostics provided by the euve and tokamak results .
these ratios also allow the determination of density in solar flare plasmas up to values of @xmath010@xmath6@xmath5 .
[ firstpage ] stars : coronae stars : flares sun : corona sun : flares
|
dark matter has become an extremely interesting area of research in both cosmology and particle physics . from the particle physics point of view
it can be thermal wimps , axions , kaluza klein states , etc .
though supersymmetric ( susy ) models predict its mass to be of the order the electro - weak scale , there are viable models of dark matter where its mass can be as low as sub - ev , for example axion - like dark matter .
especially the direct and indirect search for dark matter has narrowed down the parameter space of these well studied candidates to a large extent .
hence it is highly likely that dark matter is of much more exotic nature than thought of .
in addition , there may be a requirement for more complicated physics such as interactions with dark energy or with neutrinos .
similarity between the neutrino mass and the present dark energy density scale has inspired people to look for a connection between the two .
now we know that the normal active neutrino can not be a viable dark matter candidate because of its free - streaming ability .
but the existence of sub - ev neutrino mass might point towards richer sub - ev scale physics .
in fact there has been a few interesting works @xcite where new states of sub - ev masses are present in the dark sector .
the reason is , if the dark sector interacts only gravitationally with the standard model sector , a tev scale susy breaking in sm would predict scalar particles of mass @xmath0ev in the dark sector .
scalar dark matter of milli ev mass with a possible coupling to neutrinos has been discussed in @xcite .
also moduli of sub - ev mass can easily arise from string compactification @xcite .
if in nature dark matter arises from such a low energy scale , we would expect it to decay into light fermions like neutrinos through a yukawa type of coupling .
gev scale dark matter decay ( and annihilation ) to neutrinos has drawn lots of recent interest @xcite , especially in the context of recent cosmic ray measurements and the dama / libra experiment @xcite .
but in our case , dark matter is of sub - ev mass and the signature of its decay into neutrinos is mainly cosmological , especially in structure formation .
recently , in a different context , _ the non - thermal wimp miracle _ @xcite was introduced where a scalar of tev mass decays into a stable dark matter particle .
so the decay of a scalar particle can lead to very rich phenomenology in cosmological context .
the possible decay of scalar dark matter into light neutrinos is an effect which could potentially help to understand the apparent surplus of power on small scales in simulations containing normal cdm @xcite .
this surplus could for instance be an indication that cdm is simply clustering too much on small scales and we need some mechanism to reduce their gravitational interaction .
this is where the decay could play a role - see @xcite for a similar idea . in this paper
we study the nature of the decay and its possible signature in structure formation .
as we consider an axion - like scalar dark matter , it is undergoing a coherent oscillation and might decay into neutrinos through a parametric excitation .
in that case , the process will have many similarities with inflationary fermionic preheating @xcite .
therefore we dub the process _ preheating dark matter_. cosmology with decaying / interacting dark matter has been an interesting topic of research @xcite in recent times as it gives a probe to detect dark matter indirectly through its effects on structure formation . on top of that , if it decays into dark energy it might give rise to a unified description of dark matter and dark energy . in
most of the studies @xcite , the rate of energy transfer from dark matter to other components ( eg .
dark energy , radiation or neutrinos ) has been empirically assumed driven by mathematical simplicity . here
we present a concrete model of dark matter decay to light fermionic states like neutrinos and then study its imprint on structure formation .
in particular , we derive the decay rate of dark matter as a function of redshift using the theory of fermionic preheating . we find that in the initial stage the decay rate is faster and determined by parametric resonance .
but at late times the parametric production ceases and redshifts of fermionic modes control the decay rate .
this time varying decay rate makes the phenomenology rich and offers a prominent imprint on the structure formation , something which might be experimentally probed by near future experiments .
the plan of the paper is as follows : in section ii we discuss the particle physics aspect of our scenario . in section iii
, we incorporate the idea of inflationary preheating for scalar dark matter decaying into neutrinos . in section iv
, we identify different epochs of decay and derive the background evolution , i.e. how dark matter and neutrino energy densities evolve in presence of the decay . in section
v , we derive the perturbation equations for our scenario and in section vi we obtain temperature anisotropy and matter power spectra . finally , we summarize our work in section vii .
here we discuss how our scenario fits into a particle physics set up .
scalar fields of sub - ev masses are common in different particle physics models . in many models of
tev scale gauge mediated supersymmetry breaking , a gravitationally coupled dark sector contains a scalar of sub - ev mass .
also , string compactifications generically predict axion like scalars such as the dilaton and large numbers of moduli @xcite whose mass can easily be in the sub - ev range .
as we are interested in a sub - ev scalar field which couples to light fermions like the standard model neutrino , our set up is inspired by the models of mass varying neutrino dark energy @xcite where a sub - ev mass scalar couples to the standard model neutrino - see also @xcite . especially , we refer to a model of supersymmetric neutrino dark energy , where multiple scalars of such low mass are present and can couple to neutrinos .
in such a theory , it has been shown that the scalar potential takes the form of the well known hybrid inflation potential .
we refer readers to @xcite for details and will briefly discuss here .
the lagrangian for mass varying neutrino dark energy is given by @xmath1 where @xmath2 is the dirac mass and @xmath3 is the left chiral weyl field representing active neutrinos .
n is the righthanded heavy fermion and a is the scalar .
@xmath4 is some yukawa coupling . in supersymmetric models of neutrino dark energy where
@xmath5 is promoted to superfields @xmath6 , the superpotential takes the form @xmath7 after taking quantum corrections into account , it has been shown that this leads to a hybrid inflation kind of potential . depending on the temperature of the universe ,
the scalar either remains trapped at a metastable minimum playing the role of dark energy or it rolls off and starts oscillating coherently , behaving like cold dark matter .
following a simple model @xcite , the lagrangian looks like @xmath8 where @xmath9 and @xmath10 the terms in @xmath11 are included in order to generate a majorana mass for the neutrino in the vacuum . in this kind of theory
the superpartner sneutrinos ( here denoted by @xmath12 ) can easily be of sub - ev mass and play the role of dark matter .
also it can easily couple to light fermions ( neutrinos ) .
from now on , we will switch our focus to cosmological effects of such a model .
we are essentially interested in a light scalar field dark matter of sub ev mass which has a coupling to an ultra light fermion which for our case is the standard model neutrino .
we consider decays of such dark matter into neutrinos , though our framework is true for decay into any fermion . following the mechanism of inflationary preheating , in this section we will understand the physical nature of the decay and will clarify different regimes of decay .
we will see that the decay rate changes as the universe expands due to time evolution of a resonance parameter which controls the parametric excitation of neutrinos .
we mainly follow the work @xcite on fermionic preheating and apply that to our scenario .
so we refer to this work for detailed derivation of the equations .
briefly , to find the number density of created fermions through the preheating mechanism in an expanding background , one derives a mode equation using the original dirac equation with the friedmann - lematre - robertson - walker metric .
it has been shown that the comoving number density of created fermions can be obtained by solving for a mode function @xmath13 . for a yukawa type coupling @xmath14 ,
the mode equation is given by @xmath15 x_k = 0,\ ] ] where @xmath16 is the background solution for the time evolution of the oscillating scalar field , @xmath17 is the dimensionless fermion mass , @xmath18 , and the resonance parameter @xmath19 .
these three parameters completely determine the parametric production of fermions .
we consider the oscillation of the field with the usual quadratic potential @xmath20 as this is a good approximation around minima of any potential .
the term @xmath21 can be thought of as an effective mass of the fermion . as the scalar field oscillates
, the effective mass itself will oscillate around zero and the parametric production of fermions is enhanced when the effective mass crosses zero .
it has been shown numerically that @xmath22 oscillates and due to pauli blocking its maximum value never crosses unity .
but for decay into bosonic particles it is not bounded by unity .
we stress that the behavior of this parametric production is considerably different than the perturbative decay process @xmath23 where the decay rate is given by @xmath24 . in the above mode equation , expansion of the universe
has been neglected which may only be true at very late times where the hubble parameter drops . to get a full understanding
, one must include the expansion of the universe .
this alters two aspects .
the parameters @xmath25 and @xmath4 now become time dependent .
more specifically we get @xmath26 and the physical momentum @xmath27 where @xmath28 is the scale factor of the universe . as a result , the periodic modulation of the comoving number density does not hold anymore . for large values of the resonance parameter @xmath29 ,
the calculation of parametric production becomes , in fact , simple .
luckily , we will see later that for our case @xmath30 for large periods of ( cosmological ) time . using the method of successive scattering for fermions
, it has been shown @xcite that due to the loss of periodicity of @xmath31 the production of fermions happens through a stochastic filling of a fermi sphere up to a fermi radius @xmath32 which depends on scale factor @xmath28 and is given by @xmath33 now to find the exact dependence , one needs to know how @xmath34 changes with scale factor . as we are using a quadratic potential for the scalar ,
the solution for the scalar field for this case is well known .
oscillation of @xmath35 in this case is given by the asymptotic solution @xmath36 using this it is easy to derive @xmath37 where @xmath38 .
so , as the universe expands , the fermi sphere also expands producing more and more neutrinos .
but the resonance parameter decreases as the amplitude of oscillation drops due to hubble friction and at some point the fermi sphere stops expanding when @xmath34 becomes of the order of unity . at this regime ,
the redshift of fermionic modes due to hubble expansion is fast enough to prevent the parametric excitation .
finally , fermions will be produced with a much lower rate in the perturbative regime and perturbative processes continue unless @xmath39 .
using the above results now we can focus on production of neutrinos and its time evolution . here
we assume that parametrically produced neutrinos mix with other relic neutrinos and acquire the same temperature through thermalization .
it is instructive to note that as the mass of the scalar is way less than in the usual inflationary preheating scenario , we would get parametric excitation until very late times . from the previous discussion
, we have learnt that the resonance parameter @xmath25 is very crucial to determine the nature of the decay and @xmath25 itself is time dependent .
now we will discuss the two different decay regimes and the transition time between them for our simple model with a quadratic potential . during early stages of parametric production @xmath41 as we are interested in scalar mass of the order of @xmath42 ev , almost all over the cosmic history until today , @xmath43 .
now if the coupling constant @xmath44 is of the order of unity , we still get parametric production at very late times .
but for smaller couplings parametric excitation stops at earlier redshift when @xmath45 .
later we will take different choices of the coupling @xmath44 and study how it affects the formation of structure .
the produced neutrino number density is obtained through the volume of the fermi sphere with radius @xmath46 where @xmath47 .
this can be used to calculate the neutrino density @xmath48 where we have used @xmath49 .
it is important to note that the neutrino energy density is proportional to the local dark matter density .
using this fact and the continuity equation for the total dark matter and neutrino fluid , we can find the evolution of the dark matter energy density and hence neutrino energy density .
the continuity equation for the dark matter and neutrino as a whole reads @xmath50 where @xmath51 , @xmath52 and @xmath53 .
[ eq : continuity ] can be split up into the two components @xmath54 and @xmath55 where q represents the decay rate from dark matter to neutrinos and we have taken advantage of the fact that @xmath56 and @xmath57 . combining eqs .
[ nupro ] , [ eq : contdm ] , and [ eq : contnu ] we get the relations @xmath58 where the @xmath59 denotes the value at some fixed time(e.g . today ) and @xmath60 .
so here we clearly see that due to parametric production , dark matter no longer redshifts as @xmath61 .
its effective equation of state changes from zero to slightly higher values .
the higher the coupling @xmath44 , the higher the deviation .
this effective equation of state which corresponds to the value we would get if we did not know about the coupling between dark matter and neutrinos can be calculated from a revised version of eq .
[ eq : contdm ] @xmath62 .
the result is @xmath63 we note that for @xmath64 , it gives the right limit for the equation of state @xmath65 . in the second regime ,
parametric excitations weaken due to the drop in resonance parameter @xmath25 . during this regime ,
@xmath67 and the radius of the physical fermi sphere has approached the constant value @xmath68 .
this means the decay is controlled by the redshifts of fermi momentum due to the expansion .
as the universe expands , the fermi momentum drops , opening up space in the fermi sphere .
this space is immediately filled up by the scalar field decaying into neutrinos .
the regime may be important for structure formation if the decay into neutrinos can cause a substantial decrease in the dark matter density .
this is possible when @xmath69 , because , in this case , decay of each dm particle to a neutrino causes a significant decrease in the dark matter energy density . as the dark matter mass is of the order sub ev in our model , this can happen only at late times .
this can enhance the late isw effect thus modifying the structure formation on large scales .
again , using eq .
[ nupro ] we easily obtain @xmath70 . from the continuity equations
[ eq : contdm ] and eq .
[ eq : contnu ] we get the relations @xmath71 where @xmath72 .
so we see that the neutrino energy density is constant in this regime , only the dark matter density dilutes .
this is what we expect , because if there were no decay into neutrinos , its density would simply redshift like the standard model neutrino governed by the hubble expansion . but here , as soon as phase space opens in the fermi sea of neutrinos due to cooling of the universe , it gets refilled by the decay from dark matter thus keeping its density constant . in this regime
the effective equation of state for dark matter can be calculated to be @xmath73 which depending on model parameters can deviate an appreciable amount toward the present - see fig . [
fig : wdm ] . , @xmath74 with @xmath75 ,
dot - dashed blue : @xmath76 , @xmath77 with @xmath78 , and dashed red : @xmath79 , @xmath80 with @xmath81.,width=240 ] in this subsection we find the transition redshift between the two epochs in terms of model parameters @xmath82 and @xmath44 .
patching the two regimes together at a scale factor @xmath83 we arrive at the relations @xmath84 where the @xmath85 denotes present day values , and @xmath83 can be determined from the preheating dark matter parameters as @xmath86^{1/3}a_0 .
\label{eq : atrans}\ ] ]
in order to study the implications of preheating dark matter , we perform a cosmological perturbation analysis in the synchronous gauge in which a line element is given by @xmath87,\ ] ] where @xmath88 , @xmath89 is the minkowski space metric , @xmath90 is the perturbation to the metric and we are using comoving coordinates @xmath91 in a spatially flat background space - time .
we follow the procedure given by ref .
@xcite , in which , to linear order in the perturbations , the stress - energy tensor is given by @xmath92 where the perturbations to energy density and pressure are defined as @xmath93 and @xmath94 , @xmath95 is the anisotropic shear perturbation , and @xmath96 is the coordinate velocity of the fluid .
the latter is a small quantity and can be treated as a perturbation of the same order as @xmath97 and @xmath98 . instead of working with the velocity itself we use the divergence defined as @xmath99 .
similarly , instead of the anisotropic shear perturbation , we use the shear stress @xmath100 .
this is defined as @xmath101 .
the conservation of energy and momentum for our coupled fluid implies that the covariant derivative of the stress - energy tensor is 0 .
@xmath102 however , for the individual components in the fluid it is slightly different @xmath103 where the @xmath104 can be determined directly from eq .
[ eq : contdm ] . using the time - time(00 ) component of the stress - energy tensor from eq .
[ eq : stress ] and combining with eq .
[ eq : dq ] above , we get the equation of motion for the individual density contrasts @xmath105 .
using the space - space(ii ) components in the same way will give us the time evolution of @xmath106 .
hence we arrive at the equations of motion for the dm and neutrino components for @xmath107 @xmath108 similarly , for @xmath67 we get @xmath109 it is the effect of the term containing @xmath44 and the term containing the neutrino energy density @xmath110 in the evolution of the cdm density contrast that separates the evolution of perturbations in the preheating dark matter case from the normal case , where the decay of cdm is not permitted .
in this section we present the numerical results we obtained by using the equations from the two previous sections .
we modified the publicly available cmbfast code @xcite to include preheating dark matter to get both the matter power spectrum and the temperature anisotropy spectrum .
this code is developed to calculate the linear cmb anisotropy spectra based on integration over the sources along the photon past light cone , but also outputs transfer functions from which the linear matter power spectrum can be calculated . in our analysis
we keep the epoch of matter - radiation equality fixed , and the only free parameters are the current value of @xmath111 and the parameter @xmath44 . in addition , we keep the amount of baryons today fixed at @xmath112 and choose the normalized hubble expansion rate at the value @xmath113 .
we include one species of massless neutrinos produced in the decay as well as the three standard model neutrinos , which for simplicity are assumed to have a degenerate mass spectrum with @xmath114 ev .
we assume that the neutrinos produced in the decay mix with standard neutrinos , although this assumption makes no qualitative difference to the results .
the impact of preheating dark matter on the temperature anisotropy spectrum , can be seen in fig .
[ fig : ani ] . the most apparent difference from the spectrum of @xmath115cdm can be seen on largest angular scales , @xmath116 ( corresponding roughly to a degree ) , although for some choices of parameters the positions and relative heights of the peaks are also affected .
we generally observe an increase in power on scales , @xmath117 , whereas on scales @xmath118 we see an increase or a decrease in power depending on the model parameters . for scales
@xmath116 , the dominant contribution to the cmb temperature anisotropy spectrum comes from the integrated sachs - wolfe effect , which arises because of the evolution of the gravitational potentials encountered by the photon on its journey from the last - scattering surface .
the modification to the cosmological background because of preheating dark matter can be quite significant , as we saw in the last section .
in particular in the @xmath67 regime , where the neutrino energy density becomes constant , which leads to a second term in the evolution of the cdm perturbation in eq .
[ eq : deltatheta2 ] . on the largest angular scales
@xmath119 , the dominant contribution to the cmb temperature anisotropy spectrum comes from the late - time integrated sachs - wolfe effect ( isw ) .
this effect is a result of the universe entering an epoch of rapid expansion as it becomes dominated by dark energy . in this epoch ,
photons moving into gravitational potential wells will get a boost as the potential well is decaying and becomes slightly shallower while the photon is passing through it - and vice versa for gravitational hills .
it is clear that this effect depends intimately on many different parameters such as @xmath111 and @xmath120 . in our model ,
dark matter is being transformed into light neutrinos - i.e. radiation - most efficiently at late times , and hence we expect an effect on the largest scales . in the context of interacting dark matter - dark energy models ,
the isw effect has been studied in @xcite .
cdm with @xmath121 , @xmath122 , dotted green : @xmath123 , @xmath74 with @xmath75 , dot - dashed blue : @xmath76 , @xmath77 with @xmath78 , and dashed red : @xmath79 , @xmath80 with @xmath81.,width=240 ] turning our attention to the matter power spectra , we know that the linear matter power spectrum is extremely well determined by sdss ( see ref .
@xcite ) and wmap ( see ref .
@xcite ) on intermediate and large scales . and in addition , ly-@xmath124 forest data have some constraints on the small scales - see e.g. @xcite .
hence we normalize our matter power spectra such that they coincide with matter power spectra obtained from using normal @xmath115cdm at the largest scales ( these are also the latest to have entered the horizon ) .
the results are presented in fig .
[ fig : delta ] for different values of @xmath44 and @xmath111 .
a small damping on small angular scales seems to be generic , similar to standard models of cdm and hot dark matter , where a similar reduction in power is achieved .
we note that in order to comply with e.g. supernova data ( ref .
@xcite ) we can not change @xmath125 drastically . of course we do turn cdm into neutrinos that redshift as radiation - only slightly faster than cdm .
hence we have some room to change @xmath125 and still be in agreement with data .
the matter power spectra for the different parameter choices are agreeing relatively well with @xmath115cdm as result of the small @xmath44-value . still , we do notice the small reduction on the smallest scales which can be probed by cmbfast .
this is to be expected since part of the cdm responsible for the gravitational wells is decaying into neutrinos which undergo free - streaming on these small scales .
this prevents them from clustering and since we are creating an appreciable amount of neutrinos due to the decay of cdm , we generally expect this reduction of power on small scales . ] in the preheating dark matter scenario with the following parameter choices : solid black : @xmath115cdm with @xmath121 , @xmath122 , dotted green : @xmath123 , @xmath74 with @xmath75 , dot - dashed blue : @xmath76 , @xmath77 with @xmath78 , and dashed red : @xmath79 , @xmath80 with @xmath81 .
bottom : the difference between the preheating dark matter model and @xmath115cdm ( @xmath126(@xmath115cdm)/@xmath126(phdm ) ) as a function of k.,width=240 ] on even smaller scales we also expect a considerable effect , which will reduce the clustering ability of cdm . unfortunately , we can not probe those scales satisfactorily with cmbfast - as it is using linear perturbation analysis and we expect the evolution to be highly non - linear .
the investigation of the effect of preheating dark matter on non - linear scales is beyond the scope of the present work and will be postponed to a future paper .
in this work , we have considered a coherently oscillating scalar field of sub - ev mass which behaves as dark matter . due to its coupling
, it slowly decays into neutrinos as the universe expands until today .
the decay rate as a function of redshift has been derived following the physics of inflationary preheating of a scalar into fermions .
we find that the decay rate is modulated by pauli blocking and the expansion of the universe , giving us rich physics of dark matter decay into a neutrino sea .
we studied the effect of the decay on structure formation and obtained spectra for the anisotropies in the cosmic microwave background and matter power spectra . for the parameters proposed in this paper
, we showed that given the decay we are able to slightly reduce the amount of power on small scales in the matter power spectra - something which seems to be required from data - while at the same time being in good agreement with sdss and wmap observations .
interestingly , the proposed scenario leads to features in the temperature anisotropy spectrum - which can be seen as a prominent late isw effect and slight modifications to the second and third peaks .
consequently , as future direction , it would be interesting to study the late isw effect in detail , from which we will be able to constrain the model parameters more effectively .
in addition , we would like to do a follow - up cosmomc analysis using the newest data available as well as to examine the effect of preheating dark matter on structure formation on non - linear scales .
after such an analysis it will be more clear what the best choice of preheating dark matter parameters is such that a comparison with future planck data , for instance , will be easier .
furthermore we expect future weak lensing surveys to be useful in constraining our scenario since they can provide insight into the statistics of the dark matter distribution - hence they can ( hopefully ) shed some light on what happens on small ( and large ) scales of gravitational clustering .
we would really like to thank neal weiner for ideas and discussions and motivating this work .
oeb thanks steen hannestad and yvonne wong for helpful discussions during this work .
sd thanks kris sigurdson for some fruitful suggestions .
the research of sd is supported by the natural sciences and engineering research council of canada .
r. fardon , a. e. nelson and n. weiner , jhep * 0603 * , 042 ( 2006 ) [ arxiv : hep - ph/0507235 ] . z. chacko , l. j. hall and y. nomura , jcap * 0410 * , 011 ( 2004 ) [ arxiv : astro - ph/0405596 ] .
s. das and n. weiner , arxiv : astro - ph/0611353 .
n. kaloper and l. sorbo , phys .
d * 79 * , 043528 ( 2009 ) [ arxiv:0810.5346 [ hep - th ] ] .
j. mardon , y. nomura , d. stolarski and j. thaler , jcap * 0905 * , 016 ( 2009 ) [ arxiv:0901.2926 [ hep - ph ] ] .
m. r. buckley , k. freese , d. hooper , d. spolyar and h. murayama , arxiv:0907.2385 [ astro-ph.he ] .
d. spolyar , m. r. buckley , k. freese , d. hooper and h. murayama , arxiv:0905.4764 [ astro-ph.co ] .
r. allahverdi , s. bornhauser , b. dutta and k. richardson - mcdaniel , phys .
d * 80 * , 055026 ( 2009 ) [ arxiv:0907.1486 [ hep - ph ] ] .
l. covi , m. grefe , a. ibarra and d. tran , jcap * 1004 * ( 2010 ) 017 [ arxiv:0912.3521 [ hep - ph ] ] .
j. hisano , m. kawasaki , k. kohri and k. nakayama , phys .
d * 79 * ( 2009 ) 043516 [ arxiv:0812.0219 [ hep - ph ] ] .
j. liu , q. yuan , x. bi , h. li and x. zhang , arxiv:0911.1002 [ astro-ph.co ] .
d. hooper , f. petriello , k. m. zurek and m. kamionkowski , phys .
d * 79 * , 015010 ( 2009 ) [ arxiv:0808.2464 [ hep - ph ] ] .
j. kumar , j. g. learned and s. smith , phys .
d * 80 * , 113002 ( 2009 ) [ arxiv:0908.1768 [ hep - ph ] ]
. b. s. acharya , g. kane , s. watson and p. kumar , phys .
d * 80 * , 083529 ( 2009 ) [ arxiv:0908.2430 [ astro-ph.co ] ] .
j. f. navarro , c. s. frenk and s. d. m. white , astrophys
. j. * 490 * ( 1997 ) 493 [ arxiv : astro - ph/9611107 ] .
a. a. klypin , a. v. kravtsov , o. valenzuela and f. prada , astrophys .
j. * 522 * ( 1999 ) 82 [ arxiv : astro - ph/9901240 ] .
b. moore , s. ghigna , f. governato , g. lake , t. r. quinn , j. stadel and p. tozzi , astrophys .
j. * 524 * ( 1999 ) l19 .
v. sahni and l. m. wang , phys .
d * 62 * ( 2000 ) 103517 [ arxiv : astro - ph/9910097 ] .
p. b. greene and l. kofman , phys .
b * 448 * ( 1999 ) 6 [ arxiv : hep - ph/9807339 ] .
g. f. giudice , m. peloso , a. riotto and i. tkachev , jhep * 9908 * ( 1999 ) 014 [ arxiv : hep - ph/9905242 ] . j. garcia - bellido , s. mollerach and e. roulet , jhep * 0002 * ( 2000 ) 034 [ arxiv : hep - ph/0002076 ] .
p. b. greene and l. kofman , phys . rev .
d * 62 * ( 2000 ) 123516 [ arxiv : hep - ph/0003018 ] .
s. das , p. s. corasaniti and j. khoury , phys .
d * 73 * , 083509 ( 2006 ) [ arxiv : astro - ph/0510628 ] . c. g. boehmer , g. caldera - cabral , r. lazkoz and r. maartens , phys .
d * 78 * , 023505 ( 2008 ) [ arxiv:0801.1565 [ gr - qc ] ] .
z. k. guo , n. ohta and s. tsujikawa , phys .
d * 76 * , 023508 ( 2007 ) [ arxiv : astro - ph/0702015 ] .
g. olivares , f. atrio - barandela and d. pavon , phys .
d * 77 * , 063513 ( 2008 ) [ arxiv:0706.3860 [ astro - ph ] ] .
a. ibarra and d. tran , jcap * 0902 * ( 2009 ) 021 [ arxiv:0811.1555 [ hep - ph ] ] .
m. cirelli , p. panci and p. d. serpico , arxiv:0912.0663 [ astro-ph.co ] . a. ibarra , d. tran and c. weniger , jcap * 1001 * ( 2010 ) 009 [ arxiv:0906.1571 [ hep - ph ] ] .
j. valiviita , e. majerotto and r. maartens , jcap * 0807 * , 020 ( 2008 ) [ arxiv:0804.0232 [ astro - ph ] ] .
r. cen , arxiv : astro - ph/0005206 .
j. jaeckel and a. ringwald , arxiv:1002.0329 [ hep - ph ] .
f. piazza and m. pospelov , arxiv:1003.2313 [ hep - ph ] .
r. fardon , a. e. nelson and n. weiner , jcap * 0410 * , 005 ( 2004 ) [ arxiv : astro - ph/0309800 ] .
r. takahashi and m. tanimoto , phys .
b * 633 * ( 2006 ) 675 [ arxiv : hep - ph/0507142 ] .
o. e. bjaelde , a. w. brookfield , c. van de bruck , s. hannestad , d. f. mota , l. schrempp and d. tocchini - valentini , jcap * 0801 * ( 2008 ) 026 [ arxiv:0705.2018 [ astro - ph ] ] .
r. bean , e. e. flanagan and m. trodden , phys . rev .
d * 78 * ( 2008 ) 023009 [ arxiv:0709.1128 [ astro - ph ] ] .
s. antusch , s. das and k. dutta , jcap * 0810 * ( 2008 ) 016 [ arxiv:0807.4930 [ astro - ph ] ] .
o. e. bjaelde and s. hannestad , phys .
d * 81 * ( 2010 ) 063001 [ arxiv:0806.2146 [ astro - ph ] ] .
c. p. ma and e. bertschinger , astrophys . j. * 455 * ( 1995 ) 7 [ arxiv : astro - ph/9506072 ] .
u. seljak and m. zaldarriaga , astrophys .
j. * 469 * ( 1996 ) 437 .
m. kowalski _ et al . _ ,
astrophys . j. * 686 * ( 2008 ) 749 [ arxiv:0804.4142 [ astro - ph ] ] .
m. tegmark _ et al . _
[ sdss collaboration ] , phys .
d * 69 * ( 2004 ) 103501 [ arxiv : astro - ph/0310723 ] .
g. olivares , f. atrio - barandela and d. pavon , phys .
d * 77 * , 103520 ( 2008 ) [ arxiv:0801.4517 [ astro - ph ] ] .
e. komatsu _ et al .
_ [ wmap collaboration ] , arxiv:0803.0547 [ astro - ph ] .
s. zaroubi , m. viel , a. nusser , m. haehnelt and t. s. kim , mon . not .
* 369 * ( 2006 ) 734 [ arxiv : astro - ph/0509563 ] . h. kodama and m. sasaki , prog .
* 78 * ( 1984 ) 1 .
|
we study the possible decay of a coherently oscillating scalar field , interpreted as dark matter , into light fermions . specifically , we consider a scalar field with sub - ev mass decaying into a fermi sea of neutrinos .
we recognize the similarity between our scenario and inflationary preheating where a coherently oscillating scalar field decays into standard model particles . like the case of fermionic preheating
, we find that pauli blocking controls the dark matter decay into the neutrino sea .
the radius of the fermi sphere depends on the expansion of the universe leading to a time varying equation of state of dark matter .
this makes the scenario very rich and we show that the decay rate might be different at different cosmological epochs .
we categorize this in two interesting regimes and then study the cosmological perturbations to find the impact on structure formation .
we find that the decay may help alleviating some of the standard problems related to cold dark matter .
|
if a disc around a central object is sufficiently massive , its own self - gravity may play an important role in its evolution through the growth of the gravitational instability @xcite .
an infinitesimally thin disc is susceptible to the growth of an axisymmetric gravitational instability if the @xmath2 parameter @xcite @xmath3 where @xmath4 is the sound speed in the disc , @xmath5 is the local epicyclic frequency ( equal to the angular velocity , @xmath6 , in a keplerian disc ) , @xmath7 is the gravitational constant , and @xmath8 is the disc surface density . global , non - axisymmetric perturbations can , however , grow for @xmath2 values greater than @xmath9 , with simulations suggesting that the stability criteria in global discs is @xmath10 @xcite .
it is now , however , quite well understood that the @xmath2 parameter alone does not determine the ultimate evolution of a self - gravitating accretion disc .
the evolution is determined by both the value of @xmath2 and by the rate at which the disc is able to lose energy @xcite .
the current picture is that for long cooling times , the disc will settle into a state of marginal stability @xcite in which the instability acts to transport angular momentum outwards , allowing mass to accrete onto the central object @xcite . for short cooling times , however , the disc may become sufficiently unstable to fragment and form bound objects @xcite .
this has been suggested as a mechanism for forming gas giant planets in discs around young stars @xcite or stars in discs around supermassive black holes @xcite . two - dimensional
, shearing sheet simulations @xcite , using a specific heat ratio of @xmath11 , indicated that the boundary between fragmentation and a quasi - steady , self - gravitating state occurred at a cooling time of @xmath12 @xcite found a similar result using three - dimensional smoothed particle hydrodynamics ( sph ) simulations .
as already mentioned , in a quasi - steady state the gravitational instability acts to transport angular momentum outwards . in many instances
it is appropriate to assume that angular momentum transport is driven by disc viscosity which @xcite suggest has the form @xmath13 where @xmath14 is the disc scaleheight .
this form , however , assumes that the viscosity depends only on local disc properties .
given that disc self - gravity is inherently global , such a form is not necessarily suitable for characterising angular transport in self - gravitating discs .
however , if @xmath15 and if the disc mass is less than about half that of the central star , a local approximation appears to be a suitable representation @xcite .
since a quasi - steady state is one in which the cooling is balanced by an effective viscous heating , one can relate the viscosity to the cooling time through @xcite @xmath16 where @xmath17 is the specific heat ratio . using different values of @xmath17
, @xcite showed that rather than the fragmentation boundary depending on the cooling time , @xmath18 , it depends on the stresses in the disc , as represented by @xmath19 . in agreement with @xcite
their results indicate that self - gravitating discs can maintain a quasi - steady state if @xmath20 and will fragment if the required stress exceeds @xmath21 .
@xcite used analytic calculations and three - dimensional numerical simulations to investigate further the energy balance in self - gravitating discs and found that the perturbation amplitude , @xmath22 , is related to the cooling time through @xmath23 where @xmath24 . using two - dimensional shearing - sheet simulations
, @xcite showed , similarly , that @xmath25 .
the basic picture that has therefore been developed is that a self - gravitating accretion disc will settle into a quasi - steady state in which cooling is balanced by heating driven by the gravitational instability .
in such a state , the perturbation amplitudes will depend on the cooling rate ( or , equivalently , on the level of stress in the disc ) and if these perturbations are sufficiently large , they become non - linear , the disc is unable to maintain a quasi - steady state and instead fragments into bound objects .
what makes this general picture attractive is that there is reasonable agreement across a wide - range of different types of simulations including two - dimensional shearing sheet simulations @xcite , three - dimensional grid - based simulations @xcite , and three - dimensional sph simulations @xcite .
recent work @xcite has , however , shown that three - dimensional sph simulations that fix @xmath26 , do not converge to a well - defined fragmentation boundary as resolution is increased .
their highest resolution simulations suggested that fragmetation could occur for cooling time @xmath27 .
given that the jeans mass of a typical fragment is well - resolved in an sph simulation even for quite modest resolutions @xcite , this result is quite surprising .
there have been a number of attempts to understand this result .
@xcite suggest that numerical viscosity may influence disc thermodynamics more than originally thought and hence that simulations may require higher resolutions than indicated by earlier calculations .
@xcite use two - dimensional grid - based simulations to show that the lack of convergence could be related to edge effects in simulations with very smooth initial conditions .
there is also some suggestion that disc fragmentation may have a stochastic nature @xcite .
@xcite considered convergence in three - dimensional , grid - based , self - gravitating disc simulations .
this work , however , did nt directly address convergence of the fragmentation boundary , but instead considered convergence of the properties of a quasi - steady , self - gravitating disc .
their results were consistent with fragmentation requiring @xmath21 , but could nt really make any strong statements about convergence of the fragmentation boundary .
@xcite , extended the work of @xcite to consider simulations with radiative cooling and found that the properties of these discs did not converge in the outer , optically - thin regions .
one of their conclusions was that there may be issues with convergence in regions where the optical depth is of order unity .
it has , however , been suggested @xcite that the lack of convergence of the fragmentation boundary in the @xcite simulations was simply a consequence of the manner in which the cooling was implemented .
@xcite suggest that it may be related to the known problem - in many sph implementations - of an unphysical discontinuity in the thermal energy ( pressure ) at contact discontinuities @xcite which , if not corrected for , could lead to regions with enhanced cooling .
@xcite suggested that this could be solved by using a form of the cooling that smooths across each sph particle s neighbour sphere . in their simulations ,
fragmentation occurred only for cooling times @xmath28 but they could not claim convergence as their highest resolution simulations fragmented for a slightly longer cooling time than the value towards which the others appeared to be converging .
@xcite have recently extended this convergence work to consider how it is affected by artificial viscosity in both sph simulations and in fargo grid - based simulations .
they consider various artificial viscosity parameters and settle on the values that maximise the value of @xmath26 for which fragmentation can occur . in a quasi - steady state
the disc is in thermal equilibrium with the imposed cooling balanced by heating from both the instability and from artificial viscosity .
ideally , the artificial heating should be minimised .
the perturbation amplitudes should depend on the strength of the instability @xcite , therefore choosing artificial viscosity parameters that maximise the value of @xmath26 at which fragmentation can occur should minimise the level of artificial heating .
there are , however , a few issues with this that will be discussed in more detail in a later section . artificial viscosity is typically introduced so as to resolve shocks . in sph , however , the artificial viscosity operates even in the absence of shocks , producing an artificial dissipation that should , ideally , be minimised . the way in which the instability heats the disc is through dissipation at shocks and
so varying the viscosity parameters can reduce the artificial dissipation , but can also have an impact on the shock heating .
given that @xcite suggest that the lack of convergence seen in @xcite could be due to an unphysical structure in the shock regions , changing the shock structure could exacerbate this problem . additionally , in fargo , the artificial viscosity only operates at shocks and does nt produce any kinematic viscosity , so should nt introduce any artificial heating .
any artificial diffusion in fargo should occur at the grid scale and therefore should nt depend on the artificial viscosity parameters . in this paper
we extend the work of @xcite to show that their suggested cooling formalism does indeed appear to lead to convergence of the fragmentation boundary and that , consistent with earlier work , fragmentation requires @xmath29 for @xmath30 .
we then extend this to consider how this results depends on the artificial viscosity in sph to establish if the values suggested by @xcite are indeed optimal .
we also discuss their results obtained using fargo . in section 2
we briefly describe smoothed particle hydrodynamics ( sph ) . in section 3
we consider the sph results unsing the cooling formalism suggested by @xcite . in section 4
we address the results @xcite obtained using fargo and in section 5 we discuss these results and conclude .
sph is a lagrangian hydrodynamic formalism in which a fluid , or gas , is represented by pseudo - particles ( see e.g. , @xcite ) .
each particle is assigned a a mass ( @xmath31 ) , position ( @xmath32 ) , velocity ( @xmath33 ) and internal energy per unit mass ( @xmath34 ) .
there are many descriptions of sph ( e.g. , @xcite ) , so we wo nt repeat the details here . basically , fluid / gas density is calculated via interpolation across the mass distribution .
pressure is determined via an equation of state .
gravitational forces can either be calculated by direct summation , or - more commonly - using a tree code @xcite .
the momentum and energy equations are in a form suitable for this langrangian formalism and so the particles velocities are updated using the gravitational and pressure forces on each particle , the positions are updated using the velocity of each particle , and the internal energy changes via @xmath35 work , viscous dissipation and cooling . to investigate the evolution of self - gravitating discs , a cooling time , @xmath36 , of the of the following form
is typically used @xcite : @xmath37 this is typically added to the energy equation by assuming that the thermal energy of each particle , @xmath38 , decays with an e - folding time given by @xmath36 .
hence , the energy equation becomes , @xmath39 where the sum is over all the neighbours , @xmath40 , of particle @xmath41 , @xmath42 is the velocity difference between particle @xmath40 and particle @xmath41 , @xmath43 is the smoothing kernel used to interpolate across the neighbours of particle @xmath41 , and @xmath44 is the smoothing length that defines the volume of the neighbour sphere . as already mentioned , the unsmoothed - or individual particle - internal energy has an unphysical discontinuity at the contact discontinuity behind shock waves @xcite .
consequently , @xcite suggested that the cooling be distributed across the neighbour sphere .
their suggestion was that the standard cooling term implementation @xmath45 should be replaced with @xmath46 their suggestion was that the upper of the two equations in equation ( [ eq : smoothcool2 ] ) would be applied to particle @xmath41 , while the lower of the two equations would be applied to the neighbours @xmath40 of particle @xmath41 .
this would ensure that the cooling rate associated with particle @xmath41 would be @xmath47 in regions without discontinuities this will be the same as the cooling rate given by equation ( [ eq : basic_cool ] ) @xcite . at the discontinuities in the shock regions ,
equation ( [ eq : totalcoolj ] ) will use the physically correct interpolated value for the internal energy and will ensure that the unphysical jump in thermal energy at the contact discontinuity does not artificially enhance the cooling in that region .
the results presented in @xcite using the cooling form shown in equation ( [ eq : smoothcool2 ] ) suggested that the simulations were converging towards a fragmentation boundary between @xmath48 and @xmath49 .
their highest resolution simulation ( 10 million particles ) , however , fragmented between @xmath50 and @xmath51 and hence they could not claim convergence .
their highest resolution simulation was , however , only a ring of 4 million particles that would have had the same resolution as a full 10 million particle simulation .
strictly speaking , it was nt exactly the same conditions as the other lower - resolution simulations . here
, we present results from a single 10 million particle run using the cooling form presented in equation ( [ eq : smoothcool2 ] ) and that indicates that a stronger constraint on the convergence of the fragmentation boundary . in sph
, the artificial viscosity has two main roles ; to prevent particle interpenetration and to resolve shock waves .
the viscosity can consequently be thought of as having a shear component and a bulk component @xcite , with the bulk component acting very like a von neumann - richtmeyer viscosity used to resolve shocks in many grid - based codes .
a way in which to introduce viscosity in sph , and what is used in all the simulations presented here , is to use the following to determine the viscosity term between particles @xmath40 and @xmath41 ; @xmath52 the terms @xmath53 and @xmath54 are the average sound speed and density for particles @xmath40 and @xmath41 .
the terms @xmath42 and @xmath55 are @xmath56 and @xmath57 respectively .
the coefficients @xmath58 and @xmath59 determine the strength of the two viscosity terms and the term @xmath60 is given by @xmath61 where @xmath62 is a softening term that prevents the denominator in equation ( [ eq : mu ] ) from ever being zero .
it should be clear that equation ( [ eq : visc ] ) only operates when particles are converging and hence will prevent interpenetration and resolve shocks .
however , this form of the viscosity does not distinguish between converging flows and shear flows and hence this viscosity can also transport angular momentum and can , therefore , heat the system in the absence of shocks @xcite .
the viscosity coefficients typically satisfy @xmath63 and commonly used values , in self - gravitating disc simulations , are @xmath64 and @xmath65 .
one reason for using @xmath63 is that it ensures that the first term in equation ( [ eq : visc ] ) dominates when the convergence is slow , while the second term dominates when convergence is rapid .
the @xmath59 term is essentially necessary so as to handle high mach - number shocks @xcite .
consequently , most studies @xcite only consider the @xmath58 term when determining the dissipation due to artificial viscosity in shear flows .
optimally the @xmath58 value should be set so as to minimise artificial dissipation while still preventing particle interpenetration .
@xcite point out that the dissipation associated with the @xmath59 term is not actually negligible but is about a factor of @xmath66 smaller than that associated with the @xmath58 term when @xmath64 and @xmath65 .
they , therefore , conclude that one should optimise in terms of both @xmath58 and @xmath59 and conclude that the optimal values are @xmath64 , @xmath67 .
this was largely based on simulations that maximised the value of @xmath26 - which determines the cooling time - for which fragmentation could occur .
maximising the cooling time at which fragmentation occurs , suggests that one has minimised the amount of artificial dissipation and so is an attractive strategy to adopt .
however , as we ll discuss in more detail later , this may not necessarily be the case and so , here , we investigate how these values of @xmath58 and @xmath59 influence the fragmentation boundary when using the modified cooling form proposed by @xcite .
all of the sph simulations presented here have the same basic setup as those presented by @xcite .
they have a central star with mass @xmath68 surrounded by a disc extending from @xmath69 to @xmath70 , with a mass of @xmath71 , an initial surface density profile of @xmath72 , and with an initial minimum @xmath2 parameter of @xmath2 = 2 .
we impose a cooling of the form described by equation ( [ eq : tcool ] ) , but that is either implemented as in @xcite - which we call basic cooling - or in the modified manner suggested by @xcite - which we call smoothed cooling .
we consider various resolutions , ranging from 250000 particles to 10 million particles . in all our simulations
we take @xmath64 , but consider both @xmath65 and @xmath67 .
in the work of @xcite they considered full simulations using 250000 , 500000 , and 2 million particles , but represented a 10 million particle simulation using a simulation with 4 million particles with a mass of @xmath73 and that extended from @xmath74 to @xmath70 .
these parameters were chosen so as to have the same properties , in that region , as a full 10 million particle simulation . in @xcite the 500000 and 2 million particle simulations fragmented at between @xmath48 and @xmath49 and appeared to be converging .
the pseudo-10 million particle simulation , however , fragmented between @xmath50 and @xmath51 and hence they could not claim convergence .
we have since managed to complete a full 10 million particle simulation with @xmath50 which , after 6.5 outer rotation periods , shows no signs of fragmentation .
the state of the simulation at this time ( 5120 code units or 815 orbits at @xmath75 ) is shown in fig .
[ fig:10million ] .
there is clearly lots of spiral structure , but no evidence of fragmentation .
we also include a table with the results from the simulations of @xcite together with this new result using 10 million particles .
.list of the simulations using smoothed cooling.[tab : sims ] [ cols="^,^,^,^",options="header " , ] table [ tab : sims2 ] shows the results of the smoothed cooling simulations using @xmath67 .
[ fig : fragbound_bsph2 ] compares the results using @xmath76 ( triangles ) with those obtained using @xmath67 ( squares ) .
the symbols are located at the average of the maximum @xmath26 for which fragmentation occured and the minimum for which it did nt .
the short lines indicate the range between these two values .
although there is a difference , the results are very similar .
there is no indication that @xmath77 increases by 50% when @xmath67 , compared to that obtained when @xmath65 .
given that previous analysis @xcite indicates that @xmath67 should produce more artificial dissipation than @xmath65 , it is a little surprising that @xmath77 is nt smaller when using @xmath67 than when using @xmath65 .
it is possible that using @xmath67 does reduce some of the random noise associated with sph , but it is also clear ( from fig .
[ fig:500k_comparison ] ) that it also changes the shock structure in the disc .
maybe it is not that surprising that the results differ slightly .
these results are , however , not consistent with the suggestion @xcite that - even for large particle numbers - artificial dissipation provides @xmath78% of the heating when @xmath65 . ( squares ) and with @xmath65 ( triangles ) .
the symbols are located at the average of the maximum @xmath26 for which fragmentation occured and the minimum for which it did nt , while the lines indicates the range between these two values . ]
in addition to considering how self - gravitating discs evolve in sph simulations , @xcite have extended this to consider grid - based simulations of self - gravitation accretion discs .
they use the fargo code and show too that the fragmentation boundary does not converge as resolution increases . in this work , they also vary the artificial viscosity parameter so as to maximise the cooling time at which fragmentation occurs .
there are , however , some issues with how they have implemented these fargo simulations . as with sph ,
grid - based methods such as fargo @xcite require a form of artificial viscosity to handle shocks .
this is a direct consequence of godunov s theorem @xcite : any numerical scheme that is better than first - order accurate will introduce unphysical oscillations in the flow near shocks . since almost all numerical methods for gas dynamics aim for at least second - order accuracy , a special recipe is needed around discontinuities in the flow .
finite - difference methods like zeus @xcite and fargo employ a van - neumann - richtmeyer type of artificial viscous pressure , which , when considering an axisymmetric disc , takes the form : @xmath79 where @xmath8 denotes the surface density , @xmath80 the radial velocity and @xmath81 is the artificial viscosity parameter , which has dimensions of length . the reduced artificial viscosity parameter @xmath82 , where @xmath83 is the grid spacing , determines over how many grid points shocks will be smeared out . in this view , of course , only values of @xmath84 larger than unity make sense , and the standard value in fargo is @xmath85 . choosing a nonlinear viscous pressure as artificial viscosity results in the correct entropy jump across shocks and the correct shock propagation velocity @xcite .
the form of artificial viscosity given in equation ( [ eqartvisc ] ) has two important properties : it acts only when the flow is compressed , and it acts , for @xmath84 of order unity , only on length scales of the order of the grid scale .
this latter property implies that if the value of @xmath84 makes a difference in the outcome of a simulation , the flow _ must _ be under resolved .
we illustrate the effect of artificial viscosity on two one - dimensional ( axisymmetric ) problems below , one linear and one non - linear .
for the first problem , we take an equilibrium inviscid keplerian disc , extending from @xmath86 to @xmath87 , with @xmath88 at @xmath89 , and add a radial velocity perturbation @xmath90 note that this is a velocity perturbation equal to @xmath91 % of the sound speed . therefore ,
no shocks form in this problem , which means that entropy should be materially conserved . for simplicity , we take the surface density to be constant initially , and choose the pressure so that the initial state has constant entropy ( @xmath92 , where we take the ratio of specific heats @xmath93 ) , which means that in an ideal world , the quantity @xmath94 should remain constant . no numerical method is ideal , of course , and there are two sources of changes in @xmath95 : one is due to the finite size of the grid cells , which , unless a special entropy - conserving integration scheme is adopted , will lead to spurious changes in @xmath95 , and the other is artificial viscosity , which directly changes the entropy through the viscous heating term .
the results after integrating to @xmath96 are displayed in figure [ fig : fargolinear ] for three different grid sizes @xmath97 , with corresponding resolutions @xmath98 . the initial gaussian pulse in velocity
is resolved by @xmath99 cells for @xmath100 , @xmath101 cells for @xmath102 and @xmath103 cells for @xmath104 . as the artificial viscosity
is increased , the maximum change in @xmath95 increases due to viscous heating , as expected .
the increase in entropy at @xmath105 is due to the grid only . at the lowest resolution ,
the effects of the grid and the artificial viscosity are of similar magnitude .
fargo , like most grid - based methods , is second order accurate .
this means that after a fixed number of time steps , errors should decrease as @xmath106 as the grid is refined .
since the number of time steps required to reach @xmath96 is proportional to @xmath97 , we expect the errors due to the grid at @xmath96 to decrease as @xmath83 , which is exactly what is observed in figure [ fig : fargolinear ] .
as the resolution is increased , the differences between runs with @xmath105 and @xmath107 decrease . for all resolutions except @xmath100 , taking @xmath107 makes no difference compared to @xmath105 . in other words , for @xmath102 and @xmath104 , heating due to artificial viscosity
is completely negligible for @xmath107 . only for @xmath100
does artificial viscosity make a difference , but this is to be expected , since the extent of the initial pulse comparable to the grid scale , which means it will feel the artificial viscosity .
therefore , as expected , artificial viscosity plays no role in heating a smooth flow , unless it is not resolved . as a second test problem
, we set up a nonlinear wave characterised by initial conditions @xmath108 in the same inviscid equilibrium keplerian disc as above .
the initial pressure is set again such that the initial state is isentropic .
the solution develops a shock that leads to an increase in entropy .
the correct increase in @xmath95 was measured from a simulation using a riemann solver @xcite at @xmath104 .
the results obtained with fargo are displayed in figure [ fig : fargononlinearq ] .
it is immediately clear that simulations with @xmath109 strongly underestimate the change in entropy .
this is again not surprising , since artificial viscosity is _ necessary _ in this case because of the presence of an entropy - generating shock .
similar to sph , choosing the artificial viscosity parameter too high leads to unphysical , artificial heating : choosing @xmath110 leads to too much viscous heating for @xmath100 , but at high enough resolution a plateau emerges giving roughly the same amount of entropy generation independent of @xmath84 . in figure [ fig : fargononlinearn ]
we look at the same problem but now as a function of resolution .
simulations with @xmath111 systematically underestimate the entropy production , independent of resolution .
the situation is most severe for @xmath105 and @xmath112 , which do not even show convincing signs of convergence with resolution .
simulations with @xmath113 do seem to converge , but to a level that depends on @xmath84 .
for the largest values of @xmath84 , the entropy increase converges to a value very close to the correct one , but at the price of overestimating the entropy increase at low resolution . the standard value @xmath85 seems to be a good compromise .
it is not completely straightforward to translate the above results to two dimensions . in this case ,
an extra source of error comes from dimensional splitting , the effects of which are not entirely clear especially if the fargo algorithm is used @xcite .
moreover , shocks are no longer necessarily aligned with the grid , which in all likeliness changes the dissipation properties of the grid . however , a few general statements can be made .
the optimum value of @xmath84 is larger than unity .
for example , @xmath85 gives artificial viscosity that leads to a good estimate of entropy production in shocks ( figure [ fig : fargononlinearq ] ) , while its effect on _ resolved _ smooth flow is negligible ( figure [ fig : fargolinear ] ) .
values of @xmath114 do not give the correct amount of shock heating , and can therefore not be expected to give physical results . note that the optimum value chosen by @xcite , @xmath112 , underestimates the entropy increase in a shock by a factor of 3 ( see figure [ fig : fargononlinearn ] ) .
this will have serious consequences for the simulation results if heating is due to shock dissipation , which is the case for self - gravitating discs .
if the amount of artificial viscosity makes a difference in the results , the flow is under resolved .
this can be due to shocks , in which case there the flow can in fact not be resolved , or due to unresolved smooth flow . in the latter case , reducing the amount of artificial viscosity will in general not lead to a much better solution , since errors due to the finite size of the grid cells are likely to be as large as the error introduced by artificial viscosity , precisely because the flow is unresolved ( figure [ fig : fargolinear ] ) .
moreover , reducing the amount of artificial viscosity can only be safely done when it is absolutely certain that no shocks are present in the problem , which , for self - gravitating discs , we know is not the case . keeping the artificial parameter fixed at @xmath85 ,
there are still several avenues for investigating the problem of convergence in grid - based simulations .
a direct comparison between 2d global and local simulations @xcite , which do not appear to show convergence @xcite , is definitely warranted .
care must be taken in global simulations to avoid initial transients @xcite .
it may be that the lack of convergence in grid - based simulations is the result of the two - dimensional approximation , even though smoothing of the gravitational potential does not seem to make much of a difference @xcite .
it may also be that the inviscid problem is ill - posed , and that a finite amount of ( navier - stokes ) viscosity is needed to reach convergence .
artificial viscosity is a necessary feature in numerical simulations in order to make the code behave in a reasonable way at the smallest resolvable scale , which is the scale of the grid in grid - based codes , and the smoothing length in sph . without artificial viscosity , particle interpenetration would make the outcome of any sph simulation useless .
the use of a grid introduces its own associated viscosity , which acts on the smallest resolvable scale ( the grid scale ) and may or may not behave like a real viscosity .
this is why simulations of turbulence often employ an additional physical navier - stokes viscosity to make sure that energy dissipation on the smallest scales is well - behaved and physical ( see e.g. fromang , papaloizou , lesur , & heinemann 2007 ) .
in addition , both sph and grid - based codes like fargo need artificial viscosity to handle shocks correctly
. it should be clear from the discussion in sections [ sec : results ] and [ sec : fargo ] that care should be taken when trying to adjust the amount of artificial viscosity for a particular problem , both in grid - based codes and sph .
when the level of artificial viscosity makes a difference in the simulation outcome , this means that features close to the smallest resolvable scale ( either the grid scale or the smoothing length ) play an important role .
artificial viscosity is a way of converting bulk motion on these small scales into heat .
this can be unphysical , if the underlying flow is smooth , or physical , in the case of shocks .
of course , one would always like to reduce the amount of artificial viscosity as much as possible .
however , reducing the amount of artificial viscosity below the level required to handle shocks correctly , e.g. taking @xmath114 in fargo , can only be done if it is known that no shocks will occur in the problem . and
even then , the results in section [ sec : fargo ] indicate that for smooth flow dissipation is dominated by the grid rather than artificial viscosity , so that changing @xmath84 will not make a difference in the simulation outcome . the only way to do a better job for smooth flow in a grid - based simulation is to increase the resolution . in sph , there are more artificial viscosity parameters to play with , and it may be slightly less clear to what extent the choices for @xmath58 and @xmath59 are free . since there is no dissipation due to a grid , it is likely that the level of artificial viscosity matters more compared to grid - based codes .
typically sph simulations use @xmath63 .
this is so that the linear term dominates when particles are converging slowly , and the quadratic term dominates when the particles are converging fast enough that shocks are likely to form . using @xmath115 , as suggested by @xcite , effectively means that the quadratic term is likely to always dominate and , hence , is likely to change the properties of the simulation itself . in the case where shocks are present , the situation is more complicated , since the numerical method will always smear shocks over a few grid cells , for a grid - based code , or a few smoothing lengths for sph .
increasing the resolution therefore keeps reducing the size of shocks , which means that ever smaller scales are present in the problem .
these smallest scales can interact in non - trivial ways with larger scale structures ( clumps , waves ) , and it may not be immediately clear whether convergence can be reached and at what resolution .
recently , @xcite have suggested that , in three - dimensional sph simulations of self - gravitating accretion disc with @xmath0-cooling , the cooling time ( @xmath26 ) at which fragmentation occurs does not converge as resolution is increased .
they ve extended this work @xcite to suggest that with typical artificial viscosity parameters , the fragmentation boundary is converging towards a critical cooling time of @xmath116 .
they go on to argue , however , that adjusting the artificial viscosity parameters ( so as to maximise the critical cooling time , @xmath77 ) suggests that , with an appropriate choice of the artificial viscosity parameters , the simulations actually converge towards @xmath117 .
@xcite then continue this by considering the evolution of self - gravitating disc using the grid - based code fargo . here
they also vary the artificial visocsity parameter so as to maximise the critical cooling time and show that these simulations also do nt converge as resolution increases .
it has been suggested @xcite that the non - convergence seen in @xcite was simply a consequence of the manner in which cooling was implemented .
we ve extended the work of @xcite here to show that by implementing what they call smoothed cooling , fragmentation requires @xmath29 for all resolutions considered ( from 250000 particles to 10 million particles ) .
this is more consistent with other work @xcite and also makes physical sense given that the jeans mass is well resolved in most of the disc for simulations with 500000 particles or more , and that artificial viscosity should be providing less than 10% of the heating in such simulations , so should nt be significantly influencing the fragmentation boundary . furthermore
, if fragmentation can occur for @xmath118 this suggests that a clump can contract and become bound even though the timescale over which it is losing energy is significantly greater than the orbital period , which is likely to determine the timescale over which we d expact the clump to heat .
we also consider how the alternative artifical viscosity values suggested by @xcite influence the results when using smoothed cooling .
we find that the results are consistent with those obtained using the original artificial viscosity parameters . rather than increasing the critical cooling time , @xmath77 , by 50% @xcite
, fragmentation still requires @xmath29 , for all particle numbers considerd .
if this change to the artificial viscosity parameters was reducing the level of artificial heating ( as suggested by @xcite ) then we d expect the results to be independent of the implementation of the cooling .
that they are nt suggests that changing these parameters is influencing the simulations in some numerical way , rather than simply changing the level of artificial heating - especially as the expectation is that the changes made by @xcite should have increased , rather than reduced , the level of artificial heating .
this is also consistent with the significant difference between the spiral shock structure in the disc with @xmath67 when compared to simulations with @xmath65 .
what was quite attractive about the @xcite work was that they obtained very similar results when using the grid - based fargo code .
however , here , they also varied the artificial viscosity parameter so as to minimise the level artificial viscosity .
as discussed earlier , however , the artificial viscosity in fargo only acts on converging flows and so should not produce any artificial ( non - shock related ) heating .
additionally , in order to produce good estimates of entropy production at shocks requires that , as discussed earlier , the optimal value for the artificial viscosity in fargo be larger than unity .
@xcite claim that the optimal value is @xmath119 , much smaller than would be regarded as suitable for such a simulation .
admittedly , even their simulations with @xmath120 did not show signs of convergence , but this could be related to unsuitable initial conditions @xcite or the possibility of stochasticity @xcite and should be investigated further .
essentially , the sph simulations presented here show that if one implements the cooling so as to remove the unphysical discontinuity at the contact discontinuity behind shocks @xcite we appear to get convergence as the resolution increases , and the fragmentation boundary that we determine is consistent with earlier work ( fragmentation occuring for cooling times between @xmath48 and @xmath50 for @xmath30 ) .
although , we havent investigated @xcite s fargo results in as much detail , it seems clear that what they regard as the optimal value for the artificial viscosity parameter ( @xmath119 ) is well below what would be regarded as acceptable for such simulations .
this might suggest that their fargo results suffer from additional numerical issues . with the exception of the possibility of stochastic fragmentation @xcite
we therefore conclude that there is no real evidence that fragmentation can occur in self - gravitating discs with long cooling times and that the likely fragmentation boundary is similar to that suggested by earlier work @xcite .
consequently , this implies that - as suggested by earlier studies @xcite - gas giant planet formation via disc fragmentation is unlikely in the inner regions ( @xmath1 au ) of protostellar discs .
all simulations presented in this work were carried out using high performance computing funded by the scottish universities physics alliance ( supa ) .
wkmr and dhf gratefully acknowledge support from stfc grant st / j001422/1 .
pja ackowledges support from nasa s astrophysics theory and origins of solar systems programs through grants nnx11ae12 g and nnx13ai58 g .
durisen r. , boss a.p . , mayer l. , nelson a.f . , quinn t. , rice w.k.m . , 2007 , in reipurth b. , jewitt d. , keil k. , eds , protostars and planets v , gravitational instabilities in gaseous protoplanetary disks and implications for giant planet formation , university of arizona press kuiper g. , 1951 , in hynek j.a . ,
, proceedings of a topical symposium , c commemorating the 50th anniversary of the yerkes observatory and half a century of progress in astrophysics , mcgraw - hill , new york , p. 357
|
recently it has been suggested that the fragmentation boundary in smoothed particle hydrodynamic ( sph ) and fargo simulations of self - gravitating accretion discs with @xmath0 cooling do not converge as resolution is increased .
furthermore , this recent work suggests that by carefully optimising the artificial viscosity parameters in these codes it can be shown that fragmentation may occur for much longer cooling times than earlier work suggests .
if correct , this result is intriguing as it suggests that gas giant planets could form , via direct gravitational collapse , reasonably close to their parent stars .
this result is , however , slightly surprising and there have been a number of recent studies suggesting that the result is likely an indication of a numerical problem with the simulations .
one suggestion , in particular , is that the sph results are influenced by the manner in which the cooling is implemented .
we extend this work here and show that if the cooling is implemented in a manner that removes a known numerical artefact in the shock regions , the fragmentation boundary converges to a value consistent with earlier work and that fragmentation is unlikely for the long cooling times suggested by this recent work .
we also investigate the optimisation of the artificial viscosity parameters and show that the values that appear optimal are likely introducing numerical problems in both the sph and fargo simulations .
we therefore conclude that earlier predictions for the cooling times required for fragmentation are likely correct and that , as suggested by this earlier work , fragmentation can not occur in the inner parts ( @xmath1 au ) of typical protostellar discs .
[ firstpage ] accretion , accretion discs - gravitation - instabilities - stars ; formation - stars ;
|
little higgs models @xcite have been proposed as a solution to the hierarchy problem , the vast difference between the scales of electroweak symmetry breaking and the scale where quantum effects of gravity can no longer be neglected .
the higgs boson emerges as a pseudo - goldstone boson of spontaneously broken chiral symmetries of a strongly interacting theory in the multi - tev regime .
the quadratically divergent contributions to the higgs mass are canceled unlike in supersymmetry by particles of like statistics , chiral multiplet partners of the standard model ( sm ) fermions .
the collective breaking of these symmetries generates a quadratically - divergent correction to the higgs mass only at two - loop order , such that the models remain weakly interacting at the tev scale , thereby satisfying electroweak precision data constraints .
ideas about possible uv completions exist @xcite , but we ignore them in this paper and take little higgs models as low - energy effective theories .
various realizations of the little higgs symmetry structure have been proposed @xcite , grouped quite generally into two classes ( cf .
e.g. @xcite ) .
the original models are mostly product group models , where the electroweak gauge group is extended to a product gauge group that contains @xmath1 subgroups in at least two distinct factors . the physical @xmath2 and @xmath0 bosons are mixtures of the gauge bosons from both factors .
the global symmetry group is a simple group that includes the full gauge structure [ e.g. , @xmath3 , and the higgs is part of an irreducible multiplet of pseudo - goldstone bosons . in simple group models ,
the electroweak @xmath1 gauge group is embedded in a simple ( e.g. , @xmath4 ) gauge group , while the global symmetry group has a product structure , and the higgs is a mixture of the components of several ( at least two ) independent irreducible representations of the factor groups .
although these models remain weakly interacting at the tev scale , nevertheless electroweak precision observables pose constraints on their parameter space @xcite .
the decay constant @xmath5 of the pseudo - goldstone bosons which sets the mass scale for the additional vector bosons , scalars and fermions in little higgs models , has to lie in the range @xmath6 tev , where the upper bound comes from naturalness considerations .
usually the masses of the additional particles lie in the range of several hundred gev up to a few tev , and are therefore ( partly ) directly accessible at the lhc ( for an overview about little higgs phenomenology , cf .
the proof of the little higgs mechanism within these models relies on the discovery of the nonlinear goldstone boson structure of the higgs and the cancellation mechanism of the quadratic divergences based on measurements of the couplings .
this will not be possible in all cases at the lhc , but should be at a future international @xmath7 collider ( ilc ) .
an ilc would in any case be necessary for precision measurements .
to reveal a special group theoretical realization of little higgs models is also quite difficult and relies on precise branching ratio measurements for the higgs , the extra gauge bosons and fermionic states . a method to distinguish between simple and product group models , maybe even already at the lhc , would be highly welcome .
we propose a method based on the discovery and the properties of light pseudoscalar states in the spectrum , the little higgs pseudo - axions @xcite .
we first recall their properties , then show that a specific pseudo - axion coupling , to the higgs and the @xmath0 bosons , can occur only in simple group models .
it could thus be used as a discriminator between the two little higgs model classes .
we then discuss relevant phenomenology at the lhc and ilc .
in little higgs models , there can occur spontaneously broken ( approximate ) global @xmath8 symmetries , corresponding to diagonal generators of the overall non - abelian global symmetry group @xcite .
to each of these @xmath8 factors , which might well be anomalous , corresponds a ( pseudo- ) goldstone boson which couples to fermions like a pseudoscalar , analogous to the @xmath9 meson in chiral symmetry breaking of qcd .
such a particle is an electroweak singlet , which gets a mass through explicit symmetry breaking terms and the coleman - weinberg potential .
all couplings to sm particles are therefore suppressed by the ratio of the electroweak and the little higgs scale , @xmath10 .
all couplings to sm gauge bosons are induced by anomalous triangle loops . as a typical example for these particles ,
we show the situation for the so - called simplest little higgs or @xmath11model @xcite in fig .
[ fig : par_space ] . since this is a simple group model ,
there are two multiplets of goldstone bosons connected by a mixing angle @xmath12 as in two - higgs - doublet models .
furthermore , there is an explicit breaking of the global little higgs symmetries , analogous to the @xmath11 term in the mssm .
pseudo - axion phenomenology at the lhc and future ilc and photon colliders was discussed in @xcite .
there it was shown that in the simplest little higgs there is a tree - level coupling @xmath0@xmath13@xmath14 of the pseudo - axion to the higgs and the @xmath0 , which can only arise by electroweak symmetry breaking , and is enhanced by @xmath12 .
on the other hand , it was shown that this coupling is absent in the simplest candidate of the product group models , the littlest higgs .
we now show that the existence of such a @xmath0@xmath13@xmath14 coupling is a property of simple group models , and that it can not appear in product group models .
hence , it serves as a discriminator between the two categories .
the crucial observation is that the matrix - representation embedding of the two non - abelian @xmath1 gauge groups , and especially of the two @xmath8 factors within the irreducible multiplet of the pseudo - goldstone bosons of one simple group ( e.g. @xmath15 in the littlest higgs ) , is responsible for the non - existence of this coupling in product group models .
it is exactly the mechanism which cancels the quadratic one - loop divergences between the electroweak and heavy @xmath1 gauge bosons which cancels this coupling . in simple group models the higgs mass term cancellation
is taken over by enlarging @xmath1 to @xmath4 , and the enlarged non - abelian rank structure cancels the quadratic divergences in the gauge sector but no longer forbids the @xmath0@xmath13@xmath14 coupling . to leading order in the pseudo - axion field @xmath14 , the parameterization of the goldstone boson manifold in the effective lagrangian does not depend on the basis choice of the broken generators @xcite
. therefore one can take it proportional to the unit matrix , i.e. factor it out from the matrix of pseudo - goldstone boson .
this corresponds to a separation of the special @xmath16 group ( cf . also the discussion in refs .
@xcite and @xcite ) . here
we use @xmath17 $ ] for the pseudo - axion field and @xmath18 $ ] for the non - linear representation of the remaining goldstone multiplet @xmath19 of higgs and other heavy scalars .
then , for product group models , the kinetic term may be expanded as @xmath20 & = \ldots + f^2 ( \partial_\mu \xi ) \xi^\dagger { \mathop{\rm tr}}\left [ ( d^\mu \sigma)^\dagger \sigma \right ] + { \text{h.c.}}\nonumber\\ & = \ldots - 2f(\partial_\mu \eta)\ , { \text{im}}{\mathop{\rm tr}}\left [ ( d^\mu \sigma)^\dagger \sigma \right ] + o(\eta^2),\end{aligned}\ ] ] where we write only the term with one derivative acting on @xmath21 and one derivative acting on @xmath22 .
this term , if nonzero , is the only one that can yield a @xmath0@xmath13@xmath14 coupling .
we now use the special structure of the covariant derivatives in product group models , which is the key to the little higgs mechanism : d_= _ + a_1,^a ( t_1^a + ( t_1^a)^t ) + a_2,^a ( t_2^a + ( t_2^a)^t ) , where @xmath23 are the generators of the two independent @xmath1 groups ( extra gauge structure does not matter ) , and @xmath24 + heavy fields in a suitable normalization ( cf . ref .
@xcite ) . neglecting the heavy gauge fields and extracting the electroweak gauge bosons , we have @xmath25 & \sim w_\mu^a { \mathop{\rm tr}}\left [ ( \sigma^\dagger ( t_1^a + t_2^a)\sigma + ( t_1^a + t_2^a)^ * \right ] \notag \\ & \quad = w_\mu^a { \mathop{\rm tr}}\left [ ( t_1^a + t_2^a ) + ( t_1^a + t_2^a)^ * \right ] = 0.\end{aligned}\ ] ] this vanishes due to the zero trace of @xmath1 generators .
the same is true when we include additional @xmath8 gauge group generators such as hypercharge , since their embedding in the global simple group forces them to be traceless as well .
we conclude that the coefficient of the @xmath0@xmath13@xmath14 coupling vanishes to all orders in the @xmath26 expansion .
next , we consider the kinetic term for simple group models , where we use the following notation for the nonlinear sigma fields : @xmath27 , where @xmath28 $ ] and @xmath29 is the vacuum expectation vector directing in the @xmath30 direction for an @xmath4 simple gauge group extension of the weak group . thus , in simple group models the result is the @xmath31 component of a matrix : @xmath32 to further evaluate this term , we separate the last row and column in the matrix representations of the goldstone fields @xmath22 and gauge boson fields @xmath33 : [ matrixdecomp ] = 0 & h + h^ & 0 ,
_ = _ & 0 + 0 & 0 + the higgs boson in simple group models sits in the off - diagonal entries ( one doublet for the simplest lh and a pair of doublets for the original simple group model ) , while the electroweak gauge bosons reside in the upper left corner . with the baker - campbell - hausdorff identity
, one gets for the term in parentheses in eq .
( [ kin : simplegroup ] ) : @xmath34 the @xmath31 entry can only be nonzero from the third term on .
this can be understood as follows . in the first term
, the @xmath31 component cancels by the help of the multiple goldstone multiplets present in the simple group models ; it would be a mixing between the @xmath14 and the goldstone boson(s ) for the @xmath35 state(s ) .
if the @xmath31 component of the second term were nonzero , it would induce a @xmath0@xmath13@xmath14 coupling without insertion of a factor @xmath36 .
this is forbidden by electroweak symmetry . to see this ,
it is important to note that @xmath37 in the simple group models while the hypercharge boson @xmath38 is a mixture of the diagonal @xmath8 generator and the left - over diagonal generators in @xmath4 ( @xmath39 in @xmath40 , @xmath41 in @xmath42 , etc . ) .
the embedding of the standard model gauge group always works in such a way that hypercharge is a linear combination of the @xmath43 and @xmath8 generators .
this has the effect of canceling the @xmath44 and @xmath0 from the diagonal elements beyond the first two positions , and preventing the diagonal part of @xmath45 from being proportional to @xmath46 . by using eq .
( [ matrixdecomp ] ) one easily sees that @xmath47 is zero . the third term in the expansion yields a contribution to the @xmath0@xmath13@xmath14 coupling , ( ^)h^_h ~v h z_^ .
since in simple group models the nonlinear goldstone boson multiplets always come in pairs , there may still be a cancellation between parameters that makes the prefactor vanish ; however , this occurs only for degenerate parameter sets .
by contrast , in product group models , the absence of this coupling is a property of the symmetry structure and occurs for all possible parameter values . as a concrete example
, we derive the @xmath0@xmath13@xmath14 coupling in the simplest little higgs or @xmath11model @xcite . in this model ,
the ew group is enlarged to a gauged @xmath48 .
there are two nonlinear sigma fields , each of which parameterizes a coset space @xmath49 : _ 1 = 0 + 0 + f_1 _ 2 = 0 + 0 + f_2 where = \ { + cc 0 & 0 + 0 & 0 & h^ * + h^t & 0 } , f^2 = f_1 ^ 2 + f_2 ^ 2 , = .
the gauge boson multiplet @xmath33 consists of the @xmath50 gauge bosons and the @xmath8 gauge boson @xmath51 .
it decomposes into the electroweak gauge bosons @xmath52 , @xmath0 , the photon @xmath53 , and heavy vector bosons @xmath54 . in the covariant derivative acting on the higgs multiplets
, they enter via the matrix @xcite @xmath55 which results in the following couplings of the pseudo - axion @xmath14 : _ vh = ( h ) .
@xmath56 is defined as in ref .
@xcite : n_2 = = - as long as @xmath57 , or @xmath58 , there is a @xmath0@xmath13@xmath14 coupling as anticipated . to calculate the @xmath14 production rates via gluon - gluon fusion at hadron colliders , and
the @xmath14 branching ratio to observable final states , we also need its fermion couplings . in the @xmath11 model , these are as follows .
the couplings to top quark and heavy top partner fermion @xmath59 are @xmath60 and @xmath61 , where @xmath62 and @xmath63,@xmath64 are the top yukawa couplings of the model , chosen to produce the observed top quark mass and minimize @xmath65 ( cf . ref .
the @xmath66 coupling is @xmath67 , and has two interesting properties : it appears only at higher orders in the expansion of @xmath68 ( yet remains @xmath69 ) , and it is driven rapidly to zero the closer @xmath70 and @xmath71 are ( as @xmath72 ) .
to discuss the possible lhc @xmath0@xmath13@xmath14 coupling phenomenology , we study the simplest little higgs and adopt the parameter set of ref .
@xcite , consistent with existing ew and flavor data and the preference for a light higgs boson .
the relevant parameters are @xmath70 , @xmath71 , @xmath11 and @xmath73 . for a generic scale of @xmath74 tev
at which the little higgs effective theory breaks down , our only free parameters are then @xmath75 and @xmath11 , with the constraint that @xmath70 not be as small as the ew scale , @xmath76 tev from ew precision constraints ( primarily @xmath77 and four - fermion operators ) , and @xmath78 ( @xmath79 ) to avoid too much mixing which would lead to fermion non - universality . @xmath80
( @xmath81 ) lies right at the edge of the limits on the latter , but this point zeroes the @xmath0@xmath13@xmath14 coupling and is of trivial interest .
we vary @xmath11 over a slightly broader range than the calculated @xmath82 would suggest is allowed or favored by data . [ cols= " < , > " , ]
it is possible to perform a simple `` bump - hunting '' analysis , although it is far from optimum .
we observe significant angular correlations in the leptons and @xmath83 jets which can be used to reject the backgrounds further .
[ fig : bbll - ang ] shows the two most important correlations , the @xmath84 and @xmath85 lego plot separations .
there are distinct differences between resonant and continuum production . for the golden point and @xmath86 gev , it is possible to reduce the qcd background by a factor 3 while losing only @xmath87 of the signal
this does sculpt the total invariant mass distribution somewhat , but reduces by about half the amount of luminosity required for discovery .
in general , though , the angular cut choice needed changes with choice of input parameters , as it depends ultimately on the decay kinematics how boosted the daughters are .
we do nt attempt such a complicated analysis here , but point out the correlations for future detailed work at the detector level .
jet pair ( left ) and lepton pair ( right ) . the signal is shown in red for the golden point and @xmath88 gev using dotted , solid and dashed lines , respectively .
the qcd continuum backgrounds is shown in blue ( dot - dashed ) .
vertical lines represent cuts used for the @xmath86 case as described in the text . ] observation of the little higgs pseudo - axion .
the golden point is green , while blue ( red ) are for @xmath89 tev .
crosses represent the @xmath90 analysis for @xmath91 , while diamonds are the same for @xmath92 ( cf .
[ sub : bbll ] ) , and x symbols are for @xmath92 where the higgs boson is heavier and decays to @xmath93 ( cf . subsec .
[ sub:3l2j ] ) .
crosses for the larger @xmath75 choices are above @xmath94 . ]
we present our overall results then , without optimization using angular correlations , in fig .
[ fig : lhc - lumi ] in terms of the required luminosity for one lhc detector to make a @xmath95 observation of the @xmath90 final state above qcd backgrounds . to obtain this we used i d efficiency factors of @xmath96 for each @xmath83 jet and @xmath97 for each charged lepton , and a capture efficiency of @xmath98 for the dilepton mass window ( signal only ) .
we also include k - factors for signal and background to take into account the large qcd rate corrections : 2.3 for the signal @xcite , 1.3 for the qcd @xmath99 @xcite background , and 1.4 for @xmath100 @xcite .
our results are only estimates , not the optimal reach , and not comprehensive over parameter space .
in addition to the golden point , we show two other cases for different choices of @xmath75 . for @xmath101 tev ( @xmath102 ) ,
the collective scale @xmath103 is approximately the same , but the @xmath0@xmath13@xmath14 coupling squared is reduced by an order of magnitude . for @xmath104 tev ( @xmath105 , same as the golden point ) , the scale @xmath5 is about twice as large , but the coupling strength is unchanged . for @xmath106 , shown by crosses , large @xmath5 values drive @xmath82 to several hundred gev .
the @xmath107 br drops rapidly with increasing higgs mass , as it is out - competed by the @xmath108 modes double longitudinal - polarization enhancement .
thus a heavy higgs boson decaying to @xmath107 would likely be observable with lhc luminosity only for fairly small values of @xmath75 and @xmath11 .
the @xmath109 case suffers a somewhat different fate , as shown by the diamonds .
typically for any given @xmath75 there is only a very small range of @xmath11 which can be addressed . if @xmath11 is too large , @xmath82 is smaller than the lep limit .
only a few gev lower than this limit , the higgs boson mass lies in the region where decays to weak bosons dominate , thus the br to @xmath110 drops to zero and the @xmath111 analysis takes over . for larger @xmath75
than at the golden point , @xmath112 increases , lowering the production rate , even though the br to @xmath113 is still large .
for @xmath115 gev , the br to @xmath110 drops off dramatically , being replaced mostly by @xmath116 .
this requires a different analysis . because the @xmath2 boson leptonic br is smaller than that to quarks ( jets ) , and we already have trigger leptons in the final state from the very sharp @xmath0 boson resonance , the obvious final state to consider is the largest in terms of br : @xmath117 .
it is fully reconstructible , and typically better so than the @xmath90 case , because light - flavor jets typically give much less missing energy .
however , this channel suffers from a qcd @xmath0+jets background of several hundred fb @xcite .
a quick analysis of this channel shows that in principle it would be statistically possible for some parameter space , but the pseudo - axion resonance would peak at about the same places the qcd continuum does , with a @xmath118 ratio of about 1/50 , which makes prospects dodgy .
instead , we investigate the @xmath119 channel , which has 2/3 the br of the all - hadronic channel , but about 1/100 the background , which comes almost exclusively from @xmath120 production @xcite .
the final state is then three charged leptons , two jets and missing transverse momentum .
the basic kinematics cuts for detector acceptance and trigger are : [ eq : cuts-4jll-1 ] p_t(j ) > 20 gev , & |(j)| < 4.5 , + p_t ( ) > 15 gev , & |()| < 2.5 , + r(jj , j ) > 0.4 , & r ( ) > 0.2 , + 89.6 < m _
< 92.8 gev , & /p_t
< 30 gev .
the @xmath0-pole cut is the same as before , applied on the @xmath0 decay products explicitly ( combinatorics will be a very minor correction ) .
typically only about @xmath121 of the signal survives , but the total continuum background is now only a factor of two larger than the signal for the golden point and @xmath122 gev . with only one neutrino , it is straightforward to construct a transverse mass for the entire system , which peaks very close to the pseudo - axion mass as expected , with only minor smearing due to detector effects .
we show examples for illustration in fig .
[ fig : mt ] for the golden point and @xmath101 tev , for a few values of @xmath11 . since @xmath123 , larger peak values of @xmath65 are for larger values of @xmath11 .
the golden point would produce a signal far above the background , trivially observable , at least in the region of @xmath124 gev where the @xmath14 can be resonant and the higgs boson has at least a modest br to @xmath125 .
the @xmath101 tev cases yield a signal of the same size as the background , also quite easy to observe but naturally requiring more statistics .
this analysis is generally powerful wherever there is a decent signal rate , due to the small background . using the same i d efficiencies and signal k - factor as in sec .
[ sub : bbll ] and a generic qcd k - factor of 1.3 for the background ( @xmath126 is not known at nlo ) , we summarize our results again in fig .
[ fig : lhc - lumi ] with points using an x symbol .
the left edge of each curve is cut off by values of @xmath11 for which the @xmath14 can not be resonant , so the rate is hopelessly small . as @xmath11 increases , @xmath112 increases and @xmath82 decreases , opening up more phase space for the decay , resulting in more events passing the cuts , which in turn means less luminosity required .
however , below @xmath127 threshold the br to @xmath125 begins to falls off steeply , resulting instead in an increase in required luminosity with increasing @xmath11 .
at some point , the @xmath128 analysis becomes more powerful , and the curves cross those with the diamond points .
system in @xmath129 resonant production at the lhc .
red and magenta curves are for the golden point and @xmath101 tev , respectively , and the dashed blue curve is the @xmath130 background . ]
at a future ilc , one would not have to rely on on - shell production to prove or place a limit on the existence of a @xmath0@xmath13@xmath14 coupling .
as long as it is present and @xmath131 pair production is kinematically allowed , the pseudo - axion could be seen .
[ fig : ilc_totcross ] shows the total cross section at an ilc for various @xmath132 for this channel at the golden point for three different values of @xmath133 gev , for which @xmath134 gev and @xmath135 gev , respectively .
the maximum cross section is of the order of 0.41.2 fb @xmath136 , as shown in the left panel of fig .
[ fig : ilc_totcross ] . since in the simplest little higgs
there is a destructive interference between the sm @xmath0 boson and the @xmath35 boson , the maximum total cross section in the full model goes down by roughly a quarter , as shown in the right panel .
production at an ilc for different values of @xmath137 : the golden point with @xmath138 gev ( solid ) , @xmath139 gev ( dashed ) , and @xmath140 gev ( dotted ) .
the left panel is without the presence of a @xmath35 resonance , while the right panel includes destructive @xmath0@xmath35 interference for @xmath141 gev.,title="fig : " ] production at an ilc for different values of @xmath137 : the golden point with @xmath138 gev ( solid ) , @xmath139 gev ( dashed ) , and @xmath140 gev ( dotted ) .
the left panel is without the presence of a @xmath35 resonance , while the right panel includes destructive @xmath0@xmath35 interference for @xmath141 gev.,title="fig : " ] model pseudo - axion with ( solid line ) and without ( dotted line ) the @xmath35 resonance at the golden point .
the dashed line is the sm prediction . on
the left is the @xmath142 final state for @xmath86 gev , while the right is the @xmath143 final state for @xmath144 gev.,title="fig : " ] model pseudo - axion with ( solid line ) and without ( dotted line ) the @xmath35 resonance at the golden point .
the dashed line is the sm prediction . on
the left is the @xmath142 final state for @xmath86 gev , while the right is the @xmath143 final state for @xmath144 gev.,title="fig : " ] production in the @xmath11model pseudo - axion with ( solid line ) and without ( dotted line ) the @xmath35 resonance at the golden point .
the dashed line is the sm prediction . on
the left is @xmath144 gev , right @xmath145 gev.,title="fig : " ] production in the @xmath11model pseudo - axion with ( solid line ) and without ( dotted line ) the @xmath35 resonance at the golden point .
the dashed line is the sm prediction . on
the left is @xmath144 gev , right @xmath145 gev.,title="fig : " ] depending on the mass of the pseudo - axion , the main decay channel for the @xmath14 varies from ( cf .
[ fig : par_space ] ) @xmath110 when fairly light , to two gluons ( jets ) for intermediate masses , and finally only @xmath113 for masses above that threshold , about @xmath146 gev at the golden point . for such heavy pseudo - axions
, one would look into the @xmath142 final state .
the cross section for this process is shown in the left panel of fig .
[ fig : finalzhh ] ( this and the other ilc simulations were performed by the bsm extensions of the event generator omega / whizard @xcite ) .
the dashed curve is the sm cross section , the dotted line shows the signal process without a @xmath35 resonance or with one above 1.5 tev , while the solid line is the prediction for @xmath147 tev . for the parameter values considered here
the higgs mass is @xmath148 gev , so that the six - fermion final state of interest consists of four @xmath83 jets and a lepton pair .
the sm @xmath149 process was extensively studied to analyze the triple higgs coupling measurability at an ilc @xcite .
it was found that the search limit for this process is severely statistics - limited and hence hard to analyze .
[ fig : finalzhh ] shows that even for the presence of a destructive interference with a @xmath35 boson , the cross section in the simplest little higgs is larger than in the sm by a factor of five , allowing for a reliable analysis of the final state .
the pseudo - axion shows up as a sharp spike in the @xmath150 spectrum , since even if the @xmath113 decay is kinematically allowed the width is of the order of a few hundred mev and hence far smaller than any conceivable detector resolution .
if the pseudo - axion mass lies between roughly @xmath151 and @xmath152 gev , the @xmath14 has a large branching fraction into a gluon pair , of the order of @xmath153 , which can even become dominant close to the @xmath113 threshold . for this parameter range
, the higgs boson decays completely to @xmath125 or @xmath154 , so the complete final states are @xmath155 , @xmath156 or six jets .
higgs bosons in an intermediate mass range could be reconstructed at an ilc @xcite fairly well , and although there would be combinatorial background in the @xmath157 channel , this would not degrade the pseudo - axion search .
the right panel of fig .
[ fig : finalzhh ] shows the cross section to @xmath143 , where a cut on the dijet invariant mass has been applied to reduce the sm higgsstrahlung background which is shown as a dashed line .
again , the dotted line is the prediction without a low - lying @xmath35 , while the solid line shows the destructive @xmath0@xmath35 interference . for the golden point with @xmath158 tev and @xmath105
, there is a critical point at @xmath159 gev where the pseudo - axion and higgs boson are degenerate ( at @xmath160 gev ) .
but vector boson - mediated production can not yield two identical particles ( under the assumption of cp conservation ) ; hence this must be either a scalar and a pseudoscalar or a scalar and a vector . from the normalization of the @xmath110 peak , which does not fit to the higgs boson
, one could deduce that on one side of the event there must be a degenerate pseudoscalar present .
the best case is where @xmath110 is the dominant @xmath14 decay mode , i.e. for pseudoscalar masses below 250 gev .
the sm background is mainly higgsstrahlung with @xmath161 decay , which can be distinguished from the signal by applying a cut to the @xmath110 invariant mass . for the parameter space under consideration
, the higgs decays to a vector boson pair which can be measured semi - leptonically .
[ fig : finalhbb ] shows the cross section for the production of a higgs boson in association with a @xmath110 pair , on the left for a pseudo - axion mass of 200 gev , on the right for an extremely light @xmath14 with @xmath162 gev . for the former case one can cut out all bottom quark pairs below 150 gev to get rid of the @xmath0 background ;
while this is not a very rigorous cut , it already reduces the sm background to a negligible level . to extract the light pseudoscalar
we adopt a window between 30 and 70 gev for @xmath163 .
again , fig .
[ fig : finalhbb ] shows as a dotted line the cross section without @xmath35 interference , and solid with the @xmath164 tev . in both cases
a few hundred @xmath165 should suffice to observe @xmath13@xmath14 production in that channel and thus establish the presence of a @xmath0@xmath13@xmath14 coupling .
we have identified an important distinction between product group and simple group little higgs models crucially involving the pseudo - axion corresponding to ungauged diagonal generators in the former case , and the pseudoscalar combination of anomalous @xmath8 diagonal generators in the latter . by consideration of quantum numbers alone , a coupling between the pseudo - axion , higgs boson and @xmath0 gauge boson is allowed . however , by construction such a coupling must vanish to all orders in product ( gauge ) group models , where the _ global _ little higgs symmetry group is simple . in simple group models
it is proportional to the difference between the vevs of the chiral field which occur in pairs . observing resonant production at the lhc
, either @xmath109 or @xmath106 , or drell - yan @xmath131 production in @xmath7 collisions , would conclusively rule out a product group realization of the little higgs mechanism were a candidate little higgs found at the lhc .
note , however , that this coupling would also vanish in simple group models with an imposed t - parity @xcite , as the discrete symmetry forces @xmath80 .
the phenomenology at lhc is in principle straightforward , breaking down into three classes , depending on whether the pseudo - axion or higgs boson is heavier , and whether the higgs boson decays primarily to a pair of bottom quarks or @xmath2 gauge bosons .
we find that observation of @xmath166 is in general possible only for a very small range in @xmath11 , where the higgs boson is light enough to decay to bottom quarks but is not yet ruled out by the lep results . for small values of @xmath5 , 12 tev , this could be achieved at lhc with a reasonable few hundred inverse femtobarn of luminosity , but for large values of @xmath5 , a few tev , the @xmath14 is more massive thus harder to produce , so only the luminosity - upgraded lhc ( slhc ) would have a chance of observation of this coupling .
( this would furthermore probably require different cuts , such as a less restrictive @xmath167 cut in the @xmath90 analysis , which likely would allow much more background from @xmath100 . ) for similar values of @xmath70 and @xmath71 , the coupling is too small .
if the higgs boson is heavier , one generally loses as it preferentially decays to two longitudinal gauge bosons rather than @xmath107 , so the br drops to negligible values .
if the pseudo - axion is resonant but the higgs boson decays to gauge bosons , a multi - lepton analysis works extremely well , especially for @xmath5 around 12 tev : even a few inverse femtobarn cover much of the possible parameter space .
however , we find that lhc has large gaps in coverage , for regions where neither the higgs boson nor the pseudo - axion can be resonant .
this is precisely where a linear collider would be crucial . at a future linear collider
the detection of the pseudo - axion would be quite easy using the @xmath0@xmath13@xmath14 coupling in scalar pair production .
the cross section for the whole parameter space is large enough to give several hundred pseudo - axions for reasonable luminosity , irrespective of any of @xmath109 or @xmath106 being on - shell .
this closes the holes in the lhc discovery range .
an important point is a destructive @xmath168 interference in that channel which is most severe for the lowest @xmath5 scales but does not jeopardize the signal . in the case of an on - shell decay @xmath109 the @xmath142 final state is enhanced compared to the sm by a factor of at least two for the region favored by electroweak precision data and up to an order of magnitude for large @xmath12 .
the same holds for the @xmath169 decay mode of the pseudo - axion where the enhancement over the sm is less pronounced .
but here the pseudo - axion is easily visible even for bad signal to background ratio as a sharp peak in the @xmath170 invariant mass spectrum , whose width is completely given by the detector resolution .
so even for mixing angles as small as @xmath171 the pseudo - axion coupling @xmath0@xmath13@xmath14 coupling can be detected at an ilc .
we thank uli baur for a critical reviews of the manuscript .
this research was supported in part by the national science foundation under grant no .
phy99 - 07949 , the u.s .
department of energy under grant no .
de - fg02 - 91er40685 , and by the helmholtz - gemeinschaft under grant no .
vh - ng-005 .
n. arkani - hamed , a. g. cohen , h. georgi , phys .
b * 513 * ( 2001 ) 232 ; + n. arkani - hamed , a. g. cohen , t. gregoire , and j. g. wacker , jhep * 0208 * ( 2002 ) 020 .
e. katz , j. y. lee , a. e. nelson and d. g. e. walker , arxiv : hep - ph/0312287 ; + m. piai , a. pierce and j. wacker , arxiv : hep - ph/0405242 .
d. e. kaplan , m. schmaltz and w. skiba , arxiv : hep - ph/0405257 .
a. birkedal , z. chacko and m. k. gaillard , arxiv : hep - ph/0404197 ; + p. h. chankowski , a. falkowski , s. pokorski and j. wagner , phys .
* b598 * ( 2004 ) 252 .
n. arkani - hamed , a. g. cohen , e. katz , and a. e. nelson , jhep * 0207 * ( 2002 ) 034 . n. arkani - hamed _ et al .
_ , jhep * 0208 * ( 2002 ) 021 .
d. e. kaplan and m. schmaltz , jhep * 0310 * ( 2003 ) 039 .
i. low , w. skiba , and d. smith , phys .
d * 66 * ( 2002 ) 072001 ; + s. chang and j. g. wacker , phys .
d * 69 * ( 2004 ) 035002 ; + t. gregoire , d. r. smith , and j. g. wacker , phys .
d * 69 * ( 2004 ) 115008 ; + w. skiba and j. terning , phys .
d * 68 * ( 2003 ) 075001 ; + s. chang , jhep * 0312 * ( 2003 ) 057 .
m. schmaltz , jhep * 0408 * ( 2004 ) 056 .
t. han , h. e. logan and l. t. wang , jhep * 0601 * , 099 ( 2006 ) .
s. kraml _ et al .
_ , _ cpnsh report : the hitchhikers guide to non - standard higgs physics _ , + hep - ph/0608079 . c. cski _ et al .
_ , phys .
d * 67 * ( 2003 ) 115002 ; phys .
d * 68 * ( 2003 ) 035009 .
j. l. hewett , f. j. petriello , and t. g. rizzo , jhep * 0310 * ( 2003 ) 062 .
w. kilian and j. reuter , phys .
d * 70 * ( 2004 ) 015004 .
s. c. park and j .- h .
song , arxiv : hep - ph/0306112 ; + m .- c .
chen and s. dawson , phys .
d * 70 * ( 2004 ) 015003 ; + r. casalbuoni , a. deandrea , and m. oertel , jhep * 0402 * ( 2004 ) 032 .
a. j. buras , a. poschenrieder , s. uhlig and w. a. bardeen , arxiv : hep - ph/0607189 .
j. boersma , arxiv : hep - ph/0608239 .
t. han , h. e. logan , b. mcelrath , and l .- t .
wang , phys .
d * 67 * ( 2003 ) 095004 ; + g. burdman , m. perelstein , and a. pierce , phys .
* 90 * ( 2003 ) 241802 + [ erratum - ibid . *
92 * ( 2004 ) 049903 ] ; + m. perelstein , m. e. peskin , and a. pierce , phys . rev .
d * 69 * ( 2004 ) 075002 ; + g. azuelos _ et al .
_ , eur .
j. c * 39s2 * , 13 ( 2005 ) ; + b. c. allanach _ et al .
_ , hep - ph/0602198 ; + k. cheung , c. s. kim , k. y. lee and j. song , hep - ph/0608259 .
w. kilian , d. rainwater and j. reuter , phys .
d * 71 * , 015008 ( 2005 ) ; hep - ph/0507081 . c. g. callan , s. r. coleman , j. wess and b. zumino , phys .
rev . * 177 * , 2247 ( 1969 ) .
see e.g. k. s. cranmer , arxiv : hep - ph/0512154 .
t. stelzer , f. long , comput .
phys . commun .
* 81 * ( 1994 ) 357 .
j. campbell , r. k. ellis and d. l. rainwater , phys .
d * 68 * , 094021 ( 2003 ) .
nason , s. dawson and r. k. ellis , nucl .
b * 303 * , 607 ( 1988 ) ; + r. bonciani , s. catani , m. l. mangano and p. nason , nucl .
b * 529 * , 424 ( 1998 ) ; + n. kidonakis , e. laenen , s. moch and r. vogt , phys .
d * 64 * , 114001 ( 2001 ) ; + m. beneke _ et al .
_ , arxiv : hep - ph/0003033 ( cf .
sec . 2 ) .
atlas tdr , report cern / lhcc/1999 - 15 ( 1999 ) .
r. v. harlander and w. b. kilgore , phys .
* 88 * , 201801 ( 2002 ) ; + r. v. harlander and w. b. kilgore , jhep * 0210 * , 017 ( 2002 ) ; c. anastasiou , k. melnikov and f. petriello , phys . rev .
lett . * 93 * , 262002 ( 2004 ) ; + c. anastasiou , k. melnikov and f. petriello , nucl .
b * 724 * , 197 ( 2005 ) .
m. l. mangano , m. moretti , f. piccinini , r. pittau and a. d. polosa , jhep * 0307 * , 001 ( 2003 ) .
v. d. barger , t. han , j. ohnemus and d. zeppenfeld , phys .
d * 41 * , 2782 ( 1990 ) .
j. a. aguilar - saavedra _ et al . _
[ ecfa / desy lc physics working group ] , `` tesla technical design report part iii : physics at an e+e- linear collider , '' hep - ph/0106315 ; + s. heinemeyer _ et al .
_ , hep - ph/0511332 ( and references therein ) .
a. djouadi , w. kilian , m. mhlleitner and p. m. zerwas , eur .
j. c * 10 * , 27 ( 1999 )
. t. ohl , _ omega : an optimizing matrix element generator _ , hep - ph/0011243 ; + m. moretti , t. ohl , j. reuter , hep - ph/0102195 .
w. kilian .
_ whizard manual _ , in * 2nd ecfa / desy study 1998 - 2001 * 1924 - 1980 , + lc - tool-2001 - 039 , jan 2001 ; w. kilian , t. ohl , and j. reuter , in preparation .
|
we propose a means to discriminate between the two basic variants of little higgs models , the product group and simple group models , at the next generation of colliders .
it relies on a special coupling of light pseudoscalar particles present in little higgs models , the pseudo - axions , to the @xmath0 and the higgs boson , which is present only in simple group models .
we discuss the collider phenomenology of the pseudo - axion in the presence of such a coupling at the lhc , where resonant production and decay of either the higgs or the pseudo - axion induced by that coupling can be observed for much of parameter space . the full allowed range of parameters , including regions where the observability is limited at the lhc , is covered by a future ilc , where double scalar production would be a golden channel to look for .
|
in the following , we shall investigate the linewidth narrowing effect of the @xmath0ls .
first of all , we study the single - photon reflective properties under the resonant control field ( @xmath53 ) . and our emphasis is still put on the change of linewidth with the coupling strength , as shown in fig.3 . here
the reflective spectrum appears a multiple peak - dip - peak structure ( eit - like reflective spectrum ) because of the interference between the two transitions .
this is different from a single peak structure in the two - level system .
but with the coupling strength decreasing , the two peaks is becoming declining and narrowing .
this phenomenon is also appear in the above section .
similarly , we can not obtain narrow reflective spectrum with the intensive amplitude only by the means of decreasing coupling strength in @xmath0ls .
then we adjust the frequency of the control field and study the change of the reflective spectrum under the different detuning @xmath27 .
as we all know , the weak coupling means that the decay rate to all other channels dominates over the decay rate of the excited atom to the waveguide modes .
the obtained reflective spectrum is very faint when we decrease the coupling strength because that most of the spontaneous emission of emitters is guided into the free space or non - radiative emission .
therefore , we should suppress the effects of spontaneous decay of the excited state and decrease the coupling strength at the same time .
as we have mentioned above , the transition between the excited state @xmath11 and the metastable state @xmath25 is driven by a classical pulses with the corresponding rabi frequencies signified by @xmath26 .
here we adopts the method proposed by law and eberly@xcite . by adjusting this interaction as far - detuned , we find two dissymmetrical peaks .
one of central frequency locates at @xmath5 , whose change is the same as that of the single peak in tls .
the other peak occurs at the double - photon resonance , which is the ultra - narrow reflective spectrum we need in this letter . and
in this case , the excited state of the emitter can be adiabatically eliminated .
it should be pointed out that the reflective spectrum of @xmath0ls has double dissymmetrical peaks and the tls has only one peak in the reflective spectrum .
here we only use the peak under the case of double - photon resonance to compare with the single peak in tls .
figure 4 shows the change of linewidth as the function of the ratio of @xmath19 . in fig.4
, blue dot line , black dash line and red solid line respectively denote the reflective spectrums of tls , @xmath0ls with @xmath53 , and @xmath0ls with @xmath57 . all other parameters are chosen as the same except for the detuning @xmath27 . from fig.4 , we can see that the reflective spectrum of @xmath0ls is more intensive than that of tls under the same coupling strength . for the @xmath0ls resonant with the control field
, the linewidth of the reflective spectrum has no significant change compared with that of tls .
but when we increase the detuning @xmath27 and keep other parameter as the same , the linewidth can be narrowed greatly and the intensity can be highly preserved . for tls and @xmath0ls . for tls , the parameters are set as follows : @xmath58 . for @xmath0ls ,
the detuning @xmath27 is set to be 0 for the red solid line and 5 for the black dash line respectively . and
all other parameters are set as follows : @xmath58 , @xmath54 , @xmath56.,width=302 ] to see the effect of the detuning on the linewidth narrowing more clearly . in the following
, we shall compare the reflective spectrum under the different detuning @xmath27 , as shown in fig.5 .
we still choose the other parameters as the same besides the detuning @xmath27 . here
black dash line , blue dot line and red solid line respectively denote the reflective spectrums of @xmath0ls with @xmath57 , @xmath0ls with @xmath59 , and @xmath0ls with @xmath60 .
we can see that the reflective spectrum is getting more and more narrowing with the increase of the detuning @xmath27 .
but the reflective amplitude is always maintained about 80% .
this property can be utilized to construct a tunable narrow band source . and the needed narrowed frequency and linewidth can both be obtained by choosing approximate parameters . for @xmath0ls with different @xmath27 .
the detuning @xmath27 is set to be 5 for the black dash line , 10 for the blue dot line , and 15 for the red solid line respectively . and
all other parameters are set as follows : @xmath58 , @xmath54 , @xmath56.,width=302 ] in conclusion , we have proposed a scheme to construct a narrow - band source via a one - dimensional waveguide coupling with a three - level @xmath0-type emitter . through rigorous theoretical calculation , we find that a ultra - narrow reflective spectrum with intense amplitude can be obtained for the system work in a weak coupling regime and under a far - detuned control field . it is shown that the central frequency of narrowed spectrum can be tunable by choosing approximate parameters . and
the system should work in the weak coupling regime to obtain the narrowed reflective spectrum , which has greatly decreased the requirement of the experiment .
this approach opens up the possibility of plasmonic ultranarrow band source .
* acknowledgments : * the work is supported by the the national natural science foundation of china(11347155 , 11274112 , 11204080 ) , the fundamental research funds for the central universities(wm1214019 ) , and china post - doctoral science foundations ( grant nos.2011m500739 ) .
d. e. chang , a. s. srensen , p. r. hemmer , and m. d. lukin,quantum optics with surface plasmons " , phys.rev.lett . * 97 * , 053002 ( 2006 ) .
d. e. chang , a. s. srensen , e. a. demler , and m. d. lukin,a single - photon transistor using nanoscale surface plasmons " , nat.phys.*3 * 807 ( 2007 ) . c. ropers , c. c. neacsu , t. elsaesser , m. albrecht , m. b. raschke , and c. lienaa,grating - coupling of surface plasmons onto metallic tips : a nanoconfined light source " , nano lett .
* 7 * , 2784 - 2788 ( 2007 ) .
d. e. chang , a. s. srensen , p .
r. hemmer , and m. d .lukin,strong coupling of single emitters to surface plasmons " , phys.rev.b * 76 * , 035420 ( 2007 ) .
a. v. akimov , a. mukherjee , c. l. yu , d. e. chang , a. s. zibrov , p. r. hemmer , h. park , and m. d. lukin,generation of single optical plasmons in metallic nanowires coupled to quantum dots " , nature(london ) * 450 * 402 - 406 ( 2007 ) .
g. y. chen , y. n. chen , and d. s. chuu , `` spontaneous emission of quantum dot excitons into surface plasmons in a nanowire , '' optics letter * 33 * , 2212 - 2214 ( 2008 ) .
z. j. yang , n. c. kim , j. b. li , m. t. chen , s. d. liu , z. h. hao , and q. q. wang , `` surface plasmons amplifications in single ag nanoring , '' optics express * 18 * , 4006 - 4011 ( 2010 ) .
d. dzsotjan , a. s. srensen , and m. fleischhauer , `` quantum emitters coupled to surface plasmons of a nanowire : a green s function approach , '' phys.rev.b * 82 * , 075427 ( 2010 ) .
w. chen , g. y. chen , and y. n. chen , `` coherent transport of nanowire surface plasmons coupled to quantum dots , '' optics express * 18 * , 10360 - 10368 ( 2010 ) .
n. c. kim , j. b. li , z. j. yang , z. h. hao , and q. q. wang , `` switching of a single propagating plasmon by two quantum dots system , '' app . phys.lett . * 97 * , 061110 ( 2010 ) .
w. chen , g. y. chen , and y. n. chen , `` controlling fano resonance of nanowire surface plasmons '' optics letter * 36 * , 3602 ( 2011 ) .
a. huck , s. kumar , a. shakoor , and u. l. andersen , `` controlled coupling of a single nitrogen - vacancy center to a silver nanowire , '' phys.rev.lett .
* 106 * , 096801 ( 2011 ) .
j. li and r. yu , `` single - plasmon scattering grating using nanowire surface plasmon coupled to nanodiamond nitrogen - vacancy center , '' optics express * 19 * , 20991 - 21002 ( 2011 ) .
z. jacob and v. m. shalaev,``plasmonics goes quantum , '' science , * 334 * ( 2011 ) .
d. witthaut and a. s. srensen , `` photon scattering by a three - level emitter in a one - dimensional waveguide , '' new journal of physics * 12 * , 043052 ( 2012 ) .
y. li , l. aolita , d. e. chang , and l. c. kwek , `` robust - fidelity atom - photon entangling gates in the weak - coupling regime , '' phys.rev.lett . * 109 * , 160504 ( 2012 ) .
j. yang , g.w .
lin , y.p .
niu , and s.q .
gong , `` quantum entangling gates using the strong coupling between two optical emitters and nanowire surface plasmons , '' optics express * 21 * , 15618 - 15626 ( 2013 ) .
x. f. zang , t. zhou , b. cai , and y. m. zhu , `` controlling single - photon transport properties in a waveguide coupled with two separated atoms '' journal of physics b * 46 * , 145504 ( 2013 ) .
m. d. lukin , m. fleischhauer , m. o. scully , and v. l. velichansky,``intracavity electromagnetically induced transparency,''optics letter * 23 * , 295 ( 1998 ) .
u. d. rapol and v. natarajan,``precise measurement of hyperfine intervals using avoided crossing of dressed states , '' europhys .
* 60 * , 195 ( 2002 ) .
d. das and v. natarajan , `` hyperfine spectroscopy on the 6p(3/2 ) state of cs-133 using coherent control , '' europhys .
lett . * 72 * , 740 ( 2005 ) .
a. krishna , k. pandey , a. wasan , and v. natarajan , `` high - resolution hyperfine spectroscopy of excited states using electromagnetically induced transparency , '' europhys .
* 72 * , 221 ( 2005 ) .
t. baba , `` slow light in photonic crystals , '' nat.photonics * 2 * , 465 ( 2008 ) .
j. t. shen and s. fan,``coherent photon transport from spontaneous emission in one - dimensional waveguide coupled to a two - level system,''optics letter * 30 * , 2001 - 2003(2005 ) .
d. witthaut , m. d. lukin and a. s. srensen , `` photon sorters and qnd detectors using single photon emitters , '' europhys . lett .
* 97 * , 50007 ( 2012 ) .
j. t. shen and s. fan , `` theory of single - photon transport in a single mode waveguide . .
coupling to a cavity containing a two - level atom , '' phys.rev.a 79 , 023837 ( 2009 ) .
c. k. law and j. h. eberly , `` arbitrary control of a quantum electromagnetic field , '' phys.rev.lett . *
76 * , 1055 ( 1996 ) .
c. k. law and j. h. eberly , `` synthesis of arbitrary superposition of zeeman states in an atom , '' optics express * 2 * , 368 - 371 ( 1998 ) .
|
the spectrum width can be narrowed to a certain degree by decreasing the coupling strength for the two - level emitter coupled to the propagating surface plasmon . but the width can not be narrowed any further because of the loss of the photon out of system by spontaneous emission from the emitter . here we propose a new scheme to construct a narrow - band source via a one - dimensional waveguide coupling with a three - level @xmath0-type emitter .
it is shown that the reflective spectrum width can be narrowed avoiding the impact of the loss .
this approach opens up the possibility of plasmonic ultranarrow single - photon source . _
introduction. _ in recent years , the coupling between one - dimensional waveguide and optical emitters has attracted considerable attentions because of their wide application in many aspects@xcite .
the two - level emitter within the waveguide acts as a perfect mirror for the light field at resonance@xcite .
the @xmath0-type three - level emitter system(@xmath0ls ) can realize electromagnetically induced transparency ( eit ) mechanism to control behavior of the probe photon by applying a classical control light beam in the system@xcite .
similarly , two separated two - level emitters system can also obtain eit - like transmission spectrum by adjusting the distance between the two emitters@xcite .
as one of important applications of eit , cavity - linewidth narrowing has shown great potential in many different areas such as laser stabilization@xcite , high - resolution spectroscopy@xcite , enhanced light matter interaction , and compressed optical energy@xcite . in this letter
, we shall investigate the linewidth narrowing effect in the system of optical emitter coupling the nanowire surface plasmons . as we have known , in the two - level emitter coupling with spps systems(tls ) , the linewidth of the reflection peak is proportional to @xmath1 , where @xmath2 is the coupling strength , and @xmath3 is the group velocity of the photons and can be simplified as the velocity of the light@xcite .
so the reflection spectrum can be narrowed in the weak coupling regime@xcite .
in such a case , the linewidth can be narrowed to a certain degree for tls .
however if the ratio @xmath4 is further decreased , the narrowed reflective spectrum shall be very weak because that most of the spontaneous emission of emitters is guided into the free space or non - radiative emission . in this letter , we propose a new scheme to construct a narrow - band source via a one - dimensional waveguide coupling with a three - level @xmath0-type emitter . by applying a far - detuned classical control light beam , the reflective spectrum demonstrates two dissymmetrical peaks .
one of the central frequency of the two peaks locates at @xmath5 . in this case ,
the which equivalent to the case that the two - level emitter within the waveguide acts as a perfect mirror for the propagating single surface plasmon at resonance .
so the change of reflective spectrum is the same as that of the single peak in tls .
the other peak occurs at the double - photon resonance , which is the ultra - narrow reflective spectrum we need in this letter . and
in this case , the excited state of the emitter can be adiabatically eliminated .
it should be pointed out that the spontaneous emission of emitters can either emit into the guided surface plasmons of the nanowire with rate @xmath6 or into all other possible channels with respective rates @xmath7@xcite . in general
, @xmath7 includes contributions both from emission into free space and non - radiative emission via ohmic losses in the metal wire .
now let us comment on the above two major contributions in detail .
the major source is the imperfect coupling of the emitter to the waveguide .
here we can define the ratio @xmath4 to quantify the coupling strength .
currently , the most advanced experimental setup employs a microwave transmission line side - coupled to a flux qubit . in this experiment ,
a photon incident from the side was back - scattered with 94% propability , translated into the current setting this corresponds to a coupling of @xmath8@xcite . in the optical regime
, a strongly increased decay rate was demonstrated for single quantum emitters coupled to s surface plasmon polariton mode on a silver nanowire@xcite . for a single diamond color center , the ratio @xmath9@xcite .
the second possible channel is the photon absorption in the waveguide , losses are particularly strong in plasmonic systems .
so a rapid in- and outcoupling to a dielectric waveguide is inevitabe@xcite .
here we need the system work in the weak coupling regime to obtain the narrowed reflective spectrum , which has greatly decreased the requirement of the experiment . in this letter , we show how to overcome this problem by using @xmath0ls to obtain extremely narrow linewidth with enough intensity .
and we find that the obvious advantage of this proposal is that the central frequency of narrowed spectrum can be tunable by choosing approximate parameters .
the paper is organized as follows . in section 2
, we shall revisit the problem of photon transportation of tls under a coherent architecture .
then we shall solve the reflective spectrum of @xmath0ls based on the theory of one - photon scattering .
the effect of large values of @xmath7 on the linewidth narrowing of tls is also studied . in section 3
, we shall propose the scheme to construct a narrow - band source based on the scheme of @xmath0ls , and provide the comparison between the tls and @xmath0ls on the effect of linewidth narrowing .
it shows that the @xmath0ls can overcome the disadvantage of small intensity in the tls .
finally , our main conclusions are summarized in section 4 . _
theoretical model _ in this section , we firstly review the reflective spectrum for the photon in a quasi - one - dimensional traveling electromagnetic modes coupled to one two - level emitter .
the configuration of the system is exhibited in fig.1(a ) , and the energy level scheme of the two - level emitter is shown in fig.1(b ) .
the states @xmath10 and @xmath11 have the energy @xmath12(the energy origin)and @xmath13 respectively .
the effect of the loss of the photon out of system is also involved .
here we focuses on the effect of the coupling strength on the linewidth narrowing .
for simplicity , we fix the parameter @xmath7 and vary the different parameter @xmath14 to quantify the coupling strength @xmath4 .
-type three level emitter considered in the present paper.,width=302 ] the scattering problem for a single photon in a one - dimensional waveguide coupled to a two - level quantum emitter has been solved exactly@xcite .
here we assume that the emitter is initially in a ground state and the photon is incident with the energy @xmath15 , then the reflection coefficient for a monochromatic input state with wave number @xmath16 can be written as @xmath17 the reflection amplitude is given by @xmath18 we now discuss the change of linewidth as a function of the ratio of @xmath19 , where @xmath20 , and @xmath21 denotes the central frequency of the reflective spectrum .
it should be pointed out that the maximum value of the reflective amplitude for tls occurs at the resonance , that is to say that the central frequency @xmath21 is zero . for comparison
, we fix the parameter @xmath7 as 0.1 , and the parameter @xmath14 is set as 0.1 , 0.5 , 1 and 5 respectively . the single - photon reflective spectrum in an optical waveguide coupled with an two - level emitter
is demonstrated to be a single peak , as shown in fig.2 . for tls .
the parameter @xmath14 is chosen to be 0.1 , 0.5 , 1 and 5 respectively .
the other system parameter @xmath7 is fixed as 0.1 .
, width=302 ] from fig.2 , we shows a typical reflective spectrum for various coupling strength of tls . at resonance and strong coupling ,
most of the energy of the plasmon is reflected .
it also exhibits that the linewidth of reflective spectrum is getting more and more narrowing with the decrease of the coupling strength . but
at the same time , the amplitude of reflective spectrum is also decreased .
this is because that most of the photon energy is injected into the free space or decay via ohmic losses in the metal wire when the coupling strength is small . and
for the case of @xmath22 , the maximum of the spectrum is only 0.25 .
obviously most of the photon energy has been lost .
therefore a new scheme to obtain narrowed spectrum with enough intensity is needed . in the following ,
we propose a new scheme to overcome the small intensity of narrowed reflective spectrum by using @xmath0ls .
the configuration of the system is shown in fig.1(a ) , and the energy level scheme of @xmath0-type three level emitter is shown in fig.1(c ) .
the nanowire behaves as a 1d continuum for the coherent transport of photons , and the propagating single surface plasmon is described by annihilation operator @xmath23 .
the states @xmath10 and @xmath11 have the energy @xmath12(the energy origin)and @xmath13 respectively . the propagating single surface plasmon couples to the transition @xmath24 with the strength @xmath14 .
the excited emitter s state @xmath11 is coupled to another level @xmath25 by a classical laser beam with rabi frequency @xmath26 and detuning @xmath27 . in the spirit of the quantum jump picture , spontaneous emission to other nodes out of the one - dimensional waveguide
is modeled by attributing an imaginary part @xmath28 to the energies of the excited level @xmath11 and an imaginary part @xmath29 to the energies of the metastable level @xmath25 in @xmath30 respectively .
we firstly give a general description for the single - photon scattering problem in one dimensional waveguide based on three - level system .
the dynamics in such a one - dimensional waveguide is modeled by the total hamiltonian @xmath31 all parts of eq.(3 ) can be respectively expressed as @xmath32 @xmath33 @xmath34 where @xmath35 and @xmath36 denote the emitter s raising and lowering operators , respectively . in eq.(4 ) , the parameter @xmath3 is the group velocity of the photons and can be simplified as the velocity of the light@xcite .
the transition between the ground state and the excited state of the emitter is coupled to the propagating surface plasmon modes with the coupling constant @xmath14 .
@xmath37 represents the electronic population operator of the excited states . throughout this paper , we assume that the emitter is initially in a ground state and the photon is incident with the energy @xmath15 , a linear and nondegenerate dispersion relation holds over the relevant frequency range , and a single plasmon is coming from the left with energy @xmath38 .
then the stationary state of the system , defined by @xmath39 , can be constructed in the form @xmath40\left\vert 0,g\right\rangle
\\ + e_{k}\left\vert 0,e\right\rangle+f_{k}\left\vert 0,f\right\rangle \text{,}\ ] ] where @xmath41(@xmath42 ) is the probability amplitude of the emitter in the excited state @xmath11(metastable state @xmath25 ) , @xmath43 denotes the vacuum state with zero plasmon and the emitter being unexcited , @xmath44 denotes the vacuum field and the emitter in the exited state .
@xmath45[@xmath46 is the wave function of a right - going(a left - going ) plasmon , and can be written as @xmath47 @xmath48 \text { , } \ ] ] where @xmath49 is the step function . by solving the eigenvalue equation @xmath50
, we finally get the reflection coefficient as the follows : @xmath51 the reflection amplitude for the @xmath0ls is given by @xmath52 .
|
this paper describes two ongoing observational programs at keck unified by a common goal : to investigate the metallicity , structure , and dynamics of m31 s extended stellar halo .
the first is a low - resolution spectroscopic survey of field red giant stars in m31 s outer halo ( @xmath4@xmath550 kpc ) , while the second aims to combine low- , intermediate- , and high - resolution spectroscopy of giants in dwarf spheroidal ( dsph ) companions of m31 located at @xmath6@xmath5300 kpc .
such dwarfs have long been thought of as basic building blocks in the process of galaxy formation .
the metallicity distribution of our target stars serves as as a fossil record of the formation of m31 s halo .
moreover , these stars are good tracers for investigating the internal dynamics of the dwarfs as well as the global dynamics of the m31 subgroup , and hence the dark matter distribution within the local group .
the radial abundance gradient and abundance spread in m31 s stellar halo can potentially discriminate between competing galaxy formation models .
the dissipational collapse model@xcite predicts a strong radial metallicity gradient , whereas the accretion model@xcite predicts no strong gradient .
it is important to study both globular cluster and field star populations since they may be dynamically distinct from each other .
the andromeda galaxy ( m31 ) provides a global external perspective of a large spiral much like our own , and yet is close enough for individual stars to be resolved . ref
. shows evidence for a weak metallicity gradient in a sample of 150 m31 globular clusters , with a mean iron abundance of @xmath7=-1.2 $ ] , which is slightly higher than the mean value of @xmath7=-1.4 $ ] for galactic globular clusters .
the metallicity of m31 field halo stars remains uncertain even after a decade or more of concerted effort using ground - based telescopes .
ref . found a metallicity of @xmath7\approx-0.6 $ ] for stars located at a projected distance of @xmath8 kpc along the minor axis and ref .
estimated @xmath7\approx-1.0 $ ] at @xmath9 kpc also on the minor axis .
a study of an @xmath10 kpc field around the globular cluster g219 suggested a large metallicity spread@xcite , with a mean value @xmath7\geq-1.0 $ ] , as did the studies described in ref . , ref .
( @xmath11 kpc ) , and ref . which targeted fields around five m31 globular clusters .
recent _ hubble space telescope _
( _ hst _ ) studies have used its excellent angular resolution to separate field stars from background galaxies and find the mean metallicity at various points in the m31 spheroid to be comparable to that of 47 tuc : @xmath7\sim-0.7$]@xcite .
the ground - based studies are hindered to one degree or another by sample contamination by distant background field galaxies many of which are compact enough to be mistaken for stars at faint magnitudes .
the _ hst _
studies , on the other hand , are limited by a small field of view and the lack of comparison field control samples .
studies of the inner halo risk possible contamination by m31 disk stars , while those of the sparse outer halo have to contend with a significant fraction of foreground contaminants , milky way dwarf stars at m31 s low galactic latitude . in general ,
sample contamination makes it difficult to draw conclusions about the average [ fe / h ] of m31 s stellar halo and especially its [ fe / h ] spread .
our study is based on deep @xmath1 images , obtained with the kpno 4-meter telescope , of a @xmath12 field at @xmath0 kpc ( @xmath13 ) on m31 s se minor axis .
this field is further out on the minor axis than the ones used in previous studies , probes at least as far out into the m31 spheroid as the others ( based on a 2:1 flattened spheroid ) , and is the least susceptible to contamination by the m31 disk .
observations of a well - matched comparison field , well away from m31 but at a comparable galactic latitude , serve as a control data set .
four high resolution @xmath14-band images obtained with keck , covering most of the field , provide morphological information .
the stars in these fields ( m31 giants and milky way dwarfs ) are outnumbered by the dense background of distant faint blue field galaxies , many of which are too compact to be distinguished from stars on the basis of image morphology alone ( even with keck s @xmath15@xmath16 seeing ) .
we use an alternative method developed in ref . for discriminating between stars and faint galaxies , one that uses broadband @xmath1 color information in addition to morphological information .
the technique relies on the fact that stars , even those spanning a wide range of age and metallicity , occupy a narrow , well - defined locus in multicolor space whereas galaxies display a broad distribution due to their mix of stellar populations and redshifts . for an object to be identified as a stellar candidate
, its intensity profile must be consistent with the psf _ and _ its @xmath17 , @xmath18 , and @xmath19 colors must be consistent ( within errors ) with the stellar locus defined by model isochrones@xcite and empirical data on galactic globular cluster giants .
this morphological and @xmath1 color - based screening is done on all objects detected in the m31 halo and comparison fields .
the color - magnitude diagram ( cmd ) of surviving objects in the comparison field is well described by a superposition of foreground galactic dwarf stars ( in keeping with a standard empirical model of the galaxy ) against a backdrop of residual contaminating faint blue field galaxies , while the m31 halo field contains a clear excess of faint red objects ( @xmath20@xmath523 , @[email protected] ) in addition to these two components .
the location of this population of faint red objects in the cmd is consistent with what one would expect for red giant stars at the distance of m31 ( fig . 1 ) .
we have carried out a statistical subtraction between the m31 halo field and comparison field cmds .
a detailed description of the selection process and results is given in ref . .
-2.25truecm of the @xmath1- and morphology - selected sample of objects in our @xmath0 kpc minor axis field , the brightest ones ( @xmath22 ) serve as excellent targets for follow - up spectroscopy .
spectra of 99 m31 halo red giant candidates have been obtained with the 10-meter keck telescope and low - resolution imaging spectrometer@xcite ( lris ) around the ca ii triplet ( @xmath238700 ) .
in addition , spectra have been obtained of 12 control objects which are known to be significantly brighter than the tip of m31 s red giant branch , and which are thus likely to be foreground galactic dwarf stars .
five multislit masks were observed for @xmath24 hr each during 4 nights in the fall 1996 and 1997 observing seasons , with @xmath25@xmath525 objects per mask .
the observations used a 1200 lines mm@xmath26 grating and @xmath27-wide slitlets , for a dispersion of 0.62 pixel@xmath26 and resolution of @xmath28 km s@xmath26 over the spectral range of 75009000@xmath29 .
the radial velocity of each target object is determined from its spectrum using standard cross - correlation techniques .
this kinematic information is useful for distinguishing among m31 halo giants , m31 disk giants , foreground galactic dwarfs , and distant background galaxies .
most of the foreground contamination in this low galactic latitude m31 field ( @xmath30 ) is expected to be due to dwarf stars in the milky way s disk with heliocentric velocities close to zero , significantly displaced from m31 s systemic velocity of @xmath31 km s@xmath26 .
m31 halo red giants are expected to have a large spread of radial velocities due to their random motion in the galaxy s potential , while m31 disk giants are expected to cluster tightly around its systemic velocity ( the line - of - sight component of the disk rotation velocity is expected to be zero as our field is on the minor axis ) .
in addition to the removal of residual contaminants , spectroscopy allows direct measurement of the metallicity of each star using the near infrared ca ii absorption line strengths , @xmath32 ( sum of the three equivalent widths ) .
there has been a lot of work to develop calibration relations based on observations of the line strengths of red giants in fiducial galactic star clusters with known metallicities@xcite .
such relations connect @xmath32 and the brightness of the star @xmath33 relative to the horizontal branch to [ fe / h ] , and have been calibrated over the range @xmath34\leq0.0 $ ] .
the line strength , @xmath32 , of each of our m31 halo red giant candidates is determined using an empirical linear relation between the relative strength of the cross correlation peak and the directly - measured , but somewhat noisier statistic @xmath32 for selected stars .
this derived @xmath32 is then used to calculate the metallicity of each object .
a detailed description is given in ref . .
the surface density of red giant candidates in the @xmath0 kpc m31 halo field , as derived from the statistically - subtracted , @xmath1 color- and morphologically - selected sample , is coupled with the _ hst _ studies in ref . .
the data indicate that m31 s stellar halo is much denser and/or larger than that of the galaxy :
@xmath35 , where @xmath36 is the ratio of the radial scale lengths of m31 and the galaxy and @xmath37 is the assumed power law index of the density profile@xcite .
the density of m31 s inner halo may be even higher than this estimate : while we have adopted a single power law profile for m31 s halo , the outer profile is known to steepen both from the counts in our @xmath0 kpc field and from ref . which finds that the profile is better fit by a de vaucouleurs law than a power law . a refined estimate of the halo density profile and flattening
are underway based on a spectroscopically - selected sample of an additional 50100 secure m31 halo red giants at various locations on the major and minor axes .
radial velocities and line strengths derived from this sample should enhance our knowledge of the dynamics and metallicity of m31 s halo , respectively ( see sec .
2.4.2 and 2.4.3 below ) .
the radial velocities measured from the @xmath38 keck spectra presented in this paper display a bimodal distribution with one component centered on m31 s systemic velocity ( @xmath39 km s@xmath26 ) and another component centered at @xmath40 km s@xmath26 ( fig . 2 ) .
we use an empirical star count model of the galaxy ( iasg model)@xcite to estimate the expected distribution of line - of - sight velocities of foreground dwarfs .
the spectroscopic control sample consists of 12 stars with @xmath41 , which is brighter than @xmath42 at the distance of m31 ( fig .
1 ) ; these stars are thus exclusively foreground dwarfs .
the iasg model matches their velocity distribution very well ( dotted histogram in fig . 2 ) .
a linear combination of the iasg model @xmath43 distribution and a gaussian centered on m31 s systemic velocity is fit to the data ; the best fit gaussian width is @xmath44 km s@xmath26 and the fraction of m31 halo stars in the sample of 99 targets is estimated to be @xmath45 .
note , with this normalization of the galactic dwarf @xmath43 distribution ( see fig .
2 ) , almost all the stars with @xmath46 km s@xmath26 are likely to be foreground dwarfs . the model results may be used to estimate the foreground contamination fraction for various velocity cuts in the sample e.g .
, @xmath47 km s@xmath26 corresponds to @xmath4850% contamination in all velocity bins , while @xmath49 km s@xmath26 corresponds to @xmath5025% contamination .
the different symbols in the cmd plot ( fig . 1 ) are based on these velocity ranges .
the lack of a significant concentration of objects at the exact systemic velocity indicates a negligibly small degree of contamination by m31 disk giants .
selecting stars with @xmath49 km s@xmath26 yields a secure sample of m31 halo red giant stars .
6.3truecm the secure sample of m31 halo red giant stars with @xmath49 km s@xmath26 is particularly useful for investigating the mean [ fe / h ] and spread .
the iron abundance of each of these stars is measured using two independent methods : the star s location in the @xmath51 vs. @xmath14 cmd and the strength of the ca ii absorption lines , @xmath32 .
the cmd - based photometric method compares each star to model isochrones calculated over a range of metallicities@xcite .
the line strength @xmath32 is calibrated to [ fe / h ] using two independent calibration relations@xcite , labeled `` zw '' and `` cg97 '' , respectively , following the designation in ref . .
the photometric estimate of [ fe / h ] is in reasonable agreement with the spectroscopic estimates within the calibrated range of each method , especially for the cg97 calibration ( fig .
measuring the abundance gradient versus radius in m31 s halo involves comparison of the mean [ fe / h ] in our data set to _ hst_-based photometric determinations of [ fe / h ] at @xmath8 and 11 kpc on the minor axis@xcite and at @xmath52 kpc on the major axis@xcite .
both of our @xmath0 kpc samples of halo red giants , the large @xmath1- and morphologically - selected statistically - subtracted sample and the smaller but cleaner spectroscopically - selected secure sample , yield @xmath53\rangle$ ] in the vicinity of @xmath54 to @xmath55 using various ways to estimate [ fe / h ] ( photometric and spectroscopic ) .
this mean abundance is roughly consistent with the results of the _ hst _ studies indicating that m31 lacks a strong radial gradient in metallicity .
however , the use of heterogeneous samples , different [ fe / h ] determination methods , and possible sample contamination ( m31 disk stars ) in the inner halo fields , makes this conclusion uncertain .
we are in the process of isolating clean samples of halo red giants at a variety of radii in m31 and determining [ fe / h ] in a uniform way with the help of keck lris spectra ( sec .
2.4.1 ) .
the secure sample of m31 halo red giants in the @xmath0 kpc field displays a large spread in [ fe / h ] ( @xmath56 dex ) , independent of the metallicity measurement method ( upper panels of fig .
the metallicity distribution among m31 field halo stars is similar to the distribution of mean [ fe / h ] values of m31 globular clusters and local group dwarf satellite galaxies ( fig . 4 ) .
the observed [ fe / h ] distribution in m31 s halo can be explained in terms of chemical enrichment models with steady gas loss from the system ( see details in ref . ) .
dwarf galaxies , the most numerous type of galaxy , are considered to be major building blocks for the formation and evolution of more massive galaxies@xcite . among dwarf galaxies ,
the least luminous , least massive morphological type known is the dwarf spheroidal subclass .
almost all known dsphs are found in close proximity to more massive galaxies , and are therefore particularly useful as tracers of the gravitational potential of the `` parent '' galaxy as well as for the study of environmental effects on dwarf galaxy evolution .
while the nine dsph companions of the milky way have been studied in great detail , similar studies of the m31 dsphs have only recently become feasible thanks to large - aperture 810-meter class telescopes and efficient modern - day spectrographs@xcite .
large area surveys in search of low surface brightness objects have led to the discovery of three new m31 dsph galaxies in the last couple of years@xcite , bringing the census to six dsph galaxies .
the present set of m31 dsph galaxies cover the same range of galactocentric distances as the milky way companions .
thus it should be possible to make a direct comparison of environmental effects on these two sets of dsphs .
we present initial results from a keck spectroscopic survey of five of the six m31 dsph companions , andi , andiii , andv , andvi , and andvii .
the first four have been studied with lris using a similar multislit ca ii triplet setup as for our m31 field halo star survey ( sec .
2.3 ) ; the fifth dsph has been observed with the high - resolution echelle spectrometer@xcite ( hires)see sec .
3.2.4 below . integrated stellar velocities of the five m31 dsph companions are listed in table 1 . in each case , this is based on the mean radial velocity of 1525 member giants .
the quoted uncertainty is the @xmath57 error of the mean , derived from the observed rms dispersion of velocities among member stars ( combination of measurement error and internal velocity dispersion ) ; it does _ not _ include possible systematic error in the measurements .
note , the signal - to - noise ratio of the andiii data is substantially lower than that of the other dsphs and this is due to truncation of the exposure time .
[ h ] ' '' '' name & @xmath58 ( km s@xmath26 ) + ' '' '' andi & @xmath59 + ' '' '' andiii & @xmath60 + ' '' '' andv & @xmath61 + ' '' '' andvi ( pegdsph ) & @xmath62 + ' '' '' andvii ( casdsph ) & @xmath63 + it is interesting to compare the velocity measurements in table 1 with a recent neutral hydrogen survey described in ref . .
the hi cloud tentatively detected in the vicinity of andiii has a radial velocity of @xmath64 km s@xmath26 in the radio 21 cm line , which is within @xmath57 of the optical velocity measurement .
if this low significance hi detection at andiii is confirmed , it likely represents gas that is physically associated with the dsph .
by contrast , the 21 cm velocity of the hvc 368 cloud in the direction of andv is @xmath65 and this is clearly inconsistent with the optical value . as
originally suspected@xcite , this cloud is likely to be a fragment of a larger neighboring gas cloud ( complex ` h ' ) that happens to be superposed on andv .
figure 5 shows cmds of four of the m31 dsphs , the ones we have targeted for lris spectroscopy ( targets marked by special symbols ) .
the slope and width of the red giant branch traced by member stars are indicative of the mean metallicity and metallicity spread of the dsph , respectively .
for example , the andv red giant branch is significantly closer to vertical than andi s or andvi s indicating a lower metallicity .
this is confirmed by the spectroscopic [ fe / h ] estimates@xcite for these galaxies ( see fig .
2.5truecm perhaps an even more striking aspect of fig .
5 is the variation of the _ width _ of the red giant branch from one dsph to another .
it is convenient to compare andv to andvi since their cmds are based on photometric data of near - identical quality : short @xmath66 and @xmath14 band exposures with lris in @xmath67@xmath16 seeing conditions@xcite . while the andv member giants define a very narrow track , the andvi members display a substantial spread in color that is significantly in excess of the photometric error ; this is most easily explained in terms of a metallicity spread in andvi . the spectroscopically - determined [ fe / h ] histograms ( fig
. 6 ) support this hypothesis : despite the fact that it has the highest signal - to - noise spectra of the four , the andvi histogram shows a prominent tail towards high metallicity and a spread in excess of 1.5 dex . similarly , the andi red giant branch is broader than that of andiii ( fig . 5 ) , perhaps due to an intrinsic metallicity spread in the former galaxy .
the large spread in spectroscopic [ fe / h ] values observed for andiii ( fig .
6 ) is likely a result of measurement error caused by the low signal - to - noise ratio of its spectra .
dwarf companions can be treated as test particles in a large galaxy s gravitational potential .
ref . presents the most up - to - date and thorough dynamical analysis of the m31 halo using available radial velocity and distance data for several classes of tracers : gas and stars in the rotating disk , globular clusters , planetary nebulae , and dwarf satellites .
the recent discovery of distant dsphs around m31 , some almost 300 kpc from the parent galaxy , has breathed new life into this subject .
the most distant tracers provide the best constraints on the total mass of the halo .
an update of the m31 dynamical analysis , including radial velocity data for all of the dsphs presented in this paper@xcite , yields the best total mass estimate to date : @xmath68 .
surprisingly , this m31 value is somewhat lower than the best estimate for the total mass of the milky way halo .
however , the large uncertainty associated with each determination means that one can not rule out the possibility that the two galaxies have the same mass .
the hires observations of andvii red giants presented in this paper were carried out in an unusual manner .
special purpose multislit masks were designed and fabricated to enable simultaneous observations of multiple red giants .
each mask contains 56 slitlets , with lengths in the range @xmath27 to @xmath69 , so that the sum total of all slitlet lengths is @xmath70 in order to prevent overlap of adjacent echelle orders .
needless to say , there is not much blank `` sky '' area available on the typical slitlet . instead of the usual method of sky subtraction
, we synthesize a single sky spectrum for each multislit mask by coadding blank pixels from all the slitlets on that mask .
the experiment was made doubly difficult by the fact that our typical target star is very faint by hires standards : @xmath7123 .
nevertheless , it is possible to reliably extract velocities for most of the 21 stars ( fig .
the observed rms dispersion is @xmath72 km s@xmath26 , quadrature sum of the dsph s internal velocity dispersion and the velocity measurement error .
these data are being used to probe the dynamics and mass - to - light ratio of andvii@xcite .
* a new color - based screening technique , using deep @xmath1 photometry obtained with the kpno 4-meter telescope , is combined with traditional morphological star - galaxy separation based on high resolution @xmath14-band images obtained with the keck 10-meter / lris .
this enables us to isolate m31 halo red giant stars in a field located at a projected distance of 19 kpc from the galaxy s center along the minor axis . after screening for stars
, this field displays a clear excess population relative to a well - matched comparison field , plausibly red giants at the distance of m31 . * these data , taken together with _
hubble space telescope _ studies of the inner halo , suggest that the density of m31 s stellar halo is about @xmath73 higher than that of the milky way halo at comparable radii . *
follow - up keck lris spectra are presented for 99 m31 halo red giant candidates and a control sample of 12 foreground galactic dwarf stars .
kinematical information derived from the spectra allow elimination of residual contaminants : foreground galactic dwarfs , m31 disk giants , background galaxies .
* metallicity measurements are made on a star - by - star basis using two independent methods : the strength of the ca ii absorption lines and the location in the @xmath51 vs @xmath14 color - magnitude diagram .
various calibration schemes are tried in deriving spectroscopic [ fe / h ] estimates .
the photometric and spectroscopic [ fe / h ] estimates are in reasonable agreement with each other . *
the mean metallicity of m31 field halo giants , @xmath53\rangle$ ] , is in the range @xmath54 to @xmath55 depending on the metallicity measurement method , in rough agreement with the mean value of the milky way halo .
there does not appear to be a strong radial abundance gradient in m31 s halo : the mean [ fe / h ] in this @xmath0 kpc halo field is similar to ( possibly slightly lower than ) _
findings for the inner parts of the halo ( @xmath811kpc ) .
* there is a spread of at least 2 dex in [ fe / h ] even for the spectroscopically - selected sample of secure m31 halo red giants , independent of the method used to measure [ fe / h ] .
this is comparable to the [ fe / h ] spread among m31 globular clusters and local group dwarf satellite galaxies . *
the velocity distribution in the @xmath0 kpc minor axis halo field is well fit by an equal mix of foreground dwarfs ( drawn from a standard galactic model ) and giants in m31 s halo represented by a gaussian of width @xmath74 km s@xmath26 centered on its systemic velocity ( @xmath75 km s@xmath26 ) .
* we present spectra of about 100 red giant candidates in five m31 dsph companions : andi , andiii , andv , andvi ( pegdsph ) , and andvii ( casdsph ) . the first four have been observed with lris and the last one with hires .
this fall , we plan to carry out observations with esi , the new , intermediate - resolution , high - throughput spectrograph on keck . * the mean stellar radial velocities of the dsphs
are compared to recent velocity measurements of neutral hydrogen clouds seen in the direction of some of these dwarfs@xcite .
the stellar and hi velocities are in good agreement in the case of andiii , indicating a physical association , but not in the case of andv ( probably a chance superposition ) .
* color - magnitude diagrams of spectroscopically - confirmed dsph member giants suggests the presence of a significant metallicity spread in andvi and possibly in andi , but little or no spread in andiii and andv .
spectroscopic estimates of the metallicity distribution in these four dsph galaxies , based on the strength of the ca ii triplet in member red giant stars , appear to be consistent with this finding . *
the andvii observations are based on a novel multislit ( 56 stars at a time ) , echelle spectroscopy technique using hires .
accurate radial velocity measurements have been obtained for a total of 1520 member giants .
the observed rms velocity dispersion is about @xmath76 km s@xmath26 ( uncorrected for measurement error ) . * using our new radial velocity measurements of these outer dsph satellites in combination with other dynamical tracers , the total mass of m31
s extended halo is estimated to be a little under @xmath3 , comparable to or smaller than that of the milky way halo .
p.g . would like to thank his collaborators on various ongoing projects from which the data and results in this paper are drawn : steve majewski , jamie ostheimer , drew phillips , mark wilkinson , and especially wyn evans , linda pittroff , and steve vogt .
he is grateful to the organizers of the spie meeting in munich for their generous hospitality . e.k.g . acknowledges support by nasa through grant hf-01108.01 - 98a from the space telescope science institute .
j. couture , r. racine , w. e. harris , and s. holland , `` stellar populations in m31 : resolved v , i photometry of five halo globular clusters and surrounding fields , '' _ astron .
j. _ * 109 * , pp . 20502060 , 1995 .
e. k. grebel , `` evolutionary histories of dwarf galaxies in the local group , '' in _ the stellar content of the local group _ , p. whitelock and r. cannon , ed .
192 _ , pp . 1738 , asp , san francisco , 1999 .
p. guhathakurta , e. k. grebel , and l. c. pittroff , `` keck spectroscopy of red giants in m31 dwarf spheroidal galaxies .
metallicity distribution of andi , andiii , andv , and andvi , '' _ astron .
j. _ in preparation , 2000 .
p. guhathakurta , e. k. grebel , l. c. pittroff , d. b. reitzel , a. c. phillips , s. s. vogt , j. c. ostheimer , and s. r. majewski , `` keck spectroscopy of red giants in m31 dwarf spheroidal galaxies .
i. integrated velocities from a survey of andi , andiii , andv , andvi , and andvii , '' _ astron . j. _ submitted , 2000 .
p. guhathakurta , e. k. grebel , s. s. vogt , n. w. evans , and m. i. wilkinson , `` keck spectroscopy of red giants in m31 dwarf spheroidal galaxies .
ii . internal dynamics of andvii from multislit hires observations , '' _ astron .
j. _ in preparation , 2000 .
j. b. oke , j. g. cohen , m. carr , j. cromer , a. dingizian , f. h. harris , s. labrecque , r. lucinio , w. schall , h. epps , and j. miller , `` the keck low - resolution imaging spectrometer , '' _ pub .
pacific _ * 107 * , pp .
375385 , 1995 .
e. w. olszewski , r. a. schommer , n. b. suntzeff , and h. c. harris , `` spectroscopy of giants in lmc clusters .
i - velocities , abundances , and the age - metallicity relation , '' _ astron .
j. _ * 101 * , pp .
515537 , 1991 .
g. a. rutledge , j. e. hesser , p. b. stetson , m. mateo , l. simard , m. bolte , e. friel , and y. copin , `` galactic globular cluster metallicity scale from the ca ii triplet i. catalog , '' _ pub .
pacific _ * 109 * , pp . 883906 , 1997 .
g. a. rutledge , j. e. hesser , and p. b. stetson , `` galactic globular cluster metallicity scale from the ca ii triplet ii .
rankings , comparisons , and puzzles , '' _ pub .
pacific _ * 109 * , pp . 907919 , 1997 .
r. zinn , and m. j. west , `` the globular cluster system of the galaxy .
iii - measurements of radial velocity and metallicity for 60 clusters and a compilation of metallicities for 121 clusters , '' _ astrophys .
_ * 55 * , pp .
4566 , 1984 .
|
we present keck 10-meter / lris spectra of candidate red giants in the halo of m31 , located at a projected radius of @xmath0 kpc on the minor axis .
these spectroscopic targets have been selected using a combination of @xmath1-based and morphological screening to eliminate background galaxies .
radial velocity measurements are used to separate m31 halo giants from foreground milky way dwarf stars , m31 disk stars , and residual background galaxies .
the metallicity of each m31 halo giant is measured using standard photometric and spectroscopic techniques , the latter based on the strength of the ca ii triplet .
the various [ fe / h ] estimates are in rough agreement with one another .
the data reveal a large spread ( @xmath2 dex ) in [ fe / h ] in m31 s halo ; there is no strong radial [ fe / h ] gradient .
lris and hires spectra are also presented for red giants in five dwarf spheroidal satellites of m31 : andi , andiii , andv , andvi , and andvii . there appears to be a significant metallicity spread in andvi and possibly in andi .
the new radial velocity data on these outer dwarfs are used to constrain the total mass of m31 : the best estimate is under @xmath3 , somewhat less than the best estimate for the milky way .
|
let @xmath6 be a simple graph with vertices @xmath7 and edges @xmath8 , and let @xmath9 denote the positive integers .
a _ ( proper ) coloring _ of @xmath6 is a function @xmath10 such that @xmath11 whenever the vertices @xmath12 are adjacent .
stanley ( @xcite ; see also @xcite ) defined the _ chromatic symmetric function _ of @xmath6 as @xmath13
the sum over all colorings @xmath14 , where @xmath15 are countably infinitely many commuting indeterminates .
this definition is invariant under permutations of @xmath16 , so @xmath17 is a symmetric function , homogeneous of degree @xmath18 .
the chromatic symmetric function is a much stronger isomorphism invariant than the well - known _
chromatic polynomial _
@xmath19 , a polynomial function of @xmath20 that gives the number of colorings of @xmath6 using at most @xmath20 colors .
indeed , for any integer @xmath20 , the number @xmath19 can be obtained from @xmath17 by setting @xmath21 and @xmath22 for all @xmath23 .
it is natural to ask whether @xmath17 is a complete isomorphism invariant ; that is , whether two non - isomorphic graphs must have different chromatic symmetric functions .
the answer is no ; the smallest example , shown in figure [ samexg - figure ] , was given by stanley in @xcite .
brylawski @xcite introduced a graph invariant called the _ polychromate _ and constructed an infinite family of pairs of nonisomorphic graphs with the same polychromate .
sarmiento @xcite proved that the polychromate is equivalent to the _
@xmath24-polynomial _ studied by noble and welsh @xcite , a stronger invariant than @xmath25
; therefore , each pair in brylawski s construction shares the same chromatic symmetric functions .
( we thank anna de mier for directing our attention to these results . ) in another direction , the non - commutative version of the chromatic symmetric function , studied by gebhard and sagan @xcite , is easily seen to be a complete invariant .
stanley s question of whether @xmath17 is a complete isomorphism invariant for _ trees _ remains open .
this is equivalent to the problem of whether a tree is determined by its @xmath24-polynomial , since the formula for @xmath17 in terms of the @xmath24-polynomial ( * ? ? ?
* theorem 6.1 ) is easily seen to be reversible for trees .
stanley s question was answered in the affirmative for certain special kinds of trees by fougere @xcite and the second author @xcite , both of whom listed several other tree invariants that can be extracted from the chromatic symmetric function .
additionally , tan @xcite has verified computationally that the answer is `` yes '' for trees with 23 or fewer vertices .
( in contrast , the chromatic polynomial is nearly useless for distinguishing trees , because @xmath26 for every tree @xmath0 with @xmath27 vertices . )
our main tool is stanley s expansion of the chromatic symmetric function in the basis of power - sum symmetric functions @xmath28 ( * ? ? ?
* theorem 2.5 ) ; see equation below .
when @xmath0 is a ( possibly trivial ) tree , the coefficient @xmath29 of @xmath28 in @xmath1 has a particularly simple combinatorial interpretation . for @xmath30 , define the _ type _ of @xmath31 to be the partition whose parts are the sizes of the vertex sets of the graph with vertices @xmath32 and edges @xmath31 ( see figure [ type - figure ] for an example ) .
then , up to a sign , @xmath29 is the number of edge sets @xmath31 of type @xmath33 .
as we will see , many other invariants of @xmath0 can be recovered from @xmath1 .
recall that the _ degree _ of a vertex is the number of edges incident to it .
a _ leaf _ of a tree is a vertex of degree 1 , and the unique incident edge is called a _
leaf edge_. we define the _ subtree polynomial _ of @xmath0 by @xmath34 where the sum runs over all subtrees @xmath35 of @xmath0 with at least one edge , and @xmath36 denotes the set of leaf edges of @xmath35 . setting @xmath37 and @xmath38 in @xmath2 recovers the polynomial @xmath39 studied by chaudhary and gordon in ( * ? ? ?
* section 3 ) .
conversely , @xmath40 , so the two polynomials provide identical information about @xmath0 . for every non - empty set @xmath30 , there is a unique minimal subset @xmath41 , called the _ connector _ of @xmath31 , such that @xmath42 is a tree .
( so @xmath43 if and only if @xmath31 is itself a subtree of @xmath0 . )
the _ connector polynomial _ of @xmath0 is then defined as @xmath44 the polynomials @xmath2 and @xmath45 provide equivalent information about @xmath0 ; we will prove in proposition [ subt - conn - equivalence ] below that each of these invariants can be obtained from the other .
moreover , the path sequence and degree sequence of @xmath0 can easily be recovered from @xmath46 , as observed by chaudhary and gordon ( * ? ? ?
* proposition 18 )
. for a partition @xmath47 , and integers @xmath48 , define @xmath49 we can now state our main theorem , which asserts that the subtree and connector polynomials can be recovered from the chromatic symmetric function @xmath1 .
[ main - thm ] for every @xmath50 , and for every tree @xmath0 with @xmath27 vertices , @xmath51 and @xmath52 it follows that the chromatic symmetric function is at least as strong an invariant as the subtree and connector polynomials .
in particular , the path and degree sequences of @xmath0 can be recovered from @xmath1 , as announced previously in @xcite ; this generalizes a earlier result of fougere ( * ? ? ?
* theorem 3.3.1 ) .
section [ main - thm - section ] contains the proof of theorem [ main - thm ] , as well as explicit formulas for the path and degree sequences , and a reinterpretation of and in terms of the usual scalar product on the space of symmetric functions .
theorem [ main - thm ] implies that @xmath1 is a stronger invariant than @xmath2 .
in fact , it is _ strictly _ stronger : the two trees shown in figure [ counterexample - figure ] have different chromatic symmetric functions , but the same subtree polynomial .
( eisenstat and gordon @xcite constructed an infinite family of pairs of non - isomorphic trees with the same subtree polynomials , of which figure [ counterexample - figure ] is the smallest example . )
thus stanley s question remains open . as another application of the combinatorial interpretation of the coefficients @xmath29 , we identify some classes of trees for which the chromatic symmetric function is in fact a complete invariant .
these trees include all _ spiders _ ( trees having exactly one vertex of degree @xmath53 ) and some _ caterpillars _ ( trees from which deleting all leaves yields a path , such as those in figure [ counterexample - figure ] ) .
we prove in section [ spider - section ] that every spider can be reconstructed from its subtree polynomial , hence from its chromatic symmetric function ( generalizing results of fougere @xcite ) .
the corresponding problem for caterpillars is more difficult ; however , certain special kinds of caterpillars can indeed be reconstructed from their chromatic symmetric functions , and the methods we use to prove this may be extendible to all caterpillars .
a _ unicyclic graph _ is a graph with one cycle .
connected unicyclic graphs can be recognized as such from their chromatic symmetric functions . while the combinatorial data provided by stanley s expansion is not as fine for unicyclic graphs
as it is for trees , we can still obtain some uniqueness results for special unicyclic graphs by mimicking our results for spiders and caterpillars .
in particular , we show in section [ squid - section ] that no two _ squids _ ( unicyclic graphs with at most one vertex of degree two or more ) can have the same chromatic symmetric function , although it is not clear whether membership in the class of squids can be determined from @xmath17 .
an analogous result holds for _ crabs _ ( unicyclic graphs in which every vertex not on the cycle is a leaf ) satisfying an additional technical condition .
our collaboration began at the graduate student combinatorics conference held at the university of minnesota on april 16 and 17 , 2005 .
we thank the organizers of the conference for their efforts , and we thank franois bergeron , tom enkosky , gary gordon , brandon humpert , rosa orellana , victor reiner , bruce sagan , and stephanie van willigenburg for many helpful conversations .
the database of trees constructed by piec , malarz , and kulakowski @xcite , and john stembridge s freely available maple package sf , were invaluable for calculating examples and formulating ( and checking small cases of ) conjectures [ positivity - conjecture ] and [ integrality - conjecture ] .
finally , we thank two anonymous referees for their suggestions , and in particular for making us aware of fougere s undergraduate thesis @xcite .
we assume that the reader is familiar with basic facts about graphs and trees ( see , e.g. , ( * ? ? ?
* chapter i ) ) .
we denote a graph @xmath6 by an ordered pair @xmath54 , where @xmath55 is the set of vertices and @xmath56 is the set of edges .
all our graphs are _
simple _ ; that is , we forbid loops and parallel edges .
the _ order _ of a graph is its number of vertices .
a _ tree _ is a graph @xmath6 which is acyclic and connected and for which @xmath57 ; any two of these conditions together imply the third .
we consider the graph with one vertex and no edges to be a tree , the _
trivial tree _ ;
unless otherwise specified , all our statements about trees include this possibility .
a _ leaf _ of a tree is a vertex of degree 1 , that is , with exactly one incident edge .
every nontrivial tree has at least two leaves @xcite .
it is often notationally convenient to ignore the distinction between a graph and its edge set .
we now review some facts about symmetric functions ( for which the standard references are @xcite and ( * ? ? ?
* chapter 7 ) ) and about the chromatic symmetric function ( introduced by stanley in @xcite ) .
a _ partition _ is a sequence @xmath58 of positive integers in weakly decreasing order .
the numbers @xmath59 are called the _ parts _ of @xmath33 .
we say that @xmath33 is a partition of @xmath27 , written @xmath60 , if @xmath61 . the number @xmath62 is called the _ length _ of @xmath33 .
let @xmath15 be a countably infinite set of commuting indeterminates .
for @xmath63 , the _
@xmath64 power - sum symmetric function _ is @xmath65 and for a partition @xmath33 we define @xmath66 it is well known that @xmath67 is a basis for the @xmath68-vector space @xmath69 consisting of all symmetric functions that are homogeneous of degree @xmath27 .
let @xmath6 be a graph of order @xmath27 .
stanley ( * ? ? ?
2.5 ) proved that @xmath70 where @xmath71 is the partition whose parts are the orders of the connected components of the subgraph of @xmath6 induced by @xmath31 ( see figure [ type - figure ] for an example ) .
note that @xmath71 depends upon @xmath31 and @xmath7 , but not on @xmath8 .
we write @xmath72 for the coefficient of @xmath28 in the expansion ; that is , @xmath73 we will abbreviate @xmath74 when no confusion can arise .
the chromatic symmetric function of a graph provides much more combinatorial information when that graph is a tree . in general
, the coefficient @xmath75 does not count edge sets of type @xmath33 , because @xmath76 is not constant for all such sets @xmath31 . on the other hand ,
if @xmath77 is a tree of order @xmath27 , then every @xmath78 is acyclic , so its induced subgraph has @xmath79 connected components .
hence @xmath80 , and we obtain a useful combinatorial interpretation for the numbers @xmath75 : @xmath81 the invariants @xmath75 are far from independent ; in particular , implies that @xmath82 we next list some basic invariants of graphs that can be recovered from its chromatic symmetric function .
several of these facts were previously observed by the second author in @xcite . for notational simplicity
, we shall often omit the parentheses and singleton parts of a partition , for instance , writing @xmath83 rather than @xmath84 .
[ basic - facts ] for every graph @xmath85 , * the symmetric function @xmath17 is homogeneous of degree @xmath86 ; * @xmath87 ; and * the number @xmath20 of connected components of @xmath6 is @xmath88 .
if @xmath6 is a tree , then in addition * for @xmath89 , @xmath90 is the number of subtrees of @xmath6 with @xmath91 vertices ; and * @xmath92 is the number of leaves of @xmath6 .
assertion ( i ) is immediate from the definition of @xmath17 , and ( ii ) follows easily from . for ( iii ) , implies that @xmath93 whenever @xmath94 . on the other hand , @xmath95 , the sum over all @xmath96 with @xmath97 , where @xmath98 is the tutte polynomial of @xmath6 ( see ( * ? ? ?
x ) ) . up to sign
, this formula counts acyclic orientations of @xmath6 with exactly one sink in each component ( this follows from ( * ? ? ?
8 , p. 348 ) ) .
in particular , @xmath99 .
assertion ( iv ) holds because @xmath96 is the set of edges of a @xmath91-vertex subtree if and only if @xmath100 , and ( v ) follows because every subtree of order @xmath101 is of the form @xmath102 , where @xmath103 is a leaf . by ( i ) , ( ii ) and ( iii ) of proposition [ basic - facts ] , trees can be distinguished from non - trees by their chromatic symmetric functions .
moreover , part ( v ) implies that paths ( trees with exactly two leaves ) and stars ( trees with exactly one non - leaf ) are determined up to isomorphism by their chromatic symmetric functions .
the _ girth _ of a graph @xmath6 is defined as the length of the smallest cycle in @xmath6 , or @xmath104 if @xmath6 is acyclic . with a little more work
, we can compute the girth of @xmath6 from @xmath17 .
the idea is to find the smallest edge set for which fails .
[ girth - prop ] let @xmath85 be a graph with @xmath27 vertices and @xmath105 edges .
let @xmath20 be the largest number such that @xmath106 then the girth of @xmath6 is @xmath107 .
let @xmath108 be the girth of @xmath6 .
suppose first that @xmath109 . then @xmath110 , so every subset @xmath78 with @xmath111 edges is acyclic and hence has @xmath20 connected components . on the other hand , if @xmath112 , then the maximum size of a component of @xmath31 is @xmath113 , so @xmath31 must be acyclic and hence must have @xmath111 edges . therefore @xmath114 now suppose that @xmath115 .
we claim that @xmath78 has @xmath20 components if and only if it either has @xmath111 edges ( hence is acyclic ) or is precisely a cycle of length @xmath108 .
the `` if '' direction is evident . for
the `` only if '' direction , suppose that @xmath31 has @xmath20 components and is not acyclic , hence contains a cycle @xmath116 . by definition of @xmath6 , the length of @xmath116 can not be less than @xmath108 ; on the other hand , there are at least @xmath117 vertices that do not belong to @xmath116 ( one for each other component of @xmath31 ) , so @xmath118 .
thus @xmath116 has length exactly @xmath108 .
moreover , @xmath119 can not contain any other edge with an endpoint outside @xmath116 ( because then it would have fewer than @xmath20 components ) or an edge joining two vertices of @xmath116 ( because then @xmath6 would contain a cycle of length @xmath120 ) .
hence @xmath121 as desired .
denoting by @xmath122 the set of @xmath108-cycles of @xmath6 , we have @xmath123 as desired .
theorem [ main - thm ] expresses the subtree polynomial @xmath2 and connector polynomial @xmath45 of a tree @xmath0 in terms of the chromatic symmetric function @xmath1 . the first step is to show that @xmath2 and @xmath45 are interchangeable . in what follows , we will often abuse notation by ignoring the distinction between a tree @xmath77 and its edge set @xmath124 .
[ subt - conn - equivalence ] let @xmath0 be a tree .
then : 1 . @xmath125 .
2 . @xmath126 . for each nontrivial subtree @xmath127 , write @xmath36 for the set of leaf edges of @xmath35 .
note that @xmath128 , with equality if and only if @xmath35 consists of a single edge .
moreover , observe that @xmath129 if and only if @xmath130 . hence @xmath131 giving the first equality .
meanwhile , solving the equations @xmath132 for @xmath133 and @xmath134 yields @xmath135 , @xmath136 , giving the second equality .
we now prove the main theorem .
to do so , we establish a formula for the connector polynomial of a tree in terms of its chromatic symmetric function , then apply proposition [ subt - conn - equivalence ] to obtain a formula for the subtree polynomial . by definition ,
the coefficient of @xmath137 in @xmath138 is @xmath139 because the innermost sum vanishes unless @xmath140 , when it is @xmath141 . setting @xmath142
, we may rewrite this expression as @xmath143 and setting @xmath144 yields @xmath145 where @xmath146 the set @xmath42 is connected , so if it is a subset of @xmath147 then it must be a subset of some component of @xmath147 .
on the other hand , if @xmath148 is a ( possibly trivial ) component of @xmath147 and @xmath149 , then @xmath150 , because @xmath148 is a tree containing @xmath31 and @xmath42 is the unique minimal such tree . thus if @xmath151 then @xmath152 note that this formula is valid only if @xmath153 . substituting into
, we obtain @xmath154 ( where @xmath155 is defined by ) , giving the desired formula .
we now turn to the proof of . by proposition
[ subt - conn - equivalence ] , we have @xmath156 setting @xmath157 , we may rewrite the last expression as @xmath158 applying the binomial expansion to @xmath159 yields @xmath160 now setting @xmath161 gives @xmath162 and setting @xmath163 and rearranging gives @xmath164 which is the desired formula .
two basic invariants of a tree are its _ path sequence _ and its _ degree sequence_. the path sequence of @xmath0 is defined as @xmath165 , where @xmath166 is the number of @xmath167-edge paths in @xmath0 . the degree sequence of @xmath0 is defined as @xmath168 , where @xmath169 is the number of degree-@xmath91 vertices in @xmath0 . knowing the degree sequence is equivalent to knowing the _ star sequence _
@xmath170 , where @xmath171 is the number of @xmath20-edge stars in @xmath0 .
indeed , it is not hard to see that @xmath172 for every @xmath173 , and so @xmath174 [ degree - and - path - sequences ] the degree and path sequences of a tree @xmath0 can be recovered from its chromatic symmetric function . the key observation , due to chaudhary and gordon ( * ? ? ?
* proposition 18 ) , is that the path and star sequences of @xmath0 can be recovered from @xmath2 .
indeed , @xmath175 is the number of edges of @xmath0 , and for every @xmath176 , @xmath177 is just the coefficient of @xmath178 in @xmath46 .
meanwhile , for every @xmath179 , @xmath180 is the coefficient of @xmath181 .
we note that fougere had proved ( * ? ? ?
* theorem 3.3.1 ) that the sum of the squared vertex degrees , @xmath182 , could be obtained from the coefficient of the monomial symmetric function @xmath183 in @xmath1 .
we can rephrase the formulas for @xmath45 and @xmath2 in terms of the usual scalar product @xmath4 on the space @xmath69 of degree-@xmath27 symmetric functions ( see @xcite or @xcite ) , where @xmath27 is the order of @xmath0 . define symmetric functions @xmath184 and @xmath185 by @xmath186 then the formulas and are respectively equivalent to @xmath187 the symmetric function @xmath188 appears to have certain positivity and integrality properties , as we now explain .
the _ @xmath189 homogeneous symmetric function _
@xmath190 in indeterminates @xmath15 is the sum of all monomials of degree @xmath167 , and for a partition @xmath58 we define @xmath191 . the symmetric functions @xmath192 form a vector space basis for @xmath69 @xcite , so there is a unique list of rational numbers @xmath193 such that @xmath194 [ positivity - conjecture ] let @xmath195 be a partition , and let @xmath196 be the number of parts of @xmath197 of even length . then , for all integers @xmath198 , @xmath199 [ integrality - conjecture ] let @xmath195 be a partition .
then , for all integers @xmath198 , the number @xmath200 is an integer .
we have verified conjectures [ positivity - conjecture ] and [ integrality - conjecture ] computationally for all @xmath201 , which we think is strong evidence that they hold for all @xmath27 .
a formula for @xmath202 can be written out explicitly using the known transition matrices between bases of symmetric functions ( see @xcite ) . however , we do not know a direct combinatorial interpretation for @xmath202 or for @xmath200 .
one might hope that for every two trees @xmath203 with @xmath204 , the number of subtrees of @xmath0 isomorphic to @xmath24 might be given by a scalar product @xmath205 , where @xmath206 is some symmetric function independent of @xmath0 .
such a result would generalize corollary [ degree - and - path - sequences ] ( which covers only the case that @xmath24 is a path or a star ) and , by a theorem of harary and palmer @xcite , would imply that every tree is distinguished by its chromatic symmetric function .
in fact , it appears that such a function @xmath206 exists _ only if _ @xmath24 is a star or a path , as we have verified computationally for all @xmath24 of order @xmath207 , with one trivial exception . , the path @xmath208 , and another tree @xmath24 . since the number of four - edge subtrees of @xmath0 is just @xmath209 , we have @xmath210 . ]
theorem [ main - thm ] does not resolve stanley s question , because @xmath2 is not a complete isomorphism invariant .
indeed , the two trees @xmath211 shown in figure [ counterexample - figure ] share the same subtree polynomial ; this is a special case of a theorem of eisenstat and gordon @xcite . on the other hand , @xmath212
this inequality follows from tan s calculations @xcite , and also for the following elementary reason .
let @xmath213 be the edge set obtained by deleting the two rightmost horizontal edges in figure [ counterexample - figure ] ; then @xmath214 .
on the other hand , no subset of @xmath215 has that type .
therefore @xmath216 and @xmath217 .
the remainder of the article is devoted to identifying special classes of trees @xmath0 for which the invariants @xmath75 suffice to reconstruct @xmath0 up to isomorphism .
a tree is a _ spider _ ( or _ starlike tree _ ) if exactly one of its vertices has degree @xmath53 . by corollary [ degree - and - path - sequences ] , whether or not a tree is a spider can be determined from its subtree polynomial .
a spider may equivalently be defined as a collection of edge - disjoint paths ( the _ legs _ ) joined at a common endpoint @xmath218 ( the _ torso _ ) .
up to isomorphism , every spider on @xmath27 vertices can be described by a partition @xmath219 whose parts are the lengths of its legs ( so @xmath220 ) .
we denote the corresponding spider by @xmath221 .
note that @xmath222 equals both the number of leaves of @xmath221 and the degree of its torso .
we will show that the isomorphism type of a spider can be determined from its subtree polynomial , hence from its chromatic symmetric function .
fougere ( * ? ? ?
* chapter 2 ) had previously shown that _ forks _ ( spiders with exactly one leg of length @xmath223 and _ extended stars _ ( spiders in which every leg has length @xmath20 or @xmath224 for some @xmath20 ) could be reconstructed from their chromatic symmetric functions . before continuing ,
we describe a combinatorial problem whose solution will play a role in the proof .
let @xmath225 be nonnegative integers with @xmath226 .
suppose that we have @xmath227 distinguishable boxes , of which @xmath228 have capacity @xmath167 for each @xmath229 $ ] .
let @xmath230 be the number of ways of distributing @xmath20 indistinguishable balls among these boxes so that no box is filled beyond its capacity .
in general , it is not easy to write down a closed formula for @xmath230 , although individual instances can be computed using an inclusion - exclusion argument ( for example ) .
[ spider - thm:1 ] let @xmath219 be a partition with @xmath231 , and let @xmath232 be the corresponding spider .
then @xmath0 can be reconstructed from its subtree polynomial .
first , suppose that @xmath233 .
then @xmath234 , @xmath235 , and @xmath236 are respectively the diameter , number of edges , and number of three - leaf subtrees of @xmath221 . these invariants can be recovered from @xmath237 , and together they determine @xmath33 .
now suppose that @xmath238 .
let @xmath239 denote the number of parts of size @xmath20 in @xmath33 , and let @xmath240 denote the number of subtrees of the spider @xmath0 with @xmath167 edges and @xmath91 leaf edges ( that is , the coefficient of @xmath241 in @xmath46 ) .
we will show by induction on @xmath20 that @xmath239 can be calculated from the numbers @xmath240 .
first , suppose @xmath242 . since there is a bijection between legs of @xmath0 of length 1 ( i.e. , consisting of a single edge ) and subtrees with @xmath243 edges and @xmath244 legs ( which are formed by deleting such an edge ) .
hence @xmath245 . for @xmath246
, we can choose a subtree @xmath127 with @xmath247 edges and @xmath244 leaves as follows .
first , fix @xmath248 $ ] and delete a leg of length @xmath91 ; there are @xmath249 ways to do this .
if @xmath250 , then we still need to delete @xmath251 more edges .
it suffices to specify how many edges to delete from the end of each of the other @xmath244 legs , so the number of ways to do this is the solution to the balls - in - boxes problem described above , regarding the @xmath251 edges to be deleted as balls and each remaining leg of length @xmath167 as a box of capacity @xmath252 ( since deleting the entire leg will result in a tree with fewer than @xmath244 leaves ) .
therefore @xmath253 is given by the formula @xmath254 by induction , the @xmath240 determine @xmath255 , and equation implies that they determine @xmath239 as well .
another way of reconstructing a spider from its chromatic symmetric function will be useful in section [ squid - section ] .
define the _ two - part portion _ @xmath256 of @xmath17 by @xmath257 while @xmath258 is evidently a much weaker invariant than @xmath17 , it contains enough information to distinguish among spiders . [ spider - thm:2 ] let @xmath0 be a spider of order @xmath27 .
let @xmath259 , let @xmath260 , let @xmath261 for @xmath262 , and let @xmath263 .
then one of the following conditions holds : 1 .
the sequence @xmath264 is a partition , that is , @xmath265 .
2 . there is a number @xmath266 such that @xmath267 .
moreover , @xmath268 , and @xmath269 . in the first case ,
let @xmath270 . in the second case
, define a partition @xmath197 from @xmath264 by replacing all 2 s after the @xmath271 place with two 1 s
. then @xmath232 , where @xmath33 is the conjugate partition to @xmath197 .
for each edge @xmath272 , denote by @xmath273 the minimum of the orders of the components of @xmath274 , so that @xmath275 .
if we label the legs of @xmath0 as @xmath276 , where @xmath277 has @xmath278 edges , and label the edges of each @xmath277 as @xmath279 , starting from the leaf and reading in toward the torso , then @xmath280 .
* case 1 : * @xmath281 .
then @xmath282 for every edge @xmath283 . in this case @xmath284 , and @xmath264 is just the conjugate partition of @xmath33 .
* case 2 : * @xmath285 .
if @xmath286 , then @xmath287 , and @xmath288 for every @xmath91 .
so we can give a formula for @xmath264 in terms of @xmath33 : @xmath289 note that @xmath290 , because @xmath291 is the unique edge whose deletion contains a component of order @xmath292 .
on the other hand , @xmath293 , so @xmath294 whether @xmath27 is odd or even .
therefore , @xmath264 is not a partition , but has the form described in case ( 2 ) of the theorem .
formula implies that if we `` iron out '' @xmath264 by replacing every 2 after the @xmath295 entry with two 1 s , we will obtain the conjugate partition to @xmath33 .
we now see how to recover @xmath33 , and thus the isomorphism type of the spider @xmath221 , from the data @xmath296 .
either @xmath296 is a partition , in which case it is the conjugate of @xmath33 , or it has the form just described , in which case the `` ironing - out '' operation yields the conjugate of @xmath33 .
this is precisely the statement of the theorem .
a _ caterpillar _ is a tree @xmath0 with the property that the induced subgraph on the non - leaf vertices is a nontrivial path , called the _ spine _ of @xmath0 .
that is , every vertex of @xmath0 either lies on the spine , or is a leaf whose unique neighbor lies on the spine .
since the spine is nontrivial , a caterpillar must have at least four vertices . while eisenstat and gordon s result in @xcite rules out the possibility of distinguishing caterpillars by their subtree polynomials , there is still reason to hope that the additional information provided by the chromatic symmetric function of a caterpillar may suffice to reconstruct it up to isomorphism .
our first result is that the chromatic symmetric function distinguishes caterpillars from non - caterpillars .
the number of leaves and the diameter of a tree @xmath0 ( the maximum length of a path in @xmath0 ) can be recovered from @xmath1 by corollary [ degree - and - path - sequences ] , so it suffices to prove the following fact .
let @xmath0 be a tree with @xmath297 vertices .
then @xmath0 is a caterpillar if and only if @xmath298 .
if @xmath0 is a caterpillar , then every path of maximum length consists of the spine together with a leaf attached to each of its endpoints , hence contains all the non - leaf vertices and two other ( leaf ) vertices .
in particular , the number of edges in such a path is one more than the number of non - leaf vertices . on the other hand , if @xmath299 is a path of maximum length in @xmath0 , then the internal vertices of @xmath299 are not leaves of @xmath0 ( because each has two neighbors in @xmath299 ) but its endpoints are ( otherwise @xmath299 could be lengthened ) .
if @xmath298 , then all the vertices not lying on @xmath299 must be leaves , which is to say that @xmath0 is a caterpillar .
let @xmath0 be a caterpillar with spine vertices @xmath300 .
for each @xmath167 , let @xmath301 , where @xmath302 denotes the degree of the vertex @xmath303 .
gordon and mcdonnell ( * ? ? ?
* lemma 2 ) showed that the numbers @xmath304 are almost determined by the path sequence of @xmath0 , and are indeed determined by the path sequence when the caterpillar is _ symmetric _ ( that is , @xmath305 for @xmath306 ) .
therefore , every symmetric caterpillar is determined up to isomorphism by its chromatic symmetric function , a result proved in another way by the second author ( * ? ? ?
* theorem 4.3.1 ) .
we now describe another class of caterpillars that can be reconstructed from their chromatic symmetric functions .
we retain the labeling of the vertices of @xmath0 as @xmath300 .
let @xmath307 be the number of _ leaves _ adjacent to @xmath303 , so that @xmath308 for @xmath309 and @xmath310 for @xmath311 .
call @xmath307 the @xmath189 _ leaf number _ of @xmath0 .
in addition , call a partition @xmath33 _ singleton - free _ if all its parts are at least 2 .
[ distinct - caterpillar - thm ] let @xmath0 be a caterpillar whose leaf numbers @xmath307 are strictly positive and distinct .
then @xmath0 can be reconstructed from its chromatic symmetric function .
let @xmath312 be the set of leaf edges of @xmath0 .
since @xmath313 for all @xmath167 , every spine vertex is adjacent to at least one leaf . therefore , the edge sets @xmath314 such that @xmath71 is singleton - free are precisely those that contain @xmath312 .
in particular , @xmath315 is the unique singleton - free partition with @xmath316 parts whose coefficient @xmath29 is non - zero .
up to reordering , the parts of @xmath33 are the numbers @xmath317 .
furthermore , for each spine edge @xmath318 , the edge set @xmath319 contributes @xmath320 to @xmath321 , where @xmath322 note that the partitions @xmath323 are all distinct .
moreover , @xmath323 has @xmath324 of its parts in common with @xmath33 ; the remaining two parts of @xmath33 must be the leaf numbers of adjacent spine vertices of @xmath0 .
( this statement is valid even if the parts of @xmath323 are not all distinct . ) in this way we can recover the leaf numbers of the endpoints of every edge of @xmath0 , and this data specifies the caterpillar up to isomorphism . using the same argument
, we can relax the condition of the theorem slightly : we need only require that all leaf numbers are positive and that for each @xmath20 , the set of spine vertices with leaf number @xmath20 form a subpath of the spine .
despite the title of this article , we devote the last section to a family of graphs @xmath6 that are not trees , but rather _ unicyclic _ ; that is , @xmath6 contains a unique cycle .
equivalently , @xmath325 , where @xmath326 are the numbers of components , vertices and edges respectively , so unicyclicity can be detected from @xmath25 by proposition [ basic - facts ] .
while we can no longer interpret the coefficients @xmath75 as in , we can use stanley s expansion of @xmath17 in terms of broken circuits ( * ? ? ?
a special case of that result is the following : if @xmath6 has a unique cycle @xmath116 and @xmath327 is an edge of @xmath116 , then @xmath328 a _ squid _ is a connected unicyclic graph with a single vertex @xmath103 of degree greater than 2 .
note that @xmath103 must lie on the cycle .
a squid is described up to isomorphism by the length of its cycle and of the _ tentacles _ ( the paths from the leaf vertices to @xmath103 ) .
for example , if @xmath221 is the spider whose leg lengths are given by the parts of @xmath33 , then adding an edge between the leaves at the ends of the two longest legs produces a squid with cycle length @xmath234 and tentacle lengths @xmath329 .
it is not clear how to determine from @xmath17 whether or not a unicyclic graph @xmath6 is a squid ( for instance , we can not recover the degree sequence of an arbitrary graph from its chromatic symmetric function as we can for a tree ) .
nevertheless , the following uniqueness result does hold .
let @xmath35 be a squid with unique cycle @xmath116 .
let @xmath224 be the length of @xmath116 ; this number can be recovered from @xmath330 by proposition [ girth - prop ] .
let @xmath103 be the unique vertex of @xmath35 of degree @xmath331 , and let @xmath197 be the partition whose parts are the edge lengths of the tentacles .
label the edges of @xmath116 as @xmath332 , starting at @xmath103 and proceeding around the cycle . by and inclusion - exclusion ,
we obtain @xmath333 where the omitted terms involve edge sets @xmath31 lacking three or more edges from @xmath334 . deleting three or more edges from @xmath35
produces a graph with three or more connected components , so passing to two - part portions as in yields @xmath335 each graph @xmath336 has exactly two components , of sizes @xmath337 and @xmath338 .
removing additional edges will strictly increase the number of components , so @xmath339 where @xmath340 is the spider with legs whose lengths are @xmath167 , @xmath341 , and the parts of @xmath197 .
therefore , @xmath330 determines the quantity @xmath342 leaving the foregoing calculations aside for the moment , we note that if @xmath0 is a tree with @xmath101 edges , then to calculate @xmath343 we need only consider the edge subsets of cardinality @xmath243 .
in particular , if @xmath232 is a spider , then @xmath344 it follows that for any partition @xmath197 and numbers @xmath345 , we have @xmath346 which can be computed from @xmath330 using .
meanwhile , by theorem [ spider - thm:2 ] , we can reconstruct the spider @xmath347 from @xmath348 .
in particular , we can reconstruct the partition @xmath197 , which gives the tentacle lengths of the squid @xmath35 . just as squids can be regarded as the unicyclic analogues of spiders , the unicyclic analogues of caterpillars are _ crabs _ : connected unicyclic graphs in which every vertex not lying on the cycle
is a leaf . in analogy
to theorem [ squid - thm ] and its proof , we ask whether is it possible to use the results of section [ caterpillar - section ] to classify the chromatic symmetric functions of ( some ) crabs . as a starting point
, we prove the following analogue of theorem [ distinct - caterpillar - thm ] . the girth @xmath108 of @xmath6 can be recovered from @xmath17 by proposition [ girth - prop ] .
let @xmath116 be the unique cycle of @xmath6 , and label its vertices in cyclic order as @xmath352 .
let @xmath310 be the number of leaves adjacent to @xmath303 .
note that @xmath6 can be specified up to isomorphism by the cyclically ordered list of numbers @xmath353 .
let @xmath312 denote the set of leaf edges of @xmath6 .
the subsets of @xmath8 whose type is singleton - free are precisely those that contain @xmath312 as a subset . in particular , @xmath312 itself is the unique edge set whose type is a singleton - free partition of length @xmath108 .
thus @xmath354 can be recovered from @xmath17 .
moreover , there are precisely @xmath108 edge sets whose type is a singleton - free partition of length @xmath355 ; these edge sets are of the form @xmath356 for some @xmath357 . just as in theorem [ distinct - caterpillar - thm ] , the types of these edge sets are all distinct , and they specify which pairs of the @xmath307 correspond to adjacent vertices of the cycle
. jeremy l. martin and jennifer wagner , _ on the chromatic symmetric function of a tree _ , extended abstract for 18th annual conference on formal power series and algebraic combinatorics ( san diego , 2006 ) .
li - yang tan , _ computational verification that the chromatic symmetric function determines up to isomorphism trees of order at most 23 _ , http://www.cs.wustl.edu/~lt1/csf.html , retrieved on may 14 , 2007 .
|
let @xmath0 be an unrooted tree .
the _ chromatic symmetric function _
@xmath1 , introduced by stanley , is a sum of monomial symmetric functions corresponding to proper colorings of @xmath0 .
the _ subtree polynomial _
@xmath2 , first considered under a different name by chaudhary and gordon , is the bivariate generating function for subtrees of @xmath0 by their numbers of edges and leaves .
we prove that @xmath3 , where @xmath4 is the hall inner product on symmetric functions and @xmath5 is a certain symmetric function that does not depend on @xmath0 . thus the chromatic symmetric function is a stronger isomorphism invariant than the subtree polynomial . as a corollary
, the path and degree sequences of a tree can be obtained from its chromatic symmetric function . as another application ,
we exhibit two infinite families of trees ( _ spiders _ and some _ caterpillars _ ) , and one family of unicyclic graphs ( _ squids _ ) whose members are determined completely by their chromatic symmetric functions .
|
quantum key distribution ( qkd ) @xcite allows two parties to share a secret key which might be used for applications in cryptography .
the preferred quantum systems used for transmitting information are photons
. these can be generated , distributed and measured fairly easily with standard technology .
however , photons are usually transmitted through optical fibers and due to absorption the maximal distance where qkd is feasible is around 150 km @xcite . in order to overcome this problem the concept of quantum repeaters
can be used @xcite . for increasing the final repeater rate and the final fidelity many variations of the original protocol
have been investigated @xcite , where one of the influential generalizations is _ multiplexing _
@xcite . in fig .
[ fig : multiplexingscheme ] we show a typical set - up of a quantum repeater with multiplexing .
alice and bob have many single - photon sources which are connected to a quantum memory in the repeater station .
optical fibers are lossy , therefore after one attempt some quantum memories are filled up ( red spheres in the picture ) and some are empty ( gray spheres ) .
one possibility is to perform bell - state measurements ( bsms ) only between parallel quantum memories ; the second possibility , which is called multiplexing is to allow bsms between two arbitrary quantum memories of the two arrays . in ref .
@xcite the authors give an analytical formula for the entanglement production rate with multiplexing when quantum repeaters with two segments are considered .
the conclusion of @xcite was that multiplexing gives only a modest improvement on the rate w.r.t .
the case of parallel connections .
however , it improves significantly the requirements on memory decoherence . in ref .
@xcite a new protocol based on the duan - lukin - cirac - zoller protocol @xcite has been studied and it has been found numerically that @xmath0 where @xmath1 is the rate using multiplexing and @xmath2 is the rate using parallel connections .
other works concerning multiplexing include ref .
@xcite which studied the repeater rate and the final fidelity in the limit of large number of quantum memories , ref .
@xcite which derived an analytical formula for the average number of attempts necessary for performing the first connection and ref .
@xcite where a new protocol based on multiplexing has been proposed .
in this paper we assume a set - up with one repeater station , i.e. two segments ( see fig .
[ fig : multiplexingscheme ] ) .
we consider multiplexing when few quantum memories are used .
we introduce the _ finite - range multiplexing protocol _ ( firmp ) , which is motivated by the fact that long - range connections are experimentally demanding @xcite .
we provide analytical formulas for the repeater rate using the _ full - range multiplexing protocol _ ( furmp ) and the firmp .
then we investigate quantum memory decoherence , and we study numerically the optimal algorithm such that the memory requirements for qkd are minimized . the manuscript is organized as follows . in sec .
[ sec : protocol ] we introduce quantum key distribution and the quantum repeater protocol with finite - range multiplexing .
moreover , we describe different bell - state measurement strategies . in sec .
[ sec : reprate ] we derive analytical formulas for the repeater rate in the case of deterministic and probabilistic bell - state measurements . in sec .
[ sec : decoherence ] we show how to minimize the memory requirement such that quantum key distribution is still possible and in sec .
[ sec : scr ] we calculate the secret key rate . finally , in sec .
[ sec : conclusions ] we summarize the results and outline possible future developments .
alice and bob are two parties at a distance @xmath3 who want to create a secret key using qkd . throughout the present paper we consider that they use a quantum repeater with two segments , i.e. one repeater station .
this set - up is particularly important because alice and bob do not necessarily need entanglement sources or quantum memories . instead , single - photon sources or weak coherent pulse sources are sufficient .
this set - up resembles the measurement - device independent qkd protocol proposed in @xcite .
this protocol has been extended to the quantum repeater scenario with quantum memories in @xcite . in this paper , the sources are supposed to be single - photon sources . however , the analysis for weak coherent pulse sources can be done following the methods developed in @xcite .
we assume that the repeater station contains two arrays of @xmath4 quantum memories , where one side receives the photons sent from alice and the other one receives the photons sent from bob ( see fig .
[ fig : multiplexingscheme ] ) . .
bold vertical lines which are proportional to @xmath5 represent the instant where sources produce new photons .
the quantity @xmath6 is the frequency of the source measured in pairs per second . ] in the following we give the steps of the multiplexing protocol with finite - range connection .
we define time variables denoted by @xmath7 with integers @xmath8 and @xmath9 interpreted as follows .
the value of @xmath10 denotes the attempts of the sources to produce photons . as shown in fig .
[ fig : timescheme ] the variable @xmath11 can be always related to the elapsed time in seconds from the beginning of the experiment by using the repetition frequency @xmath6 of the source which is measured in pairs per second .
therefore @xmath12 , @xmath13 , etc .
we will call the interval between @xmath14 and @xmath15 a time - bin .
the subindex @xmath16 permits to describe instants contained in a time - bin . at time @xmath17
all quantum memories are empty and the protocol is just starting . quantum memories for each array and the connection length is @xmath18 . in @xmath19
it is not possible to perform all possible connections due to the limited connection length . in the case of furmp in @xmath19
three bsms would be possible and in @xmath20 all quantum memories would be empty . ] the steps are the following ( see fig . [
fig : scheme ] ) : 1 . : alice and bob prepare randomly and independently @xmath4 random states each according to a chosen qkd protocol where @xmath21 . ] .
they store the information regarding the preparation and they send the states to the repeater station .
2 . : arriving photons are stored and heralded in the corresponding quantum memory .
3 . : the repeater station performs the maximal number of bell - state measurements @xmath22 compatible with the maximal connection length @xmath23 . for @xmath24 only bsms in parallel are allowed and for @xmath25 any connection is possible .
4 . : the measured quantum memories are again empty .
restart from @xmath26 , i.e. @xmath27 .
the repeater station communicates to alice and bob which pairs of quantum memories were used for the bsms , as well as the measurements result .
the protocol ends after a given number of rounds .
after that , alice and bob will execute the standard qkd protocol which consists of sifting , parameter estimation , error correction and privacy amplification @xcite .
the advantage of a multiplexing protocol occurs at time @xmath19 where bsms are performed such that the number of connections is maximized . in the case of furmp
this corresponds to performing as many bsms such that one array of quantum memories is completely empty . in the case of firmp
the maximal number of connections can be found using the _ maximum cardinality bipartite matching _
@xcite , i.e. the maximum number of edges in a given bipartite graph such that each vertex has at most one neighbor .
this problem can be explicitly formulated as follows : there is a list of filled quantum memories on the left .
each quantum memory on the left may be connected to several quantum memories on the right , depending on the maximal connection length .
this defines a bipartite graph . by solving the maximum cardinality bipartite matching algorithm
we find the maximal number of possible connections .
in the case of full - range multiplexing the optimal matching will always leave one of the two arrays completely empty .
this is not the case with finite - range multiplexing .
however , it remains true that at least one of two involved quantum memories has been filled in the same time - bin when the connection is done .
obviously , there could be several possible matchings which maximize the number of connections .
the chosen matching can have an influence on the final state fidelity due to memory decoherence .
possible strategies for choosing a matching are discussed in the following .
we can view the set - up in a repeater station as a bipartite weighted graph where the filled quantum memories are the vertices and the possible connections restricted by the maximal connection length are edges . to each vertex
is assigned an integer value given by the arrival time of the stored photon .
if the quantum memory is empty it does not represent a vertex .
( see fig .
[ fig : scheme ] ) .
we consider @xmath18 . on the right three possible matching strategies
are shown .
blue lines indicate the difference between the arrival times of the photons .
it is possible to see the schemes on the right side as weighted bipartite graphs : red spheres are vertices and blue edges have indicated weights . ]
a protocol for the bsms matching consists of the following steps : 1 .
identify all possible connections between the vertices .
assign to each edge a weight @xmath28 given by the absolute value of the difference between the arrival times .
this quantity identifies the amount of decoherence that has been experienced by the older quantum memory .
this is a meaningful quantity because one of the two involved quantum memories is always fresh .
the resulting data structure is a weighted bipartite graph @xmath29 , i.e. a set of edges ( vertex - pairs ) and edge weights . 2 .
for each weighted bipartite graph @xmath30 solve the _ maximum cardinality bipartite matching _ problem .
we denote by @xmath31 the subset of @xmath30 containing graphs with exactly @xmath22 edges , where @xmath22 is the highest matching cardinality obtained over all graphs .
this subset contains all graphs which maximize the number of connections , and thus the repeater rate .
3 . select one element @xmath32 by the following strategies : 1 .
: choose with equal probability an arbitrary @xmath33 from @xmath31 2 .
: minimize the sum of the weights @xmath34 3 .
: maximize the sum of the weights @xmath34 the optimization involved in strategy 1 and 2 is known in literature as _ _ maximum weighted bipartite matching__@xcite and the optimization algorithm has complexity @xmath35 @xcite where @xmath4 is the number of quantum memories on one side .
strategy 1 connects pairs which arrived with a short time difference giving as a result the highest correlations which are possible to produce at a certain time .
the disadvantage is that older pairs remain in the memories and therefore are used at a later time , having experienced decoherence for a long time .
strategy 2 resolves this mentioned problem , connecting pairs with the largest time difference .
this strategy removes from the quantum memories older pairs as soon as possible , leaving only quantum memories which suffered decoherence for a short time .
the disadvantage is that poor correlations are produced even when perfect correlations could be obtained .
we have seen therefore , that both strategies have advantages and disadvantages . in sec .
[ sec : decoherence ] we will discuss which strategy minimizes the memory requirements for qkd .
however , note that the repeater rate is independent of the matching strategies .
in this section we derive analytical formulas for the repeater rate for the furmp and firmp .
let @xmath36 be the current time ( measured in time - bins as defined in the previous section ) , then the repeater rate is the fraction of successful bsms per memory per time - bin calculated over the whole running time of the quantum repeater protocol , i.e. @xmath37 where @xmath38 is the average number of successful bsms at time @xmath19 .
many quantum repeater protocols are based on a probabilistic bell - state measurement @xcite .
the bsm is probabilistic when implemented with linear optics @xcite or with detectors of finite efficiency @xcite .
when a measurement fails , the involved quantum memories are supposed to be emptied and this attempt is marked as unsuccessful .
let @xmath39 be the success probability of the bsm .
the probability that @xmath22 bsms are successful is given by @xmath40(t_2):=\nonumber\\&\sum_{i=\ell}^{m}\binom{i}{\ell}{\textrm{prob}}[\lambda = i](t_2)p_{bsm}^{\ell } ( 1-p_{bsm})^{i-\ell},\end{aligned}\ ] ] where @xmath41 and @xmath42 are random variables that can assume values 0,1 , ... , @xmath4 .
the random variable @xmath41 denotes the number of successful bsm and @xmath42 the number of performed bsms .
the factor @xmath43 represents the probability that @xmath22 bsm are successful and @xmath44 are not successful .
this event can happen in @xmath45 different ways .
the average number of successful bsm at time @xmath19 is given by @xmath46(t_2)\ ] ] in the following we will focus on @xmath47 $ ] .
we denote as @xmath48 one possible configuration of the quantum memories in the repeater station .
the vectors @xmath49 and @xmath50 of length @xmath4 represent the status of the quantum memories on alice s and bob s side , respectively ( see fig .
[ fig : multiplexingscheme ] ) .
each component takes the value @xmath51 if the corresponding quantum memory is empty and @xmath52 otherwise .
we define as @xmath53 the set of all configurations leading to @xmath22 bsms where @xmath23 is the maximal connection length . for example , let @xmath54 and @xmath55 be the configurations of the quantum memories as seen in fig . [
fig : scheme ] at @xmath19 , then @xmath56 and @xmath57 but @xmath58 because when @xmath18 the maximal number of connections is @xmath59 .
moreover , the set of all possible configurations is @xmath60 .
we model the whole process consisting of storage and measurement with two maps .
the _ storage map _
@xmath61 connects configurations at time @xmath14 , i.e. , before photons are received , to configurations at time @xmath62 , i.e. , after photons are received and stored .
given @xmath63 , the probability to have the configuration @xmath64 is given by @xmath65&:={\textrm{prob}}[{\bold{c}}'|{\bold{c } } ] \\ & : = \prod_{i=1}^{m}{\textrm{prob}}[c'_{i}|c_{i}]\\ & : = \prod_{i=1}^{m}{\textrm{prob}}[a'_{i}|a_{i}]{\textrm{prob}}[b'_{i}|b_{i } ] \label{eq : probccp } , \end{aligned}\ ] ] with @xmath66&:=(1-p)(1-a'_{i})(1-a_{i } ) \nonumber \\ & + p a'_{i}(1-a_{i } ) + a'_{i}a_{i } \label{eq : probaapbbp},\end{aligned}\ ] ] where @xmath67 is the probability that a photon has not been absorbed by the quantum channel .
the probability @xmath68 $ ] is defined analogously .
equation holds because the channels connecting each source to each quantum memory are independent .
equation gives the conditional probability to have a final configuration @xmath69 starting from an initial configuration @xmath70 .
the three addends on the right - hand side of eq .
are mutually exclusive , i.e. given a certain configuration at most one is not zero .
the measurement map @xmath71 relates configurations at time @xmath62 and @xmath15 , i.e. , after the quantum memories have been used for the bsms .
this map is deterministic , as the configuration @xmath64 after the measurement is uniquely determined by the matching algorithm .
coming back to the probability @xmath72(t_2)$ ] to have @xmath22 bsms we get @xmath73(t_2)&:=\sum_{{\bold{c}}'\in{{{\mathcal{h}^{m}_{w}}(\ell)}}}{\textrm{prob}}[{\bold{c}}'](t_1 ) \label{eq : probl}\\ & = \sum_{{\bold{c}}'\in{{{\mathcal{h}^{m}_{w}}(\ell)}}}\sum_{{\bold{c}}\in{{\mathcal{h}^{m}_{w}}}(0)}{\textrm{prob}}[\sigma_{\ell}({\bold{c}})={\bold{c}}']{\textrm{prob}}[{\bold{c}}](t_0),\end{aligned}\ ] ] which is the sum over all possible initial configurations @xmath74 and configurations @xmath75 at time @xmath19 of the probability that @xmath74 leads to @xmath75 weighted with the probability that the configuration @xmath74 was realized at time @xmath14 .
the probability @xmath76(t_0)$ ] is given by @xmath77(t_0)=\sum_{{\bold{c}}'\in{{\mathcal{h}^{m}_{w}}}(0)}\sum_{l=0}^{m}{\textrm{prob}}[\mu_{\ell}\circ\sigma_{\ell}({\bold{c}}')={\bold{c}}]{\textrm{prob}}[{\bold{c}}']((t-1)_0),\ ] ] i.e. , given a state @xmath78 at time @xmath79 , we calculate the probability that photon storage and measurement will lead to @xmath63 at time @xmath14 . this last probability can be rewritten as @xmath80=\sum_{{\bold{c}}''\in{{\mathcal{h}^{m}_{w}}}(\ell)}\delta_{\mu_{\ell}({\bold{c}}''),{\bold{c}}'}{\textrm{prob}}[\sigma_{\ell}({\bold{c}})={\bold{c}}''],\ ] ] where @xmath81 is the kronecker delta such that @xmath82 iff @xmath83 and @xmath84 otherwise . using the previous formulas and the initial condition @xmath85(0_0)=\delta_{{\bold{a}},\bold{0}}\delta_{{\bold{b}},\bold{0}}$ ] we have now all elements for calculating the repeater rate in eq . .
in order to do that one inserts eq . into eq . which is then inserted in eq . which is finally used for calculating eq
. through eq .
the calculation was performed in c`++ ` .
the measurement map has been implemented using the library provided at ref .
the complexity of the calculation is proportional to the number of time steps because for calculating @xmath86 it is sufficient to know quantities at time @xmath87 .
however , the set @xmath53 grows quite fast as function of @xmath4 and therefore reasonable time considerations restricted the calculation to a maximal of @xmath88 . regarding fig .
[ fig : multiplexingcomp ] the calculation ran for two days on a cluster of 10 nodes , cpu with four - cores and eight gb of ram . in the following and for the rest of the paper we consider @xmath89 which represents the transmission probability of a single photon over an optical fiber of length @xmath90 km and for an absorption coefficient @xmath91 db / km .
the relation between @xmath67 and the distance between alice and the repeater station @xmath3 is @xmath92 @xcite . as seen in fig .
[ fig : multiplexingcomp ] the repeater rate increases as function of the time reaching a plateau at time @xmath93 .
we see that this behavior persists when changing the maximal connection length @xmath23 . analyzing the dependence on the maximal connection length , we observe in fig .
[ fig : multiplexingcomp ] that the gap between @xmath24 and @xmath18 is almost the same than the gap between @xmath24 and @xmath94 which represents full - range multiplexing .
this shows that in an experimental implementation in order to profit of multiplexing it is not necessary to have long - range connection .
moreover , for our set - up with a source at @xmath52 khz , the loading time is @xmath95 s long .
this result could give a hint that in more complex quantum repeater protocols with many repeater stations , using distillation and classical communication , the loading time could play a significant role in the total time of the execution of the quantum repeater protocol . finally , we give an analytical formula for the rate when the plateau is reached . the time evolution of our system
is specified in eq . .
for @xmath96 , the rate becomes time - independent : @xmath97=\sum_{{\bold{c}}'\in{{\mathcal{h}^{m}_{w}}}(0)}\sum_{l=0}^{m}{\textrm{prob}}[\mu_{\ell}\circ\sigma_{\ell}({\bold{c}}')={\bold{c}}]{\textrm{prob}}[{\bold{c}}'].\ ] ] here @xmath76 $ ] is the unknown to be determined .
we derive an analytical form of @xmath76 $ ] in app .
[ app : fx ] . here
, we use this formula ( eq . ) for calculating the asymptotic repeater rate as function of the bsm success probability . as shown in fig .
[ fig : ratepbsm ] the largest improvement is possible to have with full - range multiplexing , but already a similar improvement is reached with maximal connection length @xmath18 , instead of @xmath94 .
moreover , the linear behavior can be justified as follows . in the case of @xmath89 and for @xmath18
we obtain @xmath98=3.9\cdot10^{-3}$ ] , @xmath99=5.5\cdot10^{-6}$ ] and @xmath100 \ll { \textrm{prob}}[\lambda=2]$ ] .
therefore , eq . becomes @xmath101 $ ] which is linear in @xmath39 . for other values of the maximal connection length
the order of magnitude of the probabilities is the same .
in case of multiplexing , even when the rate is maximized it is possible to connect pairs in different ways . in this section , we study the optimal matching algorithm for the bsm in relation to memory decoherence . for simplicity we will stick to deterministic entanglement swapping , i.e. @xmath102 in eq .
our figure of merit for optimizing the matching algorithm will be the secret fraction , which in qkd characterizes the fraction of secret bits that can be extracted the from measured qubits @xcite . in the case of the bb84 protocol
the secret fraction is given by @xcite @xmath103 where @xmath104 are the quantum bit error rates ( qber ) in base @xmath105 and @xmath106 and @xmath107 is the binary shannon entropy . for simplicity
, we consider a symmetric error model such that @xmath108 .
the qber resulting from measurements performed at time @xmath19 is @xmath109(t_2),\ ] ] where @xmath110(t_2)$ ] is the fraction of measurements of quantum memories which have experienced decoherence for a time @xmath111 .
this probability depends on the bsm strategy ( see sec . [
sec : protocol ] ) .
the qber after these measurements is given by @xmath112 .
this quantity depends on the decoherence mechanism of the quantum memories . in this paper
we consider depolarization . given @xmath113 , the state of the quantum memory at time @xmath14 , after depolarization it becomes @xmath114 where @xmath115 and @xmath116 is the decoherence time of the quantum memory . for the bb84 it holds and @xmath117 , respectively , will result in a depolarized state of fidelity @xmath118 . inserting @xmath119 and @xmath120 and using the fact that @xmath121 @xcite the result follows .
] @xmath122 the total qber is calculated between all outcomes that alice and bob get from the beginning of the protocol until time @xmath36 which is equal to @xmath123(t_2)}{\sum_{t=0}^{{t_{\mathrm{c}}}}{\left<\ell\right>}(t_2)}.\ ] ] here , the denominator is the total number of bsms from the beginning of the protocol until time @xmath36 .
the numerator is the average qber for each time - bin weighted with the total number of successful measurements for each time - bin .
the secret key rate is not zero whenever @xmath124 .
this will be used to obtain a lower bound on the necessary coherence time @xmath116 .
we have calculated eq . using numerical simulations .
it is also possible to proceed analytically as explained in sec .
[ sec : reprate ] .
however , the space of the configurations is so large that the analytical computation becomes unfeasible .
we have performed numerical simulations by repeating many times the protocol , and from the obtained connections we have calculated the averages .
the number of used experiments is about @xmath125 which permits to have a variance of the mean smaller than @xmath126 .
the simulations were performed for the strategies @xmath127 , and @xmath128 , which were introduced in sec .
[ sec : protocol ] . as shown in fig .
[ fig : minimaltau65 ] the minimal necessary coherence time is given by strategy 1 .
we see that strategy 0 is between strategy 1 and 2 and that the ordering between the strategy remains the same by changing the maximal connection length .
we then studied how the minimal coherence time scales as function of @xmath23 .
as seen in fig .
[ fig : minimaltaumanyw ] even a maximal connection length of @xmath18 has a significant impact on the minimal coherence time , compared to @xmath24 . in particular we observe that for @xmath129 the improvement from @xmath24 to @xmath18 is roughly @xmath130 .
in this section we will sum up the results of the previous two sections and we will calculate the secret key rate .
the secret key rate at time @xmath36 is calculated as the product of the repeater rate and the secret fraction , . ]
i.e. @xcite @xmath131 where the repeater rate or raw key rate @xmath132 was defined in eq . and
the qber was given in eq . .
for our calculation we have used the minimal coherence time calculated in the previous section .
in particular we have chosen @xmath133 we use @xmath134 because it is the highest achievable with our simulation ( see fig .
[ fig : minimaltaumanyw ] ) . in tab .
[ tab : coherencetime ] we report the used coherence time for each value of @xmath23 . as shown in fig .
[ fig : comparisonscr ] the finite - range multiplexing with @xmath18 leads to a similar improvement as with @xmath94 .
interestingly , we see that the secret key rate has a maximum for @xmath135 and then it slowly decreases .
this fact can be explained by observing that there are two competing behaviors : the repeater rate increases with the time and the secret fraction decreases with the time , as the qber increases with time , due to the fact that the probability that poor connections happen increases .
an improvement may be to remove very old pairs which are known not to contribute to the final secret key .
this method will certainly decrease the qber at the expense to decrease also the repeater rate .
we postpone to future investigations new possible schemes which could permit to have a secret key rate which has a monotonic behavior .
.value of the coherence time @xmath116 used for calculating the secret key rate . in order to obtain the value in seconds
it is sufficient to divide by the frequency of the source . [
cols="^,^,^",options="header " , ]
quantum repeaters offer the possibility to enlarge the distance where quantum key distribution becomes feasible . in this paper
we have considered the scenario with one repeater station in the middle , containing several pairs of quantum memories .
this is a generalization of measurement - device independent qkd with quantum memories and single - photon sources .
we have considered multiplexing as a scheme for performing the bell - state measurement .
we introduced the concept of finite - range multiplexing which originates from the experimental constraint that long range connections are demanding .
we have characterized analytically the repeater rate for the case of probabilistic and deterministic bell - state measurement .
we found that in a multiplexing protocol already short - range connections cover most of the improvement over a standard protocol .
decoherence of the quantum memories and different strategies for connecting the pairs were also studied .
we found that it is always optimal to connect pairs with the shortest time difference in arrival time : this strategy minimizes the necessary coherence time required by the quantum memories in order to extract a secret key .
moreover , we have shown that also for the figure of merit `` minimal coherence time '' short - range multiplexing is almost as good as general multiplexing .
finally , we have studied the secret key rate which characterizes the performance of quantum key distribution , finding results analogous to the previous sections .
future questions may include the case of alice and bob using weak coherent pulses .
this can be done by following @xcite .
the analysis of finite - size effects for qkd can be performed by following @xcite .
the techniques derived in our paper may also be used for addressing more complicated multiplexing protocols involving distillation and classical communication .
our work suggests that in more complex protocols the loading time may play a significant role , thus reducing the repeater rate w.r.t .
asymptotic formulas .
we would like to thank sylvia bratzik , michael epping , ren schwonnek and peter van loock for valuable and enlightening discussions .
we acknowledge financial support by the german federal ministry of education and research ( bmbf , project quorep ) .
in order to determine the left - hand side of eq . , we define a function @xmath136 where @xmath137 is the cardinality of @xmath138 .
we rewrite eq . in the following way @xmath139 where @xmath140 $ ] and @xmath141 $ ] .
the solution is the following @xmath142 with in order to see that observe that eq .
can be seen as a system of equations of the unknowns @xmath144 , with the additional condition @xmath145 which comes from the fact that @xmath146 is a probability and we sum over the whole space .
therefore we have @xmath147 with @xmath148 given in eq . .
repeating the procedure , the function @xmath146 can be expressed as @xmath149 the form in eq .
is obtained by using the additional constraint @xmath145 .
30ifxundefined [ 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 in @noop _ _ , vol .
( , ) @noop * * , ( ) link:\doibase 10.1103/revmodphys.81.1301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.5932 [ * * , ( ) ] link:\doibase 10.1103/physreva.59.169 [ * * , ( ) ] http://dx.doi.org/10.1038/35106500 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.240501 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.83.33 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.060502 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.012301 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.032301 [ * * , ( ) ] link:\doibase 10.1103/physreva.83.012323 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.108.130503 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.130502 [ * * , ( ) ] @noop ( ) http://books.google.de/books?id=tuvuaaaamaaj[__ ] ( , ) http://books.google.de/books?id=25p3mhu3ij8c[__ ] ( , ) in @noop _ _ ( ) @noop * * , ( ) @noop `` , '' ( ) , link:\doibase 10.1103/revmodphys.74.145 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreva.87.052315 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.022332 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.052305 [ * * , ( ) ] @noop ( ) ,
|
quantum repeaters represent one possible way to achieve long - distance quantum key distribution .
collins et al . in [ phys .
rev .
lett .
98 , 060502 ( 2007 ) ] proposed multiplexing as method to increase the repeater rate and to decrease the requirement in memory coherence time . motivated by the experimental fact that long - range connections are practically demanding ,
in this paper we extend the original quantum repeater multiplexing protocol to the case of short - range connection .
we derive analytical formulas for the repeater rate and we show that for short connection lengths it is possible to have most of the benefits of a full - range multiplexing protocol . then we incorporate decoherence of quantum memories and we study the optimal matching for the bell - state measurement protocol permitting to minimize memory requirements . finally , we calculate the secret key rate and we show that the improvement via finite - range multiplexing is of the same order of magnitude as via full - range multiplexing .
|
one of the seminal findings triggering the development of quantum information theory is that there exist nonlocal correlations among subparts of quantum systems that do not emerge in a classical scenario .
these nonclassical correlations , commonly identified with entanglement , can be exploited to manipulate and transmit information in novel and enhanced ways @xcite going beyond the possibilities of classical physics .
consequently , an increasing interest in their study has risen in recent years @xcite .
interestingly , signatures of correlations having no classical counterpart can be traced even in separable ( non - entangled ) states , but their nature is rather different from entanglement @xcite .
in fact , while entanglement is a consequence of the superposition principle , more general forms of nonclassical correlations arise essentially from the noncommutativity of quantum observables . when speaking about composite systems , separable states are often perceived as essentially classical . however , _
truly _ classical states , i.e. states which contain only classical correlations , represent just a subset of separable states .
moreover , it is also possible to show that almost all separable states possess a finite amount of nonclassical correlations @xcite .
this has fueled a still unsettled debate , and an active stream of research , to decide whether separable states containing nonclassical correlations can be also directly useful for quantum information tasks @xcite .
to quantitatively assess various aspects of entanglement , several entanglement measures have been adopted and extensively studied @xcite . in a similar fashion , more recently a zoology of indicators of nonclassical correlations ( in separable or entangled states )
have been introduced @xcite , among which the most popular being _ quantum discord _ @xcite .
a nice facet of discord is that it has an immediate characterization in information theory , it is endowed with operational interpretations @xcite , and its evaluation for bipartite states of simple quantum systems such as two - level systems ( qubits ) is conceptually straightforward , albeit technically hard .
studies on nonclassicality indicators are not restricted to finite dimensions , but have also been extended to quantum systems with infinite - dimensional state space , where correlations are encoded between continuous variables ( cvs ) , that is , variables with continuous spectra .
they are represented by , e.g. , modes of the electromagnetic field described by quadrature amplitudes . for these systems , a privileged role is played by the states possessing a gaussian wigner function , the so called gaussian states , as they are easy to handle both theoretically @xcite and experimentally @xcite . although gaussian states are sometimes flagged as the `` most classical '' class of continuous - variable ( cv ) states because of the positivity of their wigner function , recent results of the nonclassicality indicators show the opposite .
namely , a recently derived analytical form of quantum discord for two mode gaussian states @xcite reveals that , contrarily to the above naive categorization , all non - product bipartite gaussian states have a nonzero discord and so exhibit nonclassical correlations .
another nonclassicality indicator frequently employed in the literature is the _ measurement - induced disturbance _ or _ mid _ , introduced in @xcite .
a good property of mid is the intrinsic symmetry under swapping of the subsystems ( unlike discord ) but the main flaw is that it does not incorporate any optimization over local measurements , therefore usually returning an an overestimation of the actual amount of nonclassical correlations .
the necessity of a more faithful nonclassicality quantifier motivated the introduction of a new indicator called _ amid _ ( _ ameliorated measurement - induced disturbance _ ) which is an ameliorated synthesis of the discord and mid , being both symmetric as mid and optimized as discord .
it has been applied to characterize nonclassical correlations of arbitrary two qubit mixed states @xcite .
the amid is defined as the difference between the total ( quantum ) and the classical mutual information @xcite , where the latter quantity represents the maximum classical correlations that can be extracted from a bipartite state via local measurements on subsystems . here
an optimization over all possible bi - local measurements is involved , which represents the nontrivial part in calculation of the amid . in the present work ,
we continue the programme of characterizing quantum versus classical correlations beyond entanglement in cv systems .
specifically , we extend the definitions of mid and amid to the paradigmatic class of two mode gaussian states .
contrary to the case of discord , which has been computed for gaussian states restricting the optimization to gaussian measurements only @xcite , for the nonclassicality indicators considered here non - gaussian measurements play an important role .
the mid is in fact obtained as the state disturbance induced by local projections onto fock states ( photon counting ) , a clearly non - gaussian measurement .
this can be done analytically for simple cases ( such as pure two mode squeezed states ) and requires numerical evaluation in more general gaussian states . for the amid ,
a competition between photon counting and optimized gaussian measurements takes place in the maximization of the classical mutual information .
we define a gaussian version of amid ( gaussian amid ) , restricted to gaussian measurements in analogy with the gaussian discord @xcite , and provide the analytical framework for its computation , obtaining simple closed forms for relevant families of states , including pure states and squeezed thermal states .
we compare this measure with mid and gaussian discord , elucidating similarities and differences with the two qubit case @xcite .
similarly , we find that also for gaussian states , in most cases , mid significantly overestimates quantum correlations , and we provide evidence for states with nearly vanishing discord but arbitrarily large mid . yet , quite interestingly , we find a finite region of two mode gaussian states where mid is strictly smaller than the optimal gaussian amid , meaning that in those instances non - gaussian measurements are necessarily optimal for the calculation of the classical mutual information and for the amid .
this subset of states includes , surprisingly , pure two mode squeezed states , for which local photon countings are found to be therefore `` less disturbing '' than local homodyne detections ( which realize the optimal gaussian measurements in this case ) , and allow one to extract strictly more strongly correlated classical random variables from the pure gaussian states .
the gap between photon counting and homodyne detection in the degrees of classical mutual information and amid persists even in the limit of infinite squeezing .
such a finding is quite unexpected and certainly deserves further investigation with an eye on potential practical applications . yet
, this agrees in spirit with with a series of somehow related results : for example , the best operation to clone gaussian coherent states , as well as the best partial measurement achieving the optimal information / disturbance tradeoff for the same type of state , and the best measurement localizing maximum bipartite entanglement for gaussian states , are all known to be non - gaussian @xcite .
recall also that non - gaussian measurements are necessary for universal cv quantum computation with gaussian cluster states @xcite .
the paper is organized as follows . in section [ secprelim ]
we set up the notation and recall the main concepts of gaussian states and gaussian measurements , followed by an overview of the nonclassicality measures landscape .
the derivation of a manageable expression of mid ( associated to local fock projections ) for two mode gaussian states , in closed form for pure states , is presented in section [ secmid ] . in section [ secgamid ]
we face the evaluation of the amid for arbitrary two mode gaussian states .
we bound amid from above by the minimum between the mid and the gaussian amid .
the latter is defined by restricting the optimization to bi - local gaussian measurements , adopting the techniques used for discord in @xcite , and we provide closed analytical forms for it on special families of two mode gaussian states .
section [ secdiscuss ] presents a thorough comparison between mid , gaussian amid and gaussian quantum discord on random two
mode gaussian states , highlighting hierarchical and ordering relations between the three measures .
upper and lower bounds to the gaussian amid at fixed gaussian discord are identified .
the states where non - gaussian measurements are necessary to reach the minimum in the amid are amply discussed and characterized . in section [ secent ] , we compare the gaussian amid with the gaussian entanglement of formation ( gaussian eof ) @xcite , showing on the basis of numerical evidence that the gaussian amid is always greater or equal than the gaussian eof for all two mode gaussian states , and admits also an upper bound at fixed gaussian eof ( similarly to what observed for discord @xcite ) . a visual classification of the special class of symmetric squeezed thermal states based on their degrees of gaussian eof , gaussian discord and gaussian amid
is also provided .
finally , section [ secconcl ] summarizes the results we obtained , underlining the main implications and delivering hints on possible future applications .
we consider two modes @xmath0 and @xmath1 described by the vector @xmath2 of quadrature operators @xmath3 , @xmath4 , satisfying the canonical commutation rules that can be expressed as @xmath5=i\omega_{jk}$ ] , where @xmath6 a two mode gaussian state @xmath7 can be represented in phase space by a gaussian wigner function @xmath8 where @xmath9 is the radius vector in phase space , @xmath10 is the vector of phase - space displacements and @xmath11 is the covariance matrix ( cm ) with elements @xmath12 $ ] , @xmath13 .
any cm @xmath11 has to satisfy the constraints @xmath14 and @xmath15 ( positivity and uncertainty principle ) to ensure that it is associated to a physical gaussian state .
the cm contains a complete information about the correlations in a given gaussian state @xcite . by means of local symplectic ( unitary on the hilbert space ) operations , which leave correlations and entropic quantities invariant
, the cm of a two mode gaussian state can always be reduced to a simple standard form @xmath16 where we can assume @xmath17 without any loss of generality .
states with @xmath18 are said to be symmetric , while states with @xmath19 constitute the important family of _ squeezed thermal states_. pure two mode gaussian states are special instances of symmetric squeezed thermal states , with zero temperature ( @xmath20 ) , i.e. , they are locally equivalent to two mode squeezed vacuum states , with standard form covariances @xmath21 with @xmath22 being the squeezing parameter .
gaussian states can be produced , manipulated and detected in the laboratory with a high degree of control @xcite .
some measurements such as photon counting turn a gaussian state into a non - gaussian one . on the other hand ,
gaussian measurements play a special role as being those that map gaussian states into gaussian states .
these measurements coincide with the standard toolbox of linear optics , i.e. , can be realized using beam splitters , phase shifters , squeezers , appending auxiliary vacuum states , and performing balanced homodyne detection ( bhd ) .
any such a measurement is described by a positive operator valued measure ( povm ) of the form @xcite @xmath23 here the seed element @xmath24 is a normalized density matrix of a generally mixed single
mode gaussian state with cm @xmath25 and zero displacements , @xmath26 stands for the displacement operator , where @xmath27 and @xmath28 is a vector of certain linear combinations of the measurement outcomes of bhds .
the povm ( [ povm ] ) satisfies the completeness condition @xmath29 where @xmath30 , following from schur s lemma @xcite and the normalization condition @xmath31=1 $ ] .
here we briefly review some of the most prominent measures of quantum correlations beyond entanglement , recently proposed to identify the genuinely nonclassical portion of the total correlations in generally mixed states of bipartite quantum systems .
achieving a proper understanding of the structure and nature of correlations in relevant systems is important for gaining insights into foundational aspects of quantum theory , and is particularly imperative in view of practical applications , as quantum correlations yield the key resources to overcome classical systems in quantum information protocols . for completeness
, we also recall the definition of the entanglement of formation .
_ quantum mutual information._the total amount of ( classical and quantum ) correlations in the state of a bipartite quantum system can be reliably quantified in terms of the quantum mutual information @xcite @xmath32 where @xmath33 is the von neumann entropy and @xmath34 are the reduced states of subsystems @xmath0 and @xmath1 , respectively .
the quantum mutual information @xmath35 of a generic gaussian state @xmath36 can be easily calculated using the formula for the von neumann entropy of an @xmath37-mode gaussian state @xmath36 @xcite , @xmath38 , where @xmath39 are the symplectic eigenvalues @xcite of the cm of the state and @xmath40 for two mode gaussian states @xmath7 , the two global symplectic eigenvalues @xmath41 are defined by @xmath42 , with @xmath43 , see eq .
( [ gamma ] ) @xcite .
the mutual information of a two mode gaussian state with cm @xmath11 as in eq .
( [ gamma ] ) is thus @xmath44 with @xmath45 defined in eq .
( [ effe ] ) .
_ one - way classical correlations and quantum discord._ for any bipartite state whose correlations are purely classical , the mutual information can be equivalently expressed in two alternative forms @xmath46 with @xmath47 being the quantum conditional entropy associated with the post - measurement density matrix @xmath48/p_i$ ] , obtained upon performing the povm @xmath49 on system @xmath1 ( @xmath50 $ ] ) ; the optimization over the povms is necessary to single out the least disturbing measurement to be performed on one subsystem , so that the change of entropy on the other subsystem yields a quantifier of the correlations between the two parts . for arbitrary bipartite quantum states @xmath7 , including all entangled states and almost all separable states as well @xcite , the three quantities in eqs .
( [ mi ] ) and ( [ j ] ) evaluate to different results in general , with @xmath51 , and the @xmath52 quantities can be interpreted as ` one - way classical correlation ' measures @xcite . such a discrepancy
is now recognized as a signature of nonclassicality of the correlations in a given state , and the difference between total correlation [ eq .
( [ mi ] ) ] and one - way classical correlation [ eq . ( [ j ] ) ] defines what ollivier and zurek baptized as the ` quantum discord ' @xcite , @xmath53 quantum discord is an asymmetric measure of quantum correlations which has recently received its first operational interpretations in terms of the quantum state merging protocol @xcite and has spurred a great body of research triggered by the investigation of its potential role as the key resource for mixed - state quantum computation @xcite .
the presence of the optimization over local measurements in the definition of discord makes its analytical evaluation very hard for general bipartite states .
no closed formulas are known for @xmath54 on arbitrary two qubit mixed states , other than special cases @xcite .
very recently , a gaussian version of quantum discord has been defined @xcite , where the optimization is restricted to gaussian povms of the type eq .
( [ povm ] ) , and its closed expression has been derived for arbitrary two mode gaussian states @xcite . as a consequence of that analysis
, it was established that the only classically - correlated gaussian states are product states , which are completely uncorrelated .
however , in the limit of diverging mean energy there can exist two mode gaussian states that asymptotically approach so - called classical - quantum states @xcite , where @xmath55 while @xmath56 @xcite . a symmetrized version of quantum discord or ` two - way discord '
can be defined as @xmath57 and in this form becomes vanishing if and only if a state is purely classically correlated @xcite . throughout the paper ,
@xmath54 will in general denote the gaussian quantum discord @xcite , unless explicitly stated .
_ measurement - induced disturbance._ in order to overcome the difficulties involved in the evaluation of quantum discord , luo introduced the ` measurement - induced disturbance ' ( mid ) as an alternative nonclassicality indicator for bipartite quantum states @xcite .
mid is motivated by the observation that in classical systems , local measurements do not induce disturbance .
in particular , a bipartite state containing no quantum correlations is left invariant by the action of any bi - local complete measurement .
on the other hand , even when a state @xmath7 is a priori nonclassical , any complete bi - local measurement makes it classical as a result of a decoherence - by - measurement process @xcite .
mid is thus defined by restricting the attention to the bi - local complete projective measurement @xmath58 determined by the eigen - projectors @xmath59 of the marginal states @xmath60 ( @xmath4 ) , where @xmath61 are corresponding eigenvalues , and reads @xcite @xmath62\,,\ ] ] where @xmath63 is the post - measurement state after local measurements @xmath64 and @xmath65 and @xmath66 $ ] is the probability of obtaining the outcome @xmath67 .
the post - measurement state is obviously fully classical which implies that its quantum mutual information ( [ mi ] ) coincides with the classical mutual information of the distribution @xmath68 given by @xmath69 @xcite with @xmath70 being the shannon entropy , where @xmath71 and @xmath72 are reduced probability distributions of the distribution @xmath68 .
hence we can rephrase mid as @xmath73 the mid quantifies the quantumness of correlations in terms of the state disturbance after local measurements , but with important differences compared to quantum discord : ( i ) both subsystems are locally probed ; ( ii ) there is no optimization over the local measurements , which are chosen to be the marginal eigen - projectors for every quantum state .
although such a quantity is easily computable in arbitrary - dimensional systems , and has found widespread applications in several investigations @xcite , a number of studies have pointed out that mid is clearly an unfaithful and non - refined measure of the nonclassicality of correlations in bipartite states @xcite , being nonzero and even maximal for states approaching the classical limit , and thus severely overestimating quantum correlations . to date , the mid has not been computed for gaussian states .
_ classical mutual information and ameliorated measurement - induced disturbance._ to cure this major drawback of mid , one can define an ` ameliorated measurement - induced disturbance ' ( amid ) by incorporating into eq .
( [ mid2 ] ) an optimization ( precisely , a minimization ) over the joint bi - local povm measurement @xmath74 on subsystems @xmath0 and @xmath1 @xcite .
the amid then can be defined as @xcite @xmath75 where @xmath76 is the classical mutual information of a quantum state @xmath7 @xcite , where @xmath77 is the classical mutual information of the joint probability distribution @xmath78 $ ] of outcomes of local measurements @xmath79 and @xmath80 on @xmath7 .
the amid captures the quantumness of bipartite correlations as signaled by the minimal state disturbance after optimized local measurements . as such
, it is a symmetric , strongly faithful nonclassicality measure @xcite that vanishes if and only if a bipartite state @xmath7 is genuinely classically correlated @xcite , and it is operationally interpreted as the quantum complement to the classical mutual information [ eq .
( [ ic ] ) ] , while the latter is in turn a _ bona fide _ measure of classical correlations in general bipartite quantum states @xcite .
the amid thus englobes the nice properties of discord and mid without showing their genetic weaknesses @xcite .
the evaluation and properties of amid have been recently investigated for two qubit systems @xcite .
_ hierarchy of nonclassical correlations._the three entropic nonclassicality indicators introduced above ( discord , mid and amid ) satisfy the following hierarchical relationship on arbitrary bipartite quantum states @xcite : @xmath81 _ entanglement of formation._ for pure bipartite states @xmath82 , all the measures of nonclassical correlations introduced above ( discord , mid and amid ) reduce to the canonical entanglement measure , the ` entropy of entanglement ' , @xmath83 this shows that quantum correlations are faithfully identified with just entanglement in the special case of pure states of composite quantum systems . for bipartite mixed states @xmath7 ,
let us recall that the entanglement of formation ( eof ) @xmath84 is a well - known entanglement monotone @xcite , defined as the convex roof of the pure - state entropy of entanglement @xmath85 where the minimum is taken over all the pure - state realizations of @xmath7 , @xmath86 for general mixed bipartite gaussian states @xmath7 , one can introduce a gaussian version of the eof ( gaussian eof ) @xmath87 , which is by construction an upper bound to @xmath88 , defined as the convex roof of the entropy of entanglement restricted to decompositions of @xmath7 into pure gaussian states only @xcite .
the gaussian eof can be evaluated via a minimization over cms : @xmath89 where the infimum runs over all pure bipartite gaussian states with cm @xmath90 smaller than @xmath11 .
compact formulas for @xmath87 exist for all symmetric two mode states ( where @xmath91 as the gaussian decomposition is globally optimal ) @xcite , and in the nonsymmetric case for squeezed thermal states and so - called states of partial minimum uncertainty @xcite ; for all the other two mode states its value can still be found analytically @xcite . in the following sections , we will approach the evaluation and characterization of mid and amid for arbitrary two mode gaussian states and their interplay with entanglement , complementing the studies of refs .
@xcite on discord .
in this section we will address the calculation of the mid @xmath92 @xcite , eq .
( [ mid ] ) , for two mode gaussian states . here and in the following
we will always consider , without any loss of generality , states @xmath7 whose cm @xmath11 is in standard form , eq .
( [ gamma ] ) .
the expressions we derive can be straightforwardly recast in terms of a set of four local symplectic invariants for general two mode states , that uniquely define the standard form covariances @xcite .
the quantum mutual information of a two mode gaussian state can be computed via the formula ( [ mig ] ) . here
we will be concerned with the calculation of the mutual information after local projections onto the eigenstates of the marginal density matrices . for a generic two
mode mixed gaussian state @xmath7 in standard form , the reduced states are just thermal states @xmath93 and @xmath94 with mean number of thermal photons @xmath95 and @xmath96 , respectively . the local measurements @xmath97 entering the expression of mid , eq .
( [ mid ] ) , are then _ non - gaussian _ measurements , given by projections onto fock states ( joint photon counting ) , @xmath98 and the post - measurement state reads as @xmath99 where @xmath100 is the joint probability distribution of finding @xmath101 photons in mode @xmath0 and @xmath102 photons in mode @xmath1 .
the determination of the quantum mutual information @xmath103 $ ] of the post - measurement state ( [ pirho ] ) requires the calculation of the local von neumann entropies @xmath104 , @xmath4 of the reduced states @xmath105 $ ] as well as the global entropy @xmath106 $ ] .
since the reduced states are just equal to the local thermal states , i.e. @xmath107 , @xmath4 , we obtain immediately the local entropies equal to @xmath108 and @xmath109 , where @xmath45 is defined by eq .
( [ effe ] ) .
the global entropy can be computed from the joint photon - number probability distribution @xmath110 .
the latter can be derived using the generating function for the distribution in the spirit of ref .
@xcite .
we start by noting that any two mode quantum state @xmath7 can be represented by the complex normal quantum characteristic function defined as @xmath111,\end{aligned}\ ] ] where @xmath112 @xmath113 and @xmath114 @xmath115 are annihilation ( creation ) operators of modes @xmath0 and @xmath1 , and @xmath116 are complex parameters of the characteristic function . for a gaussian state in the standard form ( [ gamma ] ) with zero means
@xmath117 the characteristic function attains the form @xmath118,\end{aligned}\ ] ] where @xmath119 , @xmath120 , @xmath121 , @xmath122 . here
. stands for complex conjugate terms and @xmath123 .
the generating function for the distribution @xmath110 then reads @xcite @xmath124 where @xmath125 and @xmath126 are real parameters . by inserting the characteristic function ( [ cn ] ) into the integral ( [ g ] ) and performing the integration ,
we arrive at the generating function @xmath127 , with @xmath128 where @xmath129/4 $ ] , @xmath130 .
the sought photon - number distribution is obtained by differentiating the generating function ( [ g ] ) as @xmath131 which gives us explicitly @xmath132 where @xmath133!}{(\alpha+\beta - l)!}\nonumber\\ & & \times\left[-4k_{j}\frac{1+b_{1}+b_{2}+k_{j}}{(b_{1}+k_{j})(b_{2}+k_{j})}\right]^{l}.\end{aligned}\ ] ] the global posterior von neumann entropy is then given by the formula @xmath134=-\sum_{m , n=0}^{\infty}p(m , n)\ln p(m , n)\ ] ] and can be evaluated numerically for arbitrary two mode gaussian states .
we note that in refs .
@xcite it is shown that the joint photon - number distribution for multimode gaussian states can be written in terms of multivariable hermite polynomials .
recursive formulas can be derived for the calculation of the multivariable hermite polynomials , and this fact can be exploited to speed up the numerical evaluation of the mid .
the final expression for the mid of two mode gaussian states @xmath135 with a standard form cm as in eq .
( [ gamma ] ) is then @xmath136 - { \cal s}(\hat{\rho}_{ab } ) \\ & = & -\sum_{m , n=0}^{\infty}p(m , n)\ln p(m , n ) - \mathfrak{f}(\nu_+)-\mathfrak{f}(\nu_- ) \nonumber \,.\end{aligned}\ ] ] in the subcase of two mode squeezed states , with @xmath137 , we have @xmath138/4 $ ] and the distribution ( [ pmn ] ) takes the simplified form @xmath139^{j}.\end{aligned}\ ] ] for the very special instance of pure two
mode gaussian states with the standard form given by two mode squeezed vacuum states @xmath140 ) , where @xmath141 , the mid can be evaluated in closed form .
the state is in schmidt decomposition with schmidt coefficients @xmath142 .
being a pure state , @xmath143 .
the post - measurement state ( [ pirho ] ) then reads @xcite @xmath144 and its von neumann entropy is @xmath145=-\sum_{n}\lambda_{n}^2\ln\lambda_n^2 = \mathfrak{f}[\cosh(2r)]$ ] , which precisely coincides with the entropy of entanglement of the pure two mode gaussian state , eq .
( [ ee ] ) .
therefore , @xmath146&= & e[|\psi(r)\rangle_{ab}]=\cosh^{2}(r)\ln[\cosh^{2}(r)]\nonumber\\ & & -\sinh^{2}(r)\ln[\sinh^{2}(r)]\,,\end{aligned}\ ] ] as expected from the definition of mid . let us remark that this value is attained by local non - gaussian measurements ( joint photon counting ) .
an interesting question is whether there exist local gaussian povms that can result in a measurement - induced disturbance equal to the one in eq .
( [ midpuri ] ) for pure two
mode gaussian states .
surprisingly enough , we will prove in the next section that the answer is negative .
given the crucial role of gaussian states in quantum information processing , it is important to carry out a thorough analysis and comprehensive characterization of their quantum and classical correlations .
a primal step to undertake is to approach the evaluation of faithful measures , such as the classical mutual information @xmath147 and correspondingly the amid @xmath148 [ eqs .
( [ amid],[ic ] ) ] , for arbitrary two mode gaussian states @xmath7
. however , the problem appears formidable if any possible non - gaussian povm is allowed .
therefore , similarly to what has been done for discord @xcite , one can define a gaussian version of the amid , @xmath149 , where the optimization is constrained to local gaussian povms of the form ( [ povm ] ) , as @xmath150 where @xmath151 is the gaussian classical mutual information of the quantum state @xmath7 .
the true amid , optimized over general local measurements , would be then bounded from above as @xmath152 a relevant and natural question , related to the one in the closing of the previous section , is whether gaussian measurements are always optimal for the evaluation of the amid , i.e. , whether @xmath153 for all bipartite gaussian states . in the case of discord , there are subclasses of two mode gaussian states where one can prove that the gaussian discord achieves the global minimum in eq .
( [ discord ] ) even when including potentially non - gaussian measurements on a local subsystem @xcite : the properties of the gaussian discord allow us to conjecture that this might be true for arbitrary two mode gaussian states , although no rigorous investigation is available to date to support this claim .. ] remarkably , we will find instead that when both parties are locally probed , there is a finite - volume set of two mode gaussian states ( notably including pure states ) for which non - gaussian measurements are necessarily optimal for the evaluation of the amid , and the gaussian amid only provides a strict upper bound to it .
this proves that gaussian joint measurements may not be the least disturbing ones on general two mode gaussian states .
we hereby develop the framework for the determination of the gaussian amid on general two mode gaussian states , and provide closed formulas for it in some special cases .
the nontrivial part in the determination of the quantity ( [ gamid ] ) is the calculation of the classical mutual information @xmath154 requiring maximization of the shannon mutual information @xmath77 over local gaussian povms @xmath79 and @xmath80 of the form ( [ povm ] ) .
we begin by observing that in the considered optimization task we can restrict ourselves to covariant gaussian povms ( [ povm ] ) projecting onto pure states ( rank - one povms ) , similarly to discrete - variable scenarios @xcite .
the proof of this statement is provided in appendix [ secapppovm ] .
we can thus focus on local measurements of the form ( [ povm ] ) , where the seed state @xmath155 is a pure single mode gaussian state with cm @xmath25 .
let us recall that any pure - state one mode cm @xmath156 can be expressed as @xmath157 , where @xmath158 where @xmath159 and @xmath160 . in this picture ,
_ homodyne detection _ on mode @xmath161 is recovered in the limit of an infinitely squeezed pure state @xmath162 , i.e. , @xmath163 . on the other hand
, _ heterodyne detection _ on mode @xmath161 corresponds to @xmath164 .
now we want to maximize the shannon mutual information @xmath77 of the distribution ( [ pd ] ) , @xmath165,\ ] ] over all single mode pure - state cms @xmath166 . expressing the two mode state cm @xmath11 in block form as in eq .
( [ gamma ] ) , and using the formula for the shannon entropy of a gaussian distribution @xmath167 of @xmath37 variables with classical correlation matrix @xmath168 , @xmath169 $ ] @xcite , the sought mutual information can be obtained in the form @xcite : @xmath170.\ ] ] since the determinant is invariant with respect to symplectic transformations we can assume the cm @xmath11 to be in standard form , eq .
( [ gamma ] ) ; moreover , given eq .
( [ effe ] ) , eq .
( [ uv ] ) , the invariance of the determinant under orthogonal transformations , and the monotonicity of the logarithmic function , the object to be maximized reads @xmath171 where the cm @xmath90 has the @xmath172 blocks of the form @xmath173 we recall that the gaussian amid takes then the form ( [ gamid ] ) with @xmath174\,.\ ] ] the determinant in the denominator of eq .
( [ f ] ) can be expressed in terms of the local symplectic invariants @xmath175 , @xmath176 , @xmath177 and @xmath178 as @xmath179 @xcite . evidently , the function ( [ f ] ) depends on the phases @xmath180 only through invariant @xmath181 so we can optimize @xmath182 over the phases by optimizing @xmath181 . since @xmath181 is always nonnegative , the maximum of the function @xmath182 will be obtained if the phases will maximize @xmath181 .
three different cases must be distinguished depending on the values of the eigenvalues @xmath183 of the matrix @xmath184 .
+ 1 . if @xmath185 , then @xmath186 , which implies @xmath187 and therefore @xmath188 .
since also @xmath189 , we get finally @xmath190 . obviously , product states have neither quantum nor classical correlations .
actually , it is known that any non - product two mode gaussian state has nonzero quantum correlations , so we must expect be @xmath191 as soon as one of the covariances in the block @xmath184 is nonzero .
2 . if @xmath192 and @xmath193 , then for @xmath194 the optimal phases are @xmath195 and for these phases the invariant @xmath196 takes the value @xmath197 .
let us define @xmath198 we have @xmath199 and the function @xmath200 ( and consequently @xmath201 ) is obviously maximized in the limits @xmath202 ( doubly homodyne detection ) when we get @xmath203 since @xmath204 is also larger than the maximal values of the function @xmath201 on the boundaries @xmath205 or @xmath206 , we arrive at the conclusion that ( [ ghom ] ) is the optimal value of eq .
( [ f ] ) for states with @xmath193 .
, therefore the optimal gaussian measurement consists in the local homodyne detection of the quadrature with the highest intermodal correlation . ]
3 . in the general case @xmath207
, the invariant @xmath181 can be expressed as @xmath208 , where @xmath209 is a real symmetric positive - semidefinite matrix of the form : @xmath210 , with @xmath211 defined in eq .
( [ uv ] ) , and the diagonal matrices @xmath212 and @xmath213 defined in eq .
( [ blockprime ] ) .
expressing the matrix @xmath209 through eigenvalue decomposition @xmath214w^{t}(\phi)$ ] , where @xmath215 is an orthogonal matrix diagonalizing @xmath209 and @xmath216 are the eigenvalues of @xmath209 depending on the angle @xmath217 , we can maximize @xmath181 over the phase @xmath218 .
further , if we substitute into the obtained formula the explicit forms of eigenvalues @xmath219 , we can perform also the maximization over the angle @xmath217 , which finally yields the invariant @xmath181 maximized over the phases @xmath180 of local measurements of the form : @xmath220\left[b+\cosh\left(2r_{b}\right)\right]\right.\nonumber\\ & & \left.+\sinh\left(2r_{a}\right)\sinh\left(2r_{b}\right)\right\}\left(c_{1}^{2}+c_{2}^{2}\right)\nonumber\\ & & + \left\{\left[a+\cosh\left(2r_{a}\right)\right]\sinh\left(2r_{b}\right)\right.\nonumber\\ & & + \left.\left[b+\cosh\left(2r_{b}\right)\right]\sinh\left(2r_{a}\right)\right\}\left(c_{1}^{2}-c_{2}^{2}\right).\nonumber\\\end{aligned}\ ] ] the corresponding value of @xmath182 is denoted by @xmath221 . for a generic two
mode mixed gaussian state it is again convenient to express the function @xmath221 in terms of @xmath222 as in eq .
( [ gviah ] ) , being @xmath223 where @xmath224 + since @xmath225 as follows from the inequality @xmath226 , we have to maximize @xmath200 .
introducing new variables @xmath227 and @xmath228 , the extremal points of @xmath200 can be found by solving the stationarity conditions @xmath229 and @xmath230 , respectively , leading to a set of coupled polynomial equations of the form : @xmath231\mu -c_{2}^{2}(a\lambda+1)^{2}=0,\nonumber\\ c_{1}^{2}(b+\mu)^{2}\lambda^{2}+[c_{1}^{2}a(b+\mu)^{2}-c_{2}^{2}a(b\mu+1)^{2}\nonumber\\ + c_{1}^{2}c_{2}^{2}b(\mu^{2}-1)]\lambda -c_{2}^{2}(b\mu+1)^{2}=0.\end{aligned}\ ] ] upon solving the first equation as a quadratic equation with respect to @xmath232 and substituting the obtained roots into the second equation , one arrives after some algebra at a single @xmath233-order polynomial in the variable @xmath234 that we do not write here explicitly due to its complexity . by taking its real roots calculated numerically together with stationary points on the boundary and picking the one for which @xmath200 is maximized , we can finally get the optimal squeezing parameters @xmath235 of the seed elements @xmath236 of optimal local povms [ eq . ( [ povm ] ) ] maximizing the classical mutual information and thus attaining the gaussian amid , eq .
( [ gamid ] ) , of a generic two
mode gaussian state .
analytical progress in the calculation of @xmath149 can be achieved for special classes of two mode states , detailed in the following . 1 .
_ symmetric states._for symmetric states with @xmath18 the maximization of @xmath237 can be in principle performed analytically . as the optimal solution is clearly symmetric with @xmath238 , we have only one stationarity condition @xmath239 , where @xmath240 is the parameter to optimize over .
after some algebra the stationarity condition boils down to the following fourth - order polynomial equation @xmath241 where @xmath242 the equation ( [ exsym ] ) can be solved analytically using the cardan formulae but the obtained solutions are rather cumbersome and therefore we do not give them here explicitly . calculating the values of the function @xmath243 in the admissible real solutions of eq .
( [ exsym ] ) and also in the stationary points on the boundary , the point in which the function is maximal gives us the sought optimal squeezing .
2 . _ squeezed thermal states._for generally nonsymmetric squeezed thermal states with @xmath244 , the optimal gaussian povms can be derived in a simple closed form by performing the maximization of @xmath200 in eq .
( [ hh ] ) .
we find the following results .
if the state parameters @xmath245 and @xmath246 satisfy the inequality @xmath247 , then the optimality is obtained by homodyne detection ( @xmath202 ) on both modes giving eq .
( [ ghom ] ) ; conversely , if @xmath248 , then the optimality is obtained by heterodyne detection ( @xmath249 , projection onto coherent states ) on both modes giving @xmath250^{2}.\ ] ] summarizing , the gaussian amid of two mode squeezed thermal states is given by eq .
( [ gamidg ] ) with @xmath251 } , & ( a+b+1)^2\geq ab(ab - c^2 ) ; \\ g_{\rm het}\ \hbox{[{eq.~(\ref{ghet } ) } ] } , & \hbox{otherwise . }
\end{array } \right.\ ] ] 3 .
_ pure states._. for pure two mode squeezed vacuum states with @xmath252 , the first case in eq .
( [ fgsts ] ) is always satisfied , and doubly homodyne measurements are optimal for the calculation of the gaussian amid .
although this result is quite intuitive ( one could guess that in the pure - state case the optimal local gaussian measurements possessing maximum shannon mutual information for distribution of their outcomes would be homodyne detections ) , the corresponding value of @xmath149 is strictly bigger than the entropy of entanglement , eq .
( [ ee ] ) , which corresponds to the true amid @xmath148 globally optimized over joint , possibly non - gaussian local povms as in the definition ( [ amid ] ) ; and we know from eq .
( [ midpuri ] ) that the latter is indeed attained by local photon counting : @xmath253 .
namely , @xmath254 \,,\ ] ] which is strictly bigger than the expression in eq .
( [ midpuri ] ) .
the latter finding entails a novel and interesting showcase in which non - gaussian local measurements on gaussian states can lead to the extraction of larger correlations than any pair of local gaussian measurements if quantified by mutual information . in the case of pure two
mode squeezed states this behavior can not be simply attributed to the fact that , while the state encodes perfect correlations between photon numbers ( as expressed by @xmath255 for @xmath256 ) , there are imperfect correlations between quadrature operators ( as expressed by a nonzero einstein - podolsky - rosen variance @xmath257 ) .
this is because , while such a variance decays with increasing squeezing @xmath22 , the gap between the gaussian amid @xmath149 ( obtained upon local homodyne detections ) and the true amid @xmath258 ( obtained upon local fock projections ) , for pure states , is a monotonically increasing function of @xmath22 converging to @xmath259 in the limit @xmath260 . a full numerical comparison between gaussian amid , mid and gaussian discord for two mode gaussian states will be provided in the next section .
here we deploy a comprehensive comparative analysis of mid @xmath92 [ eq .
( [ mid ] ) , eq .
( [ midg ] ) ] gaussian amid @xmath149 [ eq . ( [ gamid ] ) , eq . ( [ gamidg ] ) ] , and two - way gaussian quantum discord @xmath261 [ eq . ( [ twoway ] ) , ref . @xcite ] , as tools to quantify the quantumness of correlations in arbitrary two mode gaussian states via entropic descriptions of the state disturbance following suitable local measurements on one or both local parties .
the results of the previous sections show that in ineq .
( [ boundamid ] ) , either quantity on the right hand side can be the smallest on particular instances of two mode gaussian states , suggesting that the subset of two mode gaussian states whose true amid @xmath262 , eq . ( [ amid ] )
, is necessarily optimized by non - gaussian measurements , might have a finite volume in the space of general two mode gaussian states . to confirm this interesting feature and to investigate the tightness of the hierarchy established in eq .
( [ hierardy ] ) , we have generated a large number of random two
mode gaussian states ( up to @xmath263 ) , and for each of them we have evaluated the three symmetric nonclassicality indicators @xmath261 , @xmath149 , and @xmath92 , following the prescriptions of the preceding sections . the resulting analysis is illustrated in fig .
[ figrandall ] .
panel ( a ) shows that , while the mid can be arbitrarily larger than the gaussian amid in principle , there is nonetheless a finite region in the ( @xmath149 , @xmath92 ) diagram that allocates two mode gaussian states for which even non - optimized non - gaussian measurements ( specifically , photon counting ) result in a larger classical mutual information , hence minimize the quantum correlations in the definition of the amid , compared to the optimal gaussian povms . in this study
we did not consider other non - gaussian measurements apart from photon counting ( that enters in the definition of mid because gaussian mixed states are essentially thermal states ) : therefore , we can expect that the region in which general non - gaussian measurements are optimal for the amid can be in principle much larger than the one highlighted by the present study ( that is located below the blue online segment of equation @xmath264 in fig . [ azz ] ) .
still , our finding is perhaps one of the most striking instances of an operational quantum informational measure for gaussian states that can gain a significant optimization by the use of suitable non - gaussian operations .
non - gaussian operations can sometimes reveal quantumness more accurately , thus unleashing more precisely the available nonclassical resources , than the best gaussian measurements , on certain two mode gaussian states , including quite remarkably all two mode pure gaussian states .
notice however that the rigidity in choosing the non - gaussian measurements for the evaluation of mid , excluding any optimization procedure , results in most of the cases into a very loose overestimation of the quantum correlations , as testified by the unbounded region above the straight blue line , filled by states where certainly joint photon counting is not optimal for the amid .
interestingly , pure states embody the lower bound ( dashed black curve ) in fig .
[ azz ] : they are therefore the states where the gaussian amid realizes the most dramatic overestimation of the true amid , that nonetheless can never exceed @xmath265 as computed in the previous section .
a family of states sitting on the blue line in fig . [ azz ] will be characterized shortly . before that , let us comment on the other panels of fig .
[ figrandall ] .
panel ( b ) shows as expected , and in full analogy with the case of two qubits @xcite , that in general the unoptimized mid based on photon counting is a very loose upper bound to quantum discord for two mode gaussian states ( reducing to it on pure states , depicted as dashed black again ) , unbounded from above and relentlessly approaching arbitrarily large values even for states with nearly vanishing quantum correlations as quantified by the ( gaussian ) discord .
this should discourage the usage of mid in general as it almost always provides overestimations , rather than reliable quantifications , of nonclassicality of bipartite correlations .
the last panel shows also a somehow analogous situation to the two qubit case @xcite : the gaussian amid is intimately related to discord , and admits upper and lower bounds at a given value of the two - way gaussian discord .
the lower ( blue online ) boundary in panel ( c ) accommodates states for which the two quantifiers give identical prescriptions for measuring quantum correlations .
these are states with cm in standard form , eq .
( [ gamma ] ) , given by @xmath266 in the limit @xmath260 .
they are characterized by @xmath267 .
pure states fill once more the dashed black curve , for which @xmath268 , eq .
( [ midpuri ] ) , but @xmath149 is strictly bigger , eq .
( [ gamidpuri ] ) . the upper ( red online ) boundary in fig .
[ azz3 ] can be spanned for instance by symmetric squeezed thermal states ( @xmath269 , @xmath270 ) , with @xmath271 and @xmath272 .
upper and lower boundaries ideally conjoin asymptotically for diverging discord and gaussian amid
. we can now analyze in detail the competition between the mid associated to photon counting ( typically very loose , but optimal on pure states ) and the gaussian amid ( very accurate for mixed and strongly correlated gaussian states ) to maximize the classical mutual information , hence minimizing the amid , eq .
( [ boundamid ] ) , on two mode gaussian states .
we believe it be relevant to focus on the class of two mode symmetric squeezed thermal gaussian states ( @xmath273 ) , for which the involved measures can be simply evaluated .
we plot in fig .
[ figcross ] [ ( a),(b ) ] a comparison of the three quantumness measures studied in this work as a function of the rescaled state parameters @xmath274 and @xmath246 .
we see that there is a certain threshold value @xmath275 beyond which the gaussian povms are no longer optimal for the amid , and non - gaussian measurements such as photon counting ( via mid ) provide a more accurate result , culminating in the extreme case of pure states where those specific measurements are globally optimal .
panel ( c ) depicts the threshold in the parameter space , highlighting the region where our analysis conclusively reveals the necessity of non - gaussian measurements for the global optimization of amid and classical mutual information of the considered class of two mode gaussian states .
as previously remarked , this region can be in principle ( and is likely to be so ) much larger . yet
, it certainly occupies a finite volume in the space of general two mode gaussian states .
notice that for all the gaussian states in such a region , non - gaussian measurements allow one to extract stronger correlated measurement records compared to any bi - local gaussian measurement , as the classical mutual information is maximized by non - gaussian detections .
the states attaining the threshold identified in this analysis , are an instance of states filling up the blue line in fig .
[ azz ] . for the two mode gaussian states of eq .
( [ statistrani ] ) .
these curves give origin to the dotted magenta lines in fig .
[ figrandall](a),(b ) .
all the quantities plotted are dimensionless.,width=302 ] finally , we exhibit an example family of two mode gaussian states where , on the opposite end , the mid based on non - gaussian detections is a highly inaccurate measure of quantum correlations .
these states sit on the quasi - vertical dashed ( magenta online ) curves in fig .
[ figrandall](a ) and ( b ) .
they are symmetric states with @xmath276 as apparent from fig .
[ figstrani ] , their gaussian discord and gaussian amid stay limited ( smaller than @xmath277 ) and rigorously vanish in the asymptotic limit @xmath278 . on the other hand ,
their mid arising from fock projections increases arbitrarily and diverges for @xmath278 , embodying an extreme overestimation of some vanishing quantum correlations .
clearly there will be many more families of gaussian states where such a behavior will arise .
here we present a numerical comparison between nonclassicality of correlations , measured by means of the gaussian amid [ eq . ( [ gamid ] ) ] , and entanglement , quantified by the gaussian eof [ eq . ( [ geof ] ) ] @xcite , for generally mixed two mode gaussian states .
a similar analysis was performed in ref .
@xcite , with ( gaussian ) discord used as a nonclassicality indicator . versus gaussian eof @xmath87 for @xmath279 random two
mode gaussian states . the dashed line of equation @xmath280 stands as a lower bound for the physically admitted region . refer to the main text for details of the other curves .
all the quantities plotted are dimensionless.,width=302 ] fig .
[ figrefent2d ] shows the distribution of gaussian amid versus gaussian eof for a sample of @xmath279 randomly generated two
mode gaussian states . in analogy with the case of gaussian discord vs gaussian eof , it is possible to identify upper and lower bounds on the gaussian amid @xmath149 at fixed entanglement @xmath87 .
interestingly , our numerical exploration shows that for all two mode gaussian states @xmath7 , it is @xmath281 this provides a novel hierarchical relationship between different types of nonclassical resources , entanglement @xmath87 , and more general measurement - induced quantum correlations @xmath149 : on the basis of the hereby employed measures ( both symmetric by construction and restricted to a fully gaussian scenario ) , the latters appear to always encompass and exceed entanglement itself for two mode generally mixed gaussian states .
a similar relationship does not hold for discord , which can be smaller as well as larger than entanglement of formation , even in a gaussian scenario @xcite .
we can provide two families of two mode gaussian states for which eq .
( [ majoriz ] ) becomes asymptotically tight .
one such class is provided , e.g. , by symmetric squeezed thermal states , whose standard form cm is as in eq .
( [ gamma ] ) with @xmath282 where @xmath283 the gaussian eof of these states ( equal to the true eof minimized over all possible pure - state decompositions , by virtue of the symmetry of the states @xcite ) is a simple monotonically decreasing function of the positive parameter @xmath284 , @xmath285-(1-\tilde{\nu } ) ^2 \ln\left[\frac{(1-\tilde{\nu } ) ^2}{4 \tilde{\nu } } \right]}{4 \tilde{\nu } } \,,\ ] ] if @xmath286 , and @xmath287 otherwise .
the gaussian amid can be computed analytically according to the prescription of sec .
[ secgamid ] , and for any fixed @xmath284 ( i.e. , fixed gaussian eof ) one can find the optimal value of the parameter @xmath274 , in the range defined by eq .
( [ grange ] ) , that minimizes @xmath149 .
the resulting @xmath149 as a function of @xmath87 is plotted in fig .
[ figrefent2d ] as a green curve : for this family of states , the gaussian amid approaches the gaussian eof as the latter tends to zero , tending to saturate ineq .
( [ majoriz ] ) asymptotically in the regime of infinitesimal correlations .
furthermore , let us consider another class of symmetric two mode states , whose standard form cm is as in eq .
( [ gamma ] ) with @xmath288 where the parameter range is the same as in eq .
( [ grange ] ) .
the gaussian eof of these states is still given by eq .
( [ eofsym ] ) , while in the limit @xmath278 one can show that their gaussian amid is attained by homodyne detections on both modes , yielding @xmath289 .
the corresponding gaussian amid vs gaussian eof curve for these states is depicted in fig .
[ figrefent2d ] as a blue curve : in this case , the gaussian amid approaches the gaussian eof as the latter tends to infinity , also saturating ineq .
( [ majoriz ] ) asymptotically .
it has to be underlined that the two presented gaussian families are just examples to show that the bound in eq .
( [ majoriz ] ) can be asymptotically tight , and we remark that the combination of the two presented curves does not provide a strict lower bound to gaussian amid against gaussian eof for all two mode gaussian states , as it is evident from the presence of some random points below the intersection of the two ( green and blue ) curves however above the dashed line corresponding to the bound of eq .
( [ majoriz ] ) [ see fig .
[ figrefent2d ] ] . on the other hand , a tight upper bound on the gaussian amid at fixed gaussian eof can be identified .
we found it numerically to be constituted by the maximum of two branches , @xmath290 where @xmath291 is given by eq .
( [ eofsym ] ) . the first expression in eq .
( [ gubound ] ) corresponds to the gaussian amid of the states of eq .
( [ cmivette ] ) with @xmath292 and @xmath293 , and is depicted as a black curve in fig .
[ figrefent2d ] .
it provides an upper bound for all two mode gaussian states distributed in the @xmath294 plane , in the region of moderate entanglement ; such a bound converges to 1 in the separability limit ( @xmath295 ) .
we can thus conclude that the degree of nonclassicality of correlations in separable gaussian states is always nonzero ( apart from the trivial case of uncorrelated , product states ) but stays nevertheless limited : it can at most reach unity , whether measured by the discord @xcite or by the gaussian amid . the second expression in eq .
( [ gubound ] ) corresponds instead to the gaussian amid of the states of eq .
( [ gmems ] ) in the limit @xmath278 .
it bounds from above the value of @xmath149 for all two mode gaussian states with fixed @xmath296 [ where this number is obtained by setting @xmath297 in eq .
( [ eofsym ] ) ] . in the limit of infinite entanglement ( @xmath298 ) ,
the upper bound on the gaussian amid converges to @xmath299 . combining this observation with the lower bound ( [ majoriz ] )
, we have that , interestingly , the following sandwich relation holds for all two mode gaussian states with @xmath300 , @xmath301 in the previous analysis , we have identified some similarities as well as some key differences in the quantification of nonclassical correlations versus entanglement of gaussian states when gaussian amid rather than quantum discord are employed . in order to have a visual comparison between the two nonclassicality indicators and the gaussian eof , we focus on the relevant two - parameter class of symmetric squeezed thermal states @xmath302 , with cm as in eq .
( [ gmems ] ) . for these states ,
the entanglement is given by eq .
( [ eofsym ] ) ( independently of @xmath274 ) , while the discord can be written as @xcite @xmath303\ ! \bigg\ } , \end{split}\ ] ] and the gaussian amid reads @xmath304 [ eq .
( [ eofsym ] ) ] versus gaussian discord @xmath305 [ eq .
( [ stsdis ] ) ] and gaussian amid @xmath149 [ eq . ( [ stsgam ] ) ] for two
mode symmetric squeezed thermal gaussian states , characterized by their covariance parameters @xmath274 and @xmath284 [ see eq .
( [ gmems ] ) ] .
the solid black line accommodates pure states ( @xmath306 ) .
all the quantities plotted are dimensionless.,width=302 ] fig .
[ fig3d ] shows @xmath87 plotted versus @xmath307 and @xmath149 for this particular class of gaussian states , spanned by @xmath274 and @xmath284 .
all two mode symmetric squeezed thermal states sit on a two - dimensional surface in the space of the three entropic nonclassicality indicators , providing a direct evidence of the intimate yet intricate relationship between the different aspects of quantumness in gaussian states .
one can notice the branch of separable states , in the plane @xmath308 with generally nonzero discord and gaussian amid ; while for all entangled squeezed thermal states , @xmath87 can be exactly recast as a function of @xmath309 and @xmath149 : knowledge of two nonclassicality quantifiers fixes the third one univocally .
we remark that such a strict result does not extend to more general two mode gaussian states , which distribute filling a more complex , finite - volume three - dimensional region in the space @xmath310 .
figures [ azz3 ] and [ figrefent2d ] of this paper , and figure 1(right ) of ref .
@xcite , represent the two - dimensional projections of such a region onto the planes @xmath311 , @xmath312 , and @xmath313 , respectively .
we have performed an exhaustive study of nonclassical correlations in generic two
mode gaussian states using information - theoretic nonclassicality quantifiers , in particular the ` measurement - induced disturbance ' ( mid ) and its gaussian - optimized version , the gaussian ` ameliorated measurement - induced disturbance ' ( gaussian amid ) . for a given gaussian state
the mid is a gap between its quantum mutual information quantifying the total correlations and the classical mutual information of outcomes of local fock - state detections ( detections of local eigenprojectors ) that captures a specific type of non - gaussian classical correlations in the state .
the gaussian amid is , on the other hand , a gap between the quantum mutual information and the maximal classical mutual information that can be obtained by local gaussian measurements , the latter quantifying the maximum classical correlations that can be extracted from the state by local gaussian processing .
an analytical form of the gaussian amid can be derived for the important subclasses of symmetric states and squeezed thermal states which include pure states , while for a generic mixed gaussian state one has to find roots of a higher - order polynomial in a single variable , which can be solved efficiently by numerical means .
further analysis reveals that mid is mostly larger than gaussian amid and therefore overestimates the amount of nonclassical correlations .
in fact , for a fixed value of gaussian amid , it is possible to find states with an arbitrarily large mid , even if the gaussian amid is infinitesimally small . on the other hand
, there also exists a volume of gaussian states encompassing pure states for which the mid is strictly smaller than the gaussian amid , which surprisingly unveils the importance of non - gaussian measurements for the correct assessment of the amount of nonclassical correlations in gaussian states .
we have further compared the mid and gaussian amid with the two - way gaussian discord .
we found that again there is no upper bound on mid for a fixed value of discord but a close upper bound ( as well as a lower one ) does exist for the gaussian amid .
finally , we have also compared the gaussian amid with the gaussian entanglement of formation , identifying lower and upper bounds for the former as a function of the latter . in particular
, the gaussian amid turns out to always exceed the gaussian entanglement of formation for all two mode gaussian states , enforcing a novel hierarchy between two different forms of nonclassicality .
exact relations between gaussian amid , gaussian discord and gaussian entanglement of formation can be formulated for special families of gaussian states , such as the symmetric squeezed thermal states . on a more technical side
, we have shown that symmetric measures of ( non)classicality of correlations such as the classical mutual information and the gaussian amid are tractable by analytical tools for important subclasses of states even considering optimization over local generalized gaussian measurements .
our results also demonstrate that non - gaussian processing for correct quantification of ( non)classical correlations in gaussian states is in order .
we believe that these findings will inspire further research on the characterization of quantum correlations in the gaussian scenario and beyond .
we are grateful to a. datta and m. paternostro for very fruitful discussions on the topics of this paper .
l. m. , r. t. , and n. k. acknowledge the eu grant under fet - open project compas ( 212008 ) .
l. m. has been supported by projects `` measurement and information in optics , '' ( msm 6198959213 ) and center of modern optics ( lc06007 ) of the czech ministry of education and the project of gacr no .
r. t. and n. k. are grateful for the support from supa ( scottish universities physics alliance ) .
here we prove that covariant rank - one povms , i.e. povms of the form eq .
( [ povm ] ) with a pure seed state @xmath155 , are optimal among all gaussian povms for the evaluation of the classical mutual information or , equivalently , of the gaussian amid , eq .
( [ gamid ] ) .
the measurement of local povms ( [ povm ] ) on a gaussian state @xmath7 gives the outcome @xmath314 distributed according to @xmath315.\ ] ] making use of the overlap formula for wigner functions @xcite , the distribution can be expressed as @xmath316 substituting into the formula from eq .
( [ w ] ) , where we set @xmath317 , and performing the integration , we obtain the distribution ( [ pd1 ] ) in the form : @xmath318 where @xmath166 are cms of the seed elements of povms ( [ povm ] ) and @xmath11 is the cm of the state @xmath7 . the cms @xmath166 can be expressed as @xmath319 , where @xmath320 is a pure - state cm ( @xmath321 symplectically diagonalizes @xmath156 ) and @xmath322 is a positive - semidefinite matrix ( @xmath323 is a symplectic eigenvalue of @xmath156 ) .
therefore , the outcome @xmath324 , @xmath4 of a generic povm @xmath325 , with the seed element @xmath162 being a mixed state with cm @xmath156 , can be expressed as @xmath326 , where @xmath327 is the outcome of povm with pure - state seed element having cm @xmath328 and @xmath329 are mutually uncorrelated random variables uncorrelated with @xmath330 obeying gaussian distributions with classical correlation matrices @xmath331 and @xmath332 , respectively . since such processing of variables @xmath327 can not increase their shannon mutual information due to what is now known as data processing inequality ( derived first for continuous random variables in @xcite ) , we can restrict without loss of generality to optimization over projections onto pure states .
note , that covariant measurements ( [ povm ] ) with pure - state seed elements maximize classical mutual information even within the framework of a larger class of generally noncovariant gaussian povms possessing the structure @xcite @xmath333 and satisfying the completeness condition @xmath334 .
here @xmath335 is a normalized distribution of the parameter @xmath336 , @xmath337 is a normalized gaussian state with cm @xmath338 dependent on parameter @xmath336 and @xmath339 .
upon measuring the povm ( [ noncovariant ] ) on the gaussian state @xmath7 , one finds the outcomes @xmath340 and @xmath341 to follow the distribution @xmath342 , where the distribution @xmath343 is obtained from eq .
( [ pd ] ) by replacing @xmath156 with @xmath338 . denoting the classical mutual information of the distribution @xmath344 and @xmath343 as
@xmath345 $ ] and @xmath346 $ ] , respectively , one then has @xmath347&=&\int{\cal i}[a(y_{a}):b(y_{b})]\mathop{\pi}_{j = a , b}p(y_{j})dy_{j}.\nonumber\\\end{aligned}\ ] ] hence it follows immediately that @xmath348\nonumber\\ & = & { \cal i}[a(y_{a}^{0}):b(y_{b}^{0})],\end{aligned}\ ] ] where we have to maximize over all gaussian povm elements @xmath349 and @xmath350 , @xmath4 label the povm elements @xmath351 , which maximize @xmath346 $ ] .
if we take these as seed elements of povms ( [ povm ] ) , we construct local covariant povms which give mutual information @xmath352 $ ] and therefore achieve the classical mutual information @xmath154 .
a. datta , s. t. flammia , and c. m. caves , phys .
a * 72 * 042316 ( 2005 ) ; a. datta , and g. vidal , phys .
a * 75 * , 042310 ( 2007 ) ; a. datta , a. shaji , and c. m. caves , phys . rev .
lett * 100 * , 050502 ( 2008 ) .
k. modi , m. williamson , h. cable , and v. vedral , arxiv:1003.1174 ; b. eastin , arxiv:1006.4402 ; f. f. fanchini , m. f. cornelio , m. c. de oliveira , a. o. caldeira , arxiv:1006.2460 ; a. brodutch and d. r. terno , phys .
a * 83 * , 010301 ( 2011 ) .
j. oppenheim , m. horodecki , p. horodecki , and r. horodecki , phys .
* 89 * , 180402 ( 2002 ) ; m. horodecki , j. oppenheim , and r. horodecki , _ ibid . _ * 89 * , 240403 ( 2002 ) m. piani , m. christandl , c. e. mora , and p. horodecki , _ ibid . _ * 102 * , 250503 ( 2009 ) ; k. modi , t. paterek , w. son , v. vedral , and m. williamson , _ ibid . _ * 104 * , 080501 ( 2010 ) ; b. daki , c. brukner , and v. vedral , _ ibid . _
* 105 * , 190502 ( 2010 ) ; s. luo , and s. fu , phys .
a * 82 * , 034302 ( 2010 ) .
s. l. braunstein and p. van loock , rev .
phys . * 77 * , 513 ( 2005 ) ; n. cerf , g. leuchs , and e. s. polzik ( eds . ) , _ quantum information with continuous variables of atoms and light _ ( imperial college press , london , 2007 ) .
b. m. terhal , m. horodecki , d. w. leung , and d. p.divincenzo , j. math . phys .
* 43 * , 4286 ( 2002 ) ; d. p. divincenzo , m. horodecki , d. leung , j. smolin , and b. m. terhal , phys . rev . lett . * 92 * , 067902 ( 2004 ) .
a. datta and s. gharibian , phys . rev .
a * 79 * , 042325 ( 2009 ) ; a. datta , phys . rev .
a * 80 * , 052304 ( 2009 ) ; r. srikanth , s. banerjee , and c. m. chandrashekar , phys . rev .
a * 81 * , 062123 ( 2010 ) ; a. auyuanet and l. davidovich , phys . rev .
a * 82 * , 032112 ( 2010 ) .
|
we study quantum correlations beyond entanglement in two mode gaussian states of continuous variable systems , by means of the measurement - induced disturbance ( mid ) and its ameliorated version ( amid ) . in analogy with the recent studies of the gaussian quantum discord , we define a gaussian amid by constraining the optimization to all bi - local gaussian positive operator valued measurements .
we solve the optimization explicitly for relevant families of states , including squeezed thermal states .
remarkably , we find that there is a finite subset of two mode gaussian states , comprising pure states , where non - gaussian measurements such as photon counting are globally optimal for the amid and realize a strictly smaller state disturbance compared to the best gaussian measurements . however , for the majority of two mode gaussian states the unoptimized mid provides a loose overestimation of the actual content of quantum correlations , as evidenced by its comparison with gaussian discord .
this feature displays strong similarity with the case of two qubits .
upper and lower bounds for the gaussian amid at fixed gaussian discord are identified .
we further present a comparison between gaussian amid and gaussian entanglement of formation , and classify families of two mode states in terms of their gaussian amid , gaussian discord , and gaussian entanglement of formation .
our findings provide a further confirmation of the genuinely quantum nature of general gaussian states , yet they reveal that non - gaussian measurements can play a crucial role for the optimized extraction and potential exploitation of classical and nonclassical correlations in gaussian states .
|
dense hadronic matter is of interest to a number of fundamental problems that range from nuclear physics to astrophysics .
qcd at finite baryon density is notoriously difficult : ( 1 ) the introduction of a chemical potential causes most lattice simulations to be numerically noisy owing to the sign problem ; ( 2 ) the baryon - baryon interaction is strong making most effective approaches limited to subnuclear matter densities . in the limit of a large number of colors @xmath5
, qcd is an effective theory of solely mesons where baryons appear as chiral skyrmions .
dense matter in large @xmath5 is a skyrmion crystal with spontaneous breaking of chiral symmetry at low density , and restored or stripped ( overhauser ) chiral symmetry at high density .
while some of these aspects can be studied qualitatively using large @xmath5 motivated chiral models @xcite , they still lack a first principle understanding . the ads / cft approach has provided a framework for discussing large @xmath5 gauge theories at strong coupling @xmath9 from first principles @xcite .
a particularly interesting ads / cft construction is the holographic and chiral approach proposed by sakai and sugimoto @xcite ( ss model ) . in the limit where @xmath10 , chiral qcd is obtained as a gravity dual to @xmath11 d8-@xmath1 embedded into a @xmath12 background in 10 dimensions where supersymmetry is broken by the kaluza - klein ( kk ) mechanism .
the kk scale plays the role of the chiral scale .
the ss model yields a first principle effective theory of pions , vectors , axials and baryons that is in good agreement with experiment @xcite .
the ss model at finite temperature has been studied in @xcite and the bayrons in the context of skyrmion and instanton have been worked out @xcite .
recently we have suggested that the ss model can be used to analyze dense hadronic matter at large @xmath5 and strong coupling @xmath13 @xcite . at zero temperature ,
the quark / baryon chemical potential @xmath14 ( or @xmath15 ) is introduced as the boundary value of the @xmath3 brane potential @xmath16 .
the diagonalization of the vector modes in dense matter , enforces vector dominance and yields @xmath17 throughout where @xmath18 is given in @xcite .
while the mode decomposition of @xmath18 on the brane leads to a highly oscillating @xmath19 , we have argued in @xcite that only those modes in @xmath19 below the kk scale should be retained . as a result both brane and meson properties in holographic and dense baryonic matter were discussed .
soon after the posting of this work , several studies appeared addressing the same issue of baryonic matter in holographic qcd including also temperature . in
@xcite it was suggested that the @xmath19 field is instead fixed by the equation of motion on the brane by varying the pertinent dbi action . in the absence of brane `` charges '' the authors in @xcite concluded that only a constant @xmath19 is a solution , with no baryonic effect at zero temperature .
however , the conserved baryonic charge has to be mirrored by the `` charge '' of baryon vertex .
the latter is obtained by considering brane wrapping of @xmath0 within d8-@xmath1 .
the wrapping number is the bulk conservedcharge which is at the origin of a non - constant @xmath19 in bulk .
this point was further developed in @xcite and used to discuss the phase structure of dense and hot holographic matter , albeit in a brane set up without chiral symmetry . in
@xcite it was argued that the additional `` charges '' in bulk upset the smoothness of the dbi surface in bulk , leading to a spiky structure due to the force balancing condition . as a result ,
the embedded branes should _ always _ touch the horizon even for arbitrarily small temperature and/or density , thereby altering totally the phase diagram in @xcite .
this latter point is physically unintuitive .
indeed , the `` charge '' or baryon vertex exists even at zero temperature with no need to connect to any horizon . at finite temperature through the insertion of a black hole this point is developed in details in @xcite . in this paper
we follow on the analysis in @xcite in bulk and at zero temperature ( and low temperature before deconfinement phase transition ) . the dual `` charge '' or baryon vertex
is inserted at the tip of the minimally embedded d8-@xmath1 surface . this way each of d8 , @xmath1 is shared equally , leading to a _
baryon vertex .
since no connecting string is involved , there is no spiky structure involved here . also , there is a one - to - one correspondence between the baryon vertex normalization and the wess - zumino term in the induced chiral dbi action .
thus the boundary chiral skyrmions constructed from the induced dbi action , are dual to _
static _ and _ point - like _ instantons in bulk .
these conditions will be relaxed in a sequel . in many ways ,
this approach complements the original discussion in @xcite . in section 2 , we briefly review the ss model and set up the notations . in section 3
, we introduce the @xmath3 field @xmath20 in bulk and show how the baryon charge density @xmath4 affects its minimal profile . in section 3 and 4 , we construct the bulk hamiltonian and derive the energy density as a function of the identified baryon density .
the energy density is found to grow about quadratically with the baryon density . in section 5 , we summarize the construction of the chiral effective action for pions , vectors and axials at zero density . in section 6 ,
we show how this chiral effective action is modified by the finite `` charges '' in bulk .
a number of meson properties are discussed as a function of the identified baryon number .
our conclusions are in section 7 . throughout
, the canonical formalism will be used .
in this section we summarize the d4/d8-@xmath1 set up for notation and completeness . for a thorough presentation
we refer to @xcite and references therein .
the metric , dilaton @xmath21 , and the 3-form rr field @xmath22 in @xmath5 d4-branes background are given by @xmath23 where @xmath24 , @xmath25 is the compact variable on @xmath26 .
@xmath27 and @xmath28 are the radial coordinate and four angle variables in the @xmath29 direction .
@xmath30 , where @xmath31 and @xmath32 are the string coupling and length respectively .
@xmath33 is the volume of unit @xmath0 and @xmath34 is the corresponding volume form . to avoid a conical singularity at @xmath35 the period of @xmath36 of the compactified @xmath37 direction
is set to @xmath38 in terms of which we define the kaluza - klein mass as @xmath39 the parameters @xmath40 , @xmath41 , and @xmath31 may be expressed in terms of @xmath42 , @xmath43 , and @xmath32 as @xmath44 now , consider @xmath11 probe d8-branes in the @xmath5 d4-branes background . with @xmath45 gauge field @xmath46 on the d8-branes ,
the effective action consists of the dbi action and the chern - simons action : @xmath47 where @xmath48 , the tension of the d8-brane , @xmath49 $ ] ( @xmath50 ) , and @xmath51 is the induced metric on d8-branes : @xmath52du^2 + \left(\frac{r}{u}\right)^{3/2}\!\!\!\ !
u^2 d\omega_4 ^ 2 \ , \end{aligned}\ ] ] where @xmath53 . the effective action of @xmath1 has the same form and the total action of @xmath11 d8-@xmath1-branes has a symmetry @xmath54 which is interpreted as a flavor chiral symmetry of massless quarks . in the ss model baryons
are skyrmions in @xmath2 . however , in the gravity dual they are either an effective fermion degree of freedom in 5-dimensions ( bottom - up ) or an instanton wrapping @xmath12 ( top - down ) and sourcing the baryon current through the chern - simons term @xcite . in both cases ,
the baryon can be treated as a delta function source of the brane gauge field @xmath19 ( which is @xmath55 in @xmath2 ) , which we will use explicitly .
let @xmath57 be a @xmath3 valued background gauge field in bulk .
its boundary value is related to the baryon chemical potential @xcite . in the absence of the source ,
the effective action of the d8-branes ( [ action.0 ] ) becomes @xmath58^{\frac{1}{2 } } , \label{action.1}\ ] ] where @xmath59 and the chern - simons action vanishes .
the equations of motion for @xmath60 and @xmath57 are @xcite @xmath61 = 0,\end{aligned}\ ] ] @xmath62 = 0.\end{aligned}\ ] ] in this paper we consider only the case @xmath63 , sakai - sugimoto s original embedding @xcite , where the d8-branes configuration in the @xmath37 coordinate is not affected by the existence of background @xmath56 .
this corresponds to @xmath64 , the maximal asymptotic separation between d8 and @xmath1 branes . to compare with @xcite we change the variable @xmath65 to @xmath66 through @xmath67
the action ( [ action.1 ] ) is then @xmath68 where we used @xmath63 and @xmath69 .
it is useful to define the dimensionless quantities @xmath70 in terms of which the action is written as to take into account @xmath1 branes as well as d8 branes . ]
@xmath71 where @xmath72 now we introduce the baryon source coupled to @xmath73 through the chern - simons term @xcite as mentioned before .
we assume that baryons are uniformly distributed over @xmath2 space whose volume is @xmath74 . for large @xmath13 ,
the instanton size is @xmath75 @xcite .
it can be treated as a static delta function source at large @xmath5 . for a uniform baryon distribution ,
the source is s_= n_c n_b d^4x dz ( z)_0(z ) .
the equation of motion of @xmath73 is = n_q ( z ) , [ source.action ] which yields @xmath76 where @xmath77 is the quark density and the step function @xmath78 is determined by the symmetry between d8 ( @xmath79 ) and @xmath80 . by integrating once more we get the classical solution @xmath56 @xmath81 we introduce the `` baryon charge chemical potential of a quark '' , @xmath82 , by @xcite @xmath83 this relation also defines @xmath82 as a function of @xmath84 and vice versa .
furthermore we define the baryon chemical potential as @xmath85 in fig.([fig : a0]a ) we plot the profile of @xmath86 in the @xmath87 coordinate and in fig.([fig : a0]b ) we show @xmath82 for various baryon densities .
since we work in the canonical formalism @xmath82 is more like a lagrange constraint . throughout
, the numerics will be carried using the following values @xcite : @xmath88 , @xmath89 , @xmath90mev , and @xmath91mev .
the smallest eigenvalue was calculated to be @xmath92 .
using these five values we can estimate @xmath42 , @xmath93 , @xmath94 , @xmath95 , and @xmath96 : m_kk= 950 mev , g_ym^2
n_c = f_^2 16.71 , 0.0075 , and a=3.7610 ^ 9 mev^4 , b=7.1610 ^ -6mev^-2.the definition of @xmath97 and @xmath93 are different from @xcite by a factor of @xmath98 , but it is consistent with @xcite . in all figures
@xmath4 is normalized to @xmath99 , with @xmath100 the nuclear matter density , n_0 = 0.17 fm^-3 1.310 ^ 6 mev^3 .
consider the action ( [ action.without.vector ] ) with the source term ( [ source.action ] ) , @xmath101 the @xmath56 is an auxillary field with no time - dependence .
it can be eliminated by the equation of motion ( [ eqna0 ] ) and ( [ az.solution ] ) .
the energy is @xmath102 where @xmath74 is short for @xmath103 and we may set @xmath104 . the chemical potential @xmath14 is constrained by the gibbs relation @xmath105 where @xmath106 is the helmholtz free energy which is @xmath107 at zero temperature .
thus @xmath108 which is in agreement with the solution ( [ az.solution ] ) for @xmath109 .
we note that this construction is consistent with @xcite where the grand potential is identified with the dbi action at finite @xmath82 . in terms of the baryon number density @xmath4 ( @xmath110 )
the regularized helmholtz free energy is @xmath111 \ , \label{reg.f}\end{aligned}\ ] ] after subtracting the vacuum value .
the regularized internal energy @xmath65 , pressure @xmath112 and grand potential @xmath113 as a function of baryon number density @xmath4 or the baryon chemical potential @xmath114 are @xmath115 \ , \nn \\ \frac{p(n_b)_{\textrm{reg}}}{a } & = & { \int_{-\infty}^{\infty}\!\ ! } dz\ , k^{2/3 } \left[1 - \frac{1}{\sqrt { \ , 1 + \frac{(n_c n_b)^2}{4 a^2 b } k^{-5/3 } } } \right]\ , \nn \\
\frac{\omega_{\textrm{reg } } ( { \widetilde}{\m_b})}{a v } & = & { \int_{-\infty}^{\infty}\!\ ! } dz\ , k^{2/3 } \left [ \frac{1}{\sqrt { \ , 1 + \frac{(n_c n_b({\widetilde}{\m_b}))^2}{a^2 b } k^{-5/3 } } } - 1 \right]\ , \nn \\ { \widetilde}{\m_b } & = & n_c { \int_{-\infty}^{\infty}\!\!}dz \frac{n_c n_b/4}{\sqrt{{(ab)^2k^2 } + b k^{1/3}(n_c n_b/2)^2 } } \ , \label{plot.functions}\end{aligned}\ ] ] where @xmath116 . in fig.([fig : thermodynamic.potential ] ) we present the numerical plots of these thermodynamic functions with the numerical inputs in section [ sec : chemical ] . for small baryon densities the energy density is quadratic in @xmath117 ( or @xmath118 ) . at large baryon densities
it is of order @xmath119 .
the small density limit can be qualitatively understood by noting that in bulk the @xmath19 configuration for fixed charge is obtained by minimizing the induced dbi action of d8-@xmath1 .
thus only flavor - meson mediated interactions between the point - like baryons are included . in this section
we compute the coefficients in table ( [ t2],[t4 ] ) numerically .
their physical meanings can be read off from the action ( [ s2final ] ) .
we use the same numerical inputs as detailed in section 3 .
j. m. maldacena , _ the large n limit of superconformal field theories and supergravity _ , _ adv .
* 2 * , 231 ( 1998 ) _ , _ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) _ [ arxiv : hep - th/9711200 ] ; for review , see : + o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , _ large n field theories , string theory and gravity _ , phys .
rept . * 323 * , 183 ( 2000 ) [ arxiv : hep - th/9905111 ] .
t. sakai and s. sugimoto , `` low energy hadron physics in holographic qcd , '' prog .
phys . * 113 * , 843 ( 2005 ) [ arxiv : hep - th/0412141 ] .
t. sakai and s. sugimoto , `` more on a holographic dual of qcd , '' prog .
phys . * 114 * , 1083 ( 2006 ) [ arxiv : hep - th/0507073 ] .
h. hata , t. sakai , s. sugimoto and s. yamato , `` baryons from instantons in holographic qcd , '' arxiv : hep - th/0701280 .
o. aharony , j. sonnenschein and s. yankielowicz , `` a holographic model of deconfinement and chiral symmetry restoration , '' annals phys .
* 322 * , 1420 ( 2007 ) [ arxiv : hep - th/0604161 ] ; + a. parnachev and d. a. sahakyan , `` chiral phase transition from string theory , '' phys .
lett . * 97 * , 111601 ( 2006 ) [ arxiv : hep - th/0604173 ] ; + k. peeters , j. sonnenschein and m. zamaklar , `` holographic melting and related properties of mesons in a quark gluon plasma , '' phys .
d * 74 * , 106008 ( 2006 ) [ arxiv : hep - th/0606195 ] .
s. nakamura , y. seo , s. j. sin and k. p. yogendran , `` a new phase at finite quark density from ads / cft , '' arxiv : hep - th/0611021 .
s. kobayashi , d. mateos , s. matsuura , r. c. myers and r. m. thomson , `` holographic phase transitions at finite baryon density , '' jhep * 0702 * , 016 ( 2007 ) [ arxiv : hep - th/0611099 ]
. s. nakamura , y. seo , s. j. sin and k. p. yogendran , `` baryon charge chemical potential in ads / cft , '' to appear s. k. domokos and j. a. harvey , `` baryon number - induced chern - simons couplings of vector and axial - vector mesons in holographic qcd , '' arxiv:0704.1604 [ hep - ph ] .
d. k. hong , m. rho , h. u. yee and p. yi , `` chiral dynamics of baryons from string theory , '' arxiv : hep - th/0701276 ; + d. k. hong , m. rho , h. u. yee and p.
yi , `` dynamics of baryons from string theory and vector dominance , '' arxiv:0705.2632 [ hep - th ] ; o. bergman , g. lifschytz and m. lippert , `` holographic nuclear physics , '' arxiv:0708.0326 [ hep - th ] .
j. l. davis , m. gutperle , p. kraus and i. sachs , `` stringy njl and gross - neveu models at finite density and temperature , '' arxiv:0708.0589 [ hep - th ] .
|
in the context of holographic qcd we analyze sakai - sugimoto s chiral model at finite baryon density and zero temperature .
the baryon number density is introduced through compact d4 wrapping @xmath0 at the tip of d8-@xmath1 .
each baryon acts as a chiral point - like source distributed uniformly over @xmath2 , and leads a non - vanishing @xmath3 potential on the brane . for fixed baryon charge density @xmath4
we analyze the energy density and pressure using the canonical formalism .
the baryonic matter with point like sources is always in the spontaneously broken phase of chiral symmetry , whatever the density .
the point - like nature of the sources and large @xmath5 cause the matter to be repulsive as all baryon interactions are omega mediated . through the induced dbi action on d8-@xmath1 , we study the effects of the fixed baryon charge density @xmath4 on the pion and vector meson masses and couplings .
issues related to vector dominance in matter in the context of holographic qcd are also discussed . * the chiral model of sakai - sugimoto at finite baryon density
* + keun - young kim@xmath6 , sang - jin sin @xmath7 and ismail zahed@xmath6 + @xmath6 _ department of physics and astronomy , suny stony - brook , ny 11794 _
+ @xmath8 _ department of physics , bk21 division , hanyang university , seoul 133 - 791 , korea _ +
|
recently , high energy gamma ray observations reveal some excess emissions near the galactic centre .
these excess emissions can not be easily explained by standard physical processes
. one potential origin of such emissions is due to an unusual population of millisecond pulsars @xcite .
however , @xcite point out that the large diffuse signal of gamma ray disfavours the possibility of pulsar emissions . even including both known sources and unidentified sources , the millisecond pulsars can only account no more than 10 percent of the gev excess @xcite .
in fact , the majority of discussions of the gev excess is now focused on the annihilation of dark matter particles @xcite .
it is also because the gamma ray spectrum obtained from fermi - lat data can be well fitted with @xmath1 annihilation channel of dark matter particles @xcite .
the required cross section and rest mass of dark matter particle are @xmath2 @xmath3 s@xmath4 and @xmath5 gev respectively @xcite .
this cross section is consistent with the expected canonical thermal relic abundance cross section ( @xmath6 @xmath3 s@xmath4 ) in cosmology .
furthermore , the inner slope of the radial - depedence of the gamma ray emissions is @xmath7 ( the best fit is @xmath8 ) @xcite , which is consistent with the numerical simulation of dark matter halo structure @xmath9 @xcite . in this letter ,
i show that the @xmath1 annihilation channel can also explain the energy required in x - ray emissions in the galactic centre .
this evidence can further support the dark matter annihilation model and constrain the cross section and rest mass of the dark matter particles .
in the past decade , a large amount of diffuse x - ray data had been obtained by _
chandra , bepposax , suzaku _ and _ xmm - newton _ @xcite . in particular , @xcite use the data from _ chandra _ to model the temperature of the two components within 20 pc as 0.8 kev ( soft component ) and 8 kev ( hard component ) .
the energy required to sustain the 0.8 kev and 8 kev components are @xmath10 erg / s and @xmath11 erg / s respectively @xcite .
later , @xcite point out that the cooling by adiabatic expansion may not be important if the hard component could actually be a gravitationally confined helium plasma .
therefore , the actual energy required for the hard component to balance the radiative cooling would be @xmath12 erg / s within 20 pc @xcite , but not @xmath11 erg / s . although supernova explosions are able to provide such a high energy , it is not possible for supernovae to heat the hard component plasma to such a high temperature @xcite .
also , the correlation between the hard and soft emission suggests that they are produced by related physical processes @xcite .
therefore , the required energy to balance the radiative coolings of both soft and hard component might be given by some other origins .
however , there is no widely accepted mechanism to heat and sustain the plasma to such a high temperautre @xcite .
one potential explanation is that the heating might result from the viscous friction on molecular clouds flowing toward the galactic centre @xcite . here
, i propose that the energy given out by annihilation of dark matter particles can explain the energy requirement of both soft and hard components ( @xmath13 erg / s ) within 20 pc .
although a large amount of energy from annihilation is given out in the form of gamma ray which is nearly transparent to the plasma , a large amount of high energy electrons and positrons can also be produced through the @xmath1 annihilation channel .
the spectrum of positron ( or electron ) energy @xmath14 per one annihilation is shown in fig
. 1 @xcite .
these high energy electrons and positrons would lose their energy and give their energy to the plasma mainly by three different processes : ionization loss @xmath15 , synchrotron loss @xmath16 and inverse compton scattering @xmath17 .
the corresponding energy loss rates are @xcite : @xmath18 @xmath19 and @xmath20 where @xmath21 is the radiation energy density , @xmath22 and @xmath23 are the number density of electrons and magnetic field strength in the plasma respectively .
the total number and total energy of positrons or electrons produced per one annihilation are @xmath24 and @xmath25 gev respectively for @xmath26 gev .
therefore , the average energy for one positron or electron produced is @xmath27 gev , which gives @xmath28 . since @xmath29 @xmath30 @xcite , @xmath31 ev/@xmath3 @xcite and @xmath32 g in the plasma near the galactic centre @xcite , the total energy loss rate is @xmath33 ev / s . the cooling rate would first be dominated by inverse compton scattering and synchrotron loss . when the positron or electron is cooled down to about 1 mev , the cooling rate
would be dominated by ionization loss . as a result ,
the required cooling time is @xmath34 s ( see fig . 2 ) .
for the diffusion process of the positrons or electrons , let s first consider the simple random walk model .
the stopping distance of a high energy positron or electron is @xmath35 @xcite , where @xmath36 is the larmor radius . for @xmath37
g , we have @xmath38 pc , which is very small compared with the size of our interested region ( 20 pc ) .
therefore , the required cooling time is short enough such that the diffusion process is not important .
nevertheless , @xcite suggest that the amplitude of the random magnetic field and the turbulence effect are also important to the diffusion process .
this kind of diffusion can be described by a diffusion coefficient @xmath39 and an index @xmath40 . for a large scale ( greater than 100 pc ) , the ranges of the values are @xmath41 @xmath42 s@xmath4 and @xmath43 @xcite . for the innermost region near the galactic centre ,
the picture is much more uncertain . @xcite
reveal from the analysis of @xmath44-ray observations that a significant reduction of the diffusion coefficient in the inner 10 pc region is found .
they apply two models , namely kraichnan and kolmogorov , to obtain the characteristic diffusion length @xmath45 .
they get @xmath46 pc and @xmath47 pc by using the kraichnan model and kolmogorov model respectively .
therefore , not all of the energy of positrons or electrons is lost due to the cooling process during the diffusion within @xmath45 . the diffusion length for an electron with initial energy @xmath48
is given by @xcite @xmath49^{1/2},\ ] ] where @xmath50 is the energy of the electron after the diffusion of length @xmath45 .
let s consider the lower bounds of the parameters from large scale diffusion @xmath51 @xmath42 s@xmath4 and @xmath52 .
for @xmath53 gev and @xmath54 pc , we get @xmath55 gev , which means over 65% of the energy would be lost during the diffusion process . in the following discussion
, we first neglect the effect of diffusion .
the result would not be affected if the simple random walk model is the correct diffusion model . however
, if either of the other two turbulence models is the correct diffusion model , the calculated cross section would be at most increased by a factor of 1.5 .
the total annihilation rate within @xmath56 pc is given by @xmath57 here , @xmath58 is the dark matter density profile , which is assumed to be the generalized nfw profile @xcite : @xmath59^{-3+\gamma},\ ] ] where @xmath60 gev/@xmath3 , @xmath61 kpc and @xmath62 kpc . in our calculations , we assume @xmath8 , which is the best fit of the observed gamma ray spectrum by fermi - lat @xcite . the total energy loss due to electron - positron pairs is @xmath63 . since the energy required to sustain the soft and hard components is @xmath13 erg / s , we can constrain the parameter space of @xmath64 and @xmath65 by using eq .
( 5 ) ( see fig . 3 ) . in the plot
, we see that the allowed parameter space falls within the range predicted by the annihilation model from fermi - lat data @xmath66 @xmath3 s@xmath4 for @xmath26 gev .
g and @xmath67 ev/@xmath3 .
we have assumed @xmath68 @xmath30 and @xmath69 gev initially.,width=302 ] for consistency checking , the above result also agrees with the total luminosity of gamma ray with energy greater than 500 mev within @xmath70 pc in the galactic centre detected by the egret telescope @xcite .
the total energy released per one annihilation is given by : @xmath71 since the energy carried by neutrinos is negligible @xcite , by using the predicted range of @xmath72 @xmath3 s@xmath4 from x - ray emission and @xmath26 gev , the total luminosity of gamma ray is @xmath73 erg / s , which agrees with the detected luminosity @xmath74 erg / s @xcite .
in this letter , we show that the annihilation of dark matter particles can satisfactorily explain the energy source of soft and hard components of hot plasma . the cross section and rest mass of dark matter particles calculated are consistent with the gamma ray observations and give excellent agreement with the prediction from the @xmath1 annihilation model based on the data from fermi - lat .
the predicted cross section in our calculations ( @xmath75 @xmath3 s@xmath4 ) satisfies all the constraints from cosmic microwave background and low - redshift data @xcite , and also agrees with the canonical thermal relic abundance cross section in cosmology @xmath6 @xmath3 s@xmath4 @xcite . in this model ,
the total number of positrons produced per one annihilation is @xmath76 . for @xmath77 pc
, we have @xmath78 s@xmath4 .
since a positron takes @xmath79 s to loss all its energy , we predict that the total number of high energy positrons within 20 pc is about @xmath80 ( the low energy positrons would annihilate with electrons to produce 511 kev emission line ) .
therefore , the number density of high energy positrons within 20 pc is @xmath81 @xmath30 . since the electron number density is @xmath29 @xmath30 @xcite , the positron to electron ratio becomes @xmath82 , which general agrees with the recent analysis in interstellar medium @xcite .
furthermore , this range of ratio can produce the observed 511 kev luminosity by the accretion of intergalactic material @xcite .
to conclude , we can notice that all the above evidences give a self - consistent picture in the galactic centre and strongly support the annihilation dark matter model .
the rest mass and the cross section could probably be verified by the large hadron collider experiment in the future .
i am grateful to the referee for helpful comments on the manuscript .
abazajian , k. n. , canac , n. , horiuchi , s. , and kaplinghat , m. , 2014 , phys .
d , 90 , 023526 .
belmont , r. , tagger , m. , muno , m. , morris , m. , and cowley , s. , 2005 , apj , 631 , l53 .
bergstrom , l. , bringmann , t. , eriksson , m. , and gustafsson , m. , 2005 , phys .
lett . , 94 , 131301 .
boehm , c. , hooper , d. , silk , j. , casse , m. , and paul , j. , 2004 , phys .
lett . , 92 , 101301 .
borriello , e. , cuoco , a. , and miele , g. , 2009 , phys .
d , 79 , 023518 .
calore , f. , cholis , i. , mccabe , c. , and weniger , c. , arxiv:1411.4647 .
cheng , k. s. , chernyshov , d. o. , and dogiel , v. a. , 2006 , apj , 645 , 1138 .
cirelli , m. , gaggero , d. , giesen , g. , taoso , m. , and urbano , a. , 2014 , arxiv:1407.2173 .
crocker , r. m. , bell , n. f. , balzs , c. , and jones , d. , phys .
d , 81 , 063516 .
daylan , t. , finkbeiner , d. p. , hooper , d. , linden , t. portillo , s. k. n. , rodd , n. l. , and slatyer , t. r. , 2014 , arxiv:1402.6703 .
delahaye , t. , lineros , r. , donato , f. , fornengo , n. , and salati , p. , 2008
d , 77 , 063527 .
fornengo , n. , lineros , r. a. , regis , m. , and taoso , m. , 2012 , jcap , 01(2012 ) , 005 .
fritz , t. k. _ et al .
_ , 2014 , arxiv:1406.7568 .
gordon , c. , and macias , o. , 2013 , phys .
d , 88 , 083521 .
hooper , d. , cholis , i. , linden , t. siegal - gaskins , j. , and slatyer , t. , 2013 , phys .
d , 88 , 083009 .
izaguirre , e. , krnjaic , g. , and shuve , b. , 2014 , arxiv:1404.2018 .
lacroix , t. , boehm , c. , and silk , j. , 2014 , phys .
d , 90 , 043508 .
longair , m. s. , _ high energy astrophysics _ , vol .
2 ( cambridge : cambridge university press , 1994 ) .
lopez - honorez , l. , mena , o. , palomares - ruiz , s. , and vincent , a. c. , 2013 , jcap , 07 , 046 .
madhavacheril , m. s. , sehgal , n. , and slatyer , t. r. , 2014 , phys .
d , 89 , 103508 .
mayer - hasselwander , h. a. , _ et al .
_ , 1998 , astron .
, 335 , 161 .
moore , b. , quinn , t. , governato , f. , stadel , j. , and lake , g. , 1999 , mnras , 310 , 1147 .
muno , m. p. , _ et al . _ , 2004 , apj , 613 , 326 .
navarro , j. f. , frenk , c. s. , white , s. d. m. , 1997 , apj , 490 , 493 .
regis , m. , and ullio , p. , 2008
d , 78 , 043505 .
sakano , m. warwick , r. s. , decourchelle , a. , and predehl , p. , 2004 ,
mnras , 350 , 129 .
sidoli , l. , mereghetti , s. , israel , g. l. , chiapetti , l. , treves , a. , and orlanclini , m. , 1999 , apj , 525 , 215 .
uchiyama , h. , hobukawa , m. , tsuru , t. g. , and koyama , k. , 2013 , pasj , 65 , 19 .
vecchio , a. , vincent , a. c. , miralda - escude , j. , and pena - garay , c. , 2013 , arxiv:1304.0324 .
wolfire , m. g. , tielens , a. g. g. m. , and hollenbach , d. , 1990 , apj , 358 , 116 .
|
recently , an excess of gev gamma ray near the galactic centre has been reported .
the spectrum obtained can be best fitted with the annihilation of @xmath0 gev dark matter particles through @xmath1 channel . in this letter ,
i show that this annihilation model can also solve the mysteries of heating source in x - ray plasma and the unexpected high gamma - ray luminosity .
the cross section constrained by these observations give excellent agreements with both the predicted range by using fermi - lat data and the canonical thermal relic abundance cross section .
[ firstpage ] dark matter
|
scheduling has always been an indispensable part of resource allocation in wireless networks . the seminal work of tassiulas _ et al . _
@xcite , and later @xcite , @xcite considered the case where both channel states and queue lengths are fully available to the scheduler .
it was shown that the maxweight algorithm , which serves the longest connected queue , is throughput - optimal .
subsequently , the maxweight algorithm was found to be throughput - optimal in many other settings as well ( @xcite-@xcite and the references therein ) using tools from lyapunov optimization . some other works ( see @xcite , @xcite , @xcite ) also approach the scheduling problem using convex optimization and dual decomposition techniques . @xcite
even considers the role of imperfect queue length information on network throughput , showing that the stability region does not reduce .
but , in all these cases , accurate information about channel - state is assumed as a modeling simplification . in a real - life network , _
e.g. _ , long term evolution ( lte ) @xcite or ieee 802.16e wimax , the channel - state information fed back to the transmitter can have uncertainty . the primary reason being that although resource - allocation is done at the _ finer granularity _ of a physical resource block ( prb ) , channel - state information is still fed back at the _ coarser granularity _ of a subband , which is a group of prbs .
this is done to reduce the feedback traffic from the users to the base station ( bs ) .
however , this _ averaging _ causes information loss and hence , the resulting uncertainty at the scheduler . moreover , uncertainty might be present in the channel - estimates because of the very process of estimation
some recent works have , hence , tried to model this uncertainty in the channel - estimate .
in @xcite , the authors show that infrequent channel - state measurement , unlike infrequent queue length measurement , reduces the maximum attainable throughput .
@xcite considers the effect of inaccuracy of channel estimation on throughput , but does so assuming a specific probability distribution of the channel - state and does not study the stability of the data queues either .
@xcite attempts at modeling channel- and queue - state uncertainty by considering the case where only heterogeneously delayed information is available at the scheduler .
they however assume knowledge of the channel - state transition probabilities . in @xcite , the authors study scheduling with rate adaptation in a single - hop network with a single channel under channel uncertainty .
they consider cases when the channel estimates are inaccurate but complete or incomplete knowledge of the channel - estimator joint statistics is available at the scheduler .
the authors , however , assume that the channel - estimates are independent across the channels for each user .
delay performance of various wireless systems has also been investigated by many researchers recently . among previous work in the area ,
the authors in @xcite study the problem of opportunistic scheduling of a wireless channel while also trying to minimize the mean delay .
neely @xcite has given a @xmath0 delay bound in the case of on / off channels and a @xmath1 bound for multi - rate channels for the classical maxweight algorithm for a network of size @xmath2 and any traffic input - rate vector within a @xmath3-scaled version of the stability region ( where @xmath4 ) .
subsequent work in @xcite established @xmath0 and @xmath1 delay bounds for the case of single - hop and multi - hop networks , respectively , for on / off channels and under both _
i.i.d . _ and markov modulated arrival traffic scenarios .
@xcite derives lower and upper bounds on the delay in a wireless system with single - hop traffic and general interference constraints .
our contributions of this paper are as follows : * firstly , we model the channel - estimate inaccuracy and characterize the network stability region . compared to @xcite , we allow the channel estimates to have dependence among themselves , which is a more realistic situation in a modern lte or wimax network . besides , we study a multi - channel setup whereas they consider a single channel . *
secondly , we propose two simple maxweight based scheduling schemes that achieve any rate in the interior of the stability region . * thirdly , we derive an @xmath1 delay bound for our system under one of the throughput - optimal policies we propose . *
lastly , we propose a throughput - optimal policy for the network under traffic with heterogeneous quality of service ( qos ) constraints and present some numerical results studying its performance .
the remainder of the paper is organized as follows . in section
ii , we describe the system model and the assumptions made on the arrival and channel processes .
section iii provides the network stability region .
we also discuss an example that illustrates that partial channel information may lead to a loss of throughput . in section iv , we propose two throughput - optimal policies for the network and prove their optimality .
we also present some simulation results in this section studying their performance .
section v provides a bound on the mean packet delay in the network .
section vi gives a throughput - optimal policy for the network under qos constraints and studies its performance .
we conclude with section vii .
consider a multi - user cellular downlink system with @xmath2 users and @xmath5 _ orthogonal _ channels .
the system operates with fixed - size data packets and in synchronized time slots denoted by @xmath6 .
it may , for example , be a cellular downlink orthogonal frequency - division multiple access ( ofdma ) system .
each user has a separate queue for its data and @xmath7 denotes the queue - length ( in terms of packets ) for user @xmath8 in slot @xmath9 where @xmath10 .
we assume an infinite buffer at each queue . @xmath11 denotes the number of exogenous packet arrivals for user @xmath8 in slot @xmath9 .
it is assumed that @xmath12 is _ i.i.d . _ from slot to slot with @xmath13 and with @xmath14 for all @xmath8 .
however , in a particular slot @xmath9 , @xmath11 may be dependent among themselves .
the channel - state for user @xmath8 and channel @xmath15 in slot @xmath9 is denoted by @xmath16 .
we assume that @xmath16 is _ i.i.d . _ from slot to slot and independent across users .
such an assumption holds , for example , in a wireless system like lte where our _ channel _ corresponds to a prb in the lte system .
a prb has 180 khz bandwidth which is close to the coherence bandwidth of the channel for a typical delay spread of 4 - 5 @xmath17s ( sec .
5.3.2 in @xcite ) .
we define channel - state as the maximum number of packets that can be sent over the channel successfully without suffering an outage .
we assume that @xmath18 where @xmath19 is a discrete state - space and @xmath20 .
the scheduler has access to only estimates @xmath21 of the channel - state in slot @xmath9 where @xmath10 and @xmath22 .
these estimates are used by the scheduler to schedule different channels to the users in slot @xmath9 .
the estimates for a particular user may be dependent on each other . this can be used to model the effect of _ averaging _ ( or even calculating any deterministic function , for that matter ) of the channel - gains as done in lte systems which are sent by the users to the bs to reduce the feedback traffic .
the only constraint we impose on @xmath21 is that @xmath23 , where @xmath24 is a discrete set , and that it is _ i.i.d . _ from slot to slot . as a shorthand ,
we shall use @xmath25 , @xmath26 and @xmath27 to denote the queue - length vector , arrival vector and channel - estimate matrix , respectively , at slot @xmath9 .
we use @xmath28 to denote the probability mass function of the random variable @xmath29 .
we also assume that the channel / estimator statistics given by the set of probabilities @xmath30 , @xmath31 and @xmath32 is available at the scheduler .
this can be achieved , possibly , using a mechanism that learns the statistics _ on - the - fly_. we assume that a channel can be allocated to at most one user in a particular slot .
we use the notation @xmath33 and @xmath34 , @xmath22 to denote the user scheduled on channel @xmath15 and the corresponding rate allocated to it , respectively , in the slot @xmath9 .
we can then write the queue evolution equation as : @xmath35 where @xmath36 and @xmath37 in this equation , we have assumed that the packets sent on channel @xmath15 are received successfully if and only if @xmath38 , _
i.e. _ , probability of error is _ _ negligible _ _ with any arbitrarily small probability of error ( provided @xmath16 is less than the shannon capacity ) assuming the physical - layer coding scheme supports it .
for example , we can use an appropriate turbo code or ldpc code for our purpose . ] if @xmath8 transmits at a rate less than or equal to @xmath16 on channel @xmath15 . under these conditions
, @xmath39 is a countable state markov chain . for simplicity
, we will assume it to be irreducible .
but the general case can be easily handled .
we will use the following notation : @xmath40 $ ] is the convex hull and @xmath41 $ ] is the interior of set @xmath42 , @xmath43 is the @xmath44 coordinate vector and @xmath45 and @xmath46 denote @xmath2-dimensional vectors of zeroes and ones , respectively .
let @xmath47 be a lyapunov function .
we define one - slot conditional lyapunov drift as @xmath48 in the following , we provide an upper bound on @xmath49 which will be used later on . for the quadratic lyapunov function @xmath50 , @xmath51 where @xmath52 further ,
if @xmath53 and @xmath20 for each @xmath8 and @xmath15 , there exists @xmath54 such that @xmath55
we first define the notion of stability we use in the paper . a queue @xmath7 is _ strongly stable _ if @xmath56 the network of queues is _ strongly stable _ if each individual queue is strongly stable .
_ strong stability
_ implies positive recurrence of the markov chain @xmath39 . in general , it is a stronger notion than positive recurrence . throughout the paper
, we shall use the term stability " to refer to _
strong stability_. we characterize the network stability region of the system now .
consider the set of stationary policies g that base their scheduling decisions at time @xmath9 only on @xmath57 , @xmath58 and the channel - estimator statistics .
the network stability region is defined to be the closure of the arrival rates that can be stably supported by the policies in g. the network stability region @xmath59 is given by @xmath60},\end{aligned}\ ] ] where @xmath61 ties being broken lexicographically . *
proof * the proof goes along the lines of the proof of proposition 1 in @xcite .
we have included it in appendix a for the sake of completeness .
we illustrate the loss in throughput caused by partial channel information using a simple example .
we show that scheduling schemes that naively trust the channel - estimate fed back as the _ true _ channel - state may perform much worse than the policies that do nt .
consider a system with @xmath62 user and @xmath63 channels .
the @xmath63 channel are assumed independent .
also , @xmath64 and @xmath65 .
suppose we can only observe the arithmetic average of the two channel states and not the individual states .
so , @xmath66 , for @xmath67 .
now , if we take the @xmath29 value to be the true channel - state , it can be easily shown that the mean service provided will be @xmath68 packets per slot . however , due to the special choice of the support set , the @xmath29 values give us complete information about the channel - state of both the channels .
then , it is easy to see that we can provide mean service of @xmath69 packets per slot .
we note that a careful scheduling decision can even double the mean service rate as shown in the example . though the example may appear a bit contrived , numerical studies in the next
section show that performance gains due to _ clever _ scheduling may indeed be substantial in many realistic situations .
in this section , we describe two throughput - optimal policies and also prove their optimality .
even though the _ stat _ policy described in the proof of theorem 1 is throughput - optimal , we require knowledge of the arrival rates @xmath70 for _ stat _ to be able to perform the channel - allocation .
the throughput - optimal policies described here , in contrast , just require the arrival rate vector to _ lie within _ the stability region ( without knowing @xmath70 ) and need only knowledge of the current queue - lengths .
this will be available to a downlink scheduler used at a bs .
both , of course , require knowledge of the channel - estimator statistics .
moreover , as will be seen later in the section , scheduling schemes that naively trust the channel - estimate fed back perform worse than the policies in this section . for notational simplicity
, we shall drop all the slot indices in this section .
lyapunov drift analysis techniques are used to prove the throughput - optimality of the policies in this section .
we consider the maxweight ( mw ) type policy described below . at each slot @xmath9 ,
the channel - estimate @xmath71 is observed , and the decisions @xmath72 and @xmath73 are computed separately , for each @xmath15 , as follows : 1 . to each user @xmath8 , assign rate @xmath74 such that , @xmath75 2 .
schedule the user @xmath72 that maximizes the _ rate - backlog - success - probability product _ : @xmath76 for the sake of completeness , we assume that all ties here are broken lexicographically .
the maxweight policy is throughput - optimal .
* proof * see appendix b. we now analyses an iterative version of the above maxweight policy .
these policies have also been studied in @xcite and @xcite .
the new policy will be referred to as _ imw_. we study this policy because we find that it can give a lower mean delay than _ mw _ in some networks . in the _ imw _ policy
, we allocate the channels sequentially from @xmath62 to @xmath5 in @xmath5 rounds taking into account the channels allocated so far .
the _ virtual _ queue - lengths at the beginning of round @xmath15 are considered for the allocation in round @xmath15 . to aid the analysis
, we use @xmath77 to denote the _ virtual _ queue - length of queue @xmath8 at the beginning of round @xmath15 of the allocation .
@xmath78 is defined to be @xmath79 .
we also assume that the set @xmath19 contains a largest element denoted by @xmath80 .
we can formulate the _ imw _ policy now as follows . at each slot @xmath9 ,
the channel - estimate @xmath71 is observed .
the decisions @xmath72 and @xmath73 are computed sequentially from channel @xmath62 to @xmath5 , as follows : 1 .
start with @xmath81 .
2 . for each @xmath15 , do the following : 1 . to each user @xmath8 , assign rate @xmath74 such that , @xmath75 2 .
schedule the user @xmath72 that maximizes the _ rate - backlog - success - probability product _ : @xmath82 3 . if @xmath83 , stop .
+ else , put @xmath84 , for @xmath85 , increment @xmath15 by @xmath62 and continue . here , as before ,
ties are broken lexicographically .
the iterative maxweight policy is throughput - optimal .
* proof * see appendix c. we present numerical simulation results to compare _ mw _ with _ imw _ and also show the advantage of using channel estimators instead of the average channel gains .
firstly , we consider an _ on / off _ system with @xmath86 , for all @xmath87 and @xmath88 .
we assume that each user estimates its channel - state correctly all the time but feeds back only the sum of the six channel - estimates to the scheduler .
we do this to study the effect of _ averaging _ the channel - estimates as in lte on the stability region .
the _ naive _ scheduling schemes ( maxweight@xcite and ssg@xcite ) calculate the average channel - estimate from the sum fed back , round it down to nearest integer and use that for scheduling , taking it to be the _ true _ channel - state for each of the six channels .
we consider symmetric binomial@xmath89 arrivals with equal rates for all users .
we have simulated the system for @xmath90 slots for values of @xmath91 from @xmath92 to @xmath93 .
the resulting simulated queue backlogs are shown in fig .
[ fig_first_case ] .
we see a huge gain in the stability region compared to the _ naive _ algorithms .
similar results are obtained when the channel - estimates are rounded up to nearest integer instead of down . in the second set of simulations , we consider a multi - rate system with @xmath94 , for all @xmath87 , @xmath88 and @xmath95 . here too , we study the _ averaging _ effect by making the _ naive _ scheduling schemes ( maxweight and ssg ) calculate the average channel - estimate from the sum fed back , round it up to the nearest integer and use that for scheduling .
we simulate the system for @xmath96 slots with symmetric binomial@xmath89 arrivals for values of @xmath91 from @xmath97 to @xmath62 .
[ fig_second_case ] shows the resulting mean queue backlogs .
we again see a decrease in the stability region when we use the _ naive _ schedulers . in this figure , we have also expanded the graph to show that _ imw _ performs better than _ mw _ at least under moderate traffic conditions .
similar results are obtained when the channel - estimates are rounded down instead of up .
we also simulated asymmetric systems and a similar behavior of the corresponding stability regions was observed .
the results are not reported here for lack of space .
in this section , we derive an upper bound on the mean delay for the _ mw _ policy .
assume @xmath98 $ ] .
then , the average delay in the system , denoted by @xmath99 , satisfies the following bound : @xmath100 where @xmath101 and @xmath4 and @xmath102 .
* proof * see appendix d. note that the delay bound derived is @xmath103 since @xmath104 , and @xmath105 and @xmath17 are fixed constants .
in the previous sections , there were no qos constraints on the data .
we only had to ensure that all the queues were stable . in this section ,
we consider three types of traffic : * real - time ( rt ) traffic : demands an upper bound on packet dropping ratio and delay deadline guarantees .
* rate - guarantee ( rg ) traffic : demands minimum rate guarantees . *
best - effort ( be ) traffic : demands only queue stability .
let @xmath106 , @xmath107 and @xmath108 denote the set of rt , rg and be users , respectively , in the system .
the set of all users is denoted by @xmath109 .
we have a slotted system as before .
however , the slots are grouped into frames and each frame consists of @xmath110 consecutive slots .
we assume that the rt packets have a deadline of @xmath110 slots .
these rt packets might be coming , for example , from a voice or a video source .
rg packets coming into the system only require guarantees on minimum rate .
these packets might be coming from a source with flow - control , such as by tcp protocol .
thus , the rg users may be treated , for our purposes , as packet sources with infinite backlogs . the minimum - rate guarantee sought by a rg user @xmath111
is denoted by @xmath112 .
the channel model is the same as before except that we assume @xmath113 , @xmath114 and for each channel @xmath15 .
all exogenous packets arriving into the system arrive _ only _ at the start of the frame and @xmath115 $ ] denotes the number of packet arrivals in frame @xmath116 , @xmath117 for user @xmath118 . as before , @xmath119\}$ ] is _ i.i.d . _ from frame to frame with @xmath120\}=\lambda_{n}$ ] and with @xmath120^{2}\ } < \infty$ ] for all @xmath118 . in case ,
some of the rt packets arriving in a frame could not be served by the end of the frame , they are simply dropped .
we denote the packet dropping ratio for user @xmath121 by @xmath122 , _
i.e. _ , at most @xmath122 fraction of the packets arriving for user @xmath121 can be dropped in the long term . in order to satisfy the qos constraints of the rt and rg users , we use the concept of _ virtual queues _ ( see @xcite , @xcite ) that evolve from frame to frame .
corresponding to each rt user @xmath121 and rg user @xmath111 , we have virtual queue - length processes @xmath123\}_{k=0}^{\infty}$ ] and @xmath124\}_{k=0}^{\infty}$ ] , respectively .
we can then write the queue evolution equations for the rt users as , @xmath125 = ( y_r[k ] - \mu_r[k ] + a_r[k](1-\alpha_r))^{+ } , \forall r \in \mathcal{r},\ ] ] where @xmath126=0 $ ] .
\sum_{t = kt}^{kt+t-1 } \mu_r(t)$ ] , and @xmath128 is as defined in .
similarly , the virtual queues of the rg users are updated as follows : @xmath129 = ( z_g[k ] - \mu_g[k ] + \beta_g)^{+}\ ] ] @xmath130 , where @xmath131=0 $ ] and @xmath132 $ ] is defined just like @xmath127 $ ] . unlike the rt and rg users , the queues of the be users evolve from slot to slot .
besides , be users only maintain real queues since they do not have any qos constraints whatsoever .
@xmath133 denotes the queue - length of be user @xmath134 at slot @xmath9 and its dynamics are governed by the equations , [ be_q_evolution ] @xmath135)^{+},\\ q_b(kt+t+1 ) & = ( q_b(kt+t ) - \mu_b(kt+t))^{+},\end{aligned}\ ] ] @xmath136 in frame @xmath116 .
we also define : @xmath137 = \begin{cases } y_n[k ] , & \text{if $ n \in \mathcal{r}$,}\\ z_n[k ] , & \text{if $ n \in \mathcal{g}$ , and}\\ q_n(kt ) , & \text{if $ n \in \mathcal{b}$. } \end{cases}\ ] ] we consider a modified version of the maxweight type policy described in the previous sections , which we call qos - maxweight ( qmw ) . at each slot
@xmath9 in frame @xmath116 , the channel - estimate @xmath138 is observed , and the decisions @xmath33 and @xmath34 are computed separately for each channel as follows : 1 . to each user @xmath139 and channel @xmath15 ,
assign rate @xmath140 such that , @xmath141 2 .
schedule the user @xmath33 , on channel @xmath15 , that maximizes the _ rate - backlog - success - probability product _ below : @xmath142 \mathbb{p}(x_{nj}(t ) \geqslant r_{nj } | \mathbf{s}(t)=\mathbf{s } ) r_{nj } \}.\ ] ] for the sake of completeness , we assume that all ties here are broken lexicographically .
the qos - maxweight policy is throughput - optimal . * proof * see appendix e. we investigate the effect of partial information on the network stability region using numerical simulation .
we consider a multi - rate system with @xmath94 , for all @xmath143 , @xmath88 and @xmath95 .
there are @xmath63 rt users ( with the maximum dropping ratios and deadlines being @xmath92 and @xmath144 slots for user @xmath62 , and @xmath145 and @xmath144 slots for user @xmath63 ) , @xmath146 rg users ( with minimum rate guarantees being @xmath147 , @xmath63 and @xmath62 packets - per - frame for users @xmath146 , @xmath148 and @xmath147 , respectively ) and the rest @xmath144 are be users .
rt users have binomial@xmath149 , @xmath150 arrivals .
rg users have binomial arrivals with parameters @xmath144 and the corresponding minimum rate .
we have simulated the system for @xmath151 slots with symmetric binomial@xmath89 arrivals for the be users with the value of @xmath91 varying from @xmath152 to @xmath144 .
[ fig_mqos ] shows the resulting mean sum queue backlogs .
we have considered a multi - rate wireless downlink with multiple channels and users .
the base station schedules the traffic based on partial channel information .
we have obtained its network stability region .
we have then proposed two throughput - optimal schemes to achieve the stability region and proved their optimality .
we have also derived a bound on the mean packet delay in the network .
finally , we have proposed a throughput - optimal policy for the network under qos constraints .
a natural extension of our work would be to investigate the effect of uncertainty in the queue - length information .
also , more complicated traffic models are left to be studied in greater detail .
* proof of theorem 1 : * _ sufficiency _ : we show that @xmath98 $ ] is a sufficient condition for stability .
now , since @xmath59 is convex , for each @xmath32 , there exists a scaling vector @xmath153 and a scalar @xmath154 such that @xmath155 for any user @xmath8 , and where @xmath156 .
consider the following stationary randomized policy , henceforth referred to as _
stat _ : for channel estimate @xmath138 , allocate all channels to user @xmath8 with probability @xmath157 and set the rate allocated on channel @xmath15 to it to be @xmath158 . in that case ,
the service rate of user @xmath8 , call it @xmath159 , is given by @xmath160 considering the quadratic lyapunov function defined before and using and in , we can write the lyapunov drift inequality as @xmath161 now , we note that @xmath25 evolves as a markov chain with @xmath47 bounded below by zero and @xmath162 . also , notice that the conditional drift in is negative outside the finite set @xmath163 .
thus , using theorem 2 from @xcite , we can say that the queue - length process is stable .
_ necessity _ : here we show that @xmath164 is a necessary condition for stability .
for if not so , i.e. , if @xmath165 , we have a vector @xmath166 and a scalar @xmath167 , such that for any @xmath168 , we have @xmath169 from the strict separation theorem ( see , for example , proposition b.14 in @xcite ) .
let us define the linear lyapunov function @xmath170 .
now , using definition ( [ delta ] ) and the queue evolution equation ( [ q_evolution ] ) , for any stationary policy in g , we can write @xmath171 let @xmath172 .
we now show that @xmath173 .
@xmath174\\ & = \sum_{\mathbf{s } \in \mathcal{s } } \mathbb{p}_{\mathbf{s}}(\mathbf{s } ) \mathbb{e}\left [ \mu_{i}(t)\ |\ \boldsymbol{q}(t ) ; \mathbf{s}(t ) = \mathbf{s } \right]\\ & = \sum_{\mathbf{s } \in \mathcal{s } } \mathbb{p}_{\mathbf{s}}(\mathbf{s } ) \bigg ( \sum_{j=1}^{m } \mathbb{p}(r_{j}(t ) \leqslant x_{ij}(t)\ |\ \mathbf{s}(t ) = \mathbf{s } ) \mathbf{1}(i_{j}(t)=i ) r_{j}(t ) \bigg)\\ & \leqslant \sum_{\mathbf{s } \in \mathcal{s } } \mathbb{p}_{\mathbf{s}}(\mathbf{s } ) \bigg ( \sum_{j=1}^{m } \mathbb{p}(r_{ij}^{*}(\mathbf{s } ) \leqslant x_{ij}(t)\ |\ \mathbf{s}(t ) = \mathbf{s } ) \mathbf{1}(i_{j}(t)=i ) r_{ij}^{*}(\mathbf{s } ) \bigg).\end{aligned}\ ] ] the second equality holds since the policy decisions @xmath33 and @xmath34 are completely determined by @xmath25 and @xmath27 within the class of stationary policies g. the inequality at the end holds because of the way @xmath158 is defined .
thus , @xmath173 . thus , from the strict separation theorem and ( [ delta2 ] )
, we get @xmath175 it can be shown that @xmath176 .
taking the finite set as @xmath177 and noting that @xmath178 for at least some @xmath25 , we see that the dtmc will _ not _ be positive recurrent , and hence , not _ strongly stable _ either .
thus , we conclude that the network will be unstable .
* proof of theorem 2 : * assume that we are working with @xmath98 $ ] .
mw _ policy make decisions @xmath72 and @xmath73 , and the _ stat _ policy make decisions @xmath179 and @xmath180 .
we then have , @xmath181 @xmath182 and @xmath183 , since _ mw _ maximizes the expression in the lhs over the class g. thus , summing over all @xmath8 and @xmath15 , we get @xmath184 therefore , @xmath185 where we have used the fact that the policy decisions are completely determined by @xmath186 and @xmath29 within the class g. taking expectation w.r.t .
@xmath29 , @xmath187 note that _ stat _ stabilizes the system since @xmath98 $ ] .
thus , for some @xmath154 , we can write @xmath188 where the equality holds because _ stat _ makes its decisions independent of the queue - lengths . multiplying @xmath79 on both sides and summing over all @xmath10 , we get @xmath189 where we have used ( [ deriv1 ] ) in the second inequality .
therefore , @xmath190 finally , using ( [ deriv2 ] ) in the drift inequality ( [ delta1 ] ) , we get @xmath191 now the queue evolves as a markov chain since the scheduling and rate allocation decisions are taken based on the _ current _ queue - lengths and channel - estimate .
the drift inequality above gives negative drift but for a finite set of queue - lengths .
hence , using theorem 2 from @xcite , the network is stable .
* proof of theorem 3 : * assume that we are working with @xmath98 $ ] . let the _
policy make decisions @xmath72 and @xmath73 , the _ mw _ policy make decisions @xmath179 and @xmath180 and the _ stat _ policy make decisions @xmath192 and @xmath193
. for ease of exposition , we use @xmath74 and @xmath194 as the _ intermediate _ allocation variables for _ imw _ and _ mw _ policies , respectively .
we then have , @xmath195 in the first inequality , we have used the fact that the _ virtual _ queue - lengths can not _ increase _ with allocation rounds .
the second inequality is due to the way @xmath196 does the allocation at each round .
summing both sides over all @xmath22 , we get @xmath197 therefore , @xmath198 where we have used the property of indicator functions . now , using the fact that @xmath199 , for all @xmath8 for the @xmath196 policy ( and a similar thing for @xmath200 ) , and proceeding in a way similar to the previous proof , we will get @xmath201 now , using a @xmath202 policy , as in the previous proof , to stabilize @xmath203 $ ] for some @xmath154 , and using the property of _ mw _ for the expression obtained , as done in , we get @xmath204 where we have used the inequality in ( [ deriv3 ] ) . finally ,
rearranging the terms in the above inequality and using it in ( [ delta1 ] ) , we get @xmath205 now , as before , we know that the queue evolves as a markov chain . as before , observe that the drift is negative but for a finite set of queue - lengths .
hence , the network is stable .
* proof of theorem 4 : * here too we drop the slot indices for the sake of brevity .
mw _ policy make decisions @xmath72 and @xmath73 , and any other policy @xmath206 in class g make decisions @xmath179 and @xmath180 .
then , we can write @xmath207 since @xmath98 $ ] , we can say that @xmath208 for some @xmath209 . consequently , @xmath210 . furthermore
, defining @xmath17 as in , we can say that @xmath211 where @xmath212 .
thus , owing to the convexity of @xmath59 , we have @xmath213 hence , we can find a @xmath202 policy such that @xmath214 substituting ( [ deriv5 ] ) into ( [ deriv4 ] ) , and then using ( [ delta0 ] ) , we get @xmath215 finally , using lemma 4.1 of @xcite and the previous inequality , we can write @xmath216 now , we notice that since the system evolves as an ergodic markov chain with countable state - space , the time - averages are well defined , and hence , @xmath217 can be used instead of @xmath218 .
also , note that since @xmath20 and @xmath2 is finite , we can write @xmath219 for all @xmath8 and @xmath102 .
so , @xmath220 .
using this fact and little s theorem , we thus get the delay bound in .
* proof of theorem 5 : * squaring both sides of and proceeding as in appendix f , for each rt user @xmath121 , we get @xmath221 ^ 2 - y_r[k]^2 \leqslant ( a_r[k](1-\alpha_r))^2 + \mu_r[k]^2 + 2 y_r[k ] ( a_r[k](1-\alpha_r ) - \mu_r[k]).\ ] ] similarly , for each rg user @xmath111 , from , we get @xmath222 ^ 2 - z_g[k]^2 \leqslant \beta_g^2 + \mu_g[k]^2 + 2 z_g[k ] ( \beta_g - \mu_g[k]).\ ] ] squaring both sides of , for each be user @xmath134 , we get @xmath223 ^ 2 + \mu_b(kt)^2 + 2 q_b(kt ) ( a_b[k ] - \mu_b(kt)),\ ] ] @xmath224 for @xmath225 .
now , observe that @xmath226 where a finite @xmath227 exists since we assumed that @xmath113 , @xmath114 and for each channel @xmath15 .
so , we can write @xmath228 using the above fact , we can rewrite as @xmath229 summing both sides of the previous inequality over @xmath225 , we get @xmath230 using and the above inequality , we can write @xmath231 ^ 2 + t^2 \mu_{max}^2 + 2 q_b(kt ) ( a_b[k ] - \mu_b[k]),\ ] ] where we have used the fact that @xmath232 and @xmath233 \triangleq \sum_{t=0}^{t-1 } \mu_b(kt+t)$ ] .
consider the quadratic lyapunov function @xmath234 } ) = \sum_{n \in \mathcal{n } } \phi_n[k]^2 $ ] .
we now define the one - frame conditional lyapunov drift as @xmath235 } ) \triangleq { \mathbb{e}\{l({\mathbf{\phi}[k+1 ] } ) - l({\mathbf{\phi}[k]})|{\mathbf{\phi}[k]}\}}.\ ] ] using , and , we can express the drift for the qmw policy as @xmath236 } ) & \leqslant \mathbb{e } \bigg\lbrace \sum_{r \in \mathcal{r } } \left ( a_r[k]^2 ( 1-\alpha_r)^2 + \mu_r[k]^2 \right ) + \sum_{b \in \mathcal{b } } \left ( a_b[k]^2 + t^2 \mu_{max}^2
\right ) \nonumber \\ & \quad + \sum_{g \in \mathcal{g } } \left ( \beta_g^2 + \mu_g[k]^2 \right ) \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace + 2\ \mathbb{e } \bigg\lbrace \sum_{b \in \mathcal{b } } \left ( q_b(kt ) ( a_b[k ] - \mu_b[k ] ) \right ) \nonumber \\ & \quad + \sum_{r \in \mathcal{r } } \left ( y_r[k ] ( a_r[k](1-\alpha_r ) - \mu_r[k ] ) \right ) + \sum_{g \in \mathcal{g } } \left ( z_g[k ] ( \beta_g - \mu_g[k ] ) \right ) \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace.\end{aligned}\ ] ] now , we can find a @xmath54 such that @xmath237 ^ 2 ( 1-\alpha_r)^2 + \mu_r[k]^2 \right ) + \sum_{b \in \mathcal{b } } \left ( a_b[k]^2 + t^2 \mu_{max}^2 \right ) \\ & \quad + \sum_{g \in \mathcal{g } } \left ( \beta_g^2 + \mu_g[k]^2 \right ) \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace\end{aligned}\ ] ] since @xmath238\ } < \infty$ ] and @xmath239 for each @xmath118 and channel @xmath15 .
also , @xmath119\}$ ] is _ i.i.d . _ and @xmath120\}=\lambda_{n}$ ] .
so , we can rewrite as @xmath240 } ) & \leqslant b + 2\ \mathbb{e } \bigg\lbrace \sum_{r \in \mathcal{r } } \left ( y_r[k ] ( \lambda_r(1-\alpha_r ) - \mu_r[k ] ) \right ) + \sum_{g \in \mathcal{g } } \left ( z_g[k ] ( \beta_g - \mu_g[k ] ) \right ) \nonumber \\ & \quad + \sum_{b \in \mathcal{b } } \left ( q_b(kt ) ( \lambda_b - \mu_b[k ] ) \right ) \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace.\end{aligned}\ ] ] rearranging the terms and using the definition , we get @xmath241 } ) & \leqslant b + 2\ \mathbb{e } \bigg\lbrace \sum_{r \in \mathcal{r } } y_r[k ] \lambda_r(1-\alpha_r ) + \sum_{g \in \mathcal{g } } z_g[k ] \beta_g + \sum_{b \in \mathcal{b } } q_b(kt ) \lambda_b \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace \\ & - 2\ \mathbb{e } \bigg\lbrace \sum_{n \in \mathcal{n } } \phi_n[k ] \mu_n[k ] \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace.\end{aligned}\ ] ] from the definition of the qmw algorithm , it can be easily shown that @xmath242 \mu_n[k ] \bigg| { \mathbf{\phi}[k ] } , \mathbf{s}(kt ) , \ldots , \mathbf{s}(kt+t-1 ) \bigg\rbrace \\ & \geqslant \mathbb{e } \bigg\lbrace \sum_{n \in \mathcal{n } } \phi_n[k ] \mu_n^\prime[k ] \bigg| { \mathbf{\phi}[k ] } , \mathbf{s}(kt ) , \ldots , \mathbf{s}(kt+t-1 ) \bigg\rbrace\end{aligned}\ ] ] where @xmath243 $ ] corresponds to any other stationary randomized policy . taking expectation _ w.r.t . _ the channel - estimates in the frame , we get @xmath242 \mu_n[k ] \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace \geqslant \mathbb{e } \bigg\lbrace \sum_{n \in \mathcal{n } } \phi_n[k ] \mu_n^\prime[k ] \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace.\end{aligned}\ ] ] thus , using the above fact , we can rewrite as @xmath240 } ) & \leqslant b + 2\ \mathbb{e } \bigg\lbrace \sum_{r \in \mathcal{r } } y_r[k ] \lambda_r(1-\alpha_r ) + \sum_{g \in \mathcal{g } } z_g[k ] \beta_g + \sum_{b \in \mathcal{b } } q_b(kt ) \lambda_b \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace \\ & - 2\ \mathbb{e } \bigg\lbrace \sum_{n \in \mathcal{n } } \phi_n[k ] \mu_n^\prime[k ] \bigg| { \mathbf{\phi}[k ] } \bigg\rbrace.\end{aligned}\ ] ] rearranging the terms again and using definition , we then get @xmath240 } ) & \leqslant b + 2 \sum_{r \in \mathcal{r } } y_r[k ] \mathbb{e } \ { ( \lambda_r(1-\alpha_r ) - \mu_r^\prime[k ] ) | { \mathbf{\phi}[k ] } \ }
\nonumber \\ & \quad + 2 \sum_{g \in \mathcal{g } } z_g[k ] \mathbb{e } \ { ( \beta_g - \mu_g^\prime[k ] ) | { \mathbf{\phi}[k ] } \ } + 2 \sum_{b \in \mathcal{b } } q_b(kt ) \mathbb{e } \ { ( \lambda_b - \mu_b^\prime[k ] ) | { \mathbf{\phi}[k ] } \}.\end{aligned}\ ] ] now , if the system is stabilizable , _
i.e. _ if @xmath98 $ ] , then a stationary randomized policy exists that stabilizes the system .
besides , this policy makes decisions independent of the queue - lengths . taking the @xmath243 $ ] decisions to be the service decisions corresponding to this stationary randomized policy , we have @xmath244 | { \mathbf{\phi}[k ] } \ } & < - \epsilon , \\ \mathbb{e } \ { \beta_g - \mu_g^\prime[k ] | { \mathbf{\phi}[k ] } \ } & < - \epsilon , \\ \mathbb{e } \ { \lambda_b - \mu_b^\prime[k ]
| { \mathbf{\phi}[k ] } \ } & < - \epsilon,\end{aligned}\ ] ] where @xmath154 .
thus , the previous drift inequality can the simplified as @xmath235 } ) \leqslant b - 2 \epsilon \sum_{n \in \mathcal{n } } \phi_n[k],\ ] ] where we have used definition .
now , as before , the queue evolves as a markov chain .
the drift inequality above gives negative drift but for a finite set of queue - lengths .
hence , using theorem 2 from @xcite , the network is stable .
* proof of lemma 1 : * we are given the lyapunov function @xmath50 . squaring both sides of equation ( [ q_evolution ] ) , and using the fact that @xmath245 , we have @xmath246 summing over all @xmath8 and taking @xmath247 , @xmath248 where @xmath249 is as defined in . since , @xmath53 and @xmath20 for each @xmath8 and @xmath15 , @xmath250 is bounded and independent of @xmath25 .
hence , we are done .
m. andrews , k. kumaran , k. ramanan , a. stolyar , r. vijayakumar , and p. whiting , scheduling in a queuing system with asynchronously varying service rates , " _ probability in the engineering and informational sciences _ , vol .
18 , no . 2 , 2004 s. bodas , s. shakkottai , l. ying , and r. srikant , `` scheduling in multi - channel wireless networks : rate function optimality in the small - buffer regime , '' _ in proc .
acm sigmetrics conf .
_ , jun . 2009 .
a. eryilmaz and r. srikant , joint congestion control , routing and mac for stability and fairness in wireless networks , " _ ieee journal on selected areas in communications _ , vol .
24 , no . 8 , pp .
1514 - 1524 , aug . 2006 .
k. kar , x. luo , s. sarkar , throughput - optimal scheduling in multichannel access point networks under infrequent channel measurements , " _ ieee trans .
wireless commun .
_ , vol . 7 , no . 7 , pp . 2619 - 2629 , jul
. 2008 .
m. j. neely , e. modiano , and c. e. rohrs , power allocation and routing in multibeam satellites with time - varying channels , " _ ieee / acm transactions on networking _ , vol .
138 - 152 , feb . 2003 .
s. donthi and n. b. mehta , joint performance analysis of channel quality indicator feedback schemes and frequency - domain scheduling for lte , " _ ieee transactions on vehicular technology _ ,
60 , no . 7 , pp . 3096 - 3109 , sept .
l. georgiadis , m. j. neely , and l. tassiulas , _ resource allocation and cross - layer control in wireless networks , " _ foundations and trends in networking , now publishers , vol . 1 , no
. 1 , pp . 1 - 149 , 2006 . g. r. gupta and n. b. shroff , delay analysis for wireless networks with single hop traffic and general interference constraints , " _ ieee / acm transactions on networking _ , vol .
393 - 405 , apr . 2010 .
l. b. le , k. jagannathan and e. modiano , delay analysis of maximum weight scheduling in wireless ad hoc networks , " _
43rd annual conference on information sciences and systems _ , pp .
389 - 394 , mar .
e. leonardi , m. mellia , f. neri and m. ajmone marsan ,
bounds on average delays and queue size averages and variances in input - queued cell - based switches , " _
infocom 2001 , anchorage , usa _ , vol.2 ,
1095 - 1103 , 2001 .
l. tassiulas and a. ephremides , stability properties of constrained queueing systems and scheduling policies for maximum throughput in multihop radio networks , " _ ieee trans .
automat . contr .
19361948 , dec . 1992 .
l. m. lopez - ramos , a. g. marques , j. ramos , and a. caamano , cross - layer resource allocation for downlink access using instantaneous fading and queue length information , " _ ieee proc .
mcecn at global communications conf . ,
miami , usa , _ dec . 6 - 10 , 2010 .
a. g. marques , l. m. lopez - ramos , g. b. giannakis , j. ramos , and a. caamano , optimal cross - layer resource allocation in cellular networks using channel and queue state information , " _ ieee trans . on vehic .
tech _ , vol .
61 , no . 6 , pp . 2789 - 2807 , july 2012 .
m. j. neely , e. modiano , and c. e. rohrs , dynamic power allocation and routing for time - varying wireless networks , " _ ieee journal on selected areas in communications _ , vol .
23 , issue 1 , pp .
89 - 103 , 2005 .
|
we consider a scheduler for the downlink of a wireless channel when only partial channel - state information is available at the scheduler .
we characterize the network stability region and provide two throughput - optimal scheduling policies .
we also derive a _
deterministic _ bound on the mean packet delay in the network .
finally , we provide a throughput - optimal policy for the network under qos constraints when real - time and rate - guaranteed data traffic may be present .
this work has been submitted to the ieee for possible publication .
copyright may be transferred without notice , after which this version may no longer be accessible .
|
a promising platform to quantum simulate such novel phenomena of condensed matter physics as topological insulators @xcite and superconductors @xcite is the spin - orbit coupled cold atomic systems @xcite which have drawn great attention in recent years .
raman dressed coupling between atomic pseudo - spin and its linear momentum was first realized by lin and co - workers @xcite in a two - component ( spin-1/2 ) @xmath0rb condensate , and were soon generalized to spin-1/2 degenerate fermi gases of @xmath1k @xcite and @xmath2li @xcite . in this scheme ,
photon recoil associated with the raman transition facilitated by two counter - propagating laser beams changes the center - of - mass momentum of the atom when it jumps from one spin state to the other .
very recently , the same spin - orbit coupling ( soc ) scheme was achieved in a spin-1 condensate @xcite . in general , the physics becomes richer in larger spin systems @xcite simply because more spin states are involved and more control knobs can be utilized . when the two laser beams that induce the raman transition are made to co - propagating , but possess different orbital angular momenta ( e.g. , in the form of laguerre - gaussian , or lg , beams ) @xcite , the raman transition will be accompanied by a transfer of the orbital angular momentum ( oam ) , instead of the linear momentum , to the atom
this situation has been achieved in experiment where this transfer of oam from photon to the atom was exploited to create spin textures in a spinor condensate @xcite .
several recent theoretical proposals also explored this effect to realize spin - orbit - angular - momentum ( soam ) coupling in spin-1/2 condensate @xcite , where interesting quantum states such as the half skyrmion , the meron pair and the annular striped phase , are predicted to exist . in this paper , we provide our study of the soam coupling in a weakly interacting spin-1 condensate .
we show that the interplay between the soam coupling and the quadratic zeeman shift produces a rich phase diagram , and a coherent oscillation between two different many - body quantum phases can be induced by quenching the quadratic zeeman term .
our paper is organized as follows . in sec .
ii , we present the model and discuss the single - particle properties , particularly the single - particle energy spectrum , of the system . in sec .
iii , we focus on the properties of a spin-1 condensate . both the ground state properties and the quench dynamics will be presented . in sec .
iv , we provide a brief summary .
-axis shine on a condensate .
( b ) atomic energy level structure .
( c ) the lowest energy band ( @xmath3 ) in the single - particle spectrum possesses different number of minima , which yield this phase diagram in the @xmath4-@xmath5 space . on the yellow dashed line ,
the single - particle ground state is three - fold degenerate .
( d ) solid lines represent the three degenerate ground state wave functions @xmath6
( @xmath7 , 0 , and 1 ) . here
@xmath8 and @xmath9 .
the dashed lines represent the effective potential @xmath10 experienced by different spin states .
the red numbers in the figure represent the oam quantum number is the lab frame , which equal to @xmath11 , and we take @xmath12 in all our calculations .
all quantities plotted in the figures throughout the paper are expressed in a dimensionless unit system with @xmath13.,scaledwidth=50.0% ] our theoretical model is similar to the one presented in an earlier work for spin-1/2 system @xcite , where two lg beams co - propogating along the @xmath14-axis carrying opposite angular momenta ( @xmath15 and @xmath16 ) shine on a harmonically - trapped condensate , as schematically shown in fig .
[ fig1](a ) .
the relevant atomic energy levels are shown in fig .
[ fig1](b ) .
the two laser beams induce raman transition among the three atomic hyperfine spin states ( denoted as @xmath17 and @xmath18 ) , which form a spin-1 system . in experiment , one may choose , for example , the three zeeman levels in @xmath0rb , e.g. @xmath19 , @xmath20 and @xmath21 in the @xmath22 ground hyperfine manifold @xcite .
we assume that the single - photon detuning is sufficiently large such that the atomic electronically excited states can be adiabatically eliminated . for simplicity
, we also assume that the two lg beams have identical spatial profile and light intensity .
finally , we assume that the atom is tightly confined along the @xmath14-axis which results in a quasi - two - dimensional geometry . under the rotating wave approximation @xcite , by adopting a unit system such that @xmath23 with @xmath24 and @xmath25 being respectively the atomic mass and harmonic trap frequency , the dimensionless single - particle hamiltonian can be written in polar coordinates @xmath26 in the following form @xmath27 \ , .
\label{c1}\end{aligned}\]]here @xmath28 are the spin-1 angular momentum matrices , @xmath29 ^{2}$ ] represents both the raman coupling strength and the ac stark shift , where @xmath30 is a constant proportion to the overall light intensity , @xmath31 characterizes the beam width , and @xmath32 is the generalized laguerre polynomials with azimuthal index @xmath33 determining the optical oam and the radial index @xmath34 describing the radial intensity distribution of the lg beams @xcite .
finally , the parameters @xmath35 and @xmath5 denote the effective linear and quadratic zeeman shifts , respectively .
physically , @xmath36 is related to the two - photon raman detuning , and @xmath5 can be tuned by either an external magnetic field or a microwave field and can be either positive or negative gerbier .
next we introduce a rotating frame which is related to the lab frame by a unitary transformation with the corresponding unitary operator @xmath37 . under this unitary transformation , the atomic states
@xmath38 are transformed to @xmath39 , and the transformed hamiltonian @xmath40 takes the form @xmath41where @xmath42 is the oam operator in the rotating frame , and the term proportional to @xmath43 describes the soam coupling and plays a critical role in our system .
we shall now find the energy spectrum determined by hamiltonian @xmath44 .
obviously , @xmath45 possesses rotational symmetry such that @xmath46 = 0 $ ]
. therefore all the energy eigenstates can be labelled by two quantum numbers @xmath47 where @xmath48 is the oam quantum number , and @xmath49 may be regarded as the radial quantum number . within a given @xmath48 sector
, the lowest energy eigenstate will be assigned @xmath3 .
the eigenenergies and eigenstates can be easily found numerically by taking the anstz @xmath50 it is also easy to see that different spin states @xmath51 ( @xmath7 , 0 , and 1 ) experience different effective potentials @xmath10 given by @xmath52 where the term proportional to @xmath53 arises from the centrifugal barrier .
note that , in the lab frame , the oam quantum number for spin state @xmath51 is @xmath11 .
we will take experimentally interested lg modes @xmath12 and only focus on the two - photon resonant case @xmath54 in our following calculations .
furthermore , we choose @xmath55 indicating that raman beams are red single - photon detuning .
typical energy spectra for @xmath56 and several different quadratic zeeman shift @xmath5 are presented in fig .
one can easily see that all the spectra are symmetric about @xmath57 .
this reflects an additional symmetry which is present only for @xmath54 .
this symmetry is associated with fact that the hamiltonian @xmath45 commutes with an operator @xmath58 , where @xmath59 denotes complex conjugation and @xmath60it is straightforward to show that @xmath61 and @xmath62 = 2\widetilde{t}\widetilde{l}_{z}$ ] . as a result , @xmath63 applying to an energy eigenstate @xmath47 changes the state to a degenerate eigenstate @xmath64 , i.e. , @xmath65 which yields the symmetric spectrum .
one can also observe from fig .
[ fig2 ] that , according to the value of @xmath5 , the spectrum may exhibit a single minimum at @xmath57 , two degenerate minima at @xmath66 , or three local minima at @xmath67 in the @xmath3 band .
accordingly , we plot a ` phase diagram ' in fig .
[ fig1](c ) .
( for very large @xmath68 , the strong ac stark shift confines the atoms along a thin ring and the energy spectrum is dominataed by one single minimum . )
the yellow dashed line within the region with three minima corresponds to the case where all three minima are degenerate . for @xmath56 ,
this occurs at @xmath69 , see fig .
[ fig2](b ) .
the wave functions of the three degenerate states are plotted in fig .
[ fig1](d ) .
if @xmath5 is slightly larger than this critical value , we have a global minimum at @xmath57 and two local minima at @xmath66 ; whereas for @xmath5 slightly smaller this critical value , global minima occur at @xmath70 , as shown in fig .
[ fig2](d ) . .
( a ) , ( b ) and ( c ) correspond to @xmath71 , @xmath9 , and @xmath72 , respectively , where the energy spectrum for the @xmath3 band ( connected with dots ) exhibit two , three and one minimum , respectively .
( d ) a detailed look at the @xmath3 band with three energy minima for three different values of @xmath5 . at @xmath73 ,
all three minima are degenerate.,scaledwidth=48.0% ]
in this section , we shall consider a weakly - interacting spin-1 condensate subject to the soam coupling , for which a mean - field treatment is appropriate .
we work in a parameter regime where the single - particle spectrum exhibits three minima by taking @xmath56 and @xmath74 $ ] .
this is the regime where the effects of the interatomic interaction can be most easily seen .
the interaction hamiltonian for a spin-1 condensate takes the form @xcite @xmath75 \ , , \label{c4}\]]where @xmath76 is the total particle density which obeys the normalization condition @xmath77 , @xmath78 represents the local spin texture , @xmath79 is the spin - independent two - body interaction satisfying @xmath80 ( in our following calculation , we take @xmath81 ) , and @xmath82 is the spin - dependent interaction strength . in principle ,
@xmath79 and @xmath82 can be tuned by optical feshbach resonances @xcite .
the total hamiltonian of the condensate is given by @xmath83 the ground state is obtained by minimizing the total energy .
we study this problem using two different methods the variational method and the fully numerical method based on the gross - pitaevskii ( gp ) equations derived from hamiltonian @xmath84 . for the latter method
, we performed imaginary - time evolution of the gp equations using the backward - euler and fourier pseudo - spectral discretizations toward time and space , respectively @xcite .
the two methods produced results that are in excellent quantitative agreement . in the following ,
we just present our varitional calculation . in the variational calculation
, we assume that the condensate wave function is a linear superposition of the three lowest single - particle states @xmath85 : @xmath86 where the complex amplitudes @xmath87 , @xmath88 and @xmath89 satisfy the normalization condition @xmath90 .
as it turns out , this is a very accurate approximation in the weakly - interacting regime that we are interested in .
for discussion below , we define @xmath91 , @xmath92 and @xmath93 to be the phase angles of these three amplitudes , respectively , i.e. , @xmath94 , etc .
taking @xmath87 , @xmath88 and @xmath89 as variational parameters , we minimize the total energy to obtain the ground state .
the total energy depends on these parameters in a rather complicated way . however , it depends on the three phase angles simply as @xmath95 .
hence , depending on the sign of the coefficient in front of it , the sum of the two angles @xmath96 can only take values 0 or @xmath97 ( mod[@xmath98 ) .
this calculation allows us to distinguish several phases and we present the ground state phase diagram in fig .
[ fig3](a ) .
typical spin density profiles @xmath99 and normalized local spin texture @xmath100 for different phases are presented in fig .
[ fig3](b ) and ( c ) , respectively . for a fixed @xmath101
, we plot the magnitude of the variational parameters and the total spin @xmath102 as functions of @xmath103 in fig .
[ fig3](d ) . in the @xmath103-@xmath5 parameter space we explored , five distinct phases labelled as i , ii , iii , iva and ivb in the phase diagram are found .
these phases result from the competition between the single - particle energies and mean - field interaction .
the latter favors a ferromagnetic state with finite total spin when @xmath104 , and an antiferromagnetic state with zero total spin when @xmath105 .
we describe the properties of these phases below .
phase i this phase lies in the lower left corner of the parameter space . in this region ,
the single - particle ground state is two - fold degenerate with @xmath106 , see fig .
[ fig2](d ) , and the interaction parameter @xmath104 favors a ferromagnetic state . as a result , the atoms condense into one of the single - particle ground states , and the many - body ground state is also two - fold degenerate and maintains rotational symmetry with definite oam quantum number @xmath106 , corresponding to @xmath107 , @xmath108 or @xmath109 , @xmath110 .
the total spin is finite and points along the @xmath14-axis , i.e. , @xmath111 and @xmath112 .
phase ii this phase lies in the upper right corner of the parameter space . in this region ,
the single - particle ground state is non - degenerate with @xmath113 , and the interaction parameter @xmath105 favors an antiferromagnetic state . as a result , the atoms condense into the single - particle ground state , with @xmath114 and @xmath115 , and a vanishing total spin @xmath116 .
phase iii this phase lies in the upper left corner of the parameter space .
here @xmath105 favors an antiferromagnetic state , which results in a vanishing total spin @xmath116 . in this phase ,
all three variational parameters @xmath87 , @xmath88 and @xmath89 are nonzero with @xmath117 and @xmath118 as a result , the many - body ground state does not possess a definite value of @xmath119 and spontaneously breaks the rotational symmetry .
typical spin density profiles are presented in the third column of fig .
[ fig3](b ) , from which one can see that in the phase spin-1 and ( @xmath120 components are immiscible . in this plot , we have chosen @xmath121 and @xmath122 . a different choice of the angles under the constraint of eq .
( [ iii ] ) will result in a collective rotation of all three spin denisty profiles , which is a manifestation of the goldstone mode resulting from the spontaneous symmetry breaking .
phase iv this phase lies in the lower right corner of the parameter space . here
the single - particle ground state is non - degenerate with @xmath113 , which can be regarded as an antiferromagnetic state .
but the interaction @xmath104 is ferromagnetic .
this competition again leads to all three variational parameters @xmath87 , @xmath88 and @xmath89 to be nonzero and @xmath123 similar to phase iii , the many - body ground state in phase iv does not possess a definite value of @xmath119 and spontaneously breaks the rotational symmetry .
however , different from phase iii , here spin-1 and ( @xmath120 components are miscible , as can be seen from the last two columns of fig .
[ fig2](b ) , and the total spin @xmath124 does not vanish .
the spin density profiles for phase iv in fig .
[ fig2](b ) are plotted with @xmath125 .
again , a different choice of phase angles under the constraint of eq .
( [ iv ] ) will result in a collective rotation of all spin density profiles .
in addition , phase iv is the only phase that features a non - vanishing total transverse spin @xmath126 .
the local transverse spin forms a vortex - antivortex pair , see right lower corner of fig .
[ fig3](c ) .
phase iv can be further decomposed into two subphases iva and ivb . in iva which only occupies a rather small parameter space
, we have @xmath111 . by contrast , in ivb , the amplitudes @xmath117 which , together with the phase angle constraint in eq .
( [ iv ] ) , leads to @xmath127 , i.e. , the spin-1 and ( @xmath120 components have identical density profiles , as can be seen in the last column of fig .
[ fig3](b ) . after a detailed discussion of the ground state phase diagram
, we now turn to the study of dynamics .
in particular , we will examine how a sudden change of the quadratic zeeman shift affects the system .
previous studies have shown that the quadratic zeeman shift plays an important role in the quantum dynamics of a spinor condensate without spin - orbit coupling @xcite .
the time evolution of the system is depicted in fig .
initially we prepare the system in the ground state with @xmath128 and an initial quadratic zeeman shift @xmath129 .
this state belongs to phase iii . at @xmath130
, we suddenly quench the quadratic zeeman shift to a final value of @xmath131 and the system starts to evolve .
we solve the time evolution by numerically integrating the time - dependent gp equation with the help of real - time propagation method @xcite .
( red solid curve with dots ) and @xmath132 ( blue dashed curve with hollow circles ) after the quadratic zeeman shift @xmath5 is quenched from @xmath133 to @xmath134 at @xmath130 .
( b ) time evolution of the spin density profiles .
the evolution is roughly periodic with a period of @xmath135 . here
@xmath128.,scaledwidth=48.0% ] in fig .
[ fig4](a ) , we plot the evolution of the magnitude of the total transverse spin @xmath136 which oscillates between the initial value of 0 and a maximum value , and @xmath137 ^ 2 $ ] which oscillates between a finite initial value and 0 .
note that @xmath138 implies @xmath139 , i.e. , identical density profiles for the spin-1 and ( @xmath140 ) components .
the evolution of the spin density profiles is plotted in fig . [ fig4](b ) .
it is quite evident that the evolution is periodic with a period @xmath135 .
the system starts from an initial state that belongs to phase iii . at @xmath141
, it evolves into a phase ivb state .
after another quarter period , it returns to phase iii , and the trend continues .
therefore the quench of @xmath5 induces a rabi oscillation between two many - body quantum phases .
the above behavior obtained numerically can be understood using the variational anstz eq .
( [ c5 ] ) .
the initial condensate wave function is very accurately described by eq .
( [ c5 ] ) with the amplitudes @xmath117 , the phase angles
@xmath142 , @xmath121 , and the single - particle states @xmath143 obtained at the initial quadratic zeeman shift @xmath144 . immediately after the quench , we project the condensate wave function onto the single - particle states corresponding to the final quadratic zeeman shift @xmath145 . to a very good approximation , the condensate wave function
is still dominated by the lowest band ( @xmath3 ) single - particle states with @xmath146 , i.e. , we can write @xmath147 where the amplitudes retain the relation @xmath148 , @xmath149 and @xmath150 . in the ensuing time evolution , if we neglect the weak interaction energy , the condensate wave function will evolve according to @xmath151 with @xmath152 where @xmath153 is the single - particle energy for the state @xmath154 . at @xmath155
, our calculation shows that @xmath156 with @xmath157 .
this leads to a periodic evolution of @xmath158 with period @xmath159 , which has a little discrepancy with the numerically obtained @xmath135 .
this discrepancy can mainly be attributed to the fact that we have neglected the interaction energy in our simple analysis , but the inclusion of the interaction energy would not affect the qualitative physics described here .
furthermore , the amplitudes will satisfy the condition @xmath160 and @xmath161 at @xmath162 , we then have @xmath163 and the condensate evolves into phase ivb [ see eq .
( [ iv ] ) ] , in agreement with the numerical calculation .
we have performed similar quenches starting from initial states within different phases .
the many - body rabi oscillation only occurs between phases iii and ivb .
if the initial state belongs to either phase i or ii , the state is stable in the sense that it retains the rotational symmetry and the initial oam quantum number @xmath119 .
if the initial state is within iva , the post - quench dynamics looks rather complicated , and the system would not evolve into any other phases .
finally , we remark that , instead of a quench of @xmath5 , a quench of the interaction strength @xmath103 can induce qualitatively similar dynamics .
however , from an experimental point of view , quench of @xmath5 is much more feasible , and in fact has been realized in several laboratories @xcite .
motivated by previous experiments and recent theoretical studies of soam coupling in spin-1/2 condensate , we have presented a study of soam coupling in a spin-1 condensate . as we have shown , the enlarged spin degrees of freedom gives rise to much richer physics .
focusing on a regime where the single - particle energy spectrum exhibits a three - minima structure , we mapped out the ground state phase diagram where different phases possess distinct symmetry properties , spin density profiles and spin textures .
we also investigated the dynamics induced by a sudden quench of the quadratic zeeman shift , and found an interesting many - body rabi oscillation between two of the phases .
we have presented a variational analysis , along with a full numerical investigation , to provide a simple intuitive picture that underlies the main physics .
the variational and the numerical calculations are in excellent agreement with each other .
yz is supported by nsf of china under grant nos .
11234008 and 11474189 , the national basic research program of china ( 973 program ) under grant no .
2011cb921601 , program for changjiang scholars and innovative research team in university ( pcsirt)(no .
irt13076 ) .
hp acknowledges support from us nsf and the welch foundation ( grant no .
|
we propose a simple model with spin and orbit angular momentum coupling in a spin-1 bose - einstein condensate , where three internal atomic states are raman coupled by a pair of co - propagating laguerre - gaussian beams .
the resulting raman transition imposes a transfer of orbital angular momentum between photons and the condensate in a spin - dependent way . focusing on a regime where the single - particle ground state is nearly three - fold degenerate ,
we show that the weak interatomic interaction in the condensate produces a rich phase diagram , and that a many - body rabi oscillation between two quantum phases can be induced by a sudden quench of the quadratic zeeman shift .
we carried out our calculations using both a variational method and a full numerical method , and found excellent agreement .
|
the milky way provides astronomers with a unique opportunity to explore the formation and evolution of large spiral galaxies , as well as the nature of their stellar populations and recognized structures .
the key to this understanding comes from the availability , for large numbers of individual stars , of the powerful combination of six - dimensional phase - space information ( location and velocity ) and chemical abundances .
metal - poor stars , in particular , shed light on the early stages of galaxy formation and chemical evolution , as they represent the fossil record of the first generations of stars that formed shortly after the big bang .
although theory suggests that the bulge of the galaxy may harbor numerous ancient ( though not necessarily the most metal - poor ) stars ( e.g. , tumlinson 2010 ) , the vast majority of presently recognized metal - poor stars are found in the halo system of the galaxy . according to carollo et al .
( 2007 ; c07 ) and carollo et al . ( 2010 ; c10 ) , the inner and outer halos possess different peak metallicities ( [ fe / h]@xmath23 @xmath24 ; [ fe / h]@xmath25 @xmath26 ) , as well as different spatial distributions , with the inner halo exhibiting a flatter density profile than the nearly spherical outer halo . their analysis of the kinematics of a local sample of calibration stars from the sloan digital sky survey ( sdss : york et al .
2000 ; gunn et al . 2006 ) indicated that the transition from dominance by the inner halo to the outer halo occurs in the range 15 - 20 kpc from the sun .
a similar transition range has been inferred from analysis of the `` vertical '' photometric stripes ( de jong et al .
2010 ) , obtained during the sloan extension for galactic understanding and exploration ( segue ) sub - survey of sdss - ii ( yanny et al .
2009 ) . the papers by c07 and c10 also demonstrated that the inner - halo population is essentially non - rotating , with v@xmath27 = 7 @xmath28 4 km s@xmath29 , while the outer - halo population exhibits a significant retrograde signature , with v@xmath27 = @xmath3080 @xmath28 13 km s@xmath29 ( where v@xmath27 is the galactocentric rotational velocity ) .
the velocity ellipsoids of these populations differ as well , such that ( @xmath31 , @xmath32 , @xmath33 = ( 150 @xmath28 2 , 95 @xmath28 2 , 85 @xmath28 1 ) km s@xmath29 for the inner halo and ( 159 @xmath28 4 , 165 @xmath28 9 , 116 @xmath28 3 ) km s@xmath29 for the outer halo , evaluated in a galactocentric cylindrical reference frame .
the observed differences in the nature of the spatial distributions and kinematics of the stellar populations associated with the inner- and outer - halo components suggests that , in the context of modern hiearchical cosmogonies , their progenitor mini - halos and subsequent merging and accretion scenarios differed as well .
if the inner halo formed from a limited number of moderately massive mini - halos ( see , e.g. , bullock & johnston 2005 ; schlaufman et al . 2009 , 2011 ) , while the outer halo resulted from the accretion of more numerous , but less massive ones ( c07 ; frebel et al .
2010 ; norris et al .
2010a , b , c ) , one might expect to find chemical signatures associated with the presently observed stellar populations that reflect these differences .
previous studies have provided hints that this may indeed be the case . for example
, studies of the [ mg / fe ] abundance ratios of stars thought to be associated with the inner halo appear different ( generally 0.1 dex higher ) than those associated with the outer halo ( roederer 2009 ) .
this same study also demonstrated that stars associated with the inner halo exhibit considerably lower star - to - star abundance scatter for both the iron - peak element ratio [ ni / fe ] and the neutron - capture element ratio [ ba / fe ] than found for stars of the outer halo .
the recent study by nissen & schuster ( 2010 ) demonstrated that nearby dwarfs with halo kinematics could be separated into two groups based on [ @xmath34/fe ] .
they proposed that the high-@xmath34 stars may have been born in the disk or bulge of the milky way and heated to halo kinematics by merging satellite galaxies , or else were simply members of the early generations of halo stars born during the collapse of a proto - galactic gas cloud , while the low-@xmath34 stars may have been accreted from dwarf galaxies .
schlaufman et al .
( 2009 ) report detections of numerous elements of cold halo sububstructure ( echos ) in the inner halo , essentially overdensities in radial - velocity space along the segue sightlines .
the echos are systematically more fe - rich , but less @xmath34-enhanced than the kinematically smooth component of the inner halo .
the echos are also chemically distinct from other milky way components ; they are more fe - poor than typical thick - disk stars , and both more fe - poor and @xmath34-enhanced than typical thin - disk stars .
see schlaufman et al .
( 2011 ) for a more detailed discussion .
chemically peculiar stars , such as the @xmath34-element - enhanced very metal - poor star bs 16934 - 002 ( aoki et al . 2007a ; [ fe / h ] @xmath9 ) , the low [ mg / fe ] ( @xmath35 ) , high [ ca / fe ] ( @xmath361.1 ) extremely metal - poor star sdss j2347 + 0108 ( lai et al .
2009 ; [ fe / h ] @xmath37 ) , and the low [ si / fe ] ( @xmath38 ) , low [ ca / fe ] ( @xmath39 ) extremely metal - poor star he 1424 - 0241 , identified by cohen et al .
( 2007 ; [ fe / h ] @xmath40 ) , all have inferred distances ( and metallicities ) that suggest membership in the outer - halo population .
all three of the previously recognized ultra metal - poor stars ( [ fe / h ] @xmath41 , he 0557 - 4840 ; norris et al .
2007 ) or hyper metal - poor stars ( [ fe / h ] @xmath42 , he 0107 - 5240 ; christlieb et al .
2002 , and he 1327 - 2326 ; frebel et al . 2005 ) are similarly thought to be members of the outer - halo population .
likewise , the newly discovered hyper metal - poor star sdss j102915 + 172927 appears to have an orbit consistent with outer - halo membership ( caffau et al .
these results may all be related to the star formation histories in the progenitor populations , their accretion histories , or both . in the present paper we focus on another possibly useful indicator of chemical differences between the inner- and outer - halo components , the carbon - to - iron ratio , [ c / fe ] , which we refer to as the `` carbonicity . '' in particular , we make use of the sdss / segue calibration - star sample from sdss dr7 ( abazajian et al .
2009 ) in order to search for possible contrasts between the frequency and degree of carbon enhancement for the carbon - enhanced metal - poor ( cemp ) stars from this sample that can be kinematically associated with these two halo components .
the cemp stars were originally defined as the subset of very metal - poor stars ( [ fe / h ] @xmath432 ) that exhibit elevated carbon relative to iron , [ c / fe ] @xmath13 + 1.0 ( beers & christlieb , 2005 ) @xmath44 ( aoki et al .
2007b ) . ] . in the last two decades
it has been recognized , primarily from spectroscopic follow - up of metal - poor candidates selected from objective - prism surveys ( e.g. , beers et al .
1985 , 1992 ; christlieb 2003 ) , that roughly 20% of stars with [ fe / h ] @xmath152.0 exhibit enhanced carbonicity , up to several orders of magnitude larger than the solar ratio ( marsteller et al .
2005 ; lucatello et al .
2006 ) . some recent studies ( e.g. , cohen et al . 2005 ; frebel et al .
2006 ) , have claimed that this fraction is a little lower ( 14% and 9% , respectively , for [ fe / h ] @xmath2 ) .
the frebel et al .
( 2006 ) study is of particular interest , as the authors argued that the relative fraction of cemp stars appears to increase substantially with distance above the galactic plane , suggesting a possible connection with changes in the underlying stellar populations . in any case , the fraction of cemp stars rises to 30% for [ fe / h ] @xmath45 , 40% for [ fe / h ] @xmath46 , and 75% for [ fe / h ] @xmath47 ( beers & christlieb 2005 ; frebel et al .
2005 ; norris et al . 2007 ;
caffau et al .
2011 the new [ fe / h ] @xmath48 star does not exhibit carbon enhancement ) ; definitive explanations for the origin of this increase have yet to be offered .
regardless of the ultimate reason , these results indicate that significant amounts of carbon were produced in the early stages of chemical evolution in the universe .
there exist a number of classes of cemp stars , some of which have been associated with proposed progenitor objects . the cemp - s stars ( those with @xmath49process - element enhancement ) , for example , are the most commonly observed type to date .
high - resolution spectroscopic studies have revealed that around 80% of cemp stars exhibit @xmath49process - element enhancement ( aoki et al .
the favored mechanism invoked to account for these stars is mass transfer of carbon - enhanced material from the envelope of an asymptotic giant - branch ( agb ) star to its binary companion ; it is this surviving companion that is now observed as a cemp - s star ( e.g. , herwig 2005 ; sneden et al .
2008 ; bisterzo et al .
2011 ) .
the class of cemp - no stars ( which exhibit no strong neutron - capture - element enhancements ) is particularly prevalent among the lowest metallicity stars ( fe / h @xmath152.5 ; beers & christlieb 2005 ; aoki et al .
possible progenitors for this class include massive , rapidly rotating , mega metal - poor ( [ fe / h ] @xmath50 ) stars , which models suggest have greatly enhanced abundances of cno due to distinctive internal burning and mixing episodes , followed by strong mass loss ( hirschi et al .
2006 ; meynet et al .
2006 , 2010a , b ) .
another suggested mechanism for the production of the material incorporated into cemp - no stars is pollution of the interstellar medium by so - called faint supernovae associated with the first generations of stars , which experience extensive mixing and fallback during their explosions ( umeda & nomoto 2003 , 2005 ; tominaga et al .
this model well reproduces the observed abundance pattern of the cemp - no star bd+44@xmath51493 , the ninth - magnitude [ fe / h ] @xmath52 star ( with [ c / fe ] = + 1.3 , [ n / fe ] = + 0.3 , [ o / fe ] = + 1.6 ) discussed by ito et al .
the recently reported high redshift ( @xmath53 ) , extremely metal - poor damped lyman-@xmath34 ( dla ) system by cooke et al .
( 2011 ; [ fe / h ] @xmath54 ) exhibits enhanced carbonicity ( [ c / fe ] @xmath55 ) and other elemental abundance signatures that kobayashi et al . ( 2011 ) also associate with production by faint supernovae .
it is also of interest that matsuoka et al .
( 2011 ) have reported evidence for strong carbon production in the early universe , based on their analysis of the optical spectrum of the most distant known radio galaxy , tn j0924 - 2201 , with @xmath56 .
below we seek to test if the increasing frequency of cemp stars with declining metallicity , and the suggested increase of the fraction of cemp stars with increasing distance from the galactic plane , can be explained in the context of an inner / outer halo dichotomy and the dominance of _ different carbon - production mechanisms _ ( the processes associated with the progenitors of the cemp - s and cemp - no stars ) being linked to these two populations .
this paper is outlined as follows .
section 2 describes the techniques used to estimate the atmospheric parameters ( @xmath57 , @xmath58 , and [ fe / h ] ) and carbonicity ( [ c / fe ] ) from the low - resolution sdss spectra , compares our estimates with a sample of very low - metallicity stars with available high - resolution spectroscopic determinations , and obtains first - pass estimates of the fractions of cemp stars for various cuts on [ fe / h ] .
section 3 summarizes the calibration - star sample from sdss dr7 we examine here , presents the `` as - observed '' distributions of metallicity and carbonicity for this sample as functions of height above the galactic plane , and carries out a comparison of the distance and rotational velocity distributions for cemp and non - cemp stars .
section 4 explores the global fraction of cemp stars of this sample in the low - metallicity regime , describes our adopted technique for derivation of stellar population membership probabilities for the sdss / segue dr7 calibration stars , and obtains estimates of the fractions of cemp stars associated with the inner- and outer - halo populations .
finally , section 5 summarizes our main results and considers their implications for the formation and evolution of the galactic halo populations .
estimates of the atmospheric parameters for our program stars were obtained from the segue stellar parameter pipeline ( sspp papers i - v : lee et al 2008a , b ; allende prieto et al .
2008 ; smolinski et al . 2011 ; lee et al .
typical internal errors for stars in the temperature range that applies to the majority of the calibration - star sample are @xmath59 125 k , @xmath60 0.25 dex , and @xmath61 } \sim$ ] 0.20 dex .
the external errors in these determinations are of similar size , as discussed in the sspp references listed above . in c10
, a correction was applied for the metallicity determinations of the sspp , which we adopt here as well : @xmath62}_c = -0.186 + 0.765*{\rm [ fe / h]}_a - 0.068*{\rm [ fe / h]}_a^2\ ] ] where [ fe / h]@xmath63 is the adopted metallicity from the sspp , and [ fe / h]@xmath64 is the corrected metallicity .
this polynomial has little effect on stars with metallicity greater than about [ fe / h ] = @xmath65 , but lowers the estimated metallicities for stars below this abundance by 0.1 to 0.2 dex , an offset that was shown to exist between the dr7 sspp - derived metallicities and previous high - resolution spectroscopic measurements .
carbon - to - iron abundance ratios ( [ c / fe ] ) are estimated from the ch g - band at @xmath66 4300 by matching the observed sdss spectra near this feature with an extensive grid of synthetic spectra . in order to construct the grid we employed the newodf models of of castelli & kurucz ( 2003 ) .
synthetic spectra were generated using the ` turbospectrum ` synthesis code ( alvarez & plez 1998 ) , which employs line broadening according to the prescription of barklem & omara ( 1998 ) and barklem & aspelund - johansson ( 2005 ) .
the molecular species ch and cn are provided by b. plez ( private communication ; plez & cohen 2005 ) .
the other linelists used are the same as in sivarani et al .
for the purpose of this exercise we adopted the solar abundances of asplund et al .
( 2005 ) .
the synthetic spectra cover wavelengths between 3600 and 4600 , with original resolution of @xmath67@xmath68 = 0.005 , smoothed to the sdss resolving power ( @xmath69 ) and re - binned to linear 1 pixels .
the stellar parameters of the grid cover the ranges 3500 k @xmath70 @xmath57 @xmath71 k ( steps of @xmath72 k ) , @xmath73 @xmath58 @xmath74 ( steps of 0.5 dex ) and @xmath75 [ fe / h ] @xmath70 0.0 ( steps of @xmath76 dex ) . for stars with [ fe / h ] @xmath77 , models with [ fe / h ]
@xmath78 were adopted . at the time
this analysis was carried out , lower metallicity models from this grid were not available .
however , we did have sparsely - spaced carbon - enhanced models generated by b.plez , which extended down to [ fe / h ] @xmath48 , for 4000 k @xmath79 @xmath57 @xmath79 6000 k. we carried out a number of tests of our use of the [ fe / h ] @xmath78 models extrapolated to lower metallicities , which produced results in good agreement with plez s models .
our program stars include objects with temperatures above 6000 k , where any effects due to extrapolation to lower metallicities will be lower still .
indeed masseron et al .
( 2005 ) points out that models with @xmath57 @xmath13 6000 k are not affected by enhanced carbon . in any event ,
our past experience using models of lower metallicity from other sources has indicated that very little changes when dropping below [ fe / h ] @xmath78 .
this is also indicated by the generally excellent agreement of our [ c / fe ] determinations with the high - resolution results discussed below .
the carbon abundance in the grid goes from [ c / h ] = [ fe / h ] @xmath30 0.5 to [ c / h ] = + 0.5 ( the upper limit is in agreement with agb models ) .
for example , at [ fe / h ] = @xmath302.5 , the grid covers the range from [ c / h ] = @xmath303.0 to @xmath360.5 , which corresponds to the range of carbonicity @xmath80 [ c / fe ] @xmath19 @xmath363.0 .
once constructed , we linearly interpolate within this grid , which is sufficient for the size of the steps in the parameters .
we have checked this by taking a worst - case scenario , generating test synthetic spectra at low temperatures ( @xmath57 @xmath81 k ) and over the above ranges of gravity , metallicity , and carbon abundance .
the linearly interpolated grid was able to recover the input parameters to within a few tenths of dex , which is consistent with our expected errors in the method .
estimation of carbon abundance was accomplished using chi - square minimization of the deviations between the observed and synthetic spectra in the wavelength region between 4285 and 4320 , as carried out by the idl routine amoeba ( a down - hill simplex search procedure ) .
the initial value for [ c / h ] for the the global grid search was set to the same value as the input stellar [ fe / h ] ( [ c / fe ] = 0.0 ) . during the search
, the carbon abundances are allowed to vary ; all other stellar parameters are kept constant . in order to provide some protection from falling into local minima ,
separate searches were performed with lower and higher ranges of [ c / fe ] considered . in almost all cases these converged to the same minima as found for the global search .
when not , we took the value that resulted in the best match in the region of the ch g - band , as judged from a correlation coefficient for the resulting match .
figure 1 provides an example of the spectral matching process for determination of [ c / fe ] for a warm cemp star in our sample .
the upper panel shows the input optical spectrum ( black line ) superposed with a synthetic spectrum with [ c / fe ] = 0 ( red line ) .
the middle panel shows the best matches to the ch g - band obtained from the three different ranges of [ c / fe ] considered .
the lower panel shows the final adopted match . note that our procedures are not traditional synthesis analyses , but are based on spectral matching . as part of this approach ,
the continuum - flattened observed spectra must be registered to match similarly flattened synthetic spectra . because rectification of the stellar ( and synthetic ) continua is sometimes imperfect , small deviations over localized regions of spectral range can occasionally appear .
careful inspection of the bottom panel of figure 1 reveals , for example , a slight mismatch near the red end .
we have taken care to minimize the occurrence of these mismatches to the extent feasible , with particular effort made in the range that is explored for performing the match to the ch g - band . in any case , since the atmospheric parameters are set before conducting the matching , slight registration offsets outside of the ch g - band region have no effect on our derived [ c / fe ] .
we note that most of the weak features in the region of the ch g - band are real , and not due to noise .
similar procedures have been employed in previously published papers ( beers et al .
2007 ; kennedy et al .
2011 ) , to which we refer the reader for additional discussion of these techniques .
we have also pursued a modest sanity check by carrying out full syntheses for a small number of spectra using the approach employed by norris et al .
( 2010c ) , based on the synthesis code of cottrell & norris ( 1978 ) .
these comparisons indicate that we are able to replicate derived [ c / fe ] by our spectral matching approach to within 0.1 - 0.2 dex , at the one - sigma level in the precision for the majority of our program stars .
estimates of the errors in the final determinations of [ c / fe ] were obtained based on a set of noise - injection experiments for stars over a range of temperatures and s / n ratios .
these experiments indicated that , for stars with a minimum s / n @xmath82 in the region of the g - band , [ c / fe ] could be measured over the @xmath57 range of our sample with a maximum error of 0.5 dex , decreasing to 0.05 dex for the highest s / n spectra ( s / n @xmath83 ) .
we assigned final errors to the estimate of [ c / fe ] using a linear function in s / n between these extremes .
spectra not achieving the minimum s / n level ( or which suffered from anomalies such as pixel dropouts ) were considered non - observations ( i.e. , they were dropped from the sample ) .
in addition , in order to claim a detection we required that the equivalent width of the ch g - band obtain a minimum value of 1.2 , a value again chosen based on inspection of the noise - injection experiments . in cases where this condition was not met , we consider the measured [ c / fe ] an upper limit .
we obtain a check on our determinations of atmospheric parameters and [ c / fe ] , for at least a subset of our stars , based on available high - resolution follow - up spectra .
there are three sources for our comparisons : aoki et al .
( 2008 ) , behara et al .
( 2010 ) , and aoki et al .
( 2011 , in prep . ) .
since these high - resolution spectroscopic programs were mostly interested in very and extremely metal - poor stars , our comparison sample is dominated by stars with [ fe / h ] @xmath77 .
table 1 lists the derived atmospheric parameters and carbon abundance ratios for the stars in common ( the label high indicates the high - resolution results , while the label sspp refers to our analysis of the low - resolution sdss spectra ) .
there are 23 unique objects listed , although one star has only upper limits for [ c / fe ] determined .
note that four stars in this table were reported on by two sets of authors , which provides some feeling for the level of systematic differences in the derived parameters .
the aoki et al .
( 2011 ) paper , which supplies information for most of our comparison sample , concluded that the temperature estimates provided by the sspp were sufficiently good that they simply adopted them for their analysis ( although they tested alternative methods , they could not improve upon the sspp results ) . as a result
, a comparison of @xmath57 determinations for the high - resolution and low - resolution spectroscopic analyses can not be fairly carried out . in the case of stars that were observed by multiple groups , straight averages of the listed parameters
are used ( except when the aoki et al .
2011 _ assigned _ value of @xmath58 @xmath84 , as described below , can be replaced by a _ measured _ value from the other sources ) .
figure 2 shows a comparison of the derived parameter estimates .
robust estimates of the zero - point offsets and rms scatter indicate good agreement in metallicity ( @xmath85}\rangle = -0.09 $ ] dex ; @xmath86 } ) = 0.27 $ ] dex ) , close to the random errors expected for carbon - normal stars from the sspp ( 0.2 dex ) .
surface gravity exhibits an acceptably small zero - point offset but a larger scatter ( @xmath87 dex ; @xmath88 dex ) .
larger differences in the @xmath58 determinations might be expected for several reasons .
first , the comparison spectra for the majority of our program stars ( 18 of 23 ) are based , at least in part , on `` snapshot '' high - resolution spectra ( i.e. , lower s / n ) spectra reported by aoki et al .
( 2011 ) . for these spectra
, estimates of @xmath58 were fixed to @xmath58 = 4.0 for all stars with @xmath57 @xmath13 5500 k. this was done because the usual procedure of matching fe abundances based on fe i and fe ii lines was not uniformly possible for the warmer stars , due to the weakness of the fe ii lines and the less than ideal s / n . according to aoki et al .
( 2011 ) , the true surface gravities for such stars could lie anywhere in the range @xmath58 = 3.5 to @xmath58 = 4.5 .
note that the reported surface - gravity offset and scatter above does not include the three aoki et al .
( 2011 ) stars that did not have reported estimates from other sources .
surface gravity estimates from the high - resolution spectra of giants with @xmath57@xmath89 k are determined by aoki et al .
( 2011 ) based on analysis of the fe i and fe ii lines , in the usual manner .
note that aoki et al . also identified four stars in this temperature range to be cool main - sequence stars , rather than giants , based on the weak lines of their ionized species . for these stars ,
gravity estimates are obtained by matching to isochrones for metal - poor main - sequence stars ( a similar process to that carried out by aoki et al .
2010 ) .
it is expected that the high - resolution analysis estimates of @xmath58 for the cooler stars are precise at a level no better than 0.3 dex .
secondly , difficulty in estimation of @xmath58 for cemp stars might not be suprising , given that molecular carbon bands ( in particular for later type stars ) can easily confound gravity - sensitive features in a low - resolution spectrum . however , in our analysis we have taken care to avoid regions of the spectrum where molecular carbon bands have corrupted the gravity - sensitive features . in reality
, this problem is only encountered for the coolest stars with the strongest molecular bands , which are a distinct minority of our sample .
our sample only includes stars in the temperature range 4500 k @xmath19 @xmath57 @xmath19 7000 k , and among the 813 carbon - rich stars considered in our analysis below , none of them have @xmath57 @xmath19 5000 k , and only 10% have @xmath57 @xmath19 5750 k. taking a global approximation that the high - resolution determinations of surface gravity contribute 0.4 dex to the rms scatter comparison with the low - resolution estimates , this indicates that the sspp estimates for these stars have an external error of determination of @xmath90 = 0.52 dex
. we also note that the great majority of the stars analysed in our sample have [ fe / h ] @xmath13 @xmath302.5 , for which surface gravity estimates should be better determined , due to the increasing strength of their gravity - sensitive metallic features .
the agreement in estimates of the carbonicity is quite good ( @xmath91}\rangle = + 0.05 $ ] dex ; @xmath92 } ) = 0.29 $ ] dex ) , since we expect even the high - resolution determinations to be precise to no better than about 0.15 - 0.20 dex .
this suggests that our external errors for estimates of [ c / fe ] are on the order of 0.25 dex . note that this level of agreement between [ c / fe ] determinations based on the high - resolution and sspp analyses would not be possible if the rough estimates of @xmath58 had a strong influence on our procedures .
however , at the suggestion of an anonymous referee , we have carried out explicit tests of the effect of incorrect surface - gravity determinations on the derived [ c / fe ] . in order to test the impact on the derived [ c / fe ] from the adoption of an incorrect surface gravity , we have used synthetic spectra ( with known atmospheric parameters and [ c / fe ] ) from our grid covering three fixed metallicities , [ fe / h ] = @xmath38 , @xmath93 , and @xmath65 , three fixed gravities , @xmath58 = 2.0 , 3.0 , and 4.0 , which spans the range of the majority of our program stars , and two levels of carbonicity , [ c / fe ] = 0.0 and [ c / fe ] = + 1.5 .
we then intentionally perturbed their input @xmath58 values by @xmath38 dex , @xmath94 dex , @xmath95 dex , and @xmath96 dex , and derived estimates of their [ c / fe ] following the procedures described above . as can be seen from inspection of figure 3 ( which shows the case for [ c / fe ] = 0.0 )
, the effects on estimates of [ c / fe ] never exceed 0.5 dex ( and then only in the most extreme cases of @xmath28 1.0 dex variation in @xmath58 ) , and are generally much smaller than that .
not surprisingly , the largest variations occur for the warmest stars , which have weaker ch g - band features .
however , our sample included only a small fraction of stars with @xmath57 @xmath13 6500 k ( @xmath97 % ) , so this is not expected to have a major effect .
the mean zero - point offsets and rms variations in the derived [ c / fe ] , relative to the known value across all @xmath57and @xmath58 considered , are shown for each panel in figure 3 , and are acceptably small .
similar results apply to the case when we fix the input [ c / fe ] = + 1.5 .
figure 3 considered the solar [ c / fe ] case , since we are more concerned with false positives that would cause us to count a star as carbon - enhanced when it is not .
we conclude that , while some sensitivity to estimates of [ c / fe ] may exist due to errors in estimates of @xmath58 , its impact on our results should be minimal , except for truly extreme cases .
in this work there are 31187 unique stars for which estimates of [ c / fe ] were carried out ( other objects were either repeats , in which case they were averaged , had insufficient spectral s / n ratios , were clear cases of qsos , cool white dwarfs , or very late - type stars , or had some spectral defect in the region of the ch g - band which prevented measurements being obtained all such stars are dropped from the subsequent analysis ) .
the remaining sample can be divided into two categories : stars that have a measured [ c / fe ] ( with the g - band detected ; n@xmath98 = 25647 ) , and stars for which only an upper limit on [ c / fe ] has been obtained ( g - band undetected ; n@xmath99 = 5540 ) .
we call the associated two sets of stars subsample d and subsample l , respectively .
figure 4 shows [ c / fe ] as a function of metallicity for these two categories .
the ridge lines " in both of the panels are due to grid effects in the chi - square matching procedure .
it is worth noting that most of the stars with measured [ c / fe ] ( subsample d ) exhibit carbonicity below [ c / fe ] = + 0.7 , which indeed appears to be a natural dividing threshold between carbon - normal stars and carbon - rich stars .
this limit was also established by aoki et al .
( 2007b ) , who also included in their analysis expected evolutionary effects on the definition of cemp stars .
we adopt this value of [ c / fe ] to define the carbon - rich stars in our sample , without making any luminosity correction adjustment , since our sample includes very few cool giants .
there is an evident correlation between [ c / fe ] and [ fe / h ] seen in this figure , in particular for stars with [ fe / h ] @xmath15 1.0 .
the relative number of stars with carbon excesses increases as the metallicity decreases , as does the level of carbonicity , as reported previously by several studies based on much smaller samples of stars . adopting the definition of carbon - rich stars as above , we distinguish two subcategories within subsamples d and l : stars with [ c / fe ] @xmath100 + 0.7 ( c - norm ) , and stars with [ c / fe ] @xmath13 + 0.7 ( c - rich ) . in the case of subsample l ,
the carbon status for a star can be assessed with certainty only when [ c / fe ] @xmath100 + 0.7 , and remains unknown for stars having [ c / fe ] @xmath13 + 0.7 . indeed , suppose that the limit assigned to a given star is [ c / fe]@xmath101 = + 1.5 .
then , all values below [ c / fe]@xmath101 can still be accepted for that star , including carbonicity well below [ c / fe ] = + 0.7 . when the upper limit is below [ c / fe ] = + 0.7 , however
, the carbon status of the star can be assessed as c - norm .
not surprisingly , there is a strong temperature effect in the assessment of the carbon status for stars in our sample .
for example , stars with higher @xmath57 ( above @xmath57 @xmath102 k ) would require quite high [ c / fe ] for the ch g - band to be detected ; stars without a detected ch g - band are included in the l subsample .
thus , the fractions of cemp stars we derive in this paper are lower limits to the true fractions . in a future paper
we plan to obtain an explicit correction function to account for the `` missing '' stars due to this temperature effect , in order to assess its impact on the derived frequencies of cemp stars .
taking into account the above definitions , the fraction of c - rich stars can be formulated as : @xmath103 where @xmath104 and @xmath105 are stars belonging to subsample d and having [ c / fe ] @xmath100 + 0.7 and [ c / fe ] @xmath13 + 0.7 , respectively , while n@xmath106 are stars in subsample l and having [ c / fe ] @xmath100 + 0.7 .
stars with unknown carbon status are not included in the above definition .
table 2 reports the number of stars belonging to the various categories for different ranges of metallicity , as well as the total fractions of c - rich stars .
note that , when comparing the reported fractions in this table with other fractions reported in this paper , the table 2 fractions have no restriction on whether or not acceptable proper motion and radial velocity measurements were available for a given star .
we begin with the 31187 unique dr7 calibration stars for which estimates of [ c / fe ] exist , and with distances slightly revised from those presented in c07 and c10 , as described below .
details concerning the nature of the calibration - star sample can be found in c10 ; here we recall a few facts concerning these stars that are of importance for the present analysis .
spectra of the sdss / segue calibration stars were obtained to perform spectrophotometric corrections and to calibrate and remove the night - sky emission and absorption features ( telluric absorption ) from sdss spectra .
the spectrophotometric calibration stars cover the apparent magnitude range 15.5 @xmath107 17.0 , and satisfy the color ranges 0.6 @xmath108 1.2 ; 0.0 @xmath109 0.6 .
the telluric calibration stars cover the same color ranges as the spectrophotometric calibration stars , but at fainter apparent magnitudes , in the range 17.0 @xmath107 18.5 .
the c10 paper describes the radial velocity estimates ( which have a precision of 5 - 20 km s@xmath29 , depending on the s / n ratio of the spectrum , and with a negligible zero - point offset ) , as well as photometric distance estimates , obtained by using the sspp surface - gravity estimate for luminosity classification , then following the procedures of beers et al .
since the estimated @xmath58 is used only for classification , this means that for stars redder than than main - sequence turnoff ( msto ) , the expected large differences in surface gravity for dwarfs and giants make even approximate @xmath58 estimates sufficient errors of 1 to 2 dex in @xmath58 would have to be routinely made in order to confound this procedure .
close to the msto , any method of photometric distance estimation becomes more problematic , but the difference in the derived distances for stars just above or just below the msto decreases the closer they are to it , which mitigates against these effects .
schnrich et al .
( 2011 ) have criticized the beers et al .
( 2000 ) method of distance determination , and the results of the c07 and c10 that relied on it .
they claimed that the counter - rotating halo found in c07 and c10 is a result of biases in distance estimates for main - sequence dwarfs , and furthermore that the distance derivation is influenced by sorting a subset of the stars into incorrect positions in the color - magnitude diagram of an old , metal - poor population . in a rebuttal paper , beers et al .
( 2011 ) demonstrated that the the schnrich et al .
claims concerning dwarf distances are incorrect ( due to their adoption of the wrong main - sequence absolute magnitude relationship from ivezi et al .
furthermore , the claimed retrograde tail in the rotation - velocity distribution was shown to arise from the measured asymmetric proper motions , and was not the result of the propagation of distance errors . in any event
, we have applied the procedure suggested in the rebuttal paper to reassign intermediate - gravity stars that were originally classified as main - sequence turnoff stars into dwarf or subgiant / giant luminosity classifications , in cases where their derived @xmath57 were substantially lower than expected for the turnoff . revised distances for these stars
have been adopted as well .
a reassessment of the kinematics indicates that the retrograde signal for the subsample at very low metallicity ( [ fe / h ] @xmath152.0 ) remains .
the interested reader can find more details in beers et al .
based on the discussion in that paper , we believe that our distances should be accurate to on the order of 15 - 20% .
the c10 procedures applied a series of cuts to their sample designed to better enable measurement of the kinematic and orbital properties of the various stellar populations considered .
this produced a subsample of stars in the local volume ( distance from the sun @xmath110 kpc , and with 7 kpc @xmath111 10 kpc , where @xmath112 is the galactocentric distance projected onto the plane of the galaxy ) , which they referred to as the local sample " ( n @xmath66 17000 ) .
these restrictions were made in order to mitigate against the increase in the errors in the derived transverse velocities , which scale with distance from the sun , and to improve the applicability of the simple models for the adopted form of the galactic potential . for our present analysis we do not need to redetermine the velocity parameters of the underlying stellar populations , so we can relax these cuts somewhat . as described below , in order to increase the numbers of stars in our sample with measured [ c / fe ] , we have changed the constraint on the distance from the sun , from @xmath113 4 kpc to @xmath114 10 kpc , and removed the constraint on the projected distance @xmath112 . with these relaxed cuts
the total number of stars climbs to 30874 .
we refer to this new sample as the extended sample . "
the upper panel of figure 5 shows the as - observed metallicity distribution function ( mdf ) for the dr7 calibration stars belonging to the extended sample ( black histogram ) , as well as for stars in the c10 local sample , as described above ( red histogram ) .
we use the term `` as - observed '' in order to call attention to the fact that the selection functions for the calibration stars were not intended to return a fair sample of stars , suitable for an unbiased analysis of the distribution of metallicity for stars in the galaxy .
rather , the calibration stars were selected to emphasize the numbers of moderately low - metallicity stars that might serve to best constrain the spectrophotometry and telluric line corrections carried out as part of the sdss spectroscopic pipeline reductions .
thus , these mdfs are a `` sample of convenience , '' one that is still useful for providing guidance as to the presence of various stellar populations in metallicity space , but not for obtaining estimates of their relative normalizations ( for which other samples drawn from sdss are more suitable ) . no selection for or against carbon - enhanced stars was carried out in the selecton of the calibration - star sample . the mdfs of the two samples are clearly very similar , and comprise stars with metallicities that sample all of the primary stellar components of the galaxy ( with the exception of the bulge ) .
the lower panel of figure 5 is the distribution of [ c / fe ] ( estimated as described above ) , which we refer to as the carbonicity distribution function ( cardf ) for the extended sample ( black histogram ) , and for the local sample ( red histogram ) .
the inset in the panel shows a rescaling appropriate for the high [ c / fe ] tail of this distribution . as in the case of the mdfs ,
the shape of the two cardfs are similar , with a strong peak at [ c / fe ] @xmath66 + 0.2 to + 0.3 , and two tails , a weak one at low carbonicity ( [ c / fe ] @xmath19 0 ) , and a strong one that extends to high carbonicity , + 0.5 @xmath19 [ c / fe ] @xmath19 + 3.0 . the total number of stars with high carbonicity is significantly different in the two samples
indeed , for the extended sample we find 728 stars with [ c / fe ] @xmath13 + 0.7 , while for the local sample the number is reduced to 318 .
we now examine the mdfs and the cardfs of the extended sample of calibration stars for different intervals in @xmath10z@xmath10 , with cuts chosen to ensure there remain adequate numbers of stars in each interval . in figure 6 , the first ( left - hand ) column and
the third column show the mdfs , while the second column and the fourth ( right - hand ) column are the cardfs . in the first and third columns ,
the red arrows indicate the peak metallicities of the various stellar populations considered by c10 . in the second and fourth column ,
the blue arrows show the location of the solar carbon - to - iron ratio ( [ c / fe ] = 0.0 ) , and the location of the natural threshold that divides carbon - normal stars from carbon - rich stars ( [ c / fe ] = + 0.7 ) , as identified above . examination of the first column of panels in figure 6 shows how the mdf changes from the upper - left panel , in which there are obvious contributions from the thick - disk , the metal - weak thick disk ( mwtd ) , and inner - halo components in the cuts close to the galactic plane , to the lower - left panel , with an mdf dominated primarily by inner - halo stars . in the third column of panels , with distances from the plane greater than 5 kpc ,
the transition from inner - halo dominance to a much greater contribution from outer - halo stars is clear .
this demonstration is , by design , independent of any errors that might arise from derivation of the kinematic parameters , and provides confirmation of the difference in the chemical properties of the inner- and outer - halo populations originally suggested by c07 .
the second and fourth columns show the results of the same exercise for the cardfs .
close to the galactic plane ( second column ; up to @xmath10z@xmath10 = 3 kpc ) , where the thick disk and mwtd are the dominant components , the cardf is strongly peaked at values between [ c / fe ] = 0.0 and [ c / fe ] = + 0.3 .
the cardfs of the thick disk and mwtd will be explored in a future paper .
here we simply note that in the regions close to the plane , where these components dominate the sample , there are not many stars populating regions of high carbonicity . at larger distances from the galactic plane , where the inner halo begins to be the dominant component , there appears a tail in the carbonicity distribution towards higher values , [ c / fe ] @xmath13 + 0.5 . in the fourth column of panels , where the distance from the galactic plane is @xmath10z@xmath10 @xmath13 5 kpc , and where we expect to see the beginning of the transition from inner - halo dominance to outer - halo dominance , the tails towards high [ c / fe ] values become even more evident .
the cardf exhibits a strong peak at [ c / fe ] @xmath66 + 0.3 and a long tail towards high values of carbonicity , up to [ c / fe ] = + 3.0 .
the nature of the cardf is likely to be influenced by the change of the mdf as a function of the distance from the galactic plane , due the well - known trend of increasing [ c / fe ] with declining [ fe / h ] .
however , as discussed below , there is evidence that the observed changes may reflect _ real differences _ in the chemistry of the inner- and outer - halo populations , even at a given ( low ) metallicity . by adopting the threshold of [ c / fe ] @xmath44 ,
the fraction of sdss / segue calibration stars with high carbonicity in the subsample at @xmath10z@xmath10 @xmath13 9 is 20% , ( which , as argued above , is a lower limit ) , in line with previous estimates for stars with [ fe / h ] @xmath2 .
we now consider how the nature of our extended sample differs for the low - metallicity ( [ fe / h ] @xmath1 ) cemp ( [ c / fe ] @xmath44 ) and non - cemp ( [ c / fe ] @xmath19 + 0.7 ] ) stars .
figure 7 shows two columns of panels , corresponding to the stars in our sample with different cuts on z@xmath115 ( the maximum distance of a stellar orbit above or below the galactic plane ) ; the left - hand column includes stars at all z@xmath116 , while the right - hand column only include stars satisfying z@xmath116 @xmath13 5 kpc .
the top panels show the distributions of distance , @xmath117 , while the bottom panels are the distribution of the galactocentric rotational velocity , v@xmath27 . for all panels ,
cemp stars are shown as red dot - dashed histograms ; non - cemp stars are shown as black solid histograms . as can be seen in the top panel of each column , the peak of the distance distribution in both cases is , for the non - cemp stars , @xmath118 kpc , while for the cemp stars , a softer peak is seen around @xmath119 2.5 - 4 kpc ( closer to @xmath119 4 kpc for the higher z@xmath116 cut ) , with long tails extending to the 10 kpc cutoff of the extended sample . at all distances beyond
kpc the relative fraction of cemp stars exceeds that of the non - cemp stars , while closer than @xmath119 3 kpc the relative fraction of non - cemp stars is greater than that of the cemp stars . a kolmogorov - smirnov ( k - s ) test of the distance distributions indicates that the hypothesis that the cemp and non - cemp stars are drawn from the same parent population is rejected at high significance ( @xmath120 ) for both cuts on z@xmath116 . for the case of z@xmath121 kpc , stars classified
as dwarfs represent 72% of the non - cemp sample and 42% of the cemp sample , while subgiants and giants represent 24% of the non - cemp sample and 50% of the cemp sample .
for both the non - cemp and cemp samples , main - sequence turnoff stars represent less than 10% of the samples . for the case of z@xmath122 kpc ,
dwarfs comprise 52% of the non - cemp sample and 29% of the cemp sample , while subgiants / giants represent 44% of the non - cemp sample and 63% of the cemp sample .
again , main - sequence turnoff stars comprise less the 10% of both samples .
thus , the difference in distance distribution is not solely dependent on the luminosity classes associated with the populations split on carbonicity ; dwarfs and subgiants / giants are both significantly represented in each of the non - cemp and cemp samples . from inspection of the bottom panels of figure 7 ,
the non - cemp stars exhibit a slightly asymmetric distribution of rotational velocities centered close to v@xmath123 km s@xmath29 , and a weak retrograde tail .
by contrast , the cemp stars exhibit a rather strong asymmetry extending to large retrograde velocities .
a k - s test of the rotational - velocity distributions indicates that the hypothesis that the cemp and non - cemp stars are drawn from the same parent population is rejected at high significance ( @xmath120 ) for both cuts on z@xmath116 .
it is interesting to consider the distribution of luminosity classes for the split on [ c / fe ] for the retrograde stars with v@xmath124 km s@xmath29 . for the case of z@xmath125 kpc , stars in the retrograde tail classified
as dwarfs represent 66% of the non - cemp sample and 20% of the cemp sample , while subgiants / giants represent 30% of the non - cemp sample and 73% of the cemp sample . for both non - cemp and cemp samples
main - sequence turnoff stars represent less than 10% of the samples . for the case of z@xmath126 kpc , the retrograde stars classified as dwarfs comprise 48% of the non - cemp sample and 18% of the cemp sample , while subgiants / giants represent 47% of the non - cemp sample and 75% of the cemp sample .
again , main - sequence turnoff stars comprise less than 10% of both samples . significant fractions of dwarfs and giants are present in the retrograde tail for both the non - cemp and cemp samples .
thus , the retrograde signature is unlikely to be due to a preponderance of stars with aberrant distance estimates owing to luminosity misclassifications .
if we specialize to the highly retrograde tails , we find that for @xmath127 @xmath128 kpc , the portion of the non - cemp sample in this tail is only 8% for @xmath129 @xmath130 km s@xmath29 and 4% for @xmath129 @xmath131 km s@xmath29 .
for the cemp stars , these portions are 17% for @xmath129 @xmath132 km s@xmath29 and 12% for @xmath129 @xmath131 km s@xmath29 , respectively . for the case of @xmath127
@xmath12 kpc , the portion of the non - cemp sample is 12% for @xmath129 @xmath132 km s@xmath29 and 8% for @xmath129 @xmath131 km s@xmath29 . for the cemp stars , these portions are 24% for @xmath129 @xmath132 km s@xmath29 and 19% for @xmath129 @xmath131 km s@xmath29 , respectively .
thus , in both cases , a split on the level of carbon enhancement leads directly to rather different relative population of the retrograde tails , compared to the full distributions of non - cemp and cemp stars . identifying the highly retrograde tails with the outer - halo population
, we can already infer that the outer - halo component appears to harbor a greater fraction of cemp stars than the inner - halo component .
in addition , before considering a more complete discussion below , we can make use of the @xmath129 distribution to obtain an estimate of the approximate fraction of cemp stars in the outer halo .
we proceed by asserting that the stars in the highly retrograde tail , with @xmath129 @xmath131 km s@xmath29 , are very likely to be members of the outer - halo population .
this follows because the dispersion in @xmath129 for an essentially non - rotating inner halo is on the order of 100 km s@xmath29 , and placing a cut at two sigma below the mean rotation of the inner halo excludes all but 2.5% of likely inner - halo stars . with this assumption
, we find that 11% of the stars in the highly retrograde tail ( and with @xmath127 @xmath13 5 ) kpc ) are cemp stars .
this differs from the calculations immediately above , in that we are considering the fractions of cemp stars relative to the total number of stars ( including those from the l subsample , but not those with unknown carbon status ; i.e. , we are using eqn .
2 ) . recall that this calculation applies only for stars with [ fe / h ] @xmath1 . for this subsample
we find a mean carbonicity of @xmath4[c / fe]@xmath5 = + 1.47 @xmath28 0.07 , where the error is the standard error of the mean .
this value can be taken as a first - pass estimate of the mean outer - halo carbonicity .
figure 8 shows the observed distributions of the measured proper motions in the right ascension and declination directions for the low - metallicity ( [ fe / h]@xmath133 ) stars in our sample , for two cuts on @xmath127 , represented by small blue dots .
the same panels show , represented as red stars , the objects that populate the highly retrograde tails of the distribution of rotational velocity shown in figure 7 .
note that cuts at @xmath129 @xmath15150 km s@xmath29 and @xmath129 @xmath15200 km s@xmath29 are shown in the upper and lower rows of panels , respectively . as can be appreciated from inspection of this figure ,
the proper motions in both directions for the stars assigned to the highly retrograde tail are asymmetrically distributed , and explore much larger values , relative to the rest of the sample .
this supports the reality of the highly retrograde signature seen in figure 7 , and indicates that it is due primarily to the proper motions of the participating stars .
see beers et al .
( 2011 ) for additional details concerning the veracity and interpretation of sdss proper motions for the sdss calibration - star sample .
in the relatively nearby volume explored by the sdss / segue calibration - star sample , c07 and c10 have shown that the two stellar components of the galactic halo are strongly overlapped in their spatial distribution , velocity ellipsoids , and mdfs .
thus , to explore possible differences in the frequencies and mean carbonicities of cemp stars , it is essential to assign inner- and outer - halo membership probabilities to each star in the sample that is likely to belong to the halo system .
we describe how this is accomplished in the sections below .
in the solar neighborhood , the set of parameters that best identify the presence of the main structures are the distance from the galactic plane , the rotational velocity of a star in a cylindrical frame with respect to the galactic center , and the stellar metallicity ( see c10 ) .
for consideration of the disk system of the galaxy , the distance from the galactic plane is best represented by @xmath10z@xmath10 ( the present distance of a star above or below the galactic plane ) , while for the halo components , a more suitable distance is z@xmath116 , which depends on the adopted gravitational potential .
the choice of z@xmath116 for the galactic halo is necessary because of the much larger spatial extent of its two primary components , relative to that of the disk components . before seeking a deconvolution of the inner- and outer - halo components ,
we need to check for possible correlations between all of the basic parameters we wish to employ . in c10
we have shown that z@xmath116 has a significant correlation with v@xmath134 and v@xmath135 , but not with v@xmath27 , other than that expected from the presence of the thick - disk population at high positive rotation velocity and the halo at lower rotational velocity .
a very similar behavior is confirmed for the extended sample as well ( figure 9 ) .
we conclude that the galactocentric rotational velocity , v@xmath27 , and the vertical distance , z@xmath116 , when combined with metallicity , can be used to obtain useful information on the different stellar populations present in the extended sample .
the primary chemical parameter of the present analysis is the carbonicity , [ c / fe ] , so it is important to check for its possible correlations with the basic parameters defined above .
the result of this exercise is shown in figure 10 . here
, the upper panel shows the galactocentric rotational velocity as a function of metallicity for the extended sample .
the gray dots represent the stars in the sample with [ c / fe ] @xmath136 , while the red dots denote the stars with enhanced carbonicity , [ c / fe ] @xmath44 .
it is worth noting that the stars with [ c / fe ] @xmath13 + 0.7 are _
almost all _ located in the halo components of the galaxy , with few exceptions .
this is perhaps not surprising , as the halo components of the milky way are very metal poor .
the middle panel of figure 10 shows the galactocentric rotational velocity , @xmath129 , as a function of [ c / fe ] .
this plot shows no evidence of correlation between v@xmath27 and [ c / fe ] .
finally , the lower panel of figure 10 shows [ fe / h ] as a function of [ c / fe ] . here
, there is clear evidence of a correlation between [ c / fe ] and [ fe / h ] the carbonicity increases as the metallicity decreases .
a similar trend was already noticed in past works , such as rossi et al .
( 2005 ) and lucatello et al .
we consider this result in more detail below .
the left panel of figure 11 shows the fraction of cemp stars in the extended sample , as a function of [ fe / h ] , in the metal - poor regime ( [ fe / h ] @xmath19 @xmath301.5 ) , and at vertical distance z@xmath116 @xmath13 5 kpc ( chosen to avoid thick disk and mwtd contamination ) . here
, each bin in metallicity corresponds to an interval of @xmath67[fe / h ] = 0.2 dex , with the exception of the last bin , where stars are selected in the range [ fe / h ] @xmath19 @xmath302.6 .
the fractions of cemp stars in each bin are obtained by selecting objects with [ c / fe ] @xmath13 + 0.7 , and application of eqn .
errors on the cemp fraction are evaluated through the jackknife approach .
this technique is similar to bootstrapping , but instead of sampling with replacement , it recomputes the statistical estimate leaving out one observation at a time from the sample ( wall & jenkins 2003 ) .
the dash - dotted line is a second - order polynomial fit to the data .
the increase of the fraction of cemp stars with declining metallicity pertains to the global behavior of the data in the metal - poor regime .
if metallicity alone is the driver of the carbon - enhancement phenomenon , one might wonder if the strong increase in the fraction of cemp stars at [ fe / h ] @xmath152.0 could be due to the increasing importance of the outer - halo component , which has a peak of metallicity at [ fe / h ] @xmath66 @xmath302.2 , and a long tail extending to lower metallicity .
we have evaluated the as - observed fractions of carbon - rich stars in our extended sample , for z@xmath116 @xmath13 5 kpc , over several bins in metallicity . with the adopted definition of carbon - rich stars ( [ c / fe ] @xmath13 + 0.7 ) , and following eqn .
2 , we find that 2% , 7% , and 20% of stars in the intervals @xmath137 [ fe / h ] @xmath138 , @xmath139 [ fe / h ] @xmath1 , and [ fe / h ] @xmath152.5 , are carbon - rich , respectively . for ease of comparison with previous determinations of the cemp fractions in the halo
, we note that the global fraction of cemp stars in the halo system with @xmath127 @xmath13 5 kpc and [ fe / h ] @xmath1 is 8% , for [ fe / h ] @xmath2 it is 12% , and for [ fe / h]@xmath140 it is 20% .
the right panel of figure 11 shows the mean carbonicity , @xmath4[c / fe]@xmath5 , as a function of [ fe / h ] . obviously , it is a strong function of metallicity , although there may be some sign of it leveling off at the lowest metallicities .
larger samples , in particular of lower metallicity cemp stars , are required to be certain .
figure 12 shows the fraction of cemp stars in the extended sample , as a function of distance from the galactic plane , @xmath10z@xmath10 , for stars satisfying @xmath302.0 @xmath19 [ fe / h ] @xmath151.5 and [ fe / h ] @xmath152.0 , respectively .
the intervals for the cuts on height above or below the plane have widths of @xmath67@xmath10z@xmath10 = 4 kpc , and the cemp star fractions in each bin are obtained by applying the same criteria and definitions as used for the left panel of figure 11 . in the range of metallicity @xmath302.0 @xmath19 [ fe / h ] @xmath151.5 ,
significant contamination of the sample from thick - disk stars is not expected , while the mwtd could be still present ( with metallicity peak around [ fe / h ] @xmath1411.3 ) , but only in the region close to the galactic plane ( 0 kpc @xmath19 @xmath10z@xmath10 < 4 kpc ) ; the mwtd would not be expected to contribute for the metallicity range [ fe / h ] @xmath152.0 even close to the plane .
inspection of this figure indicates a clear dependence of the cemp star fraction on distance from the galactic plane .
close to the plane , this dependence is due to the combined presence of the possible mwtd and inner - halo populations , while at @xmath10z@xmath10 @xmath13 4 kpc , the observed fractions must essentially pertain to the halo system alone .
far from the galactic plane , the observed increase of the cemp star fractions with @xmath10z@xmath10 would be difficult to understand if the halo system comprises a single population .
in such a case , one might expect the cemp star fractions to be roughly constant , as a function of @xmath10z@xmath10 , for any given cut in metallicity .
this is clearly _ not _ what the data show .
interestingly , we note that , for the same intervals in @xmath10z@xmath10 , the mean carbonicity remains approximately contstant , at a value @xmath4[c / fe]@xmath5 @xmath6 for @xmath302.0 @xmath19 [ fe / h ] @xmath151.5 and @xmath4[c / fe]@xmath5 @xmath20 for [ fe / h]@xmath2 . of course , the contribution from the inner- and outer - halo components is shifting as one progresses from low to high @xmath10z@xmath10 , and this may be smoothing out any real variations associated with the individual components .
we return to this question below .
a similar trend of increasing cemp star fraction with height above the galactic plane was previously suggested by frebel et al .
( 2006 ) , based on a much smaller sample of stars from the hamburg / eso survey ( wisotzki et al .
2000 ; christlieb 2003 ) , and proportionately larger error bars . our expanded data set now clearly indicates the existence of a strong spatial variation of the cemp star frequency _ within the halo system _ , and suggests that the observed carbon enhancement is unlikely to be purely driven by metallicity alone .
rather , it points to a real difference in cemp star fractions associated with the inner- and outer - halo populations , and opens the possibility for different carbon - production mechanisms and/or different chemical - evolution histories within their progenitors .
we return to this question below , after considering a method to probabilistically classify individual stars as likely inner- or outer - halo members .
inference of the distribution function of an observable given only a finite , noisy set of measurements of that distribution is a problem of significant interest in many areas of science , and in astronomy in particular .
the observed distribution of a parameter is just the starting point , but what is desired is knowledge of the distribution that we would have in the case of very small uncertainties of the data and with all of the dimensions of the parameter measured ; in other words , the closest representation of the underlying distribution .
usually , the data never have these two properties , and it is then challenging to find the underlying distribution without taking into account the uncertainty of the data ( bovy , hogg , & roweis , 2011 ; hereafter bhr11 ) .
the extreme deconvolution ( xd ) technique of bhr11 confronts all of these issues , and provides an accurate description of the underlying distribution of a _
d_-dimensional quantity by taking into account the potentially large and heterogeneous observational uncertainties , as well as missing dimensions .
the bhr11 paper generalized the well known mixtures - of - gaussians density - estimation method to the case of noisy , heterogeneous , and incomplete data . in this method ,
the underlying distribution of a quantity * v * is modeled as a sum of @xmath142 gaussian distributions @xmath143 where the function n(*v|m@xmath144*,*v@xmath144 * ) is the _
d_-dimensional gaussian distribution with mean * m * and variance tensor * v * and @xmath145 are the amplitudes , normalized to sum to unity ( all of these parameters are grouped together as @xmath146 in what follows ) . the data * w@xmath147 * are assumed to be noisy samples from this distribution @xmath148 where the noise is drawn from a normal distribution with zero mean and known covariance matrix * s@xmath149*. here and in what follows , we ignore the projection matrices * r@xmath147 * of bhr11 , since the data we will apply this technique to are complete . the likelihood of the model for each data point is given by the model convolved with the uncertainty distribution of that data point .
since a gaussian distribution convolved with another gaussian distribution is again a gaussian , the likelihood for each data point is a sum of gaussian distributions @xmath150 where @xmath151 the objective function is the total likelihood , obtained by simply multiplying the individual likelihoods together for the various data points @xmath152 the optimization of this objective function provides the maximum likelihood estimate of the distribution , or its parameters . in bhr11
, the authors developed a fast and robust algorithm to optimize the likelihood , based on an adaptation of the expectation - maximization algorithm ( dempster et al .
1977 ) .
the xd technique provides the best - fit values of the amplitude , mean , and standard deviation of each gaussian component , as well as the so - called _ posterior probability _ that the observed data point * w@xmath149 * is drawn from the component _ j_. the posterior probability is given by @xmath153 ( see bhr11 for the derivation of this formula ) .
the posterior probability is a powerful statistical tool to perform probabilistic assignments of stars to a gaussian component in the model distribution . in galactic studies ,
the gaussian component could represent a primary structural component such as a disk or halo , a moving group , or a spatial or velocity overdensity .
in many respects the xd technique described above is similar to the maximum likelihood technique adopted in c10 , but is more general , because it takes into account the uncertainties of the measurements and provides the membership probabilities .
the basic parameters for application of the xd approach for the analysis at hand are the galactocentric rotational velocity , v@xmath27 , the maximum vertical distance , z@xmath116 , and the metallicity , [ fe / h ] .
we employ the xd technique to determine the underlying distribution of the rotational velocity of halo stars and to determine the posterior membership probabilities for each star . in practice ,
the entries for the xd algorithm are the galactocentric rotational velocity and its uncertainty for each star , v@xmath154 and @xmath155 , respectively .
these two parameters , together with all of the other kinematic and orbital quantities , are derived using the same procedures employed by c07 and c10 ( but with revised distances for the reassigned main - sequence turnoff stars , as discussed above ) . our plan is to make use of a sample in which the constraint on the distance from the sun , @xmath117 , and on the projected galactic distance , @xmath112 , are relaxed ( in order to take advantage of the larger numbers of cemp stars in the extended sample ) . as a consequence ,
the data set becomes noisier then the local sample , because uncertainties on the transverse velocities increase with the distance @xmath117 .
in the terminology of the xd approach , the analysis of the extended sample falls into the case where all the observables are known , but some of the derived parameters have large uncertainties .
the galactic halo is assumed to be a two - component structure , comprising the inner and the outer halo , as discussed in c07 and c10 .
thus , the general expression for the likelihood takes the form : @xmath156\,,\end{aligned}\ ] ] where @xmath157 and @xmath158 are the amplitude of the inner halo and outer halo , respectively .
the velocity distributions are assumed to be gaussian , thus @xmath159 the membership probabilities then take the form : @xmath160 and @xmath161 for the _ _ i__th star .
+ in this section the posterior probabilities , derived through the application of the xd , are obtained for the extended sample ( @xmath114 10 kpc , no constraints on @xmath112 ) , selected in the range of metallicity [ fe / h ] @xmath162 , and with vertical distance z@xmath116 @xmath13 5 kpc .
the results are shown in figure 13 .
the upper panel of this figure shows the distribution of the observed rotation velocity v@xmath27 ( black histogram ) , and the green and red curves represent the resulting gaussian velocity distributions for the inner- and outer - halo components .
the middle panel shows the derived velocity distribution for the two components , obtained by weighting each star with its membership probability , and normalized such that the total area corresponds to unity .
the black histogram is the observed distribution as in the top panel , but normalized to unity as well . in the lower panel ,
the posterior probability as a function of the rotational velocity , v@xmath27 , is shown .
this probability has been obtained using eqns .
11 and 12 . for each star in the sample ,
the xd technique assigns two probability values , the first associated with the inner halo , and the second related to the outer halo . since in our model
a star is required to belong to one or the other component ( there are no orphans allowed ) , and the summed inner- and outer - halo posterior probabilities are forced to unity at each point , the two curves simply complement one another .
the values obtained for the mean rotational velocity and its dispersion are , for the inner halo , v@xmath27 = 56 @xmath28 11 km s@xmath29 , and @xmath32 = 93 @xmath28 35 km s@xmath29 ; for the outer halo , v@xmath27 = @xmath30141 @xmath28 31 km s@xmath29 , and @xmath32 = 138 @xmath28 58 km s@xmath29 . the errors on the velocity and its dispersion have been evaluated by employing the jackknife method .
a k - s test of the null hypothesis that the velocity distributions of stars belonging to the inner- and outer - halo components could be drawn from the same parent population is rejected at a high level of statistical significance ( _ p _
@xmath163 ) .
we now consider the distribution of carbonicity ( [ c / fe ] ) for the extended sample with metallicity [ fe / h ] @xmath152.0 and at z@xmath116 @xmath13 5 kpc .
these selections ensure the presence of essentially all halo stars in the sample , with little or no contamination expected from the thick disk or mwtd .
the top panel of figure 14 shows the cardf for the extended sample .
there are two clear peaks that emerge , one at [ c / fe ] @xmath164 to @xmath95 , and the second softer peak around [ c / fe ] @xmath165 , along with a tail extending towards higher values of carbonicity .
the bottom panel of figure 14 shows the weighted cardfs for the inner- and outer - halo components .
the underlying distribution of each population has been obtained by weighting the values of the cardf with the membership probabilities of each star . as usual , the green distribution denotes the inner halo , while the red distribution represents the outer halo .
a k - s test of the null hypothesis that the cardf of stars belonging to the inner- and outer - halo components could be drawn from the same parent population is rejected at high statistical significance ( _ p _ @xmath19 0.001 ) , clearly indicating that the carbonicity distribution functions of the inner- and outer - halo components are distinct .
the contrast of the cemp star fractions in the inner- and outer - halo components can be investigated by applying the xd analysis to subsamples of stars in restricted ranges of metallicity .
the stars employed are those belonging to the extended sample , selected in the low metallicity range ( [ fe / h ] @xmath19 @xmath301.5 ) and with z@xmath116 @xmath13 5 kpc .
a first attempt was made by selecting stars in intervals of metallicity such that @xmath67[fe / h ] = 0.2 dex , and applying the xd technique to the rotational velocity distribution of each subsample .
this experiment was not successful , due to the low number of stars in each bin .
a much better result has been obtained by choosing a larger interval on metallicity , @xmath67[fe / h ] = 0.5 dex .
the left panel of figure 15 shows the previously determined global trend of the cemp star fraction , as a function of metallicity ( dashed curve ) , overplotted with the derived inner- and outer - halo cemp star fractions .
the blue filled stars represent the predicted values of the cemp star fractions for each bin of metallicity , i.e. , @xmath302.0 @xmath19 [ fe / h ] @xmath166 1.5 , @xmath302.5 @xmath19 [ fe / h ] @xmath166 2.0 , and @xmath303.0 @xmath19 [ fe / h ] @xmath166 2.5 ( hereafter bin1 , bin2 , and bin3 ) .
the predicted values are derived using the global trend of cemp star fractions vs. [ fe / h ] ( second - order polynomial ; fig .
11 ) , and the posterior probability for each star in the subsamples associated with each bin of metallicity , such that : @xmath167 where p@xmath168 is the posterior probability , derived by applying the xd analysis to the subsample of stars selected in bin1 , bin2 , and bin3 , respectively . the parameter @xmath169 represents the cemp star fractions obtained in each bin of metallicity from application of the second order polynomial ( _ j _ and _ k _ denote the halo component and the bin of metallicity , respectively ) . note that the expression in eqn .
13 provides an estimate of the cemp fraction that is driven by the global trend of f@xmath170 as a function of the metallicity ( fig .
11 , left panel ) .
13 makes use of the full shape of the posterior probability , and thus , a considerable overlap between the kinematic distributions of the two components is still present . therefore , we are not expecting to find a significant difference in the predicted values for the two components in each bin of [ fe / h ] .
the predicted value of cemp star fractions in bin1 is f@xmath171 = ( 5.8 @xmath28 0.03)% , f@xmath172 = ( 5.9 @xmath28 0.03)% , for the inner halo and outer halo , respectively .
the same calculations for bin2 provide f@xmath173 = ( 11.9 @xmath28 0.1)% and f@xmath174 = ( 12.0 @xmath28 0.1)% . for the the third bin we find f@xmath175 = ( 22.9 @xmath28 0.6)% and f@xmath176 = ( 22.7 @xmath28 0.6)% .
the error on each fraction is evaluated by employing the jackknife method .
these fractions are sufficiently close to one another that only a single set of blue stars is shown in the left panel of figure 15 to represent both results
. not surprisingly , the above predicted values of the cemp star fractions in the inner and outer halo are very similar in all bins of metallicity .
the presence of a contrast in the cemp star fraction for the two components can be better investigated by reducing the overlap in the kinematic distributions .
this can be done by employing the _ hard - cut - in - probability _ method . for each halo component ,
the lower limit of the membership probability is chosen to be @xmath177 = 0.7 .
this limit is set reasonably high to reduce the contamination between the halo components .
we have selected two subsamples of stars such that p@xmath178 @xmath177 , and p@xmath179 @xmath177 ( _ i _ denotes the star ) , and for each subsample , the cemp star fraction is evaluated following eqn .
the result of this exercise is shown in the left panel of figure 15 . here
, the green ( inner ) and the red ( outer ) filled circles denote the values of the cemp star fractions obtained by employing the hard - cut - in - probability method .
the observed cemp star fractions in bin1 are f@xmath180 = ( 5.2 @xmath28 0.6)% and f@xmath181 = ( 7.7 @xmath28 1.0)% , for the inner and outer halo , respectively . in the case of bin2 , we find f@xmath182 = ( 9.4 @xmath28 0.8)% and f@xmath183 = ( 20.3 @xmath28 2.3)% . finally , in the third bin , the observed fractions are f@xmath184 = ( 20.5 @xmath28 2.4)% and f@xmath185 = ( 30.4 @xmath28 4.8)% . in the highest metallicity bin ,
@xmath14 [ fe / h ] @xmath1 , the cemp star fractions for the two halo components are close to the expected value ( @xmath66 6% ) , with overlapping error bars .
note that the inner halo is the dominant component in this range of metallicity ( [ fe / h]@xmath186 = @xmath301.6 ) , and we were not expecting to find a high contrast in the cemp star fractions . at lower metallicity ,
@xmath139 [ fe / h ] @xmath2 , where the dominant component is the outer halo ( [ fe / h]@xmath187 ) , there is evidence of a significant contrast in cemp fractions between the inner- and outer - halo components , such that f@xmath188 @xmath66 2@xmath189f@xmath182 .
this contrast is confirmed also in the lowest bin of metallicity , @xmath190 [ fe / h ] @xmath77 , even though it is less remarkable , f@xmath185 @xmath66 1.5@xmath189f@xmath191 , and has a much larger error bar , due to the smaller numbers of stars involved .
it is important to note that most of the stars in the lowest metallicity bin belong to the category for which only an upper limit of [ c / fe ] has been provided ( classified as subsample l , see section 2.4 ) .
a significant number of stars in the lowest metallicity bin have high @xmath57 ( @xmath66 65% above 6250 k ) , and thus only quite high values of [ c / fe ] are expected to be detected .
thus , it is our expectation that the quoted frequencies for cemp star fractions , at least at low metallicity , represent lower limits .
morever , the sdss / segue dr7 calibration - star sample has a limited number of stars with [ fe / h ] @xmath77
. the extremely metal - poor regime will be further investigated in a future paper using the much larger sample of such stars in the sdss / segue dr8 release ( aihara et al .
2011 ) .
interestingly , the right panel of figure 15 shows that the mean carbonicity , @xmath4[c / fe]@xmath5 , remains similar for both the inner- and outer - halo components as a function of declining metallicity .
there is a hint of a split emerging for the lowest metallicity bin , @xmath190 [ fe / h ] @xmath77 , with the outer - halo @xmath4[c / fe]@xmath5 being slightly higher than that of the inner - halo value
. however , the error bars overlap , indicating that the sample size is simply too small to be certain .
we have analyzed the sdss / segue dr7 calibration stars in order to explore the global properties of the carbonicity distribution function ( [ c / fe ] ; cardf ) of the halo components of the milky way , and the possible contrast of the cemp star fractions and mean carbonicity between the inner halo and the outer halo .
carbon - to - iron abundance ratios ( or limits ) have been obtained for 31187 stars , based on matches to a grid of synthetic spectra to the ch g - band at 4305 , with an external error for [ c / fe ] on the order of 0.25 dex .
kinematic and orbital parameters were derived employing the methodologies described in c10 .
we have considered the nature of the samples of stars considered cemp ( [ c / fe ] @xmath44 ) and non - cemp ( [ c / fe ] @xmath19 + 0.7 ) , and shown that , at low metallicity ( [ fe / h ] @xmath1 ) , their distributions on distance and rotational velocity differ significantly from one another . for the stars with a detectable ch g - band ,
a deconvolution of the inner- and outer - halo components , along with a derivation of the membership probabilities , has been obtained by applying the extreme deconvolution ( xd ) analysis to the extended sample of calibration stars , selected in the low metallicity range ( [ fe / h ] @xmath151.5 ) and with z@xmath192 5 kpc .
contrasts of the cemp star fractions and mean carbonicity have been obtained by employing the hard - cut - in - probability method to subsamples of stars selected in several bins of metallicity . * the as - observed distribution of [ c / fe ] , which we refer to as the carbonicity distribution function ( cardf ) , has been derived at various intervals on distance above or below the galactic plane . at @xmath10z@xmath10 @xmath13 5 kpc , where we expect to see the beginning of the transition from inner - halo dominance to outer - halo dominance among halo stars ,
the cardf exhibits a strong peak at [ c / fe ] @xmath66 + 0.3 , and a long tail towards high values of carbonicity , up to [ c / fe ] = + 3.0 .
the fraction of cemp stars ( defined here to mean [ c / fe ] @xmath13 @xmath360.7 ) for the subsample at @xmath10z@xmath10 @xmath13 9 kpc ( where the dominant component is the outer halo ) is 20% , in agreement with several previous estimates for stars at similarly very low metallicities , [ fe / h ] @xmath1662.0 .
* at low metallicities ( [ fe / h ] @xmath1 ) , the distribution of derived distances and rotational velocities for the cemp and non - cemp stars differ significantly from one another . * _ almost all _ of the cemp stars are located in the halo components of the milky way . *
we observe a significant increase of the fraction of cemp stars with declining metallicity in the halo system .
following eqn .
2 , we find that 2% , 7% , and 20% of stars in the metallicity intervals @xmath137 [ fe / h ] @xmath138 , @xmath139 [ fe / h ] @xmath1 , and [ fe / h ] @xmath152.5 , are carbon - rich , respectively . for ease of comparison with previous estimates ,
the global frequency of cemp stars in the halo system for [ fe / h ] @xmath1 is 8% ; for [ fe / h ] @xmath2 it is 12% ; for [ fe / h ] @xmath3 it is 20% . for reasons described above ,
we believe these fractions should be considered lower limits . * in the low - metallicity regime ( [ fe / h ] @xmath151.5 ) , and at vertical distances z@xmath193 5 kpc , a continued increase of the fraction of cemp stars with declining metallicity is found ( fig .
11 , left panel ) .
a second - order polynomial provides a good fit to the cemp star fraction as a function of [ fe / h ] .
* in the same metallicity regime
the mean carbonicity , @xmath4[c / fe]@xmath5 , increases with declining metallicity ( fig .
11 , right panel ) .
* in the low - metallicity regime , for both @xmath14 [ fe / h ] @xmath151.5 and [ fe / h ] @xmath152.0 , we find a clear increase in the cemp star fraction with distance from the galactic plane , @xmath10z@xmath10 ( fig .
12 ) . at these low metallicities ,
significant contamination from the thick disk and mwtd populations is unlikely , and would only have a small effect for the bin with 0 kpc @xmath19 @xmath10z@xmath10 @xmath19 4 kpc and @xmath14 [ fe / h ] @xmath151.5 ; this is an observed property of the halo system .
the mean carbonicity remains roughly constant as a function of @xmath10z@xmath10 , taking a value of @xmath4[c / fe]@xmath5 @xmath6 for @xmath14 [ fe / h ] @xmath151.5 , and @xmath4[c / fe]@xmath5 @xmath20 for [ fe / h ] @xmath2 .
* the extreme deconvolution technique of bhr11 has been applied to subsamples of stars selected in three bins of metallicity , @xmath302.0 @xmath19 [ fe / h ] @xmath19 @xmath301.5 , @xmath302.5 @xmath19 [ fe / h ] @xmath19 @xmath302.0 , and @xmath303.0 @xmath19 [ fe / h ] @xmath19 @xmath302.5 , and with z@xmath116 @xmath13 5 kpc .
we have successfully decomposed the inner and outer halo in all three bins , and obtained the inner- and outer - halo membership probabilities for each star .
* at low metallicity ,
@xmath14 [ fe / h ] @xmath1 , where the dominant component is the inner halo ( [ fe / h]@xmath194 ) , there is no significant difference in the cemp star fraction btween the inner and outer halo .
in contrast , at very low metallicity , @xmath139 [ fe / h ] @xmath2 , where the dominant component is the outer halo ( [ fe / h]@xmath187 ) , there is evidence for a significant difference in the cemp star fraction between the inner and the outer halo ( figure 15 , left panel ) , such that f@xmath195 @xmath66 2@xmath189f@xmath182 .
this difference is also confirmed in the lowest metallicity bin , @xmath190 [ fe / h ] @xmath77 , even though it is less remarkable , f@xmath185 @xmath66 1.5@xmath189f@xmath196 , and has a larger error bar .
we conclude that the difference in cemp frequency at very low metallicity is not driven by metallicity itself , but rather , by the stellar populations present .
* the mean carbonicity , @xmath4[c / fe]@xmath5 , remains similar for both the deconvolved inner- and outer - halo components as a function of declining metallicity .
larger samples of low - metallicity cemp stars are required to see if a split in mean carbonicity emerges . a possible scenario for the formation of the inner- and the outer - halo components has been described in c07 .
the fact that the outer - halo component of the milky way exhibits a net retrograde rotation ( and a different distribution of overall orbital properties ) , clearly indicates that the formation of the outer halo is distinct from that of the inner halo and the disk components . in c07
, it was suggested that the outer - halo component formed , not through a dissipational , angular - momentum conserving contraction , but rather through dissipationless chaotic merging of smaller subsystems within a pre - existing dark matter halo .
these subsystems would be expected to be of much lower mass , and subject to tidal disruption in the outer part of a dark matter halo , before they fall farther into the inner part . as candidate ( surviving ) counterparts for such subsystems , one might consider the low luminosity dwarf spheroidal galaxies surrounding the milky way . in the case of the inner halo , c07 argued that the low - mass sub - galactic fragments , formed at an early stage , rapidly merge into several more massive clumps , which themselves eventually dissipationally merge ( owing of the presence of gas that had yet to form stars ) .
the essentially radial merger of the few resulting massive clumps gives rise to the dominance of high - eccentricity orbits for stars that we assign to membership in the inner halo .
star formation within these massive clumps ( both pre- and post - merger ) would drive the mean metallicity to higher abundances .
this would be followed by a stage of adiabatic compression ( flattening ) of the inner - halo component owing to the growth of a massive disk , along with the continued accretion of the gas onto the galaxy .
this general picture is supported , at least qualitatively , by the most recent numerical simulations of the formation and evolution of large disk galaxies that include prescriptions to handle the presence of gas , and which also include approximate star - formation schemes and follow the evolving chemistry ( e.g. , zolotov et al . 2010 ; font et al .
2011 ; mccarthy et al .
2011 ; tissera et al .
2011 ) .
the large fractions of cemp stars in both halo components indicates that significant amounts of carbon were produced in the early stages of chemical evolution in the universe .
the observed contrast of cemp star fractions between the inner halo and outer halo strengthen the picture that the halo components had different origins , and supports a scenario in which the outer - halo component has been assembled by the accretion of small subsystems , as discussed below . in this regard ,
it is interesting that the mdf of the inner halo ( peak metallicity at [ fe / h ] @xmath141 1.6 ) is in the metallicity regime associated with the cemp - s stars , which are primarily found with [ fe / h ] @xmath197 , while the mdf of the outer halo ( metallicity peak at [ fe / h ] @xmath66 @xmath302.2 ) might be associated with the metallicity regime of the cemp - no stars , which are primarily found with [ fe / h ] @xmath77 ( aoki et al . 2007b ) .
the fact that , in the metal - poor regime , the outer halo exhibits a fraction of cemp stars that is larger than the inner halo ( cemp@xmath25 @xmath66 2@xmath198cemp@xmath23 ) , suggests that multiple sources of carbon , besides the nucleosynthesis of agb stars in binary systems , were present in the pristine environment of the outer - halo progenitors ( lower mass sub - halos ) .
these sources could be the fast massive rotators and/or the faint supernovae mentioned in the introduction .
karlsson ( 2006 ) also explored the possibility that primordial gas was pre - enriched in heavy metals by less massive sne ( 13 @xmath19 m / m@xmath199 @xmath19 30 ) , whose ejecta underwent substantial mixing and fallback , while the c ( and n , o ) originated from massive rotating stars with m @xmath200 40 m@xmath199 .
if the cemp stars in the outer halo are predominantly cemp - no stars ( which has yet to be established ) , it might suggest that non - agb - related carbon production took place in the primordial mini - halos .
the predominance of cemp - s stars in the inner halo , if found , would suggest that the dominant source of carbon was the nucleosynthesis by agb stars in binary systems .
this would place important constraints on the primordial imf of the sub - systems responsible for the formation of the two halo components .
recent efforts to model , from the population synthesis standpoint , the fractions of observed cemp stars in the halo have not managed to reproduce results as high as 15 - 20% for metallicities [ fe / h ] @xmath2 ( e.g. , izzard et al .
2009 ; pols et al .
however , these predictions are based solely on carbon production by agb stars .
while such calculations may prove meaningful for the inner - halo population , they may not be telling the full story for carbon production associated with the progenitors of the outer - halo population .
indeed , the observed increase of the cemp star fraction with @xmath10z@xmath10 ( in particular , far from the galactic plane ) we have found would be difficult to understand if the halo system represents a single population . in any event
, the lower cemp fraction in the inner halo may relieve some of the tension with current model predictions .
the fact that the cemp star fraction exhibits a clear increase with @xmath10z@xmath10 suggests that the relative numbers of cemp stars in a stellar population is not driven by metallicity alone .
the proposed coupling of the cosmic microwave background to the initial mass function ( cmb - imf hypothesis ; larson 1998 , 2005 ; tumlinson 2007 ) is one mechanism for imposing a temporal dependence on the imf .
this effect , coupled with chemical evolution models , predicts that the cemp fraction would be expected to increase as the metallicity decreases , but with similar metallicity regimes forming carbon according to the expected yields of the predominant mass range available at that time . in the hierarchical context of galaxy formation ,
star - forming regions are spatially segregated , and their chemical evolution can proceed at different rates , with stars at the same metallicity forming at different times . depending on the source(s ) of carbon , this trend could lead to a spatial variation of the cemp fraction at the same metallicity , increasing in older populations and decreasing in younger ones .
the larger fraction of cemp stars associated with the outer - halo component could be due to the nature of the progenitor low - mass mini - halos ( mass , density , etc . )
in which they formed . a natural place to look for surviving examples are the ultra - faint dwarf spheroidal galaxies surrounding the milky way discovered in the course of the sdss ( willman et al .
2005 ; belokurov et al . 2006a , b ; zucker et al .
2006 , and many others since ) .
early hints of the possible association of cemp stars with the ultra - faint galaxies came from the recognition that a ( serendipitously ) spectroscopically targeted star from sdss in the direction of the canes venatici ultra - faint dwarf , sdss j1327 + 3335 , is a carbon - rich giant with a radial velocity and inferred distance commensurate with this very metal - poor satellite galaxy ( zucker et al .
the existence of cemp stars in ultra - faint dwarf spheroidal galaxies has now been definitively established ( norris et al .
2010a , b ; lai et al . 2011 ) .
norris et al .
2010b reported the discovery of an extremely carbon - rich red giant , segue 1 - 7 , in the segue 1 system .
this star exhibits a metallicity [ fe / h ] = @xmath303.5 , carbonicity [ c / fe ] @xmath201 , and a low barium abundance ratio ( [ ba / fe ] @xmath2021.0 ) , which place it in the cemp - no category .
this discovery is consistent with the idea that the cemp - no stars may indeed be associated with the ultra - faint dwarf spheroidal galaxies ; further similar investigations should prove of great interest .
d.c . gratefully acknowledges funding from rsaa anu to pursue her research .
t.c.b . and y.s.l .
acknowledge partial funding of this work from grants phy 02 - 16783 and phy 08 - 22648 : physics frontier center / joint institute for nuclear astrophysics ( jina ) , awarded by the u.s . national science foundation .
j.b . was partially supported by nasa ( grant nnx08aj48 g ) and the nsf ( grant ast 09 - 08357 ) .
studies at anu of the most metal - poor populations of the milky way are supported by australian research council grants dp0663562 and dp0984924 , which are gratefully acknowledged by j. e. n. and d. c. + _ facilities : _ .
abazajian , k. , et al .
2009 , , 182 , 543 aihara , h. , et al .
2011 , , 193 , 29 allende prieto , c. , et al .
2008 , , 136 , 2070 alvarez , r. , & plez , b. 1998 , , 330 , 1109 aoki , w. , et al . 2007a , , 660 , 747 aoki , w. , beers , t. c. , christlieb , n. , norris , j. e. , ryan , s. g. , & tsangarides , s. 2007b , , 655 , 492 aoki , w. , et al .
2008 , , 678 , 1351 aoki , w. , beers , t. c. , honda , s. , & carollo , d. 2010 , , 723 , l201 asplund , m. , grevesse , n. , & sauval , a. j. 2005 , in cosmic abundances as records of stellar evolution and nucleosynthesis in honor of david l. lambert , asp conf .
336 , p. 25
barklem , p. s. , & omara , b. j. 1998 , , 300 , 863 barklem , p. s. , & asplelund - johansson , j. 2005 , , 435 , 373 beers , t. c. , preston , g. w. , & shectman , s. a. , 1985 , , 90 , 2089 beers , t. c. , preston , g. w. , & shectman , s. a. , 1992 , , 103 , 1987 beers , t. c. , rossi , s. , norris , j. e. , ryan , s. g. , & shefler , t. 1999 , , 117 , 981 beers , t. c. , chiba , m. , yoshii , y. , platais , i. , hanson , r. b. , fuchs , b. , & rossi , s. 2000 , , 119 , 2866 beers , t. c. & christlieb , n. 2005 , , 43 , 531 beers , t. c. , sivarani , t. , marsteller , b. , lee , y. s. , rossi , s. , & plez , b. 2007 , , 133 , 1193 beers , t. c. , et al .
2011 , , submitted ( arxiv:1104.2513 ) behara , n. t. , bonifacio , p. , ludwig , h .-
g . , sbordone , l. , gonzalez hernandez , j. i. , & caffau , e. 2010 , , 513 , 72 belokurov , v. , et al .
2006a , , 642 , l137 belokurov , v. , et al . 2006b , , 647 , l111 bisterzo , s. , gallino , r. , straniero , o. , cristallo , s. , & kppeler , f. 2011 , , in press ( arxiv:1108.0500 ) bovy , j. , hogg , d. w. , & roweis , s. t. 2011 , annals of applied statistics , 5 , no . 2 , 1657 ( arxiv:0905.2979 ) ( bhr11 ) bullock , j. s. , & johnston , k. v. 2005 , 635 , 931 caffau , e. , et al
2011 , nature , 477 , 67 carollo , d. , et al .
2007 , nature , 450 , 1020 ( c07 ) carollo , d. , et al .
2010 , , 712 , 692 ( c10 ) castelli , f. & kuruz , r. l. 2003 , modelling of stellar atmospheres , proceedings of the 210th symposium of the international astronomical union , 210 , a20 christlieb , n. , et al .
2002 , nature , 419 , 904 christlieb , n. 2003 , reviews in modern astronomy , 16 , 191 cohen , j. g. , et al .
2005 , , 633 , l109 cohen , j. g. , mcwilliam , a. , christlieb , n. , shectman , s. a. , thompson , i. , melendez , j. , wisotzki , l. , & reimers , d. 2007 , , 659 , l161 cooke , r. , pettini , m. , steidel , c. c. , rudie , g. c. , & jorgenson , r. a. 2011 , , 412 , 1047 cottrell , p. l. , & norris , j. 1978 , apj , 221 , 893 de jong , j. t. a. , yanny , b. , rix , h. -w . , dolphin , a. e. , martin , n. f. , & beers , t. c. 2010 , , 714 , 663 dempster , a. p. , laird , n. m. , & rubin , d. b. 1977 , journ .
soc . , 39 , 1 font , a. s. , mccarthy , i. g. , crain , r. a. , theuns , t. , schaye , j. , wiersma , r. p. c. , & dalla vecchia , c. 2011 , mnras , in press ( arxiv:1102.2526 ) frebel , a. , et al .
2005 , nature , 434 , 871 frebel , a. , et al .
2006 , , 652 , 1585 frebel , a. , kirby , e. n. , & simon , j. d. 2010 , nature , 464 , 72 gunn , j. e. , et al .
2006 , , 131 , 2332 herwig , f. , 2005 , , 43 , 435 hirschi , r. , frlich , c. , liebendorfer , m. , & thilemann , f. -k .
2006 , reviews in modern astronomy , 19 , 101 ito , h. , aoki , w. , honda , s. , & beers , t. c. 2009 , , 698 , l37 ivezi , . ,
2008 , , 684 , 287 izzard , r. g. , glebbeek , e. , stancliffe , r. j. , & pols , o. 2009 , pasa , 26 , 311 karlsson , t. 2006 , , 641 , l41 kennedy , c. r. , sivarani , t. , beers , t. c. , lee , y. s. , placco , v. m. , rossi , s. , christlieb , n. , herwig , f. , & plez , b. 2011 , , 141 , 102 kobayashi , c. , tominaga , n. , & nomoto , k. 2011 , , 730 , l14 lai , d. k. , rockosi , c. m. , bolte , m. , johnson , j. a. , beers , t. c. , lee , y. s. , allende prieto , c. , & yanny , b. 2009 , , 697 , l63 lai , d. k. , lee , y. s. , bolte , m. , lucatello , s. , beers , t. c. , johnson , j. a. , sivarani , t. , & rockosi , c. m. 2011 , , 738 , 51 larson , r. b. 1998 , , 301 , 569 larson , r. b. 2005 , , 359 , 211 lee , y. s. , et al . 2008a , , 136 , 2022 lee , y. s. , et al .
2008b , , 136 , 2050 lee , y. s. , et al .
2011 , , 141 , 90 lucatello , s. , beers , t. c. , christlieb , n. , barklem , p. s. , rossi , s. , marsteller , b. , sivarani , t. , & lee , y. s. 2006 , , 652 , l37 marsteller , b. , beers , t. c. , rossi , s. , christlieb , n. , bessell , m. , & rhee , j. 2005 , nucl .
phys . a , 758 , 312 matsuoka , k. , nagao , t. , maiolino , r. , marconi , a. , & taniguchi , y. 2011 , , 532 , l10 masseron , t. , plez , b. , primas , f. , van eck , s. , & jorissen , a. 2005 , in from lithium to uranium : elemental tracers of early cosmic evolution , iau symposium 228 , p. 219
mccarthy , i. g. , font , a. s. , crain , r. a. , schaye , j. , theuns , t. , & deason , a. j. 2011 , , submitted meynet , g. , ekstrm , s. , & maeder , a. 2006 , , 447 , 623 meynet , g. , hirschi , r. , ekstrm , s. , maeder , a. , georgy , c. , eggenberger , p. , & chiappini , c. 2010a , in chemical abundances in the universe : connecting first stars to planets , iau symposium 265 , p. 98
meynet , g. , hirschi , r. , ekstrm , s. , maeder , a. , georgy , c. , eggenberger , p. , & chiappini , c. 2010b , , 521 nissen , p.e . , & schuster , w.j .
2010 , , 511 , l10 norris , j. e. , christlieb , n. , korn , a. j. , eriksson , k. , bessell , m. s. , beers , t. c. , wisotzki , l. , & reimers , d. 2007 , , 670 , 774 norris , j. e. , yong , d. , gilmore , g. , & wyse , r. f. g. 2010a , , 711 , 350 norris , j. e. , gilmore , g. , wyse , r. f. g. , yong , d. , & frebel , a. 2010b , , 722 , l104 norris , j. e. , wyse , r. f. g. , gilmore , g. , yong , d. , frebel , a. , wilkinson , m. l. , belokurov , v. , & zucker , d. b. 2010c , , 723 , 1632 plez , b. , & cohen , j. g. 2005 , , 434 , 1117 pols , o. r. , izzard , r. g. , glebbeek , e. , & stancliffe , r. j. 2010 , in chemical abundances in the universe : connecting first stars to planets , iau symposium , vol .
265 , p. 117 roederer , i. u. , 2009 , , 137 , 272 rossi , s. , beers , t.c . , sneden , c. , sevastyanenko , t. , rhee , j. , & marsteller , b. 2005 , , 130 , 2804 schlaufman , k. c. , et al .
2009 , , 703 , 2177 schlaufman , k. c. , rockosi , c. m. , lee , y. s. , beers , t. s. , & allende prieto , c. 2011 , , 734 , 49 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 schnrich , r. , asplund , m. , & casagrande , l. 2011 , , 415 , 3807 sivarani , t. , et al .
2006 , , 459 , 125 smolinski , j. p. , et al .
2011 , , 141 , 89 sneden , c. , cowan , j. j. , & gallino , r. 2008 , , 46 , 241 tissera , p. b. , white , s. d. m. , & scannapieco , c. 2011 , , submitted tominaga , n. , umeda , h. , & nomoto , k. 2007 , , 660 , 516 tumlinson , j. 2007 , , 664 , l63 tumlinson , j. 2010 , , 708 , 1398 umeda , h. & nomoto , k. 2003 , nature , 422 , 871 umeda , h. & nomoto , k. 2005 , , 619 , 427 willman , b. , et al .
2005 , , 626 , l85 wall , j. v. , & jenkins , c. r. 2003 , practical statistics for astronomers , ( cambridge university press , cambridge ) wisotzki , l. , christlieb , n. , bade , n. , beckmann , v. , khler , t. , vanelle , c. , & reimers , d. 2000 , , 358 , 77 yanny , b. , et al .
2009 , , 137 , 4377 york , d. g. , et al .
2000 , , 120 , 1579 zolotov , a. , willman , b. , brooks , a. m. , governato , f. , hogg , d. w. , shen , s. , & wadley , j. 2010 , , 721 , 738 zucker , d. b. , et al . 2006 , ,
643 , l103 cccccccccrr sdss j000219.9 + 292851.8 & 2803 - 54368 - 459 & a11 & 6150 & 6184 & 4.00 & 3.24 & @xmath303.26 & @xmath302.97 & @xmath362.63 & @xmath362.73 + sdss j003602.2 - 104336.3 & & a08 & 6500 & 6475 & 4.50 & 4.59 & @xmath302.41 & @xmath302.49 & @xmath362.50 & @xmath362.21 + & & & & & & & & & & + sdss j012617.9 + 060724.8 & 2314 - 53713 - 090 & a08 & 6600 & 6841 & 4.10 & 4.73 & @xmath303.11 & @xmath302.80 & @xmath362.92 & @xmath362.67 + sdss j012617.9 + 060724.8 & 2314 - 53713 - 090 & a11 & 6900 & 6841 & 4.00 & 4.73 & @xmath303.01 & @xmath302.80 & @xmath363.08 & @xmath362.67 + & & & & & & & & & & + sdss j025956.4 + 005713.3 & 1513 - 53741 - 338 & a11 & 4550 & 4537 & 5.00 & 4.31 & @xmath303.31 & @xmath303.75 & @xmath300.02 & @xmath360.00 + sdss j030839.3 + 050534.9 & 2335 - 53730 - 314 & a11 & 5950 & 5938 & 4.00 & 3.36 & @xmath302.19 & @xmath302.41 & @xmath362.36 & @xmath362.31 + sdss j035111.3 + 102643.2 & 2679 - 54368 - 543 & a11 & 5450 & 5542 & 3.60 & 3.14 & @xmath303.18 & @xmath302.89 & @xmath361.55 & @xmath361.61 + sdss j071105.4 + 670128.2 & 2337 - 53740 - 564 & a11 & 5350 & 5252 & 3.00 & 2.70 & @xmath302.91 & @xmath302.91 & @xmath361.98 & @xmath362.16 + sdss j072352.2 + 363757.2 & 2941 - 54507 - 222 & a11 & 5150 & 5105 & 2.20 & 2.33 & @xmath303.32 & @xmath303.48 & @xmath361.79 & @xmath361.39 + sdss j074104.2 + 670801.8 & 2939 - 54515 - 414 & a11 & 5200 & 5171 & 2.50 & 2.35 & @xmath302.87 & @xmath302.92 & @xmath360.74 & @xmath360.84 + & & & & & & & & & & + sdss j081754.9 + 264103.8 & & a08 & 6300 & 6075 & 4.00 & 3.58 & @xmath303.16 & @xmath303.03 & @xmath2032.20&@xmath2031.54 + sdss j081754.9 + 264103.8 & & a11 & 6050 & 6075 & 4.00 & 3.58 & @xmath302.85 & @xmath303.03 & & @xmath2031.54 + & & & & & & & & & & + sdss j091243.7 + 021623.7 & & a11 & 6150 & 6138 & 4.00 & 3.40 & @xmath302.68 & @xmath302.76 & @xmath362.05 & @xmath362.33 + sdss j091243.7 + 021623.7 & & b10 & 6500 & 6138 & 4.50 & 3.40 & @xmath302.50 & @xmath302.76 & @xmath362.17 & @xmath362.33 + & & & & & & & & & & + sdss j092401.9 + 405928.7 & & a08 & 6200 & 6201 & 4.00 & 4.53 & @xmath302.51 & @xmath302.81 & @xmath362.72 & @xmath362.56 + sdss j103649.9 + 121219.8 & & b10 & 6000 & 5873 & 4.00 & 3.12 & @xmath303.20 & @xmath303.31 & @xmath361.47 & @xmath361.94 + sdss j124123.9 - 083725.5 & 2689 - 54149 - 292 & a11 & 5150 & 5108 & 2.50 & 2.45 & @xmath302.73 & @xmath302.90 & @xmath360.50 & @xmath360.65 + sdss j124204.4 - 033618.1 & 2897 - 54585 - 210 & a11 & 5150 & 5112 & 2.50 & 2.56 & @xmath302.77 & @xmath303.02 & @xmath360.64 & @xmath360.77 + & & & & & & & & & & + sdss j134913.5 - 022942.8 & & a11 & 6200 & 6167 & 4.00 & 4.39 & @xmath303.24 & @xmath303.16 & @xmath363.01 & @xmath362.61 + sdss j134913.5 - 022942.8 & & b10 & 6200 & 6167 & 4.00 & 4.39 & @xmath303.00 & @xmath303.16 & @xmath362.82 & @xmath362.61 + & & & & & & & & & & + sdss j161226.2 + 042146.6 & 2178 - 54629 - 546 & a11 & 5350 & 5365 & 3.30 & 2.42 & @xmath302.86 & @xmath303.19 & @xmath360.63 & @xmath360.98 + sdss j161313.5 + 530909.7 & 2176 - 54243 - 614 & a11 & 5350 & 5338 & 1.60 & 2.67 & @xmath303.32 & @xmath302.89 & @xmath362.09 & @xmath361.64 + sdss j164610.2 + 282422.2 & & a11 & 6100 & 6125 & 4.00 & 3.45 & @xmath303.05 & @xmath302.78 & @xmath362.52 & @xmath362.53 + sdss j170339.6 + 283649.9 & 2808 - 54524 - 510 & a11 & 5100 & 5120 & 4.80 & 3.58 & @xmath303.21 & @xmath303.32 & @xmath360.28 & @xmath360.43 + sdss j170733.9 + 585059.7 & 0353 - 51703 - 195 & a08 & 6700 & 6567 & 4.20 & 3.47 & @xmath302.52 & @xmath302.68 & @xmath362.10 & @xmath362.43 + sdss j173417.9 + 431606.5 & 2799 - 54368 - 138 & a11 & 5200 & 5172 & 2.70 & 2.15 & @xmath302.51 & @xmath303.18 & @xmath361.78 & @xmath362.43 + sdss j204728.8 + 001553.8 & & a08 & 6600 & 6324 & 4.50 & 3.77 & @xmath302.05 & @xmath302.28 & @xmath362.00 & @xmath361.98 cccccc @xmath301.5 @xmath19 [ fe / h ] @xmath100 @xmath301.0 & 9113 & 149 & 335 & 3 & 1.6 + @xmath302.0 @xmath19 [ fe / h ] @xmath100 @xmath301.5 & 4376 & 377 & 2651 & 226 & 5.1 + @xmath302.5 @xmath19 [ fe / h ] @xmath100 @xmath302.0 & 506 & 173 & 1387 & 458 & 8.4 + @xmath304.0 @xmath19 [ fe / h ] @xmath100 @xmath302.5 & 13 & 46 & 201 & 201 & 18.0 = 0 ( the starting value ) .
the legend at the top of this panel lists the input values determined by the sspp for @xmath57 , @xmath58 , and [ fe / h ] .
middle panels : best matches to the ch g - band obtained from the three different search ranges considered ( left : [ c / fe ] @xmath204 ; center : [ c / fe ] covering the full grid range ; right : [ c / fe ] @xmath128 ) .
the red line in each case is the best match ; the green line is the ratio of the best match to the input spectrum .
the legend above each of these panels provides the best estimate of [ c / h ] , along with a correlation coefficient between the synthetic spectrum and the input spectrum ( cc ) , and the reduced @xmath205 of the fit ( chi2 ) .
note in the left - hand panel that no value of [ c / fe ] @xmath204 provides an acceptable match .
the center and right - hand panels shown are identical , because the routine converged to the same value for the full grid search and the search restricted to [ c / fe ] @xmath128 .
lower panel : final adopted match , represented by the superposed red line .
the legend above this panel lists the resulting derived [ c / fe ] . ] , [ fe / h ] ) , and carbonicity , [ c / fe ] , as derived from the low - resolution spectra ( sspp ) and the high - resolution follow - up spectra ( high ) .
a one - to - one line is shown for each .
the open circles correspond to the stars with @xmath58 values that were assigned , than than derived ( see text ) .
right panels : histograms of the differences between the sssp and high determinations .
the dashed bins indicate the stars with @xmath58 values that were assigned , rather then derived .
no comparison is made for the @xmath57 determinations , as the high - resolution estimates for the majority of our comparison sample were taken directly from the sspp estimates ( see text ) .
also see text for discussion of the reason for the apparently large dispersion in determination of @xmath58 .,scaledwidth=70.0% ] estimates on the derivation of [ c / fe ] ( plotted as @xmath67[c / fe ] ) , as described in the text .
the top row of panels correspond to synthetic spectra with [ fe / h ] @xmath206 , the middle panels to [ fe / h ] @xmath7 , and the bottom panels to [ fe / h ] @xmath207 . from left to right , the columns of panels correspond to input perturbations in @xmath58 of @xmath38 dex , @xmath94 dex , @xmath95 dex , and @xmath96 dex , respectively .
the colored dots and lines shown in each panel corespond to input surface gravities ( prior to perturbing their values ) of @xmath58 = 2.0 , 3.0 , and 4.0 . the mean zero - point offsets and rms variations in the derived [ c / fe ] , relative to the known value across
all @xmath57 and @xmath58 considered , are shown for each panel .
the case illustrated here is for [ c / fe ] = 0.0 ; similar results pertain to the case [ c / fe]@xmath208.,scaledwidth=110.0% ] , as a function of metallicity , [ fe / h ] , for subsample d ( top panel ) and subsample l ( bottom panel ) .
the dashed red line represents the solar carbon abundance ratio [ c / fe ] @xmath209 ; the solid red line denotes the adopted limit that divides c - norm stars from c - rich stars , [ c / fe ] @xmath210 .
the `` ridge lines '' in both of these plots are due to grid effects in the chi - square fitting procedure.,scaledwidth=60.0% ] kpc , no restriction on @xmath112 ) , and for the local sample ( red histogram ; @xmath211 kpc ; 7 kpc @xmath212 kpc ) .
lower panel : as - observed carbonicity distribution function ( cardf ) for the extended sample of calibration stars ( black histogram ) , and for the local sample ( red histogram ) .
note that the extended sample contains significant numbers of dwarfs , main - sequence turnoff stars , and subgiants / giants ( 77% , 13% , and 10% , respectively ) , while the local sample primarily comprises dwarfs ( 88% ) and main - sequence turnoff stars ( 9% ) , due to the larger volume explored by the subgiants and giants .
the inset shows a rescaled view of the high - carbonicity tail . for both panels , only stars with carbon detections are used.,scaledwidth=60.0% ] z@xmath10 .
the red arrows indicate the locations of the metallicity peaks of the mdf for the thick disk ( @xmath39 ) , the mwtd ( @xmath213 ) , the inner halo ( @xmath214 ) , and the outer halo ( @xmath215 ) , respectively , assigned by c10 .
second and fourth columns : as - observed carbonicity distribution functions ( cardfs ) , for the extended sample of sdss / segue dr7 calibration stars , as a function of vertical distance from the galactic plane .
the blue arrows show the location of the solar carbon - to - iron ratio ( [ c / fe ] = 0.0 ) , and the natural threshold that divides carbon - normal from carbon - rich stars ( [ c / fe ] @xmath0 ) . the histograms represent the cardfs obtained at different cuts of @xmath10z@xmath10 . for all panels , only stars with carbon detections are used.,scaledwidth=80.0% ] ( bottom panels ) , for stars considered cemp ( [ c / fe ] @xmath44 ; dot - dashed red histograms ) , and non - cemp ( [ c / fe ] @xmath136 ; solid black histograms ) , for two cuts on z@xmath116 .
the stars are selected from the extended sample with [ fe / h ] @xmath133 .
the left - hand column of panels includes stars at all z@xmath116 ; the right - hand column of panels is only for stars satisfying z@xmath116 @xmath216 kpc . for all panels , only stars with carbon detections are used.,scaledwidth=90.0% ] @xmath1 and two different cuts on @xmath127 .
the left - hand column of panels includes stars at all z@xmath116 ; the right - hand column of panels is only for stars satisfying z@xmath116 @xmath216 kpc .
the small blue dots represent the entire sample satisfying these cuts ; the filled red stars represent stars belonging to the highly retrograde tails shown in the bottom panels of figure 7 .
the upper panels correspond to the sample of retrograde stars with @xmath129 @xmath132 km s@xmath29 ; the lower panels correspond to the sample of retrograde stars with @xmath129 @xmath131 km s@xmath29 .
clear differences in the distributions of proper motions exist between the stars in the highly retrograde tail and the rest of the sample .
the dashed lines provide a zero proper motion reference frame . for all panels , only stars with carbon detections are used.,scaledwidth=100.0% ] and the cylindrical velocity components ( v@xmath134,v@xmath27,v@xmath135 ) for the sdss / segue calibration stars in the extended sample . the strong correlation between z@xmath116 and the radial velocity component shown in the upper panel , v@xmath134 , as well as with the vertical velocity component shown in the lower panel , v@xmath135 ,
is confirmed for the extended sample .
the middle panel exhibits no strong correlation between z@xmath116 and the rotational velocity component , v@xmath217 , other than that expected from the presence of the thick - disk and halo populations .
the dashed line is a reference line at zero velocity for each component .
note , in the middle panel , the clear excess of stars with retrograde motions for z@xmath116 @xmath13 10 - 15 kpc , which we associate with the outer - halo component .
the panels include all stars , regardless of their carbon status.,scaledwidth=45.0% ] , [ fe / h ] , and [ c / fe ] , for the extended sample of sdss / segue calibration stars with detected ch g - bands .
upper panel : galactocentric rotational velocity as a function of the metallicity .
middle panel : galactocentric rotational velocity as a function of [ c / fe ] .
lower panel : metallicity vs. carbonicity , [ c / fe ] . in all panels ,
the gray dots represent stars with low carbonicity , [ c / fe ] @xmath136 ; the red dots denote the stars with high carbonicity , [ c / fe ] @xmath13 + 0.7.,scaledwidth=50.0% ] , for the low - metallicity stars of the extended sample with @xmath127 @xmath12 kpc .
each bin of metallicity has width @xmath67[fe / h ] = 0.2 dex , with the exception of the lowest - metallicity bin , which includes all stars with [ fe / h ] @xmath218 .
the error bars are evaluated with the jackknife method .
the calculation of cemp star frequency is made using eqn .
2 , which includes all stars of known carbon status , including the l subsample . a second - order polynomial fit to the observed distribution is shown by the dot - dashed line .
right panel : global trend of the mean carbonicity , @xmath4[c / fe]@xmath5 , as a function of [ fe / h ] . only those stars with detected ch g - bands are used .
the bins are the same as used in the left panel .
errors are the standard error in the mean.,scaledwidth=110.0% ] z@xmath10 , for the low - metallicity stars of the extended sample .
the stars with @xmath302.0 @xmath19[fe / h ] @xmath133 are shown as filled black circles ; those with [ fe / h ] @xmath2 are shown as magenta stars .
each bin has a width of @xmath67@xmath10z@xmath10 = 4 kpc , with the exception of the last bin , which cuts off at 10 kpc , the limiting distance of the extended sample .
errors are derived using the jackknife approach .
the calculation of cemp star frequency is made using eqn .
2 , which includes all stars of known carbon status , including the l subsample . note the clear dependence of cemp fraction on height above the galactic plane , for both metallicity regimes.,scaledwidth=80.0% ] @xmath19 @xmath302.0 and z@xmath116 @xmath13 5 kpc .
upper panel : the black histogram represents the observed distribution of v@xmath27 ; the green ( inner halo ) , and red ( outer halo ) curves show the results of the extreme deconvolution analysis .
middle panel : the observed velocity distribution function ( black histogram ) , and the inner- and outer - halo weighted velocity distribution functions ( frequencies ) , denoted by the green and the red histograms , respectively . in this panel ,
all the distributions are normalized and take into account the membership probability of each star .
lower panel : the posterior probability for the inner- and outer - halo components as a function of the rotational velocity , v@xmath27 .
note that these plots use all stars in the extended sample satisfying the stated cuts , regardless of whether or not they have detected ch g - bands.,scaledwidth=45.0% ] @xmath19 @xmath302.0 and z@xmath116 @xmath13 5 kpc and with detected ch g - bands .
top panel : cardf for the entire sample .
bottom panel : weighted cardf for the inner- and outer - halo components , shown by the green and red histograms , respectively.,scaledwidth=60.0% ] for the low - metallicity stars of the extended sample with @xmath127 @xmath12 kpc . the dashed curve is a second - order polynomial fit representing the global trend .
the blue filled stars represent the predicted values of the cemp fractions in each bin of metallicity , having a width of @xmath67[fe / h ] = 0.5 dex .
the green and red filled circles show the location of the observed cemp star fractions for the inner halo and outer halo , respectively , based on the xd method with the hard - cut - in - probability approach ( see text ) .
the error bars are evaluated with the jackknife method .
the calculation is made using eqn .
2 , which includes all stars of known carbon status , including the l subsample .
right panel : global trend of the mean carbonicity , @xmath4[c / fe]@xmath5 , as a function of [ fe / h ] . only those stars with detected ch g - bands are used .
the green and red filled circles show the mean carbonicity for the inner halo and outer halo , respectively , based on the extreme deconvolution method with the hard - cut - in - probability approach ( see text ) .
the bins are the same as used in the left panel .
errors are the standard error in the mean .
no significant difference is seen for the inner- and outer - halo subsamples.,scaledwidth=120.0% ]
|
carbon - enhanced metal - poor ( cemp ) stars in the halo components of the milky way are explored , based on accurate determinations of the carbon - to - iron ( [ c / fe ] ) abundance ratios and kinematic quantities for over 30000 calibration stars from the sloan digital sky survey ( sdss ) . using our present criterion that low - metallicity stars exhibiting [ c / fe ] ratios ( `` carbonicity '' ) in excess of [ c / fe]@xmath0 are considered cemp stars , the global frequency of cemp stars in the halo system for [ fe / h ]
@xmath1 is 8% ; for [ fe / h ] @xmath2 it is 12% ; for [ fe / h ] @xmath3 it is 20% .
we also confirm a significant increase in the level of carbon enrichment with declining metallicity , growing from @xmath4[c / fe]@xmath5 @xmath6 at [ fe / h ] @xmath7 to @xmath4[c / fe]@xmath5 @xmath8 at [ fe / h ] @xmath9 .
the nature of the carbonicity distribution function ( cardf ) changes dramatically with increasing distance above the galactic plane , @xmath10z@xmath10 . for @xmath10z@xmath10 @xmath11 kpc ,
relatively few cemp stars are identified . for distances
@xmath10z@xmath10 @xmath12 kpc , the cardf exhibits a strong tail towards high values , up to [ c / fe ] @xmath13 + 3.0 .
we also find a clear increase in the cemp frequency with @xmath10z@xmath10 . for stars with @xmath14 [ fe / h ] @xmath151.5 , the frequency grows from 5% at @xmath10z@xmath10 @xmath16 kpc to 10% at @xmath10z@xmath10 @xmath17 kpc . for stars with [ fe / h ] @xmath152.0
, the frequency grows from 8% at @xmath10z@xmath10 @xmath16 kpc to 25% at @xmath10z@xmath10 @xmath17 kpc . for stars with @xmath14 [ fe / h ] @xmath151.5 , the mean carbonicity is @xmath4[c / fe]@xmath5 @xmath18 for 0 kpc @xmath19 @xmath10z@xmath10 @xmath19 10 kpc , with little dependence on @xmath10z@xmath10 ; for [ fe / h ] @xmath152.0 , @xmath4[c / fe]@xmath5 @xmath20 , again roughly independent of @xmath10z@xmath10 .
based on a statistical separation of the halo components in velocity space , we find evidence for a significant contrast in the frequency of cemp stars between the inner- and outer - halo components the outer halo possesses roughly twice the fraction of cemp stars as the inner halo .
the carbonicity distribution also differs between the inner - halo and outer - halo components the inner halo has a greater portion of stars with modest carbon enhancement ( [ c / fe ] @xmath21 $ ] ) ; the outer halo has a greater portion of stars with large enhancements ( [ c / fe ] @xmath22 ) , although considerable overlap still exists .
we interpret these results as due to the possible presence of additional astrophysical sources of carbon production associated with outer - halo stars , beyond the asymptotic giant - branch source that may dominate for inner - halo stars , with implications for the progenitors of these populations .
|
pattern formation in ferromagnetic thin films with perpendicular anisotropy and its thermodynamical description have been the subject of intense experimental @xcite , theoretical @xcite and numerical @xcite work in the last 20 years .
magnetic order in ultrathin ferromagnetic films is very complex due to the competition between exchange and dipolar interactions on different length scales , together with a strong influence of shape and magnetocrystalline anisotropies of the sample .
these in turn are very susceptible to the growth conditions of the films @xcite . among the different magnetization patterns that have been observed in these systems , striped order ( i.e. , modulated patterns of local perpendicular magnetization with a well defined half - wavelength or stripe width @xmath0 ) at low temperatures is an ubiquitous phenomenon .
one intriguing fact is the strong variation displayed by the equilibrium stripe width @xmath0 in many of these systems , when either the temperature or the film thickness is changed@xcite .
the origin of such variation is still controversial , but recent results suggest that a key point to understand it is the role played by the interfaces ( i.e. , the domain walls ) between stripes@xcite . thus , a starting point to study this problem is to compare the energies of striped patterns with different domain wall configurations .
an accurate description of the domain walls requires to take into account not only the perpendicular component of the local magnetization , but also the in
plane component .
indeed , some experimental results@xcite are consistent with the presence of bloch domain walls , as expected for perpendicularly oriented magnetization domains@xcite .
to compute the energy contribution of domain walls it is important to consider explicitly the out of plane anisotropy , together with the exchange and dipolar interactions , whose competition is the responsible for the appearance of striped patterns .
a minimum model that contains all these ingredients is the 2d dimensionless heisenberg hamiltonian : @xmath1 - \eta \sum_{i } ( s_i^z)^2 \label{hamiltoniano}\ ] ] where @xmath2 are classical unit vectors , the exchange and anisotropy constants are normalized relative to the dipolar coupling constant ( @xmath3 ) , @xmath4 stands for a sum over nearest neighbors pairs of sites in a square lattice , @xmath5 stands for a sum over _ all distinct _ pairs and @xmath6 is the distance between spins @xmath7 and @xmath8 . in the large anisotropy limit @xmath9 this model reduces to an ising model with short range ferromagnetic and long range antiferromagnetic interactions , whose ground state is the striped one@xcite . in that limit
the stripe width increases exponentially with the exchange to dipolar coupling ratio@xcite @xmath10 . for low values of the anisotropy ,
the ground state of this model changes to a planar ferromagnetic state@xcite . in a classic work ,
yafet and gyorgy computed the energy of striped domain configurations with bloch domain walls , by considering a sinusoidal ( perpendicular ) magnetization profile at the walls and saturated magnetization inside the stripes@xcite .
they found that above certain threshold value @xmath11 a striped configuration has less energy than a uniformly ( in plane ) magnetized one . at this point
the system shows a spin reorientation transition ( srt ) . also within this approximation
the stripe width shows an exponential increase with the anisotropy strength , while the domain wall width decreases algebraically for large enough values of @xmath12 , at least for large values of @xmath13 .
this approximation is expected to work well close to the srt , where the effective anisotropy is small@xcite . however , for large values of @xmath13 and @xmath12 not too close to the srt ( i.e. , when the width of both stripes and walls are large ) one can expect a large departure in the wall energy contribution with respect to the true domain wall configuration , which should approach a hyperbolic tangent profile@xcite .
another important point concerns the in
plane magnetization component inside the stripes . in their work yafet and gyorgy considered striped solutions where the only in
plane components lay inside the walls , although they pointed out how to extend their calculations to consider non saturated magnetization inside the stripes@xcite .
based on that calculation , politi@xcite reported that , at least for large enough values of @xmath13 , the magnetization should show an abrupt saturation very close to the srt , suggesting that the in plane component inside the stripes is not relevant .
however , recently whitehead et al@xcite obtained evidences of a non saturated ground state at a relatively small value of @xmath13 for a wide range of values of the anisotropy strength @xmath12 .
their numerical results also showed a correlation between the stripe width and the in
plane component variations and suggest that this ground state configuration stabilizes at finite temperature , giving rise to what they called a canted " phase .
hence , it is important to revise the zero temperature phase diagram of this model in the whole @xmath14 space , including canted configurations , in order to determine to what extent they can be relevant to real systems and their possible influence to the stripe width variation phenomenon .
it is worth noting that the perpendicular anisotropy changes ( inversely ) with the film thickness@xcite .
it has also been pointed out that the changes in the film thickness can act as a change in the effective temperature@xcite .
hence , understanding the variation of the equilibrium stripe width and the associated domain wall structure as the anisotropy changes at zero temperature can be of great help to understand the corresponding properties a finite temperature , a far more complex problem . in this work
we analyze the complete equilibrium phase diagram at zero temperature in the @xmath14 space of hamiltonian ( [ hamiltoniano ] ) .
following yafet and gyorgy s work , we only consider straight domains , i.e. , domains in which the spin orientation can be modulated along the @xmath15 direction but is constant in the perpendicular direction @xmath16 ( @xmath17 are the coordinates on the plane of the film ) .
we also consider only bloch walls , i.e. , walls in which the magnetization stays inside the @xmath18 plane . in order to clarify notation and units , and to include some extensions , we briefly review yafet & gyorgy s approximation in appendix [ yafet ] . to obtain the phase diagram
we compute the energy of different types of magnetization profiles and compare them with simulation results obtained through a zero temperature monte carlo ( mc ) method specially designed for the present purposes .
the mc method is presented in appendix [ mcmethod ] . in section [ energysection ]
we derive a general expression for the energy of striped magnetization profiles .
the zero temperature phase diagram and associated properties are derived in section [ phasediagram ] .
a discussion and conclusions are presented in section [ discu ] .
let us consider a square lattice with @xmath19 sites , characterized by the integer indexes @xmath17 , where @xmath20 and @xmath21 , in the limit @xmath22 .
hence , the index @xmath7 in eq.([hamiltoniano ] ) denotes a pair of coordinates @xmath17 . following yg approximation@xcite
, we consider only uniformly magnetized solutions along every vertical line of sites , i.e. @xmath23 , @xmath24 and allow only bloch walls between domains of perpendicular magnetization , i.e. @xmath25 @xmath26 .
therefore , @xmath27 ^ 2 + \left[m^y(x ) \right]^2=1.\ ] ] then , for every value of @xmath15 there is only one independent component of the magnetization .
yg showed that for these types of spin configurations the energy per spin can be mapped onto the energy of a one dimensional xy model .
the energy difference between an arbitrary magnetization profile @xmath28 and a uniformly in
plane magnetized state is then given by ( see appendix [ yafet - energy ] ) : @xmath29= ( \delta-2c_2 ) - \frac{\delta'}{l } \sum_x \vec{m}(x ) .
\vec{m}(x+1 ) + \frac{1}{l } \sum_{x , x ' } \frac{m^z(x)\ , m^z(x')}{|x - x'|^2 } - \frac{\kappa'}{l } \sum_x \left[m^z(x ) \right]^2 + c \label{energia1}\ ] ] where @xmath30 , @xmath31 , @xmath32 , @xmath33 , @xmath34 and @xmath35= 2(c_2-c_1)\frac{1}{l } \sum_x m^y(x)\ , m^y(x+1)\ ] ] although small , this correction term makes a non negligible contribution when the domain walls are of the same order of the lattice constant .
this happens for small values of @xmath13 ( @xmath36 ) , where both the stripe and wall widths are of the order of a few lattice spacings . for larger values of @xmath13
it is reasonable to assume a smooth magnetization profile@xcite @xmath37 , so that @xmath38\ ] ] can be absorbed into the anisotropy term in eq.([energia1 ] ) , replacing @xmath39 .
considering now a stripe - like periodic structure of the magnetization profile with period @xmath40 , i.e. @xmath41 we can make use of a fourier expansion : @xmath42 where we have assumed @xmath43 an even function of @xmath15 just for simplicity . the anisotropy term in eq.([energia1 ] ) can be written as @xmath44 ^ 2= - \frac{\kappa'\ , m_0 ^ 2}{2 } \sum_{m=1,3,\ldots } b_m^2,\ ] ] and the dipolar term@xcite @xmath45 where@xcite @xmath46 in the general case , it is better to let the exchange term ( and the correction term @xmath47 as well ) in eq.([energia1 ] ) expressed in terms of the angle @xmath48 between @xmath28 and the @xmath49 axis : @xmath50}\\ m^y(x ) & = & \sin{[\phi(x)]}\end{aligned}\ ] ] where the angle @xmath48 has the same periodicity of @xmath28 .
we have that @xmath51}.\ ] ] putting all the terms together we get the general expression : @xmath52= ( \delta-2c_2 ) - \delta ' \frac{1}{l } \sum_x \cos{\left [ \phi(x)-\phi(x+1)\right ] } + m_0 ^ 2 \sum_{m=1,3,\ldots } b_m^2 d_m(h ) - \frac{\kappa'\ , m_0 ^ 2}{2 } \sum_{m=1,3,\ldots } b_m^2 + c .
\label{energiabm}\ ] ]
in this section we look for the minimum of eq.([energiabm ] ) for different values of @xmath53 .
we propose different striped magnetization profiles @xmath43 and compare the energies obtained by minimizing eq.([energiabm ] ) for each profile with respect to variational parameters .
we first consider a profile as proposed by yg , that is constant inside the stripes and has a sinusoidal variation inside the walls between stripes ( see fig.1 in ref . ):
@xmath54 where @xmath55 is the absolute value of magnetization inside the stripes and @xmath56 is the wall width .
in order to allow for canted profiles , we take @xmath57 , where @xmath58 is the canted angle , i.e. we define it as the minimum angle of the local magnetization with respect to the @xmath49 axis .
yafet and gyorgy solved this variational problem for @xmath59 in the continuum limit@xcite , i.e. when @xmath60 and @xmath61 , so that the profile can be considered a smooth function of @xmath15 .
while this approximation is expected to work well for large enough values of @xmath13 , it breaks down for relatively small values of it , where the discrete character of the lattice has to be taken into account .
however , the variational problem for that range of values of @xmath13 can be solved exactly ( although numerically ) by minimizing eq.([energiabm ] ) with respect to the _ integer _ variational parameters @xmath0 and @xmath56 and continuous parameter @xmath58 .
in other words , for every pair of values @xmath14 we evaluate the energy eq.([energiabm ] ) for the profile ( [ perfilyg ] ) with different combinations of @xmath62 and @xmath63 within a limited set .
for every pair of values @xmath64,we look for the value of @xmath58 that minimizes the energy with a resolution @xmath65 and compare all those energies .
details of that evaluation are given in appendix [ ygdiscrete ] .
this calculation is feasible for values up to @xmath66 , for which the maximum value of @xmath0 ( bounded by the stripe width in the @xmath67 limit ) remains relatively small ( smaller than @xmath68 ) .
some results for @xmath69 close to the srt were also obtained .
all the results of this calculation are compared against monte carlo ( mc ) simulations .
details of the mc method used are given in appendix [ mcmethod ] .
through these calculations we obtain a zero temperature phase diagram for low values of @xmath13 . before presenting the results , it is important to introduce some notations and definitions of different types of solutions .
we distinguish between four types of solutions .
if the minimum energy solution corresponds to @xmath70 and @xmath71 ( within the resolution @xmath72 ) , we call this a _ striped ising profile _ ( sip ) , i.e. a square wave like profile . if @xmath71 but @xmath73 , we call this a _
saturated state_. these states show a finite parallel component of the magnetization inside the walls .
if @xmath74 the solution is a _
canted state_. finally , if @xmath75 ( @xmath76 ) we have a _ planar ferromagnet _
( pf ) .
the zero temperature phase diagram for small values of @xmath13 ( @xmath77 ) is shown in fig.[pd1 ] .
for relatively large values of @xmath12 the minimum energy configuration is always the ising one ( sip ) , with a stripe width independent of @xmath12 . for small values of @xmath12
the minimum energy configuration is the pf , with a spin reorientation transition line ( srt ) , either to the ising state for @xmath78 ( @xmath79 ) or to a canted one for @xmath80 ( @xmath81 ) .
no saturated state configurations are observed for @xmath82 .
( color online ) zero temperature phase diagram for small values of @xmath13 .
black filled symbols and black solid lines : mc simulations .
open red symbols : swp approximation . squares and continuous black lines
correspond to transition lines between striped states of different width .
the shaded region corresponds to the canted state ( @xmath83 ) .
triangles are transition lines between planar ferromagnet ( @xmath84 ) and canted states ( spin reorientation transition line ) .
circles mark transitions between the canted and the stripes ising state ( @xmath71 and @xmath70 ) .
notice the excellent agreement between the mc and swp calculations close to the srt , while the swp approach underestimates the transition line between the canted and ising stripes states .
the dotted line corresponds to the contimuum approximation of yg for the srt ( eq.([etasrt ] ) ) . ] inside the canted region , a strong stripe width variation with the anisotropy is observed at constant @xmath13 .
note that the vertical lines that separate ising striped states with consecutive values of @xmath0 bend inside the canted region and become almost horizontal as @xmath13 increases .
hence , the exponential increase of @xmath0 with @xmath13 in the ising region ( vertical lines ) changes to an exponential increase with @xmath12 deep inside the canted region ( curved lines on the right of fig.[pd1 ] ) .
we also find an excellent agreement between the sinusoidal wall profile approximation and the mc results , except close to the transition between the ising and the canted states .
such disagreement is due to the fact that the actual wall is not well described by a sinusoidal profile far away of the srt line , as will be shown later . in fig.[energyygvsmc ] we show a comparison between the energy of the swp and the mc results as a function of @xmath12 for @xmath85 . the range of values of @xmath12 where the walls are not well described by a sinusoidal profile increases with @xmath13 .
( color online ) energy per spin ( with respect to the pf state ) vs. @xmath12 for @xmath85 within the swp ( full red line ) and mc calculations ( filled black circles ) .
open black circles correspond to the energy of a striped ising profile ( sip ) with @xmath86 ( equilibrium value for @xmath9 ) . ] for large enough values of @xmath13 the variational problem for the swp can be solved in a continuum approximation introduced by yg@xcite .
this leads to the equations ( see appendix [ ygcontinuum ] ) : @xmath87 ( 1+\sin{\theta } ) \label{ygeq2 - 2}\\ \frac{\delta\ , k^2}{2\delta } & = & - \left [ 2\gamma \left(1-\frac{\delta}{2 } \right)- \pi\ , k\ , g(\delta)\right ] \sin{\theta } , \label{ygeq3'}\end{aligned}\ ] ] where @xmath88 , @xmath89 and @xmath90 . in the limit @xmath91 ( pure sinusiodal profile )
these equations can be solved analytically and predict a srt at the line @xmath92 with @xmath93 ( see appendix [ ygcontinuum ] ) .
the line eq.([etasrt ] ) is also depicted in fig.[pd1 ] .
notice the disagreement between the continuum approximation and the exact one for @xmath77 .
this discrepancy becomes smaller than @xmath94 only for @xmath95 .
for arbitrary values of @xmath12 and @xmath13 eqs.([ygeq1 - 2])-([ygeq3 ] ) can be solved numerically . in fig.[yg - cont - num ] we show the numerical solutions for @xmath58 and @xmath0 as a function of @xmath12 for different values of @xmath13 .
we see that the range of values of the anisotropy @xmath12 for which the canted angle is appreciable different from zero within this approximation is strongly depressed as @xmath13 increases . for values
@xmath96 the canted configuration almost disappears , except very close to the reorientation line , consistently with the results reported by politi@xcite .
( color online ) ( a ) canted angle @xmath58 and ( b ) equilibrium stripe widht @xmath0 , as a function of the anisotropy @xmath12 for different values of @xmath13 from the continuum approximation of the swp eqs.([ygeq1 - 2])-([ygeq3 ] ) .
the full grey line at the left corresponds to the srt borderline given by eq.([etasrt ] ) with @xmath97 ( see appendix [ ygcontinuum ] ) . ] indeed , from our mc simulations , we observe that the range of values for which the canted state has the minimum energy gradually shrinks as @xmath13 increases , being replaced by a saturated state for values of @xmath12 above certain threshold .
this can be observed in fig.[mxsvseta ] , where we show the behavior of the canted angle and the in
plane magnetization component @xmath98 as a function of @xmath12 for @xmath99 .
the monte carlo data shows the existence of a wide range of values of @xmath12 for which the canted angle is zero while @xmath100 , meaning that the non null in
plane components are concentrated inside the walls . in other words ,
in that region we have a saturated state with thick walls @xmath101 .
notice also that the swp approach completely fails to describe those states .
moreover , we observe from our mc simulations that the swp cease to be the minimum energy solution for values of @xmath12 relatively close to the srt , well before the saturated state sets up ( see fig.[mxsvseta ] ) .
this effect becomes more marked as @xmath13 increases .
as already pointed out , the actual magnetization profile departs from the swp for large values of @xmath12 and @xmath13 .
this is expected from micromagnetic theory , which in that limit predicts that the wall structure will be dominated by the interplay between anisotropy and exchange , leading to an hyperbolic tangent shape of the wall@xcite .
this can be observed in fig.[profiletanh ] .
hence we considered a periodic magnetization profile with hyperbolic tangent walls ( hwp ) defined , for a wall centered at @xmath102 , by @xmath103 together with eq.([antiperiodic ] ) where @xmath104 as before .
in the large @xmath13 limit , assuming a smooth profile @xmath105 and @xmath106 , the anisotropy energy can be expressed as : @xmath107 ^ 2 \nonumber\\ & \approx & - \kappa\,m_0 ^ 2 \left [ 1-\frac{2\,l_w}{h}\tanh{\left ( \frac{h}{2\ , l_w } \right)}\right ] .
\label{anitanh}\end{aligned}\ ] ] ( color online ) canted angle ( circles ) and in - plane magnetization ( squares ) as a function of @xmath12 for @xmath99 . filled black symbols correspond to mc calculation .
open red symbols corresponds to the discrete swp approximation , while the red doted and dashed lines correspond to the continuum ( yg ) approximation of the swp .
continuous black and red lines are only a guide to the eye . ]
the exchange energy can be obtained in a similar way : @xmath108 solving the last integral we finally obtain @xmath109 . \label{eexchhwp2}\ ] ] the dipolar energy can be calculated using eqs.([edipinf ] ) and ( [ dm ] ) . the fourier coefficients for the profile ( [ hyperbolicprofile ] ) can be computed using the approximation ( [ perfiltanh ] ) ( see appendix [ fouriertangh ] ) .
this leads to an expression for the total energy as a function of the variational parameters @xmath0 , @xmath58 and @xmath110 that can be minimized numerically . comparing the minimum energy solution for the swp and the hwp we obtain the crossover line between sinusoidal and hyperbolic wall structure shown in fig.[pd2 ] ( dashed line ) . above that line the hwp has always less energy than the swp .
we also calculated the transition line between the canted and the saturated states by setting the condition @xmath111 , to be consistent with the criterium used in the mc calculations .
the results are shown in fig.[pd2 ] together with the srt line eq.([etasrt ] ) , and compared with mc calculations up to @xmath112 .
the excellent agreement with the mc results gives support to the analytic approximations . for large values of @xmath12
the exponential increase of @xmath0 makes it cumbersome to apply the approximation of appendix [ fouriertangh ] for the calculation of the dipolar energy . instead of
that , we can use the following heuristic argument to obtain a reasonable approximation . the main error introduced by the swp approach
is in the exchange and anisotropy contributions to the energy . since the main contribution to the dipolar energy
is given by the interaction between domains , we can assume that the dipolar contribution of the wall is relatively independent of its shape .
hence , we can approximate it by eq.([edipyg ] ) .
furthermore , taking @xmath113 ( @xmath114 is a fitting parameter of order one to be fixed later ) in the limit @xmath115 ( @xmath116 ) , @xmath117 is very well approximated by@xcite @xmath118 assuming then @xmath119,\ ] ] we compare the energy obtained with the above equation with that obtained using the approximation of appendix [ fouriertangh ] for different values of the system parameters .
we verified that the error made by the approximation eq.([dipolarenergyhwp ] ) taking @xmath120 is always smaller than @xmath94 for @xmath121 .
we also observe that the best agreement with the mc results is obtained for @xmath120 . assuming then @xmath59 , the total energy per spin ( relative to the parallel magnetized state ) for the hwp
can then be approached by @xmath122 minimizing eq.([energyhwp ] ) with respect to the variational parameters @xmath0 and @xmath110 ( using eq.([gdelta - approx ] ) ) leads to : @xmath123,\ ] ] with @xmath124 in agreement with a derivation made by politi@xcite .
( color online ) magnetization profile in a saturated state ( @xmath84 and @xmath125 ) for @xmath126 and @xmath127 .
black symbols correspond to the mc simulations ( @xmath128 and @xmath129 ) .
the full red line corresponds to the yg approximation for the swp ( @xmath130 , @xmath131 and @xmath129 ) .
the black dashed line is a fit using a hyperbolic tangent function @xmath132 ( @xmath133 and @xmath134 ) . ]
( color online ) zero temperature phase diagram for large values of @xmath13 .
the shaded region corresponds to the canted state .
symbols correspond to mc simulations and lines to theoretical results .
the dashed line correspond to the crossover between sinusoidal and hyperbolic wall structure .
the lower line ( blue ) corresponds to eq.([etasrt ] ) .
the middle line ( red ) is obtained from the hwp minimum energy solution with @xmath111 .
the upper line ( black ) corresponds to eq.([etatanhyg ] ) . ] with the previous calculation we can also estimate the transition line between the saturated and the ising striped state . in the large @xmath0 limit the energy for a sip , i.e. for @xmath135 can be easily calculated from eq.([energiabm ] ) .
the fourier coefficients can be obtained as the @xmath136 limit of eq.([bm - yg ] ) : @xmath137 using eq.([dm ] ) the dipolar energy is then given by @xmath138 where @xmath139 , @xmath140 is the euler gamma constant and @xmath141 is the digamma function@xcite .
the energy per spin respect to the in
plane magnetized state is then given by @xmath142 minimizing eq.([energiaising ] ) with respect to @xmath0 leads to the equation @xmath143 , where @xmath144 , thus recovering the known result @xmath145 . comparing the energies , we find that the hwp has less energy than the ising state for any value of @xmath12 .
eq.([htanhyg ] ) shows that the stripe width variation in the saturated state is determined by the change in the wall width as the anisotropy increases .
hence , @xmath0 will change until the wall width reaches the atomic limit , i.e. for @xmath146 , where eq.([htanhyg ] ) recovers the ising behavior @xmath145 . imposing the condition @xmath146 to eq.([lwtanhyg ] )
we obtain the transition line between the saturated and the ising stripes states : @xmath147 which is also shown in fig.[pd2 ] , in complete agreement with the mc results .
( color online ) hwp equilibrium stripe width @xmath0 vs. @xmath12 for different values of @xmath13 .
full green lines correspond to the variational solution of eqs.([anitanh ] ) and ( [ eexchhwp2 ] ) using the approximation ( [ perfiltanh ] ) for the fourier coefficients in the dipolar energy .
the dotted lines correspond to the asymptotic approximation given by eqs.([htanhyg ] ) and ( [ lwtanhyg ] ) .
the dash - dotted line corresponds to the srt borderline given by eq.([etasrt ] ) with @xmath97 .
the dashed line corresponds to the borderline between saturated and ising stripes states . ] in fig.[hvseta - hwp ] we show the variation of stripe width @xmath0 vs. @xmath12 for different values of @xmath13 in the hwp , comparing the variational solution from eqs.([anitanh ] ) and ( [ eexchhwp2 ] ) using the approximation of appendix [ fouriertangh ] and the asymptotic approximation given by eqs.([htanhyg ] ) and ( [ lwtanhyg ] ) . in fig.[hvseta - delta10 ] we compare the equilibrium stripe width @xmath0 as a function of @xmath12 obtained within the different approximations used in this work for @xmath66 and with the mc simulations .
notice that the asymptotic approximation for the hwp given by eqs.([htanhyg ] ) and ( [ lwtanhyg ] ) shows a better agreement with the mc results than using the approximation ( [ perfiltanh ] ) for the fourier coefficients in the dipolar energy .
this is because we adjusted the fitting parameter @xmath114 to optimize the agreement with the mc results at low values of @xmath13 . from fig.[hvseta - hwp ] we see that the discrepancy between both approximations becomes negligible in the large @xmath13 limit .
( color online ) comparison of the equilibrium stripe width @xmath0 vs. @xmath12 obtained within the different methods for @xmath66 .
symbols correspond to mc simulations .
full black line corresponds to the asymptotic approximation for the hwp given by eqs.([htanhyg ] ) and ( [ lwtanhyg ] ) .
the red dashed line corresponds to the variational solution of eqs.([anitanh ] ) and ( [ eexchhwp2 ] ) using the approximation ( [ perfiltanh ] ) for the fourier coefficients in the dipolar energy .
the blue dotted line corresponds to the continuous swp .
the vertical dashed line corresponds to the transition between saturated and ising stripes states given by eq.([etatanhyg ] ) . ]
the main results of this work are summarized in figs.[pd1 ] and [ pd2 ] , which display the complete zero temperature phase diagram of the model defined by the hamiltonian ( [ hamiltoniano ] ) . working upon reasonable assumptions for the ground states , like perfectly straight modulations in one dimension and bloch domain walls
, we analyzed minimum energy configurations combining a variational analysis with monte carlo results .
we find four qualitatively different kinds of solutions : a planar ferromagnet for small anisotropies , and three types of perpendicular striped states : a canted state where the local magnetization has a finite in - plane component , a saturated state in which the in - plane component is restricted to the domain walls , and an ising stripe state with sharp walls for large anisotropies .
the canted and staturated states give valuable information on the behaviour of the stripe width as the anisotropy and exchange parameters change , a still open and debated question@xcite .
we find that stripe width variation is directly associated to the presence of finite width domain walls
. for large enough values of the anisotropy @xmath12 the ground state of the system is always an ising striped state , no matter the value of the exchange coupling @xmath13 . in those states
domain walls are sharp and the stripe width is completely independent of @xmath12 .
it grows exponentially with the exchange coupling . at the srt the system always passes through a canted state as the anisotropy increases , although the range of values of @xmath12 where the canted angle is different from zero narrows as @xmath13 increases .
the exchange to dipolar coupling ratio in fcc fe based ultrathin films can be roughly estimated to be @xmath148 ( considering a cubic bilayer of fe / cu(100 ) , where@xcite the exchange coupling @xmath149 , the lattice constant @xmath150 and@xcite @xmath151 ) . for @xmath148 the anisotropy interval for the canted phase
is approximately @xmath152 .
although narrow , this suggests that the canted phase should be detectable close enough of the srt , in systems like low temperature grown@xcite fe / cu(100 ) or fe / ni / cu films@xcite ( room temperature grown fe / cu(100 ) do not exhibit srt@xcite , suggesting a rather large value of the microscopic anisotropy ) . for @xmath153 stripe width variation appears always together with a varying canted angle .
close enough to the srt domain walls present a sinusoidal shape in agreement with yg results , but as the anisotropy and the exchange increase , the wall profile changes to a hyperbolic tangent shape , as expected from micromagnetic calculations , while the magnetization inside the domains becomes fully saturated . for @xmath154
the ground state is given by the saturated state , except very close to the srt .
a similar effect ( i.e. a crossover between a sinusoidal and a saturated magnetization profile ) has been observed in room temperature grown fcc fe / cu(100 ) ultrathin films , as the temperature decreases from @xmath155 , even though those systems do not present srt@xcite . in the saturated state ,
the stripe width increase with @xmath12 is directly related to the wall width decrease through the relation @xmath156 .
the wall width in turns is determined by the competition between exchange and anisotropy . once the anisotropy is large enough that the wall width reaches the atomic limit
, @xmath0 growth stops
. one may wonder whether a similar mechanism could be behind the stripe width variation with temperature .
besides its direct application to real systems , knowing the ground state of this system for arbitrary values of the exchange coupling is of fundamental importance to have a correct interpretation of monte carlo simulation results .
being one of the more powerful tools to analyze these kind of systems at the present ( specially at finite temperatures ) , it is basically limited by finite size restrictions , which implies relatively small values of @xmath13 ( the characteristic length @xmath0 of the problem grows exponentially with @xmath13 at low temperatures ) .
we thank n. saratz for advise about experimental results on ultrathin magnetic films .
this work was partially supported by grants from conicet , foncyt grant pict-2005 33305 , secyt - universidad nacional de crdoba ( argentina ) , cnpq and capes ( brazil ) , and ictp grant net-61 ( italy ) .
we briefly review in this appendix the derivation of the main results of yg approximation@xcite . the expression for the exchange and anisotropy energies per spin from eq.([hamiltoniano ] ) in terms of the one dimensional magnetization profile
@xmath28 is straightforward : @xmath157 ^ 2\ ] ] the dipolar energy per spin can be expressed as @xmath158 , where @xmath159 is the self - interaction energy ( i.e. , the sum over @xmath15 of the interaction energy between spins belonging to the same line at @xmath15 ) and @xmath160 is the interaction energy between all different pairs of lines .
the self interaction term is given by@xcite @xmath161 ^ 2 \label{es3}\ ] ] with @xmath162 where @xmath163 is the riemann zeta function .
the interaction term can be expressed as @xmath164 where the sum in the above expression is taken over all values of @xmath165 such that @xmath166 .
the interaction energy between two lines located at @xmath15 and @xmath167 is given by @xmath168 where @xmath169^{3/2}}\ ] ] @xmath170^{5/2}}\ ] ] in the limit @xmath171 the sums in eqs.([f1 ] ) and ( [ f2 ] ) can be evaluated using a continuum aproximation@xcite giving @xmath172 and @xmath173 . for @xmath174
the error in this approximation is smaller than @xmath175 .
for @xmath176 they can be evaluated numerically giving @xmath177 and @xmath178 .
then , eq.([eint2 ] ) can be written as @xmath179 where @xmath180 and @xmath181 .
finally @xmath182 \label{es4}\end{aligned}\ ] ] where @xmath183 $ ] is given by eq.([correction ] ) . putting all these terms together
we get eq.([energia1 ] ) . in the continuum limit @xmath60 and @xmath61 ( @xmath184 ) eq.([energiabm ] ) can be written as @xmath185= \delta - \delta \frac{1}{l } \sum_x cos\left ( \phi(x)-\phi(x+1)\right ) + m_0 ^ 2 \sum_{m=1,3,\ldots } b_m^2 d_m(h ) - \frac{\kappa\ , m_0 ^ 2}{2 } \sum_{m=1,3,\ldots } b_m^2 \label{energiabm -cont}\ ] ] where @xmath186 and the functions @xmath187 are given in eq.([dm ] ) . from eqs.([perfilyg ] ) the fourier cofficients @xmath188 in this limit are given by@xcite @xmath189
and@xcite @xmath190 with @xmath191 .
the dipolar energy term in eq.([energiabm -cont ] ) can be approached by@xcite @xmath192\ ] ] where we have used eq.([sumbmyg ] ) , the cuadratic term in eq.([dm ] ) has been neglected , @xmath89 and @xmath193 assuming a smooth profile inside the walls @xmath194 the exchange term in eq.([energiabm -cont ] ) can be approached by @xmath195 taking @xmath196 in the region inside a wall and replacing the summation in eq.([eexchyg ] ) by an integral we have that @xmath197 replacing eqs.([sumbmyg ] ) , ( [ edipyg ] ) and ( [ eexchyg2 ] ) into eq.([energiabm -cont ] ) we get @xmath198 where @xmath199 . minimizing eq.([energiayg ] ) respect to the variational parameters @xmath200 we get @xmath201 \ , cos^2\ , \theta \label{ygeq2}\end{aligned}\ ] ] @xmath202 sin\ , \theta\ , cos\ , \theta \label{ygeq3}\ ] ] notice that the fully saturated state @xmath71 is never a solution of the above equations , except in the limit @xmath203 ( or @xmath136 ) , which corresponds to @xmath204 . on the other hand ,
the planar ferromagnetic state @xmath84 ( @xmath205 ) , is always solution of the above equations . for @xmath206 ,
the variational equations reduce to eqs([ygeq2 - 2 ] ) .
close to the srt ( i.e. , the transition between a state with @xmath84 and one with @xmath207 ) we can assume @xmath208 and therefore @xmath209 . replacing into eqs.([ygeq1])-([ygeq3 ] ) they become @xmath210 \label{ygeq2'}\end{aligned}\ ] ] independent of @xmath211 , while eq.([ygeq3 ] ) becomes identically zero in the limit @xmath212 ( srt ) . replacing eq.([ygeq1 ] ) into eq.([ygeq2 ] ) we find @xmath213\ ] ] both @xmath117 and the expression between square brackets in eq.([ygeq4 ] ) are monotonously decreasing functions of@xcite @xmath214 . since the maximum allowed value is @xmath215 ( which corresponds to @xmath216 .
i.e. , pure sinusoidal profile ) , the minimum value of @xmath217 for which a domain solution exists corresponds to @xmath215 . using that@xcite @xmath218 and @xmath219 we have @xmath220 from eq.([ygeq1 ] )
this corresponds to @xmath221 or @xmath97 . replacing these values into eq.([ygeq3 ] )
we see that in the limit @xmath222 we have that @xmath223 . also from eq.([energiayg ] ) we see that in this limit the planar ferromagnetic and the domain solutions become degenerated .
hence , this point corresponds to the srt and the srt line in the @xmath224 space is given by eq.([etasrt ] ) .
it is easier to carry out this calculation by considering a profile whose wall starts at @xmath102 , i. e. @xmath225 where @xmath62 and @xmath226 .
the profile eq.([perfilyg ] ) is related to the previous one by @xmath227 .
the profile ( [ perfilygdiscret ] ) can be expanded as @xmath228 so the coefficients @xmath188 of the expansion ( [ fourier - stripes ] ) are given by @xmath229\ ] ] where @xmath230 $ ] stands for the real part of @xmath49 .
the coefficients @xmath231 are given by @xmath232 the summations in eq.([cm ] ) can be carry out explicitly obtaining @xmath233 where @xmath234 with @xmath235 and @xmath236 the dipolar energy can then be evaluated from eqs.([edipinf ] ) and ( [ dm ] ) .
the anisotropy energy can be easily calculated and gives @xmath237 the exchange and correction terms in eq.([energiabm ] ) can be expressed as @xmath238 @xmath239&= & \frac{2(c_2-c_1)}{h } \ , \sum_{x=0}^{w-1 } m^y(x ) m^y(x+1 ) \nonumber \\ & + & 2(c_2-c_1 ) ( 1-\delta)\ , ( 1-m_0 ^ 2)\end{aligned}\ ] ] the summations in the above equations involve a finite number of terms that can be computed explicitly .
in order to check the different striped profiles used to minimize the energy of the system , we implemented a simulated annealing protocol , based on metropolis dynamics .
the temperature was decreased down to zero at a constant rate @xmath240 , where the time is measured in monte carlo steps .
all the simulations are made starting from a planar ferromagnetic state with @xmath241 and @xmath242 .
every simulation is repeated @xmath243 times using a different sequence of random numbers , to check for possible trapping in local minima . since we are considering only periodic straight domains with bloch walls ,
the problem is basically one dimensional and we can restrict the search to a one dimensional pattern over the @xmath15 direction fixing @xmath25 and imposing periodic boundary conditions ( pbc ) in the @xmath16 direction .
we also use pbc in the @xmath15 direction . in other words , we simulated a a lattice with @xmath244 with @xmath245 and pbc , which are implemented by means of the ewald sums technique .
for every set of values of @xmath14 we check the results for different values of @xmath246 in order to avoid artificial frustration .
we also performed some comparisons with mc results in a square @xmath247 lattice ; the results were indistinguishable .
this ansatz allows us to obtain mc results for values of @xmath13 up to @xmath126 ( for which the maximum equilibrium value is @xmath248 ) .
the function @xmath249 is very well approximated by @xmath250
|
stripe width variation in ultrathin magnetic films is a well known phenomenon still not well understood .
we analyze this problem considering a 2d heisenberg model with ferromagnetic exchange interactions , dipolar interactions and perpendicular anisotropy , relevant e.g. in fe / cu(001 ) films . by extending a classic result of yafet & gyorgy ( yg ) and
using monte carlo simulations we calculate the complete zero temperature phase diagram of the model . through this calculation
we analyze the correlation between domain walls structure and stripe width variation , as the perpendicular anisotropy changes .
in particular , we found evidences that the recently detected canted state becomes the ground state of the system close to the spin reorientation transition ( srt ) for any value of the exchange to dipolar couplings ratio .
far away of the srt the canted ground state is replaced by a saturated stripes state , in which in plane magnetization components are only present inside the walls .
we find that the domain wall structure strongly depends on the perpendicular anisotropy : close to srt it is well described by yg approximation , but a strong departure is observed in the large anisotropy limit .
moreover , we show that stripe width variation is directly related to domain wall width variation with the anisotropy .
|
under ambient conditions , amorphous silica ( sio@xmath2 ) forms a disordered tetrahedral network .
tetrahedral units with silicon atoms in the centre are connected to each other via the oxygen atoms at the four corners of each tetrahedron . under compression ,
the network structure of sio@xmath2 may change significantly . from experiments on silica glasses
@xcite , it is known that the sio coordination changes for pressures of the order of 1520gpa . then , the network structure is dominated by five- and six - fold coordinated sio units . albeit of great interest for geoscience ,
experimental studies of pure silica melts under ( high ) pressure are very rare @xcite .
this is due to the high temperatures that are required to study silica in its liquid state .
hence , computer simulations are an important tool to elucidate structural and dynamic properties of silica melts at high pressure . in many of the recent molecular dynamics ( md ) computer simulations , the pair potential proposed by van beest
, kramer and van santen , the so - called bks potential @xcite , has been used to model the interactions between the si and o ions .
et al _ @xcite showed that in bks silica the oxygen coordination around silicon atoms changes from 4 to 5 at intermediate pressures ( 10 - 15gpa ) and reaches 6 at high pressures .
subsequent studies of bks silica @xcite have indicated that transport processes first become faster with increasing pressure , in agreement with experimental studies on silicate melts @xcite . in ref .
@xcite , this was interpreted as a signature of a strong - to - fragile transition , according to angell s classification of glassforming liquids @xcite .
note that the anomalous transport in silica under pressure was already seen in a pioneering md simulation by angell _
et al _ @xcite using a different model potential .
further studies on bks silica @xcite addressed the question to what extent the properties of silica are typical for a strong glassformer , sharing its properties with other network - forming liquids such as water . in ref .
@xcite , the possibility of a phase separation between a low and high density liquid phase was discussed and a rough estimate of the free energy as a function of volume and temperature was used to locate such a transition for bks silica , finding a critical point at a density of about 2.9g/@xmath1 and a temperature of about 2000k . but recent works have called in question that simulations with simple pair potentials can give reliable results for pressure - induced phenomena in silica . trave _ et al _ @xcite claimed that empirical force fields such as the bks model `` are not sufficiently accurate to reproduce pressure - induced phenomena such as changes in coordination and increasingly strained geometries '' .
these authors used `` first - principles '' car - parrinello molecular dynamics ( cpmd ) simulations to investigate liquid silica for pressures up to about 30gpa at a constant temperature of @xmath3k . in a cpmd ,
the electronic degrees of freedom are taken into account in the framework of a density functional theory @xcite and thus it is not relying on the use of effective interatomic potentials .
whereas the cpmd calculation for silica shows good agreement with respect to the experimental equation of state , the bks model underestimates the experimental specific volume systematically by about 15% . based on cpmd calculations , tangney and scandolo ( ts ) @xcite have developed a new effective potential that nicely reproduces the experimental equation of state of amorphous silica as well as some of the properties of crystalline phases @xcite .
the ts model is a fluctuating dipole moment potential , accounting for the polarizability of the oxygen atoms .
a disadvantage of the ts model is its low efficiency . due to the expensive self - consistent computations of the dipole moments on the oxygen atoms in every time step ,
up to two orders of magnitude more computer time is needed in md simulations with the ts model than for comparable simulations with a simple pair potential such as the bks model @xcite .
the conjecture of trave _ et al _ about the validity of simple effective interaction models for silica has been recently challenged by carr _
et al _ @xcite who developed a new pair potential for silica , the so - called chik potential .
the chik potential was also parametrized from cpmd calculations and was shown to be superior to the bks model with respect to various properties of amorphous silica ( in particular the density at low pressures ) @xcite . in this work , we show that it also reproduces very accurately the experimental equation of state , various structural properties and the anomalous diffusion dynamics of silica under pressure .
this proves the transferability of the chik potential to the description of amorphous silica under high pressure .
thus , the chik model can be used to readdress the questions about the possibility of a liquid - liquid phase separation and the mechanisms of anomalous transport phenomena at high pressures .
at first glance , it might be surprising that the directional covalent bonding in silica can be described by an interaction model that depends only on the magnitude of the distance between atom pairs .
in fact , in a one - component system such as silicon one has to include at least three - body interactions to reproduce a tetrahedral network structure .
however , in a binary system such as silica there is the competition between three different pair potentials . by tuning the parameters for the repulsive oo interactions and the attractive sio interactions appropriately
, one can generate a network structure with si - o tetrahedra as the preferred structural units .
, scaledwidth=65.0% ] apart from a rather unreliable description of the equation state , the widely used bks potential @xcite has been proven successful to describe various static and dynamic properties of amorphous silica quite reliably ( see , e.g. , ref .
the parametrization of the bks model was based on hartree - fock calculations of a single sio@xmath4 tetrahedron , charge - saturated by four hydrogen atoms @xcite .
in contrast to that , the development of the chik potential was based on a cpmd simulation of a bulk system of 114 particles .
a structural fitting procedure was used for the parametrization .
the idea was to match the partial pair correlation functions , as obtained with the new effective potential , with those obtained from cpmd .
as the bks potential , the chik model consists of a long - ranged coulomb part and a short - ranged buckingham potential , @xmath5 where @xmath6 is the distance between an ion of type @xmath7 and an ion of type @xmath8 ( @xmath9 ) , @xmath10 is the elementary charge , and @xmath11 and @xmath12 are effective partial charges .
the values of the parameters @xmath13 are @xmath14ev , @xmath15@xmath16 , @xmath17ev / @xmath18 , @xmath19ev , @xmath20@xmath16 , @xmath21ev / @xmath18 , @xmath22ev , @xmath23@xmath16 , and @xmath24ev / @xmath18 .
more details on the potential can be found in ref .
@xcite .
md simulations have been performed for systems of @xmath25 particles at the densities @xmath26 , 2.4 , 2.5 , 2.7 , 2.9 , 3.0 , 3.1 , 3.3 , 3.5 , 3.6 , 3.7 , 3.9 , and 4.2g/@xmath1 in the temperature range 6100k@xmath27 2100k .
the equations of motion were integrated by the velocity form of the verlet algorithm @xcite using a time step of 1.6fs .
the long - range coulomb forces and energies were computed by the ewald summation technique @xcite . at each state , the system was equilibrated in the @xmath28 ensemble during 0.2ns to 25ns .
temperature was kept constant by coupling the system periodically to a stochastic heat bath .
after the equilibration , structural and dynamic quantities were obtained from production runs in the microcanonical ensemble .
figure [ fig1 ] displays the equation of state , @xmath29 , along different isotherms . for pressures up to about 10gpa , the pressure is almost independent of temperature in the range 4300k@xmath27 2100k .
in this pressure range , the simulation is in very good agreement with experiment @xcite .
this demonstrates the transferability of the chik model : this model has been parametrized from a single cpmd run at @xmath30k and @xmath31g/@xmath1 , without considering states at high pressure .
thus , at least as far as the equation of state is concerned , the chik potential provides a reliable description and so it appears to be appropriate for the investigation of silica under high pressure . , for different densities at the temperature @xmath32k .
the arrow indicates the cut - off @xmath33g/@xmath1 , used to identify sio bonds .
the inset shows an enlargement of the data .
the two vertical dashed lines at 1.618 and 1.647 indicate the location of the maxima for the densities @xmath26g/@xmath1 and @xmath34g/@xmath1 , respectively .
[ fig2 ] , scaledwidth=65.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] .
a ) @xmath35g/@xmath1 , b ) @xmath36g/@xmath1 , c ) @xmath37g/@xmath1 , d ) @xmath38g/@xmath1 , e ) @xmath39g/@xmath1 , f ) @xmath34g/@xmath1 .
[ fig3 ] , title="fig:",scaledwidth=40.0% ] k , 2230k , and 2100k .
[ fig4 ] , scaledwidth=65.0% ] ( @xmath9 ) for different densities at the temperature @xmath32k .
a ) @xmath40 , b ) @xmath41 , c ) @xmath42 . in panel d ) , @xmath42 for @xmath43g/@xmath1 at three different temperatures is shown , indicating that the system has crystallized at @xmath32k .
[ fig5 ] , title="fig:",scaledwidth=40.0% ] ( @xmath9 ) for different densities at the temperature @xmath32k .
a ) @xmath40 , b ) @xmath41 , c ) @xmath42 . in panel d ) , @xmath42 for @xmath43g/@xmath1 at three different temperatures is shown , indicating that the system has crystallized at @xmath32k .
[ fig5 ] , title="fig:",scaledwidth=40.0% ] ( @xmath9 ) for different densities at the temperature @xmath32k .
a ) @xmath40 , b ) @xmath41 , c ) @xmath42 . in panel d
) , @xmath42 for @xmath43g/@xmath1 at three different temperatures is shown , indicating that the system has crystallized at @xmath32k .
[ fig5 ] , title="fig:",scaledwidth=40.0% ] ( @xmath9 ) for different densities at the temperature @xmath32k .
a ) @xmath40 , b ) @xmath41 , c ) @xmath42 . in panel d ) , @xmath42 for @xmath43g/@xmath1 at three different temperatures is shown , indicating that the system has crystallized at @xmath32k .
[ fig5 ] , title="fig:",scaledwidth=40.0% ]
to analyze the change of the network structure of amorphous silica under compression , we consider the temperature dependence of the si - o as well as the o - si coordination . to this end ,
coordination numbers @xmath44 and @xmath45 are defined as the number of o ( si ) atoms surrounding a si ( o ) atom within a distance @xmath46 . for the cut - off radius we use the value @xmath33 that corresponds to the location of the first minimum of the partial pair correlation function for the si - o correlations , @xmath47 @xcite . as fig .
[ fig2 ] shows , the location of this minimum is indeed essentially independent of density .
however , one can infer from the inset of fig .
[ fig2 ] that the length of the sio bond changes slightly from about 1.62 at 2.3g/@xmath1 to about 1.65 at 3.9g/@xmath1 .
this finding is in very good agreement with results from x - ray diffraction experiments @xcite .
moreover , the periodicity of the peaks in @xmath47 changes with density . as we shall see below ,
this is due to the loss of tetrahedral order at high densities , reflected , e.g. , in the fourier transform of @xmath48 , the partial structure factor @xmath41 , by a shift of the first sharp diffraction peak to higher wavenumbers @xmath0 .
figure [ fig3 ] shows the probability distributions @xmath49 at different densities for all the relevant coordination numbers @xmath44 and @xmath45 as a function of temperature .
for the three densities @xmath35 , 2.7 , and 3.1g/@xmath1 , most of the si atoms exhibit the tetrahedral four - fold coordination by o atoms and most of the o atoms serve as bridging atoms between the tetrahedra ( @xmath50 ) . in this case , the local defect structures that correspond to @xmath51 as well as @xmath52 become less frequent with decreasing temperature such that one expects a very low probability of these structures at low temperatures .
a different behavior is seen at the density @xmath38g/@xmath1 .
now , @xmath53 has a higher probability than the tetrahedral coordination @xmath54 and one observes a relatively high probability of @xmath52 and a significant number of six - fold coordinated si atoms .
one can hardly extrapolate the curves for 3.5g/@xmath1 to lower temperatures since at this density all the coordination numbers show a rather weak temperature dependence in the considered temperature range .
the same trend is also seen at the two high densities , @xmath39g/@xmath1 and @xmath34g/@xmath1 . whereas the number of five - fold coordinated si atoms keeps approximately constant around 50% , the coordination number @xmath55 further increases , whereas the tetrahedral coordination @xmath54 becomes less and less important with increasing density . at @xmath34g/@xmath1 , there are more oxygen atoms that are three - fold coordinated by si atoms than bridging oxygens . the sio coordination as a function of pressure is shown in fig .
[ fig4 ] for @xmath56k , 2230k and 3580k .
although the qualitative behaviour of @xmath57 is similar at the different temperatures , there are significant quantitative differences between the coordination numbers at @xmath58k and those at the two lower temperatures . the maximum of @xmath59 shifts from about 30gpa at 3580k to about 25gpa at the two lower temperatures .
moreover , at low temperatures , @xmath60 decays more rapidly , accompanied by a faster increase of @xmath61 .
note that in recent simulation studies using both ab initio techniques and classical md simulation @xcite , the pressure dependence of the sio coordination has been only discussed for temperatures around 3500k .
however , fig .
[ fig4 ] demonstrates that at least in the temperature range between 2000 to 4000k the temperature dependence of @xmath57 has to be taken into account to allow for a quantitative comparison with experimental data , done at much lower temperatures .
k and b ) @xmath56k .
[ fig6 ] , title="fig:",scaledwidth=60.0% ] k and b ) @xmath56k .
[ fig6 ] , title="fig:",scaledwidth=60.0% ] the inspection of the sio and osi coordination number distributions indicates that significant changes occur in the network structure for pressures above , say 10gpa .
this finding is similar to what has been observed experimentally for silica glasses . at pressures above about 20gpa
a network with five- and six - fold coordinated si atoms is formed , similar to what has been recently found in other simulation studies @xcite .
it remains an open question whether in silica a first - order phase transition from a low density liquid to a high density liquid exists .
our data in figs .
[ fig3 ] and [ fig4 ] is also consistent with a gradual crossover from a low density phase ( with @xmath54 ) to a high density phase ( with @xmath55 ) , occurring at low temperatures . to investigate the changes in the network structure on intermediate length scales ,
we now consider the partial static structure factors , defined as @xcite @xmath62 with @xmath63 the number of atoms of type @xmath7 , @xmath64 the wavevector , and @xmath65 the distance vector between particle @xmath66 and particle @xmath67 .
[ fig5 ] shows the three @xmath68 at the temperature @xmath32k for different densities . for the density @xmath69g/@xmath1 the first peak [
also called first sharp diffraction peak ( fsdp ) ] is located at a wavenumber of about 1.6 - 1.7@xmath16 .
this peak is due to the tetrahedral order present in `` low - density '' silica .
its location is related to the length scale of two connected tetrahedra @xcite . with increasing density , the first peak shifts to higher values of @xmath0 and its amplitude decreases significantly .
the behavior of the fsdp reflects the change of the structure from an open tetrahedral network at low densities to a more densely packed network at high densities , provided by sio@xmath70 and sio@xmath71 units as well as oxygen atoms that are three - fold coordinated by silicon atoms .
note that the partial structure factors @xmath68 are the required input for calculations in the framework of the mode - coupling theory of the glass transition @xcite , mct , from which the dynamics of the melt can be predicted . a mct calculation based on the @xmath68 of our simulation
is presented elsewhere in this issue @xcite .
figure [ fig5]d shows @xmath42 for the density @xmath43g/@xmath1 at three different temperatures . at this high density ,
the system crystallizes at a temperature of @xmath32k which is reflected by the occurrence of bragg peaks in @xmath42 . at a temperature of @xmath72k
the system remains liquid ; the structure factor in this case reminds one to that of a simple liquid , showing no sign of the fsdp feature .
, scaledwidth=60.0% ] , scaledwidth=60.0% ] the structural changes in the network with increasing pressure are accompanied by an anomalous behavior of relaxation processes .
this can be inferred from the mean squared displacements of a tagged particle of type @xmath73 , @xmath74 . in fig .
[ fig6 ] only @xmath75 is shown , since the mean squared displacement for oxygen exhibits a similar behaviour . at @xmath32k ,
several time regimes can be clearly distinguished at the low density @xmath76g/@xmath1 , consisting of a ballistic regime at very short times ( @xmath77 ) , an intermediate regime with a maximum 0.2ps , a plateau - like region and finally a linear regime , @xmath78 , indicating diffusive behaviour .
note that the maximum at 0.2ps is both related to the so - called boson peak and to finite size effects present in the dynamics of silica ( for details see refs .
@xcite ) . that this feature is not seen at high density
might be related to the changes occurring in the network structure . from @xmath76g/@xmath1 to @xmath38g/@xmath1 , an acceleration of the diffusion dynamics can be inferred from fig .
[ fig6]a , followed by the opposite behaviour for a further decrease of the density .
the same behaviour as for @xmath32k is also seen for the lower temperature , @xmath56k , although the effect of a maximal speed of diffusion is now more pronounced ( fig .
[ fig6]b ) .
one can also infer from fig .
[ fig6]b that the particles become more localized with increasing density , indicated by a decrease of the plateau s height with increasing density .
the self - diffusion constants can be easily determined from the mean squared displacements using the einstein relation , @xmath79 . in fig .
[ fig7 ] , the self - diffusion constant for silicon , @xmath80 , along different isotherms is displayed as a function of density and pressure .
note that @xmath81 , albeit slightly larger , exhibits the same behaviour . along the different isotherms
a maximum is found around about 3.5g/@xmath1 , corresponding to a pressure of about 20gpa ( see fig . [
thereby , the diffusivity maximum becomes more pronounced with decreasing temperature .
the behaviour of @xmath80 is closely related to that of the coordination numbers . above about @xmath82gpa , the location of the diffusivity maximum , the number of five - fold coordinated si atoms roughly shows a saturation , while the number of six - fold coordinated si atoms further increases with increasing pressure .
it seems that a sio network with many five - fold coordinated si atoms tends to have a faster diffusion dynamics than one with many six - fold coordinated si atoms .
therefore , for pressures up to about 20gpa the network dynamics is dominated by an increasing number of five - fold coordinated si atoms , leading to an acceleration of diffusion . however , at pressures above about 20gpa , the network dynamics is dominated by an increasing number of sio@xmath71 units , leading to a decrease of the diffusion constants with pressure .
thus , the diffusivity maximum is due to an interplay between five- and six - fold coordinated si atoms .
results of an extensive md simulation have been presented to study structural and dynamic properties of liquid silica under ( high ) pressure . as an interaction potential a simple pair potential , the so - called chik potential ,
was used that leads to a reliable description of the equation of state .
we have shown that the change of the network structure with increasing pressure is related to an anomalous diffusion dynamics , characterized by a diffusivity maximum around 20gpa . elsewhere in this issue
@xcite , a calculation in the framework of mode coupling theory based on the structural input of our simulation is presented that explicitly shows that the changes in the network structure are intimately related to the anomalous transport in amorphous silica .
it is still an open question whether , at low temperatures , there is a first - order phase transition from a low - density to a high - density transition in silica .
this question will be addressed in future studies with the chik potential .
99 grimsditch m 1984 _ phys . rev . lett . _ * 52 * 2379 hemley r j , mao h k , bell p m and mysen b o 1986 _ phys .
lett . _ * 57 * 747 williams q and jeanloz r 1988 _ science _ * 239 * 902 meade c , hemley r j and mao h k 1992 _ phys . rev .
lett . _ * 69 * 1387 zha c , hemley r j , mao h , duffy t s and meade c 1994 _ phys . rev .
b _ * 50 * 13105 brazhkin v v and lyapin a g 2003 _ j. phys . : condens . matter _ * 15 * 6059 lin j - f , fukui h , prendergast d , okuchi t , cai y q , hiraoka n , yoo c - s , trave a , eng p , hu m y and chow p , 2007 _ phys .
b _ * 75 * 012201 van beest b w h , kramer g j and van santen r a 1990 _ phys .
lett . _ * 64 * 1955 tse j s , klug d d and le page y 1992 _ phys .
b _ * 46 * 5933 barrat
j - l , badro j and gillet ph 1997 _ j. comp .
_ * 20 * 17 hemley r j , badro j and teter d m 2000 in
_ physics meets mineralogy - condensed matter physics in geosciences _ ed aoki h , syono y and hemley r j ( cambridge : cambridge university press ) p 173 horbach j , kob w , binder k and angell c a 1996 _ phys .
e _ * 54 * r5897 kushiro i 1978 _ earth planet . sc .
* 41 * 87 rubie d c , ross c r , carroll m r and elphick s c 1993 _ am . mineral . _ * 78 * 574 poe b t , mcmillan p f , rubie d c , chakraborty s , yarger j and diefenbacher j 1997 _ science _ * 276 * 1245 tinker d , lesher c e and hutcheon i d 2003 _ geochim .
. acta _ * 67 * 133 angell c a 1985 in _ relaxation in complex systems _ ed ngai k and wright g b ( springfield : us dept . of commerce ) binder k and kob w 2005 _ glassy materials and disordered solids : an introduction to their statistical mechanics _
( singapure : world scientific ) angell c a , cheeseman p a and tammadon s 1992 _ science _ * 218 * 885 saika - voivod i , sciortino f and poole p h 2000 _ phys . rev
. e _ * 63 * 011202 scott shell m , debenedetti p g and panagiotopoulos a z 2002 _ phys .
e _ * 66 * 011202 sharma r , mudi a and chakravarty c , 2006 _ j. chem .
_ * 125 * 044705 le grand a , dreyfus c , bousquet c and pick r m , 2007 _ phys .
e _ * 75 * 061203 kohn w 1996 in _ monte carlo and molecular dynamics of condensed matter systems _ ed binder k and ciccotti ( bologna : societ italiana di fisica ) p 561 ; car r ibid .
p 601 trave a , tangney p , scandolo s , pasquarello a and car r 2002 _ phys .
* 89 * 245504 tangney p and scandolo s 2002 _ j. chem .
phys . _ * 117 * 8898 herzbach d , binder k and mser m h 2005 _ j. chem . phys . _ * 123 * 124711 carr a , horbach j , ispas s and kob w 2008 _ europhys
. lett . _ * 82 * 17001 vollmayr k , kob w and binder k 1996 _ phys . rev .
b _ * 54 * 15808 horbach j and kob w 1999 _ phys . rev . b _ * 60 * 3169 horbach j , kob w and binder k 2001 _ eur
j. b _ * 19 * 531 allen m and tildesley d 1987 _ compute simulation of liquids _
( oxford : oxford university press ) tsiok o b , brazhkin v v , lyapin a g and khvostantsev l g 1998 _ phys .
lett . _ * 80 * 999 gtze w 1999 _ j. phys . : condens .
11 * a1 voigtmann th and horbach j 2008 _ j. phys .
: condens .
matter _ this issue
|
the structural and dynamic properties of silica melts under high pressure are studied using molecular dynamics ( md ) computer simulation .
the interactions between the ions are modeled by a pairwise - additive potential , the so - called chik potential , that has been recently proposed by carr _
et al_. the experimental equation of state is well - reproduced by the chik model . with increasing pressure ( density ) ,
the structure changes from a tetrahedral network to a network containing a high number of five- and six - fold si - o coordination . in the partial static structure factors , this change of the structure with increasing density
is reflected by a shift of the first sharp diffraction peak towards higher wavenumbers @xmath0 , eventually merging with the main peak at densities around 4.2g/@xmath1 .
the self - diffusion constants as a function of pressure show the experimentally - known maximum , occurring around a pressure of about 20gpa .
|
hadron interactions are strong and , in principle , should be described by quantum chromodynamics ( qcd ) .
however , experimental data show that their main features originate from the non - perturbative sector of qcd .
only comparatively rare processes with large transferred momenta can be treated theoretically rather successfully by perturbative methods due to the well - known property of the asymptotical freedom of qcd .
hence , in the absence of methods for a rigorous solution of qcd equations , our understanding of the dynamics of the main bulk of strong interactions is severely limited by model building or some rare rigorous relations .
in fact , our approach to high - energy hadronic processes at present is at best still in its infancy .
as has been learned from experiment , strong interactions of colliding high - energy particles give rise to inelastic and elastic processes . some new particles ( mostly pions )
are produced in inelastic processes , which are the most probable ones , comprising 75@xmath0 to 80@xmath0 of all processes at high energies .
most created particles have comparatively small transverse momenta . at the same time , in 25@xmath0 to 20@xmath0 of events , the colliding particles do not change their nature and scatter elastically , declining at some angle from their initial trajectories . the only information about this process available from experiment is obtained by the measurement of the differential cross section ( proportional to the probability ) of elastic scattering at some angle at a given energy . in a very tiny range of extremely small angles ,
the charged particles scatter due to electromagnetic forces .
but the dominant process of elastic scattering due to strong interactions proceeds at somewhat larger angles in the so - called diffraction cone .
the differential cross sections are heavily weighted toward small transferred momenta exhibiting a huge peak .
the scattering angle is still rather small there and becomes smaller and smaller as the energy increases .
the probability of scattering at a given angle in this region decreases steeply , similarly to a gaussian exponential
. noticeably less than one percent of particles are elastically scattered to larger angles outside this diffraction cone .
the gaussian behavior is replaced there by a simple exponential one with some shoulders and ( or ) dips . at ever larger angles ( or transferred momenta ) , a power - like decrease has been observed . at angles close to @xmath2
, some additional flattening is seen .
the elastic cross section ( the integral of the differential distribution over angles or transverse momentum ) depends on the energy of the colliding partners . at high energies
, it shows a steady tendency to become larger with an increase in energy .
we note that the inelastic cross section also increases , such that their sum ( the total cross section ) increases as well .
the process of elastic scattering of hadrons has been studied experimentally in a wide energy region with different initial particles . at high energies of colliding partners ,
the most detailed results are available for the scattering of protons ( @xmath3 ) and antiprotons ( @xmath4 ) on protons .
we mainly discuss these data , sometimes referring to other colliding partners of protons such as pions and kaons . some surprises in the behavior of the differential cross sections appeared in the 1960s when the very first experimental data on elastic @xmath3 and @xmath5 scattering were obtained at energies between 6.8 and 19.2 gev in the laboratory system @xcite
( the total energy in the center - of - mass system ( cms ) is only @xmath6 4 - 6 gev ! ) .
the diffraction cone behavior changed at larger transferred momenta @xmath7 to a slower @xmath8-dependence .
somewhat later , the energy range was extended to 50 gev @xcite . with the advent of new accelerators , the data for @xmath3 scattering at energies @xmath9 19 , 20 , 23 , 28 , 31 , 45 , 53
, 62 gev were published @xcite and the data for @xmath4 at 31 , 53 , 62 , 546 , 630 , 1800 , 1980 gev @xcite appeared .
the early results are reviewed in refs @xcite .
the compilation of the data can be found in @xcite .
only recently , the results of the totem collaboration at the lhc on elastic @xmath3 scattering processes at @xmath10 = 7 tev were published @xcite . surely , these results asked for their understanding and theoretical interpretation .
the most important task is to acquire some knowledge about the internal structure of colliding particles by deciphering the information supplied by experimental data about the dependence on energy and transferred momentum .
the transferred momentum is directly related to the size and the structure of those regions inside the hadron that participate in the interaction .
many phenomenological models have been proposed .
most of them aspire to be `` a phenomenology of everything''@xmath11 related to elastic scattering of hadrons in a wide energy range .
doing so in the absence of applicable laws and methods of the fundamental theory , they have to use a large number of adjustable parameters .
the free parameters have been determined by fitting the model results to the available experimental data . even then
, their predictions often fail when a new energy domain becomes available . and
`` the verse''@xmath11 does not grow anymore !
( if not recultivated . )
independent of their success and failure , we are sure that , `` in the long run , the physical picture may be expected to be much more important than most of the detailed computations '' @xcite . in what follows , we mention and discuss many of them . the scattering of charged particles at extremely small angles is completely dominated by the coulomb amplitude .
the absolute value of the born amplitude is well known .
the phase of the coulomb amplitude varies depending on the model chosen .
however , this variation is rather mild in the considered tiny region of extremely small angles . the interference of the coulomb amplitude with the strong - interaction ( nuclear ) amplitude in the transition region where they are almost equal has been used for the experimental determination of the ratio of the real to imaginary parts of the latter .
this interference also depends on the chosen form of the nuclear amplitude .
theoretically , this ratio can be estimated with the help of dispersion relations .
we briefly discuss this problem and show how the obtained results influence our analysis of scattering at somewhat larger angles .
the most numerous group of models deals with phenomenological attempts to describe the main bulk of elastic scattering at small angles in the diffraction cone . in general , they are based on some geometrical models of particle substructure , with peripheral regions playing the decisive role . the approach using the reggeon ( pomeron ) exchanges is the most popular among them .
the approximately gaussian ( in angles ) shape of the experimentally measured differential cross section in this region has been fitted just in this way .
in addition to it , the simplest classical expressions for diffractive processes and results on the electromagnetic form factors are also used . however , the bold extension of the obtained results to larger angles is usually not very successful , even though some new parameters are introduced .
particles scattered at larger angles give insight into the deeper internal regions of particle structure .
the multiple iteration ( rescattering ) of diffractive processes may explain the region of angles that are somewhat larger than the diffractive ones .
without any additional model building , it can be described as a consequence of the unitarity condition .
the only necessary input is the experimentally known energy behavior of the diffraction cone slope and the total cross section .
it predicts the observed exponential fall - off with angles and damped oscillations imposed on it , which , depending on their amplitudes , lead to shoulders or dips of the differential cross sections . at somewhat larger angles ,
the elastic processes may be considered to be dominated by the innermost constituents of the colliding particles .
the perturbative qcd approach to hard parton scattering convoluted with some results on the parton structure of colliding particles is then used to describe experimental data .
this approach predicts the power - like angular dependence of the differential cross sections .
it has been seen in experiment .
the dimensional ( or quark ) counting of the number of participating partons has been successful .
the convolution with the internal structure of particles implies some coherence in the behavior of its constituents : all of them should coherently turn at the same angle .
a particle should not be destroyed during the collision , and its internal wave function must be left intact . therefore we can call such processes coherent large - angle scattering . at angles close to @xmath2 ,
the effects of symmetrization of the corresponding amplitudes may become important and lead to some flattening of the differential distribution .
there are no strict definitions of the lower and upper bounds of these regions .
the diffraction peak shrinks with energy , such that the exponential fall - off with squared transferred momenta @xmath8 terminates at ever smaller values .
correspondingly , the dip after it shifts to smaller values of @xmath7 as does the @xmath12-exponential . at low energies , this regime approximately occupies the interval between 0.8 and 2 gev@xmath13 , while in the lhc , it has moved to 0.4 - 1.5 gev@xmath13 . according to the qcd prejudice
, the scale for parton scattering should be set above 1 gev@xmath13 .
this is actually observed with a power - like decrease starting somewhere around @xmath141.5 - 2 gev@xmath13 at the lhc .
hence , we can speak , at least , about five subregions of elastic scattering .
we mainly discuss three of them : the diffraction cone , the orear regime and coherent hard parton scattering .
the diffraction cone is well known to us from semiclassical effects .
the regions beyond it became noticeable only at energies of colliding particles above several gev , where processes of scattering at sufficiently large angles or transferred momenta are observable .
they persist up to the present lhc energy of 7 tev . who ordered them and whether they will they survive at ever higher energies are also the questions to be discussed in this review .
its structure of this paper is as follows .
the main relations between different characteristics of elastic scattering are presented in section 2 .
then , in section 3 , their global dependences on energy and transferred momenta are discussed , together with our attempts to understand their implications within the simplest approaches .
a more detailed analysis of experimental data in the framework of different theoretical ideas and approximations is the content of section 4 .
finally , the general picture is briefly discussed in section 5 .
we do not consider the scattering of polarized particles , and the spin structure of the amplitude is ignored .
as discussed above , the measurement of the differential cross section is the only source of experimental information about a process .
hence , the main characteristics of hadron interactions directly related to the elastic scattering amplitude , such as the total cross section , the elastic scattering cross section , the ratio of the real to the imaginary part of the amplitude , and the slope of the diffraction cone , are obtained .
the first two are functions of the total energy only , while the others depend on two variables : the total energy and the transferred momentum ( or the scattering angle ) . the dimensionless elastic scattering amplitude @xmath15 defines the differential cross section as @xmath16 where the ratio of the real to imaginary parts of the amplitude is defined : @xmath17 in what follows
, we consider very high energy processes .
therefore , the masses of the colliding particles can be neglected , and we use the expression @xmath18 , where @xmath19 and @xmath20 are the energy and the momentum in the center - of - mass system .
the four - momentum transfer squared is @xmath21 with @xmath22 denoting the scattering angle in the center - of - mass system and @xmath23 being the transverse momentum .
the elastic scattering cross section is given by the integral of the differential cross section ( [ dsel ] ) over all transferred momenta : @xmath24 the total cross section @xmath25 is related by the optical theorem to the imaginary part of the forward scattering amplitude as @xmath26 elastically scattered hadrons escape from the interaction region declining mostly at quite small angles within the so - called diffraction cone is described below . ] .
therefore the main focus has been on this region .
as known from experiment , the diffraction peak has a gaussian shape in the scattering angles or exponentially decreases as a function of the transferred momentum squared : @xmath27 in view of relations ( [ sel ] ) , ( [ sigt ] ) , ( [ diff ] ) , any successful theoretical description of the differential distribution must also work in fitting the energy dependences of the total and elastic cross sections .
the diffraction cone slope @xmath28 is given by @xmath29 .
\label{bst}\ ] ] actually , the slope @xmath28 depends slightly on @xmath8 at the given energy @xmath30 , e.g. , at the lhc , its value changes by about 10@xmath31 within the cone for @xmath32 gev@xmath13 .
we neglect this in the first approximation . the normalization factor in eq .
( [ diff ] ) is @xmath33 where @xmath34 is defined as the value of the ratio of the real and imaginary parts of the amplitude in the forward direction at @xmath35 = 0 .
( [ diff0 ] ) follows from formula ( [ dsel ] ) and optical theorem ( [ sigt ] ) at @xmath36 .
according to the dispersion relations , which connect the real and imaginary parts of the amplitude , and optical theorem eq .
( [ sigt ] ) , the value @xmath34 can be expressed as an integral of the total cross section over the whole energy range . in practice
, @xmath34 is mainly sensitive to the local derivative of the total cross section . in the first approximation ,
the result of the dispersion relation can then be written in the form @xcite @xmath37\sigma _
t= \frac { 1}{\sigma _ t}\left [ \frac { \pi } { 2}\frac { d}{d\ln s } + \frac { 1}{3}\left ( \frac { \pi } { 2}\right ) ^3\frac { d^3}{d\ln s^3 } + ... \right ] \sigma
_ t. \label{rhodi}\ ] ] it follows that at high energies @xmath38 is mainly determined by the derivative of the logarithm of the total cross section with respect to the logarithm of energy . the bold extension of the first term in this series to nonzero transferred momenta would look like @xmath39 . \label{rhodit}\ ] ] if we neglect the high-@xmath7 tail of the differential cross section , which is several orders of magnitude lower than the optical point , and integrate in eq .
( [ sel ] ) using expression ( [ diff ] ) with constant @xmath28 , we obtain the approximate relation between the total cross section , the elastic cross section , and the slope : @xmath40 we can compare this formula with the upper bound obtained in ref .
@xcite : @xmath41 the phase @xmath42 of the hadronic amplitude is often defined as @xmath43 then @xmath44 these formulas are used for measuring the luminosity , which relates the cross section @xmath45 of a given process @xmath46 to the corresponding number of events @xmath47 by @xmath48 a simultaneous measurement of the total number of events @xmath49 and the number of elastic events @xmath50 is used to define the luminosity as @xmath51 the measured total cross section is independent of luminosity : @xmath52 the elastic scattering amplitude must satisfy the general properties of analiticity , crossing symmetry , and unitarity .
the unitarity of the @xmath53-matrix , @xmath54=1 , imposes certain requirements on it . in the @xmath30-channel
, we have @xmath55 [ \cos ( \theta _ 1 -\theta _ 2 ) -\cos \theta ] } } + f(p,\theta ) . \label{unit}\end{aligned}\ ] ] the region of integration in ( [ unit ] ) is given by the conditions @xmath56 the integral term represents the two - particle intermediate states of the incoming particles .
the function @xmath57 represents the shadowing contribution of the inelastic processes to the elastic scattering amplitude . following @xcite , we call it the overlap function
. it determines the shape of the diffraction peak and is completely non - perturbative .
only some phenomenological models can claim to describe it . in the forward direction
@xmath58=0 , this relation , in combination with optical theorem ( [ sigt ] ) , reduces to the general statement that the total cross section is the sum of cross sections of elastic and inelastic processes : @xmath59 unitarity relation ( [ unit ] ) has been successfully used @xcite for the model - independent description of the orear region between the diffraction cone and hard parton scattering , which became the crucial test for phenomenological models . experimentally , all characteristics of elastic scattering are measured as functions of the energy @xmath30 and transferred momentum @xmath8 .
however , it is appealing to have concrete information on the geometric structure of scattered particles and the role of different space regions in the scattering process .
we should use the fourier - bessel transform to obtain the correspondence between the transferred momenta and these space regions .
the transverse distance between the centers of colliding particles , called the impact parameter @xmath60 , determines the effective transferred momenta @xmath8 .
the amplitudes in the corresponding representations are related as @xmath61 more peripheral collisions with large @xmath62 lead to smaller transferred momenta @xmath7 .
the amplitude @xmath63 can be connected to the eikonal phase @xmath64 and to the opaqueness ( or blackness ) @xmath65 at the impact parameter @xmath60 by the fourier - bessel transformation @xmath66 the integration is over the two - dimensional space of the impact parameter @xmath60 .
assuming @xmath65 to be real and using eq .
( [ sigt ] ) , we obtain @xmath67 also , @xmath68 and @xmath69 to apply the inverse transformation , we must know the amplitude @xmath63 at all transferred momenta . therefore , it is necessary to continue it analytically to the unphysical region of @xmath8 @xcite .
this can be done @xcite .
correspondingly , the mathematically consistent inverse formulae generally contain the sum of contributions from the physical and nonphysical parts of the amplitude @xmath63 .
unitarity condition ( [ unit ] ) involves only the amplitude in the physical region ; only this part of its fourier - bessel transform is also important in the unitarity relation for the impact parameter representation .
it is written as @xmath70 where @xmath71 and @xmath72 are obtained by the direct transformation of @xmath63 and @xmath73 integrated only over the physical transferred momenta from @xmath74 to 0 .
they show the dependence of the intensity of elastic and inelastic interactions on the mutual impact parameter of the colliding particles .
analogously to relation ( [ tein ] ) , the integrals over all impact parameter values in this relation respectively represent the total , elastic , and inelastic cross sections .
it is especially simple to calculate the overlap function from algebraic eq .
( [ unib ] ) if the real part is small in some subregion , i.e. @xmath75
. then @xmath76 in the region where the transformed overlap function is small , @xmath77 , the imaginary part is also small : @xmath78 .
however , the accuracy of the unitarity condition in @xmath62-representation ( [ unib ] ) is still under discussion @xcite , because some corrections due to the unphysical region enter there , even though their role may be negligible .
moreover , the further use of the approximate formulas of the quasi - eikonal unitarization often leads to failure in describing the differential cross section outside the diffraction cone .
the average values of the impact parameters for all elastic and inelastic processes can be estimated from the amplitude @xmath63 if we assume that @xmath79 at @xmath36 @xcite : @xmath80 where , e.g. , @xmath81 nevertheless , the problem of the relative contributions of the central ( small @xmath62 ) and peripheral ( large @xmath62 ) regions under elastic hadron collisions is still widely disputed .
we must be especially careful when considering unitarity condition ( [ unib ] ) with small impact parameters for certain models .
slight variations of @xmath71 in this region may lead to strong variations of the amplitude @xmath63 at large @xmath7 .
the elastic scattering at extremely small angles allows estimating the forward ratio of the real part of the amplitude to its imaginary part @xmath34 in experiment . for completeness
, we show an approximate expression for the amplitude @xmath63 in the region dominated by the coulomb amplitude and its interference with the nuclear amplitude : @xmath82 where the upper ( lower ) sign corresponds to the scattering of particles with the same ( opposite ) electric charges , the form factors of two colliding particles @xmath83 added by hand in eq .
( [ inter ] ) take their internal composition into account , @xmath84 is the coulomb phase , and @xmath85 is the fine structure constant .
the expressions for @xmath83 and @xmath84 depend on various prescriptions for them obtained with different assumptions concerning the internal structure of a hadron .
the most popular shapes of the form factors are either the gaussian fall - off with an increasing angle , like @xmath86 , similar to that in ( [ diff ] ) , or the dipole ( power - like ) approximation , like @xmath87 , with some more complicated subleading factors .
the phase @xmath84 usually contains a term with the typical logarithmic dependence on the angle @xmath22 , which becomes large at very small angles , and some subleading terms . in both cases ,
the subleading terms have to contain additional free parameters for a more accurate description of experimental data .
as we see , the ratio @xmath88 in ( [ rho ] ) is approximated by @xmath89 in the fit ( [ inter ] ) .
this implies that both real and imaginary parts of the nuclear amplitude exhibit the same purely exponential @xmath8-dependence in the interference region ( with the dominance of the imaginary part for small @xmath34 ) .
more details can be found in @xcite .
we first discuss the asymptotic properties of fundamental characteristics such as the total cross section @xmath25 , the elastic cross section @xmath90 , the ratio of the real part to the imaginary part of the elastic amplitude @xmath91 , and the width of the diffraction peak @xmath28 at infinite energies .
then we compare this with some trends in present experimental data . more than half a century ago , it was claimed @xcite that according to the general principles of field theory and ideas about hadron interactions , the total cross section can not increase with energy faster than @xmath92 .
the upper bound was recently improved @xcite , with the coefficient in front of the logarithm shown to be half that in the earlier limit , @xmath93 where @xmath94 is a pion mass .
if estimated at present energies , this bound is still much higher than the experimentally measured values of the cross sections , with @xmath95=1 gev@xmath13 chosen as a `` natural '' scale .
therefore , this is only a functional constraint ; it forbids extremely fast growth of the total cross section , asymptotically exceeding the above limits .
both the coefficient in front of the logarithm in ( [ asymp ] ) and the constancy of @xmath95 are often questioned . in particular , some possible dependence of @xmath95 on the energy @xmath30 has been pointed out ( see , e.g. , @xcite ) .
the heisenberg uncertainty relation shows that such a regime favors an exponentially bounded spatial profile of the matter density distribution @xmath96 in colliding particles , such as @xmath97 .
because the energy density is @xmath98 and there should be at least one created particle with mass @xmath99 in the overlap region , the condition @xmath100 leads to @xmath101 and , consequently , to the functional dependence in ( [ asymp ] ) .
it was namely heisenberg who first proposed such a behavior of total cross sections @xcite .
he considered the pion production processes in proton - proton collisions as a shock wave problem governed by some nonlinear field - theory equations . to study the asymptotic regime ,
some theoretical arguments based on the general principles of field theory and the analogy of strong interactions to massive quantum electrodynamics @xcite were promoted .
the property that the limits as @xmath102 and @xmath103 ( where @xmath104 is the photon mass ) commute has been used @xcite , implying that the asymptotic domain of strong interactions coincides with the massless limit of quantum electrodynamics .
these studies led to the general geometric picture of two hadrons colliding with asymptotically high energies and interacting as lorentz - contracted black disks ( see also review paper @xcite ) . in
what follows , we discuss some other possibilities as well .
but as a starting point for further reference , we describe the predictions of this proposal .
the main conclusions are : \1 . for black ( @xmath105 ) and logarithmically expanding disks with finite radii @xmath106 ( @xmath107=const )
, it follows from ( [ cstb ] ) that @xmath25 asymptotically approaches infinity as @xmath108 \2 .
the elastic and inelastic processes give equal contributions to the total cross section : @xmath109 this quantum mechanical result differs from `` intuitive''@xmath11 classical predictions .
the width of the diffraction peak @xmath110 must shrink because its slope increases as ( see also @xcite ) @xmath111 \4 .
the forward ratio of the real part to the imaginary part of the amplitude @xmath34 must vanish asymptotically as @xmath112 this result follows directly from eq .
( [ rhodi ] ) for @xmath113 .
the differential cross section has a shape resembling the classical diffraction of light on a disk : @xmath114 ^ 2 , \label{bdis}\ ] ] where @xmath115 .
the product of @xmath25 with the value @xmath116 of @xmath7 at which the first dip in the differential elastic cross section occurs is a constant independent of the energy : @xmath117 where @xmath118 is the first zero of @xmath119 .
these are merely a few conclusions among many others , albeit model - dependent ones .
none of these asymptotic predictions have been observed yet in experiment .
surely , there is another possibility
more realistic at present energies that the black disk model is too extreme and the gray fringe always exists .
it opens the way to much speculation , with many new parameters concerning particle shape and opacity ( see , e.g. , @xcite ) .
the black disk limit might be unrealistic .
therefore , in table 1 we show the predictions of the gray disk model with the steep rigid edge described by the heaviside step - function and the gaussian disk model .
the total cross section , the slope @xmath28 , the ratio of the elastic to total cross section @xmath120 , the ratios @xmath121 and @xmath122 , and the product @xmath123 are displayed there ; @xmath124 is the diffraction profile function . the slope @xmath28 is completely determined by the size of the interaction region @xmath106 .
other characteristics are sensitive to the blackness of disks @xmath125 . in particular ,
the ratio @xmath126 is proportional to @xmath125 .
the ratio @xmath127 plays an important role for fits at larger angles , as explained in section 4.2 .
it is inversely proportional to @xmath125 .
the corresponding formulas are given by ( [ cstb ] ) , ( [ cselb ] ) and ( [ bb ] ) .
the black disk limit follows from the gray disk model at @xmath128 . for a gaussian distribution of matter
, the disk becomes nontransparent at its center in this limit .
@xmath129 the gray and gaussian disks models @xmath130 [ cols="<,<,<,<,<,<,<,<",options="header " , ] the most interesting feature of the experimental results is the minimum of the blackness parameter @xmath125 at isr energies .
it can be clearly seen in the minima of @xmath126 and @xmath122 and in the maximum of @xmath127 at @xmath10=62.5 gev .
the steady decrease of ratios @xmath126 proportional to @xmath125 and @xmath122 proportional to @xmath131 up to the isr energies and their increase at s@xmath4s , tevatron , and lhc energies means that the nucleons become more transparent up to the isr energies and more black toward 7 tev .
the same conclusion follows from the behavior of @xmath127 , which is inversely proportional to @xmath125 .
the value of @xmath127 rapidly approaches its limit for the gaussian distribution of matter in the disk .
for the gaussian shape , the parameter @xmath122 can not exceed 0.25 .
this model is excluded only at low energies . according to eq .
( [ stseb ] ) @xmath132 , which is indeed close to 0.25 within the limits of experimental errors , the estimate of @xmath133 , and slight variations of @xmath28 inside the cone in the framework of our crude model as predicted in table 1 .
this shows that our models are not bad for qualitative estimates in a first approximation . before dwelling on various fits ,
we briefly comment on some important general trends in high - energy data observed in experiment .
total cross sections increase with energy . at present energies ,
the power - like approximation is the most preferable one .
the preasymptotic behavior of @xmath25 proposed in earlier papers @xcite was @xmath134 where the numerical value of @xmath135 was estimated to be of the order of unity in strong interactions .
it was shown in @xcite to lie in the range between 0.08 and 0.2 , which is close to values obtained in recent phenomenological fits .
the power growth persists in a wide interval of energies ( see ref .
@xcite for the recent analysis of experimental data ) .
consequently , the density distribution in colliding particles is closer to a power - like dependence than to an exponential one in that energy range .
the ratio @xmath136 decreases from low energies to those of isr , where it becomes approximately 0.17 and then strongly increases up to 0.25 at lhc energies .
however , it is still quite far from the asymptotic value 0.5 , corresponding to the black disk limit .
the only higher - energy data came from the pierre auger collaboration , which recently reported @xcite a measurement of the inelastic @xmath20-air cross section @xmath137 at @xmath138 tev .
after some corrections and glauber model calculations , it results in the @xmath3 inelastic cross section @xmath139 mb . some models @xcite extrapolate their predictions for the total cross section to this energy and obtain a value of about 135 mb .
hence , the ratio of the inelastic to the total cross section could become equal to 0.67 , which is smaller than 0.75 at 7 tev . however , it is premature to reach any definite conclusions because of large errors in the cosmic ray data and the underestimated value of the total cross section predicted at 7 tev by the model @xcite .
the extrapolation to infinite energies done in the same model leads to this ratio estimated as 0.509 , which is compatible with the black - disk predictions .
still , asymptopia is but an elusive concept !
sometimes , the modified black disk limit is attributed to the sum of elastic and diffractive processes @xcite .
it may then be that @xmath140 where @xmath141 is the sum of cross sections of single and double inelastic diffraction .
the fits in ref .
@xcite suggest separately the relations @xmath142
the diffraction peak shrinks about twice from energies about @xmath143 gev , where @xmath144 gev@xmath145 , to the lhc energy , where @xmath146 gev@xmath145 . at isr energies , the slope @xmath147 increases logarithmically .
accounting for lhc data requires a stronger dependence than a simple logarithmic one .
the terms proportional to @xmath92 are usually added in phenomenological fits .
even then , predictions @xcite are not completely satisfactory . at present energies , in connection with the power - like preasymptotic behavior of @xmath25
, we could also expect a faster - than - logarithmic shrinkage of the diffraction peak .
the tendency in peak behavior at larger @xmath7 also changes with an energy increase . in the energy region up to isr , it becomes less steep near its end ( see figs 4 , 5 in ref .
@xcite ) , but its slope increases at the lhc energies . both the minimum and maximum following the peak shift to smaller @xmath7 .
as regardss the behavior of the differential cross section in the function of the transverse momentum behind the maximum , the @xmath8-exponential of the diffraction peak is replaced , according to experimental data , by the @xmath148-exponential at the intermediate angles : @xmath149 the slope @xmath150 in this region also increases with energy , and the whole orear region shifts to the ever lower transferred momenta . in this connection
, we also note an intriguing property of the ratio @xmath121 , which is closely related to the value of the slope @xmath28 . from table 2 , we see that it is about 1 at @xmath10 = 4 gev , increases to 1.5 at isr energies and then again drops to around 1 at 7 tev .
this ratio , in combination with values of @xmath91 at different angles , determines the slope in the @xmath7-region beyond the diffraction peak at any @xmath30 ( see ref . @xcite and the discussion in the subsection 4.2.2 ) . according to eqs ( [ cst ] ) , ( [ wid ] ) , @xmath127 should decrease and be asymptotically equal to 1/2 for the black disk limit , such that the relation @xmath151 be asymptotically fulfilled . at the lhc energy
7 tev , the coefficient in the right - hand side is still half as much .
however , if the preasymptotic power - like increase in the total cross section accompanied by a slower increase in the slope persists , the tendency to this limit looks quite promising .
the relation between @xmath25 and @xmath28 is also discussed in refs @xcite . in particular , the fits in @xcite correspond to the value @xmath152 at auger energies 57@xmath1536 tev , i.e. lower than 1 at 7 tev . \4 . as a function of energy ,
the ratio @xmath34 increases from negative values at comparatively low energies , crosses zero in the region of hundreds gev and becomes positive at higher energies .
this is a general tendency for collisions of any initial particles . for @xmath3 scattering ,
the prediction of ( [ rho0 ] ) with values of @xmath30 scaled by 1 gev is still somewhat higher ( about 0.177 ) than the estimates from dispersion relations ( @xmath1540.14 in refs @xcite ) , even at 7 tev , while strongly overshooting them at isr , where @xmath155 .
no logarithmic decrease is seen in these predictions , which , however , depend on the behavior of the total cross section at higher energies .
moreover , the value 0.14 can only be reached according to ( [ rho0 ] ) at the energy of 75 tev .
probably , at energies higher than 75 tev , the first signs of approach to the asymptotic regime will become visible .
no data about @xmath34 at the lhc energies exist yet . the local value of @xmath34 estimated from eq .
( [ rhodi ] ) with a power - like fit of the total cross section , proportional to @xmath156 , is @xmath157 .
that agrees quite well with the soft pomeron intercept @xmath158 .
\5 . to describe the shape of the differential cross section in the diffraction cone , significant corrections to eq .
( [ bdis ] ) must be added at present energies .
this is discussed in subsection 4.1 .
the product @xmath159 changes from 39.5 mb@xmath160gev@xmath13 at @xmath161 gev to 51.9 mb@xmath160gev@xmath13 at @xmath162 tev and strongly deviates from the predicted asymptotic value ( [ gams ] ) .
the total cross section @xmath25 increases faster than @xmath116 decreases . from the geometrical point of view , the general picture is one of protons becoming blacker , edgier and larger ( bel ) @xcite .
we conclude that even though the qualitative trends may be considered rather satisfactory , we are still quite far from the asymptotic regime , even at lhc energies .
this feature may be connected @xcite with the strong evolution of the parton content of strong interactions at present energies , revealing itself in an increase in the number of active parton pairs inside each proton with energy increase ( higher density ) and a softening of the structure functions , which leads to lower energy shares @xmath163 for each parton pair ( larger radii ) .
as always , our knowledge about particular physical processes is limited by the practical possibility of measuring their characteristics . as mentioned above , numerous experimental data on the elastic scattering of hadrons at various angles and at different energies have been obtained .
unfortunately , in some of them the available region of angles is strongly limited by the experimental setup .
therefore , a comparison with theoretical proposals is possible only in the corresponding range of angles and energies .
the data and their fits at various energies and in different intervals of transferred momenta for different participating particles are so numerous that it is impossible to show all of them in a single review paper .
therefore , from the very beginning , we use the latest results of the totem collaboration at the highest lhc energy , 7 tev , as a reference point @xcite .
the discussion of theoretical models is also concentrated near these data . the total and elastic cross sections at 7 tev are respectively estimated as 98.3 mb and 24.8 mb .. the figures from published papers @xcite demonstrating the behavior of the differential cross section as function of the transferred momentum are displayed below .
they clearly confirm the existence of the three regions discussed above . the cross section shape in the region of the diffraction cone @xcite is shown in fig .
the @xmath164-exponential behavior with @xmath16520.1 gev@xmath145 is clearly seen at @xmath166 gev@xmath13 .
the peak steepens at the end of the diffraction cone , and its slope becomes approximately equal to 23.6 gev@xmath145 in the @xmath7 interval of ( 0.36 0.47 ) gev@xmath13 .
the results at somewhat larger angles @xcite in the orear region are presented in fig .
the dip at @xmath167 gev@xmath13 with a subsequent maximum at @xmath168 gev@xmath13 and the @xmath12-exponential behavior are demonstrated . some curves , corresponding to different model predictions
, are also drawn here .
the same data as in fig .
2 are shown in fig . 3 , but with more details , including the steepend slope , the dip position , and the region of @xmath169-behavior .
the last one is ascribed to the hard parton scattering processes .
we congratulate all members of the totem collaboration with this fantastic achievement !
their efforts are truly appreciated when estimating the values of angles at which the measurements had to be done .
they were even smaller than 10@xmath170 !
detectors had to be installed at very long distances from the collision point to obtain results at low transferred momenta .
these data revived interest to elastic scattering .
theoretical models usually describe the diffraction cone and values of total and elastic cross sections related to it more or less precisely ( therefore , their fits , that are almost indistinguishable in that region , are not drawn in fig .
however , all of them fail to quantitatively predict the behavior of the differential cross section outside the diffraction cone as can be seen in fig .
the predictions of five models @xcite are drawn here .
they are very widely spread around the experimental line .
we can conclude that just this region becomes the occam razor for all models . in
what follows , we consider these models , as well as some others , in more detail .
1 . the differential cross section of elastic proton - proton scattering at @xmath10=7 tev measured by the totem collaboration ( fig . 4 in @xcite ) .
+ the region of the diffraction cone with the @xmath7-exponential decrease is shown .
the differential cross section of elastic proton - proton scattering at @xmath10=7 tev measured by the totem collaboration ( fig . 4 in @xcite ) . + the region beyond the diffraction peak is shown .
the predictions of five models are demonstrated .
3 . the differential cross section of elastic proton - proton scattering at @xmath10=7 tev measured by the totem collaboration ( fig . 3 in @xcite ) .
+ the same regions as in fig .
2 are shown with the values of the steepened slope at the outskirts of the diffraction peak , the position of the dip and the power - like behavior at the largest transferred momenta .
the three intervals of @xmath7 ( the diffraction cone , the orear regime , and the region of hard parton scattering ) are characterized by different dynamical content , as we understand it now .
they require separate approaches to their descriptions .
it seems reasonable that these regions are regulated by different but interrelated physical mechanisms . in particular , different spatial regions of overlapping colliding objects
are responsible for corresponding effects .
three subsections 4.1 , 4.2 and 4.3 are devoted to theoretical approaches to their explanation .
the internal structure of colliding , strongly interacting particles plays a crucial role in the outcome of their collisions . in high - energy hadron - hadron scattering
, each hadron behaves as an extended object .
they can be described by their size and the density of their constituents .
the simplest models are demonstrated in table 1 .
since long ago , it has been believed that hadrons contain some denser core surrounded by a meson ( pion ) cloud at their periphery .
this idea was a cornerstone of the one - pion exchange model , which was first proposed in ref .
@xcite to describe particle production in peripheral interactions .
it evolved into the well - known multiperipheral and ( multi)reggeon exchange models ( see , e.g. , @xcite for early review papers ) .
they are rather successful in describing many features of multiparticle production processes .
the multiperipheral approach developed , for instance , in the framework of the bethe - salpeter equation ( see ref .
@xcite ) can be considered an attempt to account for the @xmath8-channel unitarity .
nowadays , it is commonly believed that , at very high energies , the total cross section is dominated by peripheral events . in modern parlance , this is related to the long - range nature of the field of `` perturbatively massless '' gluons .
the exchanged boson mass may mimic a nonperturbative mass gap in qcd with the `` effective '' gluon mass of the order of 1 gev and a gluon - gluon correlation length about 0.3 fm .
the pion mass scale is rather small , and more general `` boson '' exchange is preferred .
the weight factors of different mass scales take the impact parameter distribution of the particle opacity into account .
the role of inelastic channels in describing elastic scattering can be revealed by understanding the origin and prescribing a definite shape to the overlap function @xmath171 in the @xmath30-channel unitarity condition ( [ unit ] ) or , equivalently , to its fourier transform in the impact parameter picture . the scattering is mainly diffractive , i.e.
, it is due to the absorption of incoming waves in many open inelastic channels .
its quantitative field - theory treatment presents a serious unsolved problem .
the overlap function contains the sum of products of a matrix element of the inelastic process with a particular final state and the complex conjugate matrix element with the same final particles content .
however , their kinematical difference must be taken into account , due to the fact that the two final protons are scattered at an angle @xmath22 relative to the initial ones .
correspondingly , the overlap of the momentum distributions of the intermediate inelastic @xmath172-particle states is nontrivial kinematically and , what is especially important , the phases of these matrix elements become crucial .
the phases are related to the position in space where particles are produced .
it has been pointed out in many papers @xcite that only the phase cancellation effect , which is closely related to particle correlations in inelastic processes , can lead to a realistic shape of the diffraction cone . the problem of properly accounting for them has not yet been solved . at the same time
, elastic scattering should be less peripheral because of a larger number of exchanged objects if regarded as an @xmath30-channel iteration of the overlap function .
the great difficulty in transferring large momenta reveals itself already in the sharp shape of the forward diffraction peak .
there have been numerous attempts to understand it in terms of the peripheral approach ( see , e.g. , @xcite ) .
unfortunately , no framework for commonly accounting both the @xmath30- and @xmath8-channel unitarity conditions has been developed . in general
, there have been many ideas proposed for describing elastic scattering processes , but no cogent theoretical arguments to justify the particular forms relying mainly on `` intuition '' have been offered .
the fact that they are very simple is usually the only advantage .
any strict interpretation is an idealization and as such it should not be expected to be exactly true .
the key elements of the geometric approach are : the use 1 ) of the impact parameter picture with fourier - bessel transformation ( [ eik ] ) , ( [ hsb ] ) from the transferred momenta amplitude to the spatial description , 2 ) of eikonal approximation ( [ eik ] ) , and 3 ) of unitarity condition ( [ unib ] ) .
the @xmath53-matrix in the impact parameter picture is chosen in the exponential form @xmath173 and the convolution approximation for the real opacity @xmath174 for elastic @xmath175 scattering is used : @xmath176 here , @xmath177 denotes the convolution of hadronic matter density distributions @xmath178 for @xmath15 and @xmath28 .
@xmath179 is an energy - dependent factor .
the assumptions about the validity of the eikonal approximation , the nearly imaginary character of the scattering amplitudes at low transferred momenta , the proportionality between the hadronic matter distribution and the electric charge distribution , the exponentiation of the @xmath53-matrix in @xmath62-space , and the validity of unitarity condition ( [ unib ] ) are widely used .
the droplet model @xcite for elastic collisions was the first to fully exploit all the above elements .
particles were pictured as very much similar to nuclei .
correspondingly , the notion of the density distribution @xmath178 inside a particle was introduced such that @xmath180 in potential models it corresponds to the wkb approximation . for the gaussian shape of @xmath71 , it is possible to solve for @xmath178 from ( [ dens ] ) , obtaining the function familiar in the theory of bose - einstein condensation of free particles @xcite . in the droplet model ,
the properties of the disk are independent of the energy at sufficiently high energies .
many diffractive minima in the differential cross section have been predicted .
the dipole form factors in the @xmath8-representation led to @xmath181 with a shape of the modified bessel functions , which allowed fitting differential cross sections at isr energies @xcite .
the intuitive picture of high - energy hadron collisions as two extended objects breaking in fragments ( and thus defining the overlap function ! ) has promoted the hypothesis of limiting fragmentation @xcite inspired by the droplet model .
models based on consideration of tower diagrams @xcite predict that the disk becomes larger and more absorptive as energy increases .
both the black core and gray fringe expand with energy and become more absorptive .
the first estimates of the radii of protons , pions , and kaons from their form factors @xcite showed that protons are larger than pions and kaons .
this is not surprising in view of the smaller cross sections of @xmath5 and @xmath182 interactions than those of @xmath3 .
the typical size is somewhat smaller than 1 fm .
the proton hadronic matter distribution was fitted by a dipole form similar to the electric form factor but with the energy - dependent radius .
other early attempts to consider elastic scattering of hadrons also stemmed from the analogous simple geometrical treatment of their internal structure @xcite .
later , more complicated models were used .
the main focus is , surely , on processes at small angles within the diffraction cone .
they define the bulk contribution to the elastic scattering cross section due to the steep fall - off of the distribution with increasing angles .
different models happen to fit experimental data in the cone quite well in a wide energy range .
but they fail outside the diffraction peak , as mentioned above .
large - angle scattering requires more central collisions with a lower impact parameter to probe the internal content of particles .
therefore , these regions of transferred momenta are discussed separately below .
some ideas stemmed from regularities in inelastic processes .
the multiplicity distributions of created particle are closely related to the purely geometric notion of the centrality of collisions .
when the scaling of multiplicity distributions @xcite was supported by experimental data , the proposal of the geometric scaling @xcite for the elastic amplitude was promoted .
the difficulties in accelerating the various parts of a nucleon without breaking it up had to be accounted for .
the basic idea of the geometric scaling is that , at sufficiently high energies , the amplitude @xmath63 depends on a single variable , the scaling parameter @xmath183 : @xmath184 this idea has led to several predictions at asymptotically high energies and is still actively being debated now .
such scaling was proved @xcite for cross sections increasing as @xmath185 and for an infinitesimally small ratio of the real to imaginary part of the amplitude @xmath186 at @xmath102 .
the latest results on @xmath88 discussed in subsection 4.2.5 do not support this assumption .
the purely geometrical standpoint is adopted in refs @xcite .
the three regions in the behavior of the differential cross section are clearly reflected in the three spatial scales of the internal hadronic structure considered in @xcite .
the authors of this three - scale model claim that the nucleon has an outer cloud of the quark - antiquark condensate , an inner shell of the baryonic charge density , and a still smaller internal core of massless color - singlet valence `` quarks '' surrounded by low-@xmath163 gluon clouds about 0.3 fm in size .
this picture is shown in fig .
4 . fig .
the nucleon structure according to the model @xcite .
+ the three regions of the internal structure are supposed to be directly responsible for the three regimes in the behavior of the differential cross section . the diffraction cone is described as a result of cloud - cloud interaction , represented by a class of potentials containing the sum of the modified bessel functions .
the least massive exchanged quanta are the most important ones . at larger momentum transfers , the baryonic charge at intermediate distances
is probed by the @xmath187-exchange .
the internal region filled in by the valence quarks starts playing its role in the presence of even larger transferred momenta .
the diffraction profile function , which defines the range of different densities and , correspondingly , different forces , is taken to be @xmath188 .
\label{gisl}\ ] ] the parameters @xmath189 and @xmath135 are energy dependent , @xmath190 and @xmath191 is a coupling strength ; @xmath95=1 gev@xmath13 .
these functions render the shape of the differential cross section , similar to the fraunhofer diffraction ( see subsection 4.1.2 ) , with the form factor proportional to @xmath192 ( @xmath193 is an adjustable parameter ) proposed a long time ago @xcite .
this form factor also extends somewhat to transferred momenta outside the diffraction cone .
unfortunately , the contemporary phenomenological analysis of experimental data is not able to determine the impact parameter profiles unambigously . the scattering due to @xmath187-exchange is parameterized by the product of the @xmath187-propagator and two form factors @xmath194 directly in the @xmath195-representation : @xmath196 the amplitude due to quark - quark scattering has two `` structure factors '' @xmath197 of valence quarks ( different from the above form factors ! ) , the propagator with the black disc radius @xmath198 of @xmath199 asymptotic scattering and @xmath30-dependent factors with the hard pomeron intercept equal to @xmath200 : @xmath201 in total , there are seventeen adjustable parameters in the model .
as mentioned in @xcite , the fits according to this model predict too low value of the slope @xmath28 at @xmath7=0.4 gev@xmath13 and strongly disagree with experiment at 7 tev outside the diffraction peak ( see fig .
2 ) . formulas ( [ omisl ] ) , ( [ qqisl ] ) are aimed to improve the fit just in this region , but they do not help . in general , an internal region of the nucleon where the gluons cluster around the original valence quarks resembles the valon model @xcite .
similar pictures arise in the qcd - inspired models discussed below .
surely , some care should be taken for any such model to be accepted and the geometric picture to be considered seriously , especially in view of its success or failure to describe experimental data in the whole range of transferred momenta at various energies . for a long time ( see , e.g. , ref .
@xcite ) , the formulas of classical diffraction of light on a ( black or grey ) disk with the traditional bessel functions have been used for hadronic reactions .
recently , an analogous expression for the elastic amplitude was considered in ref .
@xcite : @xmath202 . \label{ugal}\ ] ] the free parameters in arbitrarily chosen analytic expression ( [ ugal ] ) are @xmath203 .
the first term resembles the expression for the black disk ( [ bdis ] ) .
the suppression at large transferred momenta is assumed to be approximated by the form factor in front of the bessel functions . in the impact parameter representation this shape corresponds to the ordinary fermi profile used , e.g. , in refs @xcite and shown in eq .
( [ gisl ] ) : @xmath204 the second term in ( [ ugal ] ) in brackets takes the contribution due to the real part of the amplitude into account .
it should smooth the behavior of the differential cross section near zeros of the first term .
this seems to be the only difference from the first component of the previously discussed model @xcite . and , again , comparison with experimental data shows that the results of fits are satisfactory in the diffraction cone , but not outside it .
the form factor in front of common bessel functions does not fit the large @xmath7 trends of experimental distributions . throughout these developments , modifications of early guesses
have been found necessary , but the general spirit of the geometrical description remains immutable and viable .
the strongly interacting content of hadrons is often considered to be similar to their electromagnetic substructure @xcite .
similarly to the droplet model , the assumption of the proportionality between the hadronic matter distribution and the electric charge distribution is used in many models . however , in most of them , the electromagnetic form factors are used in combination with reggeon exchanges because , considered alone , they do not reproduce the energy dependence of the main characteristics .
however , the assumption about the full congruence of these distributions is not necessarily valid , since gluons do not carry an electric charge even though they play an important ( if not decisive ) role in high - energy strong interactions .
that is why the charge and matter distributions in some models are parameterized separately or some corrections are added . using the experience from calculation of tower diagrams in electrodynamics and the impact - parameter representation ,
it was proposed @xcite that the possibility of choosing the opacity @xmath205 in a factorized form be considered : @xmath206 where @xmath207 is chosen to be crossing symmetric under @xmath208 and to reproduce the energy dependence of the pomeron , considered as a fixed regge cut , @xmath209 while @xmath210 is taken as the bessel transform of @xmath211 here , @xmath212 stands for the proton `` nuclear form factor '' , parameterized like the electromagnetic form factor with two poles : @xmath213 other factors with the parameter @xmath135 are introduced `` by hand '' .
they can be treated just as a correction due to the different shapes of distributions of charge and matter .
there are six adjustable parameters in total used at high energies if the regge background is neglected .
the noticeable @xmath8-dependence of the slope @xmath214 in the diffraction cone is predicted .
however , its values at 7 tev are lower than experimental ones ( about 18 gev@xmath145 instead of 20.1 gev@xmath145 ) at @xmath215 gev@xmath13 , slightly exceed them in the tiny interval near 0.35 gev@xmath13 and do not reach the value 23.6 gev@xmath145 mentioned above .
this model is close to the totem data @xcite for the dip position and the exponential at very large @xmath7 , but predicts values of the differential cross section in the orear range , @xmath216 0.36 gev@xmath13 , about twice as large ( see fig .
in addition to the dip , some `` oscillations''@xmath11 at the transferred momenta of several gev@xmath13 are predicted ( up to the energy 6000 tev ) but not yet observed . in general ,
such structures appear as a byproduct of the eikonal approach and unitarization procedure ( see , e.g. , ref .
their energy dependence is strongly determined by the parameters used in formula ( [ ss ] ) to account for the crossing symmetry property of the amplitude .
the same parameters are crucial for the behavior of the real part of the amplitude .
it is interesting that the model predicts the dominance of the imaginary part of the amplitude even at large transferred momenta .
the real part becomes important only at zeros of the imaginary part .
the dip and oscillations are noticeable precisely there . near the cone ,
the model predicts two zeros of the real part of the full ( coulomb + nuclear ) amplitude at @xmath2170.0064 gev@xmath13 and the nuclear amplitude alone at @xmath216 0.18 gev@xmath13 , as well as one zero of the nuclear imaginary part at @xmath2170.5 gev@xmath13 . in the differential cross section , the last zero is partly filled in by the real part .
we note the difference between the power - like expression for @xmath218 and its exponential behavior in the traditional regge models .
the exponentiation of this form of @xmath218 leads to additional oscillations .
the similar but more complicated combination of the form factors has been used in refs @xcite .
the authors consider the @xmath8-dependent mellin transforms of parton distributions and claim that the first moment defines the form factor of the standard pomeron @xmath197 , while the second moment @xmath219 corresponds to interaction attributed to three nonperturbative gluons .
thus , the behavior of the differential cross section at small @xmath8 is determined by the elactromagnetic form factors and by matter distribution at large @xmath8 .
the born term of the elastic scattering amplitude is written as @xmath220 where @xmath221 @xmath222 @xmath223 gev@xmath13 , @xmath224 gev@xmath13 , @xmath225 gev@xmath13 , @xmath226 gev@xmath145 .
we note that the slope of the second term is chosen as one fourth of the first term .
the final form of the amplitude is obtained after eikonalization of the born contribution using the opacity @xmath227 the total cross section at 7 tev was predicted to be equal to 95 mb .
authors demonstrate good fits of @xmath3 and @xmath4 differential cross sections , as well as of @xmath38 , in a wide energy range , including the totem data .
only five ( three for high energies and two for low energies ) adjustable parameters are claimed to be used if all above values are regarded as fixed .
in fact , there are 10 such additional `` hidden '' parameters in total if the hard pomeron is also considered . surely , the contribution from secondary reggeons at lhc energies is negligible , i.e. , smaller than the experimental errors .
the real part of the hadron amplitude is completely determined by the complex expression for @xmath228 .
its @xmath8-dependence appears just as a byproduct of the eikonalization procedure . as a function of @xmath8
, it tends to zero at @xmath2290.16 gev@xmath13 at the energy of 7 tev .
the interesting predictions of the @xmath8-behavior of @xmath88 at nonforward transferred momenta for different energies are presented .
they are discussed in more details in subsection 4.2.4 .
the regge - pole model is beyond dispute one of the most explored .
it has already been noticed that the notion of regge trajectories has been used in the preceding subsections as well . the only reason to discuss these models there separately was their stronger inclination to the use of nonexponential electromagnetic form factors and geometric pictures in the @xmath30-channel approach . at the same time , reggeon models appeal mostly to the @xmath8-channel approach .
the amplidutes with reggeon exchanges in the @xmath8-channel are the natural candidates for explaining the exponential decrease of the differential cross section ( [ diff ] ) with the squared transferred momentum @xmath230 inside the diffraction cone .
just this shape is typical for them , because it follows from the linearity of regge trajectories .
moreover , they predict the logarithmical increase of the hadronic radii as the energy increases , i.e. , the logarithmical increase in the cone slope @xmath28 or the corresponding shrinkage of the width of the diffraction cone . this prediction is also supported by experiment . in the common regge - pole models ,
the disk becomes larger and slightly more transparent as energy increases .
the standard regge - type models @xcite use the combination of contributions due to the exchange by the ( multicomponent ) pomeron , odderon , and secondary reggeon trajectories corresponding to @xmath231 and @xmath187 mesons with or without the form factors chosen in a simple exponential form or as power - like expressions resembling the electromagnetic structure of colliding partners discussed in the preceding subsection . the price to be paid
is the increased number of adjustable parameters at each step of sophistication . to be more or less realistic , one has to use the knowledge about some of them from other ( independent ? ) experimental results . but even under this condition
, the ambiguity of their choice and sensitivity to fitted parameters leave some freedom in the conclusions .
the amplitudes of @xmath3 and @xmath4 scattering are approximated by the sum of terms corresponding to the leading ( pomeron and odderon ) and nonleading ( @xmath231 and @xmath187 meson ) regge trajectories : @xmath232 , \label{appbar}\ ] ] where the labels @xmath233 stand for the relevant contributions .
the sign in the @xmath3 and @xmath4 amplitudes differs for @xmath234-even and @xmath234-odd terms .
the contributions of the nonleading regge poles are written as @xmath235 with @xmath236 .
while the secondary trajectories are usually chosen in a standard linear way , the pomeron and odderon contributions can be regarded , for example , as dipoles with nonlinear trajectories @xcite
@xmath237 , \label{apom}\ ] ] where @xmath238 the unknown odderon contribution is assumed to be of the same form as that of the pomeron .
the parameters of the trajectories and of the absorption @xmath239 need to be adjusted .
their nonlinearity may be connected with the two - pion threshold following from the @xmath8-channel unitarity @xcite .
however , there could be double counting of the graphs with pomerons attached on both sides to the pion loop .
this is well known from old peripheral models of inelastic processes , where the self - consistent bethe - salpeter equation had to be used for the proper account of the pion - nucleon vertices .
different forms of nonlinear trajectories are in use .
for instance , the pomeron trajectory is chosen in @xcite with four free parameters as @xmath240 the more complicated nonlinearity was used in @xcite .
however , the use of the pion mass as a scale there is questionable in view of the above discussion .
the origin of the pomeron and the parameterization of its trajectory are still being debated .
there is no strict rule for choosing its shape .
the dipole and even tripole forms of unitarized pomeron have been attempted .
they mimic cut contributions @xcite .
moreover , there are arguments in favor of two pomerons with different intercepts .
even the fits with three pomerons are sometimes used @xcite .
the soft pomeron contributes a term with the energy dependence @xmath241 to hadron - hadron total cross sections , and the hard pomeron makes a small ( at present energies ) contribution with a stronger energy dependence @xmath242 .
these values of the intercepts stem from the discussions of hera data ( see , e.g. , @xcite ) .
although the hard - pomeron exchange was unnecessary for describing hadron - hadron total cross sections up to energies @xmath10 below 1 tev , it may reveal itself at lhc energies , as argued in @xcite .
the model in @xcite uses only two terms in the expansion for opacity : @xmath243 where @xmath244 stands for the contribution from single exchanges of reggeons ( two pomerons , @xmath231 , and @xmath187 ) with the adjustable parameter @xmath245 as well as for the triple - gluon exchange of the form @xmath246 needed at larger values of @xmath247 and matched at some @xmath248 to exponential shapes of the diffraction peak and to the dip region . certainly , adding such a term allows fitting the total cross section value at 7 tev , but there is a suspicion that the sharp increase of the hard - pomeron contribution will overpredict the cross sections at higher energies .
the unitarization will become mandatory once again .
the quality of the fit of the differential cross section beyond the diffraction peak is no better than of those fits shown in fig .
several variant forms of born amplitudes and different kinds of eikonalization have been attempted .
there is no consensus on their choice .
the form of the eikonal similar to ( [ dola ] ) is chosen in @xcite with the exponential suppression @xmath249 \label{golo}\ ] ] for central interactions .
the peripheral part of the pomeron interaction with the meson cloud is parameterized @xcite by a small term increasing with the energy and resulting in a @xmath250 exponential fall - off of the differential cross section .
the geometric picture corresponds to a black disk with a grey fringe , similarly to the above - described model @xcite . in general , it is not easy to estimate the total number of the adjustable parameters in different models .
there are parameters related either to @xmath8- ( @xmath62- ) or @xmath30-dependence . in some papers
, it is often assumed that part of them are known from fits of other characteristics of hadron or electromagnetic interactions at various energies and can therefore be considered known beforehand . for example , it is claimed that the model in @xcite contains about 15 parameters . in this case , it is quite difficult to find the proper minima for the matrix of @xmath251 values .
it is well known how unstable the final results can be : one has to choose the step - by - step procedure for doing this and use some special computer codes .
there are 25 adjustable parameters shown in table 1 in papers @xcite .
however , they include some values assumed to be apriori fixed in @xcite . at the same time
, additional form factors were inserted in the formulas , albeit with preliminary `` fixed '' parameters .
they were used to fit 982 @xmath3 and @xmath4 data points in a wide energy range . besides the elastic differential cross sections , the total cross sections and the ratios @xmath34 were considered .
the fit in the interference region of coulomb and hadronic amplitudes with the same parameters helped in choosing among the different coulomb phases proposed previously .
a similar situation is seen in the fits in ref .
@xcite , where it is claimed that the number of parameters is much less ( 5 only ! )
. however , there are many others ( in particular , concerning the energy behavior and form factors ) that are hidden parameters .
they are held fixed from the very beginning , as was discussed in subsection 4.1.3 . as mentioned before
, the simple exponential form of the differential cross section in the diffraction cone is quite well described .
this becomes possible mainly due to the @xmath8-shape of the pomeron trajectory ( [ apom ] ) and other reggeons contributions in ( [ areg ] ) .
the fits in this region at the different energies shown , e.g. , in @xcite , are quite impressive .
the evolution of the diffraction cone slope with energy is reproduced ( as described by @xmath252 in ( [ apom ] ) ) .
unfortunately , the variety of forms of pomeron trajectories with different intercepts , slopes , and shapes of residues unitarized in different ways and/or substituted by regge - cuts is so large that it is impossible to show all of them in this review due to the limited space .
the cuts with nonlinear trajectories mimic hard scattering @xcite .
a common problem appears in predicting them at larger angles .
the fit according to the model in @xcite seems to be most successful in predicting the position of the dip and the shape at large @xmath7 but exceeds the absolute value approximately twofold .
the model in @xcite strongly underestimates it , with the wrong position of the dip and much slower decrease at @xmath253 gev@xmath13 .
this is well demonstrated in fig . 2 and is also discussed below .
we mention that all these papers follow the general approach proposed much earlier @xcite .
they just deal with more detailed fits of newly available experimental data .
each incident particle consists of a superposition of fock states with @xmath172 partons @xcite , which are scattered instantaneously and simultaneously by the other particle .
some qcd - inspired models using this statement have been developed .
the role of partons is played by quarks and gluons .
the two competing mechanisms of hadron interactions , the increase in the density ( @xmath125 in table 1 ) and in the radius @xmath106 , determine their specific features .
in qcd , they can be respectively ascribed to the leading order solution of the bfkl equation @xcite and to the long range ( weizscker - williams ) nature of the field of massless gluons .
the density increase due to the bfkl - evolution leads to a power - like increase of the total cross section , which is nonunitary and violates froissart bound ( [ asymp ] ) .
therefore , at the critical density of the order @xmath254 , the density saturation must be taken into account @xcite .
the qcd evolution in all orders in the gluon density but in the leading logarithmic approximations is treated by the jimwlk equations @xcite . with account of multiple scattering effects ,
they can be simplified in the large @xmath255 limit to a single nonlinear bk equation for the gluon density @xcite when the induced field density is small .
the density growth effects are preasymptotic . according to @xcite ,
they are described by a hard pomeron , while the growth of the size of the black saturated regions ( the radius ) is attributed to a soft pomeron .
the hard pomeron manifests itself in small systems or in small subregions inside hadrons .
the soft pomeron appears in hadronic systems of the typical size and is related to an increased size in the impact parameter space .
only the increase due to the perturbative expansion in the transverse plane remains effective .
there is no common consensus about this scenario proposed in ref .
the soft pomeron is often used @xcite in attempts to explain the preasymptotic power growth of cross sections by an additional nonperturbative mechanism superimposed on the bfkl scenario of hard pomeron .
it is ascribed mainly to the density growth of gluon clouds around quarks and not to the spatial scale of the interaction .
even though the size of gluon clouds increases , it is still limited by a short separation from their source
. the proton looks like three valence quarks surrounded by gluon clouds or spots with mean sizes about 0.3 fm , smaller than the proton radius , of the order of 1 fm .
radiation of any additional gluon from the cloud adds the factor @xmath256 to the interaction cross section , and hence their sum gives a power - like term of the form @xmath257 with a large constant term @xmath258 and small @xmath259 . using the standard dipole form factors of protons and quasieikonal unitarization in the impact parameter space , the authors of this two - scale model @xcite fit many distributions with 10 parameters for @xmath8-dependence ( subject to 2 additional constraints ) and some parameters for the @xmath30-dependence .
such fits are , of course , aimed at high energies of colliding protons where the effects of secondary regge - trajectories die out .
they are mainly successful in the diffraction cone and , consequently , in describing the energy dependence of the total and elastic cross sections .
such a form of the total cross section with an energy - independent term @xmath258 was proposed a long time ago @xcite and actively developed later @xcite in the framework of the parton model and semihard qcd , with the gluon - gluon interaction playing the main role .
the main role of gluons is also incorporated in @xcite .
the profile is chosen in a form containing the @xmath260-terms : @xmath261 where the impact parameter distribution functions are @xmath262 and the gluon - gluon interaction cross section is @xmath263 with @xmath264 . the froissart bound for the total cross section is reproduced with @xmath265 the parameter @xmath266 describes the area occupied by gluons in the colliding protons ( the size effect ) , and @xmath267 is defined via their gluonic structure functions and , therefore , controls their soft gluon content ( the density effect ) .
again , being successful in the diffraction peak with its shape and normalization , the model in @xcite fails to predict the correct behavior of the differential spectrum outside it @xcite . its prediction is more than three times larger than the experimental value at the dip and subsequent maximum , while falling too steeply at ever higher @xmath230 above 1.5 gev@xmath13 ( see fig .
2 ) . attempts to consider the semihard scattering of quarks and gluons can be found in refs @xcite . the traditional partonic description of the process is considered in a series of papers @xcite .
the partonic approach with a hard bfkl pomeron can be merged into the domain regulated by the soft pomeron .
the transition from hard to soft is induced by absorptive multi - pomeron effects in a limited energy range .
the evolution produces parton cascades , not strongly ordered in transverse momenta , with hot spots of a relatively small size in @xmath62-space .
the saturation is driven by the enhanced multi - pomeron graphs , also regulating the high - mass dissociation .
the calculations are done with a 3-channel quasi - eikonal unitarization using the opacity formalism .
they reproduce the shapes of the differential cross sections from isr to lhc within the diffraction cone .
another picture was considered in the framework of the functional integral approach in refs @xcite using the model of the stochastic vacuum and making the assumption that the proton has a quark - diquark structure of the color dipole i.e. two quarks out of three are close together in the transverse directions .
a matrix cumulant expansion is used for vacuum expectation values of wegner - wilson loops @xcite related to hadronic amplitudes .
the qcd vacuum parameters ( the gluon condensate or the string tension , the vacuum correlation length , and the parameter due to the non - abelian tensor structure ) , as well as the hadron size , have been used .
the imaginary part of the amplitude in the @xmath62-representation was calculated . its contribution to experimentally measured quantities
was shown to describe the isr and tevatron data in the diffraction peak reasonably well .
a more phenomenological approach to the quark - diquark model was attempted in refs @xcite .
as above , the correlated quark and diquark constituents are considered . according to the detailed analysis performed in @xcite from isr to lhc energies in the region @xmath268 gev@xmath13 ,
the model is able to describe the data quite well , even outside the diffraction peak , except the narrow strip around the dip .
but it shows a much stronger dip ( by several orders of magnitude ) there than the experimentally observed one .
moreover , similarly to the abovementioned calculations , the model ignores the contributions to the real part of the elastic scattering amplitude . as we saw previously
, such contributions can smooth this dip .
if so , their shape should drastically differ from that of the imaginary part , at least in this strip , as happened , for example , in the models with electromagnetic form factors @xcite .
as long ago as the 1960s , experiments on elastic @xmath3- and @xmath5-scattering at comparatively low energies between 6.8 and 19.2 gev in the laboratory system @xcite showed that the steep exponential fall - off of the differential cross section as a function of the squared transferred momentum @xmath7 is replaced by a slower dependence at larger @xmath7 .
they showed that just after the diffraction cone a shoulder was observed and , even more surprising , a behavior exponentially decreasing with the angle or with @xmath12 , which was called the orear regime after its investigator @xcite .
the special session was devoted to these findings at the 1968 rochester conference in vienna .
the shoulder evolved later into the minimum or dip at higher isr energies .
it has also been observed at the lhc , as seen in figs 2 and 3 .
it is interesting that at fnal - isr energies , @xmath10=6 - 60 gev , the exponential fall - off with an increase of @xmath269 was observed up to quite large values of @xmath229 10 gev@xmath13 @xcite , with the exponent in the range from 6.2 to 7 gev@xmath270 ( see table 7 in ref .
it is even larger at the lhc ( about 8 - 9 gev@xmath270 ) .
the region becomes more narrow and shifts to lower values of @xmath7 from 0.5 to 1.5 gev@xmath13 .
the power - like regime already shows up at about @xmath2292 - 2.5 gev@xmath13 ( see fig .
3 ) . from the very beginning ,
it was noticed @xcite that it is possible to fit the differential cross sections at intermediate values of the momentum transfer by the dependence exponential in @xmath12 ( or @xmath22 ) except the relatively small shoulder region .
to take that into account as well , it was primarely proposed @xcite to use fits with gaussian functions with alternating signs of the coefficients directly in the expression for the amplitude .
the similar approach was later advocated in @xcite . in this way
, both the diffraction peak and larger @xmath7-behavior could be described .
no reference to any phenomenological model is given . from the geometrical point of view
, one can imagine an internal structure with envelopes of alternating density .
such an empirical approach has been recently used @xcite for fits of experimental data at isr energies .
the following parameterization of the amplitude is proposed in ref .
@xcite : @xmath271 , \label{fms11}\ ] ] where @xmath272 . the fits at different energies give information about the ratio of the real to imaginary part of the amplitude @xmath273 , besides the values of adjustable parameters @xmath274 .
two different methods were used . in total , there are 14 to 16 free parameters .
the results of nonlinear fits are rather unstable , and the conclusions are somewhat controversial .
in particular , the numbers of zeros in @xmath275 and @xmath276 differ in these methods .
the dominance of the real part of the amplitude at intermediate values of the momentum transfer in one of the methods is not confirmed when the other method is used .
a similar fit was recently attempted and applied to totem data in refs @xcite . the earlier proposal in ref .
@xcite with phenomenologically chosen two @xmath8-exponentials and the relative interference phase responsible for the dip was applied to totem data . using five parameters , it is possible to describe these data in the whole interval of transferred momenta .
we note that , similarly to the model in @xcite , the slope of the second exponential term is chosen several times smaller than that of the main term .
moreover , when the electromagnetic form factors were tried in place of simple exponentials , the fit became worse .
two exponentials without the interference term inside the diffraction peak and a tsallis - type distribution outside it were used in @xcite .
it was possible , with the help of nine free parameters , to fit the data at energies from 19.4 gev to 7 tev . in some way
, this fit business with no reference to any theoretical model looks more like art than science , especially if no conclusions about the hadron structure are obtained .
such an approach will hardly be conclusive in the future .
theoretical indications of the possibility of a new regime with an increase in transferred momenta were obtained even earlier @xcite .
it was treated as a consequence of the simple iteration of processes approximated by a gaussian within the diffraction cone .
the term @xmath277 in unitarity condition with gaussians inserted into the integrand gives rise to a gaussian with a width , that is twice as big , i.e. to a shape twice as wide as the diffraction cone .
further iterations lead to further widening .
therefore , multiple exchanges were considered .
however , the results did not fit new experimental findings .
this failure was explained as resulting from the improper treatment of the unitarity requirements and incorrect choice of the overlap function .
the droplet model relations between form factors and the elastic amplitude for hadronic scattering at infinite energy ( see eqs ( 1 ) and ( 2 ) in ref .
@xcite ) predict a series of kinks ( or zeros ) in the differential cross section , which could be related to dips . dip position movement to lower @xmath7 with a growth of the total cross section
was predicted in ref .
there is also an indication of several dips ( or shoulders ) at larger @xmath7 in the models @xcite using the electromagnetic form factors with subsequent eikonalization ( cf . figs 2 and 11 ) . in accordance with the experimental data shown in fig .
2 , only one dip is predicted by others . for example , it was described in ref .
@xcite on the basis of a modified optical model @xcite . in the framework of the geometric scaling approach @xcite the numerical integration of the relation @xmath278/(1+\rho _ 0 ^ 2 ) , \label{dk78}\ ] ] where @xmath279 was performed with @xmath183 defined by eq .
( [ ddd ] ) .
it was predicted that the dip should even disappear at energies higher than @xmath280 gev but , probably , can reappear again at ever higher energies . as we know
now , it is clearly seen at 7 tev .
the imaginary part has been chosen in such a way that it has a zero at the dip .
the absence of additional dips is explained as the deviation of the eikonal from a simple gaussian with some flattening at small impact parameters ( see subsection 4.2.3 ) .
that shows strong sensitivity to the choice of tiny details of the phenomenological eikonal and also agrees with the properties of the overlap function to be discussed in more detail below .
these results were confirmed and extended to @xmath4 collisions in ref .
@xcite .
processes described by diagrams with multiple exchange by pomerons are claimed to be responsible for the orear regime at intermediate angles according to ref .
the differential cross section is predicted to have the form @xmath281\cdot [ 1+\lambda \cos(2\sqrt { 2\pi \alpha ' ( 0 ) \vert t\vert \xi \tan(\phi /2)}+\varphi _ 0 ) ] , \nonumber \label{adya}\ ] ] where @xmath282 , and @xmath283 are adjustable parameters .
there are oscillations directly imposed on the exponential fall - off with the same exponent
. they should be well pronounced .
so far , no such oscillations have been observed . a less strong statement about some saturation of the diffraction cone due to multiple pomeron exchanges
is made in refs @xcite .
a theoretical explanation based on the consequences of the unitarity condition in the @xmath30-channel has been proposed in refs @xcite .
the careful fit to experimental data showed good quantitative agreement with experiment @xcite .
nowadays , the same approach helps explain the totem findings @xcite ( see fig . 5 below ) .
we consider the left - hand side and the integral term @xmath277 in unitarity condition ( [ unit ] ) at the angles @xmath22 outside the diffraction peak . because of the sharp fall - off of the amplitude with the angle , the leading contribution to the integral arises from a narrow region around the line @xmath284 .
therefore , one of the amplitudes should be inserted at small angles within the cone as a gaussian , while the other is kept at angles outside it .
integrating over one of the angles yields the linear integral equation : @xmath285 where @xmath286 .
it can be solved analytically ( see @xcite for more details ) with two assumptions : that the role of the overlap function @xmath57 is negligible outside the diffraction cone and the function @xmath287 can be approximated by a constant , i.e. , @xmath288=const .
both assumptions are discussed in the next subsections .
it is esay to check that the eigensolution of this equation is @xmath289 with @xmath290 this expression contains the term exponentially decreasing with @xmath22 ( or @xmath250 ) ( orear regime ! ) with oscillations strongly damped by their own exponential factors imposed on it .
these oscillating terms are responsible for the dip .
just this formula was used in refs @xcite for fits of experimental data in a wide energy range .
the ratio @xmath91 was approximated by its average values in and outside the diffraction cone , with @xmath291 , where @xmath292 is treated as the average value of @xmath91 in the orear region .
the fits at comparatively low energies @xcite are consistent with @xmath293 , i.e. , with small values of @xmath292 close to zero .
the great surprise of the fit in @xcite of totem data shown in fig .
5 was the necessity of using the negative value of @xmath294 large in modulus .
being model - independent , this solution suffers from some limitations that are inherent for the unitarity relation , in general , and for the unitarity equation ( [ linear ] ) , in particular .
first , it predicts the dependence on transferred momenta @xmath295 but not the dependence on the collision energy .
second , it is applicable in a restricted ( and not rigorously defined ) range of angles in the orear region .
the elastic scattering differential cross section outside the diffraction cone ( in the orear regime region ) is @xmath296 it has been used for the fit in fig .
only the very first oscillating term in ( [ solut ] ) is taken into account in this expression , because other terms are more strongly damped with @xmath7 .
it is important that the experimentally measured values of the diffraction cone slope @xmath28 and the total cross section @xmath25 of the same experiment mostly determine the shape of the elastic differential cross section in the orear region of transition from the diffraction peak to large - angle parton scattering .
the value @xmath121 is so close to 1 at 7 tev that the fit is extremely sensitive to @xmath287 because @xmath297 in the first term determines the slope in this region .
therefore , it becomes possible for the first time to estimate the ratio @xmath292 outside the diffraction cone directly from fits of experimental data .
the fit of the differential cross section of elastic proton - proton scattering at @xmath10=7 tev in the region beyond the diffraction peak according to the predictions of the unitarity condition @xcite .
dots - experimental data , line - theoretical approximation . moreover , in footnote 2 in ref .
@xcite it was mentioned that eq .
( [ linear ] ) is in fact an equation for @xmath298 .
the factor @xmath299 was omitted in this review and all previous papers because it was assumed that `` retaining it would exceed the accuracy of the derivation '' of the equation .
however , it will be worthwhile to take it into account in the future as well , multiplying the right - hand side of ( [ fit ] ) by @xmath300 .
this would slightly improve the fit in fig .
we note that this shape of differential cross section ( [ fit ] ) differs from formula ( [ adya ] ) , first of all , because of the suppression of oscillations by the exponential factors in front of them , which decrease much more strongly than the leading exponent . in ( [ adya ] ) ,
the exponent is common for main and oscillating terms , while in ( [ fit ] ) , the oscillations are strongly damped .
they may give rise to the dip adjusted to the diffraction cone if their amplitude is sufficiently large .
the small secondary damped oscillations at larger values of @xmath7 have been seen at comparatively low energies ( see ref .
@xcite ) but have not yet been noticed at the lhc .
we stress that the fit ( [ fit ] ) contains only three adjustable parameters : the overall normalization @xmath301 , the amplitude of oscillations @xmath302 , which determines the depth of the dip , and @xmath287 , which helps find the ratio @xmath292 outside the diffraction peak from the slope of the differential cross section there . both the overlap function and the eikonal are subject to the unitarization procedure , albeit in somewhat different approaches .
therefore , it is instructive to compare their different forms .
we discuss what shapes of the overlap function can be considered as suitable for further use .
one of the assumptions used in solving the unitarity equation was the smallness of @xmath57 in the orear region .
the results in @xcite give strong support to the validity of this assumption . the overlap function
was calculated there directly from experimental data , by subtracting the elastic contribution @xmath277 from the left - hand side of the unitarity equation without appeal to any model .
it is described by the formula : @xmath303^{1/2}k^{-1/2}(z , z_1,z_2 ) .
% \nonumber \label{overl}\end{aligned}\ ] ] here , @xmath304 ; @xmath305^{1/2}$ ] .
the result at 7 tev is shown in fig .
the overlap function at @xmath10=7 tev obtained from the unitarity condition with substitution of experimental data about the differential cross section @xcite . + it is large in the diffraction cone and negligibly small outside it .
the line nearest to the abscissa axis takes into account the real part of the amplitude .
the farthest one is computed with @xmath306 .
certainly , the shadow of inelastic processes represented by the overlap function dominates within the diffraction peak .
but it is extremely small outside .
it is even smaller at the lhc energies @xcite than at lower ones @xcite , where a similar behavior of the overlap function at large @xmath7 was observed previously .
hence , this assumption is well founded .
moreover , it is quite understandable that @xmath73 is very small at large @xmath7 in fig .
this shows that its fit by the solution of the unitarity relation has been done by the proper eigenfunction ( [ solut ] ) with the correct eigenvalues of the integral equation .
it is tempting to solve the nonlinear inhomogeneous unitarity equation ( [ unit ] ) by iterations . that has been attempted several times @xcite .
the main problem is the choice of the overlap function .
the simplest ansatz is the gaussian form at all transferred momenta .
the argument in favor of it is just that it plays the decisive role in the diffraction cone , where the elastic amplitude has a gaussian shape .
bbut the results fail to describe the orear regime .
this may be ascribed to the role of phases of inelastic processes , that determine the genuine shape of the overlap function , or / and to the improper approximation of @xmath91 by a constant outside the diffraction cone .
again , similarly to the situation in the @xmath62-representation , the tiny details of the shape prevent from the proper outcome .
no approximations for the overlap function demonstrated in fig .
6 have yet been proposed .
it is instructive to confront the shape of the overlap function @xmath73 with results obtained in the impact parameter interpretation of proton - proton scattering .
they were presented in refs @xcite for isr data and are demonstrated in figs 7 , 8 . the @xmath62-transformed amplitude @xmath71 , the overlap function @xmath72 and the eikonal @xmath181 are shown in fig . 7 at the energy @xmath307 gev @xcite .
the transformed amplitude is almost gaussian from the center to 2 fm with little flattening near the center .
there is a tail beyond 2 fm with a much flatter slope .
the flattening of the overlap function at the center is much stronger , while the eikonal is steeper there .
hence , one should not identify these three curves at small @xmath62 , even though they almost coincide beyond 2 fm .
the shapes of the amplitude , overlap function and eikonal extracted from experimental data at @xmath307 gev as functions of impact parameter squared ( borrowed from @xcite ) . in the notations of this review , the amplitude @xmath308 , the overlap function @xmath309 , the eikonal @xmath310 .
the corresponding space scales are shown in the abscissa axis .
the overlap functions at isr energies as functions of the impact parameter look similar ( borrowed from @xcite ) .
the solid line at @xmath307 gev is explained in the text .
similar features are seen in fig .
8 , taken from @xcite , where @xmath72 at the same energy is displayed .
the solid line on the logarithmic scale is a gaussian adjusted to fit at @xmath311 and @xmath312 fm .
a gaussian adjusted between 0.6 and 1.6 fm would be higher at @xmath311 and would require additional flattening .
this flattening at small @xmath62 corresponds directly to negative values of @xmath73 at large @xmath7 seen in fig .
5 . in the same way
, slight variations of eikonal @xmath181 at small @xmath62 may lead to drastic disagreement of model fits with experimental data .
therefore , their success or failure at large @xmath7 depends on the accuracy of the chosen form of the eikonal at low @xmath62 . a long tail above the solid line for large impact parameters is clearly seen in fig .
these figures demonstrate how accurate model formulas must be to correctly reproduce either the overlap function or the eikonal if the final goal is to describe the differential cross sections outside the diffraction peak .
a small `` edge '' correction to the gaussian shape of the eikonal has been claimed to be necessary for fits of experimental data at isr energies on increasing total cross sections and structures of the differential cross sections in refs @xcite .
for example , the correction factor @xmath313 with some specific dependence on the impact parameter was introduced @xcite into the overlap function @xmath72 .
it changes the shape at small @xmath62 and makes it similar to that shown in fig .
7 : @xmath314 ) .
\label{fsbk}\ ] ] it turns out that , in the @xmath8-representation , the corresponding overlap function @xmath73 has two zeros at @xmath7=0.645 and 3.83 gev@xmath13 and becomes practically indistinguishable from zero already at @xmath143.5 gev@xmath13 . the last statement is in full agreement with the conclusions in refs @xcite .
although the overlap functions in fig . 8 look quite similar to each other , there is a slight difference , which was analyzed in @xcite .
this difference reveals itself in a small increase at the level of 4@xmath31 of the overlap function , with an energy increase at the impact parameters ( radii ) about 1 fm , which implies the peripheral origin of this phenomenon . that was also discussed earlier @xcite .
moreover , in ref .
@xcite which deals with the direct analysis of experimental data at isr , a shoulder of the overlap function at 2.3 fm was noticed .
its origin is unknown .
the increase in the peripheral region about 16@xmath31 for @xmath315 data is reported in the latest review @xcite .
our recent results ( to be published ) indicate that it is twice larger at lhc .
the overlap function in the @xmath62-representation is used in ref .
@xcite to distinguish between the mechanisms of absorption and reflection with the help of the unitarity equation . in the latter case ,
the differential cross section at large momentum transfers is predicted to be 4 times larger .
the impact parameter picture used in almost all phenomenological models is very helpful for a qualitative description of the process .
however , the forms of the eikonal in the @xmath62-representation turned out to be very approximate . in our opinion , their wide use in most papers dealing with extension to larger angles suffers from this deficiency .
there are some arguments @xcite that the eikonal approximation is only valid for sums of leading terms of the tower diagrams , but it is not correct in general .
it is applicable to almost collinear processes only and does not properly take the separation due to transverse momenta into account .
that is why the quasi - eikonal models were developed where the intermediate states take inelastic diffraction processes into account , in addition to elastic ones . as a result
, formulas like ( [ gpps ] ) , ( [ gpps1 ] ) were proposed .
the eikonal does not properly reproduce the @xmath30-channel cuts of the scattering amplitude due to multiple scattering @xcite . by itself
, it does not guarantee precise unitarization .
moreover , the procedure of unitarity corrections is not well defined , because it can be implemented differently .
the accuracy of unitarity relation ( [ unib ] ) in the @xmath62-representation is also not absolutely clear , as discussed above , while its use is mandatory for interpretation of experimental data .
that is why the model predictions shown in fig . 2 fail to explain the data .
there is a drastic difference between the use of the gaussian shape for the amplitude in the @xmath30-channel unitarity condition and the same shape for the overlap function , as well as its use directly in the @xmath62-representation .
the exponential decrease ( see eq . ( [ diff ] ) ) of the differential cross section in the diffraction cone as a function of @xmath7 ( or gaussian for angles ) is an experimental observation .
it can be used anywhere within the its applicability range , as it was done , for example , in solving eq .
( [ unit ] ) .
hence , this solution is quite successful in fits of experimental data in the orear region .
the same shape can not be used for the @xmath8-dependence of the overlap function , although it plays an important role in the formation of cone behavior .
it is often argued that the fourier transform of the gaussian is a gaussian and therefore this shape can also be used in the @xmath62-representation . while the first part of the statement is correct , the second is wrong .
the tails of the differential cross sections are very sensitive to small @xmath62 .
slight variations of this shape at small impact parameters lead to crucial changes in the behavior of the amplitude at large transferred momenta .
therefore , the predictions shown in fig .
2 , which use the impact parameter profiles close to gaussian ones even in the vicinity of @xmath62=0 , are still successful inside the diffraction cone but completely fail outside it , where central collisions play an important role .
it is very difficult in a particular model to guess the proper decline from the gaussian shape at small impact parameters , which drastically influences the differential cross section at large transferred momenta .
therefore , attempts to use the non - gaussian electromagnetic form factors were of some help in improving the situation , because they are closer in shape to the eikonal demonstrated in figs 7 and 8 .
further progress in this direction is necessary in order to understand the geometric content of the interaction region in ordinary space and time .
nevertheless , it is hardly justified to blame the phenomenological model builders for their failure to predict the behavior of the differential cross sections at large transferred momenta , where it is many orders of magnitude lower than in the diffraction peak .
the great and important task of fits of the energy behavior of total and elastic cross sections , ( @xmath316)-dependence of the differential cross section , and the ratio @xmath91 in a wide interval of energies and transferred momenta can not be accomplished without free parameters and the physical intuition of model builders .
the switch to higher energies allows eliminating corrections due to secondary reggeons and improving the fits .
there is hope of gaining clearer insight into the geometrical picture of hadron interactions .
there are no reasonable arguments to neglect the @xmath8-dependence of the ratio @xmath88 in ( [ rho ] ) or of the phase @xmath42 in ( [ zeta ] ) .
this dependence seems to be important , even inside the diffraction cone , albeit the values of @xmath91 are small there . using formula ( [ rhodit ] ) and assuming that @xmath275 determines mainly the shape of the differential cross section in this region , we find that the real part must vanish at @xmath317 with the @xmath92-dependence of @xmath25 ( [ cst ] ) and @xmath147 ( [ wid ] ) and using relation ( [ sas ] ) , we have @xmath318 and hence @xmath319 at @xmath320 .
the estimates at lhc energies are @xmath321 gev@xmath13 .
notably , they agree with the results obtained in the models in @xcite .
there were several attempts to consider the behavior of @xmath88 at larger transferred momenta in refs .
the main efforts were spent on preventing differential cross sections from vanishing at those values of @xmath8 where the imaginary part of the amplitude is zero in a particular model .
the ratio @xmath273 should be infinite , e.g. , as in the models in refs @xcite .
the number of zeros of the imaginary part is sometimes greater than one .
this is typical in the fraunhofer diffraction or in models with electromagnetic form factors .
therefore , the singularities of @xmath273 appear at different @xmath8 in different models .
the real part of the amplitude fills in these kinks leaving some traces like shoulders or dips in the differential cross sections .
for example , it is predicted in ref @xcite that for @xmath3-scattering at 8 tev such traces appear at @xmath322 gev@xmath13 and at 1.5 gev@xmath13 . in refs @xcite , the dispersion relation between the phase and the modulus of the elastic amplitude considered in refs @xcite was used with some input for the modulus fitted to the experimental data at laboratory energies above 100 gev .
the conclusion was that the real part exhibits a zero in the @xmath8-distribution above 200 gev , which moves away from the forward direction as the energy increases . in ref .
@xcite , the eikonal approximation was used following the proposal in ref .
information about the interference region with a coulomb amplitude similar to that in eq .
( [ inter ] ) was inserted into the total amplitude , with the result @xmath323 . \label{klk}\end{aligned}\ ] ] the @xmath8-dependence of the phase was parameterized with the help of 5 parameters as @xmath324 the results showed that the phase ( related to @xmath91 by ( [ rzet ] ) ) increases from values close to zero at @xmath8=0 to about 0.5 in the interval @xmath325 gev@xmath13 .
this conclusion disagrees with results in refs @xcite , as well as with the arguments presented below .
a more general approach using the @xmath30-channel unitarity condition was developed in ref @xcite .
as explained above , the integral equation for the elastic amplitude is valid in the orear region .
its analytic solution ( [ solut ] ) was first obtained in the approximation where the values of @xmath91 in @xmath287 were replaced by their average values in the diffraction cone and in the orear region .
no zeros of the imaginary part of the amplitude were obtained .
the dips at 7 tev and lower energies were explained as resulting from damped oscillations .
the necessity to introduce large negative values of @xmath91 into the orear region is the main outcome and surprise of the fit in ref .
@xcite . in principle , this could happen if there were zeros of the imaginary part of the amplitude in this region , which would require very large values of @xmath326 near them .
but there seem to be no such zeros there .
we discuss this problem in more detail .
we first recall asymptotic predictions .
it was shown in @xcite that the ratio of the real and imaginary parts of the amplitude can be calculated asymptotically at nonzero transferred momenta @xmath8 as @xmath327 .
\label{rhotau}\ ] ] we consider the leading term of solution ( [ solut ] ) . with the imaginary part of the amplitude in the orear region represented as @xmath328 it is possible to calculate @xmath91 .
the very first approximation was to use the first term of the solution ( [ solut ] ) with average values of @xmath91 both in the diffraction peak ( @xmath329 ) and in the orear region ( @xmath292 ) @xcite .
then the following behavior of @xmath91 was obtained @xmath330 \label{rhot}\ ] ] where @xmath331 we note that @xmath91 passes through zero and changes sign at @xmath332 gev@xmath13 .
this agrees with the general theorem on the change of sign of the real part of the high - energy scattering amplitude , which was first proved in ref .
a similar effect is discussed in ref .
but it is difficult to obtain @xmath333 as an average of ( [ rhot ] ) over the orear region .
moreover , this behavior of an unlimited decrease in @xmath91 with @xmath7 does not look satisfactory .
it can in fact be damped if instead of replacing @xmath91 by @xmath292 in the solution , we differentiate @xmath231 according to ( [ rhotau ] ) , inserting there eq .
( [ ore ] ) , i.e. , the first term in ( [ solut ] ) .
the following differential equation is then obtained @xmath334 here , @xmath335 .
the dependence of @xmath336 can be obtained from fig .
however , one should read @xmath337 on the ordinate axis in place of @xmath273 .
i am sorry for this omission . unfortunately
, the conclusions at the lhc energies become very indefinite , because @xmath127 is very close to zero there .
the only conclusion is that @xmath273 has a single zero at @xmath338 gev@xmath13 , and it steeply changes in the orear region of @xmath339 gev@xmath13 .
the result shown in fig .
9 is another extreme approximation compared to eq.([rhot ] ) .
9 . the ratio of the real to imaginary part of the amplitude obtained from the solution of eq .
( [ dvdx ] ) that follows from the unitarity condition @xcite .
the ordinate should be read as @xmath337 in place of @xmath273 .
i am sorry for this omission . the bold use of this procedure for derivation of eq . ( [ dvdx ] ) with
@xmath273 inserted directly in the solution is , nevertheless , not satisfactory , either .
the two possibilities above should be considered as two extremes for the shapes of @xmath273 .
strictly speaking , the behavior of @xmath273 should be taken into account primarely in the integrand . then , inserting expression ( [ rhotau ] ) in place of @xmath292 in eq .
( [ linear ] ) and integrating by parts , we derive the linear integral equation @xmath340{\rm im}a(y ) \label{nonsymm}\ ] ] with @xmath341 and new variables @xmath342 .
the kernel of this equation is not symmetric .
its solution has not yet been obtained , even numerically . however
, some preliminary asymptotic estimates can be obtained from it @xcite . in the preasymptotic energy region , we obtained @xcite the orear regime @xmath343 with the exponential fall - off of the amplitude as a function of angles .
we , therefore , seek a solution of eq .
( [ nonsymm ] ) in the form @xmath344 .
the gaussian exponential shifts to @xmath345 . replacing it with the @xmath346-function of this argument
, we obtain the equation in finite differences : @xmath347 \phi ( x-\frac { a}{\sqrt { 2b } } ) .
\label{findif}\ ] ] again , we can not solve it directly , but reach an important conclusion about the zeros of the imaginary part of the amplitude .
the expression in the square brackets is equal to zero at @xmath348\approx \frac { 2}{a\rho _ 0 ^ 2}. \label{pt0}\ ] ] with the present - day values of @xmath349 , this zero would appear at extremely large @xmath350 gev .
however , zeros of the imaginary part of the amplitude in the orear region just above the diffraction cone might appear as zeros of @xmath351 itself .
this result does not contradict the above statement about the absence of zeros in the case of small oscillatory terms in the solution of a homogeneous linear integral equation .
moreover , the equation tells us that @xmath351 and , consequently , the imaginary part of the amplitude can have zeros at @xmath352 . on the @xmath23-axis , these zeros would be placed at rather short distances from one another . in the black disk limit @xmath127 tends to 0.5 .
if @xmath91 loses in the competition with @xmath127 within @xmath297 and the argument of the logarithm becomes extremely close to 1 or even less , that would mean the drastic change of the regime in the orear region @xcite .
what the outcome of the competition between decreasing @xmath127 and negative values of @xmath91 will be , poses an interesting problem .
experimental data at higher energies will be able to give the answer .
as we see , the real part of the amplitude can dominate at large transferred momenta according to the unitarity condition .
compared to the imaginary part , it can be large and negative there .
this conclusion contradicts , for example , the results of the models in @xcite with electromagnetic form factors , where the dominance of the imaginary part , on the contrary , is claimed everywhere except the tiny regions near its zeros placed in the orear region , in particular .
this disagrees with the above results .
we must remember , however , that @xmath91 is infinite at these zeros ( see fig .
. an analogous behavior of @xmath91 in the case of a single zero has been predicted in ref .
@xcite at isr energies , as shown in fig .
a similar shape of @xmath91 is obtained in ref .
@xcite at @xmath2291.5 gev@xmath13 , but for the energy @xmath10=8 tev . the real part decreases with @xmath7 .
therefore , the conclusions in different papers about the behavior of the real and imaginary parts of the elastic scattering amplitude are contradictory and require further theoretical studies and new experimental data .
the dependence of the ratio of the real part to imaginary part of the amplitude ( depicted here as @xmath353 ) on transferred momenta obtained in a definite phenomenological fit @xcite of experimental data at @xmath307 gev .
the singularity points out the position where the imaginary part is equal to zero .
we have written two formulae ( [ rhodit ] ) and ( [ rhotau ] ) for the same function @xmath88 . therefore , these two expressions must be identical . equating them , we obtain @xcite the partial differential equation @xmath354 where @xmath355 . as usual , the variables @xmath30 and @xmath7 should be regarded as scaled by the corresponding constant factors @xmath356 and @xmath357 .
( [ partial ] ) can be rewritten as @xmath358 the general solution of eq .
( [ part1 ] ) reveals the scaling law @xmath359 in the asymptotic black - disk limit @xmath360 , we obtain @xmath361 and the solution is @xmath362 this yields the scaling law for @xmath363 which implies a universal scaling dependence on a single variable @xmath364 .
we temporarily neglect the contribution of the real part of the amplitude to the differential cross section .
then the asymptotic scaling law for the differential cross section times @xmath365 should be @xmath366 we note that the additional @xmath365-factor can be replaced by an @xmath30-dependence if absorbed in the argument of the scaling function @xmath367 .
then this formula coincides with that obtained in the geometric scaling approach @xcite .
thus we have proved that the solution of partial differential equation ( [ partial ] ) with properly chosen @xmath368 leads to the results known previously about the geometric scaling . at the same time , eq .
( [ partial ] ) is more general and can be used for different assumptions about @xmath368 . in particular ,
the behavior of the total cross section at present energies is often approximated by formula ( [ kppp ] ) as a sum of a large constant term and another term that increases as some power of energy ( see @xcite for a recent reference ) . in this case ,
@xmath369 , and the equation is @xmath370 its solution is @xmath371 from here , we obtain another universal scaling dependence of the differential cross section on a single variable @xmath372 @xmath373 which could be valid at preasymptotic energies .
it follows from the expressions described above that the energy dependence of the scaling variable is determined by the behavior of the total cross section .
@xmath374 if only the first term in eq .
( [ rhodi ] ) is used .
were this scaling valid , one would be able to predict the shape of the differential cross section at a higher energy once the total cross section is known there .
the preliminary results of work with experimental data at energies from isr to lhc have shown that just this dependence best reproduces the similarity of the shapes of the corresponding lines , even though their normalization differs somewhat .
further studies are necessary
. the above scaling laws must be satisfied for the imaginary part of the amplitude times the factor @xmath375 ( see ( [ scal ] ) ) .
it follows from eq .
( [ rhodit ] ) that the real part satisfies an analogous scaling law albeit with another factor , which differs in the two cases considered above .
this would lead to the scaling violating terms when the contribution of the real part of the amplitude to the differential cross section is taken into account .
the above scaling dependences of the differential cross section are modified as @xmath376 and @xmath377 the violation of scaling laws is different in these cases .
the first law acquires a term with the coefficient depending only on the transferred momentum , while the second law aquires a term with the coefficient that depends both on energy and on the transferred momentum .
this violation of scaling laws must be negligible in the diffraction cone because the squared ratio of the real part to imaginary part which is crucial for the differential cross section is extremely small there
. it would be interesting to learn about the effect of these terms outside the cone , especially in the orear region of transferred momenta .
we note that at small values of their arguments @xmath378 , the scaling functions @xmath379 must be respectively proportional to @xmath380 and to @xmath381 , for the differential cross section to be equal to a constant at @xmath36 .
recent fits of totem data have shown @xcite that the geometric @xmath382-scaling is violated even within the diffraction cone and must be replaced at present energies by approximate @xmath383-scaling .
the energy dependence of high - energy scattering processes at a fixed center - of - mass angle is of special interest .
dimensional scaling laws have been developed as an approach to understanding it .
the large angle scattering is determined by contributions due to central interactions of internal domains inside the colliding particles .
the estimates according to the perturbative qcd become justified due to its asymptotic freedom property .
they depend on the number of constituent fields of the hadrons @xcite . at large @xmath30 and @xmath8 and a fixed ratio @xmath384
, we have @xmath385 where @xmath172 is the total number of fields in @xmath386 that carry a finite fraction of the momentum . assuming the existence of quark constituents , the @xmath102 , fixed-@xmath387 prediction for @xmath3-scattering @xcite is @xmath388 . for the elastic amplitude ,
it is @xmath389 this form can become more complicated for multiple scatterings .
for example , the lowest order graphs for @xmath99 rescatterings @xcite behave as @xmath390 and could become the leading ones .
however , due to higher - order corrections , the resulting behavior could change not so drastically , and the result would be close to the initial estimate ( [ mabr ] ) as shown in ref . @xcite .
further progress beyond the simple quark counting rules was slowed down by complications in calculating the enormous number of feynman diagrams . in parallel
, there were attempts to explain the @xmath169-regime in @xmath3 scattering by dynamical mechanisms with the help of simple feynman graphs .
for protons ( or their subregions ) consisting of three valence quarks , we can assume coherent exchange by gluons @xcite or by the color - neutral pairs of gluons @xcite between them .
the propagators of three gluons and their couplings produce an @xmath391-dependence , and two powers in the denominator are added by kinematical factors .
the general problem of these approaches is the necessity to introduce additional factors in order to preserve both protons in their initial states in large - angle scattering .
the corresponding powers of the qcd coupling constant should be included , of course , which leads to possible ( strong ? ) modifications of the simple power law .
also , the exchange by three pomerons instead of the three pairs of gluons is possible . because three colliding quarks share the total energy of the proton equally ( ? ) , their shares are smaller , and the whole process is farther from the asymptotic regime if treated at the parton level . none of these questions have been quantitatively resolved yet .
we note that the large-@xmath7 behavior of reggeons composed of two reggeized partons ( quarks , gluons ) can be calculated from the bfkl equation @xcite .
the multi - pomeron exchange for hadrons in a state with a minimum number of partons was considered in @xcite .
it was concluded that the differential cross section factores as a product of two @xmath392 representing the probability of finding the initial and final particles in a `` bare '' state and the @xmath393 describing the hard exchange interaction : @xmath394 the first factor describes the contribution of large transverse distances , and the second factor represents the contribution of small ones .
the hard exchange is determined by the pomeron vertices , which are known semiclassically : @xmath395 with @xmath396 , \label{kanc10}\ ] ] where @xmath397 are the numbers of valence quarks in colliding hadrons .
this leads to a @xmath169-behavior for @xmath3 and @xmath398 for @xmath5 .
the quantitative comparison with experimental data is more difficult because of much smaller values of the differential cross sections in this region and , correspondingly , larger error bars .
the new experimental data of the totem collaboration at the lhc about elastic scattering of protons at an energy of 7 tev have revived interest to these processes .
the picture of very short - wavelength hadron collisions has become available , adding to our insight into the spatial structure of colliding particles and providing new intrinsic information pertaining to very short - distance interactions .
the total and elastic cross sections show a stable increase with energy . the share of elastic processes increases .
the differential cross section has very intriguing properties .
the exponential @xmath7-decrease persists at small transferred momenta , analogously to lower - energy data .
but the diffraction cone slope is larger compared with low energies ; it is stable up to transferred momenta @xmath338 gev@xmath13 , then this peak steepens and a dip appears at @xmath167 gev@xmath13 , with a subsequent maximum at @xmath168 gev@xmath13 . at somewhat larger angles ,
the exponential in @xmath12-regime prevails .
it is replaced by the @xmath169-behavior at ever larger transferred momenta @xmath399 gev@xmath13 . at the same time , we are waiting for measurements at extremely small angles in the interference region of coulomb and nuclear amplitudes to gain some knowledge about the real part of the forward scattering amplitude .
it would be extremely interesting to learn its energy behavior and check our predictions from the dispersion relations .
the steeper slopes of the diffraction peak and of the orear region at higher energies , and , correspondingly , their smaller extensions clearly demonstrate that it becomes more and more difficult for a high - energy particle to preserve its identity when scattering at larger transverse momenta .
this increase in the total cross section and , especially , in the share of the elastic cross section , as well as the peculiar change of regimes in the @xmath7-behavior of the differential cross section , require a theoretical interpretation .
short of a complete theory of hadron dynamics , we have to turn our attention to phenomenological models and some rare rigorous theoretical relations .
the region of large transferred momenta became an occam razor for them , as explained above .
the geometric picture of the internal structure of protons and their collisions requires larger disk radii increasing with energy .
their blackness increases as well .
some separate subregions of different sizes and opacity are considered .
the impact parameter approach is decisive in deciphering this structure . at isr energies ,
the increase of the total cross section was attributed to some peripheral regions of nucleons .
it is important to juxtapose these findings with the lhc data .
the approach to the black - disk asymptotic limit has become very interesting .
the proposal of geometrical scaling reducing the number of independent variables is under investigation . at the same time
, the scaling law may happen to be different from the geometrical scaling .
there are many phenomenological models , at our disposal , but it is still difficult to choose any particular one among them .
most of them are quite successful , albeit with many adjustable parameters , in describing the energy behavior of the cross sections and the main bulk of the elastic processes in the diffraction cone , but fail in their predictions outside it
. the dynamical origin of many assumptions is still missing .
the small details of the suspected break at small @xmath8 , of the steepened slope and of possible weak oscillations over a smooth exponential behavior of the diffraction peak are under investigation .
there are predictions of several dips and/or visible oscillatory behavior imposed on the trend of a generally decreasing dependence on @xmath7 , which appear at larger transferred momenta . as an example , in fig .
11 borrowed from @xcite the results of some model predictions for the differential cross section of proton - proton scattering at @xmath400 tev are shown up to quite high values of @xmath401 gev@xmath13 .
they differ significantly , and further accurate experimental data expected to be obtained in 2015 2016 will surely be decisive in the choice of a model ( if any ! ) .
the experience with unsuccessful predictions at 7 tev in the region outside the diffraction cone is not very encouraging .
the problem of the behavior of the real part of the elastic scattering amplitude at nonforward transferred momenta is becoming very important . while the imaginary part of the amplitude dominates at small angles in the diffraction cone , there are indications that just the real part prevails at high transferred momenta .
the unitarity condition indicates some ways to solve this problem . however , there are other approaches with different conclusions .
another important unsolved problem is the behavior of the overlap function .
it certainly dominates in the diffraction cone , but seems to become negligibly small outside it .
the phases of matrix elements of inelastic processes must play an important role in attempts to recover its shape .
however , this presents the extremely difficult theoretical task of modeling them .
999 foley k j et al .
lett . _ * 11 * 425 , 503 ( 1963 ) cocconi g et al .
* 11 * 499 ( 1963 ) krisch a d _ phys .
lett . _ * 11 * 217 ( 1963 ) narayan d s , sarma k v l _ phys .
lett . _ * 5 * 365 ( 1963 ) orear j _ phys .
* 12 * 112 ( 1964 ) ; _ phys .
b _ * 13 * 190 ( 1964 ) baker w f et al .
* 12 * 132 ( 1964 )
cocconi g et al .
b _ * 138 * 165 ( 1965 ) orear j et al . _
b _ * 152 * 1162 ( 1966 ) allaby j v et al . _
b _ * 27 * 49 ( 1968 ) ; * 28 * 67 ( 1968 ) akerlof c w et al . _
* 14 * 2864 ( 1976 ) ayres d s et al .
d _ * 15 * 3105 ( 1977 ) asad z et al .
b _ * 255 * 273 ( 1984 ) amaldi u et al . _
b _ * 44 * 112 ( 1973 ) amendolia s et al .
b _ * 44 * 119 ( 1973 ) bohm a et al . _
lett . b _ * 49 * 491 ( 1974 ) kwak n et al . _
b _ * 58 * 233 ( 1975 ) albrow m g et al .
b _ * 108 * 1 ( 1976 ) hartmann j l et al . _
* 39 * 975 ( 1977 ) conetti s et al .
lett . _ * 41 * 924 ( 1978 ) baksay l et al . _ nucl .
b _ * 141 * 1 ( 1978 ) nagy e et al . _
b _ * 150 * 221 ( 1979 ) amaldi u , schubert k r _ nucl . phys .
b _ * 166 * 301 ( 1980 ) kuznetsov a a et al . _
_ * 33 * 74 ( 1981 ) schiz a et al . _
* 24 * 26 ( 1981 ) fidecaro g et al .
b _ * 105 * 309 ( 1981 ) faissler w et al .
* 23 * 33 ( 1981 ) rubinstein r et al .
d _ * 30 * 1413 ( 1984 ) breakstone a et al .
b _ * 248 * 253 ( 1984 ) eisenhandler e et al . _
b _ * 113 * 1 ( 1976 ) amos n a et al . _
_ * 47 * 1191 ( 1981 ) ; _ nucl .
b _ * 262 * 689 ( 1985 ) ; _ phys . lett .
b _ * 247 * 127 ( 1990 ) ambrosio m et al . _
b _ * 115 * 495 ( 1982 ) sakamoto et al .
b _ * 195 * 1 ( 1982 ) bozzo m et al . _
b _ * 147 * 385 ( 1984 ) ; * 155 * 197 ( 1985 ) breakstone a et al . _ phys . rev . lett . _
* 54 * 2180 ( 1985 ) bernard d i et al .
b _ * 171 * 142 ( 1986 ) ; * 198 * 583 ( 1987 ) augier c et al . _ phys .
b _ * 316 * 498 ( 1993 ) abe f et al . _
* 50 * 5518 ( 1994 ) abazov v m et al .
d0 collaboration arxiv:1206.0687 zotov n p , rusakov s v , tsarev v a _ elem .
nucl . _ * 11 * 1160 ( 1980 ) amaldi u arxiv:1206.3954 http://www.theo.phys.ulg.ac.be/ cudell / data/
antchev g et al .
totem collaboration _ europhys .
_ * 95 * 41001 ( 2011 ) antchev g et al .
totem collaboration _ europhys .
lett . _ * 96 * 21002 ( 2011 ) cheng h , wu t t _ phys .
* 1 * 1069 , 1083 ( 1970 ) gribov v n , migdal a a _ sov .
* 8 * 583 ( 1969 ) bronzan j b , kane g l , sukhatme u p _ phys .
b _ * 49 * 272 ( 1974 ) fischer j , kolar p _ phys .
b _ * 64 * 45 ( 1976 ) ; _ phys .
* 17 * 2168 ( 1978 ) mcdowell s w , martin a _ phys .
b _ * 135 * 960 ( 1964 ) van hove l _
nuovo cimento _ * 28 * 798 ( 1963 ) andreev i v , dremin i m _
zhetf pisma _ * 6 * 810 ( 1967 ) andreev i v , dremin i m _ sov
. j. nucl .
* 8 * 473 ( 1969 ) andreev i v , dremin i m , gramenitskii i m _ nucl .
* 10 * 137 ( 1969 ) dremin i m , nechitailo v a _ phys . rev .
d _ * 85 * 074009 ( 2012 ) adachi m , kotani t _ prog .
* 35 * 485 ( 1966 ) ; * 39 * 785 ( 1968 ) islam m m _ nucl .
b _ * 104 * 511 ( 1976 ) kundrat v , lokajicek m , krupa d _ phys .
b _ * 544 * 132 ( 2002 ) kundrat v , kaspar j , lokajicek m arxiv:0912.1188 ( 2009 ) bethe h _ ann .
phys . _ * 3 * 190 ( 1958 ) rix j , thaler r m _ phys .
d _ * 152 * 1357 ( 1966 ) ; * 172 * 1414 ( 1968 ) soloviev l d _ jetp _ * 22 * 205 ( 1966 ) locher m p _ nucl
. phys . b _ * 2 * 525 ( 1967 ) west g b , yennie d r _ phys . rev
* 172 * 1413 ( 1968 ) cahn r _
c _ * 15 * 253 ( 1982 ) selyugin o v _ phys .
b _ * 333 * 245 ( 1994 ) ; _ phys . rev
. d _ * 60 * 074028 ( 1999 ) ; arxiv:1201.4458 block m m _ phys .
* 54 * 4337 ( 1996 ) ; _ phys .
rept . _ * 436 * 71 ( 2006 ) kundrat v , lokajicek m _ z. phys .
c _ * 63 * 619 ( 1994 ) desgrolard p , jenkovszky l , struminsky b , _ eur .
j. c _ * 11 * 145 ( 1999 ) petrov v a , predazzi e , prokudin a v _ eur
. phys . j. c _ * 28 * 525 ( 2003 ) kopeliovich b z , potashnikova i k , povh b , predazzi e _ phys . rev .
_ * 85 * 507 ( 2000 ) ; _ phys .
d _ * 63 * 054001 ( 2001 ) froissart m _ phys .
rev . _ * 123 * 1053 ( 1961 ) martin a _
il nuovo cimento a _ * 42 * 930 ( 1966 ) martin a _ phys .
d _ * 80 * 065013 ( 2009 ) azimov ya i _ phys .
d _ * 84 * 056012 ( 2011 ) ; arxiv:1204.0984 ; 1208.4304 heisenberg w
_ z. phys . _ * 133 * 65 ( 1952 ) cheng h , wu t t _ phys .
* 24 * 1456 ( 1970 ) cheng h , wu t t _ phys .
b _ * 34 * 647 ( 1971 ) block m , cahn r _ rev .
_ * 57 * 563 ( 1985 ) kinoshita t , _ phys .
d _ * 3 * 2346 ( 1970 ) chou t t , yang c n _ phys .
b _ * 128 * 457 ( 1983 ) bourrely c , soffer j , wu t t _ eur .
phys . j. c _ * 28 * 97 ( 2003 ) islam m m , luddy r j , prokudin a v _ phys
b _ * 605 * 115 ( 2005 ) ; _ int
bf 21 1 ( 2006 ) desgrolard p , giffon m , martynov e , predazzi e _ eur .
phys . j. c _ * 16 * 499 ( 2000 ) golec - biernat k , wsthoff m _ phys . rev .
* 60 * 114023 ( 1999 ) petrov v a , prokudin a v _ eur .
phys . j. c _ * 23 * 135 ( 2002 ) donnachie a , landshoff p v _ nucl . phys .
b _ * 244 * 737 ( 1984 ) block m m , gregores e m , halzen f , pancheri g _ phys .
d _ * 60 * 054024 ( 1999 ) kovner a , wiedemann u a _ phys .
d _ * 66 * 034031 ( 2002 ) troshin s m , tyurin n e arxiv:1111.4454 .
merino c , shabelski yu m arxiv:1204.0769 .
chao a w , yang c n _ phys .
* 2063 ( 1973 ) amos n a et al . _
b _ * 247 * 127 ( 1990 ) cheng h , walker j k , wu t t _ phys
* 9 * 749 ( 1974 ) fagundes d a , menon m j , silva p v r g arxiv:1208.3456 auger collaboration arxiv:1107.4804 block m m , halzen f _ phys .
d _ * 72 * 036006 ( 2005 ) ; * 73 * 054022 ( 2006 ) block m m , halzen f _ phys .
d _ * 83 * 077901 ( 2011 ) ; arxiv:1205.5514 ; 1208.4086 grau a , pacetti s , pancheri g , srivastava y n _ phys .
b _ * 714 * 70 ( 2012 ) okorokov v a _ _ ; arxiv:0907.0951 campos s d , okorokov v a _ int . j. mod . phys .
a _ * 25 * 5333 ( 2010 ) fagundes d a , menon m j _ nucl .
a _ * 880 * 1 ( 2012 ) ; arxiv:1208.0510 dremin i m , nazirov m t _ zhetf pisma _ * 37 * 163 ( 1983 ) henzi r , valin p _ phys .
lett b _ * 132 * 443 ( 1983 ) dremin i m , nechitailo v a _ phys .
d _ * 70 * 034005 ( 2004 ) ; * 84 * 034026 ( 2011 ) islam m , kaspar j , luddy r _ mod .
* 24 * 485 ( 2009 ) jenkovsky l , lengyel a , lontkovskyi d _ int
a _ ( 2012 ) dremin i m , chernavsky d s _ zhetp _ * 38 * 229 ( 1960 ) amati d , cini m , stanghellini a _
nuovo cimento _ * 30 * 193 ( 1963 ) akimov v n et al .
b _ * 14 * 285 ( 1969 ) dremin i m , dunaevsky a m _ phys .
c _ * 18 * 159 ( 1975 ) fukuda h , iso c _ nuovo cim .
_ * 43 * 43 ( 1966 ) zalewski k , van hove l _
nuovo cim .
_ * 46 * 806 ( 1966 )
michejda l _ nucl .
b _ * 4 * 113 ( 1967 ) giffon m , hama y , predazzi e
_ z. phys .
c _ * 19 * 341 ( 1983 ) sopkovich n j _
nuovo cim . _ * 26 * 186 ( 1962 ) dar a , kuyler m , dothan y , nussinov s _ phys .
* 12 * 82 ( 1964 )
durand l , chiu y t _ phys .
* 12 * 399 ( 1964 ) arnold r c _ phys .
rev . _ * 136 * b1388 ( 1964 ) jackson j d _ rev .
* 37 * 484 ( 1965 ) wu
t t , yang c n _ phys .
rev . _ * 137 * b708 ( 1965 ) byers n , yang c n _ phys .
rev . _ * 142 * 976 ( 1966 ) chou t t , yang c n _ phys .
_ * 170 * 1591 ( 1966 ) ; _ phys .
lett . _ * 20 * 1615 ( 1968 ) benecke j , chou t t , yang c n , yen e _ phys .
_ * 188 * 2159 ( 1969 ) chou t t _ phys .
d _ * 11 * 3145 ( 1975 ) chou t t _ phys .
* 19 * 3327 ( 1979 ) chou t t , yang c n _ phys .
b _ * 244 * 113 ( 1990 ) krisch a d _ phys .
* 135 * b1456 ( 1964 ) islam m m _ nuovo cim .
a _ * 48 * 251 ( 1967 ) cheng h , wu t t _ phys .
_ * 22 * 666 ( 1969 ) dias de deus j _ nucl . phys .
b _ * 59 * 231 ( 1973 ) koba z , nielsen h b , olesen p _ nucl
b _ * 40 * 317 ( 1972 ) auberson g , kinoshita t , martin a _ phys .
d _ * 3 * 3185 ( 1971 ) buras a j , dias de deus j _ nucl .
phys . b _ * 71 * 481 ( 1974 ) frahn w e , venter r h _ ann
. phys . _ * 24 * 234 ( 1963 ) frahn w e _ nucl .
phys . a _ * 302 * 267 ( 1978 ) ; * 302 * 281 ( 1978 ) sprung d w l , martorell j _ j. phys .
* 30 * 6525 ( 1997 ) ; * 31 * 8973 ( 1998 ) hwa r c _ phys .
* 22 * 759 , 1593 ( 1980 ) hwa r c , zahir m s _ phys .
* 23 * 2539 ( 1981 ) singh v , roy s m _ phys .
* 1 * 2638 ( 1970 ) uzhinsky v , galoyan a arxiv:1111.4984 brogueira p , dias de deus j _ eur .
j. c _ * 37 * 075006 ( 2010 ) bourrely c , soffer j , wu t t _ phys .
* 19 * 3249 ( 1979 ) bourrely c , soffer j , wu t t _ phys .
_ * 54 * 757 ( 1985 ) bourrely c , soffer j , wu t t _ eur .
j. c _ * 71 * 1601 ( 2011 ) cudell j r , selyugin o v _ phys .
lett . b _ * 662 * 417 ( 2008 ) selyugin o v , teryaev o v _ phys
lett . _ * 102 * 032003 ( 2009 ) ; _ found .
_ * 40 * 1042 ( 2010 ) selyugin o v arxiv:1205.5867 jenkovsky l
_ rivista nuovo cim . _ * 10 * 1 ( 1987 ) kontros j , kontros k , lengyel a arxiv:0006141 fleming h , predazzi e _
nuovo cim .
* 4 * 556 ( 1970 ) giffon m , hama y , predazzi e _
c _ * 25 * 129 ( 1984 ) anselm a , gribov v _ phys . lett.b
_ * 40 * 487 ( 1972 ) fiore r et al .
* 24 * 2551 ( 2009 ) khoze v a , martin a d , ryskin m g _ eur .
phys . j. c _ * 18 * 167 ( 2000 ) bertini m et al . _
rivista nuovo cim . _ * 19 * 1 ( 1996 ) lukaszuk l , nicolescu b _ nuovo cim
* 8 * 405 ( 1973 ) jenkovszky l l , martynov e s , struminsky b v _ phys . lett .
b _ * 249 * 535 ( 1990 ) cudell j r , lengyel a , martynov e _ phys . rev
. d _ * 73 * 034008 ( 2006 ) martynov e _ phys .
d _ * 76 * 074030 ( 2007 ) donnachie a , landshoff p v _ phys .
b _ * 437 * 408 ( 1998 ) donnachie a , landshoff p v arxiv:1112.2485 goloskokov s v , kuleshov s p , selyugin o v _ z. phys .
c _ * 50 * 455 ( 1991 ) donnachie a , landshoff p v _ phys .
b _ * 123 * 345 ( 1983 ) ; _ nucl . phys .
b _ * 267 * 690 ( 1985 ) ; _ phys . lett . b _ * 185 * 403 ( 1987 ) feynman r p , the behavior of hadron collisions at extreme energies , in _ high energy collisions _ , proceedings of the third international conference , stony brook , new york , edited by c. n. yang et al . , pages 237 - 249 ( gordon and breach , new york , 1969 ) .
kuraev e a , lipatov l n , fadin v s _ sov .
. jetp _ * 45 * 199 ( 1977 ) ; balitsky ya ya , lipatov l n _ sov .
_ * 28 * 21 ( 1978 ) gribov l v , levin e m , ryskin m g _ phys .
* 100 * 1 ( 1983 ) jalilian marian j , kovner a , leonidov a , weigert h _ nucl .
b _ * 504 * 415 ( 1997 ) balitskii ya ya _ nucl . phys .
b _ * 463 * 99 ( 1996 ) ; kovchegov yu _ phys . rev .
d _ * 60 * 034008 ( 1999 ) kharzeev d , levin e m _ nucl .
b _ * 578 * 351 ( 2000 ) kopeliovich b z , potashnikova i k , povh b , arxiv:1208.5446 cline d , halzen f , luthe j _ phys
lett . _ * 31 * 491 ( 1973 ) ellis s d , kislinger m b _ phys .
* 9 * 2027 ( 1974 ) gaisser t k , halzen f _ phys .
lett . _ * 54 * 1754 ( 1985 ) afek y , leroy c , margolis b , valin p _ phys .
lett . _ * 45 * 85 ( 1980 ) durand l , pi h _ phys
_ * 58 * 303 ( 1985 ) ; _ phys .
d _ * 40 * 1436 ( 1989 ) margolis b et al .
b _ * 213 * 1615 ( 1988 ) block m m et al . _
d _ * 58 * 017503 ( 1998 ) kaspar j et al .
b _ * 843 * 84 ( 2011 ) martin a d et al . _
j. c _ * 70 * 189 ( 2009 ) ryskin m g , martin a d , khoze v a _ eur .
j. c _ * 71 * 1617 ( 2011 ) ; arxiv:1201.6298 martin a d , khoze v a , ryskin m g arxiv:1202.4966 berger e r , nachtmann o _ eur .
phys . j. c _ * 7 * 499 ( 1999 ) berger e r _ nucl . phys
suppl . _ * 74 * 96 ( 1999 ) dosch h g , simonov yu a _ phys .
b _ * 205 * 339 ( 1988 ) dosch h g , ferreira e , krmer a _ phys
d _ * 50 * 1992 ( 1994 ) bialas a , bzdak a _ phys .
b _ * 649 * 263 ( 2007 ) ; _ phys . rev .
c _ * 77 * 034908 ( 2008 ) nemes f , csrgo t arxiv:1202.2438 carvalho p a s , menon m j _ phys .
d _ * 56 * 7321 ( 1997 ) avila r f , menon m j _ eur .
j. c _ * 54 * 555 ( 2008 ) fagundes d a , menon m j , silva g l p _ eur .
j. c _ * 71 * 1637 ( 2011 ) troshin s m , tyurin n v arxiv:1203.5137 philips r j n , barger v d _ phys .
b _ * 46 * 412 ( 1973 ) fagundes d a , menon m j , silva p v r g arxiv:1204.5646 cottingham w n , peierls r f _ phys .
b _ * 137 * 147 ( 1965 ) chou t t , yang c n _ phys
lett . _ * 20 * 1213 ( 1968 ) chou t t , yang c n _ phys .
lett . _ * 46 * 764 ( 1981 ) olsen s l _ z. phys . c _ * 13 * 215 ( 1982 ) dias de deus j , kroll p _ acta phys .
polonica b _ * 9 * 157 ( 1978 ) fischer j , jakes p , novak m _ acta phys .
polonica b _ * 14 * 807 ( 1983 ) anselm a a , dyatlov i t _ phys . lett .
b _ * 24 * 479 ( 1967 ) ; _ yad . fiz . _
* 6 * 591 ( 1967 ) ( _ sov . j. nucl .
phys . _ * 6 * 430 ( 1968 ) ) khoze v a , martin a d , ryskin m g _ eur .
phys . j. c _ * 18 * 167 ( 2000 ) gotsman e , levin e m , maor u _ phys . lett .
b _ * 353 * 526 ( 1995 ) andreev i v , dremin
i m , steinberg d n _ sov .
phys . _ * 11 * 261 ( 1970 ) henyey f s , tuan r h , kane g l _ nucl .
b _ * 70 * 445 ( 1974 ) amaldi u , schubert k r _ nucl
b _ * 166 * 301 ( 1980 ) henzi r , valin p _ phys . lett .
b _ * 48 * 119 ( 1974 ) henzi r _ nucl
b _ * 104 * 52 ( 1976 ) henzi r , valin p _ nucl
b _ * 148 * 513 ( 1979 ) henzi r , margolis b , valin p _ phys
lett . _ * 32 * 1077 ( 1974 ) troshin s m , tyurin n e arxiv:1111.4454 landshoff p v , polkinghorne j c _ phys .
rev . _ * 181 * 1989 ( 1969 ) kroll p _ nucl .
b _ * 82 * 510 ( 1974 ) grein w , guigas r , kroll p _ nucl
b _ * 89 * 93 ( 1975 ) kundrat v , lokajicek m v , krupa d _ phys
d _ * 35 * 1719 ( 1987 ) dremin i m _ jetp lett .
_ * 96 * 277 ( 2012 ) odorico r _
nuovo cim .
a _ * 54 * 96 ( 1968 ) alvarez - estrada r _ ann . of phys . _
* 68 * 196 ( 1971 ) cahn r
_ z. phys .
c _ * 15 * 253 ( 1982 ) martin a _ lett .
nuovo cim . _ * 7 * 811 ( 1973 ) dremin i m _ nucl
a _ * 888 * 1 ( 2012 ) dremin i m , radovskaya a a _ epl _ * 100 * 61001 ( 2012 ) , arxiv:1209.1935 dremin i m , nechitailo v a _ phys .
b _ * 720 * 177 ( 2013 ) dias de deus j , kroll p _ j. phys .
g _ * 9 * l81 ( 1983 ) matveev v a , muradyan r m , tavkhelidze a n _ lett .
nuovo cim . _ * 7 * 719 ( 1973 ) brodsky s j , farrar g r _ phys .
lett . _ * 31 * 1153 ( 1973 ) ; _ phys .
d _ * 11 * 1309 ( 1974 ) landshoff p v _ phys .
d _ * 10 * 1024 ( 1974 ) mueller a h _ phys .
rep . _ * 73 * 237 ( 1981 ) donnachie a , landshoff p v _ z. phys . c _ * 2 * 55 ( 1979 ) donnachie a , landshoff p v _ z. phys .
c _ * * ( 1996 ) dremin i m , nazirov m t _ yad .
fiz . _ * 31 * 1606 ( 1980 ) sotiropoulos m g , sterman g _ nucl .
b _ * 419 * 59 ( 1994 ) ; * 425 * 489 ( 1994 ) kwiecinski j _ phys
d _ * 26 * 3293 ( 1982 ) kirschner r , lipatov l n _ z. phys .
c _ * 45 * 477 ( 1990 ) kancheli o v arxiv:1012.5385
|
colliding high energy hadrons either produce new particles or scatter elastically with their quantum numbers conserved and no other particles produced .
we consider the latter case here .
although inelastic processes dominate at high energies , elastic scattering contributes considerably ( 18 - 25@xmath0 ) to the total cross section .
its share first decreases and then increases at higher energies .
small - angle scattering prevails at all energies .
some characteristic features are seen that provide informationon the geometrical structure of the colliding particles and the relevant dynamical mechanisms .
the steep gaussian peak at small angles is followed by the exponential ( orear ) regime with some shoulders and dips , and then by a power - law drop .
results from various theoretical approaches are compared with experimental data .
phenomenological models claiming to describe this process are reviewed .
the unitarity condition predicts an exponential fall for the differential cross section with an additional substructure to occur exactly between the low momentum transfer diffraction cone and a power - law , hard parton scattering regime under high momentum transfer .
data on the interference of the coulomb and nuclear parts of amplitudes at extremely small angles provide the value of the real part of the forward scattering nuclear amplitude .
the real part of the elastic scattering amplitude and the contribution of inelastic processes to the imaginary part of this amplitude ( the so - called overlap function ) at nonforward transferred momenta are also discussed .
problems related to the scaling behavior of the differential cross section are considered .
the power - law regime at highest momentum transfer is briefly described .
* elastic scattering of hadrons * if only you knew what trash gives rise + to verses that are nt ashamed uprise + a. akhmatova + @xmath1 * 1 introduction . * * 2 the main relations . * * 3 where do we stand now ? * * 4 experimental data and phenomenological models . *
* 4.1 the diffraction cone and geometrical approach .
* * 4.1.1 the geometry of the internal hadron structure . * * 4.1.2 the modified fraunhofer diffraction . * * 4.1.3 electromagnetic analogies .
* * 4.1.4 reggeon exchanges . * * 4.1.5 the qcd - inspired models . *
* 4.2 the intermediate angles : the dip and the orear regime . * * 4.2.1 the gaussian fits . * * 4.2.2 the phenomenological models . * * 4.2.3 the unitarity condition . *
* 4.2.4 the overlap function and the eikonal .
* * 4.2.5 the real part of the elastic scattering amplitude at nonzero transferred momenta . * * 4.3 scaling laws . *
* 4.4 the hard scattering at large angles .
* * 4.4.1 the dimensional counting .
* * 4.4.2 coherent scattering . * * 5 discussion and conclusions .
* * references . * pacs numbers : 13.85.dz elastic scattering ; 13.75.cs nucleon - nucleon interactions key words : cross section , amplitude , slope , diffraction , orear regime , partons
|
the aharon - vaidman game @xcite is a conceptually simple example of how quantum mechanics can be both beneficial and counter - intuitive . in the classical analog ,
alice puts a particle in one of three boxes such that when bob , who in next turn is allowed to check only two of them , is most likely to find it .
alice who can either accept or not accept a particular game trial wins whenever she accepts a trial in which bob has also discovered the particle .
hence , it is obvious that alice will not use the box that bob does not have access to and therefore her chance to win is @xmath0 .
however , the result of the game can be totally different when the particle is described by laws of quantum mechanics and alice prepares it in equal superposition of being in each of the boxes .
then her chance to win can reach @xmath1 if she takes a specific projective measurement after bob s turn . in this article we present an experimental realization of the av quantum game using a single photon as the incident particle and a system of three slits in lieu of the boxes .
the original three box paradox was proposed by aharonov and vaidman in ref .
the quantum game @xcite was conceived much later and exhibits a clear quantum advantage if the game rules are followed with care .
the setup comprising of a single photon source ( heralded parametric down conversion source or attenuated laser ) , triple slit @xcite and single photon detectors allowed us to perform optimized quantum tomography to characterize the qutrit states and to play the game correctly in the next step . some quantum communication protocols that can be considered as quantum games such ascoin tossing @xcite and the byzantine agreement @xcite
have been demonstrated previously
. however , the av game is a specific example of a quantum game where one can demonstrate the quantum advantage playing the game with a single particle at a time in contrast to the entanglement based games @xcite .
we will begin with the introduction of the concept of the qutrit encoded in spatial degrees of freedom of a single photon @xcite . in the next step
we discuss the av quantum game @xcite and its experimental implementation .
presented implementation of the game is conceptually similar to the young experiment , where multilevel quantum systems can be encoded in paths related to a photon passing through the slits . recently
this type of quantum state encoding has drawn much interest .
in particular , taguchi _ et .
used parametric down conversion source to prepare two qubit @xcite and two qutrit @xcite states . in turn ,
lima _ et . al .
_ in ref .
@xcite demonstrated the seven and eight dimensional state encoding .
alternative approaches resort to state encoding in various hybrid ways such as energy - time @xcite or polarization - orbital angular momentum @xcite .
those implementations were useful to perform a bell test for energy - time entangled qutrits @xcite and to demonstrate the optimal cloning strategy @xcite , respectively .
more recently , the noncontextuality of quantum mechanics was tested based on similar scheme @xcite .
the young type qutrit is realized using triple slits and a single photon source(sps ) .
the photon s initial spatial mode is gaussian with the characteristic diameter much larger than the size of the slits and with the peak intensity coincident with the slit area , see inset in fig .
[ fig : experiment ] . under these conditions
one can consider the state of a photon of wavelength @xmath2 in the position @xmath3 to be a plane wave @xmath4 propagating in direction given by the wave vector @xmath5 of length @xmath6 .
moreover the photon s initial propagation direction is assumed to be paraxial and the distance between the slits larger than their characteristics widths .
this allows us to approximate the phase to be constant at each of the slit .
hence , the spatial wave function of the photon passing through the @xmath7th slit can be written as @xmath8 , where @xmath9 , and @xmath10 stands for the transmission probability amplitude , which we assume is constant on the slit and @xmath11 elsewhere .
this means that the total wave function of the photon passing through three slits comprises of three orthogonal contributions .
each of them can be written in momentum representation as @xcite : @xmath12 , where @xmath13 , @xmath14 is the slit width and @xmath15 is the distance between the slits .
these definitions allow us to write the state of the transmitted photon as : @xmath16 which accounts for the basic definition of a young type qutrit .
here amplitudes @xmath17 , @xmath18 and @xmath19 depend on the transmission functions @xmath10 .
the projective measurements are determined by the laws of propagation and the geometry of the setup . for simplicity
, we chose to detect in the positions corresponding to near and far field .
this can be done using a lens and placing a detector in the focal plane ( far field ) and in the plane where the image of the slits is formed ( near field ) . in the near field , if the active area of the detector is larger than the image of each slit , the probability to detect a photon prepared in the state @xmath20 as defined above in the position corresponding to @xmath7th slit image is proportional to @xmath21 .
hence it is easy to see that each of the three positions can be associated with the measurement operator defined as @xmath22 where @xmath23 is the normalization factor to be specified later and subscript nf stands for near field .
the interpretation of measurements in the far field needs more attention . a detection in the position @xmath24 in the focal plane corresponds to the projector onto @xmath25 , which is related to the plane wave propagating in the direction given by the transverse wave vector @xmath26 .
hence the probability to detect a photon @xmath27 can be seen as proportional to @xmath28 , where we introduced @xmath29 .
based on this observation we can define the measurement operator in the far field as : @xmath30 where @xmath31 is the normalization factor , the phase parameter reads as @xmath32 and the subscript ff stands for far field . the measurement operators @xmath33 and @xmath34 can be used to construct rank 7 positive value operator measure ( povm ) set allowing one for reconstruction of arbitrary pure state . for this reason
we take three near field measurements @xmath35 , @xmath9 and six far field operators @xmath36 corresponding to @xmath37 .
this specific choice requires renormalization , which can be done when @xmath38 .
nm ) , laser power controller ( lpc ) and neutral filter ( nf ) .
the pdc - sps ( @xmath39 nm ) is based on ppktp crystal pumped by blue continuous wave laser .
heralding photon is detected by detector d3 .
the single photons from both sources are coupled to single mode fibers ( smf ) .
a qutrit is prepared using the blocking mask and three slits .
next the measurement part of the setup comprises of a 2 inch diameter @xmath40 mm lens ( l3 ) , 2 inch diameter pellicle beamsplitter ( bs ) , color filters ( f ) and two detection systems for far ( d1 ) and near filed ( d2 ) , each comprised of multimode fiber mounted on a precise motorized stage ( thorlabs zst13 ) and a perkin elemer avalanche photodiode . ]
the ability to encode and measure qutrit states can be utilized to demonstrate aharon and vaidman s quantum game @xcite .
a classical strategy allows alice for at most @xmath0 chance to win . on the other hand , when she uses quantum particles her chance rises above this limit and ideally reaches @xmath1 , when she chooses her initial state to be @xmath41 in the first turn of the game . in the second turn
, assuming bob has access to slits ( boxes ) @xmath11 and @xmath42 , if he decides to check if the photon is passing though slit number @xmath11 , he does a projective measurement on the state @xmath43 .
if he finds the particle , then his state becomes @xmath44 , otherwise @xmath45 .
the photon detection results in losing the photon from the system , otherwise the photon goes through opened slits .
this can be simulated by blocking slit number 0 , which will allow us to simulate all those cases when bob s detector did not click . on the other hand ,
the case of finding a photon in slit number 0 can be simulated by closing all others .
next , in the third turn of the game alice makes a projective measurement on @xmath46 .
this can be done by placing the detector d1 in the far field plane in the position corresponding to povm element @xmath47 .
if alice detects a particle , she accepts the game trial , and if she does not , she cancels it .
now it is clear that alice can not lose as whenever bob does not detect a photon , the state after the second turn is @xmath48 and alice s detector never clicks as @xmath49 . if bob found the particle in a slit @xmath11 and tried to leave no trace of that , alice has @xmath50 chance to detect it after that .
the same reasoning holds if bob chooses the slit number @xmath42 .
the experimental setup is depicted in fig . [ fig : experiment ] .
we used two single photon sources : heralded parametric downconversion ( pdc ) source based on periodically polled potassium titanyl phosphate ( ppktp ) crystal ( pdc - sps ) and attenuated hene laser ( al - sps ) . in order to fulfill the assumption of the plane wave incidence at the slits we used single mode fibers ( smfs ) and optics to set the characteristic spatial mode diameter to approximately @xmath51 mm .
we control the state of the qutrit using blocking mask ( b ) to change the configuration of opened slits and tilting slightly the mirror ( m ) to change the incidence angle @xmath52 . under these simplifying assumptions
the experimentally possible states are in the following form @xmath53 .
the far and near field measurements were implemented by photon counting in the transverse planes at distances of @xmath54 and @xmath55 mm , respectively .
large @xmath42 inch pellicle beamsplitter ( bs ) was introduced to reduce the disturbance of the setup while changing the planes of detection .
each detector system ( d1,d2 ) was comprised of a multimode fiber mounted on precise motorized stage and perkin elemer avalanche photodiode .
step motors were used to control the transverse position of the fiber with an accuracy of @xmath56 m .
counts were registered by the field programmable gate array ( fpga ) logic system . before simulating the av game and characterizing prepared states
the setup was calibrated .
it was done by opening all slits and setting the initial direction of photon propagation to @xmath57 , which corresponded to preparing the state @xmath58 .
next we measured the photon count rates as a function of the detector position in far- and near - field planes .
the results together with the best fits are presented as insets in fig .
[ fig : experiment ] .
blue ( online ) dots represent experimental data , while the continuous line is a theoretical fit .
the positions corresponding to the far-(near- ) field part of povm are marked with bigger red dots on inset next to detector d1(d2 ) .
note that the smoothed shape of the slits image ( near field ) is attributed to the finite size of a detector . .
photon count probabilities were measured in the positions related to projective measurements @xmath34 and @xmath33 for three typical states @xmath59 , @xmath60 , @xmath61 . [ cols="^,^,^,^,^",options="header " , ] we characterized the prepared qutrit , which has been done resorting to povm set described earlier and quantum state tomography methods . for this reason
the photon counts were measured by placing detectors in positions related to measurements operators @xmath34 and @xmath33 .
those positions are marked with red dots on inset plots in fig .
[ fig : experiment ] . in order to justify the results of the quantum game we took three typical states : @xmath62 , @xmath63 , @xmath64 .
for the first state all slits were open , for the second one a slit number 0 was closed and for the last state the propagation direction of the photon was modified in order to introduce a phase @xmath65 .
these measurements were done using al - sps , its outcomes are gathered in table [ tab : counts ] and the results of tomographic reconstruction using the maximal likelihood method are depicted in fig .
[ fig : density ] .
it is seen in fig .
[ fig : density](a , b ) that for the states @xmath66 and @xmath67 the real part dominates as there is no phase present . on the other hand
changing the initial photon direction @xmath52 it was possible to introduce the phase as is apparent in fig .
[ fig : density](c ) as the imaginary bars are significant . ideally , the imaginary part of the density matrix for states @xmath66 and @xmath67 is zero . here
, the nonzero height is attributed to the noise originating from dark counts , stray light and imperfect positioning . .
left ( right ) column depicts real ( imaginary ) part of the reconstructed density matrix .
the reconstructed phase related to state @xmath68 was @xmath69 . ] for the quantum game the qutrit was prepared in @xmath70 , which was characterized before , see fig .
[ fig : density](a ) .
we simulated all possible scenarios of bob s measurement using pdc - sps and al - sps .
the measured photon counts are presented in table as an inset of fig .
[ fig : qg ] . in the perfect case one expects no counts when two slits are open and alice sets her detector in the far field plane in the position related to @xmath47 .
this corresponds to the first local maximum marked with red dot on plot related to far field inserted in fig .
[ fig : experiment ] . here , the measured counts are attributed to the finite size of the multimode fiber core , dark counts and stray light .
we estimate that the former two contribute approximately 3 coincidence counts per 2 seconds . despite the background noise from
stray light and dark counts , and the slight non - ideal properties of alice s measurements , she won in 87% of the accepted trials using pdc - sps and in 82% of the accepted trials using al - sps .
note that the overall efficiency of the game , which is due to experimental deficiencies including the photon collection , the detection efficiencies and number of alice s detectors , will only limit the number of accepted trials , but not the percentage of winning trials .
s ( @xmath71 min , coincidence window @xmath71 ns ) . ]
in conclusion , we experimentally presented a simple way to implement a qutrit system into a single photon s spatial degree of freedom , which allowed us to perform state tomography and simulate the av quantum game .
the encoding part resorted to the young type experiment , where a photon passes through 3 slits , which defined its state . by controlling an initial propagation direction of a photon and configuration of the slits it was possible to encode a certain class of states .
our state reconstruction technique was based on a small number of measurements over a short period of time , which makes our method stable and time efficient in contrast to ref .
the authors acknowledge insightful discussions with c. fuchs , m. graydon and g. noel tabia from perimeter institute and funding from nserc ( cgs , quantumworks , discovery , usra ) , ontario ministry of research and innovation ( era program ) , cifar , industry canada and the cfi .
pk acknowledges fruitful discussions with r. demkowicz - dobrzanski from warsaw university , support by the foundation for polish science team project cofinanced by the eu european regional development fund and the mobility plus project financed by polish ministry of science and higher education .
ac acknowledges support by micinn project no .
fis2008 - 05596 and the wenner - gren foundation .
|
the aharon - vaidman ( av ) game exemplifies the advantage of using simple quantum systems to outperform classical strategies .
we present an experimental test of this advantage by using a three - state quantum system ( qutrit ) encoded in a spatial mode of a single photon passing through three slits .
the preparation of a particular state is controlled as the photon propagates through the slits by varying the number of open slits and their respective phases .
the measurements are achieved by placing detectors in the specific positions in the near and far fields after the slits .
this set of tools allowed us to perform tomographic reconstructions of generalized qutrit states , and to implement the quantum version of the av game with compelling evidence of the quantum advantage .
|
a subgame perfect equilibrium is one of the fundamental solution concepts in game theory .
its characterization is very simple : a strategy profile is a subgame perfect equilibrium if it constitutes a nash equilibrium of every subgame of the original game . in extensive form games with perfect information ,
subgame perfect equilibrium coincides with backward - induction and can be easily determined . in this work
we study the distribution of the subgame perfect equilibrium outcome in random two - player extensive form games . in these games there
are two potential natural degrees of randomness that may be considered .
the first is random payoffs where the payoffs at the leaves of the tree are randomly drawn .
the second is random assignment of players to nodes , where one of the two decision makers at every node is drawn at random in accordance with a certain distribution .
we restrict attention to binary trees with two players and focus on the case where the payoffs are i.i.d .
uniform random draws over a given domain @xmath0 .
our analysis combines the two potential sources of randomness and relates properties of the subgame perfect equilibrium outcome to a solution concept in the theory of bargaining .
the existing literature on random games focuses on normal form games and studies properties such as the expected number of nash equilibrium @xcite , the distribution of pure nash equilibria @xcite or the maximal number of equilibria @xcite .
here we restrict our attention to random extensive form games .
we consider binary random games and study the asymptotic properties of the subgame perfect equilibrium outcome when the depth is increasing .
the fundamental natural question that arises is under which condition the limit exists and whether in some sense the subgame perfect distribution _ concentrates _ around a certain point , i.e. , whether the subgame perfect equilibrium outcome is close to a certain point with probability close to 1 , for sufficiently deep trees . such a concentration point ,
if exists , may serve as a benchmark solution for a bargaining problem with @xmath0 as its feasible set .
indeed this point has quite solid strategic justification : for _ _ almost all _ _ for the correct dimension @xmath1 ) .
note that analyzing probability under i.i.d .
uniform distribution is equivalent to analyzing the lebesgue measure . ]
strategic interactions of a quite natural type with outcomes in @xmath0 , the solution of the interaction will be close to this concentration point , i.e. , for large enough perfect information games over binary trees ( with a particular assignment of players to control the nodes ) and with payoffs in @xmath0 . hence we are interested in characterizing the cases where concentration occurs , and these cases will induce interesting incites into bargaining theory . our main theorem introduces the _ random extensive form _ bargaining solution that naturally arises from random extensive form games ( see section [ section : atb ] ) .
we characterize our solution concept by introducing a sequence of increasing random extensive form games for which concentration around a point holds .
this point defines our solution concept and satisfies the three core axioms of a standard solution concept in bargaining : symmetry , efficiency and scale invariance . starting with nash @xcite in 1950
, the bargaining problem has been widely studied from different perspectives .
two complementary trusts of the bargaining theory are the _
axiomatic approach _ and _ implementation theory_. the axiomatic approach aims to understand what should be the solution of a bargaining problem by introducing basic properties that it should satisfy .
however , this approach provides no strategic reasoning for how this solution arises .
the so - called nash program " aims to support solutions in a non - cooperative framework .
this is exactly where implementation theory comes to play , where the goal is to design a game ( i.e. , a bargaining mechanism ) whose equilibrium is the solution .
many axioms have been suggested in the context of bargaining . among the suggested axioms , the _ standard
_ axioms that are satisfied by most of the classical solutions ( e.g. , nash solution @xcite , kalai - smorodinski solution @xcite , and others ) are pareto efficiency , symmetry , and invariance with respect to positive affine transformations .
the terminology of kalai - smorodinsky @xcite ( see also @xcite ) , defines a solution to be _ standard _ if it satisfies these three axioms
. our focus will be on implementing a standard solution .
implementation theory in bargaining , and in particular implementation in subgame perfect equilibrium , has been widely studied .
several mechanisms @xcite have been suggested for implementing standard bargaining solutions .
all these mechanisms are based on the idea that players sequentially offer an allocation that should be approved by the other players .
we take a different approach to the implementation problem , and we pose the question : can a random extensive form game serve as an implementation to a standard bargaining solution ?
theorem [ theo : bar ] constitutes a positive answer to this question .
this approach differs from the above @xcite in the following aspects .
first , the designed game is _ not _ based on offers and approvals .
second , it has minimalistic requirements about the abilities of the designer : it is sufficient that the designer will be able to draw outcomes from the bargaining set uniformly at random ( as outcomes of the game ) in order to design a game whose equilibrium outcome is ( close to ) a standard solution .
third , our result , in some sense , answers the following type of questions : how hard is it to implement a standard solution ?
how structured should the designed game be ?
our result can be interpreted as follows , _ almost all _ binary choice large enough extensive form games ( with a particular random structure of the decision maker over the nodes ) implement a standard bargaining solution . for ease of clarity of the presentation , throughout the paper
we focus on a specific simple setting .
as shown in comment 1 of section [ sec : discussion ] , our analysis holds in much more general settings .
we let @xmath2 be a compact convex set is not essential ; see comment 1 of section [ sec : discussion ] . ] .
let @xmath3 be the uniform distribution over with positive density over @xmath0 ; see comment 1 of section [ sec : discussion ] . ]
the game is played over a complete binary tree of depth @xmath4 .
the @xmath5 payoffs at the leaves are random independent draws in accordance with @xmath3 .
we consider two benchmark models of assigning players to control nodes in the tree : 1 . _ the alternating game .
_ player 1 controls all nodes at depths @xmath6 player 2 controls all nodes at depths @xmath7 .
such a game ( with random payoffs ) will be denoted by @xmath8 .
2 . _ the random controlling player game . _
each player @xmath9 controls each node with probability @xmath10 , where the controlling player at each node is drawn independently across the nodes and independently of the payoffs . such a game ( with random controlling players and random payoffs )
will be denoted by @xmath11 . in section
[ sec : bar ] we consider additional assignments of players to nodes , which combines these two models . for both players @xmath9 , both random games @xmath8 and
@xmath11 have no two equal payoffs for player @xmath12 with probability one .
therefore , both @xmath8 and @xmath11 have almost surely unique subgame perfect equilibrium .
we denote by @xmath13 and @xmath14 the random variable of the _ games value _ , which is the payoff profile at the unique subgame perfect equilibrium .
we are interested in the asymptotic properties of the values @xmath13 and @xmath14 when @xmath15 .
in particular are interested in the phenomena of _ concentration of the value _ which holds if there exists a point @xmath16 such that the probability that @xmath17 lies in a given neighborhood of @xmath18 goes to one as the depth of the tree goes to infinity .
formally , for a sequence of random games @xmath19 , we say that @xmath17 _ concentrates around the point _
@xmath20 if the sequence @xmath17 weakly converges to the dirac measure on the singleton @xmath18 , i.e. , @xmath21 .
we start our analysis with the zero - sum case , namely @xmath22 .
the results for random zero - sum games are as follows : * for the alternating game , @xmath13 concentrates around the point @xmath23 , where @xmath24 ( theorem [ th : zs ] ) .
* for the random controlling player model , however , we do not have a concentration result .
in fact , we have @xmath25 , i.e. , the value is distributed according to the distribution of the initial payoffs for all @xmath4 ( proposition [ pro : zs ] )
. then we turn to the non - zero sum case where @xmath0 is a convex set with non - empty interior .
the results for this case are as follows : * for the alternating game , @xmath13 concentrates around a pareto efficient point ( theorem [ th : gd ] ) .
* for the random controlling player game , although we do not have a concentration around a point , we do have a concentration of @xmath14 around the pareto efficient boundary ( proposition [ pro : gen ] ) .
our next target is to examine whether random extensive form games can serve as an implementation to a bargaining solution .
the set @xmath0 can be considered as the feasible set of alternatives .
a solution concept assigns a unique point for every such feasible set .
therefore , the first property that should be satisfied for such a random game is 1 .
concentration of the value around a point .
another two natural properties that are common to most solution concepts in bargaining ( the _ standard axioms _ ) are : 1 .
pareto efficiency : the unique solution should be pareto efficient .
2 . symmetry : for every symmetric sets @xmath0 the solution should be symmetric . by considering the results mentioned so far , we see that none of the suggested random extensive form games enjoys all three properties .
the alternating game concentrates around a pareto efficient point ( theorem [ th : gd ] ) .
however , the concentration point does not satisfy symmetry ( theorem [ th : zs ] ) ; player 1 who acts last has an advantage .
the random controlling player game is symmetric by nature ; however it fails to satisfy concentration ( proposition [ pro : zs ] ) .
it turns out that a random extensive form game , which satisfies all the above desirable properties can be obtained by a hybrid " of the two discussed cases ( the alternating and the random ) . in the _ hybrid game
_ all nodes at depth @xmath26 for odd @xmath1 are controlled by player @xmath27 with probability @xmath28 and by player @xmath29 with the remaining probability @xmath30 .
all nodes at depth @xmath26 for even @xmath1 are controlled by player @xmath29 with probability @xmath28 and by player @xmath27 with probability @xmath30 .
such a game ( with random controlling players and random payoffs ) will be denoted by @xmath31 .
note that @xmath32 and @xmath33 .
we show in theorem [ theo : bar ] that for small values of @xmath34 the hybrid game @xmath31 , indeed ( approximately ) satisfies the above - mentioned desirable property : 1 . for every fixed @xmath35
, the value of @xmath31 concentrates around a point .
2 . for every fixed @xmath35 ,
the concentration point of @xmath36 is pareto efficient .
3 . for a symmetric set @xmath0 , given @xmath37 we can set @xmath35 such that the concentration point of @xmath36 will be @xmath38-close to the diagonal ( i.e. , @xmath38-close to the symmetric solution ) .
we first consider the case of random zero - sum games , where @xmath39 , i.e. , player 1 s payoff is uniformly distributed in @xmath40 $ ] . for convenience of notations , along this section we drop the payoff of player 2 , which is equal to minus player s 1 payoff .
the following theorem states that the value of the alternating zero - sum game converges to the golden ratio minus one .
[ th : zs ] @xmath13 concentrates around point @xmath41 where @xmath42 .
we denote @xmath43 . by the backward induction procedure we can deduce a recursive formula for the sequence of measures @xmath44 .
the value of a game of depth @xmath45 is a function of the values of the two subgames of depth @xmath4 ( which correspond to the two possible actions of the player that controls the root ) .
note that the two values of the two subgames are two independent draws according to @xmath46 .
we denote by @xmath47 and @xmath48 the minimum and the maximum of two independent draws according to @xmath49 , and we have the following recursive formula @xmath50 because player 1 will choose the maximal value ( among the two ) , and player 2 will choose the minimal . for @xmath51 $ ]
we denote @xmath52)$ ] .
for @xmath53 we have @xmath54 because the maximum of two independent random variables is below @xmath55 iff both of them are below @xmath55 .
note also that @xmath56 the first equality follows from the inclusion - exclusion principle .
the second inequality follows from equation .
let @xmath57 , and let @xmath58 be @xmath1 times composition of @xmath59 on itself , i.e. , @xmath60 it readily follows by induction from equation that @xmath61 note that for @xmath62 , @xmath63 is a monotonically increasing polynomial with three fixed points : @xmath64 .
in addition , @xmath65 for @xmath66 , and @xmath67 for @xmath68 ( see figure [ fig:2x2-x4 ] ) . [ fig:2x2-x4 ] ( -0.1,0 ) ( 1.1,0 ) node[below ] @xmath55 ; ( 0,-0.1 ) ( 0,1.1 ) node[left ] @xmath59 ; ( 1,0.02 ) ( 1,-0.02 ) node[below ] @xmath27 ; ( 0.62,0.02 ) ( 0.62,-0.02 ) node[below ] @xmath41 ; ( 0.02,1 ) ( -0.02,1 ) node[left ] @xmath27 ; plot ( , ) ; plot ( , 2 * - ) ; for every @xmath69 we consider the sequence @xmath70 , which is monotonically decreasing because @xmath71 let @xmath72 .
the sequence is monotonically decreasing ; therefore , @xmath73 is well defined and @xmath74 .
note also that @xmath73 is a fixed point of @xmath59 because @xmath75 hence @xmath76 . using similar arguments
we can show that for @xmath77 the sequence @xmath78 is monotonically increasing and its limit is the fixed point 1 .
summarizing , we have @xmath79 this implies that @xmath80 converges to a limit dirac measure concentrated in @xmath41 .
the fact that the sequence @xmath81 converges to the same limit readily follows from equation .
we have proved that the cdf of @xmath82 pointwise converges to the cdf of the dirac measure , which implies weak convergence .
the following observation demonstrates that unlike alternating random games the value of a random zero - sum game with random structure does not concentrate .
[ pro : zs ] for every @xmath4 , @xmath14 is the uniform distribution over @xmath40 $ ] . using similar arguments to those in the proof of theorem [ th : zs ] ,
we can write a recursive formula for the sequence of probability measures @xmath83 .
we argue that @xmath84 this follows from the fact that with probability @xmath10 player 1 controls a certain vertex at depth @xmath85 , and then he chooses the maximal payoff . with probability @xmath10 player 2 controls this vertex ,
and then he chooses the minimal payoff . from this recursive formula
we deduce that @xmath86)=\frac{1}{2}(\nu_k([0,x]))^2+\frac{1}{2}[2\nu_k([0,x])-(\nu_k([0,x]))^2]=\nu_k([0,x]).\end{aligned}\ ] ] namely , the sequence of measures @xmath87 remains constant for every initial probability measure @xmath88 .
now we consider the case where @xmath89 is a general compact convex set with non - empty interior .
[ def : c ] for a point @xmath90 we denote @xmath91 the first quadrant with the origin at point @xmath55 .
similarly we denote @xmath92 , @xmath93 , and @xmath94 the second , third , and fourth quadrants respectively , see figure [ pic : defc ] .
more generally we denote @xmath95 where @xmath96 . , @xmath92 , @xmath93 , and @xmath94 .
] as in the zero - sum case , the value of the alternating game concentrates around a point .
moreover the concentration point is pareto efficient .
[ th : gd ] the value of the alternating game @xmath13 concentrates around a pareto efficient point . for @xmath97
we denote @xmath98 , similarly , for @xmath99 we denote @xmath100 .
we let @xmath101 be the cdf of the marginal of @xmath102 to player 1 s payoffs and @xmath103 to player 2 s payoffs .
let @xmath104 be the number @xmath55 for which @xmath105 and similarly @xmath106 is defined by the number @xmath107 for which @xmath108 .
[ mon : lemma ] both @xmath109 and @xmath110 are increasing sequences . in order to show that @xmath111 is increasing with @xmath4
it is sufficient to show that @xmath112 since if holds , then @xmath113 , and by the monotonicity of the cdf we have that @xmath114 .
first , note that @xmath115 and hence @xmath116 ; this follows directly by a similar consideration as in equation .
now we consider step @xmath117 .
clearly , @xmath118 is bounded from above by the probability that two i.i.d .
draws that are distributed according to @xmath119 lies in @xmath120 ( because @xmath121 is essentially obtained by a selection rule among i.i.d .
outcomes that are distributed according to @xmath119 ) .
therefore we have @xmath122 a symmetric consideration applied to player @xmath29 shows that @xmath123 is increasing .
let @xmath124 , and @xmath125 .
we first contend that if @xmath126 then @xmath127 . to see this note that , by definition @xmath128 for some @xmath1 . a similar derivation to those applied in the proof of theorem [ th : zs ] yields that @xmath129 where the last equality follows from the fact that @xmath130 .
from this we can deduce that @xmath131 , because for odd steps we can bound @xmath132 .
similar arguments show that for every @xmath133 , @xmath134 we next show that @xmath135 lies on the pareto efficient boundary .
equations and imply that @xmath136 is nonempty for every @xmath137 and @xmath133 ( see definition [ def : c ] ) , and therefore @xmath138 is nonempty . assume by way of contradiction that @xmath135 is pareto dominated .
let @xmath139 be any point such that @xmath140 @xmath141 and @xmath142 has a nonempty interior .
in particular @xmath143 .
we claim first that for every @xmath4 , @xmath144 because with probability at least @xmath145 the value of the left subtree is in @xmath146 and the value of the right subtree is in @xmath142 ; in such a case player 1 will choose the outcome from @xmath142 .
however , we also have the disjoint event where the left and right realizations are flipped ( this event is disjoint because @xmath147 ) where again the chosen outcome is in @xmath142 . therefore @xmath148 is a lower bound on @xmath149 .
a similar consideration applied to player @xmath29 shows that for every @xmath4 , @xmath150 together equations and imply that for every @xmath4 , @xmath151 since @xmath143 this stands in contradiction to the fact that @xmath152 for every @xmath4 .
we must therefore have that @xmath135 lies in the pareto efficient boundary . by equations and for every @xmath153 and @xmath154 we have that , @xmath155 .
since every open set @xmath156 in @xmath0 that contains @xmath135 must also contain @xmath157 for some @xmath158 and @xmath154 , we must have that @xmath159 .
this shows that @xmath82 weakly converges to the dirac measure @xmath160 and establishes the proof of the theorem .
we denote by @xmath161 the set of points within a distance at most @xmath34 from the pareto efficient boundary .
we say that a sequence of measures @xmath102 _ concentrates on the pareto efficient boundary _
if @xmath162 for every @xmath35 .
the following proposition shows that although @xmath14 does not concentrate around a point , the value is ( approximately ) pareto efficient with probability close to 1 .
[ pro : gen ] the value of the random controlling player game concentrates on the pareto efficient boundary . similar to the zero - sum case ( equation [ eq : rec - sym - zs ] ) , we can deduce the following recursive formula on @xmath163 .
@xmath164 where @xmath165 denotes the maximum over the @xmath12th coordinate of two independent realizations of @xmath102 .
we let @xmath166 be the set of points with a distance of precisely @xmath167 from the efficient boundary .
we will show first that for every @xmath168 it holds that @xmath169 ( see definition [ def : c ] ) .
if not , there exists some @xmath170 and @xmath37 for which @xmath171 for infinitely many @xmath4 s .
as before @xmath172 represents the probability under @xmath102 that the realised outcome for player @xmath27 would be less than @xmath55 and similarly @xmath173 for player @xmath29 .
we bound from below the probability @xmath174 using the recursive formula .
there are three events under which the realization of @xmath175 lies in @xmath176 .
1 . both realizations of @xmath177 lie in @xmath176 .
the probability of this event is @xmath178 .
player 1 controls the root
one realization of @xmath102 is in @xmath176 , and the other realization is in @xmath179 .
the probability of the first of the two is clearly @xmath180 the probability of the second is @xmath181 since the realizations can be flipped ( each can appear in the left subgame or the right subgame ) .
overall this event holds with probability @xmath182 .
3 . player 2 controls the root one realization of @xmath102 lies in @xmath176 , and the other realization lies in @xmath183 .
the probability of this event is @xmath184 .
these three events are disjoint ( note that ( 2 ) and ( 3 ) are disjoint because the controlling player defers ) .
therefore , we deduce that for every @xmath4 @xmath185 we can rewrite equation as follows @xmath186 since @xmath187 we conclude that @xmath188 increases with @xmath4 .
also if @xmath171 then @xmath189 hence since @xmath171 for infinitely many @xmath4 we have that for every @xmath1 there exists @xmath190 such that for every @xmath191 , @xmath192 this stands in contradiction to the fact that @xmath193 .
fix @xmath194 .
we show that there exists @xmath190 such that for every @xmath191 , @xmath195 choose @xmath196 to be a finite sequence of points , dense enough such that if @xmath197 satisfies @xmath198 then @xmath199 , i.e. , @xmath200 . by the above we can find large enough @xmath190 such that for every @xmath191 , and every @xmath201 , @xmath202 we therefore conclude that for every @xmath191 , @xmath203 since @xmath200 we conclude that @xmath195 this completes the proof of the proposition .
as we did previously , we consider as a warm - up the zero - sum case where @xmath204 $ ] .
we establish a generalization of theorem [ th : zs ] which states that the value of the random hybrid game @xmath31 converges to @xmath205 .
this is indeed a generalization of theorem @xmath27 because if we set @xmath206 then the random game @xmath32 and @xmath207 .
@xmath36 concentrates around the point @xmath205 .
the proof uses similar ideas to the proof of theorem [ th : zs ] . for every @xmath4 ,
let @xmath208 be the cdf of @xmath36 .
for every @xmath4 we have , @xmath209,\ ] ] and @xmath210.\ ] ] using equations and we get , @xmath211+(2\varepsilon)^3[-(f_{2n}(x)-(f_{2n}(x))^2)^2].\end{aligned}\ ] ] let @xmath212+(2\varepsilon)^3[-(x - x^2)^2].\ ] ] simple algebraic computations show that the function @xmath213 has four fixed points @xmath214 among these fixed points only @xmath215 and @xmath216 are in the segment @xmath40 $ ] .
we denote @xmath217 , and we note that @xmath218 and @xmath219 .
note also that @xmath220 for @xmath221 ( for instance because @xmath222 ) and @xmath223 for @xmath224 ( for instance because @xmath225 ) .
figure [ fig : phi - eps ] illustrates the function @xmath213 for @xmath226 . [ fig:2x2-x4 ] ( -0.1,0 ) ( 1.1,0 ) node[below ] @xmath55 ; ( 0,-0.1 ) ( 0,1.1 ) node[left ] @xmath59 ; ( 1,0.02 ) ( 1,-0.02 ) node[below ] @xmath27 ; ( 0.56,0.02 )
( 0.56,-0.02 ) node[below ] @xmath227 ; ( 0.02,1 ) ( -0.02,1 ) node[left ] @xmath27 ; plot ( , ) ; plot ( , -0.125 * -0.25*+0.625*+ 0.75 * ) ; one can see by equation that @xmath228 similar arguments to those applied in theorem [ th : zs ] imply that @xmath229 weakly converges to @xmath230 as @xmath4 grows . the following theorem studies the asymptotic properties of @xmath229 for the general convex domain @xmath0 . [
theo : bar ] for every @xmath34 the value of the hybrid game concentrates around a pareto efficient point @xmath231 .
moreover , if the domain @xmath0 is symmetric then @xmath232 ; namely , the concentration point approaches the diagonal when @xmath34 goes to @xmath233 . before establishing the proof of the theorem ( section [ sec : bar - pr ] ) we present more formally the applications of this result to bargaining and implementation . given a bargaining set
we define the _ random - extensive - form ( ref ) solution _ to be @xmath234 note that the ref solution can also be derived from a more natural version of a random extinctive form game @xmath11 , where the ref solution is the limit point of the medians of the marginal distributions ( see proposition [ pro : second - part ] ) .
we recall that a solution is called _ standard _ if it satisfies pareto efficiency , symmetry , and invariance with respect to positive affine transformations .
the ref solution is a standard solution . pareto efficiency and symmetry follow immediately from theorem [ theo : bar ] .
for invariance with respect to positive affine transformations we note that the uniform distribution remains uniform after operating such a transformation .
in addition , the subgame perfect equilibria outcomes are invariant to such transformations . beyond the fact that the ref solution satisfies several desirable axioms ,
the main advantage of the ref solution is the fact that this solution has a very clean probabilistic ( approximate ) implementation .
we set small @xmath34 and large @xmath4 .
now we simply randomize the decision makers and the outcomes in the game as it is done in the hybrid game @xmath31 . with high probability
the subgame perfect equilibrium outcome of the game is close to the ref solution .
we start with a proof of the first part of theorem [ theo : bar ] ( regarding the concentration around a pareto efficient point ) .
in fact , we derive a stronger statement which provides also a type of characterization for the concentration point .
we denote by @xmath235 the probability distribution of @xmath236 and for @xmath9 we let @xmath237 be the cdf of the marginal distribution of @xmath238 over the payoffs to player @xmath12 .
we define @xmath239 to be the unique number @xmath55 such that @xmath240 .
similarly , @xmath241 is the number @xmath107 such that @xmath242 .
[ pro : first - part ] the sequences @xmath239 and @xmath243 have the following properties . 1 .
@xmath244 and @xmath245 are monotonically increasing in @xmath4 .
2 . @xmath246 is a converging sequence .
3 . @xmath247 , namely , the limit of the sequence is the concentration point of @xmath236 .
4 . @xmath231 lies on the pareto efficient boundary of @xmath0 . by applying identical considerations to those in the proof of theorem [ th : gd ] , and replacing @xmath41 by @xmath227 we get exactly all these properties .
now we get to the proof of the second statement in theorem [ theo : bar ] . the proof is based on a connection between the sequences of measures @xmath14 and @xmath36 ( see proposition [ pro : second - part ] ) . in order to state this connection we need an additional convergence statement about the random structure games @xmath14 .
we denote by @xmath248 the cdfs of the marginal distribution of @xmath249 over the payoffs to player @xmath250 .
we denote by @xmath104 and @xmath106 the medians of @xmath251 and @xmath252 ( respectively ) .
[ pro : sym ] the sequences @xmath104 and @xmath106 are monotonically increasing . moreover
the limit point @xmath253 lies on the pareto efficient boundary of .
nevertheless , the proposition states that the limit point @xmath254 is well defined , and is pareto efficient .
] @xmath0 . for @xmath255
we denote @xmath256 . in order to prove that @xmath257 is monotonically increasing it is sufficient to show that @xmath258
. with probability @xmath10 player 1 controls the node , and in such a case , the result ( at the @xmath259th iteration ) is in @xmath260 iff both i.i.d .
realizations ( according to @xmath102 ) are in @xmath260 .
the probability of such an event is @xmath261 .
with probability @xmath10 player 2 controls the node , and in such a case , a necessary condition for the result ( at the @xmath259th iteration ) to be in @xmath260 is that at least one realization is in @xmath260 .
the probability of such an event is @xmath262 . summarizing , over the two cases we get @xmath263 .
similar arguments show that the sequence @xmath264 is monotonically increasing .
now we turn to the proof of statement ( 2 ) .
assume by way of contradiction that @xmath254 lies in the interior of @xmath0 , and then @xmath265 ( see definition [ def : c ] ) . by equation ( which holds for the random controlling player case ) we have that for every @xmath266 @xmath267 since @xmath268 ( and @xmath269 ) we have @xmath270 .
hence , @xmath271 which contradicts the fact that @xmath272 for every @xmath4 .
assume by way of contradiction that @xmath254 lies out of the @xmath0 .
then for some finite @xmath273 @xmath274 is out of the set @xmath0 .
we denote @xmath275 and @xmath276 .
note that @xmath277 , and @xmath278 , which implies that @xmath279 .
this contradicts the fact that the density of @xmath102 remains strictly positive everywhere in @xmath0 for every finite @xmath4 .
therefore @xmath254 must lie on the pareto efficient boundary .
the connection between the sequences of measures @xmath14 and @xmath36 is given by following proposition , which states that the concentration point of @xmath36 is close to the limit of medians of @xmath14 from proposition [ pro : sym ] .
[ pro : second - part ] @xmath280 .
we emphasize that @xmath36 and @xmath14 behave very differently for large @xmath4 ( the first one concentrates around a point , whereas the latter does not ) .
therefore it is somewhat surprising that we succeeded to connect the concentration point of @xmath36 with some object that is derived from the sequence @xmath281 .
let @xmath282 be a global lipschitz constant such that for every two points @xmath283 and @xmath284 on the pareto efficient boundary of @xmath0 holds will suffice for the arguments that we present here in the proof . ]
@xmath285 we set @xmath37 , and we shall prove that there exists @xmath286 such that for every @xmath287 holds @xmath288 . the way we do it is by considering the sequences @xmath289 and @xmath246 . by proposition
[ pro : sym ] @xmath290 , so there exists @xmath190 such that latexmath:[\[\begin{aligned } \label{eq : xn - x } we emphasize that @xmath190 does not depend on @xmath34 ( because @xmath190 depends only on a symmetric random structure where @xmath34 is not involved ) .
we denote by @xmath292 and @xmath293 the medians of @xmath294 and @xmath295 . by lemma [ lem : str - con ]
the measures @xmath296 and @xmath297 are strongly continuous with respect to @xmath34 .
in particular , it follows that the medians of @xmath298 converge to the medians of @xmath49 .
namely , @xmath299 and @xmath300 .
therefore , there exists @xmath301 such that for every @xmath302 holds @xmath303 and @xmath304 are continuous with respect to @xmath55 and @xmath107 ( respectively ) .
note also that @xmath305 .
therefore by the definition of @xmath306 and @xmath243 there exists @xmath307 such that for every @xmath308 holds @xmath309 then inequalities , , and guarantee that @xmath310 ( at the beginning of the proof ) we know that @xmath231 pareto dominates the point @xmath311 .
in addition , by proposition [ pro : sym ] we know that @xmath254 is pareto efficient . by the definition of @xmath312 together with the above three properties we get that @xmath288 .
[ lem : str - con ] for every constant @xmath190 the measure @xmath313 is strongly continuous ( in the total variation distance ) with respect to @xmath34 around @xmath314 .
the random variable @xmath315 can be obtained by the following procedure . 1 .
we randomize @xmath316 i.i.d .
random outcomes according to @xmath317 .
for each node @xmath318 we choose player @xmath9 to control node @xmath318 with probability @xmath10 .
3 . for each node
@xmath318 we draw a bernoulli random variable @xmath319 with @xmath320 .
4 . for each odd - depth node
@xmath318 , if @xmath321 we set player 1 to control the node ( irrespective of who was chosen to control it at step ( 2 ) ) .
5 . for each even - depth
node @xmath318 , if @xmath321 we set player 2 to control the node ( irrespective of who was chosen to control it at step ( 2 ) ) .
. we apply backward induction on the resulted game .
note that @xmath322= ( 1 - 2\varepsilon)^{2^{n_0}-1}$ ] . in such a case
( of @xmath323 for all @xmath318 ) we have exactly the random variable @xmath324 . therefore the total variation distance is bounded by @xmath325 , i.e. , @xmath326 .
now theorem [ theo : bar ] follows immediately from propositions [ pro : first - part ] and [ pro : second - part ] . the first part of the theorem is proved in proposition [ pro : first - part ] items ( 3 ) and ( 4 ) . regarding the second part , if @xmath0 is symmetric then @xmath327 for every @xmath4 , and therefore obviously @xmath328 .
so proposition [ pro : second - part ] completes the proof .
* 1 . general settings for which all the results hold . * by looking on the proofs of the results one can observe that the properties of the domain @xmath0 which were used in the proofs are as follows : 1 .
@xmath0 is a compact set with a non - empty interior .
2 . for every not pareto efficient point @xmath329 the set
@xmath136 has a non - empty interior .
3 . the pareto efficient boundary forms a connected path which has the lipschitz property .
this requirement is needed only for the analysis of the hybrid game . again by looking at the proofs of the results one can observe that the only property on the initial distribution @xmath3 that was used is the fact that @xmath3 is a non - atomic measure with strictly positive density over all @xmath0 . the only change in the results ( but not in the proofs ) that is caused by changing the uniform distribution by a general distribution @xmath3 is theorem [ th : zs ] : the concentration point is obviously not the point @xmath330 but the point @xmath18 for which @xmath331)=b$ ] . *
2 . games over ternary trees .
* the paper focuses on the case where the game tree is a complete binary tree .
the techniques derived in the paper can be applied also for the case where the game tree is ternary .
we overview here the results for this case ( without proofs ) .
as we can see part of the results are similar but part are not : * the value of the alternating game concentrates around a pareto efficient point . in the case of a zero - sum game
the concentration is around @xmath332 , where @xmath41 is the unique solution of @xmath333 in the segment @xmath334 .
these results are similar to the case of binary trees . *
the value of the random controlling player game converges to the distribution @xmath335 which assigns a probability of @xmath10 to both points @xmath336 , where @xmath337 is the best point in @xmath0 for player @xmath12 .
this statement holds for both , the zero sum case and the case of a general two - dimensional set @xmath0 .
these results _ defer _ from the binary case where the limit measure has positive density over the entire pareto efficient boundary . * for small enough values of @xmath34 ,
the value of the hybrid game converges to @xmath338 , where @xmath338 assigns a probability @xmath34 close to @xmath10 to @xmath339 ( from the previous bullet ) and the remaining probability to @xmath340 . in particular , the hybrid game fails to have the desirable property of concentration around a point .
namely , a different random structure than the hybrid game is required in order to implement a standard solution in ternary games .
|
we consider two - player random extensive form games where the payoffs at the leaves are independently drawn uniformly at random from a given feasible set @xmath0 .
we study the asymptotic distribution of the subgame perfect equilibrium outcome for binary - trees with increasing depth in various random ( or deterministic ) assignments of players to nodes .
we characterize the assignments under which the asymptotic distribution concentrates around a point .
our analysis provides a natural way to derive from the asymptotic distribution a novel solution concept for two - player bargaining problems with a solid strategic justification .
|
hydrodynamics is one of the main tools for studying the collective flow in high - energy nuclear collisions . in this approach
, it is assumed that , after a complex process involving microscopic collisions of nuclear constituents , at a certain early instant a hot and dense matter is formed , which would be in local thermal equilibrium .
after this instant , the system would expand hydrodynamically , following the well known set of differential equations .
the initial conditions ( ic ) for the hydrodynamic expansion are usually parametrized as smooth distributions of thermodynamic quantities and four - velocity .
however , since our systems are small , important _ event - by - event fluctuations _ are expected in the ic of the real collisions . moreover
, each set of ic should presents strongly _
inhomogeneous structure_. our purpose here is to discuss some of the effects caused by such fluctuating and bumpy ic on observables .
the main tool for our study is called nexspherio .
it is a junction of two computational codes : nexus and spherio .
the nexus code @xcite is used to compute the ic .
it is a microscopic model based on the regge - gribov theory and , once a pair of incident nuclei and their incident energy are chosen , it can produce , in the event - by - event basis , detailed space distributions of energy - momentum tensor , baryon - number , strangeness and charge densities , at a given initial time @xmath2fm .
we show in fig.[ic ] an example of such a fluctuating event , produced by nexus event generator , for central au + au collision at @xmath3gev . as seen , the energy - density distribution is highly irregular and in a transverse plane ( left panel ) it presents several high - density blobs , whereas in a longitudinal plane ( right panel ) it presents a baton - like structure . when averaged over many events , these bumps disappear completely giving smooth ic , as those commonly used in hydro calculations .
however , this bumpy structure gives important effects as we will show below .
the spherio code @xcite , based on smoothed particle hydrodynamics ( sph ) algorithm @xcite , is well suited to computing the evolution of such systems , so complex as the one shown in fig.[ic ] . a gev .
left : energy density distribution is plotted in the @xmath4 plane .
right : corresponding plot in the @xmath5 fm plane.,title="fig:",width=207 ] a gev . left : energy density distribution
is plotted in the @xmath4 plane .
right : corresponding plot in the @xmath5 fm plane.,title="fig:",width=245 ]
in a previous paper @xcite , we discussed some effects of the bumpy structure of ic , as shown in fig.[ic ] , on @xmath0 spectra and @xmath1 coefficient .
the main consequence of baton structure is a violent and cylindrically isotropic expansion of the batons , which produces additional isotropic high-@xmath0 particles .
as clearly shown in fig [ dndpt ] , this implies that high-@xmath0 part of the @xmath0 spectrum becomes enhanced as compared with the smooth averaged ic case , making it more concave and closer to data .
distributions ( including those from resonance decays ) computed in two different ways .
the solid line indicates result for fluctuating ic , whereas the dotted line the one for the averaged ic .
data points @xcite are also plotted for comparison . ] as for the anisotropy of the transverse flow , we illustrate in fig.[v2pt ] that the elliptic - flow coefficient @xmath1 suffers reduction as we go to high-@xmath0 region , due to the additional high-@xmath0 isotropic component included now .
the result is closer to the available data .
dependence of @xmath6 in the centrality window and @xmath7 interval as indicated , compared with data @xcite .
the solid line indicates result for fluctuating ic , whereas the dotted one that for the averaged ic .
the curves are averages over phobos centrality sub - intervals with freeze - out temperatures as indicated .
, width=245 ] as for the @xmath7 dependence of @xmath8 , we know that the average matter density decreases as @xmath9 increases as reflected in the @xmath7 distribution of charged particles , so the isotropic expansion effect of the batons becomes more efficacious there and , therefore , @xmath8 suffers a considerable reduction in those regions .
dependence of @xmath6 for three centrality windows .
the solid lines indicate results for fluctuating ic , whereas the dotted lines the ones for the averaged ic .
data points@xcite are also plotted for comparison .
@xmath10 has been taken as in fig.[v2pt].,width=245 ] the ic fluctuation also implies a large @xmath1 fluctuation . in preliminary works @xcite on au+au collisions at @xmath11 gev
, we showed that this actually happens .
the results , with qgp included , have indeed been confirmed in recent experiments@xcite .
more recent computations for au+au at @xmath12gev @xcite gave similar results . in fig .
[ flv2 ] , we compare the latter with those data . computed for au+au collisions at @xmath3gev , compared with data .
upper : @xmath13 is given as function of the impact parameter @xmath14 and compared with the star data @xcite .
lower : the same results are expressed as function of participant nucleon number @xmath15 and compared with the phobos data , @xcite . , width=245 ] [ flv2 ] another effect , which is produced naturally by the longitudinal baton structure of ic , as shown in fig.[ic ] , is the so - called _ ridge phenomenon _ which has been experimentally seen in high-@xmath0 nearside correlations @xcite . since batons in fluctuating ic which are close to the surface produce longitudinally correlated high-@xmath0 particles , always in the same side of the hot matter , the ridge structure naturally appears .
this is shown in fig.[ridge ] , as computed with nexspherio by j. takahashi _ et al .
_ @xcite the ridge phenomenon has been discussed also in connection with glasma flux tubes @xcite .
particle correlation , computed with event - by - event fluctuating ic .
in this paper , we presented several consequences of event - by - event fluctuating ic with baton structure , computed with nexspherio code .
the main results are : 1 ) the baton structure of ic produces more concave @xmath0 spectra , as compared with smooth ic ; 2 ) it reduces @xmath6 both in the high-@xmath0 and large-@xmath16 regions ; 3 ) it also produces the _ ridge structure _ in the nearside correlations ; 4 ) large @xmath1 fluctuations occurs , in good agreement with data .
this work was partially supported by fapesp , cnpq , faperj and pronex .
we are grateful to jun takahashi for providing us with fig.[ridge ] t. osada , c.e .
aguiar , y. hama , and t. kodama , in proc . of the vi ranp workshop , eds . t. kodama _ et al .
_ , world scientific , singapore ( 2001 ) , pg .
174 [ nucl - th/0102011 ] ; c.e .
aguiar , y. hama , t. kodama and t. osada , nucl .
a*698 * , 639c ( 2002 ) .
|
we show effects of the event - by - event fluctuation of the initial conditions ( ic ) in hydrodynamic description of high - energy nuclear collisions on some observables .
such ic produce not only fluctuations in observables but , due to their bumpy structure , several non - trivial effects appear .
they enhance production of isotropically distributed high-@xmath0 particles , making @xmath1 smaller there . also , they reduce @xmath1 in the forward and backward regions where the global matter density is smaller , so where such effects become more efficacious
. they may also produce the so - called _ ridge effect _ in the two large-@xmath0 particle correlation .
|
bosonic part of the lagrangian for the @xmath1 supergravity action in five dimensions is given by @xmath6 where @xmath7 and @xmath8 .
also , @xmath9 a bosonic field content of the theory is the following .
we have weyl multiplet which contains the fnfbein @xmath10 , the two - form auxiliary field @xmath11 , and the scalar auxiliary field @xmath12 .
there are @xmath13 vector multiplets enumerated by @xmath14 , each containing the one - form gauge field @xmath15 ( with the two - form field strength @xmath16 ) , and the scalar @xmath17 .
scalar fields @xmath18 , which are belonging to the hypermultiplet , can be gauge fixed and the convenient choice is given by @xmath19 , @xmath20 .
lagrangian ( [ l0susy ] ) can be obtained from 11-dimensional sugra by compactifying on six - dimensional calabi - yau spaces .
then @xmath17 have interpretation as moduli ( volumes of @xmath21-cycles ) , and @xmath22 as intersection numbers .
condition @xmath23 is a condition of real special geometry .
action ( [ l0susy ] ) is invariant under susy variations , which when acting on the purely bosonic configurations are given with @xmath24 where @xmath25 is gravitino , @xmath26 auxiliary majorana spinor ( weyl multiplet ) , @xmath27 gaugino ( vector multiplets ) , and @xmath28 is a fermion field from hypermultiplet . in
@xcite a four - derivative part of the action was constructed by supersymmetric completion of the mixed gauge - gravitational chern - simons term @xmath29 .
the bosonic part of the action is @xmath30 + f^{iab } \left [ \frac{1}{6 } v_{ab } d - \frac{1}{2 } c_{abcd } v^{cd } + \frac{2}{3 } e^{-1 } \epsilon_{abcde } v^{cd } \hat{\mathcal{d}}_f v^{ef } \right .
\nonumber \\ & & \left .
+ e^{-1 } \epsilon_{abcde } v^{c}_{\;f } \hat{\mathcal{d}}^d v^{ef } - \frac{4}{3 } v_{ac}v^{cd } v_{db } - \frac{1}{3 } v_{ab } v^2 \right ] \right\}\end{aligned}\ ] ] where @xmath31 are some constant coefficients sugra they are topological numbers connected to second chern class .
] , @xmath32 is the weyl tensor , and @xmath33 is the conformal covariant derivative .
we are interested in extremal black hole solutions of the action obtained by combining ( [ l0susy ] ) and ( [ l1susy ] ) : and for the string tension @xmath34 . ]
@xmath35 the action ( [ lsusy ] ) is quartic in derivatives and generally probably too complicated for finding complete analytical black hole solutions even in the simplest spherically symmetric case .
but , if one is more modest and interested just in a near - horizon behavior ( which is enough to find the entropy ) of _ extremal _ black holes , there is a smart way to do the job - sen s entropy function formalism @xcite . for five - dimensional spherically symmetric extremal black holes near - horizon geometry
is expected to be @xmath36 , which has @xmath37 symmetry .
if the lagrangian can be written in a manifestly diffeomorphism covariant and gauge invariant way , it is expected that near the horizon the complete background should respect this symmetry . in our case
it means that near - horizon geometry should be given with @xmath38 where @xmath39 , @xmath40 , @xmath17 , @xmath41 , and @xmath12 are constants .
all covariant derivatives are vanishing .
if one defines @xmath42 where right hand side is evaluated on the background ( [ efhere ] ) , then equations of motion are equivalent to @xmath43 derivatives over electric field strengths @xmath40 are giving ( properly normalized ) electric charges : @xmath44 the entropy ( equal to the wald formula @xcite ) is given with @xmath45 it is immediately obvious that though the system ( [ seom ] ) , ( [ chgdef ] ) is algebraic , it is in generic case too complicated to be solved in direct manner , and that one should try to find some additional information .
such additional information can be obtained from supersymmetry .
it is known that there should be 1/2 bps black hole solutions , for which it was shown in @xcite that near the horizon supersymmetry is enhanced fully .
this means that in this case we can put all variations in ( [ svar ] ) to zero , which one can use to express all unknowns in terms of one . to fix remaining unknown
we just need one equation from ( [ seom ] ) , where the simplest is the one for @xmath12 . typically one is interested in expressing the results in terms of charges , not field strengths , and this is achieved by using ( [ chgdef ] ) .
one gets @xcite @xmath46 where we introduced scaled moduli @xmath47 .
the entropy becomes @xmath48
a virtue of this presentation is that if one is interested only in entropy , then it is enough to consider just ( [ mbareq ] ) and ( [ sentm ] ) , in which the sole effect of the higher derivative terms are just constant shifts of charges @xmath49 .
it was shown in @xcite that ( [ sentm ] ) agrees with the osv conjecture @xcite , after proper treatment of uplift from @xmath5 to @xmath2 is made .
we shall be especially interested in the case when prepotential is of the form @xmath50 where @xmath51 is a regular matrix with an inverse @xmath52 . in this case , which corresponds to @xmath53 11-dimensional compactifications , it is easy to show that the entropy of bps black holes is given with @xmath54 when additionally @xmath55 , @xmath56 , our action is equivalent to the ( consistently truncated ) tree - level effective action of heterotic string compactified on @xmath57 .
one especially interesting ( and simple ) case is given with @xmath58-prepotential @xmath59 , which corresponds to heterotic string on @xmath60 .
black hole solutions are caracterised by three integer charges usually denoted as @xmath61 , @xmath62 and @xmath63 . and
@xmath64 are momentum and winding number on @xmath65 , and @xmath66 is the magnetic charge of antisymmetric @xmath67 field . ]
constructed bps black hole solutions are physically acceptable for @xmath68 and @xmath69 .
the entropy is now @xmath70 it is remarkable that here non - bps solutions ( for almost all values of charges ) were also analytically constructed @xcite .
for example , for @xmath71 , @xmath72 , @xmath73 the entropy is @xmath74 properties of these non - bps solutions suggest possibility that they are descending from bps states either in @xmath75 and/or @xmath76 @xcite . motivated by results from @xmath5
, we have studied solutions when @xmath0-correction is given purely by gauss - bonnet density .
we obtained that the entropy is different from the one following from @xmath0 susy action ( [ sek3 ] ) .
this mismatch was not present in @xmath5 .
in view of above results , it is interesting to perturbatively calculate entropy of large 5-dimensional 3-charge extremal black holes up to @xmath77-order using low energy effective action of heterotic string ( which is fully known only up to @xmath77-order ) .
the main virtue is that this is a straightforward calculation giving unambiguous results . using entropy function formalism , and taking special care for non manifestly covariant gravitational chern - simons term ( using technique from @xcite )
, we obtained for the entropy of bps black holes @xcite @xmath78 which is in agreement with the supersymmetric result , i.e. , with ( [ ssusyb ] ) after being expanded in @xmath79 . for non - bps black holes we obtained the entropy @xmath80 which disagrees with both susy ( [ ssusyb ] ) and gauss - bonnet results already at @xmath4-order .
instead , our result ( [ spertn ] ) suggests the following formula @xmath81 furthermore , if we take bps formula ( [ ssusyb ] ) for granted , then we have been able to show that @xmath82 term in the non - bps entropy formula ( [ spertn ] ) must be @xmath83 , which is again in agreement with the conjectured expression ( [ segzn ] ) .
now , using ads / cft arguments , from ( [ spertb ] ) and ( [ spertn ] ) one infers that central charges satisfy @xmath84 , which is indeed what is expected @xcite .
extremely interesting is what happens when one takes @xmath85 in ( [ sek3 ] ) . for the @xmath86 heterotic compactifications
the entropy becomes @xmath87 on the other hand , for the action with gauss - bonnet @xmath0 term we obtain @xmath88 these black holes are _ small _ , meaning that their horizon is generated ( regularized ) by higher - derivative terms in the action .
contrary to the large black holes discussed before , for these small black holes microscopic stringy description _ is _ known ( in the special 2-charge case of @xmath89 heterotic compactification microstates are well - known perturbative dabholkar - harvey states ) for which statistical entropy was calculated @xcite .
for bps states microscopic entropy ( for @xmath90 ) is _ exactly _ equal to the black hole entropy obtained from the action supplemented with just gauss - bonnet @xmath0-correction ( [ smallgb ] ) , and disagrees with the entropy obtained by using supersymmetric @xmath0-correction ( [ smallk3 ] ) .
one needs in the action also higher gauss - bonnet densities @xcite . ]
we have analysed near - horizon solutions for ( both bps and non - bps ) static spherically symmetric black holes of @xmath1 sugra actions with @xmath0-corrections , which are effective actions of tree - level heterotic string compactified on @xmath3 and @xmath89 .
in addition , we have also made calculations by taking for @xmath0-correction just the gauss - bonnet density . in @xmath2 ,
contrary to a situation in @xmath5 ( see @xcite for a review ) , for these two types of higher - derivative corrections formulae for the entropy are _ not _ matching . for _ large _ black holes , where full stringy microscopic description is not known , obtained entropies of bps black holes are _ equal _ to the one obtained from osv conjecture and topological string , and are consistent with perturbative results up to @xmath77-order .
the exact matchings obtained by using just the @xmath0-corrections in effective actions ( which has infinite expansion ) is surprising . for susy correction
it has partial explanation through ads@xmath91/cft@xmath92 correspondence @xcite , but for small black holes where simple gauss - bonnet correction does the job ( and susy correction fails ) it is still a complete mistery .
we believe that these issues deserve further investigation .
this work was supported by the croatian ministry of science , education and sport under the contract 119 - 0982930 - 1016 .
was also supported by alexander von humboldt foundation , and m.c . by central european initiative .
k. hanaki , k. ohashi and y. tachikawa , prog .
* 117 * , 533 ( 2007 ) .
a. sen , jhep * 0509 * , 038 ( 2005 ) .
wald , phys .
d * 48 * , 3427 ( 1993 ) ; v. iyer and r.m .
wald , phys .
d * 50 * , 846 ( 1994 ) .
chamseddine , s. ferrara , g.w .
gibbons and r. kallosh , phys .
d * 55 * , 3647 ( 1997 ) .
a. castro , j.l .
davis , p. kraus and f. larsen , jhep * 0706 * , 007 ( 2007 ) .
m. alishahiha , jhep * 0708 * , 094 ( 2007 ) . m. cvitan , p. dominis prester , s. pallua and i. smoli , jhep * 0711 * , 043 ( 2007 ) arxiv:0706.1167 [ hep - th ] .
a. castro , j.l .
davis , p. kraus and f. larsen , jhep * 0709 * , 003 ( 2007 ) .
h. ooguri , a. strominger and c. vafa , phys .
d * 70 * , 106007 ( 2004 ) .
b. sahoo and a. sen , jhep * 0701 * , 010 ( 2007 ) .
m. cvitan , p. dominis prester and a. ficnar , arxiv:0710.3886 [ hep - th ] .
p. kraus , f. larsen and a. shah , arxiv:0708.1001 [ hep - th ] .
a. dabholkar , phys .
* 94 * , 241301 ( 2005 ) .
x . huang , a. klemm , m. marino and a. tavanfar , arxiv:0704.2440 [ hep - th ] .
p. prester , jhep * 0602 * , 039 ( 2006 ) .
a. sen , arxiv:0708.1270 [ hep - th ] .
p. kraus and f. larsen , jhep * 0509 * , 034 ( 2005 ) .
|
we review recent results on near - horizon static black hole solutions and entropy in @xmath0-corrected @xmath1 sugra in @xmath2 , focusing on actions connected to heterotic string compactified on @xmath3 .
comparison with @xmath4-perturbative results , results obtained by using simple gauss - bonnet @xmath0-correction , osv conjecture and microscopic stringy description ( for small black holes ) shows that situation in @xmath2 is , in a sense , even more interesting then in @xmath5 .
|
rs cvn binaries were first noted for their sinusoidal optical light curve variability , interpreted as large scale analogs of sunspots ( hall , 1976 ) .
subsequent studies showed them to have optical ( hydrogen and calcium ; huenemoerder & ramsey 1984 ) and ultraviolet ( carbon , silicon , oxygen , nitrogen ; simon and linsky 1980 ) emission lines also analogous to the solar chromosphere and transition region .
they were later found to be luminous in x - rays ( walter & bowyer , 1981 ) , extending the solar analogy to the corona , but with 1000 times the activity level . eclipsing systems like
ar lac permit us to study coronal phenomena for stars of known mass , size , and rotational velocity .
emission line spectroscopy provides temperature and density diagnostics , which together with geometrical constraints , can provide insights regarding sizes of magnetic loops .
eclipse and rotational modulation provide further information on the size and distribution of the emitting structures .
together in the context provided by other stars and long term trends , we hope to learn about the fundamental physical processes which heat stellar coronae .
we observed ar lac with the chandra high energy grating spectrometer ( hetgs ) for 97 ks , divided into 6 separate observations which sampled eclipse and quadrature phases .
the hetgs has a maximum spectral resolution ( @xmath0 ) of about 1500 , covering the range of 1.8 to 30 with two complementary gratings , the medium , and the high energy gratings ( meg and heg , respectively ) .
approximately 20 years prior to these chandra observations , walter , gibson , and basri ( 1983 ) mapped x - ray emission with the _ einstein _ observatory , and found both extended and compact emission regions with an asymmetric distribution .
neff et al . ( 1989 ) mapped the chromosphere using ultraviolet lines .
we now have high resolution spectra which resolve the emission into a multitude of emission lines from highly ionized elements .
a light curve was made by binning diffracted photons in orders -3 to + 3 ( excluding 0 ) in both heg and meg arms .
figure [ fg : lcurve ] shows the count - rate against orbital phase , which is also rotational phase , since the system is tidally synchronized ( the period of ar lac is 1.98 days ) .
obsid 7 shows a strong flare , caught during the rise and part of the decay .
obsid 8 has a high rate , but is not obviously rising or falling .
the light curve shows no primary eclipse , indicating that the g2 iv star ( smaller component ) is x - ray dark , at least on one hemisphere .
the orbital radial velocity amplitude of ar lac is 230 km / s .
solar flares have been seen to have velocity fields of up to 400 km / s . in figure
[ fg : doppler ] , we compare line profiles for the quadrature times ( obsids 9 and 6 ) to profiles during times of conjunction , which also coincidentally have flares ( obsids 7,8,10 , & 11 ) .
we have looked at lines near the highest instrumental resolution : o viii 19 meg , ne x 12 heg ( blended w/ fe xvii ) , and fe xvii 15 ( heg and meg ) .
it is clear that lines during quadrature are broader than during conjunction , which is expected if both components are x - ray bright .
future work will examine line centroids versus time versus order to better quantify the origin of the broadening and shifts .
the helium - like triplet lines provide density diagnostics which are independent of relative abundance ratios and which are relatively independent of temperature .
the ne ix region is shown in figure [ fg : density ] for flare and low - flux intervals . the forbidden line ( 13.7 ) to intercombination line ( 13.55 ) ratio is density sensitive and decreases for densities above about @xmath1 ( cgs ) .
it appears likely that the ratio decreases during flares .
we have derived a provisional differential differential emission measure model by measuring line fluxes in the total integrated spectrum , and then fitting to them a differential emission measure ( dem ) with variable abundances .
the model is considered as provisional because of the great variability of the light curve and because the system is composite .
after an initial temperature distribution was derived , the dem and abundances were scaled to give an approximate match to line - to - continuum ratios .
the fit is not unique , and we will be fitting shorter , more stable intervals of the spectrum . at low temperatures ,
the fit is unconstrained because there are no lines from plasma cooler than about @xmath2 .
the hot end is constrained by emission from high ionization stages of fe and by the continuum .
the provisional dem is shown in figure [ fg : dem ] .
counts spectra are extracted for the entire 97 ks observation ( obsids 6 - 11 ) , which is shown in figure [ fg : spec ] .
plus and minus 1st orders were summed and smoothed by 0.008 ( medium energy grating , meg ) or 0.004 ( high energy grating , heg ) for display .
the marked lines are the brightest features for the adopted model .
the model was binned from the aped ( astrophysical plasma emission database ; smith et al 2001 ) and folded through the instrumental response .
the hetgs spectra show that the coronae of ar lac are highly variable , present on both stellar components , and asymmetrically distributed .
ionic species detected indicate a broad range in plasma temperatures .
relative abundances are subject to further more detailed analysis , but preliminary results fairly reliably indicate that iron is depleted by about a factor of three relative to neon , based upon ne ix and fe xvii ratios , since these ions have similar emissivity functions with temperature .
we also detected a possible density enhancement during the flare state .
|
we observed ar lac , an eclipsing rs cvn binary star , with the chandra high energy grating spectrometer for a total of 100 ks divided into six intervals covering both quadratures and eclipses .
we repeated observations at each phase .
at least two flares were seen , in which the flux increased by factors of two and four .
the flares occurred near eclipse phases and compromised detection of eclipse modulation .
the quadrature fluxes were the most stable , but they also show non - repeating trends .
quadrature line profiles are broadened relative to eclipse profiles , presumably due to orbital velocity doppler shifted emission from each binary component .
the spectrum appears to be iron poor and neon rich , similar to hr 1099 , but not as extreme as ii pegasi .
we will examine line strengths , widths , and positions vs. phase , and present preliminary differential emission measure models and abundance determinations using the aped emissivity database .
this work is supported by nasa contract nas8 - 38249 ( hetg ) and sao sv1 - 61010 ( cxc ) to mit .
|
from the publicly available data , we consider the gamma - ray bursts ( grb ) with known redshifts and detected radio afterglows through march , 2009 . tables 14 collect the grbs with these properties which satisfy certain criteria .
we demand that the energy of the initially detected gamma - ray burst in its rest frame be compatible with the energy range of the fermi gamma - burst monitor .
this excludes the x - ray flashes grb 080109a and grb 020903a . to determine whether an observation of the radio flux is indeed a detected radio afterglow , we demand that the radio flux detection be within the error box of the location of the observed grb , that it not be a site of previously observed radio activity , and that it be significantly non - zero .
we use the criterion employed by ref .
@xcite and thus require that the radio observation be nonzero by three standard deviations or more to be termed a detection .
such considerations exclude grb 030277a , grb 980425a , and grb 011130a .
consequently , we find 53 grbs to consider , for which we report all detected radio frequencies of 75 ghz or less .
we report the time of the initial detection of the grb and the time of the detection of a particular observation frequency @xmath128 in the associated radio afterglow ( ra ) .
we calculate the time lag between the detection of a particular radio frequency and the initial detection of the grb from these quantities .
remarks concerning the collected data are in order . a redshift marked with an means that no certainty has been reported in the work which determines it . in those cases ,
we assume the uncertainty in the redshift to be plus or minus one unit in the last significant figure reported .
the redshift we report for grb 021004a is the average of those reported in refs .
@xcite . for the redshifts marked with a
, the associated publication states that the reported redshift is either a lower limit or the `` likely '' value . in the particular case of grb 970508a , work in refinement of the upper bound on the red shift has been reported in ref .
@xcite . for the frequency marked with a double ,
observations were made in the frequency range from 14.5 ghz to 17.5 ghz .
we also note which radio detections have lesser significance , so that detections with a significance between 3@xmath129 and 4@xmath129 are marked with a , whereas detections with a significance between 4@xmath129 and 5@xmath129 are marked with a . for time lags marked with a
, we assume the uncertainty in the time lag to be plus or minus one unit in the last significant figure of the reported detection time of a particular radio frequency .
if a radio frequency detection reports an observation time interval , the time lag is calculated from the midpoint of the time interval , and we take one half of the observation time as the uncertainty in the time lag . the ra time reported in these cases
is that of the midpoint of the observation time .
to realize the limits reported in the main body of our paper , we fit to eq .
( 2 ) of that source .
we find that large statistical scale factors must be employed to realize good fits to the data .
this may stem , in part , from the circumburst environment , and , more generally , from time delay effects which arise from neither source effects nor propagation effects across the expanse of space .
the nature of the circumburst environment of grb 050904a is discussed in ref .
@xcite . to study
how our limits rely on the details of our data set , we compare fits with frequency in the source rest frame .
we also study how the fits change with the significance of the included radio observations , with omitting radio observations with reported integration times , and with omitting more poorly determined red shifts .
we also try to study source evolution effects with @xmath3 . as to the frequency dependence
, differences do emerge if we include points at the observed frequency of @xmath130 ghz .
this portion of the data set admits time delays in the grb rest frame of a month and more , and larger scale factors are required to yield fits of comparable quality .
fitting data satisfying @xmath131 ghz , for which with a scale factor of 450 applied to the error in @xmath92 we find @xmath97 with @xmath132 day ghz@xmath94 mpc@xmath95 and @xmath133 day , or @xmath98 day ghz@xmath94 mpc@xmath95 at 95% cl .
were we to fit all the data points we would require a scale factor of 685 to yield a fit of @xmath97 from which we would find @xmath134 day ghz@xmath94 mpc@xmath95 at 95% cl .
we thus restrict our discussion henceforth to the portion of the data set which satisfies @xmath131 ghz .
if we now restrict our fit to points in which the significance of the radio observation is @xmath135 or more , then our limit on the slope is still @xmath98 day ghz@xmath94 mpc@xmath95 at 95% cl , though we must increase the scale factor to 512 to recover a fit of comparable quality .
some of the ra measurements have reported integration times and hence determinable uncertainties , whereas others do not .
repeating the fits without the points with reported integration times yields no significant difference in the fit , though we must increase the scale factor to 467 to yield a fit of comparable quality .
this may well follow as the dropped points do have larger errors and hence play a lesser role in the original fit .
if we now return to our original data set and omit those points for which the determined redshift is a lower limit or merely `` likely , '' the limits do weaken .
for example , with a scale factor of 419 , we find @xmath136 with @xmath137 day ghz@xmath94 mpc@xmath95 and @xmath138 , so that the limit on the slope at 95% cl is roughly a factor of 3 worse .
this can be rationalized as these poorly known redshifts possess large @xmath3 .
finally , to study the effects of cosmological evolution on the grbs , we repeat our fit with @xmath139 and find similar results .
that is , for a scale factor of 539 we determine @xmath140 with @xmath141 day ghz@xmath94 mpc@xmath95 and @xmath142 , to yield @xmath143 day ghz@xmath94 mpc@xmath95 at 95% cl .
thus the effects of cosmological evolution appear to be small .
.gamma ray burst and radio afterglow data from mid-2005 through march , 2009 .
[ cols= " < , < , < , < , < , < " , ]
, s. b. _ et al . _ .
9053 , http://gcn.gsfc.nasa.gov/gcn3/9053.gcn3 ( 2009 ) .
|
the dark - matter candidates of particle physics invariably possess electromagnetic interactions , if only via quantum fluctuations . taken en masse , dark matter
can thus engender an index of refraction which deviates from its vacuum value .
its presence is signaled through frequency - dependent effects in the propagation and attenuation of light .
we discuss theoretical constraints on the expansion of the index of refraction with frequency , the physical interpretation of the terms , and the particular observations needed to isolate its coefficients .
this , with the advent of new opportunities to view gamma - ray bursts at cosmological distance scales , gives us a new probe of dark matter and a new possibility for its direct detection .
as a first application we use the time delay determined from radio afterglow observations of distant gamma - ray bursts to realize a direct limit on the electric - charge - to - mass ratio of dark matter of @xmath0 at 95% cl .
some twenty - three percent of the universe s energy budget is in dark matter @xcite , yet , despite its abundance , little is known of its properties .
a number of methods have been proposed for the detection of dark matter .
such studies typically rely , for direct searches , on dark - matter nucleus scattering , and , for indirect searches , on two - body annihilation of dark - matter to standard model ( sm ) particles ; constraints follow from the nonobservation of the aftermath of particular two - body interactions .
in contrast , we probe dark matter in bulk to infer constraints on its particulate nature .
we search for dark matter by studying the modification of the properties of light upon passage through it .
one can study either polarization @xcite or propagation effects ; we focus here on the latter .
the resulting constraints are most stringent if dark matter consists of sufficiently low mass particles , be they , e.g. , warm thermal relics or axion - like particles , that its number density greatly exceeds that of ordinary matter .
we thus consider dark _
matter_. matter effects are signaled by dispersive effects in the speed or attenuation of light .
we study this by introducing an index of refraction @xmath1 , whose deviation from unity is controlled by the light
dark - matter scattering amplitude in the forward direction , i.e. , the forward compton amplitude , as well as by the angular frequency @xmath2 of the light and the redshift @xmath3 at which the matter is located .
a dark - matter particle need not have an electric charge to scatter a photon ; it need only couple to virtual electromagnetically charged particles to which the photons can couple .
the scattering amplitude is related by crossing symmetry to the amplitude for dark - matter annihilation into two photons , so that any dark - matter model which gives rise to an indirect detection signal in the two - photon final - state @xcite can also drive the index of refraction of light from unity .
its real part is associated with the speed of propagation , and we search for its deviation from unity by searching for frequency - dependent time lags in the arrival of pulses from distant gamma - ray bursts ( grbs ) .
the limits on the non - observation of frequency - dependent effects in the speed of light are severe .
the best limits from terrestrial experiments control the variation in the speed of light @xmath4 with frequency to @xmath5 @xcite , but the astrophysical limits are much stronger .
the arrival time difference of pulses from the crab nebula bound @xmath6 @xcite , and still better limits come from the study of grbs @xcite .
grbs are bright , violent bursts of high - energy photons lasting on the order of thousandths to hundreds of seconds , and their brightness makes them visible at cosmological distances .
time delays which are linear in the photon energy can occur _ in vacuo _ in theories of quantum gravity ; the special features of grbs make them particularly well - suited to searches for such signatures of lorentz violation @xcite .
the detection of photons of up to @xmath7 gev in energy from grb 090510 severely constrains this scenario , placing a lower limit on the energy scale at which such linear energy dependence occurs to @xmath8 gev @xcite .
although vacuum lorentz violation and light dark - matter interactions can each induce dispersive effects , their differing red shift and frequency dependence render them distinct .
a model - independent analysis of the deviation of the refractive index from unity is possible if we assume that the photon energy is small compared to the energy threshold required to materialize the electromagnetically charged particles to which the dark matter can couple . in models of electroweak - symmetry breaking which address the hierarchy problem , a dark - matter candidate can emerge as a by - product . in such models
the inelastic threshold @xmath9 is commensurate with the weak scale , or crudely with energies in excess of @xmath10 , as the new particles are produced in pairs .
if the photon energy @xmath2 satisfies the condition @xmath11 , we can apply the techniques of low - energy physics to the analysis of the forward compton amplitude . under an assumption of lorentz invariance and other symmetries , we expand the forward compton amplitude in powers of @xmath2 and give a physical interpretation to the coefficients of the first few terms as @xmath12 @xcite .
in particular , the term in @xmath13 is controlled by the dark matter particles charge and mass , weighted by preponderance , irrespective of all other considerations save our assumption of lorentz invariance @xcite .
the relationship between the index of refraction @xmath14 and the forward scattering amplitude @xmath15 for light of angular frequency @xmath2 is well - known @xcite , where we relate @xmath15 to the matrix element @xmath16 of quantum field theory to connect to particle physics models of dark matter . using standard conventions @xcite , we determine @xmath17 , in the matter rest frame @xcite , so that @xmath18 and @xmath19 with @xmath20 the mass density of the scatterers and @xmath21 the particle mass . in our analysis
we assume @xmath22 , where @xmath23 is the temperature of the dark matter at the red shift of the observed gamma - ray burst .
since @xmath23 should be a factor of some @xmath24 smaller than @xmath25 at the moment of its production or decoupling , our limits are not restricted to cold dark matter exclusively . moreover , even in the latter case , the candidate mass can be as light , e.g. , as light as @xmath26 ev in the axion model with bose - einstein condensation of ref . @xcite . under the assumptions of causality or ,
more strictly , of lorentz invariance , as well as of charge - conjugation , parity , and time - reversal symmetry in the photon dark - matter interaction , we have @xcite @xmath27 , where @xmath28 is the spin operator associated with the dark - matter particle and @xmath29 ( @xmath30 ) is the polarization vector associated with the photon in its initial ( final ) state .
the functions @xmath31 and @xmath32 are fixed in terms of the dark - matter electric charge and magnetic moment , respectively , as @xmath12 @xcite without further assumption it does not even matter if the dark - matter particle is composite .
the amplitude @xmath33 is implicitly a @xmath34 matrix in the photon polarization , and its diagonal matrix elements describe dispersion in propagation and attenuation @xcite .
the @xmath32 term describes changes in polarization with propagation , so that we need not consider it further . under analyticity and unitarity , expanding @xmath31 for @xmath11 yields a series in positive powers of @xmath35 for which the coefficient of every term of @xmath36 and higher is positive definite @xcite .
thus a term in @xmath14 which is linear in @xmath2 , discussed as a signature of lorentz violation @xcite , does not appear if @xmath37 and the medium is unpolarized . we parametrize the forward compton amplitude as @xmath38 , where @xmath39 @xcite and the dark - matter millicharge is @xmath40 .
the terms in @xmath36 and higher are associated with the polarizabilities of the dark - matter candidate .
dispersive effects in light propagation are controlled by the group velocity @xmath41 , so that the light emitted from a source a distance @xmath42 away has an arrival time of @xmath43 .
for very distant sources we must also take the cosmological expansion into account @xcite , so that as we look back to a light source at redshift @xmath3 , we note that the dark - matter density accrues a scale factor of @xmath44 , whereas the photon energy is blue shifted by a factor of @xmath45 relative to its present - day value @xmath46 @xcite .
thus the light arrival time @xmath47 is @xmath48 with the hubble rate @xmath49 .
we employ the cosmological parameters determined through the combined analysis of wmap five - year data in the @xmath50cdm model with distance measurements from type ia supernovae ( sn ) and with baryon acoustic oscillation information from the distribution of galaxies @xcite .
thus the hubble constant today is @xmath51 , whereas the fraction of the energy density in matter relative to the critical density today is @xmath52 and the corresponding fraction of the energy density in the cosmological constant @xmath50 is @xmath53 @xcite .
we find that the time delay is characterized by powers of @xmath35 and unknown coefficients @xmath54 .
different strategies must be employed to determine them .
the @xmath55 term incurs no frequency - dependent shift in the speed of light , so that to infer its presence one needs a distance measure independent of @xmath3 , much as in the manner one infers a nonzero cosmological constant from type ia supernovae data .
interestingly , as @xmath56 it has the same phenomenological effect as a nonzero cosmological constant ; the longer arrival time leads to an inferred larger distance scale .
cosmologically , though , its effect is very different as it scales with the dark - matter density ; it acts as grey dust .
the remaining terms can be constrained by comparing arrival times for differing observed @xmath46 .
the determination of @xmath57 and @xmath58 require the analysis of the grb light curves at extremely low and high energies , respectively , and probe disjoint dark - matter models . as a first application of our method
, we use radio afterglow data to determine @xmath57 and thus to yield a direct limit on the electric - charge to mass of dark matter .
this quantity gives insight into the mechanism of dark matter stability .
if dark matter possesses an internal symmetry , e.g. , it can not decay to lighter particles and conserve its hidden charge .
such dynamics can also conspire to give dark matter a slight electric charge @xcite , which , no matter how small , reveals the existence of its hidden interactions and the reason for its stability . to determine @xmath57 we consider grbs with known redshift in which a radio afterglow is also detected .
we collect the data and describe the criteria used in its selection in the supplementary material @xcite .
the time lag between the initial detection of the grb at some energy and the detection of the radio afterglow is @xmath59 .
if we first observe the grb at kev energies and compare with the observed arrival time in radio frequencies , then the terms in positive powers of @xmath46 , as well as the term in @xmath60 , are negligible ; we let @xmath61 .
in order to assess reliable limits on @xmath57 we must separate propagation effects from intrinsic source effects .
statistically , we expect time delays intrinsic to the source to be independent of @xmath3 , and the time delay from propagation to depend on @xmath3 and @xmath2 in a definite way .
such notions have been previously employed in searches for lorentz invariance violation @xcite .
we separate propagation and emission effects , respectively , via @xmath62 where @xmath63 depends on the cosmological past through the hubble rate @xmath64 and @xmath65 allows for a frequency - dependent time lag for emission from the grb in the grb rest frame .
the frequency @xmath66 , and @xmath67 contains the millicharge - to - mass ratio @xmath68 , i.e. , @xmath69 with @xmath70 gev/@xmath71 @xcite and @xmath72 the fine - structure constant . to provide a context , we first consider the value of @xmath73 which would result were we to attribute the time lag associated with the radio afterglow of one grb to a propagation effect .
choosing the grb with the largest value of @xmath74 , we have a time lag of @xmath75 day associated with grb 980703a at @xmath76 measured at a frequency of 1.43 ghz . with eq .
( [ lag ] ) , setting @xmath77 , and noting that @xmath78 if the errors in its inputs are uncorrelated , the measured time lag fixes @xmath79 .
since there are no known examples of a radio afterglow preceding a grb , this single time lag in itself represents a conservative limit . turning to our data sample of 53 grbs , we plot the measured time lag versus @xmath74 in fig .
[ fig1 ] and make a least - squares fit of eq .
( [ lag ] ) to determine @xmath67 and @xmath80 .
we require @xmath81 as demanded by our model .
fitting to the points with frequencies of @xmath82 ghz in the grb rest frame , we determine @xmath0 at 95% cl , which is comparable to our limit derived from a single observation of grb 980703a .
the dependence of our fit results on the selected frequency window , as well as the stability of our fits to the significance of the radio afterglow observation , to evolution effects in @xmath3 , and to the more poorly determined red shifts and radio afterglows is discussed in the supplementary information @xcite . determined from the observation of a grb and its radio afterglow plotted as a function of @xmath74 with @xmath83 , employing the data reported in the supplementary material @xcite .
the points correspond to frequency windows of @xmath84 ghz ( @xmath85 , green ) , @xmath86 ghz ( @xmath87 , maroon ) , and @xmath88 ghz ( @xmath89 , blue ) in the grb rest frame .
points with @xmath90 ghz do not appear within the chosen frame of the figure .
the fit of eq .
( [ lag ] ) to the data with @xmath91 ghz with a scale factor in the uncertainty in @xmath92 of 450 , to compensate for environmental effects in the vicinity of the emission from the grb , yields @xmath93 day ghz@xmath94 mpc@xmath95 and @xmath96 day with @xmath97 .
thus @xmath98 day ghz@xmath94 mpc@xmath95 at 95% cl to yield @xmath0 at 95% cl .
the statistical scale factors are not shown explicitly .
for clarity of presentation we display time lags in the grb rest frame of less than 10 days only . ]
we have found a direct observational limit on the dark - matter electric - charge - to - mass ratio .
our study probes for a charge imbalance averaged over cosmological distance scales , without regard to its sign , at distance scales shorter than the wavelengths of the radio observations in our data set .
our bound rules out the possibility of charged `` q - balls '' @xcite of less than 100 kev in mass as dark - matter candidates .
our limit holds regardless of the manner in which the dark matter is produced , though we can compare it to limits arising from the nonobservation of the effects of millicharged particle production . for example , for @xmath99 , they are crudely comparable to the strongest bound from laboratory experiments @xcite , for @xmath100 for @xmath101 @xcite . in comparison
the model - independent bound arising from induced distortions in the cosmic microwave background ( cmb ) radiation is @xmath102 for @xmath103 , though model - dependent constraints reach @xmath104 for @xmath105 @xcite .
cosmological limits also arise from observations of the sunyaev - zeldovich effect for which , e.g. , @xmath106 for @xmath107 @xcite . for these light masses
our limit is stronger , which shows that millicharged particles of such mass and charge are not the primary constituents of dark matter .
limits also arise from stellar evolution and big - bang nucleosynthesis constraints , for which the strongest is @xmath108 for @xmath109 @xcite , as well as from the manner in which numerical simulations of galactic structure confront observations @xcite .
we offer a visual summary of this discussion in fig .
[ fig2 ] .
indirect limits can be evaded : for example , in some models , the dynamics which gives rise to millicharged matter are not operative at stellar temperatures @xcite ; other models evade the galactic structure constraints @xcite . , in units of @xmath110 with candidate mass @xmath21 with limits from other sources stemming from millicharged particle production .
our limit derives from observations of grbs and their radio afterglows and is marked `` grb '' ( solid , black ) , so that the region above that line is excluded .
the condition @xmath111 , as discussed in text , sets the lower endpoint of the solid line . since eq .
( [ lag ] ) persists in qed , for which @xmath112 , we expect our limit to persist for lighter , cold dark matter as well , as indicated by the dashed line .
the strongest laboratory limits , which are for fermions , are marked `` pol '' ( solid , blue ) @xcite , and the strongest limits from induced distortions in the cmb , which are also for fermions , are marked `` cmb '' ( long - dashed , maroon ) the upper curve is the model - independent limit , whereas the lower curve is the model - dependent ( md ) limit @xcite .
constraints on @xmath113 emerge from limits on novel energy - loss mechanisms in stars and supernovae ; such limits also fail to act if @xmath113 is too large .
the limit from plasmon decay in red - giants is marked by `` rg '' ( short - dashed , red ) @xcite , the same limit in white dwarfs is marked by `` wd '' ( dot - dot - dashed , indigo ) @xcite , and the limit from sn 1987a is marked by `` sn '' ( dot - dashed , orange ) @xcite .
the rg limit acts if @xmath114 @xcite .
we have also reported the limit from big - bang nucleosynthesis , marked by `` bbn '' ( dotted , green ) , from ref .
@xcite as well . ]
our limit also significantly restricts the phase space of models with hyperweak gauge interactions and millicharged particles which can arise in string theory scenarios @xcite as viable dark - matter candidates .
we estimate that our limit can be improved considerably before the dispersive effects from ordinary charged matter become appreciable @xcite . the largest such contribution to @xmath67 should come from free electrons .
we estimate the cosmological free electron energy density @xmath115 to be no larger than @xmath116 @xcite , where @xmath117 is the fraction of the energy density in baryons with respect to the critical density today and @xmath118 and @xmath119 are the electron and proton mass , respectively . replacing @xmath120 with @xmath115 and @xmath68 with @xmath121 in @xmath122
we find that our limit would have to improve by @xmath123 before the contribution from free electrons could be apparent .
we set our limit of @xmath0 from existing radio observations at no less than 4 ghz in the grb rest frame , so that our limit is certainly operative if @xmath124 ghz , or crudely , if @xmath125 ev .
studies of the polarizability in qed @xcite , for which @xmath112 , also reveal the analytic structure in @xmath2 we have assumed for the forward compton amplitude . thus we believe our limit to be of broader validity , so that the lower limit on the mass can be less than @xmath126 ev , though it is model dependent and set by the lee - weinberg constraint @xcite , much as the minimum mass of @xmath127 ev is determined in the axion model of ref .
@xcite .
forward scattering is coherent irrespective of whether the photon wavelength is large compared to the interparticle spacing , so that we expect our results to persist in the dilute particle limit as well , as supported by laboratory studies @xcite .
one further comment : at a frequency of 4 ghz
our limit implies that we probe the average net charge of dark matter , with no constraint on its sign , at length scales of no longer than 8 cm .
our limits can be significantly bettered through grb radio afterglow studies at longer wavelengths .
we thank keith olive for an inspiring question and scott dodelson , rene fatemi , wolfgang korsch , and tom troland for helpful comments .
sg would also like to thank stan brodsky for imparting an appreciation of the low - energy theorems in compton scattering and the institute for nuclear theory and the center for particle astrophysics and theoretical physics at fermilab for gracious hospitality .
this work is supported , in part , by the u.s .
department of energy under contract de - fg02 - 96er40989 . 1 d. n. spergel _ et al . _
[ wmap collaboration ] , astrophys .
j. suppl .
* 148 * , 175 ( 2003 ) .
m. tegmark _ et al . _
[ sdss collaboration ] , phys .
rev .
d * 69 * , 103501 ( 2004 ) .
e. komatsu _ et al .
_ [ wmap collaboration ] , astrophys . j. suppl . * 180 * , 330 ( 2009 ) .
s. gardner , phys .
rev .
lett .
* 100 * , 041303 ( 2008 ) .
p. ullio , l. bergstrom , j. edsjo , and c. g. lacey , phys .
rev .
d * 66 * , 123502 ( 2002 ) .
j. l. hall , in _ atomic masses and fundamental constants _ , edited by j. h. sanders and a. h. wapstra ( plenum , new york , 1976 ) , p. 322 .
b. warner and r. nather , nature * 222 * , 157 ( 1969 ) .
b. e. schaefer , phys .
rev .
lett . * 82 * , 4964 ( 1999 ) .
g. amelino - camelia _ et al .
_ , nature * 393 * , 763 ( 1998 ) .
a. a. abdo _
et al . _
[ fermi lat collaboration ] , nature * 462 * , 331 ( 2009 ) .
m. gell - mann , m. l. goldberger , and w. e. thirring , phys .
rev .
* 95 * , 1612 ( 1954 ) . m. l. goldberger , phys . rev . * 97 * , 508 ( 1955 ) .
t. r. hemmert , b. r. holstein , j. kambor , and g. knchlein , phys .
rev .
d * 57 * , 5746 ( 1998 ) .
w. e. thirring , philos . mag . * 41 * , 1193 ( 1950 ) .
e. fermi , _ nuclear physics , revised edition _
( university of chicago press , chicago , il , 1974 ) , p. 202 .
r. g. newton , _ scattering theory of waves and particles , 1st edition _
( mcgraw - hill , new york , 1966 ) , p. 24ff .
m. e. peskin and d. v. schroeder , _ an introduction to quantum field theory _
( addison - wesley publishing company , reading , ma , 1995 ) .
p. sikivie and q. yang , phys .
rev .
lett . * 103 * , 111301 ( 2009 ) .
l. i. lapidus and chou kuang - chao , j. exptl .
theoret .
phys .
( ussr ) * 39 * , 1286 ( 1960 ) [ sov .
phys .
jetp * 12 * , 898 ( 1961 ) ] . s. j. brodsky and j. r. primack , ann .
phys .
* 52 * , 315 ( 1969 ) .
j. r. ellis _ et al .
_ , astrophys .
j. * 535 * , 139 ( 2000 ) .
u. jacob and t. piran , jcap * 0801 * , 031 ( 2008 ) .
b. holdom , phys .
lett .
b * 166 * , 196 ( 1986 ) .
d. feldman , z. liu , and p. nath , phys .
rev .
d * 75 * , 115001 ( 2007 ) .
a. kusenko and p. j. steinhardt , phys .
rev .
lett . * 87 * , 141301 ( 2001 ) .
see appended supplementary information
. i. m. shoemaker and a. kusenko , phys .
rev .
d * 78 * , 075014 ( 2008 ) .
s. davidson , s. hannestad , and g. raffelt , jhep * 0005 * , 003 ( 2000 ) .
m. ahlers _ et al .
_ , phys .
rev .
d * 77 * , 095001 ( 2008 ) .
a. melchiorri , a. polosa , and a. strumia , phys .
lett .
b * 650 * , 416 ( 2007 ) . c. burrage , j. jaeckel , j. redondo , and a. ringwald , jcap * 11 * , 002 ( 2009 ) .
b. a. gradwohl and j. a. frieman , astrophys . j. * 398 * , 407 ( 1992 ) . l. ackerman , m. r. buckley , s. m. carroll , and m. kamionkowski , phys .
rev .
d * 79 * , 023519 ( 2009 ) .
j. l. feng , m. kaplinghat , h. tu , and h. b. yu , jcap * 07 * , 004 ( 2009 ) .
e. mass and j. redondo , phys .
rev .
lett . * 97 * , 151802 ( 2006 ) .
c. p. burgess _ et al .
_ , jhep * 07 * , 073 ( 2008 ) .
l. bombelli and o. winkler , class .
quant .
grav .
* 21 * , l89 ( 2004 ) .
e. llanta and r. tarrach , phys .
lett .
b * 78 * , 586 ( 1978 ) .
b. w. lee and s. weinberg , phys .
rev .
lett . * 39 * , 165 ( 1977 ) .
g. k. campbell _ et al .
_ , phys .
rev .
lett . * 94 * , 170403 ( 2005 ) .
|
little string theories with @xmath8 supercharges @xcite are obtained by decoupling limits of the type ii strings near @xmath1 ns5-branes .
they are non - local theories without gravity . depending on whether we start from type iia or iib ns5-branes , the system has @xmath9 or @xmath10 super - poincare symmetry , respectively
. since ns5-branes are one of the most difficult nonperturbative objects to study in string theory , it would be very desirable to have better understanding on these strings . also , the type iia little string theory has interesting low energy limit given by interacting @xmath9 superconformal field theories .
little strings have similarities with critical strings , and also differences .
the fact that it does not contain gravity would be the main difference , with far - reaching implications .
however , being non - local theories , it inherits from the type ii strings various stringy properties , such as the t - duality .
so after circle compactification , the two little string theories are supposed to be t - dual to each other .
unlike critical strings , noncritical little strings are difficult to study . some approaches to study them are : holographic approach @xcite , discrete lightcone quantization @xcite , the double scaling limit @xcite .
in particular , the dlcq approach considers the little string theory compactified on a small circle , in which one studies a sector with definite momentum which is decoupled from the rest . via t - duality , the dlcq description can be obtained by a large radius compactification of the t - dual strings with definite winding number . in this paper , we study the qfts living on the little strings macroscopically extended on @xmath11 .
they describe 2d decoupled degrees living on these strings at low energy . such theories are studied in detail in the literature @xcite .
one starts from 2d @xmath2 gauge theories , which flow to interacting cfts and describe these strings .
compactifying these strings on large circles , the ground energy is proportional to the radius times the winding quantum number , much larger than the momentum energy scale .
so we can consider a low energy decoupled sector with fixed winding quantum numbers .
they also have direct relevance to the study of dlcq little strings under t - duality , in which the momentum is fixed . in this paper
, we make a modest contribution to constructing such uv theories for the little strings , on the iia side starting from @xmath5 gauge theories .
the system is proposed to flow to a cft with enhanced @xmath12 susy .
compared to the @xmath12 gauge theories discussed in @xcite , the new description has an advantage of manifestly having certain ir symmetries in uv , which is very crucial for computing some protected ir observables such as the elliptic genus .
our @xmath13 uv qfts are similar to those for the self - dual strings of the @xmath9 superconformal field theory , called ` m - strings ' @xcite . with t - duality , the spectrum of the circle compactified theories would be the same for iia and iib little strings .
we would like to probe this t - duality with the above gauge theory descriptions for macroscopic strings . in general , these descriptions are valid only when the compactification radii are large . as the t - duality exchanges the iia and iib radii as @xmath14 , the two gauge theory descriptions will never be simultaneously reliable .
however , one naturally expects that the protected bps spectrum would be reliable all the way to small radii . in this paper
, we study the t - duality of little strings in the bps sector , from the uv gauge theory descriptions . in particular ,
being able to compute the elliptic genus indices on both iia and iib sides , thanks to our new gauge theory descriptions , we can directly compare their bps spectra .
we find , in fugacity expansions to highly nontrivial orders , that the two elliptic genera precisely map to each other via t - duality .
ns5-branes and study the massive spectra . ]
apart from confirming the naturally expected t - duality , our finding is establishing a very nontrivial identity between the elliptic genera computed from the type iia and iib sides , so that alternative expressions can be used to extract various properties which would have been very difficult to see from the other viewpoints .
for instance , we explain in section 5 how one can easily understand the @xmath15 transformation properties of the elliptic genus , for the complex structure and kahler parameters of the torus , by using our t - dual expressions for the elliptic genus .
the rest of this paper is organized as follows . in sections 2 and 3 ,
we explain the 2d gauge theory descriptions of the iib and iia little strings , respectively , and study their elliptic genera . in section 4
, we study the t - duality of the two elliptic genera , as well as extended duality / triality properties . in section 5
, we study the @xmath16 transformation properties of the elliptic genus in various fugacities .
section 6 concludes with brief discussions .
we first consider the type iib little strings , which are the type iib fundamental strings bound to the ns5-branes . at low energy , the world - volume description of iib ns5-branes
is given by 6d maximally supersymmetric yang - mills theory , with @xmath10 supersymmetry and @xmath3 gauge group .
the fields consist of the gauge field @xmath17 , 4 scalar fields @xmath18 , and fermions .
these degrees are provided by the d - strings ending on the ns5-branes .
the bosonic symmetry of the theory is @xmath19 .
@xmath20 is the lorentz symmetry on the ns5-branes , and @xmath21 is the symmetry on their transverse directions , which rotates @xmath22 .
the yang - mills coupling constant is given by @xmath23 fundamental strings form threshold bounds with the ns5-branes .
they are identified as the instanton strings in the 6d sym .
the instanton string tension is given by @xmath24 agreeing with the tension of the fundamental string .
the coupling constant is independent of the 10d string coupling constant , @xmath25 .
so one can take the little string theory limit , in which we take @xmath26 with fixed @xmath27 .
all the gravitational degrees of freedom are decoupled .
we shall consider @xmath28 macroscopically extended little strings , extended along @xmath11 part of @xmath29 .
we are interested in the dynamics of the degrees of freedom supported on these macroscopic strings , decoupled from the rest of the 6d degrees at low energy .
the system of @xmath28 f1 and @xmath1 ns5-branes admit a uv gauge theory description given by a @xmath30 gauge theory with @xmath2 supersymmetry .
the field theory is identical to that living on the d1-d5 system via s - duality , and has been studied extensively in the literature , e.g. @xcite .
this 2d theory at low energy can also be regarded as the worldsheeet description of the instanton strings of the 6d sym theory .
the gauge theory has the @xmath30 @xmath2 vector multiplet , an adjoint hypermultiplet , and @xmath1 fundamental hypermulitiplets which host @xmath3 global symmetry .
these fields are shown in table [ iib - multiplet ] , and more details about this theory is explained in appendix [ appendix:2d - action ] . for later convenience , we also show the supermultiplet structure with respect to the right - chiral @xmath13 susy .
.@xmath2 supermultiplets for @xmath28 iib strings .
[ cols="^,^,^,^,^",options="header " , ] a 2d @xmath5 uv gauge theory is engineered from this brane setting , with supercharges given by @xmath31 .
the fields can be characterized again by a circular quiver of fig .
[ iiaquiver ] .
each circular node involves @xmath5 @xmath32 gauge multipletiplet @xmath33 , and a @xmath5 adjoint hypermultiplet @xmath34 , denoted by the solid lines .
@xmath35 s are the number of the d2-branes suspended between adjacent ns5-branes .
thick dashed lines between two circular nodes denote the bi - fundamental twisted hypermultiplets @xmath36 .
thin dashed lines between two circular nodes denote the bi - fundamental fermi multiplets @xmath37 .
d6-brane introduces extra fields : fundamental hyper multiplets @xmath38 and fermi multiplets @xmath39 .
these fields are are summarized in table .
[ multiplet - iia ] . as explained in the previous paragraph , and just like @xcite , the chemical potentials for @xmath40 and @xmath41 are locked as @xmath42 , so that one just has one @xmath43 .
compared to the previous @xmath12 gauge theory for the iia strings , the @xmath13 fields on the first and third lines of table .
[ multiplet - iia ] are forming the @xmath12 vector multiplet , which we decomposed as above since the system does not preserve @xmath12 susy .
also , the fields on the fourth and fifth lines form the previous @xmath12 hypermultiplet .
they again make a twisted higgs branch , which represents the degrees of freedom of fully winding d2-branes leaving the ns5-branes . the coulomb branch of the @xmath12 theory is replaced here by the higgs branch formed by the second and third lines , which is our main interest to study the iia little strings
. the susy action of the @xmath13 gauge theory can also be easily constructed . from the @xmath44 supersymmetric formalism
, one has to determine the holomorphic potentials @xmath45 , @xmath46 for each fermi multiplet @xmath47 , ensuring the @xmath13 susy enhancement .
for instance , see @xcite for how this can be done . here , following @xcite , we simply write down these potentials for our theory . let us call the @xmath44 fermi multiplet from the @xmath13 vector multiplet as @xmath48 , which is made of @xmath49 and @xmath50 . then one should first take @xmath51-\xi_{\mathbb{c}}\ , \ \
e_{\lambda_i}=\phi_{i+1}\tilde{\phi}_{i+1}-\tilde\phi_{i}\phi_{i}\ , \ ] ] for @xmath13 susy @xcite . here and below
, we use the chiral superfield notation @xmath52 , @xmath53 , @xmath54 for a while .
we also inserted the fi parameter @xmath55 for later use , which corresponds to turning on worldvolume @xmath56 field on @xmath57 directions .
the above @xmath58 should be accompanied by other @xmath58 functions for other fermi fields , to satisfy @xmath59 after summing over all fermi multiplets @xmath47 .
this is another requirement from susy . to meet the last condition
, one should turn on the following potentials for other fermi multiplet fields : @xmath60 the bosonic potential is @xmath61 with @xmath62 given by @xmath63 + [ \tilde{b}_i,\tilde{b}_i^\dag]-\phi_i^\dag\phi_i+\tilde\phi_i\tilde\phi_i^\dag + \phi_{i+1}\phi_{i+1}^\dag-\tilde\phi_{i+1}^\dag\tilde\phi_{i+1}-\xi_{\mathbb{r}}\ .\ ] ] after some rearrangement , one obtains @xmath64-\xi^m\right)^2+\left ( ( \sigma^i)^{a}_{\ \ b}\phi_{ia}\bar\phi_i^b- ( \sigma^i)^{a}_{\ \ b}\bar\phi_{i-1}^b\phi_{i-1,a}\right)^2\nonumber\\ & & \hspace{1.3cm}\left.+|\phi_{ia}q_{i\dot\alpha}|^2 + |\phi^\dag_{i+1,a}q_{i\dot\alpha}|^2+|\phi_{ia}a_{i\alpha\dot\beta}- a_{i-1,\alpha\dot\beta}\phi_{ia}|^2\right]\end{aligned}\ ] ] where @xmath65 and @xmath66 , with manifest @xmath67 symmetry .
note that with nonzero @xmath68 , @xmath69 fields are required to be nonzero at low energy , which lift the twisted higgs branch of @xmath70 .
namely , even if @xmath71 , they can not combine and leave the ns5-branes unlike the @xmath2 model . also , the previous @xmath12 coulomb branch fields @xmath72 form @xmath13 higgs branch fields , together with new degrees @xmath73
. the @xmath13 setting will thus be computing the correct little string elliptic genus .
however , the @xmath13 elliptic genus will also capture a subtle trace of the presence of a d6-brane from the sector with @xmath71 , in which case the d2-branes make full windings along @xmath74 .
this can be easily accounted for and factored out , after which we shall be obtaining the iia little string index .
we shall explain this in section 4 .
we define the index of iia little string theory wrapping a spatial circle along @xmath75 , as follows , @xmath76\;.\ ] ] @xmath77 are charges of the self - dual tensor fields , supported on each m5-brane , with the chemical potentials , @xmath78 .
@xmath79 are the fugacity variable counting the number of momentum , and @xmath80 the winding fugacity of the iia little strings .
@xmath81 are the positions of the 5-branes along @xmath74 .
an m2-brane suspended between the @xmath82th interval between the m5-branes , @xmath83 , carries nonzero charges @xmath84 and @xmath85 .
the charges @xmath86 form the simple roots of the @xmath87 algebra , for @xmath88 .
the last charge @xmath89 , is accompanied by an extra winding , so the whole @xmath1 roots become the simples roots of @xmath4 .
the fugacity variables corresponding to these simple roots are given by @xmath90 where @xmath91 .
for convenience , we introduce @xmath92 , where @xmath93 . for large @xmath94 ,
the low energy degrees living on the winding strings decouple from the 6d degrees on the 5-branes .
so the index of the iia little string theory on @xmath95 factorizes as @xmath96 @xmath97 comes from the momenta on @xmath1 separated m5-branes wrapping a spatial circle .
unlike the iib perturbative index @xmath98 which had massive w - boson contributions , the iia 5-brane does not have extra massive particle states in it ( because it only has strings ) .
so this contribution should factorize into @xmath1 single 5-brane contributions .
it can be computed either from @xmath1 abelian tensor multiplet , or equivalently from the multiple d0-brane index bound to a single d4-brane @xcite .
the result is @xmath99\;,\ ] ] where @xmath100 with @xmath101 .
the contribution @xmath102 comes from the elliptic genera of the @xmath13 gauge theory theory that we have explained in the previous subsection .
this elliptic genus can be computed by the contour integral using the jeffrey - kirwan residues @xcite , or the refined topological vertex method with @xmath103-fivebrane web obtained by t - dualizing the branes along @xmath104 @xcite . by summing up the elliptic genera over all possible @xmath35 numbers , one obtains @xmath105 .
the result is labeled by sets of @xmath1 young diagrams , @xmath106 , where @xmath107 , @xmath108 @xmath109 is the elliptic genus with fixed @xmath35 s , which is given by @xmath110 where @xmath111 @xmath112 denotes the position of each box in a young diagram .
@xmath113 is the length of the @xmath114th row of the young diagram @xmath115 .
@xmath116 is the length of the @xmath114th column of @xmath115 . as we emphasized earlier in this section ,
the contribution from the @xmath12 higgs branch ( or the @xmath13 twisted higgs branch ) formed by @xmath117 is not completely decoupled .
we shall explain at the beginning of section 4 what contribution we expect to get from this decoupled sector .
the iib little string theory on a circle is supposed to be t - dual to iia little string theory on the dual circle , with the radiii related by @xmath118 .
the winding iib little strings on @xmath119 is dual to the momentum on @xmath120 , and vice versa .
their bps masses agree with each other , since @xmath121 the fractional momenta of iib little string theory are dual to the fractional winding numbers of iia little strings , @xmath122 where @xmath91 .
t - duality between two little string theories is demonstrated by fig .
[ t - duality ] .
t - duality between iia and iib little string theories would naively imply @xmath123 as we stated at the end of section 3.1 , using an alternative 2d @xmath13 gauge theory to compute the iia elliptic genus ( and thus the iia little string index ) will leave a subtle trace of the fact that we made a uv deformation of the gauge theory by putting an extra d6-brane .
a spectrum change will happen in a sector with full wound d2-branes along @xmath74 , namely with states carrying the factors of fugacities @xmath124 but not @xmath125 s .
let us first explain this small subtlety in our iia calculation . on the iib side ,
consider the single particle states with zero electric charges ( no dependence on @xmath125 ) and zero winding ( @xmath126 order ) .
this contribution is contained in the @xmath127 factor . in particular
, it comes from the @xmath1 cartan modes of the 6d @xmath3 sym .
their partition function is given by @xmath128\ .\ ] ] in the iia side , this will correspond to a sector with fully wound d2-branes along @xmath74 , at @xmath129 order with @xmath71 . at @xmath130 in section 3.1
, there is an extra twisted higgs branch which meets the higgs branch of our interest .
the former sector will represent the little strings leaving ns5-branes .
the two sectors would decouple in ir , but the 2d gauge theory contains both in its hilbert space .
now by turning on the fi term @xmath68 , the continuum of twisted higgs branch will be lifted . however , after turning on @xmath68 , it often happens that there appear extra bound states of the continuum degrees with the remaining 2d strings of our interest .
for instance , see @xcite and references therein for many occasions in which extra bound states occur at nonzero fi parameters . from the @xmath13 computation of the iia side , we shall find @xmath131 = pe\left[ni_+(\epsilon_{1,2},m)\frac{w^\prime}{1-w^\prime}\right]\cdot z_{\rm extra}(w^\prime)\ ] ] in the same sector , instead of ( [ iib - neutral ] ) , where @xmath132 .
we shall give an account for why @xmath133 should be appearing due to our @xmath13 deformation of the uv theory . with this understood
, we should define the true iia index as the expression computed in section 3.2 divided by @xmath133 .
we call this @xmath134 we shall find that @xmath135 which we checked for the cases with @xmath136 . this will establish the t - duality of the strong - coupling little string spectra via the elliptic genus calculus . we first explain ( or at least heuristically understand ) how @xmath133 would be appearing in our @xmath13 calculus . consider the sector with @xmath137 , forming @xmath138 full winding branes which have the right quantum number to leave the ns5-branes .
we weight @xmath138 windings by @xmath139 , and relax the constraint on fixed @xmath138 .
we would like to count the bps bounds of these strings with @xmath1 ns5-branes directly , not using the elliptic genus formula of @xcite .
we shall do so without and with one d6-brane , to clearly compare . for convenience , we t - dualize along the @xmath74 circle , and obtain many d1-branes along @xmath140 , @xmath1-centered taub - nut on @xmath141 and @xmath142 circle , and optionally a d5-brane along @xmath143 with @xmath56 ( fi term ) on @xmath57 .
firstly , without d5 , any number @xmath138 of wrapped d1-branes can form a bound state of multiply wound single string .
for each massive particle of this sort , we study its ground state wavefunction on the @xmath1-centered taub - nut .
this space has @xmath1 normalizable harmonic forms , so that there could be @xmath1 possible bound states of the original @xmath1 ns5-branes with this particle .
the index for this particle is thus @xmath144 , where @xmath1 comes from @xmath1 normalizable harmonic forms .
the @xmath145 factor appears because this is exactly the same type of bound states as the half - bps w - bosons in sym , as in the iib setting .
summing over @xmath138 and considering the multi - particle hilbert space , one exactly obtains ( [ iib - neutral ] ) with @xmath146 replaced by @xmath124 . note that we have arrived at this conclusion by a direct counting , without any deformation by continuous parameters , so this should be part of the iia little string index .
now we consider the same problem after placing one d5-brane with fi parameter ( @xmath56 background ) .
the setting of section 3 was that d6 and @xmath1 ns5-branes are placed at the same point of @xmath147 in the @xmath141 directions .
now t - dualizing along @xmath74 , one finds a d5-brane on top of the @xmath148 singularity of unresolved taub - nut .
now , among the @xmath1 normalizable harmonic forms of @xmath1-centered taub - nut , @xmath149 of them are supported at the @xmath150 singularity , where d5 is sitting . since the fully winding d1-branes are forced to be bound to d5 at the tip due to the fi parameter , d1-branes stuck to d5 can still assume one of these @xmath149 bound state wavefunctions .
the multi - particle index of the bounds is @xmath151 $ ] .
however , the last normalizable harmonic form of taub - nut is not localized at the tip , so d1-branes confined to d5 can not be in this bound state .
( the forbidden wavefunction is in the twisted higgs branch . )
this accounts for the second term on the left hand side of ( [ iia - neutral ] ) .
now , note that @xmath138 d1-branes can also form threshold bounds with single d5-brane , whose partition function is given by @xmath152 $ ] @xcite .
( this extra contribution is also from the twisted higgs branch , since the d2-d6 bounds still exist after displacing d6-ns5 s . )
this explains the first term of ( [ iia - neutral ] ) , and thus the origin of @xmath133 . by the discussions of this paragraph
, it clearly comes from having d6-brane and nonzero fi parameter , causing extra bound states or destroying some in the twisted higgs branch . so with this understood
, t - duality would imply ( [ t - duality - final ] ) .
we start by considering the index of the @xmath153 iib theory , although this should be a free theory .
the perturbative contribution is given by @xmath154.\ ] ] the @xmath153 instanton string partition function is given by @xmath155 where @xmath156 with @xmath157 the full index of the @xmath153 theory is given by @xmath158 to further explain this index , consider the single instanton string index given by @xmath159 where @xmath160 . in terms of @xmath161 , we find that the multi - instanton string index is given by the hecke transformation of @xmath162 , @xmath163\ ] ] where @xmath164 .
this is checked up to high orders in @xmath165 and @xmath146 .
the partition function given by the hecke transformation appears , for instance , in conformal field theories on symmetric product target spaces .
this is closely related to the fact that the moduli - space of @xmath153 multi - instantons is a symmetric product of the single instanton moduli space @xmath166 .
more precisely , the symmetric product cft was suggested to be the theory at nonzero world - sheet theta angle @xmath167 @xcite .
since the elliptic genus would be insensitive to the continuous parameters , away from @xmath168 , @xmath169 , it is natural to have ( [ hecke ] ) . on the iia side
, the 2d @xmath5 quiver gauge theory itself has an enhanced @xmath2 susy , and becomes precisely the same to the 2d @xmath2 adhm gauge theory for iib strings .
therefore , @xmath170 the extra factor @xmath171 on the iia side is given by @xmath172= z^{\rm iib}_{\rm pert}(\epsilon_\pm , m , q^\prime ) z_{\rm extra}(q^\prime)\ .\ ] ] so the t - duality relation ( [ t - duality - final ] ) is equivalent to @xmath173 being invariant under the exchange of @xmath79 and @xmath80 .
the last property is in fact true , which can be understood as the geometric duality of the 5-brane web obtained by t - dualizing our iia brane setting along @xmath104 @xcite , as shown in fig .
[ fig : rank1 - 5branes ] . fivebranes web dual to rank 1 little string theory .
b ) triality between three khler parameters , @xmath174 , @xmath175 , and @xmath176.,width=529 ] if we write @xmath173 as @xmath177\;,\ ] ] @xmath178 is given by @xmath179 { \nonumber}\\ & + q'^{3}+w'^{3}+(q'^{2}w'+q'w'^{2})\left[t^2 u^2+\frac{t^2}{u^2}+\frac{u^2}{t^2}+\frac{1}{t^2 u^2}+t^2+\frac{1}{t^2}-t u^2 y-\frac{t y}{u^2}-\frac{t u^2}{y}-\frac{t}{u^2 y } \right.{\nonumber}\\ & -\frac{y}{t u^2}-\frac{u^2}{t y}-\frac{1}{t u^2 y}-\frac{u^2 y}{t}+t u+\frac{t}{u}+\frac{1}{t u}+\frac{u}{t}-2 t y-\frac{2 t}{y}-\frac{2}{t y}-\frac{2 y}{t}+2 u^2+\frac{2}{u^2}-u y-\frac{y}{u } { \nonumber}\\ & \left.-\frac{u}{y}-\frac{1}{u y}+y^2+\frac{1}{y^2}+4\right]+(q'^{4}+w'^{4})+(q'^{3}w'+q'w'^{3})\left[t^3 u^3+\frac{t^3}{u^3}+\frac{u^3}{t^3}+\frac{1}{t^3 u^3}+t^3 u+\frac{t^3}{u } \right .
{ \nonumber}\\ & + \frac{u}{t^3}+\frac{1}{t^3 u}-t^2 u^3 y-\frac{t^2 y}{u^3}-\frac{t^2 u^3}{y}-\frac{t^2}{u^3 y}-\frac{u^3 y}{t^2}-\frac{y}{t^2 u^3}-\frac{u^3}{t^2 y}-\frac{1}{t^2 u^3 y}+t^2 u^2+\frac{t^2}{u^2}+\frac{u^2}{t^2 } { \nonumber}\\ & + \frac{1}{t^2 u^2}-2 t^2 u y-\frac{2 t^2 y}{u}-\frac{2 t^2 u}{y}-\frac{2 t^2}{u y}-\frac{2 u y}{t^2}-\frac{2 y}{t^2 u}-\frac{2 u}{t^2 y}-\frac{2}{t^2 u y}+2 t^2+\frac{2}{t^2}+2 t u^3+\frac{2 t}{u^3 } { \nonumber}\end{aligned}\ ] ] @xmath180 { \nonumber}\\ & + ( q'^{2}w'^{2})\left[u^4 t^4+u^2 t^4+\frac{t^4}{u^2}+\frac{t^4}{u^4}+t^4+u^3 t^3 + 2 u t^3-u^4 y t^3 - 2 u^2 y t^3 - 2 y t^3+\frac{2 t^3}{u}-\frac{2 y t^3}{u^2 }
\right . { \nonumber}\\ & + \frac{t^3}{u^3}-\frac{y t^3}{u^4}-\frac{u^4 t^3}{y}-\frac{2 u^2 t^3}{y}-\frac{2 t^3}{y}-\frac{2 t^3}{u^2 y}-\frac{t^3}{u^4 y}+2 u^4 t^2 + 7 u^2 t^2+u^2 y^2 t^2+\frac{y^2 t^2}{u^2}+y^2 t^2 { \nonumber}\\ & -2 u^3 y t^2 - 5 u y t^2-\frac{5 y t^2}{u}+\frac{7 t^2}{u^2}-\frac{2 y t^2}{u^3}+\frac{2 t^2}{u^4}-\frac{2 u^3 t^2}{y}-\frac{5 u t^2}{y}-\frac{5 t^2}{u y}-\frac{2 t^2}{u^3 y}+\frac{u^2 t^2}{y^2}+\frac{t^2}{y^2 } { \nonumber}\\ & + \frac{t^2}{u^2 y^2}+9 t^2 + 5 u^3 t+u^3 y^2 t+4 u y^2 t+\frac{4 y^2 t}{u}+\frac{y^2 t}{u^3}+15
u t - u^4 y t-7 u^2 y t-12 y t+\frac{15 t}{u } { \nonumber}\\ & -\frac{7 y t}{u^2}+\frac{5 t}{u^3}-\frac{y t}{u^4}-\frac{u^4 t}{y}-\frac{7 u^2 t}{y}-\frac{12 t}{y}-\frac{7 t}{u^2 y}-\frac{t}{u^4 y}+\frac{u^3 t}{y^2}+\frac{4 u t}{y^2}+\frac{4 t}{u y^2}+\frac{t}{u^3 y^2 } + \frac{2 u^4}{t^2 } { \nonumber}\\ & + \frac{u^4}{t^4}+2 u^4+\frac{u^3}{t^3}-u y^3+\frac{7 u^2}{t^2}+\frac{u^2}{t^4}+12 u^2+\frac{u^2 y^2}{t^2}+2 u^2 y^2+\frac{y^2}{t^2}+\frac{4 y^2}{t u}+\frac{2 y^2}{u^2}+\frac{y^2}{t^2 u^2}+\frac{y^2}{t u^3 } { \nonumber}\\ & + 5 y^2+\frac{2 u}{t^3}-4 u^3 y-14 u y-\frac{2 u^3 y}{t^2}-\frac{5 u y}{t^2}+\frac{9}{t^2}-\frac{u^4 y}{t^3}-\frac{2 u^2 y}{t^3}-\frac{2 y}{t^3}+\frac{1}{t^4}-\frac{y^3}{u}-\frac{14 y}{u}+\frac{15}{t u } { \nonumber}\\ & -\frac{5 y}{t^2 u}+\frac{2}{t^3 u}+\frac{12}{u^2}-\frac{7 y}{t u^2}+\frac{7}{t^2 u^2}-\frac{2 y}{t^3 u^2}+\frac{1}{t^4 u^2}-\frac{4 y}{u^3}+\frac{5}{t u^3}-\frac{2 y}{t^2 u^3}+\frac{1}{t^3 u^3}+\frac{2}{u^4}-\frac{y}{t u^4}+\frac{2}{t^2 u^4 } { \nonumber}\\ & -\frac{y}{t^3 u^4}+\frac{1}{t^4 u^4}-\frac{4 u^3}{y}-\frac{14 u}{y}-\frac{u^4}{t y}-\frac{7 u^2}{t y}-\frac{12}{t y}-\frac{2 u^3}{t^2 y}-\frac{5 u}{t^2 y}-\frac{u^4}{t^3 y}-\frac{2 u^2}{t^3 y}-\frac{2}{t^3 y}-\frac{14}{u y}-\frac{5}{t^2 u y } { \nonumber}\\ & -\frac{7}{t u^2 y}-\frac{2}{t^3 u^2 y}-\frac{4}{u^3 y}-\frac{2}{t^2 u^3 y}-\frac{1}{t u^4 y}-\frac{1}{t^3 u^4 y}+\frac{2 u^2}{y^2}+\frac{5}{y^2}+\frac{u^3}{t y^2}+\frac{4 u}{t y^2}+\frac{u^2}{t^2 y^2}+\frac{1}{t^2 y^2}+\frac{4}{t u y^2 } { \nonumber}\\ & + \frac{2}{u^2 y^2}+\frac{1}{t^2 u^2 y^2}+\frac{1}{t u^3 y^2}-\frac{u}{y^3}-\frac{1}{u y^3}+22+\frac{5 u^3}{t}+\frac{u^3 y^2}{t}+\frac{4 u y^2}{t}+\frac{15 u}{t}-\frac{u^4 y}{t}-\frac{7 u^2 y}{t } { \nonumber}\\ & \left.-\frac{12 y}{t } \right]+ \cdots.\end{aligned}\ ] ] where @xmath181 , @xmath182 , @xmath183 .
we checked the symmetry of @xmath184 exchange up to 5th orders in @xmath79 and @xmath80 .
furthermore , defining the following variables , @xmath185 triality of exchanging @xmath186 has been discovered in @xcite .
this is also a geometric duality of fig .
[ fig : rank1 - 5branes ] .
triality is simply realized on the universal covering of the torus , as a subgroup of @xmath187 duality . to deal with @xmath186 in equal footing ,
we redefine the index , including extra perturbative contributions at @xmath188 , as @xmath189z_{\rm iia}\ .\ ] ] @xmath190 is given by @xmath191 writing @xmath192 as @xmath193,\ ] ] @xmath194 is given by @xmath195 reconfirming the expected triality of @xcite . it is curious to note that the triality implies the t - duality of iia / iib strings .
the index of @xmath196 iib little string theory is given by @xmath197 where @xmath198 = pe\left[i_{+}\frac{v_{1}+v_{2}+2v_{1}v_{2}}{1-v_{1}v_{2}}\right],\ ] ] with @xmath199 , and @xmath200 .
@xmath201 is given by eq.([i+ ] ) .
@xmath202 is given by @xmath203 @xmath204 is obtained from eq.([instanton ] ) .
expanding @xmath205 with @xmath165 , @xmath206 , @xmath207 , one obtains @xmath208 the index for the rank 2 iia little string theory is given by @xmath209 where we inserted @xmath210 , @xmath211 .
@xmath212 is given by @xmath213 = pe\left[\frac{v_{1}v_{2}}{1-v_{1}v_{2}}\right]\;.\ ] ] @xmath214 is given by @xmath215\;.\ ] ] @xmath216 takes the form of @xmath217 note that @xmath218 , from the symmetry of the quiver .
@xmath219 can be easily obtained from ( [ iiastring ] ) .
for instance , @xmath220 @xmath221 we write the indices of the iia / iib little string theories as @xmath222\;,\ ] ] @xmath223\;,\ ] ] where @xmath224 is given by eq.([icom ] ) .
the coefficients @xmath225 are polynomials of @xmath181 , @xmath226 , and @xmath227 .
it is easily checked that @xmath228 .
t - duality implies that @xmath229 .
we check t - daulity by comparing @xmath230 and @xmath231 .
we checked the agreements for @xmath232 @xmath233 @xmath234 @xmath235 @xmath236 @xmath237 @xmath238 @xmath239 @xmath240 @xmath241 @xmath242 and further up to @xmath243 .
let us define the following variables , @xmath244 we can check @xmath245 exchange symmetry of the index .
this is an analog of the triality exchanging @xmath246 at @xmath247 .
it can be understood as a geometric duality of the dual @xmath103-fivebrane web diagram fig.[fig : rrank2 - 5branes ] of the rank 2 little string theory .
fivebranes web dual to rank 2 little string theory , width=188 ] namely , let us define the index @xmath248z_{\rm iia}\ .\ ] ] we find that @xmath192 is invariant under the @xmath245 exchange , to some high orders in fugacities .
the index of @xmath249 iib little string theory is given by @xmath250 with @xmath251 { \nonumber}\\ & \times pe\left[i_{+}\left(v_{1}+v_{1}^{-1}+v_{2}+v_{2}^{-1}+v_{1}v_{2}+v_{1}^{-1}v_{2}^{-1}\right)\frac{q}{1-q}\right ] { \nonumber}\\ & = pe\left[i_{+}\frac{v_{1}+v_{2}+v_{3}+v_{1}v_{2}+v_{1}v_{2}+v_{2}v_{3}+3v_{1}v_{2}v_{3}}{1-v_{1}v_{2}v_{3}}\right].\end{aligned}\ ] ] where @xmath252 , @xmath253 , and @xmath254 .
@xmath204 s appearing in @xmath202 are obtained from eq.([instanton ] ) .
the index of the rank 3 iia little string theory is given by @xmath255 @xmath212 is given by @xmath256 = pe\left[\frac{v_{1}v_{2}v_{3}}{1-v_{1}v_{2}v_{3}}\right]\ ] ] @xmath214 is given by @xmath257\;.\ ] ] note that @xmath258 is invariant under the permutation of @xmath259 , @xmath260 and @xmath261 , from the symmetry of the quiver .
the elliptic genera of the iia fractional little strings , @xmath258 , are obtained from eq.([iiastring ] ) .
for example , @xmath262 @xmath263
@xmath264 we write the iib / iia indices as @xmath265\;,\ ] ] @xmath266\;,\ ] ] where @xmath224 is given by eq.([icom ] ) . the coefficients @xmath267 are polynomials of @xmath268 , @xmath269 , and @xmath176 , satisfying @xmath270 .
t - duality implies that @xmath271 .
we checked this for higher orders of the fugacity variables .
@xmath272 are given by @xmath273 @xmath274 @xmath275 @xmath276 @xmath277 @xmath278 @xmath279 @xmath280 @xmath281 @xmath282 @xmath283 @xmath284 @xmath285 and so on .
for the rank 3 indices , we can also check the duality of exchanging @xmath286 and @xmath176 .
the index of the winding iib little strings is given by @xmath288 where @xmath289 @xmath146 and @xmath165 are given by @xmath290 where @xmath291 is the complex structure on the torus , @xmath292 is the khler parameter of it , and @xmath293 is the radius of the temporal circle .
the modular transformation of the jacobi s theta function is given by @xmath294 using this property , the s - duality transformation of @xmath202 in @xmath295 is given by @xmath296 transforming the fugacity variable for the winding number , @xmath165 by @xmath297 @xmath298 is invariant under the following transformation , @xmath299 the elliptic genus of the iia strings is given by @xmath300 where @xmath301 and @xmath302 .
@xmath303 is given by @xmath304 where @xmath305 upon t - duality transformation , the complex structure and the khler parameter are exchanged .
the modular transformation of @xmath303 in @xmath306 is given by @xmath307 \cdot z^{(n_{1},\cdots , n_{n})}_{\mathrm{string}}\left(\frac{\epsilon_{\pm}}{\rho},\frac{m}{\rho};-\frac{1}{\rho}\right ) \;.\end{aligned}\ ] ] where @xmath308 . via t - duality relation
, this would imply a definite s - duality transformation of @xmath309 in @xmath310 , which would have been difficult to obtain directly without knowing the t - dual expression .
note that the above s - duality transformation becomes paraticularly simpler when @xmath71 : @xmath311 \cdot z^{(n_{1},\cdots , n_{n})}_{\mathrm{string}}\left(\frac{\epsilon_{\pm}}{\rho},\frac{m}{\rho};-\frac{1}{\rho}\right ) \;.\ ] ] the prefactor can be absorbed into a scaling of @xmath211 fugacity , conjugate to @xmath138 .
this expression might be useful to understand the dlcq of type iib little strings , in which @xmath312 gauge theory description was used @xcite .
in this paper , we explored the 2 dimensional @xmath2 and @xmath5 gauge theory descriptions of macroscopic iia / iib little strings .
in particular , we proposed a new @xmath13 gauge theory which enables the computation of the iia strings elliptic genera .
we used these elliptic genera to study the little string t - duality .
the elliptic genus is enjoying @xmath15 symmetry on the complex structure @xmath295 and kahler parameter @xmath310 of the torus .
interesting extended dualities were studied in @xcite for 6d maximal sym theory compactified on @xmath313 , from its seiberg - witten curve .
it will be interesting to see whether a larger duality than what we explored here is realized in the elliptic genera .
it will also be interesting to see if one can study the t - duality of elliptic genera for the heterotic little string theories , living on the heterotic 5-branes in the @xmath314 and @xmath315 theories . just like
our iia strings are closely related to the ` m - strings ' of 6d @xmath9 cft , the @xmath315 little strings would be closely related to the so - called e - strings of the 6d @xmath316 cft , with @xmath317 global symmetry @xcite .
the e - string elliptic genera have been recently studied in @xcite , from 2d @xmath13 gauge theories .
finaly , the self - dual string elliptic genera in 6d scfts turn out to be related to other interesting observables , such as the superconformal indices @xcite .
it will be interesting to see if the elliptic genera for little strings also find similar interesting applications .
0.5 cm 0.2 cm this work is supported in part by the national research foundation of korea ( nrf ) grants no .
2012r1a1a2042474 ( jk , sk ) , 2012r1a2a2a02046739 ( sk ) , 2006 - 0093850 ( kl ) , 2009 - 0084601 ( kl ) .
the lagrangian of the 2d @xmath318 gauge theory for iib strings is given by @xmath319 @xmath320 is given by @xmath321^{2 } \right .
{ \nonumber}\\ & + \frac{i}{2}(\bar{\lambda}_{a}^{\dot{\alpha}})^{\dag}(d_{t}+d_{s})\bar{\lambda}^{\dot\alpha}_{a } + \frac{i}{2}(\bar{\lambda}^{a \dot{\alpha}})^{\dag}(d_{t}-d_{s})\bar{\lambda}^{a \dot{\alpha } } + \frac{i}{2}(\lambda^{a}_{\alpha})^{\dag}(d_{t}+d_{s})\lambda^{a}_{\alpha } + \frac{i}{2}(\lambda_{a \alpha})^{\dag}(d_{t}-d_{s})\lambda_{a \alpha } { \nonumber}\\ & + \frac{1}{2}d^{i}d^{i } - d^{i}\big(\bar{q}^{\dot{\alpha}}q_{\dot{\beta}}(\tau^{i})^{\dot\alpha}{}_{\dot\beta}+\frac{1}{2}(\tau^{i})^{\dot{\alpha}}{}_{\dot{\beta}}[a^{\dot{\beta}\alpha},a_{\alpha \dot{\alpha}}]-\zeta^{i}\big ) + \frac{1}{2}d^{i'}d^{i ' } - d^{i'}\big ( \frac{1}{2}(\tau^{i'})^{a}{}_{b}[\varphi^{ba},\varphi_{a a } ] \big ) { \nonumber}\\ & -\frac{i}{\sqrt{2}}(\lambda_{a \alpha})^{\dag}[a_{\alpha \dot{\beta}},\bar{\lambda}^{\dot{\beta}}_{a}]-\frac{i}{\sqrt{2}}(\lambda^{a}_{\alpha})^{\dag}[a_{\alpha \dot{\beta}},\bar{\lambda}^{a \dot{\beta } } ] + \frac{i}{\sqrt{2}}(\bar\lambda^{a \dot{\alpha}})^{\dag}[a^{\dot\alpha \beta},\lambda^{a}_{\beta } ] + \frac{i}{\sqrt{2}}(\bar\lambda_{a}^{\dot{\alpha}})^{\dag}[a^{\dot\alpha \beta},\lambda_{a \beta } ] { \nonumber}\\ & \left.+ \frac{i}{\sqrt{2}}(\bar{\lambda}^{\dot\alpha}_{a})^{\dag}[\varphi_{a a } , \bar{\lambda}^{a \dot{\alpha } } ] + \frac{i}{\sqrt{2}}(\bar{\lambda}^{a \dot\alpha})^{\dag}[\varphi^{aa } , \bar{\lambda}^{\dot{\alpha}}_{a } ] - \frac{i}{\sqrt{2}}(\lambda_{a \alpha})^{\dag}[\varphi_{aa},\lambda^{a}_{\alpha}]-\frac{i}{\sqrt{2}}(\lambda^{a}_{\alpha})^{\dag}[\varphi^{aa},\lambda_{a \alpha } ] \right ] .{\nonumber}\\\end{aligned}\ ] ] @xmath322 is given by , @xmath323\;.\end{aligned}\ ] ] in the higgs branch , the theory describes iib strings bound to the ns5-branes , whose target space is the @xmath28 instanton moduli space .
m. berkooz , m. rozali and n. seiberg , phys .
b * 408 * , 105 ( 1997 ) [ hep - th/9704089 ] .
n. seiberg , phys .
b * 408 * , 98 ( 1997 ) [ hep - th/9705221 ] .
o. aharony , m. berkooz , d. kutasov and n. seiberg , jhep * 9810 * , 004 ( 1998 ) [ hep - th/9808149 ] .
a. giveon , d. kutasov and o. pelc , jhep * 9910 * , 035 ( 1999 ) [ hep - th/9907178 ] . o. aharony , class . quant . grav .
* 17 * ( 2000 ) 929 [ hep - th/9911147 ] .
e. witten , jhep * 9707 * , 003 ( 1997 ) [ hep - th/9707093 ] .
o. aharony and m. berkooz , jhep * 9910 * , 030 ( 1999 ) [ hep - th/9909101 ] .
e. witten , commun .
phys . * 109 * , 525 ( 1987 ) .
r. flume and r. poghossian , int .
j. mod .
a * 18 * , 2541 ( 2003 ) [ hep - th/0208176 ] .
u. bruzzo , f. fucito , j. f. morales and a. tanzini , jhep * 0305 * , 054 ( 2003 ) [ hep - th/0211108 ] . h. c. kim , s. kim , e. koh , k. lee and s. lee , jhep * 1112 * , 031 ( 2011 ) [ arxiv:1110.2175 [ hep - th ] ] .
a. gadde and s. gukov , jhep * 1403 * , 080 ( 2014 ) [ arxiv:1305.0266 [ hep - th ] ] .
f. benini , r. eager , k. hori and y. tachikawa , lett .
phys . * 104 * , 465 ( 2014 ) [ arxiv:1305.0533 [ hep - th ] ] .
f. benini , r. eager , k. hori and y. tachikawa , arxiv:1308.4896 [ hep - th ] .
d. tong , jhep * 1404 * , 193 ( 2014 ) [ arxiv:1402.5135 [ hep - th ] ] .
t. j. hollowood , a. iqbal and c. vafa , jhep * 0803 * , 069 ( 2008 ) [ hep - th/0310272 ] .
e. witten , nucl .
b * 471 * , 195 ( 1996 ) [ hep - th/9603150 ] .
a. klemm , p. mayr and c. vafa , in * la londe les maures 1996 , advanced quantum field theory * 177 - 194 [ hep - th/9607139 ] .
j. kim , s. kim , k. lee , j. park and c. vafa , arxiv:1411.2324 [ hep - th ] .
h. c. kim and s. kim , jhep * 1305 * , 144 ( 2013 ) [ arxiv:1206.6339 [ hep - th ] ] ; g. lockhart and c. vafa , arxiv:1210.5909 [ hep - th ] ; h. c. kim , j. kim and s. kim , arxiv:1211.0144 [ hep - th ] ; h. c. kim , s. kim , s. s. kim and k. lee , arxiv:1307.7660 ; j. qiu and m. zabzine , phys .
d * 89 * , no .
6 , 065040 ( 2014 ) [ arxiv:1312.3475 [ hep - th ] ] ; j. qiu , l. tizzano , j. winding and m. zabzine , arxiv:1403.2945 [ hep - th ] .
|
we study the 2d @xmath0 gauge theory descriptions of little strings on type ii ns5-branes .
the iib strings on @xmath1 ns5-branes are described by the @xmath2 gauge theories , whose higgs branch cfts on @xmath3 instanton moduli spaces are relevant .
the iia strings are described by @xmath2 @xmath4 quiver theories , whose coulomb branch cfts are relevant .
we study new @xmath5 quiver gauge theories for the iia strings , which make it easier to study some infrared observables .
in particular , we show that the elliptic genera of the iia / iib strings precisely map to each other by t - duality .
addtoresetequationsection + + * little strings and t - duality * jungmin kim@xmath6 , seok kim@xmath6 , kimyeong lee@xmath7 _ @xmath6department of physics and astronomy & center for theoretical physics , + seoul national university , seoul 151 - 747 , korea .
_ + _ @xmath7school of physics , korea institute for advanced study , seoul 130 - 722 , korea . _
+ e - mails : [email protected] , [email protected] , [email protected]
|
oscillons are localized and oscillating concentrations of energy in a nonlinear field that decay very slowly over many of their oscillation periods @xcite .
they are thought to be relevant in the dynamics of cosmological phase transitions @xcite and inflation scenarios @xcite , where they provide localized stores of energy .
in contrast to cosmic strings @xcite or other topological defects that can only be destroyed by annihilation , isolated oscillons can decay by radiation over long time scales . closely related to oscillons
are breathers , which are time - periodic localized solutions of a nonlinear field equation @xcite .
breathers only exist when resonance between the breather frequency and extended linear waves is avoided .
this is known to happen only in integrable nonlinear wave equations like the sine - gordon ( sg ) equation @xcite and the nonlinear schrdinger equation @xcite , or in nonlinear lattices @xcite due to the occurrence of band gaps in the linear wave spectrum .
oscillons are thus the more generic localized and oscillating nonlinear excitations . in most nonlinear field equations ,
oscillon solutions are not known in closed form and the investigation of oscillon properties relies on numerical simulations .
however , due to the integrability of sg equation in @xmath0 dimensions @xcite , the closed form solutions of sg breathers as well as solitons can be derived @xcite .
the fundamental solitons of the sg equation are localized topological excitations with a topological charge of @xmath1 known as kinks and anti - kinks , respectively .
sine - gordon breathers are localized in space and periodic in time and can be understood as the bound states of kink and anti - kink .
since their total topological charge is zero , breathers are not protected by their topological properties but rather by the integrable nature of the sg equation .
slight modifications of the equation that break the integrability typically destroys the breather solutions .
however , numerical methods can still find long - lived spatially localized and oscillating concentrations of energy , namely oscillons @xcite .
linearly coupled becs have been proposed as a model system to simulate the sine - gordon equation in a variety of contexts @xcite .
this has aroused considerable interest in investigating the excitations of linearly coupled bose gases , in particular josephson vortices , which are closely related to sg kinks @xcite . in a recent paper several of us
have simulated the spontaneous formation and decay of josephson vortices within the kibble - zurek scenario of a rapid quench through the bec phase transition in quasi - one - dimensional coupled bose gases @xcite .
the finding that post - quench dynamics and in particular the collision and annihilation of josephson vortices contribute decisively to the breakdown of kibble - zurek scaling laws has motivated the current study . as shown in sec .
[ s : collison ] , the collision of slow and oppositely charged josephson vortices generically produces oscillons . it has further been suggested that breather - like excitations ( oscillons ) with a finite lifetime can be spontaneously formed in coupled becs via a dynamical instability triggered by parametric amplification of quantum fluctuations @xcite . this emergence of localized excitations from amplified quantum fluctuations bears close analogy with the formation of oscillons after inflation of the early universe predicted within relativistic cosmological models @xcite .
the coupled becs may thus provide an experimentally accessible model to simulate the evolution of the early universe @xcite .
while recent works on simulating the sg model using coupled becs have focussed entirely on the relative phase dynamics @xcite , this picture neglects the coupling to the symmetric degrees of freedom in the system , such as the total phase and densities of the condensates . realistically
, such couplings can serve as channels of decoherence that can significantly alter the relative - phase dynamics and then lead to the instability of non - topological excitations , such as breathers .
a similar circumstance arises , for example , in annular josephson junctions , where an effective dissipation is introduced by carefully eliminating electronic degrees of freedom @xcite . by the same token
, it has been shown that the breathers are not robust against an ambient perturbation and hence decay by continuously radiating phonons @xcite thus forming oscillons .
this motivates us to carry out a detailed study of the full dynamics of breather - like excitations in becs .
specifically , we investigate the time evolution of an initial sg - breather state by numerically solving the gross - pitaevskii equation over broad ranges of coupling energy and imprinted sg breather frequency .
we assess the stability of the long - lived oscillon , and present a simple and feasible experimental scheme to create an oscillon by phase - imprinting method in a double - ring bec system .
the organization of this paper is as follows . in sec .
ii , we give a brief account of the system consisting of two linearly coupled becs .
we outline the emergent sg sector in the weak - coupling limit , and its kink solutions . in sec .
iii , we numerically demonstrate that the oscillon forms by imprinting the sg breather onto the initial state of the becs , and the stability of the ocillon is studied for broad ranges of coupling energy and imprinted sg breather frequency . in sec .
iv , we demonstrate the creation of an oscillon from the collision of two josephson vortices , and outline experimental constraints for a realistic phase imprinting protocol in a double - ring geometry .
we conclude in sec .
we consider a system consisting of two linearly coupled becs , which are tightly confined in the transverse directions but loosely confined in the longitudinal direction .
the grand - canonical gross - pitaevskii ( gp ) energy functional for this quasi - one - dimensional system is given by @xmath2\notag\\ & & -\tilde{\nu}\int dz\left[\psi_{1}^{\ast}\left(z\right)\psi_{2}\left(z\right)+\psi_{2}^{\ast}\left(z\right)\psi_{1}\left(z\right)\right],\label{grand - egp}\end{aligned}\ ] ] where @xmath3 is the order parameter of the _ _ -th condensate , @xmath4 the chemical potential , @xmath5 the coupling strength characterizing the tunnelling energy between the two systems , and @xmath6 is the effective nonlinear interaction strength .
the equation of motion for each @xmath3 can be derived via the hartree variational principle , namely , @xmath7 . for computational simplicity
, we let the length and time to be scaled in units of @xmath8 and @xmath9 respectively , so that the coupled gp equations are expressed in the dimensionless form @xmath10 where @xmath11 is the dimensionless coupling energy .
the ground state of the coupled becs can be determined by minimizing the energy functional , eq .
( [ grand - egp ] ) , where we seek the minimum of @xmath12-\nu\left[\psi_{1}^{*}\psi_{2}+\psi_{2}^{*}\psi_{1}\right]$ ] . the symmetry @xmath13 imposes a common amplitude for the ground state fields .
taking @xmath14 , @xmath15 , and @xmath16 , the minimum of @xmath17 occurs at @xmath18 and @xmath19 . to find the low - lying excitations above the ground state
, we can replace @xmath3 with @xmath20 in eq .
( [ eq : gpe ] ) and retain the expressions up to the linear order in @xmath21 and @xmath22 , which then leads to the bogoliubov - de genes ( bdg ) equation @xmath23\left[\begin{array}{c } \alpha_{q}^{1}\\ \beta_{q}^{1}\\ \alpha_{q}^{2}\\ \beta_{q}^{2 } \end{array}\right]=0,\label{bdg}\ ] ] where @xmath24 .
the low - lying excitation spectrum is determined by solving the eigenvalue problem , eq .
( [ bdg ] ) . as a result ,
we obtain two distinct dispersion relations for the excited modes : @xmath25 with the eigenvector @xmath26 @xmath27 , and @xmath28 with the eigenvector @xmath26 @xmath29 .
equation ( [ gapless eigenvalue ] ) represents a gapless mode , corresponding to the in - phase excitation of the two components of condensates , which manifests itself as the bogoliubov sound wave propagating at a speed @xmath30 . on the other hand , eq .
( [ gapped eigenvalue ] ) indicates a gapped mode , which corresponds to out - of - phase excitation .
it should be noted that these two modes are decoupled in the linear approximation .
moreover , the gapped excitation possesses a relativistic energy dispersion @xmath31 with sound speed , @xmath32 and the rest energy , @xmath33 . the gapless branch accounts for the relative phase dynamics which can be described by the sine - gordon equation @xcite .
before studying the oscillon excitation in linearly coupled becs , we investigate the low - energy dynamics of the symmetric degrees of freedom . in general , the order parameters of the coupled becs can be represented by @xmath34 and @xmath35 , where @xmath36 and @xmath37 are respectively the total and relative phase , @xmath38 the total density , and @xmath39 is the density mixing angle . substituting @xmath40 and @xmath41 into eq .
( [ grand - egp ] ) with the assumption that the coupling between the two becs is sufficiently weak , and following the arguments used in ref .
@xcite , we obtain the relevant hamiltonian density accounting for the sine - gordon dynamics @xmath42 where the total density @xmath38 is replaced by the sum of the ground state density of the two components , namely , @xmath43 . using the variational principle , the euler - lagrange equation of the relative phase
is derived @xcite , @xmath44 which is equivalent to the sg equation .
it is well known that the sg equation possesses the topologically protected kink and antikink solutions @xcite , i.e. , @xmath45,\label{soliton}\end{aligned}\ ] ] where the @xmath46@xmath47 sign corresponds to the kink ( antikink ) and @xmath48 and @xmath49 are the velocity and the initial location of the ( anti)kink , respectively . remarkably , the kink and antikink in eq .
( [ soliton ] ) may form a bound kink - antikink pair .
such a solution does exist for the sg model , which is spatially localized but temporally oscillating , and is thus called breather , @xmath50.\label{breather}\end{aligned}\ ] ] here @xmath51 is a positive constant that parametrises the amplitude and frequency of the breather by @xmath52 and @xmath53 respectively .
in contrast to kink and antiknik , the breather is not protected by topology but is protected by the integrability of the sg equation .
the breather becomes unstable once the integrability is broken . from eq .
( [ breather ] ) , we can define the full width at half maximum ( fwhm ) of the breather , @xmath54 .
note that the amplitude ( frequency ) decreases ( increases ) with increasing @xmath51 . with eq .
( [ breather ] ) , it is straightforward to show that the energy of a breather is given by @xmath55 with @xmath56 @xcite .
in addition to the kink and breather , the sg equation also possesses extended phonon excitations @xcite .
the hamiltonian , eq .
( [ sg - gp - subset ] ) , represents an emergent sg sector in the weak coupling limit of the total gp hamiltonian , where the latter supports excitations described by both asymmetric and symmetric degrees of freedom .
excitations in the sg sector such as the kink , breather , and phonon , depend only on the asymmetric degrees of freedom . on the other hand , off the sg sector , there are other types of excitations depending on the symmetric degrees of freedom , such as the bogoliubov phonon ( gapless excitation ) , and the gp dark ( grey ) soliton which has a @xmath57@xmath58 phase jump in the total phase . under the framework of the full gp formalism ,
theses excitations of different degrees of freedoms would couple to each other and these couplings would play an important role in the dynamics of the condensates .
therefore , it is expected that the sg breather is not stable in the presence of these couplings and would decay by slowly radiating energy , thus forming a so - called oscillon . in the following section ,
we investigate dynamics of the oscillon excitations in the linearly coupled becs .
in order to create an ocillon , we begin with a phase imprinted sg breather , with relative phase given by eq .
( [ breather ] ) , initially imprinted into a system of two linearly - coupled becs in a double - ring geometry .
the couplings between different degrees of freedom lead to the instability of the initially imprinted breather and result in energy radiation in terms of sound waves .
we expect that the propagation of the emitted sound waves may introduce extra effects to the dynamics of the oscillon , and our aim is to investigate how the oscillon evolves under the influence of the evolving background condensates .
a technical issue arrises , namely that , since we impose periodic boundary condition corresponding to a double - ring geometry for the numerical computations , whenever the outgoing sound waves reach one of the boundaries they will re - enter the system from the boundary at the opposite side .
these re - entering waves will interfere with those outgoing ones and also interact with the oscillon , complicating the dynamics in an uncontrollable way . to suppress the reentry of the emitted sound waves , we impose the absorption boundary layers by including a position - dependent damping coefficient @xmath59 in eq .
( [ eq : gpe ] ) , which turns eq .
( [ eq : gpe ] ) into a damped gp equation @xcite @xmath60,\label{eq : dgpe-1 } \\ i\partial_{t}\psi_{2 } & = & \left(1-i\sigma\right)\left[\left(-\partial_{z}^{2}/2+\left\vert \psi_{2}\right\vert ^{2}-1\right)\psi_{2}-\nu\psi_{1}\right].\label{eq : dgpe-2}\end{aligned}\ ] ] the damping coefficient @xmath61 varies with @xmath62 according to the prescription , @xmath63/2 $ ] , where @xmath49 and @xmath64 are the center and the width of the error function , respectively . in the simulations
, @xmath49 is chosen to locate far from the origin and @xmath65 is set to be sufficiently small , such that @xmath61 is increasing smoothly over the boundary layer and the reflected sound waves are greatly reduced . from a physical point of view , the absorber means that we are effectively considering a very large ring system , where the feedback from acoustic emissions lags the oscillon dynamics by a significant timescale . in what follows ,
we present the main results obtained by numerically integrating eqs .
( [ eq : dgpe-1 ] ) and ( [ eq : dgpe-2 ] ) over a variety of initial conditions , where the space integration by highly accurate fourier pseudospectral method , and the time integration is by the adaptive runge - kutta method of orders 4 and 5 ( rk45 ) . in order to study the dynamics and stability of oscillons in the coupled becs
, we imprint the phase profile of the sg breather , @xmath66 , which is defined in eq .
( [ breather ] ) , onto the homogeneous background density @xmath67 , @xmath68 and evolve this initial state in accordance with eqs .
( [ eq : dgpe-1 ] ) , ( [ eq : dgpe-2 ] ) . and @xmath69 [ eq . ( [ breather ] ) ] : ( a ) the density profile of the two superposed atomic fields , @xmath70 .
( b ) the spatial distribution of the relative phase , @xmath71 .
( c ) the spatial distribution of the total phase , @xmath72 .
( d ) the density profile of the grand canonical gp energy @xmath73 .
note that the imprinted sg breather is unstable which instantly decays into two gp solitons . ]
we first consider the case of a weak - coupling energy @xmath74 , and @xmath75 for the initially imprinted sg breather profile .
the created oscillon can be easily identified in the pattern of superposition , @xmath70 and in the time - varying spatial distribution of the relative phase , @xmath71 , as shown in fig .
[ nu1e-2-u-5e-1](b ) . continuous emission of bogoliubov sound waves from the oscillon is revealed by the appearance of the wavefronts in fig .
[ nu1e-2-u-5e-1](a ) and ( c ) , that corresponds to the up - chirping of the oscillon frequency as shown in fig .
[ nu1e-2-u-5e-1](e ) . to characterize the evolution of the oscillon
, we measure the frequency ( period ) of the relative phase oscillation shown in fig .
[ nu1e-2-u-5e-1](e ) . to verify the degree of deviation of the oscillon from the sg breather
, we compare the energy of the oscillon with those calculated by integrating the gp energy functional ( [ grand - egp ] ) and sg hamiltonian ( [ sg - gp - subset ] ) within the region of localization of the oscillon , @xmath76 . as shown in fig .
[ nu1e-2-u-5e-1](f ) , the energies @xmath77 of eq .
( [ breather - energy ] ) obtained from the measured oscillon frequency ( blue circle ) , the gp energy @xmath73 of eq .
( [ grand - egp ] ) ( black line ) , and the sg hamiltonian @xmath78 of eq .
( [ sg - gp - subset ] ) ( red line ) agree closely with each other , implying that the resulting oscillon is almost identical to the sg breather .
furthermore , the red line shows small - amplitude oscillation which is owing to the coupling of the asymmetric degrees of freedom to the symmetric ones and this suggests that the observed localized energy oscillation is in the context of oscillon .
we also perform the simulations for much smaller couplings and the oscillation of sg energy and the emission of the bogoliubov sound are suppressed which suggests that for sufficiently weak coupling ( @xmath79 ) the coupled becs support long - lived oscillon - type excitations .
studies of both the @xmath80 and the perturbed sg models have predicted the decay of the oscillon excitation via phonon emission @xcite . however in the weakly coupled bec system we observe a slightly different decaying behaviour where the oscillon - type excitations in the asymmetric degrees of freedom lose energy to the symmetric degrees of freedom by emitting bogoliubov phonons . unlike the @xmath80 and sg model , the decay which we observed could only occur in the coupled becs since the coupled becs support different degrees of freedom . in the strong - coupling regime ,
the relative phase dynamics is beyond the sg description so the dynamical properties of the oscillon are expected to be different from those of the sg breather . to study the oscillon dynamics in the strong coupling regime ,
we follow the same simulation procedure in the weakly coupled becs by imprinting the sg breather onto the ground state .
we first consider a stronger coupling energy of @xmath69 and an initially imprinted sg breather of @xmath75 . the evolution is shown in fig .
[ nu-5e-1-u-5e-1 ] , where the imprinted sg breather is unstable and decays into two gp solitons instantly .
we see that the two gp grey solitons , which carry the majority of the excitation energy , move towards the boundaries of the coupled condensates .
however , we find that the oscillon could remain long - lived by increasing @xmath51 of the initially imprinted sg breather .
now let us consider the same coupling energy but with an otherwise different initial imprinted sg breather of @xmath81 .
the time evolution of this initial state is shown in fig .
[ nu5e-1-u10 ] . in fig .
[ nu5e-1-u10](a ) , the imprinted sg breather emits sound waves initially to reduce its energy and then form the long - lived oscillon excitation .
in contrast to the ocillon in the sg regime , the oscillon in the strongly coupled becs exhibits different dynamical behaviour - the oscillon only emits sound waves at the beginning and subsequently the total phase manifests localized breathing oscillation as shown in figs . [
nu5e-1-u10](a ) and ( b ) .
similar to the weak coupling case , we also compare the energies of the oscillon obtained by eqs .
( [ grand - egp ] ) , ( [ sg - gp - subset ] ) , and ( [ breather - energy ] ) , and find that all these results are noticeably different .
we see that the gp energy density [ fig .
[ nu5e-1-u10](d ) ] shows a different behaviour from the weak coupling case [ fig .
[ nu1e-2-u-5e-1](d ) ] . the latter exhibits a temporally oscillatory behaviour but
the former does not .
this oscillation implies the energy transfer between the asymmetric degrees of freedom and the symmetric degrees of freedom which also accounts for the emergence of the oscillon in both the total and relative phases .
furthermore , the deviation of gp energy from the breather energy obtained by inserting the measured frequency into eq .
( [ breather - energy ] ) [ fig .
[ nu5e-1-u10](e ) ] implies that the relative dynamics in the present case can not be properly described by the sg equation .
we have also considered a much stronger coupling energy and the result shows similar behaviour except a much larger amplitude oscillation in the total phase .
the numerical results show the possibility to study the oscillon - type excitation in the strongly coupled becs . from the previous simulations ,
we conclude that the long - lived oscillon excitation can exist in both weak- and strong coupling regimes .
in general , the stability of the oscillon depends on two parameters , @xmath82 and @xmath51 .
we notice that @xmath51 is not directly related to system parameters , but it can be expressed in terms of @xmath82 and @xmath83 , the frequency of the initial sg breather .
therefore , it is more practical to characterize the stability of the oscillon as a function of the coupling @xmath82 , and the frequency @xmath83 of the imprinted breather .
we take the relative energy , @xmath84 , which is the ratio of remnant gp energy to the gp energy of the imprinted initial state as a measure of the stability of the oscillon . here
the gp energy of the oscillon is evaluated by integrating the energy density over the interval @xmath76 . in order to avoid the transient effects of the initial phase imprinting ,
the relative energy is obtained by calculating the ratio of the gp energy at @xmath85 to that at @xmath86 , where @xmath87 is the period of the imprinted sg breather . in the simulation , @xmath82 is varied from 0.01 to 0.6 , and for each @xmath82 , the imprinted frequency @xmath88 is varied from 0.005 to @xmath89 .
the limiting value @xmath89 is the maximally possible frequency of the sg breather of eq .
( [ breathfreq ] ) and indicates the boundary beyond which the sg breather can not be imprinted .
numerically we also find an energy threshold for stability given by @xmath90 , i.e. , the oscillon is unstable if @xmath91 and stable if @xmath92 .
the results are shown in fig .
[ phase - diagram](a ) , where we see that the oscillon as the remnant of the imprinted sg breather is always long - lived ( @xmath93 ) as long as @xmath94 . in fig .
[ phase - diagram](a ) , the upper green area marks the forbidden domain where no sg breather can be phase - imprinted and the black area indicates the domain where the imprinted sg breather would instantly decay into two gp solitons as shown in fig . [ nu-5e-1-u-5e-1 ] .
moreover , we observe that the oscillon energy loss is * as a function of @xmath88 and @xmath82 ( see text ) .
( b ) the relative frequency as a function of @xmath88 and @xmath82 . the upper green regions in panels ( a ) and ( b ) indicate where the sg - breather initial condition can not be imprinted .
the unstable ( black ) region in ( a ) corresponds to the lower green region in ( b ) , as @xmath95 is undefined . ]
accompanied by an up - chirping of its oscillation frequency with time .
hence , the frequency chirping of the oscillon can also serve as a measure of stability . to this end , we consider the relative frequency , @xmath96 , where @xmath97 is defined as the mean frequency of the oscillon averaged over the time interval @xmath98 ; this interval includes several oscillations , while being much shorter than the chirp timescale . the results of the frequency chirping are shown in fig .
[ phase - diagram](b ) . besides the green area in the energy diagram
, there are two stable domains having relative energy above the threshold .
the upper stable domain in the energy diagram shows a lower chirping rate frequency comparing to that of the lower ( narrow ) stable domain .
this suggests that the energy decay is stronger in the lower stable domain than that in the upper stable domain .
we also observe that , while increasing the frequency of the imprinted sg breather at a given coupling energy , the two grey gp solitons , originating from the decay of the imprinted sg breather , would gradually speed up in the black area of the energy diagram .
the decay into two grey gp solitons arises in the lower stability region as well , yet the emitted solitons acquire a speed very close to that of bogoliubov sound , indicating that the solitons are very unstable in this domain . note that in the upper stable domain , there are no grey solitons , but instead two outgoing small density bumps appear when the oscillon is stablized as shown in fig .
[ nu5e-1-u10 ] .
we note that , since the dynamical properties of the oscillon in strong coupling regime are different from those of the sg breather , the frequency of the imprinted sg - breather may not serve as the relevant parameter to describe the stability of oscillon in strong coupling regime , which leads to the appearance of the bifurcated stable domain in the strong coupling regime ( starting around @xmath99 in fig .
[ phase - diagram ] ) .
we have shown how the oscillon originating from a sg breather loses its energy in both weak- and strong - coupling cases .
the influence of external perturbations upon the dynamics of topology - protected excitations @xcite , so to speak , is broadly analogous to the influence of integrability breaking on oscillons and solitons @xcite , namely , that the breakdown of strict integrability couples the excitation to additional degrees of freedom that can extract its energy .
intriguingly , we find that the formation of oscillons could be related to the collision of josesphson vortices and in next section we shall address such a possibility .
as josephson vortices are topologically stable excitations of the coupled bec system , they can spontaneously form during the bec phase transition via the kibble - zurek mechanism @xcite , and may then collide to form an oscillon .
a josephson vortex pair could also be phase imprinted and then allowed to collide . in this section
we study the josephson vortex collision process as a basic prototype of oscillon formation .
we simulate the collision of two josephson vortices of opposite handedness , namely , kink and antikink , which provides a possibility of generating an oscillon for eq .
( [ eq : dgpe-1 ] ) . ):
( a ) the density profile of the two superposed atomic fields , @xmath70 .
( b ) the spatial distribution of the relative phase , @xmath71 .
the two counter - moving josephson vortices with velocities @xmath100 collide at the instant @xmath101 forming an oscillon . as shown in panel ( a ) , the quasi - breather continuously radiates bogoliubov sound waves to the background superfluid , causing an up - chirp of the oscillon s oscillation frequency . ]
we recall that the static josephson vortices in the linearly coupled becs exist when @xmath102 , and closed form expressions are known from ref .
@xcite . to obtain a solution for moving josephson vortices , we assume solutions propagating with constant velocity and solve the resulting coupled gpes numerically , increasing the velocity in small steps starting from zero @xcite .
the initial state for the time - dependent simulation is made up of two counter - moving josephson vortices with speed @xmath103 , initially located at both sides with equal distance from the origin . as can be seen from fig .
[ collision ] , the head - on collision of two josephson vortices takes place at the origin . remarkably , the two josephson vortices do not annihilate each other but bind together to form a oscillon excitation after the collision .
as soon as the oscillon is formed , it continuously emits bogoliubov sound waves .
since the coupling is very weak , it is expected that the relative phase dynamics is essentially governed by the sg equation , and the oscillon emerging from the collision of two josephson vortices ( sg kinks ) appears very close to a sg breather , as the one depicted in fig .
[ nu1e-2-u-5e-1 ] . as shown in fig .
[ collision](a ) , the frequency of the oscillon increases with time . according to eq .
( [ breather - energy ] ) , increasing @xmath83 will lower the value of @xmath104 , implying that the oscillon in fig .
[ collision](a ) gradually loses its energy .
this decay is due to the fact that the relative phase @xmath37 can couple to the other degrees of freedom in the gp energy , namely , the total phase @xmath105 and density @xmath38 .
although the oscillon is not an exact sg breather and undergoes damping , it has a very long lifetime comparing to its breathing period , suggesting that oscillon excitations in linearly coupled becs may be observable in bec experiments . , and
( b ) and ( d ) show the relative phase @xmath71 for two different sets of initial parameters : @xmath106 , @xmath74 , @xmath107 [ ( a ) and ( b ) ] and @xmath81 , @xmath69 , @xmath107 [ ( c ) and ( d ) ] . in both cases
a long - lived breather is created as seen in the relative phase . ]
we have systematically studied the oscillon in the coupled becs at different regimes of coupling strength . a possible way to realize
the oscillon is to phase - imprint the relative phase profile by shining a gaussian laser beam onto the coupled bec system , so that the stark effect generated by the laser can cause a controllable position - dependent phase difference between the two becs . in order to create the required relative phase profile
, the laser beam should be focused between the two becs and the focusing point should sit slightly away from the centerline between the two becs . assuming that the phases imprinted on the two becs take the form @xmath108,\phi_{2}^{i}=\epsilon a\exp\left[-z^{2}/\kappa^{2}\right],\label{imprinted - phase}\ ] ] where @xmath109 is the width of the gaussian beam , @xmath110 the geometric factor due to the departure of focal point from the centerline between the two condensates , and @xmath111 is proportional to the product of laser intensity and time duration of the exposure . as a consequence ,
the relative phase is given by @xmath112 $ ] .
matching the amplitude and width of @xmath113 to those of the sg breather given by eq .
( [ breather ] ) , the width @xmath109 and the amplitude @xmath111 can be determined after some algebra as @xmath114 where @xmath115 is the fwhm of the sg breather defined previously .
note that parameter range of the imprinted gaussian phase is limited since eq .
( [ width ] ) becomes imaginary when @xmath116 . to demonstrate that the aforementioned method is practical ,
we simulate eq .
( [ eq : dgpe-1 ] ) by phase imprinting the gaussian phase profile with @xmath117 and @xmath118 and the width @xmath109 and the amplitude @xmath111 are given according to eqs .
( [ width ] ) and ( [ intensity ] ) .
note that according to the assumption in eq .
( [ imprinted - phase ] ) , a non - zero total phase will also be imprinted , which causes disturbance in the total density . as shown in fig .
[ nu-1e-2-u-2-gaussian ] , this disturbance is detectable only in the atomic interference pattern [ panels ( a ) and ( c ) ] despite the creation of an oscillon excitation in the relative phase [ panels ( b ) and ( d ) ] .
we have theoretically studied the dynamics of an oscillon in the relative phase of two linearly coupled bose - einstein condensates .
the stability of the oscillon has been examined over broad ranges of coupling energy and imprinted sg breather frequency . in the weak coupling limit , the system exhibits sg dynamics predominantly , and thus an oscillon forms a long - lived metastable excitation , despite gradually losing energy to bogoliubov phonons . in the strong coupling regime ,
the system dynamics is no longer governed by the sg equation for the relative phase due to the increasingly prominent interplay between the symmetric and asymmetric degrees of freedom .
a systematic study of parameter space reveals large regions where the lifetime of the oscillon exceeds its period by several orders of magnitude .
the experimental realisation of oscillons may enable the possibility to use bose - einstein condensates as an analogue model to simulate and characterize oscillon formation in the dynamics of the early universe .
we are indebted to sophie shamailov for the provision of numerical code for generating moving josephson vortices used for sec .
[ s : collison ] and for useful comments on the manuscript .
ab was supported by a rutherford discovery fellowship administered by the royal society of new zealand .
of was supported by the marsden fund ( project no . mau 1205 ) , administrated by the royal society of new zealand .
sws , scg , and ikl were supported by the ministry of science and technology , taiwan ( grant no .
most 103 - 2112- m-018 - 002-my3 ) .
scg was also supported by national centre for theoretical scientists , taiwan . 0ifxundefined
[ 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 http://jetpletters.ac.ru/ps/1809/article_27642.pdf [ * * , ( ) ] link:\doibase 10.1103/physrevd.49.2978 [ * * , ( ) ] http://books.google.com/books?id=dckiefvh2kmc&pgis=1[__ ] ( , , ) link:\doibase 10.1103/physrevd.52.1920 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.241302 [ * * , ( ) ] link:\doibase 10.1088/0034 - 4885/58/5/001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.30.1262 [ * * , ( ) ] link:\doibase 10.1007/bf01017105 [ * * , ( ) ] link:\doibase 10.1016/s0370 - 1573(97)00068 - 9 [ * * , ( ) ] @noop ( ) @noop * * , ( ) @noop _ _ ( , , ) link:\doibase 10.1016/0167 - 2789(83)90290 - 7 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.063621 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.174511 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.085304 [ * * , ( ) ] link:\doibase 10.1002/andp.201300113 [ * * , ( ) ] , link:\doibase 10.1103/physreva.71.011601 [ * * , ( ) ] link:\doibase 10.1103/physreva.73.013627 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.011602 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/45/3/035004 [ * * , ( ) ] @noop ( ) link:\doibase 10.1103/physrevlett.110.215302 [ * * , ( ) ] , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.81.023630 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/12/5/055018 [ * * , ( ) ] link:\doibase 10.1103/physreva.81.033606 [ * * , ( ) ] @noop * * , ( )
|
long - lived , spatially localized , and temporally oscillating nonlinear excitations are predicted by numerical simulation of coupled gross - pitaevskii equations .
these oscillons closely resemble the time - periodic breather solutions of the sine - gordon equation but decay slowly by radiating bogoliubov phonons .
their time - dependent profile is closely matched with solutions of the sine - gordon equation , which emerges as an effective field theory for the relative phase of two linearly coupled bose fields in the weak - coupling limit .
for strong coupling the long - lived oscillons persist and involve both relative and total phase fields . the oscillons decay via bogoliubov phonon radiation that is increasingly suppressed for decreasing oscillon amplitude .
possibilities for creating oscillons are addressed in atomic gas experiments by collision of oppositely charged bose - josephson vortices and direct phase imprinting .
|
the discovery of cosmic acceleration @xcite has stimulated the study of different models of dark energy @xcite which may be responsible for such a phenomenon .
models of dark energy include those based on different perfect fluids , having negative pressure , on minimally and non - minimally coupled scalar fields and on fields having non - standard kinetic terms kinetic , tachyons .
the latter ones include as a subclass the models based on different forms of the born - infeld - type action , which is often associated with the tachyons arising in the context of string theory @xcite .
tachyonic models with relatively simple potentials were confronted with observational data in @xcite . compared to the standard klein - gordon scalar field cosmological models the dynamics of tachyon models can be much richer due to the non - linearity of the dependence of the tachyon lagrangians on the kinetic term of the tachyon field . in a recent paper @xcite a particular one - parameter family of tachyon models was considered , which has revealed some unexpected features . at some values of the parameter of the model a long period of accelerated quasi - de sitter expansion
is followed by a period of cosmic deceleration culminating , after a finite time , in an encounter with a cosmological singularity of a new type , which was named _ big brake_. this singularity is characterized by an infinite negative value of the second time derivative of the cosmological radius of the universe , while its first time derivative and the hubble variable vanish , and the radius itself acquires a finite value .
this singularity belongs to the class of soft ( sudden ) cosmological singularities @xcite which have been rather intensively studied during the last years .
here it is worth mentioning that in the context of the scrutiny of candidates for the role of dark energy , some other singularities attract the attention of cosmologists . among them
a special place occupies the big rip singularity @xcite , arising in some models where phantom dark energy @xcite is present .
the possibility of existence of a phase of contraction of the universe , ending up in the standard big crunch cosmological singularity was also considered in the literature @xcite .
recently , @xmath1-singularities were also proposed @xcite . an attractive peculiarity of the tachyon model studied in paper we - tach is the fact that there the big brake singularity is not put in by hands , but arises naturally as a result of the cosmological evolution , provided some initial conditions are chosen .
therefore it is a consequence of the dynamics , rather than a pure kinematical possibility . such evolution leading to the big brake coexists with another type of evolution describing an infinite expansion of the universe . in other words ,
a small change of initial conditions can have drastic consequences for the future of the universe .
actually , in spite of it being somewhat exotic , we show that the cosmological model @xcite does not contradict observations . to this aim
we compare the cosmological evolutions predicted in @xcite with the data coming from the supernovae type ia observations .
we select the compatible initial conditions by studying the backward evolution in comparison with the luminosity - redshift diagrams for the supernovae type ia standard(izable ) candles . then , choosing initial conditions which are compatible at the 1@xmath2 level with the data , we study the forward evolution and show that a deceleration period following the present accelerated expansion is possible , and when it is so , we estimate how long it is expected to last . the structure of the paper is the following . in sec .
ii we introduce the model and its basic equations ; in sec .
iii we find a subset of initial conditions which are compatible with the observational data by integrating numerically the dynamical equations backwards in time ; in sec .
iv we study numerically the cosmological evolutions for the selected initial conditions by numerical integration forward in time .
we end with some concluding remarks .
( @xmath3).,height=226 ] we consider the flat friedmann universe with the metric @xmath4 , filled with a spatially homogeneous tachyon field @xmath5 evolving according to the lagrangian @xmath6the energy density and the pressure of this field are , respectively @xmath7and @xmath8the equation of motion for the tachyon is @xmath9we consider the following tachyon potential @xmath10 @xcite : @xmath11where @xmath12 is a positive constant and @xmath13 .
taking into account the friedmann equation @xmath14 , where the hubble variable @xmath15 is defined as @xmath16 , and the newtonian constant is normalized as @xmath17 , we obtain the following dynamical system : @xmath18@xmath19when the parameter @xmath20 is negative , the evolution of the system ( system1)-([system2 ] ) is confined inside the rectangle @xmath21@xmath22the system has only one critical point : @xmath23which is an attractive node corresponding to a de sitter expansion with hubble parameter @xmath24all cosmological histories begin at the big bang type cosmological singularity located on the upper ( @xmath25 ) or lower ( @xmath26 ) side of the rectangle ( [ rectangle])-([rectangle1 ] ) , the individual history being parametrized by the initial value of @xmath5 satisfying the inequality ( rectangle1 ) .
they all end up in the node ( [ crit ] ) . in the case
@xmath27 the situation is more complicated .
first of all , the real potential @xmath28 is well - defined only in the interval @xmath29 where @xmath30 @xmath31 ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42
the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes . , title="fig:",height=302 ] ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42
the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes . , title="fig:",height=302 ] ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42
the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes . , title="fig:",height=302 ] ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42
the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes .
, title="fig:",height=302 ] ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42
the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes . , title="fig:",height=302 ] ( upper left ) , @xmath32 ( upper right ) , @xmath33 ( middle left ) , @xmath34 ( middle right ) , @xmath35 ( lower left ) , @xmath36 ( lower right ) , in the parameter plane ( @xmath37 , @xmath38 ) .
the white areas represent regions where the bounds on the model are not satisfied .
the contours refer to the @xmath39 ( 1@xmath40 ) and @xmath41 ( 2@xmath40 ) confidence levels . for increasing values of @xmath42 the well - fitting regions are increasingly smaller .
the colour code for @xmath43 is indicated on the vertical stripes . , title="fig:",height=302 ] the dynamical system ( [ system1])-([system2 ] ) has three fixed points : the node ( [ crit ] ) and the two saddle points with coordinates @xmath44 and , respectively , @xmath45 which give rise to an unstable de sitter regime with hubble parameter @xmath46 .
the most striking feature of the model under consideration with @xmath47 consists in the fact that now the cosmological trajectories do cross the corners of the rectangle ( [ rectangle]),([interval ] ) .
indeed , the direct analysis of the system of differential equations in the vicinity of the points @xmath48 and @xmath49 ( see fig .
[ fig1 ] ) shows that these points are not singular points of the system @xcite .
moreover , there is no cosmological singularity in these points @xcite .
that means that the cosmological evolutions must be continued through them .
an apparent obstacle to such a continuation is the fact that the expression under the square root in the formula for the potential ( poten ) changes sign when @xmath5 becomes smaller than @xmath50 or greater than @xmath51 .
however , the expression under the square root for the kinetic term @xmath52 also changes sign at the same time .
then , since the lagrangian of the theory is the product of these square roots , these simultaneous changes of sign leave the lagrangian and the corresponding expressions for the energy density ( [ energy ] ) and the pressure ( pressure ) real . the equation of motion for the tachyon field ( eq - of - motion ) also conserves its form .
the sign , which we prescribe for the product ( or for the ratio ) of the square roots is uniquely determined by the friedmann equation . in analyzing the behavior of our dynamical system in the regions where @xmath53 it is convenient to use the new potential @xmath54and to substitute in all expressions the term @xmath55 by @xmath56 . in doing so
the energy density and pressure have the form @xmath57and @xmath58being both positive .
the procedure of continuation of the trajectories through the corners of the rectangle is described in detail in @xcite . here ,
for the convenience of the reader we reproduce the phase portrait of the dynamical system from @xcite with some brief comments .
the rectangle in the phase space @xmath59 should be complemented by four infinite stripes ( see fig .
[ fig1 ] ) .
the left upper stripe ( the right lower stripe ) corresponds to the initial stages of the cosmological evolution , while the right upper stripe ( the left lower stripe ) corresponds to the final stages .
there are five classes of qualitatively different cosmological trajectories .
the trajectories belonging to classes i and ii end their evolution with an infinite de sitter expansion , while the trajectories of classes iii , iv and v encounter a big brake singularity .
the curves @xmath60 and @xmath61 are separatrices , dividing different classes of trajectories .
we end this section with the following remark . like
the other tachyon or dbi cosmological models ( for example , models displaying the power - law or exponential potentials ) the model based on potential ( [ poten ] ) possesses a wide class of cosmological evolutions ending up in an infinite accelerated expansion .
in addition , for small values of @xmath5 , this potential behaves as @xmath62 , a behavior which has been widely studied in the literature .
so far , so good . on the other hand , because of the more complicated structure of the potential ( [ poten ] ) , our model exhibits another class of trajectories with a qualitatively very different behavior and , in our opinion , this is precisely the feature which makes it particularly interesting .
in this section we select , at the confidence level of 1@xmath2 , and for a given choice of values of the parameter @xmath20 the set of initial conditions ( @xmath63 ) for the system ( [ system1])([system2 ] ) , which are compatible with the supernovae type ia data taken from paper @xcite . to this purpose , for the numerical analysis of the model it is convenient to rescale the relevant variables introducing the following dimensionless quantities : @xmath64where @xmath65 is the present value of the hubble parameter @xmath66 .
in addition we find it convenient to replace the variable @xmath5 with the new variable @xmath67also to switch from the time derivative to the derivative with respect to the redshift @xmath68 : @xmath69and denote @xmath70 with a prime .
then , the system of equations ( [ system1])([system2 ] ) in terms of the new variables @xmath71 ( all depending on @xmath68 ) becomes : @xmath72where @xmath73 and @xmath74 are given by @xmath75 ^{1/2}}{1-y^{2}}\ , \label{v } \\
\hat{v}_{,\hat{t } } & = & \frac{3\omega _ { \lambda } \sqrt{\ \omega _ { \lambda } \left ( 1+k\right ) } y\left [ k-1+\left ( 1+k\right ) y^{2}\right ] } { % 2(1-y^{2})^{3/2}\left [ 1-\left ( 1+k\right ) y^{2}\right ] ^{1/2}}\ .
\label{vt}\end{aligned}\ ] ] since @xmath76 , the present day values of the variables @xmath77 and @xmath78 satisfy the constraint @xmath79 } { \left [ 1-y^{2}\left ( 0\right ) % \right ] ^{2}}}\ .\]]we can avoid double coverage of the parameter space ( the model being invariant under the simultaneous change of signs @xmath80 and @xmath81 ) by replacing @xmath82 by the new variable @xmath83 the luminosity distance function for a flat friedmann universe @xmath84gives for the dimensionless luminosity distance @xmath85 the equation @xmath86 we are now in a position to compare our model with the supernovae type ia data @xcite . following ref .
@xcite we introduce the distance modulus type quantity @xmath87 , with @xmath88 a constant offset between the data and the theoretical expression .
the comparison involves computing @xmath89 ^{2}\ , \]]where the sum is over the supernovae in the data set and @xmath90 are the experimental errors in @xmath91 . the distance luminosity function
@xmath92 depends on the initial condition @xmath93 and @xmath94 .
we minimize this expression with respect to @xmath88 obtaining @xmath95with@xmath96 \ , \\ d & = & \sum_{i=1}^{n}\frac{1}{\sigma _ { i}^{2}}\ .\end{aligned}\ ] ] in table [ ys ] are listed the values @xmath97 , of the variable @xmath98 corresponding to the values @xmath99 of the variable @xmath5 given in formulas ( [ t3])([t2 ] ) for the chosen positive values of @xmath20 .
.the values of @xmath100 ( corresponding to the @xmath99 ) for some positive values of @xmath20 . [ cols="<,<,<,<",options="header " , ] @xmath101y_0@xmath102w_0@xmath102z_@xmath102 t _ ( 10 ^ 9yrs ) @xmath102z_bb@xmath102t_bb ( 10 ^ 9yrs ) @xmath103 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - 0.521@xmath10210@xmath102 - 0.542@xmath10211@xmath104 - [email protected]@xmath102 - 0.551@xmath10211@xmath102 - 0.571@xmath10212@xmath104 - [email protected]@xmath102 - 0.587@xmath10212@xmath102 - 0.605@xmath10213@xmath104 - [email protected]@xmath102 - 0.756@xmath10219@xmath102 - 0.766@xmath10220@xmath104 - [email protected]@xmath102 - 0.837@xmath10225@xmath102 - 0.845@xmath10226@xmath105 @xmath101y_0@xmath102w_0@xmath102z_@xmath102 t _ ( 10 ^ 9yrs ) @xmath102z_bb@xmath102t_bb ( 10 ^ 9yrs ) @xmath103 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - [email protected]@xmath102 - [email protected]@xmath104 - [email protected]@xmath102 - 0.533@xmath10210@xmath102 - 0.554@xmath10211@xmath104 - [email protected]@xmath102 - 0.557@xmath10211@xmath102 - 0.577@xmath10212@xmath104 - [email protected]@xmath102 - 0.584@xmath10212@xmath102 - 0.603@xmath10213@xmath104 - [email protected]@xmath102 - 0.616@xmath10213@xmath102 - 0.633@xmath10214@xmath104 - [email protected]@xmath102 - 0.658@xmath10215@xmath102 - 0.673@xmath10216@xmath104 - [email protected]@xmath102 - 0.693@xmath10216@xmath102 - 0.707@xmath10217@xmath104 - [email protected]@xmath102 - 0.733@xmath10218@xmath102 - 0.745@xmath10219@xmath104 - [email protected]@xmath102 - 0.779@xmath10221@xmath102 - 0.789@xmath10222@xmath104 - [email protected]@xmath102 - 0.814@xmath10223@xmath102 - 0.822@xmath10224@xmath104 - [email protected]@xmath102 - 0.865@xmath10228@xmath102 - 0.872@xmath10229@xmath104 - [email protected]@xmath102 - 0.927@xmath10236@xmath102 - 0.930@xmath10237@xmath104 - [email protected]@xmath102 - 0.955@xmath10243@xmath102 - 0.957@xmath10244@xmath105 finally we have evolved numerically backward in time some of the trajectories crossing the 1@xmath2 domain , until they reached one of the big bang singularities of the model .
all trajectories we have checked originate from the singularity at @xmath106 . in other words , they start from the horizontal boundaries of the rectangle in the phase plane ( @xmath107 ) , and depending on whether they evolve into an infinite de sitter expansion or reach the big brake singularity , they belong to either type ii or iii .
in this paper we have shown that the tachyon cosmological model of ref .
we - tach allows for a consistent set of trajectories which are compatible with the supernovae type ia data .
we have found that , among these , for positive values of the parameter @xmath20 of the model , there is a subset of evolutions which end up into a big brake singularity and , for the latter , we have computed the relevant big brake parameters @xmath108 and @xmath109 .
the compatibility of cosmological evolutions possessing soft cosmological singularities with the supernovae type ia data was studied in @xcite .
curiously , it was found in ref .
@xcite that a sudden singularity may take place in already a very close future , even less then 10 million years .
however this analysis was purely kinematical , and we also note that the parameters in our model ( as given by the tachyonic dynamics ) near the big brake singularity fall outside the range considered in @xcite . the problem of stability of a cosmological evolution in the vicinity of such singularities was studied in @xcite .
finally , we may ask why the model proposed in @xcite is worth studying .
first , the soft ( sudden ) cosmological singularity of the big brake type arises in our model in a very natural way as a particular class of solutions of the dynamical system ( [ system1])([system2 ] ) .
second , the model has another interesting feature . a subtle interplay between geometry and matter , induces a change of the very nature of the latter : it transforms from tachyon into a `` pseudo - tachyon '' field ( see @xcite for details ) .
we point out that a similar effect was observed also in scalar - phantom cosmological models @xcite .
phenomena of this kind represent a distinguishing feature of general relativity @xcite : the requirement of self - consistency of einstein equations can impose the form of the equations of motion for the matter .
thus , in spite of being a toy model , the tachyon cosmological model we - tach can serve as a prototype of realistic ( i.e. compatible with observational data ) cosmological models which may lead to a final fate of the universe , different from the infinite quasi - de sitter expansion of the @xmath12cdm model .
what will actually happen in the future is left to our far away descendants to experience !
we thank gy . szab for discussions in the early stages of this project .
we are grateful to j.d .
barrow and m.p .
dabrowski for useful correspondence .
was supported by the otka grant 69036 ; l..g . was supported by the otka grant 69036 , the london south bank university research opportunities fund and the polnyi program of the hungarian national office for research and technology ( nkth ) ; a.k .
was partially supported by rfbr grant no .
08 - 02 - 00923 and by the grant lss-4899.2008.2 . v. sahni and a.a .
starobinsky , int .
d * 9 * , 373 ( 2000 ) ; * 15 * , 2105 ( 2006 ) ; t. padmanabhan , phys . rep . * 380 * , 235 ( 2003 ) ; p.j.e .
peebles and b. ratra , rev .
phys . * 75 * , 559 ( 2003 ) ; e.j .
copeland , m. sami and s.tsujikawa , int .
d * 15 * , 1753 ( 2006 ) .
a. sen , jhep * 0207 * , 065 ( 2002 ) , g.w .
gibbons , phys .
b * 537 * , 1 ( 2002 ) , a. feinstein , phys .
d * 66 * , 063511 ( 2002 ) , t. padmanabhan , phys .
d * 66 * , 021301 ( 2002 ) , a. frolov , l. kofman and a. starobinsky , phys .
b * 545 * , 8 ( 2002 ) ; n. barnaby , jhep * 0407 * , 025 ( 2004 ) ; r. herrera , d. pavon and w. zimdahl , gen .
rel . grav .
* 36 * , 2161 ( 2004 ) ; j. m. aguirregabiria and r. lazkoz , mod .
a * 19 * , 927 ( 2004 ) ; a. sen , phys .
scripta * t117 * , 70 ( 2005 ) ; e. j. copeland , m. r. garousi , m. sami and s. tsujikawa , phys . rev .
d * 71 * , 043003 ( 2005 ) ; r. lazkoz , int .
j. mod .
d * 14 * , 635 ( 2005 ) ; v. gorini , a. kamenshchik , u. moschella , v. pasquier and a. starobinsky , phys .
d * 72 * , 103518 ( 2005 ) ; e. elizalde and j. q. hurtado , int .
j. mod .
d * 14 * , 1439 ( 2005 ) ; c. quercellini , m. bruni and a. balbi , class
. quant .
* 24 * , 5413 ( 2007 ) ; j. martin and m. yamaguchi , phys .
d * 77 * , 123508 ( 2008 ) .
j. s. bagla , h. k. jassal and t. padmanabhan , phys .
d * 67 * , 063504 ( 2003 ) ; g. calcagni and a. r. liddle , phys . rev .
d * 74 * , 043528 ( 2006 ) ; r. j. yang , s. n. zhang and y. liu , jcap * 0801 * , 017 ( 2008 ) . y. shtanov and v. sahni , class . quant . grav . * 19 * , l101 ( 2002 ) ; j. d. barrow , class .
* 21 * , l79 ( 2004 ) ; j. d. barrow , class .
* 21 * , 5619 ( 2004 ) , a. kamenshchik , c. kiefer and b. sandhofer , phys . rev .
d * 76 * , 064032 ( 2007 ) .
r. kallosh , a. linde , s. prokushkin and m. shmakova , phys .
d * 66 * , 123503 ( 2002 ) , r. kallosh and a. linde , jcap * 0302 * , 002 ( 2003 ) , u. alam , v. sahni and a. a. starobinsky , jcap * 0304 * , 002 ( 2003 ) , r. kallosh , j. kratochvil , a. linde , e. v. linder and m. shmakova , jcap * 0310 * , 015 ( 2003 ) .
|
we compare the existing observational data on type ia supernovae with the evolutions of the universe predicted by a one - parameter family of tachyon models which we have introduced recently in paper @xcite . among the set of the trajectories of the model which are compatible with the data
there is a consistent subset for which the universe ends up in a new type of soft cosmological singularity dubbed big brake .
this opens up yet another scenario for the future history of the universe besides the one predicted by the standard @xmath0cdm model .
|
the perceptron was introduced in the 50s as a simple model of how neurons learn @xcite .
it has been extremely influential , counting both deep learning architectures and support vector machines amongst its descendants .
unfortunately however , the perceptron and related models badly misrepresent important features of neurons .
in particular , they treat outputs symmetrically ( @xmath3 ) rather than asymmetrically ( 0/1 ) .
this is crucial since spikes ( 1s ) are more metabolically expensive than silences ( 0s ) .
further , by - and - large neurons only update their synapses after spiking .
by contrast , perceptrons update their weights after every misclassification . to build a tighter link between learning algorithms and neurocomputational models , we recently discretized standard models of neuronal dynamics and plasticity to obtain the selectron @xcite .
this section shows that , suitably regularized , the selectron is almost identical to linear regression .
the difference is a _ selectivity _ term arising from the spike / silence asymmetry that encourages neurons to specialize .
[ [ the - selectron . ] ] the selectron .
+ + + + + + + + + + + + + + let @xmath4 denote the set of inputs and @xmath5 the set of possible synaptic weights .
let @xmath6 , where threshold @xmath7 is fixed . given @xmath8 , neuron @xmath9 with synaptic weights @xmath10 outputs 0 or 1 according to @xmath11 we model neuromodulatory signals via @xmath12 where @xmath13 , with positive values corresponding to desirable outcomes and conversely .
signals may arrive after a few hundred millisecond delay , which we do not model explicitly .
a threshold neuron is a selectron if its reward function takes the form @xmath14.\ ] ] the reward function is continuously differentiable ( in fact , linear ) as a function of @xmath15 everywhere except at the kink @xmath16 where it is continuous but not differentiable .
we can therefore maximize the reward via gradient ascent to obtain synaptic updates @xmath17 [ t : limit ] discretizing gerstner s spike response model @xcite yields .
discretizing spike - timing dependent plasticity ( stdp ) @xcite yields .
finally , stdp is gradient ascent on a reward function whose discretization is @xmath18 . if rewards are more common than punishments then leads to overpotentiation ( and eventually epileptic seizures ) .
neuroscientists therefore introduced a depotentiation bias into stdp .
alternatively , @xcite , introduced an @xmath19-constraint on synaptic weights enforced during sleep .
below , we interpolate between the two approaches by replacing the constraint with a regularizer . [
[ linear - regression . ] ] linear regression .
+ + + + + + + + + + + + + + + + + + we first recall linear regression to aid the comparison .
given data @xmath20 , regression finds parameters @xmath21 minimizing the mean squared error @xmath22.\ ] ] one way to solve this is by gradient descent using @xmath23.\ ] ] [ [ selective - linear - regression . ] ] selective linear regression .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + now , if we regularize the reward in as follows @xmath24\\ \notag = & \operatorname*{argmin}_{{\mathbf{w}}_j } \hat{{{\mathbb e}}}\big [ \frac{1}{2}\big ( \underbrace{\mu - \langle{\mathbf{w}}_j,{{\mathbf x}}\rangle_\vartheta}_{\text{regression } } \cdot \underbrace{f_{{\mathbf{w}}_j}({{\mathbf x}})}_{\text{selectivity } } \big)^2\big]\end{aligned}\ ] ] the result is _ selective _ linear regression : a neuron s excess current predicts neuromodulation _ when it spikes .
_ since synaptic weights are non - negative , the neuron will not fire for @xmath25 such that @xmath26<0 $ ] .
computing gradient ascent obtains @xmath27 = \sum_{\{\iota|n_j \text { spikes}\}}\big ( \mu^\iota-\langle{\mathbf{w}}_j,{{\mathbf x}}^\iota\rangle_\vartheta\big)\cdot{{\mathbf x}}^\iota.\ ] ] the synaptic updates are @xmath28 . in other words , if neuron @xmath9 receives spike @xmath29 and produces spike @xmath30 , then it modifies synapse @xmath31 proportional to how much greater the rescaled neuromodulatory signal is than the excess current .
the selectivity term in makes biological sense .
there are billions of neurons in cortex , so it is necessary that they specialize .
neuromodulatory signals are thus ignored unless the neuron spikes providing a niche wherein the neuron operates .
[ [ multi - view - learning - in - cortex . ] ] multi - view learning in cortex .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + there are two main types of excitatory synapse : ampa and nmda .
so far we have modeled ampa synapses which are typically feedforward and `` driving '' they cause neurons to initiate spikes .
nmda synapses differs from ampa in that @xcite * they _ multiplicatively modulate _ synaptic updates and * they prolong , but do not initiate , spiking activity .
we model the two types of synaptic inputs as ampa , @xmath25 , and nmda , @xmath32 , views with synaptic weights @xmath33 and @xmath15 respectively . in accord with the observations
above , we extend by adding a _ multiplicative _ modulation term . the nmda view is encouraged to align with neuromodulators and is regularized the same as ampa .
the nmda view has no selectivity term since it does not initiate spikes .
finally , we obtain a ( discretized ) neuronal co - optimization algorithm , which simultaneously attempts to maximize how well each view predicts neuromodulatory signals and aligns the two views on unlabeled data : @xmath34.\end{gathered}\ ] ] the next section describes a semi - supervised regression algorithm inspired by .
this section translates the multiview optimization above into a workable learning algorithm .
[ [ from - neurons - to - machine - learning . ] ] from neurons to machine learning .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we make three observations about cortical neurons : * each neuron s inputs are preprocessed by millions of upstream neurons , providing a non - linear feature map analogous to the kernel trick @xcite .
* a neuron s synaptic contacts ( not weights ) are random to first approximation @xcite . *
the `` _ _ modulation _ _ - _ regularizers _ '' terms in resemble canonical correlation analysis @xcite : @xmath35 } { \sqrt{\hat{{{\mathbb e}}}[\langle{\mathbf{v}},{{\mathbf x}}\rangle^2 ] \cdot \hat{{{\mathbb e}}}[\langle{\mathbf{w}},{{\mathbf z}}\rangle^2 ] } } = \frac{\text{``\emph{modulation}''}}{\text{``\emph{regularizers}''}}.\ ] ] * input : * labeled data : @xmath36 and unlabeled data : @xmath37 sample @xmath38 uniformly from the dataset , compute the eigendecompositions of the sub - sampled kernel matrices @xmath39 and @xmath40 which are constructed from the samples @xmath41 and @xmath42 respectively , and featurize the input : @xmath43}{^{\top}}\text { for } \mu\in\{1,2\}.\ ] ] compute cca bases @xmath44 , @xmath45 and canonical correlations @xmath46 for the two views and set @xmath47 solve @xmath48 * output : * @xmath49 the observations suggest neurons perform an analog of * kernelized multiview regression * with random views and * a cca penalty . to check these form a viable combination , we put the pieces together to develop correlated nystrm views ( ` xnv ` ) , see algorithm [ alg : xnv ] and @xcite . reassuringly
, ` xnv`beats the state - of - the - art in semi - supervised learning @xcite . [ [ multiview - regression . ] ] multiview regression .
+ + + + + + + + + + + + + + + + + + + + + the main ingredient in ` xnv`is multiview regression , which we now describe .
suppose the loss of the best regressor in each view is within @xmath50 of the best joint estimator .
@xmath51 introduce the canonical norm @xmath52 where @xmath53 are orthogonal solutions to with correlation coefficients @xmath54 .
multiview regression is then @xmath55.\ ] ] penalizing with the canonical norm biases the estimator towards features that are correlated across both views ( the signal ) and away from features that are uncorrelated ( the noise ) .
multiview regression is thus a specific instantiation of the general co - training principle that good regressors agree across views whereas bad regressors may not .
the multiview estimator s error , , compared to the best linear predictor @xmath56 , is bounded by @xmath57-\operatorname{loss}(f ) \leq 5\epsilon + \frac{\sum_i \lambda_i^2}{n}.\ ] ] according to the theorem , a slight increase in bias compared to ordinary regression is potentially more than compensated for by a large drop in variance .
the reduction in variance depends on how quickly the correlation coefficients decay .
for example , in the trivial case where the two views are identical , there is no benefit from multiview regression . to work well
, the algorithm requires sufficiently different views ( where most basis vectors are uncorrelated ) that nevertheless both contain good regressors ( that is , the few correlated directions are of high quality ) .
[ [ randomization . ] ] randomization .
+ + + + + + + + + + + + + + to convert multiview regression into a general - purpose tool we need to construct pairs of views for any data satisfying two requirements .
first , they should contain good regressors .
second , they should differ enough that their correlation coefficients decay rapidly .
a computationally cheap approach that does not depend on specific properties of the data is to generate random views .
to do so , we used the nystrm method @xcite and random kitchen sinks @xcite .
a recent theorem of bach implies nystrm views contain good regressors in expectation @xcite and a similar result holds for random kitchen sinks .
although there are currently no results on correlation coefficients across random views , recent lower bounds on the frobenius norm of the nystrm approximation suggest `` medium - sized '' views ( a few hundred dimensional ) differ sufficiently @xcite .
empirical performance is discussed below . [ [ performance . ] ] performance .
+ + + + + + + + + + + + table [ t : performance ] summarizes experiments evaluating ` xnv`on 18 datasets , see @xcite for details .
we do not report on random kitchen sinks , since they performed worse than nystrm views .
performance is evaluated against kernel ridge regression ( ` krr ` ) and a randomized version of ` sssl ` , a semi - supervised algorithm with state - of - the - art performance @xcite .
, the randomized version of ` sssl ` , performs similarly to the original in a fraction of the runtime . ] ` xnv`typically outperforms ` sssl`@xmath58 by between 10% and 15% , with about 30% less variance .
both semi - supervised algorithms achieve less than half the error of kernel ridge regression .
.average performance of ` xnv`against ` krr ` and randomized ` sssl ` on 18 datasets . [ cols="<,>,>,>,>,>",options="header " , ] importantly , ` xnv`is _
fast_. for @xmath59 points , ` xnv`runs in @xmath60 on a laptop whereas ( unrandomized ) ` sssl ` takes @xmath61 . for @xmath62 points ,
` xnv ` s runtime is @xmath63 whereas ` sssl ` takes unfeasibly long .
we describe work in progress on multiview neuronal regression . [ [ selective - co - regularized - least - squares . ] ] selective co - regularized least squares .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the multiview optimization in can be rewritten , essentially , as co - regularized least squares @xcite with a selectivity term encouraging specialization : @xmath64 \\
\notag + \ ; & \hat{{{\mathbb e}}}_{\text{unlabeled}}\big [ \alpha_{co}\cdot \underbrace{\big(\langle{\mathbf{v}}_j,{{\mathbf x}}\rangle_\vartheta -\langle{\mathbf{w}}_j,{{\mathbf z}}\rangle\big)^2}_{\text{unsupervised co - regularization } } \big].\end{aligned}\ ] ]
the model closely resembles ` xnv ` , with a penalty that is easier for neurons to implement . the selectivity term in ensures that neurons only predict the neuromodulatory signals when they spike . in other words , neurons have the flexibility to search for an ampa view containing a good regressor .
the nmda weights are then simultaneously aligned with the neuromodulatory signal and the ampa weights by the remaining two terms .
[ [ guarantees . ] ] guarantees .
+ + + + + + + + + + + theoretical guarantees for co - regularization are provided in @xcite .
as above , they depend on having good regressors in sufficiently different views .
we briefly sketch how these apply .
the benefit from co - training depends on the extent to which co - regularizing with unlabeled shrinks the function space applied to the labeled data . in short ,
it depends on the rademacher complexity of @xmath65 + \hat{{{\mathbb e}}}_{\textrm{unl}}\big[\alpha_{co}\cdot|f({{\mathbf x}})-g({{\mathbf z}})|^2\big]\leq 1\big\}.\ ] ] denote the gram matrices of the two views by @xmath66 where @xmath67 and @xmath68 are dot - products of unlabeled data on the respective views , and other blocks are similarly constructed using mixed and labeled data .
it is shown in @xcite that @xmath69 the last term is of particular interest .
each column @xmath70 represents a labeled point in the first view by its dot - product with unlabeled points , and similarly for @xmath71 in the second view .
the greater the _ difference _ between the representations in the two views , as measured by , the lower the rademacher complexity and the better the generalization bounds on co - training .
it remains to be seen how selective multiview regression performs empirically , and to what extent provides a good guide to the improvement in generalization performance of the original _ undiscretized _ models .
co - training and randomization are two simple , powerful methods that complement each other well and which neurons appear to use in conjunction .
no doubt there are more tricks waiting to be discovered .
it is particularly intriguing that nell , one of the more ambitious ai projects in recent years , uses various co - training strategies as basic building blocks @xcite . despite a large body of research on how humans learn categories and relations
, it remains unknown how ( or whether ) individual neurons learn categories .
although the results sketched here are suggestive , they fall far short of the full story . for example , since neurons learn online and only when they spike they face similar explore / exploit dilemmas to those investigated in the literature on bandits . it will be interesting to see if new ( randomized ) bandit algorithms can be extracted from models of synaptic plasticity .
|
despite its size and complexity , the human cortex exhibits striking anatomical regularities , suggesting there may simple meta - algorithms underlying cortical learning and computation .
we expect such meta - algorithms to be of interest since they need to operate quickly , scalably and effectively with little - to - no specialized assumptions .
this note focuses on a specific question : how can neurons use vast quantities of _ unlabeled _ data to speed up learning from the comparatively rare labels provided by reward systems ? as a partial answer , we propose randomized co - training as a biologically plausible meta - algorithm satisfying the above requirements . as evidence
, we describe a biologically - inspired algorithm , correlated nystrm views ( ` xnv ` ) that achieves state - of - the - art performance in semi - supervised learning , and sketch work in progress on a neuronal implementation .
although staggeringly complex , the human cortex has a remarkably regular structure @xcite .
for example , even expert anatomists find it difficult - to - impossible to distinguish between slices of tissue taken from , say , visual and prefrontal areas .
this suggests there may be simple , powerful meta - algorithms underlying neuronal learning .
consider one problem such a meta - algorithm should solve : taking advantage of massive quantities of unlabeled data .
evolution has provided mammals with neuromodulatory systems , such as the dopamenergic system , that assign labels ( e.g. pleasure or pain ) to certain outcomes .
however , these labels are rare ; an organism s interactions with its environment are typically indifferent .
nevertheless , mammals often generalize accurately from just a few good or bad outcomes .
our problem is therefore to understand how organisms , and more specifically individual neurons , use unlabeled data to learn quickly and accurately .
next , consider some properties a semi - supervised _ neuronal _ learning algorithm should have
. it should be : * fast ; * scalable ; * effective ; * broadly applicable ( that is , requiring few - to - no specialized assumptions ) ; and * biologically plausible .
the first four requirements are of course desirable properties in any learning algorithm .
the fourth requirement is particularly important due to the wide range of environments , both stochastic and adversarial , that organisms are exposed to .
regarding the fifth requirement , it is unlikely that evolution has optimized all of the cortex s @xmath0 connections , especially given the explosive growth in brain size over the last few million years .
a simpler explanation , fitting neurophysiological evidence , is that the macroscopic connectivity ( largely the white matter ) was optimized , and the details are filled in randomly .
[ [ contribution . ] ] contribution .
+ + + + + + + + + + + + + this note proposes _ randomized co - training _ as a semi - supervised meta - algorithm that satisfies the five criteria above .
in particular , we argue that randomizing cortical connectivity is not only necessary but also beneficial . [
[ co - training . ] ] co - training .
+ + + + + + + + + + + + a co - training algorithm takes labeled @xmath1 and unlabeled data @xmath2 consisting of two views @xcite .
examples of views are audio and visual recordings of objects or photographs taken from different angles .
the key insight is that good predictors will agree on both views whereas bad predictors may not @xcite .
co - training algorithms therefore use unlabeled data to eliminate predictors disagreeing across views , shrinking the search space used on the labeled data resulting in better generalization bounds and improved empirical performance @xcite . the most spectacular application of co - training is perhaps never - ending language learning ( nell ) , a semi - autonomous agent that updates a massive database of beliefs about english language categories and relations based on information extracted from the web @xcite .
however , despite its conceptual elegance , co - training remains a niche method .
a possible reason for the lack of applications is that it is difficult to find naturally occurring views satisfying the technical assumptions required to improve performance .
constructing randomized views is a cheap workaround that dramatically extends co - training s applicability .
[ [ outline . ] ] outline .
+ + + + + + + + section [ s : selectron ] shows that discretizing standard models of synaptic dynamics and plasticity @xcite leads to a small tweak on linear regression . incorporating nmda synapses into the model as a second view leads to a neuronal co - training algorithm .
section
[ s : xnv ] reviews recent work which translated the above observations about neurons into a learning algorithm .
we introduce correlated nystrm views ( ` xnv ` ) , a state - of - the - art semi - supervised learning algorithm that combines multi - view regression with random views via the nystrm method @xcite .
finally , [ s : rcc ] returns to neurons and sketches preliminary work analyzing the benefits of randomized co - training in cortex .
|
lattice is represented with its _ generator matrix _ @xmath0 , whose entries are real numbers .
let @xmath1 and @xmath2 denote the number of rows and columns of @xmath0 respectively with @xmath3 .
the rows of @xmath0 , which are @xmath4 , are called _ basis vectors _ and are assumed to be linearly independent vectors in @xmath5 .
the lattice of dimension @xmath1 is defined as the set of points @xmath6 this paper is about methods to find the _
closest point _ in a lattice to a given vector @xmath7 , hereafter called _ received vector , _ which requires minimization of the metric @xmath8 over all lattice points @xmath9 with @xmath10 . in 1981
, pohst @xcite suggested a way of finding the closest point in lattices , which later on was complemented by fincke and pohst in @xcite .
the general method has later become known as _ sphere decoding . _
the implementation details of the fincke pohst ( fp ) enumeration method were first presented by viterbo and biglieri in @xcite . in 1999 ,
viterbo and boutros applied the fp enumeration method to maximum likelihood ( ml ) detection for finite constellations @xcite . later on , agrell _ et al .
_ in @xcite illustrated that the schnorr - euchner ( se ) refinement @xcite of the fp enumeration strategy improves the complexity of the sphere decoder algorithm . during the last decade , a lot of work has been done to improve the efficiency of sphere decoder algorithms @xcite , due to the significant usage they have found in numerous types of applications . in communication theory ,
the closest point problem arises in ml detection for multiple - input multiple - output ( mimo ) channels @xcite , ml sequence estimation @xcite , quantization @xcite , vector perturbation in multiuser communications @xcite , and joint detection in direct - sequence multiple access system @xcite .
the closest point search algorithms can be modified to find the ml point in finite constellations @xcite , which has an important application in mimo channels . assuming a system with @xmath1 transmit and @xmath2 receive antennas , the new set of points @xmath11 is defined by replacing @xmath12 in with the finite range of integers @xmath13 the transmit set can be mapped to an @xmath14-pam constellation with @xmath15 .
the received vector after an additive white gaussian noise ( awgn ) channel with double - sided noise power spectral density @xmath16 is @xmath17 where @xmath18 , @xmath7 , @xmath19 , and @xmath20 is a vector of independent and identically distributed ( i.i.d . )
gaussian noise with variance @xmath16 . in this case , ml detection is equivalent to minimization of the metric @xmath8 over all possible points @xmath9 with @xmath18 . in mimo systems where usually quadrature amplitude modulation ( qam ) is used , the @xmath21-qam signal constellation can be viewed as two real - valued @xmath14-pam constellations with @xmath22 , @xmath23 , @xmath24 , and @xmath25 . for both types of applications , lattices or finite constellations ,
the calculations can be implemented based on @xmath0 , as in the original fp algorithm and its numerous refinements , notably @xcite , or based on @xmath26 @xcite . in this paper , we draw attention to a hitherto unnoticed problem with the standard algorithms .
it is illustrated that the standard sphere decoder algorithms based on fp @xcite and se @xcite enumeration strategies perform many excessive numerical operations .
a method is proposed to avoid these unnecessary computations .
however , the revision proposed is not related to choosing a more accurate upper bound on @xmath27 or scanning set of feasible point @xmath9 in a different order .
we believe that the se strategy is the best way in this regard .
our modifications instead change how lattice vectors are recursively constructed from lower - dimensional lattices ( for @xmath0-based implementations ) or how the received vector @xmath28 is recursively projected onto the basis vectors ( for @xmath29-based implementations ) , which accounts for most of the floating point calculations in sphere decoding . with the proposed methods ,
not a single value would be calculated twice or remain without any use .
standalone implementations of the new ( and old ) algorithms are given in fig .
[ fig : algorithms ] .
without loss of generality , we assume that @xmath0 is a square lower - triangular matrix with positive diagonal elements @xcite . consequently , @xmath26 is also square with positive diagonal elements .
the decription of the sphere decoding principle in this section takes the @xmath29-based approach .
every lattice can be divided into layers of lower - dimensional lattices .
the diagonal elements of @xmath29 illustrate the distances between these layers , such that @xmath30 represents the distance between the @xmath31-dimensional layers in an @xmath32-dimensional layer .
thus , @xmath33 is the distance between the lattice points in a one - dimensional layer .
[ fig:1 ] illustrates an @xmath1-dimensional hypersphere with radius @xmath34 centered on a vector @xmath28 .
all lattice points inside this hypersphere lie on @xmath35-dimensional layers , which are also hyperspheres .
the basis vector @xmath36 is in the same direction as the hypotenuse of right triangles @xmath37 and @xmath38 , while all the other basis vectors @xmath39 lie in the subspace spanned by one of these @xmath35-dimensional layers .
starting from dimension @xmath1 , the received vector @xmath40 is projected onto the lattice basis vectors @xmath41 .
this is done by a simple matrix multiplication @xmath42 , where @xmath43 . for known @xmath44 and @xmath45 the corresponding range for the integer component @xmath46
is @xcite @xmath47 where @xmath48 and @xmath49 denote the round up and round down operations respectively , which is also intuitively conspicuous from fig .
[ fig:1 ] .
[ cc][][0.75]a [ cc][][0.75]b [ cc][][0.75]c [ cc][][0.75]d [ cc][][0.75]e [ cc][][1]@xmath28 [ cc][][1]@xmath50 [ cc][][1]@xmath36 [ cc][][1]@xmath51 [ cc][][1]@xmath34 [ cc][][1]@xmath52 [ cc][][1]@xmath53 [ cc][][1]@xmath46 -dimensional hypersphere , divided into a stack of @xmath35-dimensional hyperspheres ( layers).,title="fig:",width=302 ] for each @xmath35-dimensional layer @xmath54 that is to be examined , the orthogonal displacement @xmath51 from the received vector @xmath28 to this layer is calculated , which is shown with line @xmath55 in fig .
[ fig:1 ] .
this displacement follows from the congruence of @xmath37 and @xmath38 : @xmath56 in order to calculate @xmath57 , which will be used later on to calculate the range of @xmath58 and the displacement @xmath59 , the received vector @xmath28 is first projected onto the examined @xmath35-dimensional layer and then to the lattice basis vectors .
we use the notation @xmath60 for the projected received vector @xmath28 , where @xmath61 denotes the dimension of the layer that the received vector is projected on .
thanks to the lower - triangular representation , the orthogonal projection of @xmath28 onto the @xmath35-dimensional layer currently being investigated affects only the last component of @xmath28 .
thus , it is sufficient to subtract @xmath51 from the @xmath1th element of @xmath28 to obtain @xmath62 this positions @xmath60 exactly on the perpendicular vertex of @xmath38 .
projecting the vector @xmath60 onto the lattice basis vectors can also be done by the multiplication @xmath63 where @xmath64 .
the important element here is @xmath57 , which is the value that should be multiplied to the lattice basis vector @xmath65 to create the projected vector @xmath60 .
this element determines the corresponding range for @xmath58 @xcite @xmath66 where @xmath67 and @xmath68 is the squared radius of the examined @xmath35-dimensional layer .
the sphere decoder is applied recursively to search this @xmath35-dimensional layer .
thereafter the next @xmath46 value in is generated and a new @xmath35-dimensional layer is searched .
generalizing , the closest point in an @xmath32-dimensional layer is found by dividing the layer into @xmath31-dimensional layers , searching each of these separately , and then proceeding to the next @xmath32-dimensional layer .
we will refer to this process of decreasing and increasing @xmath32 as _ moving down and up the layers , _ resp .
we derive for @xmath69 @xmath70 where @xmath71 is the received vector @xmath28 projected onto an @xmath32-dimensional layer , and @xmath72 gives the coefficients of @xmath71 expressed as a linear combination of the lattice basis vectors .
( in a zero - dimensional layer , which is a lattice point , @xmath73 and @xmath74 . ) assuming an @xmath32-dimensional sphere similar to fig
. [ fig:1 ] , the orthogonal displacement between the projected vector @xmath71 and the examined @xmath31-dimensional layer is @xmath75 based on a lower - triangular form and the interpretation that @xmath76 only affects the @xmath32th component of @xmath71 , for @xmath77 @xmath78 similarly , the bounds for every @xmath32-dimensional layer are @xmath79 where @xmath80 is the squared distance from the received vector @xmath28 to the projected vector @xmath81 and @xmath82 is the squared radius of the examined @xmath32-dimensional layer . hence , @xmath83 denotes the euclidean distance between the received vector @xmath28 and a potential closest point @xmath84 . finally , the range of @xmath85 for @xmath86 is @xcite @xmath87 where the _ projection value _
@xmath88 is the value that should be multiplied with the lattice basis vector @xmath89 to create the projected vector @xmath71 .
in this section , we claim that most of the arithmetic operations in standard sphere decoders are redundant and we propose methods to avoid them , thus increasing the decoding speed .
the redundant operations are of two types : for @xmath29-based implementations , numerous quantities are calculated which are never used , and for @xmath0-based implementations , some quantities are calculated more than once . in both cases ,
the source of the problem is the way the projection values are calculated .
most of the numerical operations carried out in standard sphere decoders based on @xmath29 are related to the projection of the received vector @xmath28 , or its lower - dimensional counterpart , onto the lattice basis vectors as in . defining a matrix @xmath90 whose rows are @xmath91 , it follows from that all elements of this matrix are updated from the elements immediately below .
however , the only values that are required in the sphere decoder algorithms are the diagonal elements @xmath88 , used in and .
thus , the elements located above the diagonal of @xmath90 are not required to be calculated .
they correspond to @xmath92 values that have already been calculated in previous stages of the algorithm , see .
the sphere decoder proposed in @xcite always updates the first @xmath32 elements of @xmath93 simultaneously .
for instance , if we are in an @xmath32-dimensional layer after computing @xmath88 , we update @xmath94 for all @xmath95 .
these values may be used later to update @xmath96 for some @xmath97 after moving down the layers . but why should one project the entire vector @xmath71 to the lattice basis vectors , and calculate the @xmath94 for all @xmath95 , when they are not supposed to be used at that stage of the algorithm , and possibly not at all ?
the answer to this question inspires an intelligent algorithm to manage the projection of @xmath28 and updating the @xmath98 values , based on following criteria : @xmath99 as explained in sec .
[ sec:3.1 ] , we are just interested in elements located in the lower triangular form of @xmath90 .
@xmath99 the last row of @xmath90 , @xmath100 , is just calculated once since there exists just a single @xmath1-dimensional layer .
@xmath99 according to and , updating an element @xmath98 ( with @xmath101 ) requires knowledge of both @xmath102 and @xmath103 .
@xmath99 unlike the row - wise updating method in @xcite , we propose updating @xmath90 column - wise , i.e. , updating @xmath104 for a suitable value of @xmath105 before calculating the desired value of @xmath88 .
@xmath99 if we move to an @xmath32-dimensional layer , the first @xmath106 elements of @xmath93 and of other @xmath107 vectors above that row will be affected , since we are projecting the received vector @xmath28 to this new @xmath32-dimensional layer .
however , the elements below @xmath93 will remain unaffected .
@xmath99 our main target at each stage , when we are moving towards the lower - dimensional layers , is just to update the @xmath88 values .
the other @xmath98 values for @xmath108 will be updated if and only if they are needed to calculate the @xmath88 values .
@xmath99 the algorithm should track of the movement down and up the layers in order to avoid the recalculation of values that remain unchanged , see sec .
[ sec:3.3 ] . in sec .
[ sec:5 ] , we demonstrate by simulations how the complexity of sphere decoder algorithms , for both lattices and finite constellations , is reduced due to the method outlined above for projecting of the received vector @xmath28 . also in the @xmath0-based implementations , the time - consuming step is to calculate the projection values , which we denote with @xmath88 in @xmath29-based implementations , as discussed in sec .
[ sec:3.1 ] , and @xmath109 in @xmath0-based implementations . according to @xcite , which uses the same recursions as @xcite ,
the projection value is calculated as @xmath110 , where @xmath111 and @xmath112 moving further down the layers in order to calculate the @xmath113 projection value for @xmath97 , one can notice that part of the sum in is already calculated and does not need to be recalculated if stored in memory .
hence , we define @xmath114 for @xmath115 and @xmath116 for @xmath117 . as a result , we can calculate @xmath118 for @xmath119 and @xmath120 for @xmath117 , which requires fewer operations than .
we collect the elements @xmath121 in a lower - triangular matrix @xmath122 , which is completely irrelevant to the matrix @xmath90 discussed in sec .
[ sec:2 ] and [ sec:3.1 ] .
however , the optimized projection method proposed in sec . [ sec:3.1 ] to update the @xmath98 values , with some minor modifications , can be similarly applied herein to update the @xmath121 values .
the changes are as follows : @xmath99 the last row of @xmath122 is the zero vector .
@xmath99 according to , updating @xmath121 requires knowledge of both @xmath123 and @xmath124 values .
@xmath99 if we move to an @xmath32-dimensional layer , all the elements of the @xmath32th row of @xmath122 , and of other rows above that row , will be affected , since we are investigating a new @xmath32-dimensional layer . however , the elements below that row will remain unaffected .
@xmath99 our main target at each stage , when we are moving towards the lower - dimensional layers , is just to update the @xmath125 values .
the other @xmath121 values for @xmath108 will be updated if and only if they are needed to calculate the @xmath125 values .
@xmath99 similarly to sec .
[ sec:3.1 ] , we should keep track of the movement up and down the layers .
based on the preceding criteria , one can avoid starting from the @xmath1th layer and updating all @xmath126 elements located in the @xmath32th column of @xmath122 before calculating the objective @xmath125 value .
while this significantly reduces the complexity of the algorithm , the memory write operations are increased .
standalone representations of the old and new algorithms , @xmath0-based and @xmath29-based versions , for lattices and finite constellations , are given in fig .
[ fig : algorithms ] , all based on the se enumeration strategy .
the specifications are intended to be sufficiently detailed to allow a straightforward implementation , even without knowledge of the underlying theory . as starting points , we use the @xmath0-based algorithm called `` algorithm ii '' in @xcite , labeled with 2 in fig . [
fig : algorithms ] , and the @xmath29-based algorithm `` decode '' in @xcite , here labeled with 3 .
the loops have been restructured for consistency between the algorithms , but the calculations in fig .
[ fig : algorithms ] are exactly the same as in @xcite .
indeed , all algorithms for lattice decoding ( algorithms 1 , 3 , 5 , and 7 ) visit the same layers @xmath85 , in the same order , and return the same result @xmath127 , although they calculate different intermediate quantities .
a similar note holds for decoding finite constellations ( algorithms 2 , 4 , 6 , and 8) .
after the initialization , the algorithms are divided into three parts . in the first part
, we move down the layers ( decrease @xmath32 ) , as long as the squared euclidean distance @xmath80 between the received vector @xmath28 and the projected vector @xmath81 is less than the squared euclidean distance @xmath44 between the received vector @xmath28 and the closest lattice point detected so far . in the second part , we move up in the hierarchy of layers ( increase @xmath32 ) as long as @xmath128 .
moreover , before leaving each of these parts , we store the minimum and maximum level @xmath32 that has been visited .
these values are used in the last part of the algorithm , which only belongs to the new algorithms . the method to manage the recursive projection of @xmath93 or the calculation of @xmath121 is proposed in the last part .
the value of @xmath129 for @xmath130 denotes the starting point for the recursions in and in order to update the objective @xmath88 or @xmath125 values .
for instance , @xmath131 indicates that in order to update @xmath88 , we should start the projection from @xmath132th layer , where @xmath133 , and calculate @xmath98 for @xmath134 . due to the well - documented performance gain that the se enumeration strategy brings to sphere decoders ,
we apply herein the proposed refinement only to the se strategy .
however , the same refinement can be applied to the original fp enumeration strategy .
it is also applicable to most , or all , of the numerous sphere decoder variants , optimal as well as suboptimal , that have been developed in the last decade .
herein , we evaluate the effectiveness of the proposed smart vector projection technique on the sphere decoder algorithms based on se enumeration strategy , for both lattices and finite constellations .
all eight algorithms are implemented according to the pseudocode presented in fig .
[ fig : algorithms ] .
we base our performance comparison measure on counting the number of floating point operations ( flops ) and integer operations ( intops ) that each algorithm carries out to reach the closest lattice point .
both types of operations include addition , subtraction , multiplication , division , and comparison , but not @xmath135 loop counters , whose role differs between programming languages .
the @xmath136 operation is counted as a single floating point operation , and @xmath137 in sec .
[ sec:5.2 ] is counted as one floating point operation for 2-pam and two for 4-pam . to compare the complexity of two algorithms , typically an old and a new one , we generate @xmath138 random generator matrices @xmath139 , and for each @xmath140 we generate @xmath141 random
received vectors @xmath142 .
the same vectors are decoded using both algorithms and the number of operations @xmath143 is counted , which could be either flops or intops .
the average gain with the new algorithm is reported as @xmath144 we generate the lattice generator matrices with random numbers , drawn from i.i.d .
zero - mean , unit - variance gaussian distributions .
the random input vectors are generated uniformly inside a voronoi region according to @xcite .
our simulation results are based on averaging over @xmath145 different generator matrices .
the number of input vectors @xmath141 depends on the dimension @xmath1 of the lattices .
fewer input vectors are examined in high dimensions , to the extent that we ensure that the plotted curves are reasonably smooth .
[ fig : flops ] compares the number of flops for the standard @xmath0- and @xmath29-based algorithms ( algorithms 1 and 3 in fig .
[ fig : algorithms ] ) with the new algorithms proposed in this paper ( algorithms 5 and 7 ) .
it can be seen that the @xmath0- and @xmath29-based implementations have about the same complexity , but both can be significantly improved .
[ cc][][0.8]dimension [ ] [ ] [ 0.8]flops [ cc][][0.8 ] [ cc][][0.7]@xmath0-based new [ cc][][0.7]@xmath0-based old [ cc][][0.7]@xmath29-based new [ cc][][0.7]@xmath29-based old [ cc][][0.8]@xmath146 [ cc][][0.8]@xmath147 [ cc][][0.8]@xmath148 [ cc][][0.8]@xmath149 [ cc][][0.8]@xmath150 [ cc][][0.8]@xmath151 [ cc][][0.8]@xmath152 [ cc][][0.8]@xmath153 [ cc][][0.8]@xmath154 [ cc][][0.8]@xmath155 [ cc][][0.8]@xmath156 [ cc][][0.8]@xmath157 [ cc][][0.8]@xmath158 [ cc][][0.8]@xmath159 [ cc][][0.8]@xmath160 [ cc][][0.8]@xmath161 [ cc][][0.8]@xmath162 [ cc][][0.8]@xmath163 [ cc][][0.8]@xmath164 [ cc][][0.8]@xmath165 [ cc][][0.8]dimension [ ] [ ] [ 0.8]gain [ cc][][0.8 ] [ cc][][0.7]@xmath0-based flops without reduction [ cc][][0.7]@xmath0-based flops with lll reduction [ cc][][0.7]@xmath29-based flops without reduction [ cc][][0.7]@xmath29-based flops with lll reduction [ cc][][0.7]@xmath0&@xmath29-based intops without reduction [ cc][][0.7]@xmath0&@xmath29-based intops with lll reduction [ cc][][0.8]0 [ cc][][0.8]5 [ cc][][0.8]10 [ cc][][0.8]15 [ cc][][0.8]20 [ cc][][0.8]25 [ cc][][0.8]30 [ cc][][0.8]35 [ cc][][0.8]40 [ cc][][0.8]45 [ cc][][0.8]50 [ cc][][0.8]55 [ cc][][0.8]60 [ cc][][0.8]0.5 [ cc][][0.8]1 [ cc][][0.8]1.5 [ cc][][0.8]2 [ cc][][0.8]2.5 [ cc][][0.8]3 [ cc][][0.8]3.5 [ cc][][0.8]4 [ cc][][0.8]4.5 a preprocessing stage was applied to each lattice , replacing the generator matrix with another generator matrix for the same lattice via the so - called lenstra lenstra lovsz ( lll ) reduction @xcite .
the operations needed for the reduction were not counted , since the preprocessing is only done once for each lattice , regardless of the number of received vectors .
the gain with new algorithms increases linearly with dimension , while the reduction does not change the ratios substantially .
the drawback is a somewhat larger number of intops , but the penalty converges to a mere 15@xmath166 increase at high dimensions . in simulations
it was observed that most of the operations in the algorithms are flops , especially as the dimension increases .
for instance , at dimension 60 with the old @xmath29-based algorithm , the flops are roughly 10 times more than the intops .
hence , flops dominate the complexity of the algorithms and intops have a relatively small effect on the overall complexity .
we also measured the running time for the algorithms . as expected , the gain increases roughly linearly with the dimension , similarly to the flops curves in fig . [ fig:2 ] .
however , the slope of the curve varies significantly between different processors and compilers , which is why we did not include running time in fig .
[ fig:2 ] . at dimension 60
, the gain ranged from 1.7 ( amd processor , visual c++ compiler ) to 2.7 ( intel processor , gcc compiler ) , for the @xmath29-based algorithm without reduction .
we can thus safely conclude that the reduced number of operations translates into a substantial speed gain , but how much depends on the computer architecture
. the channel model in for an @xmath14-pam constellation is considered , where the average symbol energy of the constellation , @xmath167 , is calculated from the signal set @xmath168 and the snr is defined as @xmath169 , where @xmath170 is the average energy per bit and @xmath16 is the double - sided noise spectral density .
[ cc][][0.8]dimension [ cc][][0.8]gain [ cc][][0.8 ] [ cc][][0.675]@xmath0-based snr=0db [ cc][][0.675]@xmath0-based snr=5db [ cc][][0.675]@xmath0-based snr=10db [ cc][][0.675]@xmath29-based snr=0db [ cc][][0.675]@xmath29-based snr=5db [ cc][][0.675]@xmath29-based snr=10db [ cc][][0.8]0 [ cc][][0.8]10 [ cc][][0.8]20 [ cc][][0.8]30 [ cc][][0.8]40 [ cc][][0.8]50 [ cc][][0.8]60 [ cc][][0.8]70 [ cc][][0.8]80 [ cc][][0.8]1 [ cc][][0.8]1.5 [ cc][][0.8]2 [ cc][][0.8]2.5 [ cc][][0.8]3 [ cc][][0.8]3.5 [ cc][][0.8]4 [ cc][][0.8]4.5 [ cc][][0.8]5 [ cc][][0.8]dimension [ cc][][0.8]gain [ cc][][0.8 ] [ cc][][0.675]@xmath0-based snr=0db [ cc][][0.675]@xmath0-based snr=5db [ cc][][0.675]@xmath0-based snr=10db [ cc][][0.675]@xmath29-based snr=0db [ cc][][0.675]@xmath29-based snr=5db [ cc][][0.675]@xmath29-based snr=10db [ cc][][0.8]0 [ cc][][0.8]5 [ cc][][0.8]10 [ cc][][0.8]15 [ cc][][0.8]20 [ cc][][0.8]25 [ cc][][0.8]30 [ cc][][0.8]35 [ cc][][0.8]40 [ cc][][0.8]45 [ cc][][0.8]50 [ cc][][0.8]1 [ cc][][0.8]1.5 [ cc][][0.8]2 [ cc][][0.8]2.5 [ cc][][0.8]3 [ cc][][0.8]3.5 [ cc][][0.8]4 [ cc][][0.8]4.5 the gain in flops is presented in figs .
[ fig:5][fig:6 ] for 2-pam and 4-pam constellations , resp . , averaged over 100 random channel matrices @xmath0 with i.i.d .
zero - mean , unit - variance elements .
the same general conclusion as for lattices holds for finite constellations too : the new algorithms provide a substantial complexity gain , and the gain increases linearly with the dimension .
however , in contrast to lattice decoding , the gains are here higher for @xmath0-based implementations .
furthermore , the gains increase at low snr , and 4-pam offers slightly higher gains than 2-pam .
u. fincke and m. pohst , `` improved methods for calculating vectors of short length in a lattice , including a complexity analysis , '' _ mathematics of computation _ , vol .
44 , no . 170 , pp .
463471 , apr .
1985 .
w. k. ma , b. n. vo , t. n. davidson , and p. c. ching , `` blind ml detection of orthogonal space - time block codes : efficient high - performance implementations , '' _ ieee trans . signal process .
_ , vol .
54 , no . 2 ,
738751 , feb .
2006 . a. burg , m. borgmann , m. wenk , m. zellweger , w. fichtner , and h. blcskei , `` vlsi implementation of mimo detection using the sphere decoding algorithm , ''
_ ieee j. solid - state circuits _ , vol .
40 , no . 7 , pp .
15661577 , july 2005 .
b. m. hochwald , c. b. peel , and a. l. swindlehurst , `` a vector - perturbation technique for near - capacity multiantenna multiuser communication part ii : perturbation , '' _ ieee trans .
_ , vol .
53 , no . 3 , pp . 537544 ,
|
most of the calculations in standard sphere decoders are redundant , in the sense that they either calculate quantities that are never used or calculate some quantities more than once . a new method , which is applicable to lattices as well as finite constellations , is proposed to avoid these redundant calculations while still returning the same result .
pseudocode is given to facilitate immediate implementation .
simulations show that the speed gain with the proposed method increases linearly with the lattice dimension . at dimension 60 ,
the new algorithms avoid about 75 % of all floating - point operations .
|
recent advances in infrared and x - ray array technologies have greatly increased the spatial coverage and resolutions in near - ir and x - ray observations , providing powerful tools to compare high - resolution radio , ir and x - ray images of the complex region of the galactic center .
sgr a@xmath0 is a bright compact radio source which is considered to be the strongest candidate for a 2.5@xmath10 black hole at the dynamical center of the galaxy ( eckart and genzel 1997 ; ghez et al .
this source is surrounded by diffuse ionized gas as well as clusters of hot and cool stars seen in near - ir wavelengths .
detailed comparison of radio and near - ir images was recently made with a new approach by menten et al .
( 1997 ) , who detected a number of sio ( 0.7 cm ) and h@xmath11o ( 1.3 cm ) masers associated with known compact stellar sources seen in the diffraction limited 2.2 @xmath12 m infrared images .
astrometric measurements between radio and 2 @xmath12 m images of this region were used to determine the position of sgr a@xmath0 with respect to other well - known near - ir sources .
although the relative position of sgr a@xmath0 is accurately measured within 30 mas in this technique , the absolute position of sgr a@xmath0 was known with an accuracy of 0.2@xmath13 .
a more accurate position of this source and its subsequent x - ray counterpart in future sub - arcsecond observations has important implications on the nature of high energy activity predicted from models of this source ( melia 1994 ; narayan et al .
1995 ) .
the absolute positions of radio sources are routinely obtained to an accuracy of a few milliarcseconds using centimeter wavelength vlbi techniques .
however , the large scattering diameter of sgr a@xmath0 ( yusef - zadeh et al . 1994 ) makes the application of centimeter wavelength vlbi less feasible , as the source is completely resolved on longer baselines .
thus , shorter baseline arrays and/or shorter wavelengths are needed resulting in reduced positional accuracy .
the most recent measurements of the location of sgr a@xmath0 with the smallest error bar is given as @xmath2 , @xmath3 [ 1950 ] = @xmath14 29314@xmath15 , @xmath16 ( rogers et al .
this position is determined with respect to the reference source nrao 530 using vlbi techniques at 3 mm . to improve the absolute position of sgr a@xmath0
, we obtained archival vla data sets at a number of frequencies taken over the last 13 years . by averaging the position of sgr a@xmath0 at different frequencies and at different epochs , the new position of sgr a@xmath0
is determined with an accuracy of an order of magnitude better than earlier measurements .
the averaging of the data over such a long span of time and frequency can be justified .
the proper motion measurements of this source is too small with respect to reference extragalactic radio sources ( reid et al .
1998 ; backer 1996 ) to affect individual measurement . also , the change in the position of sgr a@xmath0 as a function of frequency is too small to be significant compared to other systematic errors .
the frequency dependence of the size of sgr a@xmath0 due to scattering of radio waves will not change the centroid of the position of sgr a@xmath0 as a function of frequency ( yusef - zadeh et al .
it is worth mentioning that the detection of frequency dependence of the position of sgr a@xmath0 at a milliarcsecond level , due to its intrinsic radio emission , has the potential to support the asymmetrical bondi - hoyle accretion model ( melia 1994 ) .
table 1 shows the list of 15 data sets that were based on continuum observations made with the very large array of the national radio astronomy observatory at 6 ( 5 ghz ) , 3.6 ( 8.4 ghz ) , 2 ( 15 ghz ) , 1.2 ( 20 ghz ) , and 0.7 cm ( 44 ghz ) between 1986 and 1999 .
all had a bandwidth ranging between 2 @xmath17 50 and 2 @xmath17 12.5 mhz in each right and left circular polarization .
each of these original data sets had relatively good uv coverage and none were done in snapshot mode with less than one hour of integration time on the source . with the exception of 0.7 cm data which was taken in the c configuration ,
all the observations were done in either the a or the b configurations of the vla to obtain the highest resolution images of sgr a@xmath0 with minimal contamination or confusion by ionized features in the vicinity of sgr a@xmath0 .
the low resolution images of sgr a@xmath0 at 20 and 90 cm suffer from this source confusion as well as optical depth effects due to intervening thermal and nonthermal continuum sources .
thus , 20 and 90 cm observations were not used . in all observations , nrao530 and 3c286 were chosen as the phase ( astrometric ) and flux ( photometric ) calibrators , respectively .
all processing was done using the nrao aips analysis package ; the standard calibration was applied and the field around sgr a@xmath0 was imaged .
since the observations were all made in the b1950 coordinate system , the data were analyzed in this system .
the vla calibrator b1950 coordinate system is referred to epoch 1979.5 .
no self calibration was done on any of the data as this procedure loses the absolute position of a source in order to minimize atmospheric phase errors .
prior to self - calibration , no significant emission from sgr a@xmath0 was detected at 0.7 cm in the a and b configurations , so these data were excluded .
vla continuum data consists of two independent frequency bands .
the final images based on combining these two frequency bands were used to determine the position of sgr a@xmath0 with aips routine jmfit by fitting a gaussian to the source .
fits done using images from the individual frequency bands showed no significant positional shift . at short wavelengths ,
the phase fluctuations degrading the positional accuracy are due mainly to variable refraction in the troposphere ; at longer wavelengths the ionosphere adds increasingly important contributions .
variable ionospheric refraction is particularly significant for observations made at 6 cm during the day when the ionosphere is densest .
solar heating of the troposphere also increases the variations in refraction .
so , we have assigned weights of 1 and 2 to day and night - time observations , respectively . for observations that were carried out between night and day ,
the assigned weight was 1.5 .
the atmospheric effects are aggravated by the low declination of the source and the calibrator and the resultant low elevation angles of the observations .
these effects are minimized by restricting the _ uv _ data to near transit .
the data sets selected within one to three hours of the highest elevation improved considerably the dispersion in measured position of sgr a@xmath0 among the different epochs .
errors in atmospheric refraction affect mainly the measurements of the declination of sgr a@xmath0 .
the astrometric calibrator used , nrao530 , is about 15@xmath18 north of sgr a@xmath0 .
since the observations were restricted to those near transit , this corresponds to a difference primarily in elevation .
corrections to the atmospheric refraction model are determined from the calibrator observations which is at a significantly higher elevation than sgr a@xmath0 ( i.e. at lower air mass ) ; any unmodeled differences in atmospheric refraction between the calibrator and sgr a@xmath0 will appear primarily as errors in declination .
a further potential source of systematic error is the assumed position of the calibrator , nrao530 .
our measurements are of position relative to that of the calibrator whose absolute position is known ; any errors in the assumed position of nrao530 will directly affect the derived position of sgr a@xmath0 . in all our measurements ,
the position of nrao530 was constant and was assumed to be at ( b1950 epoch 1979.5 ) @xmath2 , @xmath3 = @xmath19 , @xmath20 .
this position is consistent to within 5 mas of the vlbi source position given by johnston et al .
( 1995 ) and by us naval observatory s ( usno ) astrometry list ( ftp://casa.usno.navy.mil/navnet/n9798.sor ) after the correction between the vla and usno frames is made .
the position of nrao530 assumed is an absolute position measured by vlbi techniques which are insensitive to structure larger than a few tens of milliarcseconds .
nrao530 is suspected of having structure on the scale of several hundred milliarcseconds which would not affect the vlbi measurements but will cause an apparent shift of the position when viewed with the lower vla resolution .
any difference between the centroid of the position of nrao530 at vlbi and vla resolutions will appear as an error in the derived position of sgr a@xmath0 .
table 1 lists 15 data sets , three of which at 1.3 cm , four at 2 cm , five at 3.6 cm , two at 6 cm , and one at 0.7 cm . the date of each observation , the gaussian fitted position with 1 @xmath21 error , duration of the observation and the corresponding weight in the final averaging are also given .
the errors of the fitted positions depend on the signal - to - noise ratio and are much less than the errors introduced by the atmosphere .
the rms scatter within the data set , as described below , is about ten times the fitted noise , thus we have not used the errors of the fitted positions in any of our analysis .
the mean position of sgr a@xmath0 , both weighted and unweighted , as well as the position determined by rogers et al .
1994 are also listed at the bottom of the table .
the most descripant point at 1.3 cm was not included in the final averaging .
the rms scatter of the individual measurements in table 1 ( either weighted or unweighted ) is about 0.035@xmath13 in right ascension and 0.055 `` in declination .
this is consistent with the assumption that the errors are dominated by uncertainties in atmospheric refraction which should produce a larger scatter in declination .
this is further supported by the two pairs of positions measured on the same day , in which the declinations differed by about 0.030 '' , or nearly half of the day - to - day scatter .
the one sigma errors in the mean position are taken to be the rms scatter of the sample divided by the square root of the number of observations minus one .
figure 1 shows the right ascension and declination of sgr a@xmath0 with one sigma error given by rogers et al .
( 1994 ) against 15 positions determined from our measurements .
all the new positions fall to the sw quadrant of the position derived by rogers et al .
the weighted position of sgr a@xmath0 with the weighted mean epoch of our observation being 1992.4 is ( b1950 ) @xmath2 , @xmath3 = @xmath4 293076@xmath5 , @xmath22 or ( j2000 ) @xmath2 , @xmath3 = @xmath8 400383@xmath5 , @xmath23 .
these measurements are in good agreement with the less precise measurement of rogers et al .
1994 as well as with the position of sgr a@xmath0 determined from recent vlbi measurements ( reid et al .
1999 ) .
reid , m. j. , readhead , a. c. s. , vermulen , r. c. & treuhaft , r. n. 1998 , in the central parsecs , galactic center workshop 1998 , asp conf .
ser . , eds . h. falcke , a. cotera , w. huschl , f. melia , and m. rieke ( in press )
|
the absolute position of the compact radio source at the dynamical center of the galaxy , sgr a@xmath0 , was known only to an accuracy of @xmath1 in spite of its accurate location with respect to near - ir stellar sources to within 30 milliarcsecond ( mas ) . to remedy this poor positional accuracy ,
we have selected 15 high - resolution , high - frequency vla observations of sgr a@xmath0 carried out in the last 13 years and determined the weighted average position with the average epoch 1992.4 to be at @xmath2 , @xmath3[1950 ] = @xmath4 293076@xmath5 , @xmath6 , or @xmath2 , @xmath7[2000 ] = @xmath8 400383@xmath5 , @xmath9 which agrees with earlier published values to within the @xmath1 error bars of the earlier measurements . an accurate absolute position of sgr a@xmath0 can be useful for its identification with sources at other wavelengths , particularly , in soft and hard x - rays with implications for the models of a massive black hole at the galactic center .
submitted to the editor of the apj letters on february 4 , 1999 i v i
|
the anomalous thermodynamic and kinetic behavior of water is known to play a fundamental role not only in many physical and chemical processes in materials science , but also in biological , geological and terrestrial processes in nature @xcite .
for this reason , a lot of effort has been devoted to rationalizing water s anomalous behavior in a coherent and simple physical picture , but no consensus has yet emerged .
one of the breakthroughs in this endeavor was the discovery of water s polyamorphism , i.e. the existence of amorphous coexisting phases in the supercooled region of the phase diagram .
distinct states have indeed been found in glassy water , called low - density ( lda ) , high - density ( hda ) and very high - density ( vhda ) @xcite amorphous ices , which can interconvert with each other by the application of pressure .
it is believed that the transition between the amorphous ices connects to a liquid - liquid first - order phase transition line above the glass transition temperature ( @xmath1 ) , and terminates at a critical point @xcite , but the fundamental nature of this transition is still being debated @xcite .
the verification of the liquid - liquid critical point scenario ( llcp ) is hindered by water s crystallization at large supercooling @xcite , so that much of the evidence comes from computer simulation studies @xcite , and only indirectly from experiments where crystallization is suppressed either by strong spatial confinement @xcite or by mixing an anti - freezing component @xcite . one way to understand water s polyamorphic behavior is to introduce the concept of locally favored structures @xcite , which are defined as particular long - lived molecular arrangements which correspond to some local minima of the free energy . in this view , water s polyamorphism comes from the competition between two different types of molecular arrangements @xcite : one in which the different tetrahedral units form open structures , and the other with a smaller specific volume due to a high degree of interpenetration .
the presence of two amorphous fluid phases has indeed been observed in computer simulations of some models of water , where the freezing transition can be avoided @xcite .
but whether locally favored states exist above the critical region is still a matter of debate .
many physical quantities exhibit behaviors suggestive of two states in liquid water , such as infrared and raman spectra @xcite , and the presence of an isosbectic point in raman spectra has been regarded as a clear indication supporting a mixture model since its finding by walrafen @xcite .
some evidence for the inhomogeneous structure of water was also reported in experimental studies of x - ray absorption spectroscopy , x - ray emission spectroscopy and x - ray small angle scattering @xcite , but these results are highly debated @xcite , especially since the majority component at room temperature was proposed to be associated with the break - up of the tetrahedral structure . despite these pieces of evidence supporting a two - state picture , the lack of a clear connection between these experimental observables and the amount of locally favored structures has made it difficult to estimate the fractions of the two states in a convincing manner . from a phenomenological standpoint ,
two - state models have been extremely successful in describing the anomalies of water using a restricted number of fitting parameters @xcite , and recent simulations by cuthbertson and poole @xcite have opened the way for a quantitative assessment of these models from miscroscopic informations , but only for state points around the widom line , i.e. , at temperatures not accessible to experiments and far from the anomalies .
the essential difficulty in defining locally favored states is finding a structural order parameter that directly correlates with water s anomalies .
several attempts have been made , each differing in the microscopic definition of the states involved .
examples include states based on ice polymorphs @xcite , tetrahedral order @xcite , or relative distance between neighbors @xcite .
to overcome this difficulty , we introduce a new structural order parameter that quantifies the degree of translational order in the second shell .
we show that , for one of the most reliable computer models , water anomalies are a consequence of translational ordering of the second shell and can be described very well by a two - state model .
we further identify the structural characteristics of the locally favored structures of water and their link to the unusually large degree of supercooling of the liquid phase .
first we explain a key idea to identify the relevant structural order parameter for characterizing the water structure .
our locally favored structures correspond to local structures with low energies @xmath2 , high specific volumes @xmath3 and low degeneracy @xmath4 . in the following
we denote these structures as the @xmath5 state .
in contrast , thermally excited states are characterized by a high degree of disorder and degeneracy , low specific volumes and high energies .
we label these structures as the @xmath6 state . in formulas , @xmath7 , @xmath8 and @xmath9 . to identify the @xmath5 state we introduce a new order parameter @xmath10 which measures local translational order in the second shell of neighbors .
the importance of the second shell structure was notably pointed out by soper and ricci @xcite .
translational order is a measure of the relative spacing between neighboring particles , and it is one of the fundamental symmetries broken at the liquid - to - solid transition @xcite .
locally , a molecule is in a state of high translational order if the radial distribution of its neighbors is ordered .
a liquid , by definition , can not have full translational order , but it might display translational order on shorter scales .
water molecules , even at high temperatures , displays a high level of tetrahedral symmetry , meaning that a high degree of translational order is always present up to the first shell of nearest neighbors . in this sense , tetrahedral order itself is not enough to describe water s anomalies . to determine locally favored states
, we focus instead on `` translational order of second nearest neighbors '' .
the operational definition goes as follows ( a schematic representation is given in fig .
[ fig:1]a ) . for water molecule @xmath11 ( labeled * 0 * in fig .
[ fig:1]a ) we order its neighbors according to the radial distance @xmath12 of the oxygen atoms ; the order parameter @xmath13 is then defined as the difference between the distance @xmath14 of the first neighbor not hydrogen bonded to @xmath11 ( with label * 5 * in the figure ) , and the distance @xmath15 of the last neighbor hydrogen bonded to @xmath11 ( labeled * 4 * ) .
as we will show , the @xmath5 state having high translational symmetry , is characterized by large values of @xmath10 , with a clear separation between first and second shell ( for example when the fifth molecule is in position * 5 * in fig .
[ fig:1]a ) . but
these structures should not be confused with local crystalline structures , as they generally lack orientational order , i.e. , the neighbors in the second shell are not oriented according to the crystal directions ( with well defined eclipsed and staggered configurations ) , due to their embedding in water s disordered network . in the @xmath6 state the second shell
is collapsed , with a distribution of @xmath10 values roughly centered around @xmath16 ( in fig .
[ fig:1]a this state is obtained for example when the fifth neighbor is in position * 5 * ) and comprising many configurations with negative values of @xmath10 , resulting from the penetration of the first shell from an oxygen belonging to a distinct tetrahedra . at ordinary thermodynamic conditions ,
these two states are mixed and the free energy ( @xmath17 ) of the mixture takes the form of a regular solution , i.e. , the simplest non - ideal model of liquid mixture @xcite .
@xmath18+js(1-s),\ ] ] where @xmath19 is the fraction of the @xmath5 state , @xmath20 ( @xmath21 ) is the free energy of the pure component , @xmath22 , and @xmath23 is the coupling between the two states , i.e. , the source of non - ideality . unlike ordinary regular solutions ,
the fraction @xmath19 is not fixed externally , but by the equilibrium of the conversion reaction between the two states , @xmath24 , obtained by equating their chemical potentials ( @xmath25 ) @xmath26 if @xmath27 the mixing is endothermic , and the model displays a critical point at @xmath28 , below which the system undergoes a liquid - liquid demixing transition .
to test our model we conducted molecular dynamics simulations of the tip4p/2005 model of water , one of the best models of the liquid state @xcite .
simulations were run for many state points covering a large area of the liquid state , with temperatures ranging from @xmath29 k to @xmath30 k , and pressures ranging from @xmath31 kbar to @xmath32 kbar . to identify hydrogen bonds we adopted the definition found in ref .
@xcite and widely adopted in simulations of liquid water .
temperature is always expressed in k , pressure in bar , distance in nm , density @xmath6 in g/@xmath33 , and isothermal compressibility @xmath34 in bar@xmath35 .
for more details on the simulation methods refer to the methods section .
, @xmath36-dependence of the structural order parameter .
( _ a _ ) schematic representation of the local environment around a water molecule ( labeled * 0 * ) , showing the position of the hydrogen - bonded molecules that form the first shell ( with labels from * 1 * to * 4 * ) .
also depicted is the closest non - hydrogen - bonded molecule for two different states : the s state , when the fifth neighbor is in position * 5 * , so that there is a clear separation between first and second shell , with the order parameter @xmath10 representing the distance between them ; the @xmath6 state , with the fifth neighbor in position * 5 * so that the second shell is collapsed onto the first one .
( _ b _ ) distribution function of the order parameter @xmath10 at ambient pressure ( @xmath37 bar ) and at different temperatures .
the temperatures range from @xmath29 k to @xmath30 k in steps of @xmath38 k ( the arrows indicate increasing temperatures ) .
the inset shows the distribution function at @xmath39 k obtained from the simulations ( circle symbols ) , and the decomposition in two gaussian populations : the @xmath6 population in dashed line , the @xmath5 population in dashed - dotted line , and the sum of both populations in the continuous ( red ) line .
( _ c _ ) values of the fraction of the @xmath5 state ( @xmath19 ) as a function of temperature for all simulated pressures .
the symbols are the values obtained by the decomposition of the order parameter distribution , @xmath40 , at the corresponding state point .
continuous lines are fits according to the two - state model , with the following parameters defined in the text : @xmath41 , @xmath42 , @xmath43 , @xmath44 and @xmath45 . the full diamond shows the location of the critical point , @xmath46 k , @xmath47 bar @xcite.,width=302 ] figure [ fig:1]b shows the probability distribution of the order parameter @xmath10 at ambient pressure ( @xmath37 bar ) and for different temperatures , ranging from @xmath29 k to @xmath30 k. the change of the distribution shows a remarkable non - monotonic change ( of both height and width ) with temperature .
this behavior can be rationalized by decomposing the distribution function in two gaussian populations , each varying monotonically with temperature ( see methods for the details ) .
an example of this decomposition is shown in the inset of fig .
[ fig:1]b for the distribution function at @xmath39 k , which is close to the state point of density maximum . the @xmath6 population ( dashed line in the inset of fig .
[ fig:1]b ) is centered in proximity of @xmath16 , meaning that there is a large fraction of configurations in which the first shell ( defined as the hydrogen bonded molecules to the central molecule ) is being penetrated by oxygen atoms belonging to different tetrahedral units .
the @xmath5 population is instead characterized by high values of @xmath10 , having a well formed second shell .
the fitting procedure produces a reliable decomposition for state points having @xmath48 ( @xmath19 is the fraction of the @xmath5 state in the model of eq . ( [ eqn : two_state ] ) ) , where the fitting parameters are well behaved .
this covers the whole region of the phase diagram accessible to experiments .
but for deeply supercooled states at low pressures ( below @xmath49 k at ambient pressure ) , the fraction of the @xmath6 state becomes small , and the estimation of @xmath19 from unconstrained fits becomes more difficult ( see methods ) .
nonetheless , as we will show later , the predictions of the model for the deeply supercooled states are still in good agreement with simulations . from the decomposition we can extract the fraction of the @xmath5 state in the liquid , i.e. , the parameter @xmath19 in the two - state model of eq .
( [ eqn : two_state ] ) .
the values of @xmath19 extracted from all simulated state points are shown in fig .
[ fig:1]c . for any pressure ,
the fraction of the @xmath5 state increases monotonically by lowering the temperature , and the increase is steeper at lower pressures . at low temperatures , we can see a big jump in the values of @xmath19 between @xmath50 bar and @xmath51 bar .
this signals the close presence of a liquid - liquid critical point .
in fact , previous studies of tip4p/2005 water ( and with the same system size ) have determined its location at @xmath46 k and @xmath47 bar @xcite , even if the exact location of the critical point is still ambiguous @xcite .
next we fit the two - state model of eq .
( [ eqn : s_value ] ) to our simulation results ( see methods for the details of the fitting ) .
the results of the fit are shown as continuous lines in fig .
[ fig:1]c , demonstrating that the two - state model provides a very good representation of the results obtained from simulations ( represented by symbols in the same figure ) .
note that the agreement holds to very high temperatures , and for all pressures , suggesting a strong microscopic basis for the relevance of our order parameter .
this is in stark contrast to previous attempts to obtain a two - state description of water from microscopic information @xcite , where the agreement was restricted to the deeply supercooled region .
we note that most of previous models have unphysical saturations of the value of @xmath19 at high temperature around 0.5 or even higher , unlike our model where @xmath52 at high temperatures ( see also ref .
@xcite for a review ) .
it is also worth mentioning that the only region of significant discrepancy is limited around the critical point , which is due not only to the higher uncertainty in accessing @xmath19 around the critical region , but possibly also to critical fluctuations which are not incorporated in the present mean - field two - state model ( but which is possible with crossover theory @xcite ) .
bar : dashed lines are simulation results , while symbols are two - state model predictions .
the curves in the inset are traslated on the @xmath53-axis to improve readability .
( _ b _ ) phase diagram of the two - state model .
the widom line ( dashed line ) is by definition the line where @xmath54 .
the compressibility maximum ( squares ) from simulations lies close to the widom line , eventually converging at the critical point ( full diamond ) .
below the critical point the two components undergo macroscopic phase separation .
the spinodals of the transition are denoted with full lines : green for the @xmath6 liquid , and red for the @xmath5 liquid .
circles show the locus of density maximum , which continuously increases as water is stretched , in agreement with experiments at extreme negative pressures @xcite .
dotted lines show the location of the coexistence lines between the liquid and different forms of ice taken from ref .
@xcite , width=302 ] we will now check whether the two - state picture extracted so far can account for water s anomalies . from the free energy of eq .
( [ eqn : two_state ] ) it is possible to derive the anomalous contribution to density , which takes the form @xmath55 with @xmath56 , where @xmath57 is the volume of @xmath58 molecules of the pure @xmath6-state and @xmath59 is the volume difference between @xmath5-state and @xmath6-state . on the @xmath60-@xmath36 dependence of @xmath59 and @xmath57 ,
see methods .
figure [ fig:2]a shows density isobars measured in simulations ( continuous lines ) which are compared to the results from the two - state model ( symbols ) .
we can see that the two - state model correctly represents the intensity and location of the density anomaly for all studied pressures ( with the anomaly disappearing at high pressures ) .
the agreement is more remarkable in the relevant region of the anomalies , while it is approximate at very low temperatures ( for which we know the estimation of @xmath19 being subject to higher uncertainty ) . for example the model predicts a density minimum at around @xmath61 k at ambient pressure ( full circles in fig .
[ fig:2]a ) , which is found instead at @xmath62 k in simulations .
the inset shows the isothermal compressibility @xmath34 for pressures @xmath63 bar . as in the main figure ,
continuous lines are direct simulation results , showing the rapid increase at low temperatures , while symbols are predictions from the model ( see methods ) .
isothermal compressibility anomalies are much harder to describe accurately , both because compressibility is a second derivative of the model s free energy , @xmath17 ( see methods ) , thus suffering bigger uncertainty , and also because its anomaly is located at very low temperatures , where it is more difficult to get reliable estimation of the fraction @xmath19 . nonetheless , as shown in the inset of fig .
[ fig:2]a , the model predicts within @xmath38 k the location of the compressibility maximum , and also its increased intensity at higher pressures ( eventually diverging at the critical point ) . in the appendix
we show that , with minor approximations , it is possible to obtain the fraction of the @xmath5 state not only from the distribution of @xmath10 but also directly from the experimentally measurable o - o radial distribution function .
this opens up a possibility to estimate the structural order parameter , i.e. , the degree of translational order in the second shell , from scattering experiments .
we summarize the phase behavior of the two - state model in the phase diagram of fig .
[ fig:2]b . the widom line of the model , where @xmath54 , lies close to the compressibility maximum line obtained from simulations ( open squares ) , with the two lines converging at the critical point . here
we note that the location of the widom line is determined by the condition @xmath64 , i.e. , the two - state feature without cooperativity ( see eq .
( [ eqn : s_value ] ) ) .
this indicates that the isothermal compressibility anomalies in this temperature range are not due to critical phenomena associated with llcp , but due to the sigmoidal change in @xmath19 characteristic of the two - state model ( schottky - type anomaly ) .
also shown in the phase diagram are the spinodals ( or , stability limits ) of the @xmath5 liquid and the @xmath6 liquid , which in the liquid - liquid critical point scenario are usually called the low density liquid ( _ ldl _ ) and the high density liquid ( _ hdl _ ) respectively .
we now investigate the microscopic features of locally favored states .
we have shown that the @xmath5 state can be identified with configurations having a high degree of translational order , meaning that second nearest neighbors are at approximately the same distance from the central oxygen atom .
we now discuss the orientational order of second nearest neighbors .
crystalline configurations are characterized by full orientational order , with second nearest neighbors occupying the characteristic eclipsed and staggered orientations present in the stable ice @xmath65 and @xmath66 polymorphs ( see fig .
[ fig : s_structure]a ) . to investigate the structure of the @xmath5 state
, we consider only oxygens atoms for which our order parameter @xmath10 is in the range @xmath67 $ ] , where the @xmath5 population peaks ( see fig . [ fig:1 ] ) .
we also exclude from the analysis particles which are identified as belonging to small crystals that spontaneously form and dissolve in a supercooled melt . to identify crystalline particles we use standard order parameters based on steinhardt rotational invariants @xcite . for each of the oxygens atoms
previously defined , we determine the optimal rotation that minimizes the following root mean square deviation @xmath68 , where @xmath69 ( @xmath70 ) are the unitary vectors joining the central oxygen atom ( @xmath71 ) with its four nearest neighbors , and @xmath72 ( @xmath73 ) are the directions of a reference tetrahedron , given by @xmath74 , @xmath75 , @xmath76 , @xmath77 and @xmath78 .
the same rotation is applied to second nearest - neighbors , defined as all oxygen atoms whose distance from first nearest neighbors is within @xmath79 times the average oxygen - oxygen distance .
we then compute the probability distribution for the position of second nearest neighbors in spherical coordinates , according to the usual transformations : @xmath80 , @xmath81 and @xmath82 , where @xmath83 is the axis connecting the central oxygen atom with the closest vertex @xmath72 of the regular tetrahedron . a schematic representation of the coordinate system is shown in fig .
[ fig : s_structure]a . the probability to find a crystalline particle with hexagonal planes pointing in the @xmath84 direction
is then given by @xmath85 .
typical results for equilibrium configurations are shown in fig .
[ fig : s_structure ] . to aid the understanding of these plots , we first report in fig .
[ fig : s_structure]d the results for the hexagonal ice crystal .
the six well defined peaks correspond to the possible positions of second nearest neighbors in the crystal , showing full orientational order : larger peaks correspond to staggered configurations , smaller peaks to eclipsed configurations . in fig .
[ fig : s_structure]b we plot the results for the @xmath5 state of liquid water at @xmath29 k and @xmath37 bar .
we first notice that the full orientational order found in the crystal is lost in the @xmath5 state .
nevertheless , one can still identify structural patterns that characterize the @xmath5 state , and that are marked by the dashed lines in fig .
[ fig : s_structure]b . the first pattern ( * d * ) corresponds to staggered arrangements of molecules , providing strong evidence that the @xmath5 state is a precursor of crystallization , i.e. that it is along the microscopic pathway that water undergoes when transforming from liquid to solid . the second prominent structural pattern found in the @xmath5 state is denoted by * e * in fig . [ fig : s_structure]b and it is not found in stable ice crystals .
this pattern is centered around eclipsed configurations , but it is characterized by fluctuations which are distinct from the one found in the hexagonal crystal . while second nearest neighbors in crystalline configurations are involved in loops of six hydrogen - bonded oxygen atoms , the pattern in * e * is instead due to five - membered rings .
this is shown in fig .
[ fig : s_structure]e , where only oxygen atoms belonging to five - membered rings are plotted , displaying the same pattern found in the @xmath5 state .
pentagonal rings , loops of five water molecules bonded to each other through hydrogen bonding , thus act as the source of frustration against crystallization @xcite . in order to crystallize ,
the @xmath5 structure needs first to break an hydrogen bond and then orient its neighbors along the crystal s directions .
the @xmath5 state is energetically stable with respect to the disordered @xmath6 state ( since each pentagon ring adds one hydrogen bond to the structure ) , but pays a high entropic cost , due to the missing degrees of freedom when closing a ring .
pentagon rings could then be responsible for the high degree of supercooling reachable with water , stabilizing the @xmath5 state and frustrating the crystallization transition .
finally , the structures denoted by * f * represent four - member rings , which are plotted in fig .
[ fig : s_structure]f , and are present in far less extent than five - member rings .
thus , the @xmath5-state is characterized by mixed structural signatures , one of which is consistent with the crystal structure and the other is not . .
( _ b _ ) fraction of pentagonal rings in the @xmath5 structures , as a function of @xmath60 for all pressures considered in the work .
the order of increasing pressure is indicated by the dashed arrow.,width=226 ] we investigate the statistics of five - membered rings in fig . [ fig : rings ] .
the first panel shows some snapshots of pentagonal rings for configurations having respectively one , two , three and four five - membered rings . for each @xmath5 structure
the maximum number of pentagonal rings is six . in fig .
[ fig : rings]b we show how the fraction of pentagonal rings ( @xmath86 ) in the @xmath5 structures changes with temperature , for different values of the pressure .
the number of five - membered rings increases with decreasing temperature and pressure .
a decrease in temperature increases the population of the @xmath5 state , so favoring crystallization , but this is partly counterbalanced by an increase five - membered rings .
to conclude , we have provided a microscopic description of water s anomalies based on locally favored states defined as structures with local translational order .
microscopically , these states reflect the underlying crystallization behavior of water . due to its strong directional bonding
, the crystallization pathway of water can be approximated in two steps : in the first step , water develops translational order , which is reflected in the population of @xmath5 states ; in the second step , it develops orientational order where the hydrogen bonds in the shell of second nearest neighbor acquire the staggered and eclipsed configurations which characterize the hexagonal and cubic forms of ice . the @xmath5-state is a locally favored state ( energetically stabilized ) with a high abundance of pentagonal rings , loops of five particles bonded to each other through hydrogen bonding .
five - membered rings increase with lowering the temperature and thus act as a source of frustration against crystallization @xcite .
the model which results from this framework is compatible with water s anomalies over a wide range of temperatures and pressures .
moreover it can be fully consistent with the liquid - liquid critical point scenario , where the @xmath87 and @xmath88 states are respectively the @xmath5-dominant and @xmath6-dominant states below the critical point .
in fact , the order parameter @xmath10 includes the local configurations responsible for water s two - lengthscales average interaction @xcite .
we note that , within the two - state model , the presence of a critical point associated with demixing is not a necessary condition for the existence of thermodynamic anomalies , and the anomalies survive even if the mixture lacks a critical point ( @xmath89 in eq .
( [ eqn : two_state ] ) ) @xcite .
such a possibility has recently been suggested for tip4p/2005 water @xcite .
we stress that the parameters of the model were obtained only from microscopic information , and then its predictions were compared to the anomalies of water
. it would be possible also to use the model in a phenomenological way , by fitting the anomalies to improve the model parameters , for example improving the estimates in the deeply supercooled region ( obtaining better estimates for the density minimum and the isothermal compressibility maximum ) . in this work
we have avoided such an approach to show that a two - state description of the phase behavior of water is possible from microscopic information .
once this is confirmed , precision fitting of the anomalies can provide additional insights @xcite .
for this reason we have also provided an approximate scheme to extract the model s parameters from experimentally accessible measurements ( see appendix ) .
knowledge of the oxygen - oxygen radial distribution function is in fact accessible with both neutron and x - ray scattering methods ( see , e.g. , ref .
@xcite ) . performing these measurements over a wide range of temperature and pressures could provide important information , such as the nature of the non - idealities of the mixture ( for example whether they have an energetic or entropic origin @xcite ) . here
it may be worth noting the roles of locally favored structures in crystallization .
we show that locally favored structures in water not only have translational order in the second shell , but also contain five - membered rings of hydrogen - bonded molecules , indicating a mixed character of their roles in crystallization : the former helps crystallization , whereas the latter causes frustration against crystallization . this frustration effect may be related with the rather large degree of supercooling of water before homogeneous crystal nucleation of hexagonal or cubic ices takes place .
finally , the validity of two - state models can be assessed not only in experimental studies of pure water , but also in water mixtures , where a liquid - liquid transition of the water component has been observed @xcite , or in water - salt solutions , where the effect of salt on the structure of water is similar to the effect of pressure @xcite .
we can further speculate that the two - state model based on the translational order of the second shell may be relevant also to other tetrahedral liquids ( si , ge , silica , germania ) @xcite , which play a crucial role in materials science .
* molecular dynamics simulations .
* molecular dynamics simulations were run using the gromacs ( v.4.5 ) molecular dynamics simulation package .
the isothermal - isobaric @xmath90 ensemble was sampled through a nos - hoover thermostat and an isotropic parrinello - rahman barostat .
lennard - jones interactions have a cutoff at @xmath91 nm , and cutoff corrections are applied to both energy and pressure .
electrostatic interactions are calculated through ewald summations , with the real part being truncated at @xmath91 nm , and the reciprocal part evaluated using the particle mesh method .
the systems consist of @xmath92 molecules of water , and the total simulation time varied from @xmath62 ns for the high temperature simulations , up to @xmath93 @xmath94s for the lowest temperatures .
the water force field is tip4p/2005 @xcite .
hydrogen bonds are located through geometric constraints on the relative positions of the donor ( d ) , acceptor ( a ) and hydrogen ( h ) atom .
two oxygen atoms are considered hydrogen bonded if their distance is within @xmath95 nm , and the angle hda is less than @xmath96 @xcite .
* analysis of the distribution function @xmath40 . *
the distribution function @xmath40 is decomposed in two gaussian populations . in order to distinguish unambiguously two gaussian populations with large overlap
, we take into account the fact that one of the two populations ( corresponding to the @xmath5 state ) should be characterized by fully formed translational order up to second shell , so its distribution should be vanishingly small for @xmath97 .
this means that the @xmath5 state is characterized by good translational order on both first and second shells . imposing this constraint on the fitting we are able to decompose the distribution in two populations for a large region of the phase diagram .
the fitting function satisfying the above constraint is given by @xmath98 where @xmath99 are the fitting parameters .
both @xmath100 and @xmath101 are monotonically decreasing with temperature , showing that states become more and more structured at low temperatures , while @xmath102 and @xmath103 are increasing with temperature , as expected by the increase of thermal fluctuations .
the fitting performs well for state points characterized by @xmath104 ( so throughout the experimentally accessible region ) , but at very low temperatures and pressures the fraction of the @xmath6 state becomes small , increasing the uncertainty of the fit .
one possible way to overcome these difficulties is by noting that the @xmath6 fluid should behave like a fluid without anomalies , and thus its parameters ( @xmath100 and @xmath102 ) should be well behaved functions of @xmath36 and @xmath60 .
we then choose to estimate the value of @xmath100 with a quadratic extrapolation from the values obtained at state points with @xmath104 .
this procedure degrades the accuracy of the model only at the lowest temperatures and for small pressures ( as can be seen in the estimates for the isothermal compressibility maximum and the minimum in density in fig .
[ fig:2]a ) , but the agreement with simulations is still satisfactory . * fittings of the @xmath60-@xmath36 dependence of the order parameter , density , and isothermal compressibility . *
the two - state model predictions are based on the free energy expression given by eq .
( [ eqn : two_state ] ) . to fully specify the model , an expression for the free energy difference between the two bulk states , @xmath105 , has to be provided . in our model
this difference is expressed as a second order expansion around the known location of the liquid - liquid critical point : @xmath106 where @xmath107 and @xmath108 .
we employ @xmath46 k and @xmath47 bar , which are reported for the same system @xcite .
this also fixes the value of @xmath109 .
the coefficients of the expansion @xmath110 are given in the caption of fig .
[ fig:1 ] .
the resulting free energy allows the calculation of the anomalous term of any thermodynamic property @xmath111 , i.e. the difference between the value of @xmath111 in the mixture and its value in a pure state ( the background ) , @xmath112 or @xmath113 .
we define the anomalous term , @xmath114 , with the following expression : @xmath115 . for example
, the expressions for the density and compressibility anomalies are found respectively with first and second order derivatives of the free energy with respect to pressure @xmath116 the calculation of the absolute value of the thermodynamic properties requires a reasonable assumption on the @xmath60-@xmath36 dependence of the background part , which is usually written as a low - order polynomial .
for example @xmath117 .
we stress that the success of a two - state model depends on the ability to determine the location and intensity of the anomalies , which do not depend on the knowledge of the background terms .
the microscopic approach described in the main text requires the knowledge of the instantaneous positions of both oxygen and hydrogen atoms in the system , thus being suitable only for computational studies .
here we show that , with minor approximations , it is possible to obtain the fraction of the @xmath5 state not only from the distribution of @xmath10 but also from experimentally measurable quantities , which can probe the degree of translational order .
the definition of @xmath10 involves the difference between the distance of the first non - hydrogen bonded oxygen and the last hydrogen bonded oxygen .
the hydrogen bond interaction is a limited range interaction , and breaks when the oxygens involved in the bond are too far apart .
we call this cutoff distance @xmath118 . in our simulation study
we have set this distance at @xmath119 nm @xcite .
this means that if we look at two oxygen atoms whose distance is about @xmath118 , these atoms ( with high probability ) do not share an hydrogen bond . at the same time
we know that in the @xmath5 state the first non - hydrogen bonded atom has a vanishingly small probability to be at distances close to the first neighbors shell ( @xmath16 ) .
it follows that the contribution to the radial distribution function at distance @xmath118 comes predominantly from the @xmath6 state . to clarify this point
, we consider the radial distribution functions of oxygen atoms , @xmath120 , plotted in fig .
continuous lines show the radial distribution function for different state points along the widom line ( where by definition @xmath54 ) : the distributions are all very close to each other , with an isosbestic point approximately at @xmath119 nm .
the reason for the isosbestic point is that , as noted above , the value of the radial distribution function at @xmath118 is proportional to the fraction of the @xmath6 state , and this is constant at the widom line .
coincidentally , the value of the radial distribution function at @xmath118 is approximately @xmath121 , suggesting a direct proportionality between the fraction of the @xmath6 state and @xmath122 .
this is confirmed by plotting the radial distribution functions of state points with a low fraction of the @xmath6 state ( @xmath123 bar , @xmath29 k , dotted - dashed line ) and a high fraction of @xmath6 state ( @xmath124 bar , @xmath30 k , dashed line ) , where the value of @xmath122 is found to be close to @xmath125 and @xmath93 respectively .
we thus write the following relation between the height of @xmath120 and @xmath19 ( fraction of the @xmath5 state ) : @xmath126 .
figure s1b compares the results of this relation with the results obtained by fitting the distribution function of the order parameter @xmath10 . in the inset
these two values are plotted for all state points considered in this work , showing that indeed the relation @xmath127 holds to a good approximation .
the main panel in fig .
s1b shows the temperature dependence of these two values , showing again the good agreement between the two calculation methods .
note that the relation @xmath127 seems to hold better for @xmath128 , which is the same range that the experiments are able to access . from the radial distribution function of oxygen atoms , @xmath120 .
( _ a _ ) radial distribution function @xmath120 for state points along the widom line ( continuous lines ) .
the functions were obtained by linear interpolation of radial distribution functions of the closest simulated state points . also plotted are the radial distribution functions of two state points with respectively a high fraction of the @xmath5 state ( @xmath123 bar , @xmath29 k , dotted - dashed line ) and a high fraction of @xmath6 state ( @xmath124 bar , @xmath30 k , dashed line ) the dotted lines denote the isosbestic point close to @xmath119 .
( _ b _ ) comparison between the values of @xmath19 : symbols represent values obtained from the relation @xmath126 , while lines are the values obtained from the distribution function of the order parameter @xmath10 ( the lines in this figure represent the same data as the symbols in fig .
the inset shows the comparison between the value of @xmath19 obtained from the distribution function of the order parameter @xmath10 and the value of @xmath122 at the same state points .
, width=264 ]
this work was partially supported by a grant - in - aid for scientific research ( s ) from jsps , aihara project , the first program from jsps , initiated by cstp , and a jsps postdoctoral fellowship .
|
water is a complex structured liquid of hydrogen - bonded molecules that displays a surprising array of unusual properties , also known as water anomalies , the most famous being the density maximum at about @xmath0c .
the origin of these anomalies is still a matter of debate , and so far a quantitative description of water s phase behavior starting from the molecular arrangements is still missing .
here we provide its simple physical description from microscopic data obtained through computer simulations .
we introduce a novel structural order parameter , which quantifies the degree of translational order of the second shell , and show that this parameter alone , which measures the amount of locally favored structures , accurately characterizes the state of water .
a two - state modeling of these microscopic structures is used to describe the behavior of liquid water over a wide region of the phase diagram , correctly identifying the density and compressibility anomalies , and being compatible with the existence of a second critical point in the deeply supercooled region .
furthermore , we reveal that locally favored structures in water not only have translational order in the second shell , but also contain five - membered rings of hydrogen - bonded molecules .
this suggests their mixed character : the former helps crystallization , whereas the latter causes frustration against crystallization .
|
although the magnification tensor is symmetric , the relative magnification tensor is not and thus will have four independent components .
it is customary to refer to the relative magnification tensor in terms of its eigenvalues @xmath69 and corresponding eigenvector position angles @xmath70 ( _ e.g. _ fgs ) . with unresolved images ,
it is not possible to measure the full relative magnification tensor .
however , the flux ratio of the two images gives the magnitude of the determinant of @xmath71 $ ] .
observational constraints [ obscons ] image positions and magnifications [ imdat ] the highest - resolution images of 0957 + 561 have been obtained by means of radio vlbi . shortly after the discovery of the source , vlbi observations ( @xcite ) revealed similar core - jet structure in the radio components a and b , reinforcing the lensing hypothesis for the system .
et al . _ ( 1988a ) obtained improved @xmath2213 cm vlbi maps and resolved the a and b jet images into three gaussian components each .
they were thereby able to construct the magnification tensor @xmath72 $ ] relating the a and b images .
implicit in their calculation of @xmath72 $ ] was the assumption that the tensor remains the same both in the qso core and in the jet region .
@xcite have recently obtained even more accurate vlbi maps of 0957 + 561 at @xmath2218 cm and have been able to measure variations of @xmath73 $ ] along the jet .
@xcite fit the a and b vlbi images of 0957 + 561 with six gaussian components each , denoted as @xmath74 and @xmath75 . @xmath76 and
@xmath77 correspond to the respective core components while @xmath78 and @xmath79 are successive blobs in the jet .
the relative positions of the brightest jet components , @xmath80 and @xmath81 , with respect to the core components , @xmath76 and @xmath77 , are measured to within 0.1 mas ( table [ impostab ] ) . in our lens modeling , we use these two image offsets as constraints . because @xcite do not give the separation between images a and b , we take as @xmath82 the value @xmath83 reported by gorenstein _
et al . _ ( 1988a ) . the improved resolution of the @xmath2218 cm vlbi map allowed g94 to measure the change in the relative magnification tensor along the axis of the jet .
this gradient is effectively measured between components 1 and 5 in the two images .
because of their limited signal to noise , @xcite were forced to set two of the magnification tensor gradient components to zero , and they evaluated the gradients only of the other two components .
thus , @xcite provide a total of six constraints on the magnification tensor , _ viz . _ four matrix elements corresponding to the transformation from @xmath80 to @xmath81 , plus gradients of the two matrix eigenvalues along the long axis of the jet .
the six constraints are summarized in table [ magtab ] .
third image flux [ flux3 ] although gravitational lensing theory predicts a third image of 0957 + 561 near the center of the lensing galaxy , no such image has been seen down to a @xmath84 limit of 0.6mjy , which is @xmath85 of the flux density of image b ( @xcite ) .
this flux limit provides an additional constraint on the lens model . because there is some ambiguity in the non - detection of the third image flux ( @xcite ; @xcite ; @xcite ) , we have chosen not to treat this constraint in the standard fashion .
we instead adopt a weighting function that assigns no penalty to a lens model if it predicts a third image flux below the @xmath84 detection limit , but where the penalty increases steeply if the predicted flux is greater than this limit .
thus , if @xmath86 is the flux ratio between the third image and the b image according to a lens model , we take its contribution to the @xmath6 to be @xmath87 this unorthodox penalty assignment is a compromise between what we view as two unrealistic extremes . on the one hand , treating the @xmath85 flux ratio upper limit as a true @xmath84 contribution to the overall @xmath6 excessively penalizes models for which the third image flux is say only at the 2 or @xmath88 level and thus would likely have been dismissed as noise in the large area scanned for the third image . on the other hand , treating the @xmath85 flux ratio as a @xmath89 penalty allows the model too much freedom to produce unrealistically bright third images .
although the particular @xmath6 assignment given above results in non - gaussian uncertainties in the model parameter values , we feel that this penalty function is the fairest representation of the unclear observational status of the third image .
we emphasize this point because different choices of the third image penalty lead to significantly different model limits on the g1 core radius .
however , there is very little effect on our results for the hubble constant .
position of the g1 center of mass [ obsg1 cm ] although the angular separation of the a and b images in 0957 + 561 is known extremely accurately , the positions of these images with respect to the center of mass of g1 are not so well constrained ( see table [ centers ] ) .
stockton ( 1980 ) reported an optical center of brightness for g1 with a 30 milliarcsecond ( mas ) uncertainty .
vla observations of the region ( @xcite ) revealed a point - like source g with 1 mas uncertainty in the position .
however , vlbi mapping of the same region ( @xcite ) found no source coincident with the vla detection , but did detect a weak ( 0.6 mjy ) point source @xmath90 some 30 mas away , again with 1 mas uncertainty . both the vla and vlbi sources are consistent with the optical center , but they are inconsistent with each other by many standard deviations . in view of the discrepancy between the vla and vlbi detections ,
we take the optical center of brightness of g1 and its error bars as the reference for the model fitting .
note that this only affects two coordinates , namely the two components of ( @xmath91g1 ) , since all other image positions are measured as offsets with respect to either @xmath77 or @xmath76 .
our treatment of the galaxy center differs from that adopted by @xcite , who selected the vlbi point source @xmath90 as their center of mass position with 1 mas uncertainty .
lens mass models[mmod ] most of our calculations are based on a five - parameter model of the 0957 + 561 lens system where we represent the cluster as a simple astigmatic deflector and the galaxy as a power - law deflector with a core radius .
we also test the particular five - parameter model used by fgs , who have the same form for the cluster contribution , but represent the galaxy as a king potential with an additional compact nucleus modeled as a point mass .
we shall describe the lensing properties of these two models in detail below .
the softened power - law sphere[splsmod ] our primary mass model for the lensing galaxy is an extension of the power - law mass distribution , @xmath92 , used previously in the modeling of lens systems ( @xcite ) .
we describe the galaxy by the following spherically symmetric volume density profile , @xmath93 where @xmath94 is the core radius . for @xmath95 ,
the mass of the galaxy varies as @xmath96 which corresponds to @xmath97 .
the parameter @xmath2 is restricted to lie between @xmath98 , which corresponds to a modified hubble profile ( see below ) , and @xmath99 , which corresponds to a constant surface mass density sheet .
this family of softened power - law sphere ( spls ) models includes the singular isothermal sphere , which corresponds to @xmath100 , @xmath101 .
although the spls density profile does not yield an analytic potential @xmath102 or included mass @xmath103 , the deflection angle does have an analytic form . to show this
, we first obtain the surface mass density profile implied by equation [ [ density ] ] : @xmath104 where @xmath105 is the standard euler beta function .
this surface density is a generalization of the modified hubble profile ( cf .
@xcite ) , which is the specific case @xmath98 .
we then find that equation [ [ sigeq ] ] yields an analytic projected mass @xmath61 of the form @xmath106 , \qquad m_0 \equiv \left(\frac{2\pi}{\eta}\right ) \sigma_0\,r_c^2.\ ] ] combining the deflection formula ( eq . [ [ sphdef ] ] ) for a spherically symmetric deflector with equation [ [ mass ] ] ,
we then obtain the deflection function @xmath50 : @xmath107 \gb{\theta},\qquad \alpha_0 \equiv \left({4 g m_0 \over c^2 d}\right ) ^{1/2}.\ ] ] for our model - fitting we adopt an equivalent but more convenient expression for the deflection function ( eq . [ [ alph1 ] ] ) which remains well - behaved even when @xmath108 and the central density @xmath0 diverges : @xmath109\gb{\theta } \mbox{,\quad } \alpha_{\rm e}\equiv \alpha_0^{2/(2-\eta)}\,\theta_c^{-\eta/(2-\eta)}.\ ] ] the parameter @xmath110 represents an approximate einstein radius of the galaxy , insofar as @xmath111 for @xmath112 , and @xmath113 for @xmath114 .
note that the basic spls galaxy model described here has three adjustable parameters : @xmath115 , @xmath1 , @xmath2 . for the particular calculations described in [ splsbh ]
we add a point mass at the center of the galaxy to investigate if the additional freedom would allow us to obtain a better fit of the observations . in these cases , the mass model has four parameters . the fgs galaxy deflector [ fgsmod ] in this model the lensing galaxy consists of a smooth , circularly symmetric , king - type surface density profile parameterized by its angular core radius @xmath1 and velocity dispersion @xmath116 .
the deflection function employed by fgs for this profile is an analytic approximation introduced by young _
et al . _ ( 1981 ) : @xmath117 } & = & \left({\sigma_v^2\over c^2}\right )
\left({\gb{\theta}\over\theta}\right ) \alpha_*(\theta ) , \\ \alpha_*(\theta ) & = & 53.2468\,f\left({1.155 { \theta\over\theta_c}}\right ) - 44.0415\,f\left({0.579 { \theta\over\theta_c}}\right),\nonumber\\ f(x ) & = & { { \left({1 + x^2}\right)^{1/2 } - 1 } \over x}.\nonumber\end{aligned}\ ] ] this formula reproduces the deflection of a king profile with cutoff radius @xmath118 and is accurate to @xmath119 within @xmath120 of the center . because 0957 + 561a lies @xmath121 from the center of the galaxy g1 ( we will hereafter refer to the point at the center as g1 ) , model - fitting with this approximation is reliable only for @xmath122 .
fgs found that a pure king - type galaxy has too `` soft '' a potential to reproduce the observed magnification ratio between the a and b quasar images satisfactorily .
furthermore , the potential does not sufficiently demagnify the third image near the galaxy center . for both reasons
fgs added to their mass model a compact nucleus at g1 .
this component is modeled as a point - mass in the lensing equations , with a deflection law given by @xmath123 where @xmath124 is the einstein radius of the compact nucleus .
the einstein radius is related to the mass @xmath125 of the compact nucleus via @xmath126 although treated as a point mass , the compact nucleus postulated for the 0957 + 561 lens need only have a radius small in comparison with the @xmath127 separation between image b and g1 . as in the case of the spls model described earlier
, the fgs model again has three adjustable parameters : @xmath116 , @xmath1 , @xmath128 .
external sources of deflection[sheet ] the lensing galaxy in 0957 + 561 is a member of a cluster of galaxies , and because of the large angular separation of the a and b images it would appear that part of the deflection in this lens is produced by the cluster .
following fgs , we model the cluster deflection by means of a convergence @xmath4 and shear @xmath5 with position angle @xmath129 .
such a model is reasonable so long as the projected mass density of the cluster is relatively constant over the angular scale of the image separation .
we consider the effect of deviations from this model in [ sisgalsec ] .
the convergence @xmath4 is the ratio of the local mass surface density of the cluster to the _ critical density _
, @xmath130 ( @xcite ) .
the angular deflection due to a constant convergence takes the simple form @xmath131 . since this corresponds to isotropic focusing ,
there is a degeneracy in lens models which was noted by falco _
et al . _ ( 1985 ) .
given any model of the lensing galaxy which fits the observations , it is possible to find a new model where a constant convergence @xmath4 of arbitrary magnitude is included and the mass in the galaxy is at the same time scaled down by a factor @xmath132 . for the particular galaxy models described earlier , the parameters which are modified are : @xmath133 , @xmath134 , @xmath135 .
all observables are invariant under this transformation , except the time delay which scales as @xmath132 .
the effect on the time delay translates to a change in the derived hubble constant : @xmath136 .
since the surface mass density of the cluster has to be a positive quantity , we have the constraint @xmath137 . because of the existence of the simple tranformation described above , it is not necessary to include @xmath4 explicitly in the model .
the shear , however , must be included .
it transforms with variable @xmath4 as follows : @xmath138 .
therefore , in effect , it is the _ scaled _ shear @xmath139 which we fit in our models .
the deflection law due to shear takes the form @xmath140\ , \gb{\theta}\hbox{,\quad where\quad } \left[{{\bf t}(\phi)}\right ] \equiv \left(\begin{array}{cc}\cos2\phi & \sin2\phi \\
\sin2\phi & -\cos2\phi \end{array}\right).\ ] ] purely from measurements of the lensed images it is not possible to break the model degeneracy due to a variable @xmath4 .
however , direct measurements of the mass either in the galaxy or the cluster do offer the possibility of breaking the degeneracy .
we discuss this in [ getkappa ] .
deriving @xmath12 from a lens model [ hoderiv ] to determine a value for @xmath12 from 0957 + 561 , it is necessary to compute the difference in light propagation time from the source to the observer along the two image paths .
this quantity depends on the image positions and relevant redshifts as well as the mass distribution of the lens . in the following discussion , we assume @xmath141 for simplicity , and include the relevant factor of @xmath132 only at the end . to obtain the relative time delay we first determine the delay of the path corresponding to a given image location @xmath51 relative to the `` unlensed '' light path to the observer from the source at location @xmath43 .
this takes the form @xmath142 where the first term is the geometric time delay and the second is the gravitational time delay due to the `` shapiro effect '' ( shapiro 1964 ) .
the factor @xmath143 accounts for the expansion of the universe since the deflection of the rays at redshift @xmath144 . the relative time delay between the two images
is given by @xmath145 .
given an spls mass model and a fitted source position @xmath43 , we then have the following relation for @xmath146 : @xmath147 ^{\gb{\theta}_i}_{\gb{\theta}_j } \nonumber \nobreak\\ & & - \,\alpha_{\rm e}^{2-\eta}\int_{\gb{\theta}_j}^{\gb{\theta}_i } { { { \left({\theta^2 + \theta_c^2}\right)^{\eta/2 } - \theta_c^\eta}\over{\theta}}\,d\theta } - \alpha_{\rm bh}^2 \ln{{|\gb{\theta}_i| } \over { |\gb{\theta}_j|}},\end{aligned}\ ] ] where @xmath148 and @xmath149 are the @xmath23- and @xmath24-components of angular position vector @xmath44 in our coordinate system .
we have cast the time delay equation in the above form to show the separate dependencies on model parameters ( on the right ) and on the effective lens distance @xmath38 ( on the left ) .
all other lensing variables appearing in this equation ( @xmath144 , @xmath51 , @xmath150 ) are observables .
as the king - type galaxy approximation used by fgs has an analytic potential , the time delay equation for the fgs model has the following closed form : @xmath151 ^{\gb{\theta}_i}_{\gb{\theta}_j } \nonumber \nobreak\\ & & { } - \theta_c \left({\sigma_v\over c}\right)^2 \left [ { 27.6636\,g\!\left({1.155 { |\theta|\over\theta_c}}\right ) - 45.6437\,g\!\left({0.579 { |\theta|\over\theta_c}}\right ) } \right ] ^{\gb{\theta}_i}_{\gb{\theta}_j } \nonumber \nobreak\\ & & { } - \alpha_{\rm bh}^2 \ln{{|\gb{\theta}_i| } \over { |\gb{\theta}_j|}},\\ g(x ) & = & \sqrt{1 + x^2 } - \ln\left({1 + \sqrt{1 + x^2}\!}\right).\nonumber\end{aligned}\ ] ] the dependence of the predicted @xmath146 on @xmath12 is embedded in @xmath38 , as can be seen from equation [ [ angdi ] ] .
thus , once the model parameters are determined with sufficient precision via model - fitting , we obtain an estimate of the quantity @xmath152 .
a measurement of @xmath146 then leads to an estimate of the hubble constant . actually , because of the @xmath4 degeneracy ( [ sheet ] )
, the formula for @xmath146 given above should be multiplied by the undetermined factor @xmath132 .
therefore , technically we can only determine the quantity @xmath153 from the model .
it is convenient to define a dimensionless number @xmath154 , which allows for all of these factors : @xmath155 each lens model gives a unique value of @xmath154 . the value of @xmath12 derived from this , however , depends on the measured time delay @xmath146 and on the unknown magnitude of @xmath4 .
since we know that @xmath137 , we see that the lens method can provide an upper bound for the hubble constant even when there is no independent determination of @xmath4 .
although the discussion here has focused on 0957 + 561 , the same ideas are applicable to any lensing system in which ( i ) we know the source and deflector redshifts ( in order to express @xmath38 in terms of @xmath12 ) , ( ii ) we are able to measure a time delay , and ( iii ) we obtain a well - constrained mass model for the lens . if we have a lens system with @xmath156 quasar images , we can in principle measure @xmath157 independent time delays , all of which should be consistent with a single value for the hubble constant .
results [ results ] the various observations detailed in [ obscons ] furnish a total of fifteen constraints on the mass model .
both the spls model and the fgs model fit these constraints with nine parameters four source emission coordinates , two each for blob 1 and blob 5 , and five variables specifying the lens mass distribution .
we also study a variant of the spls model that includes a compact nucleus , for a total of ten parameters . the third image flux constraint discussed in [ flux3 ] only comes into play for models without a g1 compact nucleus .
the predicted flux of the third image is zero when there is any substantial mass in the nucleus .
thus the third image flux upper limit does not influence the fgs model or the spls model with compact nucleus .
accordingly , we have @xmath158 degrees of freedom for the basic spls model , @xmath159 d.f . for the fgs model , and @xmath160 d.f . for the spls with compact nucleus
we discuss the results from the spls model in [ nobh ] and those from the fgs model in [ fgstest ] . in
[ splsbh ] , we discuss the results of adding a compact nucleus to the spls model .
we also investigate the degree to which our spls model results vary if we add perturbations due to the ellipticity of g1 ( [ ellsec ] ) and the influence of nearby cluster galaxies ( [ sisgalsec ] ) .
fitting to a softened power - law sphere [ nobh ] goodness of fit [ errsubsec ] we employed the amoeba non - linear minimization algorithm ( press _ et al . _
1992a ) , based on the downhill simplex method , to optimize the lens model by minimizing the @xmath6 of the fit .
our best - fit model gives @xmath161 for a chi - square per degree of freedom of @xmath162 . in table
[ chitable ] we compare the model predictions of the various observables with the measured data . in general , it is seen that the model agrees well with the data ; most deviations are under 1@xmath15 and the largest discrepancies are under @xmath163 .
however , despite this good agreement , the formal @xmath164 value is quite large .
this is primarily the result of the very large correlations which @xcite quote between the errors on the various observables ( table [ impostab ] ) .
for instance , if we ignore the correlations and define @xmath6 to be the straight sum of the squares of the values in the last column of table [ chitable ] , then our fitted parameter set gives @xmath165 , corresponding to @xmath166 . throughout this paper
we always include the correlations when calculating @xmath6 .
@xcite noted that the value of @xmath167 predicted by the model of @xcite is discrepant with the value which they measured .
our best - fit spls model has similar difficulty the predicted @xmath168 is almost two standard errors below @xcite s value . an interesting feature of table [ chitable ] is that almost all of our model @xmath6 from fitting the image positions arises from the @xmath169 constraint . as noted in
[ obsg1 cm ] , we have adopted the larger ( 30 mas ) error bars of the observed optical center of brightness ( @xcite ) .
we find that the model takes advantage of this freedom in the center of mass position to better satisfy the other , much tighter constraints on qso image separations from vlbi .
our best model ( without compact nucleus ) chooses a g1 center of mass some 62 mas from the measured optical center g1 , 44 mas from the vlbi source @xmath170 , and 64 mas from the vla source g. clearly this separation would have been prohibitive if we had used the 1 mas error bars of the radio sources as the uncertainty in g1 center of mass position .
we experiment with forcing the galaxy center of mass to be coincident with @xmath170 to the 1 mas precision of vlbi .
this was the procedure of @xcite in their analysis . in this case
the spls model fits very poorly to the data , never obtaining a @xmath6 below 442 ( @xmath171 ) .
the contribution to the overall @xmath6 from the magnification tensor constraints only marginally increases ( to 22.7 , up from 21.5 ) , with essentially the entire extra @xmath6 coming from badly fit image positions .
our model has the most difficulty in simultaneously fitting both @xmath172 and @xmath173 , which are each almost 13 standard errors from the vlbi measured values and account for two - thirds of the image position @xmath6 . the fgs model ( see [ fgstest ] ) fares better , but still has an unacceptably large @xmath164 value of 43 . adding a compact nucleus to the spls ( see [ splsbh ] )
achieves similar results , @xmath174 . in both cases
the @xmath6 contribution from magnification constraints is 60% , as compared with @xmath175 for the basic spls .
fitted spls model parameters[splsnums ] in table [ bestnobh ] we show the best - fit spls model parameters and their associated 95% @xmath176 confidence limits . because of our poor @xmath164 , we have estimated the @xmath163 bounds such that they correspond to @xmath177 rather than @xmath178 . of particular interest are the limits on @xmath1 and @xmath2 .
the best - fit core radius is zero , and models with core radii in excess of @xmath179 are excluded at the @xmath163 level . at the lens redshift of 0.36
, this corresponds to a @xmath163 upper limit of @xmath180 pc on the linear core radius of the lensing galaxy .
the limit on @xmath1 arises primarily from the limit on the flux of the third image ( _ cf . _
@xcite ) . the predicted third image flux increases rapidly with core radius , exceeding the @xmath84 detection limit of @xmath181 for @xmath182 mas , or @xmath183 pc .
the best - fit value of the radial power law index @xmath2 is @xmath184 , which makes the density profile of the lens slightly shallower than isothermal @xmath185 .
if @xmath2 is pushed lower , the model compensates by raising the core radius .
this is only effective down to @xmath186 , at which point the core radius becomes large enough that the third image flux exceeds the @xmath84 detection limit . because of this , an isothermal profile is ruled out quite strongly .
our best `` isothermal '' model has @xmath187 , which is unacceptably large . in the other direction , we find that the fit degrades rapidly for @xmath188 , with @xmath189 excluded at the @xmath190 level .
we illustrate the bounds on @xmath1 and @xmath2 in figure [ rcbfig ] , where we display contours of constant @xmath6 corresponding to best - fit models with fixed values of these two parameters .
the most striking feature of the figure is the fact that the range of models which fall within the @xmath163 contour is limited to a narrow valley covering quite a small range of the two parameters .
this is despite the fact that we have conservatively defined the @xmath163 limit as @xmath191 . in other words ,
the 0957 + 561 observations do an excellent job of constraining the parameters of our mass model despite the large @xmath6 for the best - fit .
even fairly small deviations in the model parameters about their optimal values give a far worse fit to the data .
the reason is that many of the data constraints have extremely small quoted errors so that the model has to be just right even to fit within @xmath163 .
the large correlations between the magnification matrix elements quoted by g94 ( table [ impostab ] ) make the problem more acute .
note from figure [ rcbfig ] that the best - fit model has @xmath112 and therefore lies at one edge of the allowed parameter space . because the surfaces of constant @xmath6 are truncated at this edge
, we do not have gaussian - distributed uncertainties for our parameter estimates .
generally , most of the model parameters are tightly constrained to one side of their best - fit value and much less constrained in the opposite direction , corresponding to the narrow @xmath6 `` valley '' seen in figure [ rcbfig ] . despite the non - gaussian errors , we have chosen to use @xmath192 as our definition of the `` 95% '' confidence limits for the parameters .
these are the values listed in table [ bestnobh ] .
implications for @xmath12[h15lims ] given a set of fitted model parameters , we can estimate the dimensionless hubble parameter @xmath154 as described in [ hoderiv ] .
figure [ hocontour ] shows contours of @xmath154 overlaid on contours of @xmath6 .
we see that the @xmath154 contours are roughly parallel to the long axis of the valley of good solutions .
because of this , only a narrow range of values of @xmath154 is allowed .
we quantify the limits on @xmath154 as follows . given a parameter set @xmath193 with corresponding goodness - of - fit @xmath194 and derived hubble parameter @xmath195 , we solve for the function @xmath196 via the method of lagrange multipliers .
we obtain points on the @xmath196 curve by finding the parameter set @xmath193 which minimizes the function @xmath197 for various values of the lagrange multiplier @xmath22 .
each choice of @xmath22 generates a pair of values for @xmath6 and @xmath154 which represent the least @xmath6 for the subset of parameter combinations which yield the given @xmath154 .
the results of this procedure are shown in figure [ hochiplot ] .
the best - fit model gives @xmath198 , with the @xmath163 interval given by @xmath199 . substituting this value into equation [ [ h1.5eq ] ]
, we then obtain @xmath200 testing the fgs model[fgstest ] we have also tested the family of models used by @xcite and described in [ fgsmod ] .
we obtain a best - fit @xmath6 of 28.4 , corresponding to @xmath201 .
this is significantly worse than the @xmath202 fit we obtained with the spls model . in table
[ fgsparm ] we list the best - fit parameter values we obtain with the fgs model , and compare these with the values previously obtained by @xcite using their more limited data . for almost all parameters ,
our new estimates deviate significantly from the old ones .
this is somewhat worrisome since it suggests that this lens model may not be very robust .
incidentally , if we use the original model parameters as given by @xcite with our data constraints , we obtain an extremely poor @xmath6 of @xmath203 .
our estimate for @xmath154 from the fgs models is 0.732 , with 95% confidence limits given by @xmath204 .
the @xmath163 lower limit obtained here overlaps the @xmath163 upper limit of the spls model and therefore the two models may be considered to be marginally consistent .
however , between the two , the models span quite a wide range of @xmath154 .
this is disturbing since it suggests that the data are still unable to constrain the lens model very well .
@xcite made a similar point with a smaller data set .
another disturbing feature is that the value of @xmath154 suggested by our current best - fit fgs model differs considerably from the value @xmath205 which fgs obtained with their orginal fits .
this is a reflection of the large changes in the parameters of the model as shown in table [ fgsparm ] .
once again it implies that the fgs model is not well - constrained .
one odd feature of the fgs model which must be mentioned is the extremely massive compact nucleus required with this model . at a mass of @xmath206 billion @xmath207 @xmath208 ,
this compact nucleus of the model is unlikely to represent a supermassive black hole at the center of the galaxy .
it is unclear what this mass represents .
results for spls models with compact nucleus [ splsbh ] because @xcite found that including a compact nucleus at the center of g1 greatly improved their fit to the data , we have tested modified spls models where we add a compact central nucleus of mass @xmath125 . for the expanded spls models with compact nucleus we find that the best fit model has @xmath209 and @xmath210 , for @xmath211 .
this reduced @xmath6 is similar to the @xmath212 we obtain with the fgs model - fitting .
we see that both models with a g1 compact nucleus do a significantly worse job than the basic spls . in table
[ bestwbh ] we list the parameter values of the best - fit spls with a compact nucleus .
these may be compared with the parameters of the basic spls model ( @xmath213 ) listed in table [ bestnobh ] .
we see that @xmath2 has increased from 1.16 to 1.38 , so that the deviation from isothermality is larger .
most of the other parameters are almost unchanged .
the derived hubble parameter is significantly smaller , however , at @xmath214 . as we increase the mass of the compact nucleus beyond the biest - fit @xmath215 while optimizing the remaining parameters , we find that the @xmath6 increases very slowly , not reaching @xmath216 until @xmath217 . over this range ,
the core radius steadily rises and the power - law exponent drops toward zero . beyond @xmath218
the @xmath6 becomes worse very quickly . at @xmath219 ,
the model can do no better than @xmath220 .
the transition happens at the point where the einstein radius @xmath124 of the nucleus becomes comparable to the angular separation between g1 and b. models with more massive nuclei have great difficulty fitting the b image and therefore give large @xmath6 . near the limiting mass ,
the core radius of the galaxy becomes large , comparable to the g1-b separation , and the index @xmath2 approaches zero .
table [ bestwbh ] gives parameter values corresponding to a few specially selected spls models with compact nucleus .
in addition to the best - fit model which we have already discussed , we show the best - fit isothermal ( @xmath8 ) model , where @xmath221 , and an fgs - like spls where we fix @xmath125 equal to the optimum value obtained with the fgs model . generally , we find that up to @xmath222 , the core radius is zero and @xmath2 increases . above this mass
, @xmath2 starts decreasing and the core radius goes up .
the variations are particularly rapid as @xmath125 approaches the limiting value .
the derived values of the hubble parameter shown in table [ bestwbh ] reveal a similar behavior . until @xmath223 , @xmath154 decreases , going down to about 0.5 . for more massive nuclei , @xmath154 turns around , increasing almost to 0.75 near the fgs mass .
an interesting result is that the value of @xmath154 we obtain with the modified spls model is very similar to that with the fgs model at the same mass of the nucleus .
therefore , the question of whether or not to take seriously the value of @xmath224 obtained with the fgs model boils down to whether or not a nucleus with a mass of @xmath206 billion @xmath207 is reasonable .
we ourselves find this mass uncomfortably large .
a more reasonable nucleus of a few billion @xmath207 , corresponding say to an agn - like supermassive black hole , gives results very similar to those of the basic spls model described in
[ nobh ] and therefore suggests @xmath225 .
considering an elliptical g1 mass distribution [ ellsec ] all of the above models for the 0957 + 561 lens system assume azimuthal symmetry of the galaxy g1 .
while azimuthal symmetry is expedient for calculations , there is evidence even in the early observations of g1 by young _
et al . _ ( 1980 ) that the galaxy has elliptical isophotes .
@xcite disputed the young _ et al . _
ellipticity measurement of @xmath226 , suggesting that contamination from the inner quasar image led to an underestimate of the true ellipticity . from their own observations
, @xcite reported a more substantial ellipticity @xmath227 .
both studies agreed that the position angle of the g1 major axis is @xmath228 east of north .
if the g1 mass distribution is as flattened as the isophotes , one might worry that the poor @xmath164 of the various models discussed above stems from their common assumption of g1 azimuthal symmetry .
in addition , we would like to be assured that the derived @xmath12 for the spls model is not significantly affected if the g1 mass distribution includes some ellipticity .
unfortunately , there are very few instances in which the deflection function for elliptical mass distributions can be expressed in closed form .
kassiola & kovner ( 1993 , hereafter @xcite ) pointed out that the softened isothermal ( @xmath8 ) homoeoid is one such instance in which the lensing properties are analytic .
we have found that the deflection function for singular ( @xmath233 ) power - law homoeoids may also be expressed in closed form . in [ splhmods ] we provide a more detailed discussion of the lensing properties of these particular cases . as we noted in
[ errsubsec ] , the introduction of an adjustable radial power - law index @xmath2 is the distinctive feature of our mass model that allows us to fit the 0957 + 561 system without a supermassive black hole and with a much improved @xmath6 over forced - isothermal models .
we therefore hesitate to restrict ourselves to an isothermal homoeoid , as do kormann , schneider , & bartelmann ( 1994 ) in their unsuccessful attempt to model the gravitational lens producing the quadruple image b1422 + 231 .
we also do not wish to restrict ourselves to only singular elliptical power - law mass distributions because we have seen that the data accomodate finite - core spls models ( _ cf . _ fig .
[ rcbfig ] ) , even if our best - fit spls has a vanishing core radius .
the only alternative that allows us to work with lens equations in closed form is to approximate the splh with a suitable elliptical potential model . as we show in [ plummer ] , the tilted plummer family of elliptical potentials possesses the appropriate parameters to approximate the behavior of the splh .
the approximation is quite good for the 0957 + 561 system because we are fitting g1 with a core radius much smaller than the galaxy - image separations , and we assume that the mass distribution is no more elliptical than the observed isophotal ellipticity @xmath234 mentioned above . @xcite
have shown that for such small ellipticity , an elliptical potential is a very accurate representation of the true potential of an elliptical mass distribution .
we refer the reader to [ plummer ] for a detailed explanation of the tilted plummer potential approximation . taking the g1 mass distribution
to be as flattened as the surface brightness , @xmath235 , we find that the splh fits slightly better than the spls .
we summarize the results of our model - fitting with an elliptical g1 in table [ realmodtab ] .
the splh gives @xmath236 in contrast to @xmath202 for the spls , both with six degrees of freedom .
almost all of the @xmath6 reduction comes from a better fit to the observed @xmath237 .
the elliptical g1 does not help the problematic @xmath167 , which still remains 1.8 standard deviations below the value observed by @xcite . comparing table [ realmodtab ] with table [ bestnobh ]
, we see that adding ellipticity to the g1 density profile makes little change to the galaxy mass model parameters .
in particular , the elliptical model best - fit core radius @xmath238 mas , einstein radius @xmath239 , and power - law exponent @xmath240 are all well within the @xmath163 confidence limits of the spls fitting .
while the best - fit scaled shear remains exactly the same at @xmath241 , we find that the optimal shear position angle rotates substantially to @xmath242 .
this value is far outside the @xmath163 bounds @xmath243 we obtain when fitting to the spls .
these results are not entirely surprising , as we might expect the shift to an elliptical g1 to have the most impact upon the fitted external shear and position angle .
just as @xmath5 compensates for the shear induced by the cluster , so too does it compensate for shear caused by the lensing galaxy and not adequately reproduced by a spls model . now introducing a separate ellipticity due to g1
, we find that the external shear adjusts so as to model only the cluster shear , but to first order none of the g1 parameters are affected .
of course , we have assumed that the mass ellipticity of g1 is the same as its isophotal ellipticity .
if the mass is significantly more distorted , then we expect bigger changes . despite the significant shift in external shear direction ,
the best - fit elliptical g1 model gives a hubble parameter of @xmath244 .
this is a deviation of less than two percent from the @xmath198 we obtained with the spls , and well within the @xmath163 spls confidence interval of @xmath245 .
we therefore conclude that asphericity of the mass in g1 does not significantly alter the results we have obtained by fitting an spls , particularly with regard to the derived hubble constant .
considering perturbations to the cluster model [ sisgalsec ] up to this point , all of our 0957 + 561 lens models have assumed that we may add to the potential of g1 a locally quadratic potential due to the cluster , characterized by fixed convergence , shear , and shear position angle . as we have seen in
[ ellsec ] , this `` external '' quadratic potential may also compensate for failings in our particular choice of the g1 mass distribution . even if we were to have chanced upon the perfect parameterization of the g1 potential , we may nonetheless obtain a poor match to the observations if there were too much `` clumpiness '' in the local cluster potential around g1 for our quadratic external potential to handle .
this possibility is heightened by the large ( @xmath246 ) separation between the two images of 0957 + 561 .
assuming for the moment that the cluster dark matter potential is locally quadratic about g1 , we may then ask if there are other galaxies sufficiently close to g1 that their differential lensing properties from image a to image b might cause problems .
angonin - willaime , soucail , & vanderreist ( 1994 ) surveyed the 0957 + 561 region , obtaining photometry complete to @xmath247 in a 45 field as well as spectroscopy of 34 galaxies in a 6 field .
their galaxy redshifts confirmed the existence of a cluster at mean redshift @xmath248 containing g1 , which had been suggested by earlier redshift measurements from garrett , walsh , & carswell ( 1992 ) . from a sample of 21 member galaxies , angonin - willaime _
_ concluded that the cluster containing g1 is extended and poor , with a large ( @xmath249% ) spiral fraction .
of that sample , only two galaxies ( numbered 20 and 21 in table 2 of angonin - willaime _
et al . _ 1992 ) are located within @xmath250 of either 0957 + 561a or b. both galaxies , hereafter referred to as g20 and g21 , are faint ellipticals at redshift @xmath251 . both are also sufficiently near the quasar images for concern about their contribution to higher - order terms in the expansion of the potential about g1 .
the nearest of the two is g20 , with magnitude @xmath252 and offset from 0957 + 561b of @xmath253 , @xmath254 .
g21 has magnitude @xmath255 and 0957 + 561b offset of @xmath256 , @xmath257 . in order to test
what effect these nearby galaxies might have on our results , we include them as singular isothermal spheres .
the singular isothermal sphere is an attractive candidate not only because of its simple lensing properties , but also because of its relatively slow density falloff at large radii .
this allows us to be conservative in estimating the perturbative effect of g20 and g21 on the lensing at g1 . to assign isothermal velocity dispersions @xmath15 to the galaxies
, we use the faber - jackson relation @xmath258 ( faber & jackson 1976 ) .
following kochanek ( 1993 ) we adopt the values @xmath259 and @xmath260 km s@xmath29 appropriate for e / s0 galaxies . as noted by kochanek , 245
km s@xmath29 is at the high end of the estimated @xmath261 range of 183248 km s@xmath29 from dynamical estimates .
therefore we are being still more conservative with regard to the possible degree of lensing perturbation from the nearby galaxies .
we arrive at isothermal velocity dispersions of 193 km s@xmath29 for g20 and 165 km s@xmath29 for g21 .
these may be compared with the g1 value of 330 km s@xmath29 given in @xcite , also obtained with the faber - jackson relation .
the corresponding einstein radii are 0646 for g20 and 0472 for g21 .
we point out that these einstein radii are small fractions of the galaxies distances from the 0957 + 561 images less than 10% for g20 and less than 5% for g21 .
we summarize in table [ realmodtab ] the results of our model - fitting with a perturbed cluster .
the additional galaxies bring the reduced chi - square of the fit down to 3.4 , a significant improvement over the original spls and slightly superior to the splh .
better fitting of the quasar image positions is responsible for almost all of the @xmath6 reduction , as is the case in our study of g1 ellipticity ( see [ ellsec ] ) .
similarly , we find little improvement in either the overall magnification @xmath6 or the persistent discrepancy between model - predicted and observed @xmath168 .
the best - fit g1 einstein radius and cluster scaled shear are both down by @xmath262 .
the reduction of the einstein radius is because the the two external galaxies contribute an effective mass density , or convergence @xmath4 , in the vicinity of the lensed images . as we have already discussed in
[ sheet ] , any external @xmath4 leads to a corresponding reduction in the mass of the primary galaxy g1 .
the magnitude of the shear decreases partly for the same reason and partly because the external galaxies themselves produce some of the shear needed to explain the geometry of 0957 + 561 .
the shear position angle rotates by a significant amount ( to @xmath263 ) , again indicating that the external galaxies have absorbed a fraction of the required shear .
the g1 core radius and power - law index show no significant change , implying that these parameters are quite robust .
the best - fit hubble parameter @xmath154 is 0.582 , a drop of some 4% from the 0.605 value obtained using the plain spls .
this shift is compatible with the @xmath163 uncertainty we quoted for the spls model .
the fact that @xmath12 goes down rather than up is easily explained by the fact that the two external galaxies have introduced an effective convergence into the model ( see [ sheet ] ) . in conclusion
, there are no qualitative surprises from these calculations .
quantitatively , we find that the two nearest galaxies in the cluster have insignificant effect on our estimate of the hubble constant . eliminating the cluster degeneracy [ getkappa ] the focus to this point has been to show how recent vlbi observations of the lensed images 0957 + 561a , b limit the range of possible lens mass models , thereby limiting the model - dependent uncertainty in the determination of @xmath12 using this system .
quantitatively , the result is expressed in the bounds on the hubble parameter @xmath154 given in the previous section . in order to obtain @xmath12 , we see from equation [ [ h1.5eq ] ] that we need a measurement of the relative time delay @xmath13 and a determination of the cluster convergence @xmath4 .
considerable work has gone into the former ( _ e.g. _ @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , and it is only a matter of time before a precise ( @xmath264 ) value of @xmath13 will be settled upon for 0957 + 561 .
constraining the cluster convergence is a more difficult problem . as described in
[ sheet ] , the factor @xmath132 in the @xmath12 equation [ [ h1.5eq ] ] can not be eliminated purely by observations of the lensed images . however , as falco _
et al . _ ( 1991 ) showed , it is possible to estimate this factor by measuring the velocity dispersion of the lensing galaxy .
we apply this method in [ veldi ] below .
alternatively , a measurement of the velocity dispersion of the cluster may be used ( [ measclust ] ) .
velocity dispersion of the lensing galaxy g1 [ veldi ] the falco _
et al . _ ( 1985 ) degeneracy arises because all image observables are unchanged if the lensing galaxy mass is lowered by a factor @xmath132 and replaced by a mass sheet with convergence @xmath4 .
however , when such a transformation is made , the velocity dispersion of the galaxy will be lower than in the original model since the galaxy now has less mass .
turning this around , a measurement of the velocity dispersion of the lensing galaxy allows us to normalize the lens mass model and thereby constrain @xmath4 . because the mass of the galaxy scales linearly with the square of the line - of - sight velocity dispersion , @xmath265 , we may rewrite equation [ [ h1.5eq ] ] as @xmath266 where @xmath267 is the expected velocity dispersion of the model lens galaxy in the limit that the surrounding cluster has zero convergence .
falco _ et al .
_ ( 1991 ) applied this method to obtain their result of @xmath268 our analogous result from fitting the fgs model to the new vlbi data ( [ fgstest ] ) is @xmath269 in practice , one does not measure the dark matter velocity dispersion @xmath116 , but rather the velocity dispersion @xmath270 of the _ luminous _ matter in the lensing galaxy , which need not be equal to @xmath271 .
thus it is dangerous simply to take a measured g1 velocity dispersion and substitute it for @xmath271 in the above equations . in the following calculations with the spls model
, we use the virial theorem to estimate the model stellar velocity dispersion @xmath267 directly , for use in equation [ [ kappatov2 ] ] . for notational convenience
we define @xmath272 . in order to obtain a dispersion estimate that most directly compares with observation , our calculations below account for possible anisotropic orbits of the stars in the galaxy and also for the finite aperture of the slit used in the velocity dispersion measurements . by taking a suitable moment of the jean s equation , kochanek ( 1993 )
has shown that the line - of - sight velocity dispersion of the stars in a spherical galaxy satisfies @xmath273 where @xmath103 is the enclosed total mass at radius @xmath274 , and @xmath275 is the volume luminosity density .
note that equation [ [ veleq ] ] explicitly allows for the possibility that the luminous matter may have a different distribution than the total mass .
for the spls lens model , @xmath103 is given by @xmath276 where @xmath277 is the hypergeometric function ( _ cf .
_ @xcite ) .
we numerically evaluate @xmath277 by making use of the fact that the function solves the hypergeometric differential equation @xmath278{{df}\over{dz}}.\ ] ] to obtain @xmath275 we start with the observed surface brightness profile of g1 , which is well - fit by a de vaucouleurs profile with radius @xmath279 ( @xcite ) .
we then take @xmath275 to be the spherically symmetric function which produces the de vaucouleurs ( 1948 ) surface brightness function @xmath280\right\}$ ] .
we compute @xmath275 via abel inversion ( it cf .
@xcite ) of the de vaucouleurs profile @xmath281 according to @xmath282 two further issues need to be considered before using equation [ [ veleq ] ] to estimate @xmath283 .
first , the derivation of equation [ [ veleq ] ] assumes that the _ entire _ luminosity distribution is sampled the integrals of equation [ [ veleq ] ] are evaluated out to infinite radius from the galaxy center . in practice
, a velocity dispersion measurement is taken through a narrow - slit mask , and thus gives disproportionate weighting to the central region of the galaxy luminosity distribution .
kochanek ( 1993 ) addresses this issue and gives a corrected form of equation [ [ veleq ] ] taking into account finite - aperture effects .
we have applied kochanek s method assuming that the velocity dispersion of the galaxy is measured with a long slit of angular width 1 .
the second issue concerns the degree of isotropy of the stellar orbits . whereas equation [ [ veleq ] ] when integrated out to infinity is valid
regardless of the shapes of the orbits , the predicted dispersion measured through a finite aperture varies with orbit anisotropy .
kochanek ( 1993 ) has considered this issue as well and has presented results for a stellar system with a constant anisotropy factor , @xmath284 , relating radial and tangential velocity dispersions @xmath285 of orbits at any given radius : @xmath286 assuming isotropic @xmath287 orbits in g1 and a 1 slit aperture , we find that the best - fit spls model without compact nucleus ( table [ bestnobh ] ) predicts @xmath288 km s@xmath289 , with 95% confidence limits .
we show in figure [ veldisp ] a map of the predicted velocity dispersion as a function of core radius and power - law exponent . from this plot
we see that contours of constant @xmath290 are remarkably parallel to the contours of @xmath6 , so that @xmath290 varies by only @xmath291 over the range of allowed lens model parameters .
a somewhat larger uncertainty is present if we allow for anisotropic orbits .
figure [ trivel ] shows @xmath6 versus @xmath290 corresponding to three values of @xmath284 : @xmath292 , 0 , 0.2 .
we find that the shape of the curve is virtually unchanged over this range of anisotropy , but the best - fit velocity scales approximately as @xmath293 .
if we assume that the bright cluster elliptical g1 has @xmath294 , the fractional uncertainty in @xmath283 due to uncertainty in @xmath284 is no more than 6% .
this is reassuring , as we expect the uncertainty in the measured dispersion to be at least this high .
thus , we finally obtain the following result for the hubble constant : @xmath295 where we now have absorbed the model parameter uncertainty in @xmath283 into the error estimate on the coefficient . we have also included ( in quadrature ) the aforementioned 6% uncertainty due to the unknown g1 anisotropy @xmath284 . equation [
[ h_0final ] ] shows that with good measurements of the relative time delay and of the velocity dispersion of the galaxy in 0957 + 561 , it is possible to obtain a comparatively precise determination of the hubble constant .
the main residual uncertainty would arise from our incomplete coverage of model space by restricting ourselves to the spls model .
we discuss this issue in 7 .
direct measurement of the cluster potential [ measclust ] in addition to measuring the velocity dispersion of g1 , we may also break the degeneracy between the lensing galaxy and the cluster by making direct measurements of the surrounding cluster .
one approach is to measure the core radius @xmath296 and velocity dispersion @xmath297 of the cluster .
assume that the cluster potential @xmath298 corresponds to a softened isothermal sphere with the following simple form , @xmath299 where @xmath300 is the critical radius of the cluster .
if @xmath301 represents the angular separation of the 0957 + 561 images from the cluster center , the expansion of equation [ [ clpot ] ] at the galaxy center then yields a local convergence @xmath4 of the form ( @xcite ) @xmath302 this relation allows us to estimate @xmath4 knowing the velocity dispersion of the cluster @xmath297 , its core radius @xmath296 , and the position of the cluster center . if we take @xmath303 , @xmath304 , and @xmath305 as suggested by rhee
_ et al . _ ( 1995 ) , we obtain @xmath306 .
the present uncertainty in these cluster parameters places the cluster convergence in the approximate range @xmath307 .
equation [ [ kapclust ] ] is valid only if the cluster is smooth and relaxed ( see [ sisgalsec ] ) .
alternately one may try to infer the cluster potential directly through its lensing effects on background galaxies .
kaiser & squires ( 1993 ) showed that the shearing of background galaxies by a galaxy cluster may be inverted to recover a map of the projected cluster mass density , though subject to an overall degeneracy ( @xcite ) similar to the falco _
et al . _ ( 1985 ) degeneracy discussed above .
the degeneracy can be removed by measuring magnifications of background galaxies in addition to shear distortions ( @xcite ; @xcite ) .
preliminary results using a variant of the kaiser & squires method have been obtained for the cluster in 0957 + 561 ( @xcite ; rhee , fischer , & tyson 1995 ) and it is hoped that more detailed information on the cluster potential will soon be available .
such studies will at the very least provide a check on whether or not the quadratic cluster potential model introduced by fgs and used in this paper is valid .
an additional test would be to compare the scaled shear and its orientation as determined by our model with the direct estimates of these quantities from the cluster mass reconstruction .
summary and discussion the main result of this paper is that we have developed a new and more general model of the lensing mass in the double quasar 0957 + 561 .
the model consists of two parts : ( i ) a three parameter softened power - law sphere ( spls ) mass model ( [ splsmod ] ) for the primary lens galaxy g1 , which is more general than previous models in that it allows for a variable power - law radial dependence , and ( ii ) a two - parameter model ( [ sheet ] ) of the form proposed by fgs to describe the shear due to the surrounding cluster of galaxies . by using a larger set of data constraints
than in previous analyses , and by making use especially of the recent vlbi observations of garrett _ et al . _ ( 1994 ) , we are able to constrain the parameters of our model quite tightly .
the best - fit values of the five model parameters and their 95% confidence limits are shown in table [ bestnobh ] .
we obtain quite a strong upper limit on the angular core radius of the galaxy .
this limit is set primarily by the requirement that the model be consistent with the absence of a third image of the quasar near the center of the lensing galaxy .
the limit on the linear core radius is @xmath308 pc , which is similar to limits obtained by wallington & narayan ( 1993 ) and @xcite .
the main difference is that our result refers to a specific galaxy whereas wallington & narayan and @xcite carried out a statistical analysis of the ensemble of lens galaxies .
we are able to constrain the radial index @xmath2 of the lens model ( _ cf . _ eq . [ [ density ] ] ) to within a 10% range .
the allowed range of @xmath2 is close to the isothermal value , @xmath8 .
however , exact isothermality seems to be ruled out and our model suggests that the density falls more slowly than @xmath309 at large radius . the tight constraint on @xmath2 plus the fact that the model makes use of images located as far as 5 from the galaxy center makes this a fairly significant result .
the light from the lens galaxy fits a de vaucouleurs profile ( @xcite , @xcite ) , which falls off at large radii much more steeply than a mass - traces - light isothermal distribution .
our lens model therefore indicates a substantial amount of dark matter out to at least @xmath310 kpc from the center of the lens .
a similar result was obtained by kochanek ( 1995 ) for the lensing galaxy in the radio ring source mg1654 , but out to a somewhat smaller radius . using a different approach , but again based on gravitational lensing , brainerd , blandford , & smail ( 1995 ) found evidence for isothermal halos in galaxies extending over 100 kpc from the center .
these studies illustrate the unique advantages of gravitational lensing for studying the mass distributions of distant galaxies .
in addition to the spls model , we also test an fgs - like model of the lensing galaxy which consists of an approximate king potential ( with two parameters ) plus a compact central mass .
the idea here is to repeat the analysis of fgs using our larger data set to better constrain the model .
the best - fit fgs parameters are shown in table [ fgsparm ] , and turn out to be quite different from the values published by fgs . adding more data thus seems to have modified the model significantly , suggesting that the parameter values determined by fgs were not robust .
one reason could be that fgs had only one degree of freedom in their fit .
another noteworthy feature of the fgs model is that it requires an enormous central mass of @xmath311 .
such a large mass is unlikely to be a nuclear black hole and it is not clear exactly what it represents . as a result of the large central mass , the fgs model also has a large core radius in excess of the 0957 + 561b separation from the galaxy center .
one of the advantages of the spls model is that it does not require a central mass . nevertheless , in analogy with the fgs model
, we try to see what would be the effect of adding a point mass to the spls model .
the results are discussed in [ splsbh ] and shown in table [ bestwbh ] . in brief
we find that the reduced @xmath6 worsens by 1.2 when a central mass is added , which suggests that the data do not favor the inclusion of such a mass
. one troubling feature of our best - fit model is that the reduced @xmath6 is quite large , @xmath312 . perhaps our model , despite being more general than previous ones ,
is still too simple to fit all the data . incorporating reasonable perturbations due to g1 ellipticity ( [ ellsec ] ) and nearby cluster galaxies ( [ sisgalsec ] ) results in modest @xmath164 improvement , but never do we obtain @xmath313 for any of our spls variants .
efforts are under way to map the 0957 + 561 cluster mass distribution using weak distortions of background galaxies ( @xcite ) .
these studies should help to show whether an fgs - like quadratic potential is sufficient to account for the cluster lensing local to g1 .
if our cluster model is adequate , and we simply have not hit upon the correct shape for g1 , we must look to alternate mass models for the lensing galaxy . to this end
, we plan to explore more general mass - traces - light models , such as the following circularly - symmetric , four - parameter model proposed by tremaine ( 1995 ) : @xmath314^ { ( \gamma-\beta)/\alpha},\ ] ] where @xmath315 is the galaxy core radius .
surface brightness profiles of many galaxies are fit by this functional form with @xmath316 ; @xmath317 ; and @xmath318 .
the poor fit we find with the fgs model is also troubling . using the fgs model ,
we obtain a best - fit @xmath319 , rather worse than the spls and much worse than the @xmath320 quoted by fgs in their earlier fit to lower - resolution data .
we are surprised that neither the spls nor the fgs - type family of models can accommodate the recent vlbi data of @xcite .
why is @xmath164 large for all models tested ?
part of the reason seems to be that some of the data are measured with extraordinary precision , _
e.g. _ the positions of the quasar jet emission blobs are known to 0.1 mas ( table [ impostab ] ) .
the models are therefore penalized even for very small errors .
in fact , had we adopted the falco _
et al . _ ( 1991 ) convention of 1 mas vlbi positioning of the g1 center of mass , both the spls and the fgs models would have been grossly inconsistent with the data ( [ errsubsec ] ) . as noted in @xcite ,
having the galaxy center of mass coincident with the vlbi source provides a natural explanation that the radio emission originates in the core of this bright cluster elliptical .
our current model has no explanation for this observed radio emission .
another reason for the large @xmath164 of all our models is the large correlations quoted by @xcite among the elements of the 0957 + 561 relative magnification matrix and its gradient .
these correlations ( table [ magtab ] ) cause even fairly good fits of the measurements ( table [ chitable ] ) to contribute large amounts to the overall @xmath6 .
further vlbi observations of the jets in 0957 + 561 would be desirable both to confirm the present results and to improve the quality of the constraints . in view of the large @xmath164 values
, we have been conservative in setting confidence limits on the various parameter estimates ( [ splsnums ] ) we have taken the 95% limits to correspond with a @xmath6 increase of @xmath321 rather than 4 . despite this conservative approach ,
the spls mass model parameters are well constrained . moreover ,
when we perturb the fitting , either by giving the g1 mass distribution a modest ellipticity corresponding to the observed g1 isophotes ( [ ellsec ] ) or by including the nearest observed cluster members explicitly in the mass model ( [ sisgalsec ] ) , we see little change in most fitted model parameters and insignificant difference in the derived hubble constant .
the primary aim of developing a mass model for the lensing galaxy in 0957 + 561 was to use the model to estimate the hubble constant @xmath12 .
our main result is given in equation [ [ honovel ] ] .
the tight constraints that we obtain for the parameters of the lens model translate via equation [ [ hoeq ] ] to a correspondingly tight constraint on the relation for @xmath12 in terms of the relative time delay @xmath13 . the coefficient in @xmath12 equation [ [ honovel ] ]
has an uncertainty of only about @xmath322 .
the allowed range of @xmath12 is asymmetric with respect to the optimum value for reasons discussed in [ h15lims ] .
note that the results quoted in this paper correspond to a flat universe ( @xmath36 ) .
our derived value of @xmath12 increases almost linearly with decreasing @xmath37 , with a total increase less than 10% for @xmath323 .
as equation [ [ honovel ] ] shows , 0957 + 561 could be used to obtain a useful estimate of @xmath12 provided the relative time delay @xmath13 is measured with sufficient precision and the convergence @xmath4 due to the cluster is estimated .
many years of work have gone into collecting the required data for estimating @xmath13 , both in optical ( @xcite ; @xcite ) and radio ( @xcite ) .
estimates of @xmath13 have however varied , with two distinct values emerging from different data sets and analyses : @xmath324 days ( @xcite ; @xcite ; @xcite ) and @xmath325 days ( @xcite ) .
the two estimates individually have very small formal errors , so that they are seriously inconsistent with each other .
ongoing work is expected to resolve this problem shortly .
meanwhile , in this paper , we have explicitly presented our results according to both claimed time delays , using scaling factors @xmath326 and @xmath327 respectively .
the factor @xmath132 in equation [ [ h1.5eq ] ] arises because of a degeneracy in lens models discovered by falco _
et al . _ ( 1985 ) and gorenstein _ et al . _ ( 1988b ) .
these authors showed that any lens model can be modified by reducing the mass in the lens by an arbitrary factor and substituting a constant density mass sheet of appropriate convergence @xmath4 .
in such a transformation , all image observables except the time delay remain invariant .
this means that a given set of lens observations can not provide a unique mass model but rather a one - parameter family of models parametrized by @xmath4 .
since the value of @xmath4 modifies the predicted time delay , this unfortunately means that we can not obtain a unique estimate of @xmath12 from a given lens unless we independently estimate @xmath4 .
note , however , that @xmath4 must necessarily be a positive number since it is proportional to the surface mass density of the sheet .
therefore , we can always obtain an upper bound on @xmath12 ( narayan 1991 ) . from equation [ [ honovel ] ] we see that the 95% upper bound is @xmath328 for @xmath329 yr and @xmath330 for @xmath331 yr .
fgs showed that it is possible to eliminate the @xmath4 degeneracy if we could measure the line - of - sight velocity dispersion @xmath332 of the lensing galaxy .
the idea is that the family of models with different values of @xmath4 have different amounts of mass in the primary lensing galaxy .
this mass can be scaled to @xmath333 through the virial theorem .
we discuss this approach in [ veldi ] and show how the method would work in the case of the spls model . equation [
[ h_0final ] ] gives the final result .
rhee ( 1991 ) measured @xmath15 of the stars in the lensing galaxy in 0957 + 561 to be @xmath334 .
unfortunately , the measurement is not sufficiently precise to provide a significant constraint on @xmath12 . a more precise measurement of @xmath15 should be possible with a large optical telescope and is highly desirable , as it would eliminate this last uncertainty in the modeling of 0957 + 561 .
the velocity dispersion @xmath15 refers only to the stars in the lensing galaxy , whereas the gravitational lensing is done by the total mass .
there has been some uncertainty as to how the measured stellar @xmath15 should be related to the @xmath15 of the total mass , which is the relevant quantity for normalizing the galaxy mass distribution .
the straightforward approach is to assume that the velocity dispersion of the stars and that of the dark matter particles are equal .
however , turner _
et al . _ ( 1985 ) argued that in many circumstances the @xmath15 of the stars would be lower than that of the total mass by a factor of @xmath335 .
this makes quite an important difference to the results .
for instance , narayan ( 1991 ) derived on the basis of the fgs model , assuming @xmath336 days ( @xcite ) and @xmath337 ( @xcite ) , that @xmath338 if the dark matter has the same dispersion as the stars and @xmath339 if the correction factor of @xmath335 is applied . the approach used in this paper , based on the work of kochanek ( 1993 ) , avoids the ambiguity since it is based on a fundamental application of the virial theorem .
our model here gives , for the same parameters as the ones employed by narayan ( 1991 ) , @xmath340 .
narayan ( 1991 ) has discussed an interesting additional benefit that one obtains by measuring @xmath15 . once the mass of the lens has been normalized through such a measurement , it is possible to obtain an estimate of @xmath12 that is independent of the source redshift . in other words ,
the distance to the source drops out of the relations .
this is , of course , not a particular advantage since most often the source redshift is known .
however , a corollary of the theorem is that if there are additional mass sheets with shear between the lens and the source , say due to other clusters , the formula for @xmath12 is transparent to their presence provided the additional clusters are describable by quadratic potentials over the angular extent of the lensed images .
this theorem is quite useful .
there is evidence for a second cluster at redshift 0.51 in the field of 0957 + 561 ( @xcite , @xcite ) , and there may well be other clusters at higher redshift .
it is advantageous to be able to estimate @xmath12 independently of these complications .
another interesting result is that with a measurement of @xmath15 one directly obtains the angular diameter distance @xmath341 to the lens regardless of cosmological model , i.e. the result is independent of the values of @xmath342 and @xmath42 ( @xcite ) .
the @xmath4 degeneracy can also be eliminated by estimating the mass surface density of the cluster directly .
one simple method is to measure the core radius of the cluster , its velocity dispersion , and the location of the lens relative to the cluster center .
we describe in [ measclust ] how this information can be translated into an estimate of @xmath4 . using the parameter values given by rhee _
et al . _ ( 1995 ) , we estimate that @xmath343 for the lensing cluster in 0957 + 561 .
this translates to @xmath344 for @xmath345 and @xmath346 for @xmath347 .
a more ambitious undertaking is to map the two - dimensional surface density of the cluster using the weak distortions of background galaxies .
the idea for this method goes back to tyson , wenk , & valdes ( 1990 ) and kaiser & squires ( 1993 ) .
et al . _ ( 1995 ) are currently applying the method to the field around 0957 + 561 and results are awaited .
what is the future for lens - based measurements of the hubble constant ?
we believe 0957 + 561 will deliver a result soon .
it is only a matter of time before the controversy over the time delay is settled . our model ( _ cf .
_ eq . [ [ h1.5eq ] ] ) will then directly provide an upper bound on @xmath12 or a direct estimate if we take the value of @xmath348 mentioned above .
measuring @xmath15 and thereby obtaining a more reliable estimate of @xmath4 is more challenging , but efforts are under way and once again we are optimistic . with measurements of both @xmath13 and @xmath15 , our model should provide an estimate of @xmath12 which would be quite competitive with other determinations .
there is unfortunately one major remaining uncertainty , namely whether our model captures the mass distribution of the lens sufficiently well .
the poor reduced @xmath6 of our model is certainly a concern .
perhaps a mass - traces - light model will improve matters , or perhaps future observations will resolve our spls discrepancies . in any case , it is desirable to study additional lenses in order to obtain other independent estimates of @xmath12 .
it is generally agreed that 0957 + 561 is not the best source for estimating @xmath12 since the presence of the cluster adds an extra layer of complication .
several other good candidates are available where the lensing galaxy appears to be more or less isolated .
such systems are easier to model .
the radio ring sources are particularly promising since the modeling of these is likely to be more reliable than with multiply - imaged optical quasars .
the experience with 0957 + 561 shows that having information on the full relative magnification matrix and its spatial derivative is invaluable for constraining the mass model .
therefore , resolved sources which can be mapped and modeled in detail are likely to be much superior to unresolved sources . to conclude , we emphasize that the lens - based method of estimating @xmath12 is completely independent of all other methods and works directly on high redshift sources without using any intermediate distance ladders . furthermore , the method is based on very basic geometry and physics .
these are substantial advantages , and we feel that the method deserves to be pursued seriously .
we are grateful to e. falco , m. garrett , c. kochanek , j. lehr , p. schechter , r. schild , and i. shapiro for discussions and comments , and to the referee for useful suggestions .
this research was supported in part by grant ast 9423209 from the nsf . approximating a softened power - law homoeoidal lens lensing properties of the splh [ splhmods ]
as we often observe galaxies to have elliptical isophotes , it would be appealing to model their lensing properties with a mass distribution having elliptical isodensity contours .
a simple example is the homoeoid , whose isodensity contours are concentric ellipses of constant ellipticity and position angle .
the surface density of the homoeoid varies only as a function of the elliptical `` radius '' @xmath349 from the center of the distribution : @xmath350 where the parameter @xmath231 reflects the degree of flattening . in the above equation and throughout the remainder of this appendix
, we assume that the major axis of the elliptical profile is aligned with the @xmath23-axis of our coordinate system . from equation [ [ remeq ] ]
, we relate the asphericity parameter @xmath231 to the axis ratio of isodensity contours : @xmath351 we emphasize that @xmath231 is _ not _ equivalent to the ellipticity @xmath352 , defined by @xmath353 .
bourassa & kantowski ( 1975 ) found that the lensing properties of generic elliptical mass distributions may be compactly expressed by adopting a complex angular notation , where vector angles @xmath44 map to the complex plane according to @xmath354 .
they derived the complex ray deflection @xmath355 for a general homoeoidal density profile @xmath356 : @xmath357 with @xmath358 as the complex conjugate of @xmath359 .
here @xmath360 denotes the branch of the complex square root for which @xmath361 and @xmath358 lie in the same quadrant of the complex plane ( bray 1984 ) .
as noted by schramm ( 1994 ) , closed solutions of equation [ [ compdef ] ] for general elliptical lenses have rarely been found ( narasimha 1982 , @xcite ) and are by construction restricted to density profiles with homoeoidal symmetry .
@xcite have shown that equation [ [ compdef ] ] has a closed form for softened isothermal homoeoids , whose density profiles are characterized by central convergence @xmath362 and core radius @xmath1 : @xmath363 the complex ray deflection for the softened isothermal homoeoid is @xmath364 } \right\},\ ] ] where @xmath365 , @xmath366 , and @xmath367 is the asymptotic einstein radius for @xmath368 .
we have found that a closed form of the ray deflection equation [ [ compdef ] ] also exists for the singular power - law homoeoids .
the density profile of the singular power - law homoeoid is parameterized by asymptotic ( @xmath369 ) einstein radius @xmath110 and power - law index @xmath2 of radial mass increase : @xmath370 substituting the surface density from equation [ [ singplh ] ] into the integrand of equation [ [ compdef ] ] , we obtain the complex ray deflection of the singular power - law homoeoid : @xmath371 where @xmath372 and @xmath373 is the complex hypergeometric function . in the limit of azimuthal symmetry ( @xmath374 , @xmath375 and the hypergeometric function becomes unity .
equation [ [ singplhdef ] ] therefore reduces to @xmath376 , which may be compared with the deflection of a singular power - law sphere ( eq . [ [ goodalpheq ] ] with @xmath233 )
. tilted plummer elliptical potentials [ plummer ] to overcome the difficulties of elliptical mass distributions , we turn to elliptical potentials .
such potentials vary only as a function of the elliptical radius @xmath377 $ ] , characterized by asphericity parameter @xmath378 .
although the lensing properties of elliptical potentials are easily expressed in closed form , their associated isodensity contours are sometimes unphysical .
elliptical potentials with more than a moderately high asphericity @xmath379 require isodensity contours that are dumbbell - shaped ( @xcite ) . in certain cases , highly flattened elliptical potentials may even require _ negative _ isodensity contours ( @xcite ) .
however , elliptical potentials provide quite an accurate representation of elliptical mass distributions for ellipticities @xmath380 .
fortunately for this study , the 0957 + 561 lensing galaxy g1 has an isophotal ellipticity compatible with this limit .
isothermal ( @xmath101 ) elliptical potentials have been popular in the modeling of aspherical galaxies ( _ e.g. _ blandford & kochanek 1987 ; kochanek & blandford 1987 ; @xcite ) , but we seek an elliptical potential which has the radial power - law generality of the splh .
the family of tilted plummer potentials fits this description : @xmath381^{\eta/2},\ ] ] where the potential @xmath382 , expressed in units of radian@xmath383 , is equivalent to @xmath384 in the notation of [ approxcoordsec ] . here
@xmath385 represents the core radius of the potential , and we recycle the spls parameters @xmath110 and @xmath2 to represent an effective einstein radius and radial power - law index , respectively . taking the gradient of equation [ [ ellpoteq ] ] , we find that the ray deflection of the tilted plummer model is given by @xmath386 comparing the asymptotic @xmath387 behavior of equation [ [ ellpotdef ] ] with the spls analogue ( eq . [ [ goodalpheq ] ] ) , we see that the tilted plummer model parameters @xmath115 and @xmath2 and their spls counterparts are equivalently defined .
@xcite show that the tilted plummer potential and splh potential approximately coincide when the asphericities and core sizes are sufficiently small .
in particular , the sizes of the respective tangential caustics coincide .
the g1 core sizes compatible with nondetection of a third qso image ( @xmath399 ) are indeed small , and even at a g1 mass ellipticity of 0.30 there is only minor deviation between the models isodensity contours ( _ cf . _
@xcite fig . 4 ) .
we therefore conclude that our treatment of g1 as an splh is not compromised by approximating the lensing properties with a suitably chosen tilted plummer potential .
lr@ @xmath400 lr@ @xmath400 lr@ @xmath400 lemission & & & + component & & & + @xmath76 & 14.2 & 0.1 & & + @xmath80 & 10.6 & 0.2 & 48.3 & 0.1 & 19.9&0.1 + @xmath77 & 11.4 & 0.1 & & + @xmath81 & 7.0 & 0.5 & 58.8 & 0.1 &
17.8&0.1 + lrrrrrr covar & @xmath401 & @xmath402 & @xmath408 & @xmath409 & @xmath410 & @xmath167 + @xmath401 & 1.00 & & & & & + @xmath402 & 0.46 & 1.00 & & & & + @xmath408 & @xmath411 & @xmath412 & 1.00 & & & + @xmath409 & @xmath413 & @xmath414 & 0.21 & 1.00 & & + @xmath410 & 0.96 & 0.40 & @xmath415 & @xmath416 & 1.00 & + @xmath168 & 0.70 & 0.79 & @xmath417 & @xmath418 & 0.70 & 1.00 + foo abramowitz , m. , & stegun , i. a. 1964 , handbook of mathematical functions , applied mathematics series , vol .
55 ( washington : national bureau of standards ; reprinted 1968 by dover publications , new york ) angonin - willaime , m .- c . ,
soucail , g. , & vanderreist , c. 1994 , , 291 , 411 bartelmann , m. , & narayan , r. 1995 , , 451 , 60 bernstein , g. m. , tyson , j. a. , & kochanek , c. s. 1993 , , 105 , 816 binney , j. , & tremaine , s. d. 1987 , galactic dynamics ( princeton : princeton university press ) blandford , r. d. , & kochanek , c. s. 1987 , , 321 , 658 blandford , r. d. , & narayan , r. 1992 , , 30 , 311 bonometti , r. j. 1985 , ph.d .
thesis , massachusetts institute of technology borgeest , u. , & refsdal , s. 1984 , , 141 , 318 bourassa , r. r. , & kantowski , r. 1975 , , 195 , 13 brainerd , t. g. , blandford , r. d. , & smail , i. 1995 , , in press bray , i. 1984 , , 208 , 511 broadhurst , t. j. , taylor , a. n. , & peacock , j. a. 1995 , , 438 , 49 dahle , h. , maddox , s. j. , & lilje , per b. 1994 , , 435 , l79 de vaucouleurs , g. 1948 , ann.dap .
, 11 , 247 faber , s. m. , & jackson , r. e. 1976 , , 204 , 668 falco , e. e. , gorenstein , m. v. , & shapiro i. i. 1985 , , 289 , 1 falco , e. e. , gorenstein , m. v. , & shapiro i. i. 1991 , , 372 , 364 garrett , m. a. , walsh , d. , & carswell , r. f. 1992 , , 254 , 27 garrett , m. a. , calder , r. j. , porcas , r. w. , king , l. j. , walsh , d. , & wilkinson , p. n. 1994 , , 270 , 457 gorenstein , m. v. , shapiro i. i. , cohen , n. l. , corey , b. e. , falco , e. e. , marcaide , j. m. , rogers , a. e. e. , whitney , a. r. , porcas , r. w. , preston , r. a. , & ruis , a. 1983 , science , 219 , 54 gorenstein , m. v. , shapiro i. i. , rogers , a. e. e. , cohen , n. l. , corey , b. e. , porcas , r. w. , falco , e. e. , bonometti , r. j. , preston , r. a. , rius , a. , & whitney , a. r. 1984 , , 287 , 538 gorenstein , m. v. , cohen , n. l. , shapiro i. i. , rogers , a. e. e. , bonometti , r. j. , falco , e. e. , bartel , n. , & marcaide , j. m. 1988a , , 334 , 42 gorenstein , m. v. , falco , e. e. , & shapiro i. i. 1988b , , 327 , 693 greenfield , p. d. , roberts , d. h. , & burke , b. f. 1985 , , 293 , 370 kaiser , n. 1992 , , 388 , 272 kaiser , n. & squires , g. 1993 , , 404 , 441 kassiola , a. , & kovner , i. 1993 , , 417 , 450 kochanek , c. s. , & blandford , r. d. 1987 , , 321 , 676 kochanek , c. s. 1991 , , 382 , 58 kochanek , c. s. 1993 , , 419 , 12 kochanek , c. s. 1995 , , 445 , 559 kormann , r. , schneider , p. , & bartelmann , m. 1994 , , 286 , 357 lehr , j. , hewitt , j. n. , roberts , d. h. , & burke , b. f. 1992 , , 384 , 453 narasimha , d. , subramanian , k. , & chitre , s. m. 1982 , , 200 , 941 narayan , r. 1991 , , 378 , l5 pelt , j. , hoff , w. , kayser , r. , refsdal , s. , & schramm , t. 1994 , , 286 , 775 pelt , j. , kayser , r. , refsdal , s. , & schramm , t. 1995 , , in press porcas , r. w. , booth , r. s. , browne , i. w. a. , walsh , d. , & wilkinson , p. n. 1981 , nature , 289 , 758 press , w. h. , flannery , b. p. , teukolsky , s. a. , & vetterling , w. t. 1992 , numerical recipes ( second edition , cambridge : cambridge univ . press ) press , w. h. , rybicki , g. b. , & hewitt , j. n. 1992b , , 385 , 404 press , w. h. , rybicki , g. b. , & hewitt , j. n. 1992c , , 385 , 416 refsdal , s. 1964 , , 128 , 295 refsdal , s. 1966 , , 132 , 101 rhee , g. 1991 , nature , 350 , 211 rhee , g. , fischer , p. , & tyson , j. a. 1995 , in aip conference proceedings no . 336 :
dark matter , college park , maryland , 1994 , ed .
s. s. holt & c. l. bennett ( new york : aip press ) roberts , d. h. , greenfield , p. e. , hewitt , j. n. , burke , b. f. , & dupree , a. k. 1985 , , 293 , 356 schild , r. e. 1990 , lect . not .
phys . , 360 , 102 schild , r. e. & thomson , d. j. 1993 , in gravitational lenses in the universe : proceedings of the 31st lige international astrophysical colloquium , june 21 - 25 , 1993 , ed . j. surdej _
_ schneider , p. , ehlers , j. , & falco , e. e. 1992 , gravitational lenses ( heidelberg : springer ) schneider , p. & seitz , c. 1995 , , 294 , 411 schramm , t. 1994 , , 284 , 44 shapiro i. i. 1964 , , 13 , 789 stockton , a. 1980 , , 242 , l141 thomson , d. j. , & schild , r. e. 1993 , in proceedings of the international conference on applications of time series analysis in astronomy and meteorology , september 6 - 10 , 1993 , padua tremaine , s. d. 1995 , private communication turner , e. l. , ostriker , j. p. , gott , j. r. i. 1984 , , 284 , 1 turner , e. l. 1990 , , 365 , l43 tyson , j. a. , wenk , r. a. , & valdes , f. 1990 , , 349 , 1 vanderreist , c. , schneider , j. , herpe , g. , chevreton , m. , moles , m. , & wlrick , g. 1989 , , 215 , 1 vanderreist , c. , gosset , e. , remy , m. , swings , j .-
1992 , lect . not .
phys . , 406 , 107 wallington , s. , & narayan , r. 1994 , , 403 , 517 walsh , d. , carswell , r. f. , & weymann , r. j. 1979 , nature , 279 , 381 wambsganss , j. , & paczynski , b. 1994 , , 108 , 1156 young , p. , gunn , j. e. , kristian , j. , oke , j. b. , & westphal , j. a. 1980 , , 241 , 507
|
we present a simple mass model for the lensing galaxy in the gravitationally lensed quasar 0957 + 561 .
we represent the galaxy as a softened power - law sphere ( spls ) , a generalization of the singular isothermal sphere with three parameters @xmath0 , the central density , @xmath1 , the angular core radius , and @xmath2 , the radial index which is defined such that mass increases as @xmath3 at large radius . as in previous studies we approximate the galaxy cluster surrounding the lensing galaxy by means of a quadratic potential described by its convergence @xmath4 and shear @xmath5 .
a feature of the model is that it does not require a large central compact mass .
we fit the model to a recent high resolution vlbi map of the two images of 0957 + 561 .
the data provide a number of independent constraints and the model - fit has six degrees of freedom , which is a significant improvement over previous models .
although the reduced @xmath6 of the best - fit model is only 4.3 , nevertheless we obtain a tight constraint on the radial index , @xmath7 , at the 95% confidence level .
thus , the galaxy has mass increasing slightly more rapidly than isothermal ( @xmath8 ) out to at least @xmath9 kpc .
since the light from the galaxy follows a de vaucouleurs profile , we deduce that the mass - to - light ratio of the galaxy increases rapidly with increasing radius .
we also obtain an upper limit on the core radius , namely @xmath10 or linear core radius @xmath11 pc .
we use the model to calculate the hubble constant @xmath12 as a function of the time delay @xmath13 between the two images .
we obtain @xmath14 once @xmath13 is measured , this will provide an upper bound on @xmath12 since @xmath4 can not be negative .
in addition , the model degeneracy due to @xmath4 can be eliminated if the one - dimensional velocity dispersion @xmath15 of the lensing galaxy is measured . in this case
we find that @xmath16 we find that these results are virtually unchanged when we investigate the effects of ellipticity in the lensing galaxy and clumpiness in the lensing cluster .
# 1#1 = cmmib10 at 12pt = cmmib8 = cmmib7 99 9 `` 710b ' ' 710c `` 710d ' ' 710e `` 710f ' ' 7110 `` 7111 ' ' 7112 `` 7113 ' ' 7114 `` 7115 ' ' 7116 `` 7117 ' ' 7118 `` 7119 ' ' 711a `` 711b ' ' 711c `` 711d ' ' 711e `` 711f ' ' 7120 " 7121 introduction in two seminal papers , refsdal ( 1964 , 1966 ) showed that there should exist a time delay between flux variations of multiple images of a lensed background source , and demonstrated that the time delay is inversely proportional to the hubble constant @xmath12 .
gravitational lenses thus can be used to determine @xmath12 independently of traditional distance - ladder techniques .
the lens method of estimating @xmath12 requires a measurement of the time delay and a determination of the mass distribution of the lens .
a particularly promising candidate for this technique is the first gravitational lens discovered , the so - called `` double quasar '' 0957 + 561 ( walsh , carswell , & weymann 1979 ) , which consists of a pair of lensed images of a @xmath17 quasar separated by @xmath18 on the sky .
the galaxy responsible for the lensing , denoted g1 , was discovered by stockton ( 1980 ) .
this galaxy is a bright cluster elliptical at redshift @xmath19 , residing in a cluster of galaxies which also contributes to the lensing ( garrett , walsh , & carswell 1992 ) .
long - term monitoring of the optical ( @xcite ; @xcite ) and radio ( @xcite ) light curves of the two quasar images has provided strong evidence for a time delay , though there is still some uncertainty regarding the actual value of the delay ( press , rybicki , & hewitt 1992b , c ; @xcite ; @xcite ) .
the observational uncertainties in the other observables such as the image positions , the relative image magnification , and the source and lens redshifts are small , leaving the lens mass distribution as the chief remaining obstacle to estimating @xmath12 using this system ( @xcite ) .
the mass distribution of the lens is difficult to measure directly , and has to be constrained using the observations of the lensed images .
it is customary to postulate a simple functional form for the lens mass profile , and to adjust the parameters of the model so as to obtain the best fit to the observables , namely the quasar image positions and relative image magnifications .
such calculations have been done by various authors in the past ( @xcite ; @xcite ; greenfield , roberts , & burke 1985 ; @xcite ; falco , gorenstein , & shapiro 1991 , hereafter @xcite ; bernstein , tyson , & kochanek 1993 , hereafter @xcite ) using the data available at the time .
we present here a new model of 0957 + 561 which improves on previous work in two respects .
first , we employ a new parameterization of the lens mass , namely the softened power - law sphere ( spls ) model , which allows us to explore a wider range of radial mass profiles than in earlier studies .
our model space includes the standard isothermal sphere model ( e.g. turner , ostriker , & gott 1984 ) as a particular case , and one of our aims is to use the data to determine the degree to which the galaxy deviates from an isothermal radial profile .
we also include a core radius for the galaxy and represent the surrounding cluster by means of a quadratic potential as proposed by fgs .
second , we make use of a larger set of data constraints than in any previous study .
fgs for their modeling used high - resolution vlbi maps of 0957 + 561 ( @xcite ) which resolved features of an inner radio jet extending @xmath20 milliarcseconds from the quasar core of each image .
this permitted them to derive a @xmath21 relative magnification matrix between the two quasar images which they used to constrain their model .
recent @xmath2218 cm vlbi mapping of 0957 + 561a , b by garrett _
et al . _ ( 1994 , hereafter @xcite ) revealed even greater detail in the fine structure of the quasar images .
the a and b inner jet regions are now resolved into five centers of emission with sub - milliarcsecond accuracy in the positions .
in addition , @xcite report a gradient in the relative magnification tensor between the quasar core and the end of the inner jet .
this provides two additional data constraints which we include in our model - fitting . as a result of the additional data
, we have a better - constrained lens model , with six degrees of freedom in the data fits , as against the previous studies by fgs and kochanek ( 1991 ) which had only one degree of freedom . in 2 of the paper
we briefly introduce the notation and approximations employed in this study , as well as the basic lens equations that allow us to test mass models against constraints from observation . in
3 we review the observations of 0957 + 561 , paying particular attention to the high - resolution vlbi mapping by @xcite which provides the majority of our model constraints .
we introduce the spls lens model in 4 and derive its lensing properties .
we then employ the lensing equations of 2 to show explicitly the dependence of model - predicted observables , including the time delay , upon our various model parameters . for comparison
, we do the same also for the fgs lens model . in 5 we describe the results of our model - fitting , both with our spls model and with the fgs model , and including the effects of ellipticity in g1 and clumpiness in the cluster .
we derive confidence limits on the lens parameters and obtain bounds on @xmath12 . in 6 we address one of the chief difficulties in obtaining tight @xmath12 bounds from 0957 + 561 , namely the uncertainty regarding the amount of lensing contributed by the cluster .
we describe how this uncertainty can be removed by measuring the velocity dispersion of the galaxy , as shown by fgs , or by measuring the shape and velocity dispersion of the cluster .
we summarize the paper in 7 and discuss prospects for further improvements in the method , both with 0957 + 561 and with other lens systems .
lensing geometry and notation [ approxcoordsec ] we employ a cartesian coordinate system on the sky with the origin at the center of mass of the lensing galaxy g1 , the @xmath23-axis taken positive to the east , and the @xmath24-axis positive to the north .
we define position angles to be zero toward the north and increasing eastward .
we employ the standard angular diameter distance appropriate to a friedmann universe . for an observer at redshift @xmath25 and a source at redshift @xmath26 ,
the angular diameter distance is given by @xmath27 where @xmath28 km s@xmath29 mpc@xmath29 is the hubble constant . a proper length @xmath30 at @xmath26 subtends an angle @xmath31 at @xmath25 .
it is customary to use the abbreviations @xmath32 , @xmath33 , and @xmath34 when referring to the angular diameter distances from observer to deflector , observer to source , and deflector to source , respectively .
we also introduce the _ effective lens distance _
@xmath35 which appears in the lensing equations below . for simplicity , we assume @xmath36 in what follows .
the sensitivity of the results to @xmath37 is rather small ( @xcite ) . for the 0957 + 561 lens ,
@xmath38 increases approximately linearly with decreasing @xmath37 , to a value @xmath39 larger for @xmath40 than the fiducial value for @xmath41 .
a non - zero cosmological constant @xmath42 can have a more important effect on the results ( _ e.g. _ turner 1990 ) , but we do not explore the dependence in this paper . using standard notation ( _ e.g. _ @xcite ; schneider , ehlers , & falco 1992 ) , we represent angular positions at the distance of the background source ( the `` source plane '' ) with the vector @xmath43 and angular positions at the distance of the deflector ( the `` image plane '' ) with @xmath44 .
the lens mass deflects light rays at the image plane through an angle which we represent with @xmath45 .
these quantities are then related via the lens equation , @xmath46 where the _ reduced _ deflection angle @xmath47 is related to the true deflection angle @xmath48 at the image plane by @xmath49 the ray deflection function @xmath50 may in general allow multiple solutions @xmath51 to the lens equation [ [ lenseq ] ] for a given @xmath43 .
if a source happens to lie at such a @xmath43 , we observe multiple images of the source at positions @xmath51 , a gravitational `` mirage '' .
the lens deflection @xmath48 is sensitive only to the surface mass density of the deflector . with @xmath52
representing linear position in the image plane , we may express the lens deflection angle as the two - dimensional gradient of a potential @xmath53 , @xmath54 the potential @xmath55 is related to the lens surface mass density @xmath56 according to @xmath57 .
equivalently , we may express the deflection in terms of the green s function of the poisson operator , @xmath58 for a radially symmetric surface mass density profile @xmath59 , the above equation simplifies to @xmath60 where @xmath61 is the projected mass of the deflector within cylindrical radius @xmath30 .
|
quantum coherence is considered to be one of the most important resources in quantum physics .
it has many significant applications in various subjects , such as quantum optics @xcite , quantum information @xcite , quantum biology @xcite , and thermodynamics @xcite .
however , a general criterion to quantify the coherence of quantum states in information theoretic terms has only been proposed recently @xcite , although the topic has ever been attempted in early papers @xcite .
the general criterion has triggered the community s great interest @xcite , and based on the criterion , various concrete measures of coherence have been given .
the @xmath0 norm of coherence and the relative entropy of coherence , which satisfy the general criterion of quantifying the coherence of quantum states , were first suggested as two coherence measures based on distance in ref.@xcite . the coherence measure based on skew information @xcite , and the coherence measure based on entanglement @xcite
were then proposed .
more expressions were subsequently shown to fulfill the general criterion of quantifying coherence , such as the coherence of formation , the distillable coherence , and the coherence cost @xcite .
there have been several different expressions of coherence measures , as mentioned above .
however , these different expressions , which are based on various physical contexts , result in different values of coherence for a state in general . in this situation
, an interesting question is whether these coherence measures give the same ordering of states .
that is , for two coherence measures @xmath1 and @xmath2 , does @xmath3 imply @xmath4 for any two states @xmath5 and @xmath6 ?
as two effective measures of coherence , one may expect @xmath1 and @xmath2 to give the same ordering for all the states even if numerical values of them are not quantitatively equal . in the paper
, we address this issue .
our discussions focus on the two well - known coherence measures , the @xmath0 norm of coherence and the relative entropy of coherence , which have been widely used as a tool to investigate various aspects of coherence , such as the freezing phenomena of quantum coherence @xcite , the complementarity relations for coherence measures @xcite , and the cohering and decohering power of quantum channels @xcite .
we will show that there are the states for which the two measures do not give the same ordering . as an extension of our analyses
, we will also briefly discuss the other coherence measures , the @xmath0 norm of coherence and the coherence of formation as well as the relative entropy of coherence and the coherence of formation , showing that they do not give the same ordering too .
the paper is organized as follows . in sec .
2 , we review some notions related to quantifying coherence and recall the three coherence measures under considered , i.e. , the @xmath0 norm of coherence , the relative entropy of coherence , and the coherence of formation . in sec . 3 , we present our main results , i.e. , we show that there are the states for which the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering , and also give examples to show that the @xmath0 norm of coherence and the coherence of formation , as well as the relative entropy of coherence and the coherence of formation , do not give the same ordering too .
4 is our conclusions .
we review some notions related to quantifying coherence , such as incoherent states and incoherent operations , and recall the three coherence measures , the @xmath0 norm of coherence , the relative entropy of coherence , and the coherence of formation .
let @xmath7 be the hilbert space for a @xmath8-dimensional quantum system .
a particular basis of @xmath7 is denoted as @xmath9 , which is chosen according to the physical problem under discussion .
coherence of a state is then measured based on the basis chosen @xcite .
a state is called an incoherent state if and only if its density operator is diagonal in the basis , and the set of all the incoherent states is usually denoted as @xmath10 . therefore , a density operator @xmath11 is of the form @xmath12 .
all other states , which can not be written as diagonal matrices in the basis , are called coherent states .
hereafter , we use @xmath13 to represent a general state , a coherent state or an incoherent state , and use @xmath14 specially to denote an incoherent state .
a completely positive trace preserving map , @xmath15 , is said to be an incoherent completely positive trace preserving ( icptp ) map , or an incoherent operation , if the kraus operators @xmath16 satisfy not only @xmath17 but also @xmath18 , i.e. , each @xmath16 maps an incoherent state to an incoherent state .
a functional @xmath19 can be taken as a measure of coherence if it satisfies the four postulates @xcite : + ( c1 ) @xmath20 , and @xmath21 if and only if @xmath22 ; + ( c2 ) @xmath23 for all icptp maps @xmath24 ; + ( c3 ) @xmath25 for all @xmath26 with @xmath27 and @xmath18 ; + ( c4 ) @xmath28 for any set of states @xmath29 and any @xmath30 with @xmath31 . + postulate ( c1 ) imposes coherence measures to be a non - negative functional .
( c2 ) and ( c3 ) show the monotonicity of coherence measures under incoherent operations and that under selective incoherent operations .
( c4 ) means that coherence measures are nonincreasing under mixing of states .
it is obvious that ( c3 ) and ( c4 ) imply ( c2 ) .
these four postulates comprise a general criterion of defining a coherence measure . in accordance with the general criterion ,
several coherence measures have been put forward . out of them , there are the @xmath0 norm of coherence , the relative entropy of coherence , and the coherence of formation , which are considered in this paper .
the @xmath0 norm of coherence is defined as @xmath32 where @xmath33 are entries of @xmath13 . the coherence measure defined by the @xmath0 norm
is based on the minimal distances of @xmath13 to the set of incoherent states @xmath10 , @xmath34 with @xmath35 being the @xmath0 norm , @xmath36 .
the upper bound is attained for the maximally coherent state @xmath37 @xcite .
the relative entropy of coherence is defined as @xmath38 where @xmath39 is the quantum relative entropy , @xmath40 is the von neumann entropy , and @xmath41 is the diagonal part of @xmath13 .
the coherence measure defined by the relative entropy is based on the minimal distances of @xmath13 to @xmath10 , @xmath34 with @xmath35 being the the relative entropy , @xmath42 .
the coherence measure expressed by eq.([entropy ] ) can also be derived based on the notion of distillation process , i.e. , from the process that distills the maximally coherent states @xmath43 from @xmath13 under incoherent operations .
it is the maximal rate at which the maximally coherent states can be obtained from the given state @xcite .
the coherence of formation , is defined as @xmath44 where @xmath45 is any decomposition of @xmath13 into pure states @xmath46 with @xmath47 .
expression ( [ cost ] ) , as the coherence of formation , was first given in ref .
@xcite , and was proved to be a coherence measure , i.e. , satisfying the four postulates , in ref .
@xcite . the coherence measure expressed by eq.([cost ] ) can also be derived based on the notion of coherence cost , i.e. , from the process that prepares the state @xmath13 by consuming maximally coherent states @xmath43 under incoherent operations .
it is the minimal rate at which the maximally coherent states have to be consumed to prepare the given state under incoherent operations @xcite .
before proceeding further , we first specify the notion of ordering states with coherence measures . for a coherence measure @xmath19 , the coherence of a state , @xmath48 , is always a non - negative number , and therefore according to the numbers , all the states can be ranked in numerical order . for two coherence measures @xmath1 and @xmath2 , we say that they give the same ordering if the following relations are satisfied for all the states @xmath5 and @xmath6 , @xmath49 otherwise , we say that they do not give the same ordering .
relation ( [ order ] ) means that @xmath3 if and only if @xmath4 for all the states @xmath5 and @xmath6 , or equally , @xmath50 if and only if @xmath51 . similar notions of ordering states have been widely used in entanglement measures @xcite and other quantum - correlation measures @xcite .
we will show that the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering of states , i.e. , there are the states for which the two measures give a different ordering . to this end , we need to find the states for which @xmath52 and @xmath53 , or @xmath54 and @xmath55 .
hereafter , we refer to the pair of states for which the two coherence measures give a different ordering as the ordering - different pair for convenience .
we first consider @xmath56-dimensional quantum systems .
the density operators of a @xmath56-dimensional system can be generally written as @xmath57 with @xmath58 .
@xmath59 can be further expressed as @xmath60 , where @xmath61 with @xmath62 and @xmath63 , and @xmath64 is a diagonal and unitary matrix with @xmath65 .
it means that @xmath59 and @xmath66 can be transformed into each other by an incoherent operation , which further implies that they correspond to the same coherence value for all the coherence measures due to postulate ( c2 ) .
therefore , we only need to consider the family of density operators with the form @xmath66 . by substituting eq .
( [ rho2 ] ) into eqs .
( [ l ] ) and ( [ entropy ] ) , we obtain the @xmath0 norm of coherence , @xmath67 and the relative entropy of coherence , @xmath68 where @xmath69 is the binary shannon entropy function .
@xmath70 is only dependent on @xmath71 , while @xmath72 is dependent on both @xmath71 and @xmath73 .
since @xmath72 is an even function of @xmath73 , we can further restrict our discussion to the range @xmath74 without loss of generality .
specially , when @xmath13 is a pure state , there is @xmath75 due to @xmath76 . in this case
, @xmath77 is an increasing function of @xmath71 just like @xmath78 , and therefore the @xmath0 norm of coherence and the relative entropy of coherence always give the same ordering for all the @xmath56-dimensional pure states .
there is no ordering - different pair in this case . to find the ordering - different pairs for which the @xmath0 norm of coherence and the relative entropy of coherence give a different ordering
, we further analyze the expression of @xmath79 without the restriction of pure states . from eq .
( [ crtz ] ) , we have @xmath80 and @xmath81 for @xmath82 , which means @xmath79 is an increasing function of @xmath71 as well as @xmath73 . with these knowledge , we may now try to find a pair of states @xmath5 and @xmath6 , such that the ordering of @xmath83 and @xmath84 are different from that of @xmath85 and @xmath86 . for this
, we let @xmath87 without loss of generality , we assume that @xmath88 , i.e. , @xmath89 .
the question then turns to find @xmath90 and @xmath91 such that @xmath53 , i.e. , @xmath92 for a given pair of @xmath88 , @xmath90 and @xmath91 satisfying eq .
( [ inequality ] ) may not exist in general .
for instance , at @xmath93 and @xmath94 , there do not exist @xmath90 and @xmath91 fulfilling eq .
( [ inequality ] ) . since @xmath79 is an increasing function of @xmath71 and @xmath73 , the maximal value of the left side of eq .
( [ inequality ] ) is appearing at @xmath95 , and the minimal value of the right side is appearing at @xmath96 .
hence , we obtain the necessary and sufficient condition of @xmath97 and @xmath98 , @xmath99 under which @xmath90 and @xmath91 exist .
it means that @xmath90 and @xmath91 satisfying the inequality ( [ inequality ] ) exist if and only if @xmath97 and @xmath98 ( @xmath100 ) satisfy the relation ( [ tcondition ] ) . and
@xmath91 satisfying @xmath101 . here , we have taken @xmath102 and @xmath103 .
all the pairs of states @xmath5 and @xmath6 defined by @xmath90 and @xmath91 in the domain for @xmath104 fulfill @xmath105 and @xmath53 .
, width=317 ] therefore , to find @xmath5 and @xmath6 that satisfy both @xmath105 and @xmath53 , one may first choose @xmath97 and @xmath98 ( @xmath106 ) with the aid of eq .
( [ tcondition ] ) and then find @xmath90 and @xmath91 ( @xmath107 ) by using eq .
( [ inequality ] ) .
for example , if we take @xmath102 and @xmath103 , which satisfy eq.([tcondition ] ) , all the @xmath90 and @xmath91 in the domain with @xmath101 are given in fig .
1 . clearly , there are infinitely many pairs of states satisfying @xmath108 .
an explicit expression of @xmath5 and @xmath6 reads @xmath109 for which , @xmath110 but @xmath111 .
+ we now consider high dimensional quantum systems .
the above discussion is only on @xmath56-dimensional states .
it shows that there are the pairs of states for which the @xmath0 norm and the relative entropy of coherence do not give the same ordering .
this conclusion can be easily extended to the case of high dimensional density operators .
in fact , any two @xmath112-dimensional states defined by @xmath113 where @xmath114 and @xmath115 are @xmath116-dimensional incoherent states , necessarily satisfy @xmath117 and @xmath118 as long as @xmath5 and @xmath6 do .
this is because @xmath119 and @xmath120 , @xmath121 . it is worth noting that although the @xmath0 norm of coherence and the relative entropy of coherence give the same ordering for all the @xmath56-dimensional pure states , it is not true for the @xmath122-dimensional case
there are pairs of high dimensional pure states for which the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering .
for example , in the @xmath123-dimensional case , @xmath124 is a pair of states for which the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering , since @xmath125 but @xmath126 . with the aid of the @xmath123-dimensional case
, we can find pairs of higher dimensional pure states for which the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering . indeed , any pairs of @xmath127-dimensional pure states defined by @xmath128 with @xmath129 and @xmath130 necessarily satisfy @xmath131 and @xmath132 as long as @xmath133 and @xmath134 do ( see the appendix for the proof ) .
after having shown that the @xmath0 norm of coherence and the relative entropy of coherence do not give the same ordering of states , we now extend our discussion to one more coherence measure , the coherence of formation . we wonder whether the @xmath0 norm of coherence and the coherence of formation , as well as the relative entropy of coherence and the coherence of formation , give the same ordering of states .
we will briefly discuss the issue in this subsection .
in fact , with the aid of the above calculations , it is very easy to confirm that there are infinitely many ordering - different pairs for these coherence measures too .
we first address the @xmath0 norm of coherence and the coherence of formation .
( [ entropy ] ) and ( [ cost ] ) , we have @xmath135 for all pure states @xmath136 , which implies that all the pure states that satisfy @xmath137 and @xmath138 must fulfill @xmath137 and @xmath139 .
therefore , all pairs of the @xmath123-dimensional states defined by eq .
( [ psi12 ] ) and the @xmath127-dimensional states defined by eq .
( [ lyt ] ) are the ordering - different pairs for the @xmath0 norm of coherence and the coherence of formation too . note that there is no ordering - different pair in the @xmath56-dimensional case , since both @xmath140 and @xmath141 @xcite are increasing functions of @xmath71 and therefore they give the same ordering for all the @xmath56-dimensional states .
we now address the relative entropy of coherence and the coherence of formation .
since the coherence of formation and the @xmath0 norm of coherence give the same ordering for all the @xmath56-dimensional states as shown above , i.e. , @xmath142 if and only if @xmath52 , all the @xmath56-dimensional states that satisfy @xmath105 and @xmath53 must fulfill @xmath143 and @xmath53 . therefore , all the @xmath56-dimensional ordering - different pairs for @xmath144 and @xmath145 are the ordering - different pairs for @xmath146 and @xmath145 too .
as pointed out in the case of the @xmath0 norm of coherence and the relative entropy of coherence , there are infinitely many such pairs of states , and one may obtain them with the aid of eqs .
( [ inequality ] ) and ( [ tcondition ] ) .
an explicit expression of @xmath5 and @xmath6 is given in eq .
( [ ccl ] ) , for which @xmath147 and @xmath111 . besides
, just like the result that any two entanglement measures coinciding on pure states are either identical or give a different ordering @xcite , we can also obtain a similar result for coherence measures .
let us consider two coherence measures @xmath1 and @xmath2 that coincide on all pure states .
note that for any state @xmath13 and any sufficiently small @xmath148 , there must exist pure states @xmath149 and @xmath150 such that @xmath151 .
if @xmath1 and @xmath2 give the same ordering on all states , we have @xmath152 .
as @xmath1 and @xmath2 coincide on pure states , we get @xmath153 , which further leads to @xmath154 . taking @xmath155 to zero ,
we finally obtain @xmath156 , which implies that any two coherence measures that coincide on pure states are either identical or give a different ordering .
this result can be used to determine whether some coherence measures , for instance @xmath145 and @xmath146 , give the same ordering .
however , it is not applicable to many other coherence measures , for example @xmath144 and @xmath145 , which do not coincide on pure states .
the issue of ordering states with coherence measures is first discussed in this paper .
we have shown that the two well - known coherence measures , the @xmath0 norm of coherence and the relative entropy of coherence , do not give the same ordering of states .
there are infinitely many ordering - different pairs for which the @xmath0 norm of coherence and the relative entropy of coherence give a different ordering .
detailed calculations show that the ordering - different states include the @xmath56-dimensional mixed states , the @xmath157-dimensional mixed states and the @xmath157-dimensional pure states , but exclude 2-dimensional pure states since all the coherence measures give the same ordering for 2-dimensional pure states .
our analysis can be extended to other coherence measures , and as an illustration of the extension we have shown that the @xmath0 norm of coherence and the formation of coherence , as well as the relative entropy of coherence and the coherence of formation , do not give the same ordering too .
our results indicate that at least each pair of the three coherence measures , the @xmath0 norm of coherence , the relative entropy of coherence and the coherence of formation , give a different ordering of states , although it remains an open question to examine whether all other coherence measures give a different ordering of states . noting that any two states in an ordering - different pair can not be transformed into each other by incoherent operations
, our result may be helpful in determining what kinds of states can not be transformed into each other by incoherent operations .
we show that any two @xmath127-dimensional pure states defined by eq .
( [ lyt ] ) necessarily satisfy @xmath131 and @xmath132 as long as @xmath158 and @xmath159 do . to this end , we only need to prove the following theorem : if the two @xmath160-dimensional states expressed as @xmath161 satisfy @xmath162 and @xmath163 , then the two @xmath8-dimensional states defined by @xmath164 with @xmath165 and @xmath166 , satisfy @xmath131 and @xmath132 . by the definition of the relative entropy of coherence , we have @xmath172 and @xmath173 where @xmath69 is the binary shannon entropy function , and similarly , @xmath174 and @xmath175 eqs .
( [ eq : crphid-3 ] ) and ( [ al ] ) show that @xmath132 if @xmath176 .
this completes the proof of the theorem . with this theorem , it is easy to obtain the conclusion related to eq .
( [ lyt ] ) .
99 m. o. scully and m. s. zubairy , _ quantum optics _ ( canbrudge university press , cambridge , 1997 ) .
m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _
( canbrudge university press , cambridge , 2000 ) .
n. lambert , y .-
n . chen , y .- c .
cheng , c .- m .
li , g .- y .
chen , and f. nori , _ nat .
phys . _ * 9 * , 10 ( 2013 )
. s. f. huelga and m. b. plenio , _ nat .
* 10 * , 621 ( 2014 ) . c. a. rodrguez - rosario , t. frauenheim , and a. aspuru - guzik , arxiv:1308.1245 .
j. berg , _ phys .
lett . _ * 113 * , 150402 ( 2014 ) . m. lostaglio , k. korzekwa , d. jennings , and t. rudolph , _ phys .
x _ * 5 * , 021001 ( 2015 ) .
t. baumgratz , m. cramer , and m. b. plenio , _ phys .
lett . _ * 113 * , 140401 ( 2014 ) .
j. berg , arxiv:0612146 .
b. toloui , g. gour , and b. c. sanders , _ phys .
a _ * 84 * , 022322 ( 2011 ) . i. marvian and r. w. spekkens , _ new j. phys . _
* 15 * , 033001 ( 2013 ) .
f. levi and f. mintert , _ new j. phys . _
* 16 * , 033007 ( 2014 ) .
a. monras , a. chcinska , and a. ekert , _ new j. phys . _
* 16 * , 063041 ( 2014 ) .
w. vogel and j. sperling , _ phys .
a _ * 89 * , 052302 ( 2014 ) . c. smyth and g. d. scholes , _ phys .
a _ * 90 * , 032312 ( 2014 ) .
d. girolami , _ phys .
lett . _ * 113 * , 170401 ( 2014 ) .
t. r. bromley , m. cianciaruso , and g. adesso , _ phys .
lett . _ * 114 * , 210401 ( 2015 ) .
a. streltsov , u. singh , h. s. dhar , m. n. bera , and g. adesso , _ phys .
_ * 115 * , 020403 ( 2015 ) .
shao , z. xi , h. fan , and y. li , _ phys .
a _ * 91 * , 042120 ( 2015 ) .
d. p. pires , l. c. celeri , and d. o. soares - pinto , _ phys .
a _ * 91 * , 042330 ( 2015 ) .
u. singh , m. n. bera , h. s. dhar , and a. k. pati , _ phys .
rev . a _ * 91 * , 052115 ( 2015 ) .
s. du , z. bai and y. guo , _ phys .
a _ * 91 * , 052120 ( 2015 ) .
x. yuan , h. zhou , z. cao , and x. ma , _ phys .
a _ * 92 * , 022124 ( 2015 ) .
y. yao , x. xiao , l. ge , and c. p. sun , _ phys . rev .
a _ * 92 * , 022112 ( 2015 ) .
a. mani and v. karimipour , _ phys .
rev . a _ * 92 * , 032331 ( 2015 ) .
s. cheng and m. j. w. hall , _ phys . rev .
a _ * 92 * , 042101 ( 2015 ) . m. n. bera , t. qureshi , m. a. siddiqui , and a. k. pati , _ phys .
rev . a _ * 92 * , 012118 ( 2015 ) .
s. du and z. f bai , _ ann .
_ * 359 * , 136 ( 2015 ) . z. xi , y. li , and h. fan , _ sci .
rep . _ * 5 * , 10922 ( 2015 ) .
a. winter and d. yang , arxiv:1506.07975 .
j. eisert and m. b. plenio , _ j. mod
. opt . _ * 46 * , 145 ( 1999 ) .
k. yczkowski , _ phys . rev
. a _ * 60 * , 3496 ( 1999 ) .
s. virmani and m. b. plenio , _ phys .
a _ * 268 * , 31 ( 2000 ) .
k. yczkowski and i. bengtsson , _ ann .
_ * 295 * , 115 ( 2002 ) .
t. c. wei , k. nemoto , p. m. goldbart , p. g. kwiat , w. j. munro , and f. verstraete , _ phys . rev .
a _ * 67 * , 022110 ( 2003 ) .
t. c. wei and p. m. goldbart , _ phys .
a _ * 68 * , 042307 ( 2003 ) .
a. miranowicz , _
a _ * 37 * , 7909 ( 2004 ) .
a. miranowicz , _ phys .
a _ * 327 * , 272 ( 2004 ) . f. verstraete , d. porras , and j. i. cirac , _ phys .
lett . _ * 93 * , 227205 ( 2004 ) .
a. miranowicz and a. grudka , _ phys .
a _ * 70 * , 032326 ( 2004 ) .
a. miranowicz and a. grudka , _ j. opt .
b _ * 6 * , 542 ( 2004 ) . m. ziman and v. bzek , _ phys .
rev . a _ * 73 * , 012312 ( 2006 ) .
y. kinoshita , r. namiki , t. yamamoto , m. koashi , and n. imoto , _ phys .
a _ * 75 * , 032307 ( 2007 ) .
r. horodecki , p. horodecki , m. horodecki , and k. horodecki , _ rev .
phys . _ * 81 * , 865 ( 2009 ) .
a. sen de and u. sen , _ j. mod .
opt . _ * 50 * , 981 ( 2003 ) .
s. luo , _ phys .
rev . a _ * 77 * , 042303 ( 2008 ) .
y. yeo , j .- h . an , and c. h. oh , _ phys .
rev . a _ * 82 * , 032340 ( 2010 ) .
m. ali , a. r. p. rau , and g. alber , _ phys .
rev . a _ * 81 * , 042105 ( 2010 ) .
a. al - qasimi and d. f. v. james , _ phys .
a _ * 83 * , 032101 ( 2011 ) .
d. girolami and g. adesso , _ phys .
a _ * 83 * , 052108 ( 2011 ) .
d. girolami , m. paternostro , and g. adesso , _ j. phys .
a _ * 44 * , 352002 ( 2011 ) . m. d. lang , c. m. caves , and a. shaji , _ int .
j. quantum inf .
_ * 09 * , 1553 ( 2011 )
. m. okrasa and z. walczak , _ europhys .
lett . _ * 98 * , 40003 ( 2012 ) .
k. modi , a. brodutch , h. cable , t. paterek , and v. vedral , _ rev .
phys . _ * 84 * , 1655 ( 2012 ) .
|
the quantification of quantum coherence has attracted a growing attention , and based on various physical contexts , several coherence measures have been put forward . an interesting question is whether these coherence measures give the same ordering when they are used to quantify the coherence of quantum states . in this paper
, we consider the two well - known coherence measures , the @xmath0 norm of coherence and the relative entropy of coherence , to show that there are the states for which the two measures give a different ordering .
our analysis can be extended to other coherence measures , and as an illustration of the extension we further consider the formation of coherence to show that the @xmath0 norm of coherence and the formation of coherence , as well as the relative entropy of coherence and the coherence of formation , do not give the same ordering too .
|
rabi oscillations are periodical transitions of a two - level quantum system between its stationary states under the action of an oscillatory driving field , see e.g. @xcite .
the phenomenon was theoretically predicted by rabi on nuclear spins in radio - frequency magnetic field @xcite and was firstly observed by torrey @xcite .
afterwards , rabi oscillations were discovered in various physical systems , such as electromagnetically driven atoms@xcite ( including the case of rydberg atomic states @xcite ) , semiconductor quantum dots ( qds ) @xcite , and different types of solid - state qubits ( superconducting charge qubits based on josephson junctions @xcite , spin - qubits @xcite , semiconductor charge - qubits @xcite ) . in real physical systems the ideal picture @xcite of rabi effect can be essentially modified by additional features , such as the time - domain modulation of the field - matter coupling constant @xcite , the phonon - induced dephasing @xcite and the local - field effects @xcite .
new phenomena appear in rabi oscillators with broken inversion symmetry @xcite and in systems of two coupled rabi oscillators @xcite . in spatially extensive samples with a large number of oscillators
the propagation effects come into play . as a result ,
the mechanism responsible for rabi oscillations causes also a number of nonstationary coherent optical phenomena , such as optical nutation , photon echo , self - induced transparency , etc @xcite . in low - dimensional systems
the propagation effects also take place . for example , the numerical modeling of the coherent intersubband rabi oscillations in a sample comprising 80 algaas / gaas quantum wells @xcite shows that the population dynamics depends on the quantum well position in the series .
this result demonstrates strong radiative coupling between wells and , more generally , significant difference in the picture of rabi effect for single and multiple oscillators .
another aspects of such a difference , namely , the effects of quantum interference and correlations between photons in multiatom fluorescence , are demonstrated in @xcite ( on the example of atomic chains in nanofibers ) . from practical point of view , the effect of rabi oscillations is a key ingredient for realization of binary logic and optical control in quantum informatics and quantum computing .
the theoretical analysis of rabi oscillations is highly diversified both in form and content .
the common feature is the impossibility of consideration of electromagnetic field influence as a small perturbation .
different ways of description are used in the analysis of the model problem of rabi oscillations in a single two - level atom @xcite .
one such a way , the probability amplitude method , consists in solving of the schrdinger equation for wavefunction @xmath0 , which is the superposition of various atom - photon states .
the second one is the heisenberg operator method which is based on the analysis of the photonic and atomic operators time evolution . and finally , the third way is the unitary time - evolution operator technique .
as is demonstrated in @xcite , all these methods lead to identical solutions ; the choice of the concrete one is determined by the convenience considerations .
taking into account the quantum nature of the electromagnetic field has the special significance in the analysis of rabi oscillations in complex systems .
two different cases can be marked out .
the first could be called quasiclassical one @xcite . in this case external field
has classical nature .
however , the true field is supposed to be concordant with the quantum motions of the particles .
so , the field should contain the contribution of the induced polarization , which has a quantum nature .
another case is really quantum and takes into consideration the photon structure of the electromagnetic field @xcite . during the process of energy - level transitions in the atom the photon structure of the field
is also transformed , therefore the atom - photon dynamics should be considered self - consistently . in this paper
we build for the first time a theoretical model of a distributed system of coupled rabi oscillators and predict a new physical effect : propagation of rabi oscillations in space in the form of traveling waves and wave packets .
quantum oscillators in the system that exhibit rabi - wave propagation interact strongly enough to stare and exchange the e - h pair excitation between light emission and light absorption acts . often in quantum optics
an `` all - matter '' picture is employed , where the dynamics of electromagnetic field is integrated out , for example in the optical bloch equations @xcite . in @xcite
the light - matter interaction is treated in an `` all - light '' picture ( lippmann - schwinger equation approach ) . in our model the quantum nature of light
is also fully accounted . as the theoretical approach the _ probability amplitude method _ ,
generalized for the case of 1d - chain , is convenient .
the paper is organized as follows . in sec.[sec : theory ] the model is formulated and equations of electron - photon dynamics both for discrete chain and continuous limit are obtained . in sec.[sec : trav_w ] the plane rabi waves are considered ( their dispersion equations , eigenmode structure , dispersion curves properties ) .
sec.[sec : wave_pack ] is devoted to the rabi wavepackets investigation , their spatial - temporal dynamics is analyzed , experimental observability of rabi waves is discussed . in sec.[sec : class_lim ] the rabi waves in the case of classic light are examined and distinctive features of semiclassical consideration are treated of . in sec.[sec : local_field ] the role of the local - field effects and conditions of this effects negligibility are discussed . in sec.[sec : correlators ] the space - time structure of electron - electron and electron - photon correlators is considered .
the main results of the work are formulated in sec.[sec : conclusion ] .
schematic picture of the qd - chain interacting with single - mode electromagnetic field ( a ) and the energy levels diagram for @xmath1-th qd ( b).,scaledwidth=45.0% ] consider an infinite periodical one - dimensional chain of identical qds , containing an electron in the size - quantized conduction band .
it is assumed that in each @xmath1-th qd there are at least two one - electron orbital states , ground @xmath2 and excited @xmath3 , with transition frequency @xmath4 .
neighboring qds are coupled via electron tunneling @xcite , i.e. the electron can go from state @xmath3 to state @xmath5 and from @xmath2 to state @xmath6 .
transitions between ground state and excited state belonging to different qds are neglected : @xmath7 .
let the qd chain be exposed to a plane wave of quantum light , which is incident on the chain at an angle @xmath8 , @xmath9 , where @xmath10 , @xmath11 are photon creation and annihilation operators , @xmath12 , @xmath13 is the normalizing volume , @xmath14 is the unit polarization vector , wavenumber @xmath15 .
the dependence of the field on transverse coordinates is negligibly small because of electrical smallness of qds .
the case of oblique propagation is of interest due to the possibility of wavenumber variation with variation of incidence angle @xmath8 at constant frequency @xmath16 .
then the hamiltonian of the system in the rotating wave approximation @xcite reads @xmath17 where @xmath18 is hamiltonian of qd - chain in the absence of electron tunneling and qd - field interaction , @xmath19 , @xmath20 is the hamiltonian of the free electromagnetic field@xcite .
the component @xmath21 describes qd - field interaction , where @xmath22 is interaction constant , @xmath23 is the qd dipole moment .
note that all transition dipole moments in the chain are assumed to be real values and to have fixed orientation .
the operators @xmath24 and @xmath25 are the transition operators from ground to excited states and vice versa for @xmath1-th qd .
the term @xmath26 takes into account interdot electron tunneling ; @xmath27 are the electron tunneling frequencies for the excited ( @xmath28 ) and ground ( @xmath29 ) states of the qds .
the term @xmath30 corresponds to the local - field effects originated from the dipole - dipole electron - hole intradot interaction @xcite ; in the mean - field approximation this term is given by @xmath31 here @xmath32 is the depolarization tensor ( see eq . (
18 ) in ref .
@xcite ) , @xmath33 is the volume of qd .
the state vector of the `` qd - chain+light '' system may be represented in terms of the eigenstates of isolated qds as @xmath34 here @xmath35 , @xmath36 , where @xmath37 is the light fock state with @xmath38 photons , @xmath39 , @xmath40 are the probability amplitudes . in the interaction picture
the evolution of the system is described by the nonstationary schrdinger equation @xmath41 , where the interaction hamiltonian has the form @xmath42 .
this schrdinger equation leads to the following equations for the probability amplitudes @xmath43 where @xmath44 is the local - field induced depolarization shift @xcite . obtaining this equations
we have taken into account that the interaction can cause the transitions between the states @xmath45 , @xmath46 only .
as it is seen from eqs. , two competitive mechanisms manifest themselves in the light - qd - chain coupling : the local - field induced nonlinearity and the dispersion spreading due to the tunneling . to this point
we have taken no account of the decay processes inside the qds . to take this processes into account one
should allow for processes of interaction of e - h pair in qd with phonon bath .
such interaction can be described by conception of quantum trajectories @xcite of e - h pair , which are the superpositions of deterministic evolutions and random jumps under the action of lindblad operators . in this case
the hamiltonian should be replaced by the efficient non - hermitian hamiltonian @xmath47 , where @xmath48 is lindblad operator for @xmath1-th qd .
suppose that the density of phonon states of the bath has the form of lorentz line with frequency @xmath4 and width @xmath49 .
if the bath is weakly coupled with qd then @xmath50 ( see refs @xcite ) .
it means that the decay can be taken into account by substitution @xmath51 into eq.[system_diskr_a ] and @xmath52 into eq.[system_diskr_b ] .
let us restrict here consideration to linear regime of the carrier motion and omit the terms @xmath53 in . in the sec.[sec : local_field ] we shall analyze the limitations imposed by omission of nonlinear terms . by this means , omitting the terms @xmath53 , we arrive at the set of difference - differential equations for vectors @xmath54 $ ] as follows : @xmath55 \bm{\psi}_{p , n}\\ + i\hat{\xi}\left(\bm{\psi}_{p-1,n}+\bm{\psi}_{p+1,n}\right ) , \end{split}\ ] ] where @xmath56 denotes 2d unit operator , @xmath57 $ ] , @xmath58/2 $ ] .
coefficients in are expressed in terms of pauli matrices @xmath59 acting on vectors @xmath60 .
equations are the recurrent ordinary differential equations . in some cases
it is more convenient to transform them into the system of partial differential equations . to do this
we should turn to the continuous limit , making the standard substitutions ( see , for example , @xcite ) @xmath61 , @xmath62 , @xmath63 , @xmath64 .
then the system leads to @xmath65\bm{\psi}_n \\ + ia^2\hat{\xi}\partial^2_x\bm{\psi}_n . \end{split}\ ] ] formula represents the system of partial differential equations with the variable coefficient @xmath66 .
this is the basic system for most of our further calculations .
consider elementary solution of system in the form of damping traveling waves : @xmath67 , @xmath68 , where @xmath69 is a given wave number and @xmath70 is the eigenfrequency to be found , @xmath71 , @xmath72 are the unknown constant coefficients . substituting them into equations and omitting common factors , we obtain @xmath73 where @xmath74 , @xmath75 , % \\ % \label{theta_2 } % \vartheta_2(h)=\xi_2[2-a^2(h - k/2)^2].\end{aligned}\ ] ] system has a nontrivial solution if its determinant is equal to zero , i.e. we obtain quadratic equation which connects @xmath70 and @xmath69 .
solving the equation with respect to @xmath70 , we determine the eigenfrequencies of system as @xmath76.\ ] ] here @xmath77 it follows from that there are two eigenmodes for each value of photon number @xmath38 , namely @xmath78 where @xmath79 is the number of the eigenmode , and coefficients @xmath80 , @xmath81 are defined by the expressions @xmath82 and @xmath83 where @xmath84 are normalizing constants , @xmath85 , the values @xmath86 are expressed by the eqs .
, , respectively .
each mode corresponds to the electron - photon entangled state , which partial amplitudes oscillate both in time and space .
the expression with or describes the _ state of radiation - dressed qd - chain _ since the entangled state includes field states with different photon numbers ( @xmath38 and @xmath87 ) .
this states are the generalization of the single atom dressed states @xcite for the case of distributed system .
the qualitative distinction of this case is the space - time modulation of dressing parameter : it propagates along the qd - chain according to the traveling wave law @xmath88 $ ] . entanglement and dressing
are caused by the interaction of light with qd chain and vanish in the limit of @xmath89 . in that case , describes qd - chain electron - hole pairs in equilibrium and inverse states , respectively .
spatial oscillations of the partial amplitudes are due to qd - coupling .
they vanish in the limit of @xmath90 .
generally , both the modes are excited simultaneously .
however , any of them can be excited separately by a proper choice of initial conditions . similarly to the other coherent excitations in condensed matter , rabi waves and introduce a new family of quasi - particles ( we name them _ rabitons _ ) .
one can apply to them the standard secondary quantization technique . similarly to the effect of the self - induced transparency , the rabi wave propagation can be interpreted as the motion of a precessing pseudo dipole @xcite .
however , the coherence mechanisms in these two cases are principally different : in the rabi wave the coherence is settled by the dispersion relation while in the case of the self - induced transparency it has solitonic character .
note that the eigenmodes ( [ mode1 ] ) and ( [ mode2 ] ) each comprise traveling waves with _ different _ wave numbers @xmath91 .
physically , it means that the rabi wave propagates in an effective periodically inhomogeneous medium formed by spatially oscillating ( with period @xmath92 ) electric field
. therefore , the diffraction appears in the system . in the limit @xmath93 ,
the medium turns homogeneous and the diffraction effect vanishes .
reflections of rabi waves and their mutual transformations at the field inhomogeneities become possible .
thus one obtain a unique possibility to control the processes of the reflection and transmission of rabi waves by varying the spatial structure of the light .
note , that for the waves under consideration _ the reciprocity conditions break down _ : @xmath94 .
this is due to the presence of preferential direction , which is determined by the direction of photon mode propagation along the qd - chain ( sign of the value @xmath95 ) .
the form of the dispersion characteristics for typical values of parameters is presented on fig.[fig_disp_char ] .
dispersion curves for rabi waves for different parameters values .
a ) @xmath96 , @xmath97 , @xmath98 , @xmath99 ; b ) @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath104 ; c ) @xmath100 , @xmath105 , @xmath106 , @xmath98 , @xmath99 . ,
title="fig:",scaledwidth=48.0% ] dispersion curves for rabi waves for different parameters values .
a ) @xmath96 , @xmath97 , @xmath98 , @xmath99 ; b ) @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath104 ; c ) @xmath100 , @xmath105 , @xmath106 , @xmath98 , @xmath99 . ,
title="fig:",scaledwidth=48.0% ] dispersion curves for rabi waves for different parameters values .
a ) @xmath96 , @xmath97 , @xmath98 , @xmath99 ; b ) @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath104 ; c ) @xmath100 , @xmath105 , @xmath106 , @xmath98 , @xmath99 .
, title="fig:",scaledwidth=48.0% ] to interpret them one should note that in the limit @xmath107 both of branches have cross points , two in general case and one at @xmath108 .
modes under examination are characterized by continuous spectrum ( the value @xmath69 varies continuously ) .
the values @xmath109 are real for all real @xmath69 .
therefore , the system being investigated is stable @xcite .
dispersion equation for - can be solved with respect to wave number @xmath110 for given real - valued frequency @xmath70 . in this case for every mode there are such critical values @xmath111 that for @xmath112 the values @xmath113 become complex .
physically , it means opacity of the qd - chain for rabi waves with such frequencies@xcite . for @xmath114
the qd - chain is opaque only for one of the modes and for @xmath112 it is non - transparent for both of them .
one can choose the system parameters in such a way that in some range of @xmath70 the all real solutions @xmath115 are either negative or positive ( as an example see fig.[fig_disp_char]a ) .
this implies one - way opacity of the system ( rabi waves could propagate in one direction only ) , which is caused by nonreciprocity of the qd - chain .
the critical frequencies are determined from the condition @xmath116 , where @xmath117 it is significant , that critical values @xmath111 depend on photon number @xmath38 ( see fig.[critical_fr ] ) .
this gives an important result , namely : _ the conditions of transparency of the qd - chain are different for rabi waves with different photon numbers_. note , that @xmath118 increases and @xmath119 decreases with increasing of @xmath38 .
dependence of critical frequencies of rabi waves on the photon number , for the input parameters as follows : @xmath100 , @xmath120 , @xmath121 , @xmath98 .
, scaledwidth=48.0% ] usually both rabi waves demonstrate the normal dispersion and propagate in the _
same direction_. but for a special choice of the parameters the _ abnormal dispersion region _ for one of the modes appears . for this mode phase and group velocities
are oppositely directed ( see fig.[fig_disp_char]a ) .
the dispersion characteristics have special behavior for absolutely nontransparent ground - state barrier : @xmath122 ( fig.[fig_disp_char]c ) . in this case
the group velocity @xmath123 of one of the waves is very small for all @xmath69 , except for the narrow vicinities of the cross - points . due to this fact the modes exchange their places when passing the cross points ( e.g. on fig.[fig_disp_char]c @xmath124 for @xmath125-mode between cross points @xmath126 and for @xmath127-mode at all other values of @xmath69 ) .
thus , in current case for a wide range of @xmath69 one of the modes practically does not transfer the energy along the qd - chain .
to find the general solution of the system let us introduce the new variable @xmath128 .
it leads to the system of partial differential equations with coefficients independent of qd - position @xmath129 : @xmath130&\bm{\phi}_n\\ + a^2k\hat{\sigma}_z\hat{\xi}\partial_x\bm{\phi}_n - ia^2&\hat{\xi}\partial_x^2\bm{\phi}_n=0 , \end{split}\ ] ] where @xmath131 .
this system can be solved by using the fourier transform with respect to @xmath129 : @xmath132 substituting eq . into eq .
, we obtain the system of ordinary differential equations @xmath133 where @xmath134/2 $ ] .
the system can be easily integrated , the solution is given by @xmath135 where @xmath136t}e^{i\hat{\sigma}_z \delta t } e^{i[\bm{m}\hat{\bm{\sigma}}\omega_n(h)]t},\ ] ] @xmath137 , @xmath138 is the unit vector of the form @xmath139 . the expression can be written in matrix form as follows : @xmath140,\ ] ] @xmath141 , \end{aligned}\ ] ] @xmath142 and @xmath143 are determined by and , correspondingly , and @xmath144 can be obtained from initial conditions by the inverse fourier transform : @xmath145 thus , for the vector of probability amplitudes @xmath146 we have the following expression : @xmath147t/2\ } } } dh .
\end{split}\ ] ] expression is the central result of this section .
it allows one to characterize such observables as inversion , as well as the exciton - exciton and exciton - photon correlators for different initial states of light and qds . in this section
we will analyze the propagation of rabi waves with spatially localized initial state .
the qd - chain issupposed to be initially either in the ground state ( @xmath148 for all @xmath38 ) , or in the excited state ( @xmath149 for all @xmath38 ) , or in the mixed state ( arbitrary superposition of ground and excited states ) .
we describe the spatial - temporal dynamics of rabi oscillations by the spatial density of the inversion ( the inversion per single qd ) @xmath150.\ ] ] space - averaged temporal evolution of rabi oscillations is characterized by the integral inversion @xmath151 let us approximate the initial spatial distributions by the gaussian beams @xmath152 $ ] , where @xmath153 are normalization constants , @xmath154 are the positions of the beams and their widths , respectively ( the indexes @xmath155 refer to excited and ground state of the electron in the qd - chain , respectively ) . the rabi wavepacket evolution can be qualitatively illustrated with the help of the dispersionless approximation for eqs .. according to we obtain for @xmath156 $ ] : @xmath157 where @xmath158 is an arbitrary photonic distribution .
one can see from that the spatial spectrum of @xmath159 is given by two gaussian peaks with peak widths @xmath160 and positions of the peak centers @xmath161 , respectively .
if @xmath162 are sufficiently large , the largest contribution to the integration in comes from the narrow vicinities of points @xmath163 .
assuming @xmath164 and @xmath165 ( the dispersion effects neglecting ) and doing standard calculations , we obtain from eq .
: @xmath166e^{-ikx/2}\right.\\ & + \left.\left[b_{n+1}(x+v_1^+t,0)\eta_1 ^ -e^{i\theta_1^+t}+b_{n+1}(x+v_1 ^ -t,0)\eta_1^+e^{i\theta_1 ^ -t}\right]e^{ikx/2}\right\},\\ \label{appr_sol_b } \nonumber b_{n+1}(x , t)&=e^{-i(kx-\omega t)/2}e^{-\lambda t } \left\{\left[a_{n}(x+v_2^+t,0)\eta_2 ^ -e^{i\theta_2^+t}+a_{n}(x+v_2 ^ -t,0)\eta_2^+e^{i\theta_2 ^ -t}\right]e^{-ikx/2}\right.\\ & + \left.\left[b_{n+1}(x+v_1^+t,0)\zeta_1 ^ -e^{i\theta_1^+t}+b_{n+1}(x+v_1 ^ -t,0)\zeta_1^+e^{i\theta_1 ^ -t}\right]e^{ikx/2}\right\},\end{aligned}\ ] ] where the velocities @xmath167 are defined as @xmath168 @xmath169 and @xmath170 and @xmath171 , @xmath172 are introduced to denote the frequency shifts and amplitude factors , respectively : @xmath173-v^{\pm}(h_{1,2}^{(0 ) } ) , \ \\
\label{zeta } \zeta_{1,2}^{\pm}&=\frac{\omega_n(h_{1,2}^{(0 ) } ) \pm \delta_{eff}(h_{1,2}^{(0)})}{2\omega_n(h_{1,2}^{(0 ) } ) } , \ \\
\label{eta } \eta_{1,2}^{\pm}&=\pm \frac{g\sqrt{n+1}}{\omega_n(h_{1,2}^{(0)})}. \ \\end{aligned}\ ] ] as is apparent from , generally , any probability amplitude in the rabi - wave packet is the _ superposition of four subpackets_. two of them ( first and second terms in expressions )
correspond to the excited initial state and two another ( third and fourth terms in ) correspond to the ground initial state .
the partial subpackets in eqs .
are characterized by the different frequency shifts @xmath170 and different velocities of motion @xmath167 , given by , .
this velocities coincide with the group velocities of traveling rabi waves when @xmath174 : one can elementary verify , that @xmath175 , @xmath176 .
it is essential that velocities @xmath167 as well as the frequency shifts @xmath170 depend on photon number @xmath38 .
it means that the spatial propagation of wavepacket is accompanied by the change of quantum light statistics ( for example , in initially coherent light incoherent component appears ) . under the some specific conditions the number of subpackets could decrease .
two mechanisms of such decrease are possible : the first one is the tending to zero the subpacket amplitude and the second one is the confluence of the subpackets velocities . as is seen from , if @xmath177 for one pair of the subpackets the velocity synchronism condition @xmath178 is fulfilled .
notice that eq .
is the analogue of the ordinary synchronism condition in single two - level system @xmath179 .
the velocity synchronism condition can be fulfilled by fitting the value of the detuning @xmath180 .
if this is the case , there are three subpackets instead of four them due to the confluence mechanism acting .
note , that @xmath181 and @xmath182 can not be equal to zero together at @xmath183 .
if the system is initially prepared in the stationary state , and the synchronism condition is fulfilled in the corresponding point ( for example @xmath184 for any @xmath38 and @xmath185 ) , only one subpacket is preserved . in this case
the expression for the inversion density becomes rather simple : @xmath186.\ ] ] expression sums up contributions of different photonic states ( terms with different @xmath38 ) .
each contribution is the product of two multipliers describing different optical processes .
the second multiplier describes rabi oscillations with the frequencies @xmath187 , occurring in the qd - chain , while the first one shows that the region of rabi oscillations moves therewith in space with velocity @xmath188 .
one should note , that in the case of the exact synchronism @xmath189 does not depend on @xmath38 .
that is why in this partial case the quantum statistics of light is not distorted with propagation of the wavepacket . in particular ,
the initial coherency of quantum light persists in time .
it is immediately follows from eq . and normalization condition that the integral inversion @xmath190 corresponding to the inversion density oscillates in time similarly to the rabi - oscillations in the single two - level system ( see eq.(6.2.21 ) in @xcite at @xmath102 ) . generally , each pair of the subpackets propagates due to the tunneling through its own potential barrier
the velocities of the subpackets motion are proportional to the barrier transparency ( @xmath191 ) .
if one of the barriers becomes absolutely opaque , the corresponding pair of subpackets does not move : @xmath192 at @xmath193 . the light propagation _ along the qd - chain _ is another necessary condition of rabi subpackets motion : @xmath192 at @xmath194 . for more detail analysis of rabi wavepackets dynamics ( in particular for taking into account the diffraction spreading ) the integrals need to be calculated numerically .
the results of calculations is analysed in the following section .
consider the case of an excited initial state with space distribution in form of single gaussian beam : @xmath195{\pi\sigma^2}$ ] , @xmath184 .
assume that initially light is prepared in coherent state , so photon distribution is given by the poisson law : @xmath196 , where @xmath197 defines the average photon number .
the spatial - temporal dynamics of inversion density for this case is depicted on fig.[fig1]a . as is seen from it , original gaussian packet oscillates in time and moves along the chain .
oscillations collapse to zero quickly , but after a while they revive _ in another area of space_. the phenomenon of rabi oscillations collapses and revivals , caused by discreteness of photon distribution , is well studied @xcite , but the spacing the collapse and subsequent revival is qualitatively new effect .
space - time distribution of the inversion density ( a ) and temporal dependence of the integral inversion ( b ) in the qd chain for a coherent initial state of light ( @xmath198 ) .
the initial state of qd - chain is a single gaussian wavepacket .
@xmath195{\pi\sigma^2}$ ] , @xmath184 , @xmath199 , @xmath200 , @xmath121 , @xmath201 , @xmath202 , @xmath203.,title="fig:",scaledwidth=48.0% ] space - time distribution of the inversion density ( a ) and temporal dependence of the integral inversion ( b ) in the qd chain for a coherent initial state of light ( @xmath198 ) .
the initial state of qd - chain is a single gaussian wavepacket .
@xmath195{\pi\sigma^2}$ ] , @xmath184 , @xmath199 , @xmath200 , @xmath121 , @xmath201 , @xmath202 , @xmath203.,title="fig:",scaledwidth=48.0% ] space - time distribution of the inversion in the qd chain for the case of two counterpropagating wavepackets .
@xmath204/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath205/\sqrt[4]{4\pi\sigma^2}$ ] .
the light is initially in coherent state ( @xmath206 ) , @xmath207 , @xmath208 , @xmath209 .
a ) @xmath210 , @xmath211 ; b ) @xmath202 , @xmath203.,title="fig:",scaledwidth=48.0% ] space - time distribution of the inversion in the qd chain for the case of two counterpropagating wavepackets .
@xmath204/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath205/\sqrt[4]{4\pi\sigma^2}$ ] .
the light is initially in coherent state ( @xmath206 ) , @xmath207 , @xmath208 , @xmath209 .
a ) @xmath210 , @xmath211 ; b ) @xmath202 , @xmath203.,title="fig:",scaledwidth=48.0% ] although a variation of the inversion density , depicted in fig.[fig1]a , in arbitrary point of the space occupied by the rabi wavepacket is not too large , the integral inversion , presented in fig.[fig1]b oscillates between -1 and 1 , thus indicating presence of strong light - qd coupling .
temporal dependence of the integral inversion for a coherent initial state of light at the input parameters as follows : @xmath206 , @xmath202 , @xmath203 ( solid line ) ; @xmath206 , @xmath210 , @xmath211(dotted line ) @xmath212 , @xmath202 , @xmath211(dashed line ) . in all cases , @xmath208 , @xmath207 , @xmath209 .
, scaledwidth=48.0% ] the asymmetry of the inversion space - time distribution for two counterpropagating wavepackets in case of the coherent state of light with small average number of the photons @xmath212 .
@xmath204/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath205/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath208 , @xmath207 , @xmath209 , @xmath202 , @xmath211.,scaledwidth=48.0% ] let us consider now the case of a mixed initial state ( both @xmath213 and @xmath214 are nonzero ) .
as it has been noted above , the synchronism condition can not be fulfilled for both pair of subpackets ( except the case when one of the pairs is motionless , see below ) . as a consequence
, we have four subpackets ( or three ones if @xmath215 for one of the pairs ) . if @xmath216 ( potential barriers with equal transparency ) and @xmath179 , then both packet pairs move with pairwise equal velocities in opposite directions ( @xmath217 ) .
if the initial spatial distributions of probability amplitudes for ground and excited states are spaced , the subpackets will collide ( see fig.[double ] ) .
initial gaussian profiles of packets deform with time due to the difference in subpacket velocities .
one should note that for the case of quantum external field in coherent state qd - chain can not be saturated , i.e. the integral inversion could not be a constant ( see fig .
[ int_inv2 ] ) .
this is due to the fact that @xmath218 and @xmath219 are asymmetrical with respect to @xmath38 and as a result @xmath220 is not a constant for any form of space dependence of @xmath218 , @xmath219 .
this asymmetry exists due to the presence of vacuum term in the sum and increases with decreasing of the average photon number @xmath197 ( compare fig.[double ] and fig.[double_3 ] ) .
in other limiting case , for @xmath122 ( fully opaque ground - state barrier ) , the velocities @xmath221 and one pair of subpackets do not moves along the chain .
if , moreover , @xmath222 , then @xmath223 , and the synchronism condition is fulfilled for both packet pairs ( see fig.[double2 ] ) .
space - time distribution of the inversion in the qd chain for the case of the fully opaque ground - state barrier .
@xmath204/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath205/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath199 , @xmath122 , @xmath224 , @xmath201 , @xmath202 , @xmath203 .
the light is initially in the coherent state with @xmath206,scaledwidth=48.0% ] let us now consider the interaction of the qd - chain with another important state of initial electromagnetic field , namely , the fock qubit state , i.e. a superposition of two fock states : @xmath225 , where @xmath226 is an arbitrary fixed number , @xmath227 satisfy the normalization conditions . as in the previous case , for initial spatial distributions the gaussian beams will be used .
the spatial - temporal dynamics of the inversion density for the initial state of field @xmath228 is shown in fig .
[ fig_fock_qubit]a .
it can be seen from the figure that rabi wavepacket does not collapse and revive , but oscillates in a complicated manner ( see also integral inversion dynamics , fig.[fig_fock_qubit ] ) .
space - time distribution of the inversion ( a ) and temporal dependence of the integral inversion ( b ) in the qd chain for the field in the fock qubit initial state .
@xmath228 , @xmath229{\pi\sigma^2})$ ] @xmath230 , @xmath148 for @xmath231 , @xmath149 for all @xmath38 , @xmath199 , @xmath200 , @xmath121 , @xmath202 , @xmath209 , @xmath203.,title="fig:",scaledwidth=48.0% ] space - time distribution of the inversion ( a ) and temporal dependence of the integral inversion ( b ) in the qd chain for the field in the fock qubit initial state .
@xmath228 , @xmath229{\pi\sigma^2})$ ] @xmath230 , @xmath148 for @xmath231 , @xmath149 for all @xmath38 , @xmath199 , @xmath200 , @xmath121 , @xmath202 , @xmath209 , @xmath203.,title="fig:",scaledwidth=48.0% ] in the end of this section we consider the vacuum initial state of external electromagnetic field .
it is well known that for this case the vacuum harmonic rabi oscillations occuring due to spontaneous emission can take place @xcite .
the corresponding rabi - wave packet in the initially excited qd - chain is illustrated at fig.[fig_fock_qubit ] .
this excitation couples single mode electromagnetic vacuum state with one - photon state only and can be imagined as a wave beam which is described by the monochromatic rabi - frequency spectrum and continuous spatial spectrum at the same time .
however , it should be noted that one need to consider an interaction of the qd - chain with a multi - mode field for full treatment of such effects @xcite . this problem is a subject for future considerations .
space - time distribution of the inversion in the qd chain for the vacuum initial state of the field .
@xmath232{\pi\sigma^2})$ ] @xmath230 , @xmath148 for @xmath233 , @xmath149 for all @xmath38 , @xmath234 , @xmath235 , @xmath121 , @xmath210 , @xmath201 , @xmath203.,scaledwidth=48.0% ]
the case of a coherent state of electromagnetic field with the large photonic number @xmath38 is especially interesting for us . in this case
one can neglect the quantum nature of the light and consider electromagnetic field as a classical one replacing the electric field operator with its expectation value @xmath236\}$ ] . in doing this
we neglect any changes of the field state . omitting the local - field effects as before ,
we can write the hamiltonian of the system `` qd - chain - electromagnetic field '' in the form @xmath237 , where the term @xmath238\ ] ] describes rabi oscillations in non - interacting qds and @xmath239 is the rabi frequency @xcite , the interdot interaction mechanism is independent on light properties , so the term @xmath240 is defined as before by equation .
the state vector of the system has the form : @xmath241 where @xmath242 , @xmath243 are unknown functions .
now the equations of motion for them can now be obtained using approach of sec.[sec : mot_eq ] . introducing the decay factor @xmath49 into the nonstationary schrdinger equation and making continuous limit transition in the qd - chain similar to the sec.[sec : cont_lim ] , we obtain the system of equations as follows : @xmath244\bm{\psi}\\+ia^2\hat{\xi}\partial^2_x\bm{\psi } , \end{split}\ ] ] where @xmath245 $ ] .
note , that spontaneous emission can be taken into account in the same way as in [ sec : mot_eq ] , i.e. by describing it as interaction of qds with photon bath . in this case
the damping constant @xmath49 in eq . will be the sum of two partial constants .
the system can be considered as a particular case of the general system and can be solved in the same way .
the solution reads @xmath246t/2\ } } } dh .
\end{split}\ ] ] where the values @xmath247 , @xmath248 , @xmath249 , @xmath250 , @xmath251 , @xmath143 are determined , respectively , by equalities , , , , , if one makes the substitution @xmath252 ( in doing that the photonic number dependence in @xmath247 , @xmath253 , @xmath254 , @xmath255 disappears ) .
the vector function @xmath144 is determined by the initial conditions similar to eq .. space - time distribution of the inversion in the qd in the classical light chain for a single gaussian wavepacket @xmath256{\pi\sigma^2})$ ] @xmath230 , @xmath257 , @xmath258 , @xmath259 , @xmath121 , @xmath209 , @xmath260 .
a ) @xmath261 ; b ) @xmath262,title="fig:",scaledwidth=48.0% ] space - time distribution of the inversion in the qd in the classical light chain for a single gaussian wavepacket @xmath256{\pi\sigma^2})$ ] @xmath230 , @xmath257 , @xmath258 , @xmath259 , @xmath121 , @xmath209 , @xmath260 .
a ) @xmath261 ; b ) @xmath262,title="fig:",scaledwidth=48.0% ] a typical space - time distribution of the inversion density @xmath263 $ ] is shown in fig . [ fig2 ] .
as is seen from the figure , the space - time dynamics of rabi wavepacket is similar to the one for the case of vacuum oscillations ( but it is of substantially different physical nature : electromagnetic field is in coherent state with photon number @xmath264 and this state does not change when the system oscillates between ground and excited states in contradiction to the case of vacuum rabi oscillation ) . since the photon distribution does not change , collapses and revivals are absent .
however there is another phenomenon of similar nature .
one can see in fig .
[ fig3 ] , where the plots of the integral inversion are presented , that the oscillations of this quantity at @xmath265 , @xmath266 decrease with time even at @xmath267 , whereas such a damping is absent at @xmath268 , @xmath108 and integral inversion oscillates harmonically in the range from -1 to 1 ( dotted curve in fig .
[ fig3 ] ) .
temporal dependence of the integral inversion in the classical light for the input parameters as follows : @xmath102 , @xmath268 , @xmath269 ( dotted line ) ; @xmath270 , @xmath271 , @xmath269 ( solid line ) ; @xmath121 , @xmath271 , @xmath272 , @xmath273 ( dashed line ) . in all cases
@xmath274 , @xmath211 . , scaledwidth=48.0% ]
such a behaviour is the consequence of dependence of the effective detuning @xmath143 ( and therefore the rabi frequency @xmath251 ) on @xmath69 .
but if @xmath268 and @xmath108 , then @xmath275 and @xmath255 does not depend on @xmath69 . the damping rate is controlled by values of wavenumber @xmath95 and coupling constants @xmath27 .
if the initial state of the system is the excited state ( @xmath276 ) , the damping rate is predominantly determined by product @xmath277 ( and damping is practically absent if @xmath278 ) , for the ground initial state ( @xmath279 ) it is defined by @xmath280 . in the case of a mixed initial state
all three quantity are important .
note , that more narrow wavepacket spreads faster , than the wider one ( compare fig .
[ fig2]a and fig . [ fig2]b ) . in the weak coupling limit the indicated dephasing mechanism is analogous to the landau damping in plasma
space - time distribution of the inversion density in the qd chain in the classical light for two counterpropagating identical gaussian wavepackets : @xmath281/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath282/\sqrt[4]{4\pi\sigma^2}$ ] , @xmath102 , @xmath209 , @xmath261 , @xmath283 , @xmath260.,scaledwidth=48.0% ] interaction of two counterpropagating identical gaussian rabi wavepackets colliding at @xmath284 is shown in fig . [ fig4 ] . unlike the case of quantum field ,
the space - time distribution of the inversion density is fully symmetrical with respect to the line @xmath284 and integral inversion does not oscillate : this quantity equals zero for all @xmath285 and arbitrary values of @xmath286 .
local fields are depolarization fields arising inside qds due to the dipole - dipole electron - hole interaction . as a result , the field acting on the e - h pair differs from mean field @xcite . the local - field action is described by eq . and leads to nonlinear terms in the equations of motion ( last terms in eqs .
the presence of nonlinearity results in the qualitatively new features of rabi oscillations in single qds @xcite .
the influence of the local - fields on rabi waves is the subject for future investigations . in this section
we consider some aspects of this influence only and discuss some sufficient conditions of the local - field negligibility .
let us consider finite ( of the length @xmath287 ) qd - chain interacting with classical light .
the relaxation processes are neglected ( i.e. @xmath211 ) .
the equations of motion follow from eqs.- in the same way as eq .
reduce to eq . and
have the form @xmath288 we use the periodic boundary conditions ( born - von karman conditions ) : @xmath289 suppose that electromagnetic field are also satisfy periodic conditions , i.e. @xmath290 , where @xmath291 is integer number .
then the problem under consideration has a clear physical interpretation : we are dealing with ringed qd - chain of the radius @xmath292 .
qd - chain interacts with electromagnetic mode of whispering gallery type , which is excited in microcavity with rotation axis ( @xmath293 , @xmath294 , @xmath295 is the azimuthal coordinate ) .
let us seek partial solution of the system - in the form of traveling wave : @xmath296 , @xmath297 , where @xmath298 , @xmath299 are constant amplitudes satisfying the normalization condition for wavefunction , @xmath69 and @xmath70 are real values . ssubstituting @xmath300 and @xmath301 into eqs.-
we obtain @xmath302u_0 + \frac{\omega_r}{2}v_0=0 , \\
\rule{0in}{4ex } \label{char_eq_2b } & \left[\nu+\phi_2-\delta\omega|u_0|^2\right]v_0 + \frac{\omega_r}{2}u_0=0,\end{aligned}\ ] ] where @xmath303 . from boundary conditions
one can obtain the quantization condition for @xmath69 : @xmath304 , @xmath38 is integer . expressing @xmath299 from eq . and
substituting into eq .
, we obtain @xmath305 the wavefunction normalization condition results in @xmath306 eqs.- form the closed system of equations with respect to @xmath70 and @xmath307 , which defines the dispersion law of rabi waves @xmath308 with regard to the local - field effects .
it should be noted that only solutions with real - valued @xmath70 have physical meaning ( otherwise the system - contradicts to the initial equations - ) . for @xmath309 , @xmath310 the system - reduces to the dispersion equation @xcite for rabi waves in the infinite qd - chain interacting with classical light .
dispersion curves for rabi waves subject to local - field effects .
a ) @xmath311 , b ) , c ) @xmath312 . in all cases
@xmath269 , @xmath313 , @xmath102 , @xmath314 . ,
title="fig:",scaledwidth=48.0% ] dispersion curves for rabi waves subject to local - field effects .
a ) @xmath311 , b ) , c ) @xmath312 . in all cases
@xmath269 , @xmath313 , @xmath102 , @xmath314 . , title="fig:",scaledwidth=48.0% ] dispersion curves for rabi waves subject to local - field effects . a ) @xmath311 , b ) , c ) @xmath312 . in all cases
@xmath269 , @xmath313 , @xmath102 , @xmath314 . ,
title="fig:",scaledwidth=48.0% ] the dispersion characteristics of rabi waves for two essentially different values of @xmath287 but identical @xmath315 and other parameters are shown in fig.[nonlin_disp ] .
the curves depicted in fig.[nonlin_disp]a corresponds to the qd - chain of large radius .
the nonlinear effects in this case are negligibly small .
dispersion curves do not feel nonlinear effects and are practically identical to obtained in @xcite .
for qd - chain of small radius the situation becomes principally different due to nonlinear effects .
the _ closed loop _ appears in the vicinity of any cross point and the number of the modes increases to four ( fig.[nonlin_disp]b ) .
the size of this loop decreases with decreasing the ratio @xmath316 .
the detailed structure of closed loop is shown in fig.[nonlin_disp]c .
when @xmath69 ( and , correspondingly , @xmath38 ) changes adiabatically slow the motion along the loop takes place ( it is shown by arrows in fig.[nonlin_disp]c ) . the transitions from one branch to another
are followed by simultaneous changes of rabi frequency and rabi wave group velocity .
the trajectories of motion in different directions proves to be nonidentical .
thus , the loop has an hysteresis character .
interaction of the local - fields and rabi waves leads to effect of bistability of quantum states in the qd - chain .
the detailed analysis of different dispersion branches stability is the subject for future investigations .
let us now discuss the conditions of the local field effects neglecting in equations .
note , that for @xmath317 the system transforms to equations ( 24 ) in @xcite , where rabi oscillations in single qd are described with account of local field effects .
an analysis of equations ( 24 ) in @xcite has shown that local field effects are negligibly small for the case of sufficiently large rabi frequencies , namely , @xmath318 , where @xmath319 , @xmath197 is average photon number .
if this condition is fulfilled for coherent state , the conventional picture of rabi oscillations ( the alternation of collapses and revivals ) takes place . in the opposite case
qualitatively new regime of oscillations arises .
this regime is due to the local field effects and is described in detail in@xcite . in this work
we are interested only in first regime . in our case
the role of @xmath320 is played by frequencies @xmath321 , which are determined by eq .. therefore it is reasonable to suggest that for traveling rabi waves the local fields effects are negligible under the condition @xmath322 where @xmath315 is determined by eq .. in the case of rabi - wavepackets general line of reasoning is the same as for eq . and leads to the following condition of the local - field negligibility : @xmath323 .
the space - time dependence of normalized correlation function @xmath324 for light in fock state with @xmath325 and the input parameters as follows : @xmath326 , @xmath327 , @xmath274 , @xmath328{\pi\sigma^2}$ ] , @xmath329.,scaledwidth=48.0% ] in this section we will calculate exciton - exciton and exciton - photon multi - time correlation functions . for this purpose
it is convenient to pass to the heisenberg picture and to use the discrete model of qd - chain ( transition to the continuous limit will be made on the final stage ) .
thus , we should look for a time dependence of operators @xmath330 , i.e. we shall seen to find the evolution operator
. it can be obtained from ( for details see appendix [ sec : evol_operator ] ) and has the form @xmath331\\ & \left.+{\varphi}_n^+ ( h , t ) e^{i\delta_h^-t } \left|b_p , n+1\right\rangle\left\langle b_q , n+1\right| \right\}e^{ih(p - q)a}dh . \end{split}\ ] ] here @xmath332 , @xmath333 , @xmath334 are given by formulas , , .
quantities @xmath335 , @xmath143 are defined by expressions , if one substitutes @xmath336 $ ] .
note , that if one makes approximation @xmath337 \cong1-(h\pm k/2)^2a^2/2 $ ] ( this is the usability condition for continuous limit ) , the quantities @xmath335 , @xmath143 coincide with expressions , . from expression
we obtain by using the formula @xmath338 the exciton - exciton correlation functions @xmath339 as follows : @xmath340\\ [ \varphi_{n+1}^+(h , t)\varphi_{n+1}^-(g , t')u_{l , n+1}^*(0)u_{r , n+1}(0)+ \psi_{n+1}(h , t)\psi_{n+1}(g , t')v_{l , n+2}^*(0)v_{r , n+2}(0 ) \\ \rule{0in}{3ex } \nonumber+\varphi_{n+1}^+(h ,
t)\psi_{n+1}(g , t')u_{l , n+1}^*(0)v_{r , n+2}(0)- \psi_{n+1}(h , t)\varphi_{n+1}^-(g , t')v_{l , n+2}^*(0)u_{r , n+1}(0)]dh'dhdg,\end{gathered}\ ] ] where @xmath341 , @xmath342
. the continuous limit transition can be made by means of the following substitutions : @xmath343 , @xmath344 , @xmath345 , @xmath346 .
one can see , that in general case @xmath347 can not be represented as @xmath348 or @xmath349 .
it indicates the absence both spatial and time homogeneity .
let us consider case of @xmath350 and @xmath351 .
then @xmath352 , @xmath353 and taking into account that @xmath354 , we obtain rather simple expression for correlation function : @xmath355 or , in continuous limit , @xmath356 plot of normalized correlation function @xmath357 for @xmath358 , @xmath351 is presented on fig.[correl ] .
the external field is supposed to be in a single - photon initial state and the electron in qd - chain is in the excited initial state ( @xmath184 ) .
the correlation function of the polarization operator @xmath359 with the operator of the external field @xmath360 ( exciton - photon correlators ) appears as @xmath361 \\
\nonumber -\psi_{n+1}(h',t)\sqrt{n+2}\left[\psi_n(h , t)v^*_{r , n+1}(0)u_{l , n}(0)+\varphi_n^+(h , t)v^*_{r , n+1}(0)v_{l , n+1}(0)\right ] \ } e^{ipa(h - h')}e^{i(h'r - hl)a}e^{i(kpa-\omega_0 t)}e^{i\omega(t - t')}.\end{gathered}\ ] ]
the main results , obtained in this work , can be summarized as follows : \i ) in qd - chain with tunneling coupling in strong electron - photon coupling regime the space propagation of rabi oscillations ( rabi waves ) takes place . for propagation of rabi waves the wave vector of the photon mode must have a nonzero component along the chain .
characteristics of the rabi waves depend strongly on relations between parameter of electron - photon coupling , frequency deviation and transparency factors of tunneling barriers for both of levels .
\ii ) traveling rabi waves are quantum states of the qd - chain dressed by radiation .
the qualitative distinction of this states from states of single dressed atom @xcite is the space - time modulation of dressing parameter by traveling wave law .
traveling rabi waves can be interpreted as entangled states of e - h pair and photons .
these states are characterized by the dependence of energy on quasi - momentum and can be treated as new type of quasiparticles ( in this paper they are called rabitons ) .
\iii ) there are two traveling rabi modes with different frequencies of rabi oscillations at a given value of wavenumber .
the range of rabi oscillations frequencies is limited by the critical value , which is different for both modes .
the qd - chain is opaque for rabi modes with frequencies lesser than critical one .
the critical frequencies as well as dispersion characteristics depend on number of photons in the mode .
\iv ) in general case the propagation of rabi wavepackets occurs in the form of four partial packets , which are characterized by different amplitudes and velocities of motion .
rabi wavepackets transfer energy , inversion , quasimomentum , electron - electron and electron - photon quantum correlations along the chain .
particularly , in the case of qd - chain interacting with quantum light in the coherent state the known collapses - revivals picture @xcite is drastically modified due to propagation effect : collapses and revivals take place in different points of space .
\v ) one pair of rabi wavepackets exists due to the ground state contribution in initial state of the system , another one appears due to the contribution of excited state ( generally , initial state is arbitrary superposition of ground and excited states ) .
each pair of packets moves through its own potential barrier
. if one of the barriers is completely opaque , then corresponding pair of wavepackets does not move and does not spread .
\vi ) rabi wavepackets propagation along the qd - chain is followed by the transformation of quantum light statistics ( for example , in initially coherent light incoherent component appears ) .
if specific conditions are fulfilled then incoherent part can be decreased to zero .
\vii ) in particular cases a number of rabi wavepackets can be diminished .
there are two mechanisms of such diminution : tending to zero of partial wavepacket amplitude and merging of the packets due to their velocities equality .
rabi waves can take place in a number of other distributed systems strongly coupled with electromagnetic field .
the example are superconducting circuits based on josephson junctions , which are currently the most experimentally advanced solid - state qubits @xcite .
it is evident , for example , that the chain of qubits placed inside a high - q transmission - line resonator due to the qubit - qubit capacitance coupling will support the rabi waves propagation similar to described in this article .
rabi waves effect can be practically used in nanoelectronics , quantum computing , quantum informatics .
the work of g. ya .
slepyan was partially carried out during the stay at the institut fr festkrperphysik , tu berlin , and supported by the deutsche forschungsgemeinschaft ( dfg ) .
authors are grateful to dr
. s. a. maksimenko , dr . d. s. mogilevtsev , dr .
j. haverkort and dr .
a. m. nemilentsau for stimulative discussions .
in this appendix , the evolution operator for a discrete qd - chain ( eq . ) is obtained from eq .
let us introduce new variables @xmath362 , @xmath363 .
it leads to the system coefficients in this system do not depend from qd - number @xmath1 .
it allows us to seek the solution in the form @xmath365 , @xmath366 , @xmath367 $ ] .
then the equations for different qds become independent and may be written as @xmath368 u_{p , n}-i g\sqrt{n+1}v_{p , n+1 } e^{i\delta t},\\ \label{system_diskr4_b } & \partial_tv_{p , n+1}=2i\xi_2 \cos [ ( h - k/2)a]v_{p , n+1}-i g\sqrt{n+1}u_{p , n } e^{-i\delta t}.\end{aligned}\ ] ] expressing @xmath369 from and substituting into , one obtains the set of independent second - order ordinary differential equations for the variables @xmath370 @xmath371 where @xmath372 + 2\xi_2 \cos [ ( h - k/2)a]$ ] . the solution of reads @xmath373t/2}+c_2^n(0)e^{i\left[\delta_h-\omega_n(h)\right]t/2}\ ] ] with the constants @xmath374 to be determined from initial conditions .
writing the latter in the form @xmath375 , @xmath376 , we rewrite as @xmath377t}+\beta_ne^{i\left[\delta_h^++\omega_n(h)/2\right]t}\right ] e^{ihpa}\ ] ] and @xmath378t}+\delta_ne^{-i\left[\delta_h^--\omega_n(h)/2\right]t}\right ] e^{ihpa},\ ] ] where @xmath379 we now can represent the required wavefunction as fourier integral with basis functions @xmath381 : @xmath382 where @xmath383 is unknown weighting function . to determine it
let us take into account , that substituting into , we obtain @xmath389 , with @xmath390 given by the eq .. these calculations allow one to pass from schrdinger picture to the heisenberg one .
they are used in the calculations of different types of correlators . in hamiltonian @xmath391 the constant term describing the photon vacuum contribution
is omitted . in singlemode
assumption its influence identically equal to zero . in
so doing we are not interested in effects of spontaneous decay and lamb shift .
this effects are caused by interaction of the system with photon vacuum and can be taken into account in multimode assumption only .
|
we predict and theoretically investigate the new coherent effect of nonlinear quantum optics spatial propagation of rabi oscillations ( rabi waves ) in one - dimensional quantum dot ( qd ) chain .
qd - chain is modeled by the set of two - level quantum systems with tunnel coupling between neighboring qds .
the space propagation of rabi waves in the form of traveling waves and wave packets is considered .
it is shown , that traveling rabi waves are quantum states of qd - chain dressed by radiation .
the dispersion characteristics of traveling rabi waves are investigated and their dependence on average number of photons in wave is demonstrated .
the propagation of rabi wave packets is accompanied by the transfer of the inversion and quantum correlations along the qd - chain and by the transformation of quantum light statistics .
the conditions of experimental observability are analyzed .
the effect can find practical use in quantum computing and quantum informatics .
|
the only way by which the age of the sun can be estimated to a useful degree of precision is by accepting the basic tenets of solar - evolution theory and measuring those aspects of the structure of the sun that are predicted by the theory to be indicators of age .
the structure measurements must be carried out seismologically , and evidently one expects greatest reliability of the results when all the available helioseismic data are employed . however , the most relevant modes are those of lowest degree , because it is they that penetrate most deeply into the energy - generating core where the relic helium - abundance variation records the integrated history of nuclear transmutation .
moreover , it is also only they that can be measured in other stars .
therefore , there has been some interest in calibrating theoretical stellar models using only low - degree modes .
the prospect was first discussed in detail by ( * ? ? ?
* ; * ? ? ?
* christensen - dalsgaard ( 1984 , 88 ) ) , @xcite and @xcite , although prior to that it had already been pointed out that the helioseismic frequency data that were available at the time indicated that either the initial helium abundance @xmath2 or the age @xmath3 , or both , are somewhat greater than the generally accepted values ( * ? ? ? * ( gough 1983 ) ) , an inference which is consistent with our present findings .
subsequent , more careful , calibrations were carried out by @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and bonanno , schlattl & patern ( 2002 ) .
not all of them addressed the influence of uncertainties in @xmath2 on the determination of @xmath3 .
as a main - sequence star ages , helium is produced in the core , increasing the mean molecular mass @xmath4 , preferentially at the centre , and thereby reducing the sound speed .
the resulting functional form of the sound speed @xmath5 depends not only on age @xmath3 but also on the relative augmentation of @xmath6 , which itself depends on the initial absolute value of @xmath4 , and hence on @xmath2 .
@xcite tried to separate these two effects using the degree dependence of the small separation @xmath7 of cyclic frequencies @xmath8 , where @xmath9 is order and @xmath10 is degree .
this is possible , in principle , because modes of different degree and similar frequency sample the core differently .
however , that difference is subtle , and the sensitivity to the relatively fine distinction between the effects of @xmath3 and @xmath2 on the functional form of @xmath5 in the core is low .
consequently the error in the calibration produced by errors in the observed frequency data is uncomfortably high .
this lack of sensitivity can be overcome by using , in addition to core - sensitive seismic signatures , the relatively small oscillatory component of the eigenfrequencies induced by the sound - speed glitch associated with helium ionization ( * ? ? ?
* ( gough 2002 ) ) , whose amplitude is close to being proportional to helium abundance @xmath11 ( * ? ? ?
* ( houdek & gough 2007a ) ) .
the neglect of that component in the previously employed asymptotic signature @xmath12 had not only omitted an important diagnostic of @xmath11 but had also imprinted an oscillatory contamination in the calibration as the limits @xmath13 , where @xmath14 , of the adopted mode range was varied ( * ? ? ? * ( gough 2001 ) ) .
it therefore behoves us to decontaminate the core signature from glitch contributions produced in the outer layers of the star ( from both helium ionization and the abrupt variation at the base of the convection zone , and also from hydrogen ionization and the superadiabatic convective boundary layer immediately beneath the photosphere ) . to this end
a helioseismic glitch signature has been developed by @xcite , from which frequency contributions @xmath15 can be computed and subtracted from the raw frequencies @xmath8 to produce effective glitch - free frequencies @xmath16 to which a glitch - free asymptotic formula ( [ eq : asymp ] ) can be fitted .
the solar calibration is then accomplished by interpolating the theoretical seismic signatures computed on a grid of solar models to the observations , using a standard grid to compute derivatives with respect to @xmath3 and @xmath2 , and a carefully computed reference solar model designed to be close to the sun .
the result of the first preliminary calibration by this method , using bison data , has been reported by @xcite . here
we enlarge on our discussion of the analysis , and we augment our results with a calibration based on a second reference solar model .
, of low - degree solar frequencies from bison .
top right : the symbols are second differences @xmath17 of adiabatic pulsation eigenfrequencies of solar model s of @xcite .
the solid curves in both panels are the diagnostics ( [ eq : delnu ] ) ( [ eq : secdiff ] ) , whose eleven parameters have been adjusted to fit the data optimally .
bottom : the symbols denote contributions @xmath18 to the frequencies produced by the acoustic glitches of the sun ( left panel ) and model s ( right panel ) . ]
any abrupt variation in the stratification of a star ( relative to the scale of the inverse radial wavenumber of a seismic mode of oscillation ) , which here we call an acoustic glitch , induces an oscillatory component in the spacing of the cyclic eigenfrequencies of seismic modes .
our interest is principally in the glitch caused by the depression in the first adiabatic exponent @xmath19 ( where @xmath20 , @xmath21 and @xmath22 are pressure , density and specific entropy ) caused by helium ionization , which imparts a glitch in the sound speed @xmath5 . the deviation @xmath23 where @xmath24 , of the eigenfrequency @xmath25 from the corresponding frequency @xmath26 of a similar smoothly stratified star is the indicator of @xmath11 that we use in conjunction with the indicators of core structure to determine the main - sequence age .
approximate expressions for the frequency contributions @xmath27 arising from acoustic glitches in solar - type stars were recently presented by @xcite .
here we improve them by adopting the appropriate airy functions ai@xmath28 that are used as comparison functions in the jwkb approximations to the oscillation eigenfunctions , as did @xcite .
the complete expression for @xmath27 is then given by @xmath29 where @xmath30\cr & & \hspace{-8pt } \times\bigl[\tilde\mu\tilde\beta\int_0^t\kappa^{-1}_{\rm i } { \rm e}^{-(\tau-\tilde\eta\tau_{\rm ii})^2/2\tilde\mu^2\delta^2_{\rm ii}}|x|^{1/2 } & & \;\;+\;\int_0^t\kappa^{-1}_{\rm ii } { \rm e}^{-(\tau-\tau_{\rm ii})^2/2\delta^2_{\rm ii}}|x|^{1/2 } \label{eq : delgamnu}\end{aligned}\ ] ] arises from the variation in @xmath31 induced by helium ionization , and @xmath32 \!-\!(16\pi^2\tilde{\tau}_{\rm c}^2\nu^2\!+\!1)^{1/2 } \right\}\ , \label{eq : delcnu}\end{aligned}\ ] ] arises from the acoustic glitch at the base of the convection zone resulting from a near discontinuity ( a true discontinuity in theoretical models using local mixing - length theory with a non - zero mixing length at the lower boundary of the convection zone ) in the second derivative of density .
here , @xmath33 is a constant , being a representative polytropic index in the expression for the approximate effective phase @xmath34 appearing in the argument of the airy function , and @xmath35 , @xmath36 and @xmath37 are constants of order unity which account for the relation between the acoustic glitches caused by the first and second stages of ionization of helium ( ( * ? ? ?
* houdek & gough 2007a ) ) ; @xmath38 is acoustic depth beneath the seismic surface of the star , and @xmath39 is the total acoustic radius of the star ; @xmath40 and @xmath41 are respectively the acoustic width of the glitch and its acoustic depth beneath the seismic surface .
the argument of the airy function is @xmath42 , where @xmath43 \hspace{23pt}{\rm if}\ \tilde\tau>\tau_{\rm t}\,,\ ] ] and @xmath44 \hspace{10pt}{\rm if}\ \tilde\tau\le\tau_{\rm t}\,,\ ] ] in which @xmath45 , with @xmath46 , and @xmath47 is the location of the upper turning point of the mode ; also @xmath48^{1/2}\,,$ ] and @xmath49 and @xmath50 .
in addition @xmath51+\pi/4,\,\ ] ] where @xmath52 and @xmath53 .
the seven coefficients @xmath54 , @xmath40 , @xmath41 , @xmath55 , @xmath56 , @xmath57 , @xmath58 are found by fitting the second difference @xmath59 to the corresponding observations by minimizing @xmath60 using the value of @xmath61 obtained by fitting to ( [ eq : asymp ] ) , where @xmath62 is the @xmath63 element of the inverse of the covariance matrix c@xmath64 of the observational errors in @xmath65 , computed , perforce , under the assumption that the errors in the frequency data @xmath25 are independent .
the last term in equation([eq : secdiff ] ) approximates smooth contributions arising , in part , from wave refraction in the stellar core , from hydrogen ionization and from the superadiabaticity of the upper boundary layer of the convection zone , introducing four more fitting coefficients @xmath66 , @xmath67 , @xmath68 .
the covariance matrix @xmath69 of the errors in @xmath70 were established by monte carlo simulation .
the outcome of the fitting to the bison data ( ( * ? ? ?
* basu et al . 2007 ) ) and to the adiabatically computed eigenfrequencies of solar model s ( ( * ? ? ?
* christensen - dalsgaard et al . 1996 ) ) is displayed in figure1 : the upper panels display the second differences , together with the fitted formula ( [ eq : secdiff ] ) , the lower panels display the corresponding contributions @xmath27 to the frequencies of oscillation from the acoustic glitches .
all the frequencies displayed in the figure have been used in equation ( [ eq : minsecdiff ] ) for fitting ( [ eq : secdiff ] ) . to the resulting glitch - free frequencies
@xmath26 , derived from equation ( [ eq : nudiff ] ) , of both the solar observations and the eigenfrequencies of the reference solar model , is fitted the asymptotic expression @xmath71 by minimizing @xmath72 , where @xmath73 and c@xmath74 is the covariance matrix of the observational errors in @xmath26 , from which we obtain both the coefficients @xmath75 , @xmath76 , and the covariance matrix @xmath77 of the errors .
following @xcite , we carry out this fitting in the frequency range given by @xmath78 , where @xmath14 and @xmath79 , and we vary @xmath80 and @xmath81 .
each of the parameters @xmath82 represents an integral of a function of the equilibrium stratification .
the integrals @xmath83 and @xmath84 are of particular importance to our analysis , because @xmath85 and @xmath84 are dominated by conditions in the core , and , although the contributions to @xmath86 from the core and the rest of the star are roughly equal in magnitude ( and potentially have opposite signs ) , the latter is relatively insensitive to @xmath3 and @xmath2 .
the integrands in the remaining integrals are either more evenly distributed throughout the sun or are concentrated near the surface .
we have carried out age calibrations using combinations of the parameters @xmath87 where @xmath88 is a measure of the maximum depression in @xmath31 in the second helium ionization zone .
presuming , as is normal , that the reference model is parametrically close to the sun , we consider the reference value @xmath89 to be approximated by a two - term taylor expansion of @xmath90 about the value @xmath91 of the sun : @xmath92 where @xmath93 and @xmath94 are the deviations of age @xmath3 and initial heavy - element abundance @xmath95 from the reference model , and @xmath96 are the formal errors in the calibration parameters whose covariance matrix @xmath97 can be derived from @xmath77 and @xmath69 .
a ( parametrically local ) maximum - likelihood fit then leads to the following set of linear equations : @xmath98 in which @xmath99 , @xmath100 , is the solution vector subject to ( correlated ) errors @xmath101 , @xmath102 , and the partial derivatives @xmath103 $ ] , @xmath104 .
a similar set of equations is obtained for the formal errors @xmath101 : @xmath105 from which the error covariance matrix @xmath106 can be computed from @xmath97 .
the partial derivatives @xmath107 were obtained from the two sets of five calibrated evolutionary models for the sun that were used in a similar calibration by @xcite , computed with the evolutionary programme by @xcite , and adopting the livermoore equation of state and the opal92 opacities .
one set of models has a constant value for the heavy - element abundance @xmath108 but varying age ; the other has constant age but varying @xmath95 .
note that , for prescribed relative abundances of heavy elements , the condition that the luminosity and radius of the sun agree with observation defines a functional relation between @xmath109 and @xmath3 .
the values of the partial derivatives @xmath107 are listed in table1 .
.partial derivatives @xmath107 obtained from two sets of calibrated evolutionary models for the sun .
values with respect to age @xmath3 are in units of gy@xmath110 . [ cols="^,^,^,^,^,^",options="header " , ]
to illustrate the effect of taking @xmath27 into account , we compare in figure2 a first assessment of @xmath111 using the glitch - free frequencies @xmath26 ( left panel ) with that obtained from the raw frequencies @xmath112 ( right panel ) .
recall that @xmath86 represents a functional of the equilibrium structure of the star , and should not vary with @xmath80 and @xmath81 .
the range of values for @xmath86 is the lower for @xmath26 , as we had anticipated .
we believe that the upturn of @xmath86 for low values of @xmath80 in the left panel of the figure is a result of the failure of the asymptotic formula ( [ eq : delnu])([eq : delcnu ] ) when @xmath25 is low .
the dipping of @xmath86 at high values of @xmath80 and low values of @xmath81 occurs because the frequency range is too small for a reliable determination of the fitting coefficients @xmath82 .
we therefore adopt intermediate values for @xmath80 and high values for @xmath81 , for which @xmath86 is insensitive to the selected frequency range .
( see equation [ eq : asymp ] ) as functions of @xmath80 and @xmath81 ( @xmath14 ) .
results are shown for fitting ( [ eq : asymp ] ) to the glitch - free frequencies @xmath26 ( left panel ) and to the raw frequencies @xmath25 ( right panel ) . ]
age calibrations using different combinations of the parameters @xmath90 and two different reference models are summarized in table2 .
the younger reference model is ` model s ' ( * ? ? ?
* ( christensen - dalsgaard et al . 1996 ) ) which has age @xmath113gy ; the second is ` model t ' , which has age @xmath114gy .
the same physics was adopted in the evolutionary calculations of both models .
we notice in table2 that the calibration for the combination @xmath115 , i.e. without @xmath116 , is less stable to a change in the reference model than are the calibrations with combinations in which @xmath116 is included , and therefore is less reliable , as we have explained in the introduction .
if we ignore in table2 the results for @xmath115 and combine the others , we obtain including the calibrations with @xmath115 does not change the outcome . error contours corresponding to the calibration from model s in the first row of table2 are plotted in figure3 .
corresponding contours for model t are the same , except that their centres are displaced to ( 4.677gy , 0.0170 ) .
one can adduce from our description of the analysis in section2 that our current treatment of the errors is not completely unbiassed ; however , the potential bias is of the order of only @xmath119 , which is small .
the age we have found is greater than currently accepted values .
the values of @xmath95 are somewhat smaller than those of models s and t ( 0.01963 ) , but we hasten to point out that they should not be regarded strictly as statements about the initial heavy - element abundance , but rather as measures of the opacity in the radiative interior .
@xcite have argued that the photospheric abundances of c , n and o had previously been overestimated , suggesting that the actual total heavy - element abundance is even lower than had previously been believed .
however , that can not imply that the opacity in the solar interior is necessarily comparably lower because it has been implicitly calibrated here ( by accepting the tenets of solar - evolution theory , and the opal opacity calculations upon which the models are based ) , and indeed the opacity has already been determined seismologically from a broader spectrum of modes than has been adopted here ( ( * ? ? ? * gough 2004 ) ) . the matter raised by asplund etal .
therefore challenges either the opacity calculations , the nuclear reaction rates , or the basic physics of stellar evolution , not helioseismology as some spectators have surmised .
as we know already from seismological structure inversions , the solar models are not accurate by helioseismological standards .
therefore the properties inferred from these calibrations could be more contaminated by systematic error than by errors in the observed frequencies .
lccccccc @xmath90 & @xmath3 ( gy ) & @xmath95 & @xmath3 ( gy ) & @xmath95 & @xmath120 & @xmath121 & @xmath122 + @xmath123&4.679&0.0169&4.677&0.0170&0.017&-0.0023&0.0005 + @xmath124&4.658&0.0177&4.673&0.0171&0.023&-0.0037&0.0007 + @xmath125&4.673&0.0165&4.676&0.0169&0.017&-0.0019&0.0007 + @xmath126&4.700&0.0169&4.680&0.0170&0.028&-0.0029&0.0005 + 2004 , _ a&a _ 417 , 751 2007 , _ apj _ 655 , 660 2002 , _ a&a _ 390 , 1115 1982 , _ mnras _ 199 , 735 1984 , in : mangeney a. , praderie f. , ( eds ) , _ space research prospects in stellar activity and variability _ , paris observatory press , paris , p.11 1988 , in : christensen - dalsgaard j. , frandsen s. , ( eds ) , _ proc .
iau symp.123 , advances in helio- and asteroseismology _ , reidel , dordrecht , p.295 1996 , _ sci _ 272 , 1286 1999 , _
a&a _ 343 , 990 1983 , in : shaver p.a .
, kunth d. , kjr k. , ( eds ) , _ primordial helium _
, southern observatory , p.117 1987 , _ nat .
_ 326 , 257 2001 , in : von hippel t. , simpson c. , manset n. , ( eds ) , _ asp conf .
vol.245 , astrophysical ages and timescales _ , gough d. o. 1987 , _ nat .
_ 326 , 257 2002 , in : favata f. , roxburgh i.w . , gadal - enrquez d. , ( eds ) , _ proc 1st eddington workshop : stellar structure and habitable planet finding _ , esa sp-485 , noordwijk , p.65 2004 , in : elebonovi v. , dppen w. , gough d. o. , ( eds ) , _ aip conf .
vol.731 , equation - of - state and phase - transition issues in models of ordinary astrophysical matter _ , am .
phys . , melville , p.119 1990 , _ solar phys .
_ , 128 , 143 1989 , _ apj _ 339 , 1156 1997 , _ apj _ 484 , 937 2007a , _ mnras _ 375 , 861 2007b , in : stancliffe r.j . , dewi j. , houdek g. , martin r.g . , tout c.a . , ( eds ) , _ aip conf . proc . : unsolved problems in stellar physics _ , american institute of physics , new york , p. 219
1986 , _ apj _ 306 , l37 1998 , _ a&a _ 332 , 215
|
we report on an ongoing investigation into a seismic calibration of solar models designed for estimating the main - sequence age and a measure of the chemical abundances of the sun
. only modes of low degree are employed , so that with appropriate modification the procedure could be applied to other stars .
we have found that , as has been anticipated , a separation of the contributions to the seismic frequencies arising from the relatively smooth , glitch - free , background structure of the star and from glitches produced by helium ionization and the abrupt gradient change at the base of the convection zone renders the procedure more robust than earlier calibrations that fitted only raw frequencies to glitch - free asymptotics . as in the past , we use asymptotic analysis to design seismic signatures that are , to the best of our ability , contaminated as little as possible by those uncertain properties of the star that are not directly associated with age and chemical composition . the calibration itself , however , employs only numerically computed eigenfrequencies .
it is based on a linear perturbation from a reference model .
two reference models have been used , one somewhat younger , the other somewhat older than the sun .
the two calibrations , which use bison data , are more - or - less consistent , and yield a main - sequence age @xmath0gy , coupled with a formal initial heavy - element abundance @xmath1 .
the error analysis has not yet been completed , so the estimated precision must be taken with a pinch of salt .
|
let @xmath0}$ ] , where @xmath1 is a field and @xmath2 is considered with its standard @xmath3 grading ( multigrading ) .
a monomial ideal @xmath4 is called _ stable _ if for every monomial @xmath5 , the monomial @xmath6 for each @xmath7 , where @xmath8 .
the class of stable ideals , introduced by eliahou and kervaire in @xcite is arguably the most well - studied class of monomial ideals . in their work ,
eliahou and kervaire construct the minimal free resolution of an arbitrary stable monomial ideal by identifying basis elements of the free modules present in the resolution which they call admissible symbols and by describing how the maps within the resolution act on these admissible symbols . among others ,
the class of stable ideals contains the so - called borel - fixed @xcite ideals as a subclass , which are fundamental to grbner basis theory .
certain homological properties of stable ideals are analyzed by herzog and hibi @xcite , where it is shown that stable ideals are componentwise linear .
more recently , distinct topological methods have been used to describe the minimal free resolution of an arbitrary stable ideal @xcite .
we construct the minimal free resolution of an arbitrary stable ideal @xmath9 using the theory of _ poset resolutions _ , introduced by the author in @xcite .
specifically , we define a poset @xmath10 on the admissible symbols of eliahou and kervaire by taking advantage of a decomposition property unique to the monomials contained in stable ideals . in our first main result ,
theorem [ stabletheorem ] , we recover the eliahou - kervaire minimal free resolution of a stable ideal as a poset resolution .
the value of this technique is that the maps in the resolution are shown to act on the basis elements of the free modules in a way that mirrors the covering relations present in @xmath11 .
taking the lattice - linear ideals of @xcite into consideration , poset resolutions therefore provide a common perspective from which to view the minimal resolutions of three large and well - studied classes of monomial ideals ; stable ideals , scarf ideals @xcite and ideals having a linear resolution @xcite .
an additional advantage of the method described herein is that for a fixed stable ideal , combinatorial information from the poset of admissible symbols is transferred to topological information of a cw complex .
specifically , the poset @xmath11 is a _
cw poset _ in the sense of bjrner @xcite , which allows its interpretation as the face poset of a regular cw complex @xmath12 . in our second main result ,
theorem [ stablecellulartheorem ] , we show that @xmath12 supports a minimal cellular resolution of the stable ideal @xmath9 . by utilizing this combinatorial connection , we are able to provide a minimal cellular resolution of @xmath9 in a considerably more explicit manner than in three previous methods ; the construction of batzies and welker @xcite which uses the tools of discrete morse theory , the result due to mermin @xcite that the eliahou - kervaire resolution is cellular and the technique of sinefakopoulos @xcite which produces a polyhedral cell complex that supports the minimal free resolution of ( the more restricted class of ) borel - fixed ideals generated in one degree . in addition ,
corso and nagel in @xcite , construct cellular resolutions of strongly stable ideals generated in degree two and separately , nagel and reiner in @xcite , construct cellular resolutions of strongly stable ideals generated in one degree . to the author s knowledge ,
connections between these methods and the one contained in this paper have not been formally studied . in section [ pres ] , we review the fundamentals of poset resolutions and define the poset of admissible symbols @xmath11 .
section [ pnshellable ] provides the details of the ( dual ) shellability of @xmath11 . the intrinsic cw poset structure and properties of the sequence of vector spaces and maps which is produced by the methods of @xcite
are exhibited in section [ propertiesofcpn ] .
the proof of theorem [ stabletheorem ] is given in section [ prstableproof ] and section [ stablecellular ] contains the description of the regular cw complex @xmath12 and the proof of theorem [ stablecellulartheorem ] . throughout the paper
, we assume that the topolological and combinatorial notions of posets , order complexes , cw complexes and face posets are familiar to the reader .
many thanks to jeff mermin , whose comments were of considerable help in improving the clarity of the exposition and in correcting an omission of certain cases which appear in the proofs of theorem [ pnshellable ] and lemma [ cyclelemma ] .
thanks as well to amanda beecher and alexandre tchernev for helpful questions and discussions .
[ pres ] let @xmath13 be a finite poset with set of atoms @xmath14 and write @xmath15 if @xmath16 and there is no @xmath17 such that @xmath18 .
we say that @xmath19 is _ covered by _
@xmath20 in this situation . for @xmath21 , write the order complex of the associated open interval as @xmath22 . in @xcite ,
the collection of simplicial complexes @xmath23 is used to construct a sequence of vector spaces and vector space maps @xmath24 in this construction , @xmath25 , a one - dimensional vector space . for @xmath26 ,
the vector space @xmath27 is defined as @xmath28 where @xmath29 . in particular
, the vector space @xmath30 has its nontrivial components indexed by the set of atoms @xmath14 , and the map @xmath31 is defined componentwise as @xmath32 . for @xmath33 , the maps @xmath34 are defined using the mayer - vietoris sequence in reduced homology for the triple @xmath35 where @xmath36 $ ] for all @xmath37 . for notational simplicity , when @xmath37 let @xmath38 so that @xmath39 is the map induced in homology by the inclusion map and @xmath40 is the connecting homomorphism from the mayer - vietoris sequence in homology of ( [ mvseq ] ) . for @xmath33 the map @xmath41 is defined componentwise by @xmath42 where @xmath43 is the composition @xmath44 .
we now describe the process by which the sequence of vector spaces @xmath45 is transformed into a sequence of multigraded modules . for a monomial @xmath46
we write @xmath47 and @xmath48 for @xmath49 .
assuming the existence of a map of partially ordered sets @xmath50 , the sequence of vector spaces @xmath45 is _ homogenized _ to produce @xmath51 a sequence of free multigraded @xmath2-modules and multigraded @xmath2-module homomorphisms .
indeed , homogenization of @xmath45 is carried out by constructing @xmath52 and multigrading the result with @xmath53 for each @xmath54 .
similarly , for @xmath26 , we set @xmath55 where the grading is defined as @xmath56 for each @xmath57 . the differential in this sequence of multigraded modules is defined componentwise in homological degree 1 as @xmath58 and for @xmath26 , the map @xmath59 is defined as @xmath60 where @xmath61 takes the form @xmath62 for @xmath37 .
the sequence @xmath63 approximates a free resolution of the multigraded module @xmath64 where @xmath65 is the ideal in @xmath2 generated by the monomials @xmath66 whose multidegrees are given by the images of the atoms of @xmath67 .
@xcite if @xmath63 is an acyclic complex of multigraded modules , then we say that it is a _ poset resolution _ of the ideal @xmath65 . throughout the remainder of the paper
, @xmath9 will denote a stable monomial ideal in @xmath2 and we write @xmath68 as the unique minimal generating set of @xmath9 . for a monomial @xmath69 set @xmath70 and @xmath71 to describe further the class of stable ideals , let @xmath72=\{1,\ldots , d-1\}$ ] , for @xmath73 $ ] let @xmath74 and write @xmath75 . in lemma 1.2 of @xcite , eliahou and kervaire
prove that a monomial ideal @xmath9 is stable if and only if for each monomial @xmath76 there exists a unique @xmath77 with the property that @xmath78 and @xmath79 .
we adopt the language and notation introduced in the paper of eliahou and kervaire , and refer to @xmath80 as the _ unique decomposition _ of the monomial @xmath81 .
following their convention , we encode this property in a _ decomposition map _ where @xmath82 is the collection of monomials of @xmath9 and @xmath83 .
an _ admissible symbol _ is an ordered pair @xmath84 which satisfies @xmath85 , where @xmath86 and @xmath73 $ ] .
the _ poset of admissible symbols _ is the set @xmath11 of all admissible symbols associated to @xmath9 , along with the symbol @xmath87 which is defined to be the minimum element of @xmath11 .
the partial ordering on @xmath11 is @xmath88 when both symbols are admissible . in the case
when @xmath89 and @xmath90 for some @xmath91 , then we write @xmath92 to describe the covering that occurs in @xmath11 .
as constructed , we have @xmath93 for every @xmath86 .
we are now in a position to state our first main result .
[ stabletheorem ] suppose that @xmath9 is a stable monomial ideal with poset of admissible symbols @xmath11 and define the map @xmath94 so that @xmath95
. then the complex @xmath63 is a minimal poset resolution of @xmath96 . in order to prove theorem [ stabletheorem ]
, we first describe the combinatorial structure of @xmath11 and then exhibit the connection between the complex @xmath63 and the minimal free resolution of the stable ideal @xmath9 constructed by eliahou and kervaire in @xcite .
[ pnshellable ] we begin this section by recalling some general facts regarding the shellability of partially ordered sets .
recall that a poset @xmath67 is called _ shellable _ if the facets of its order complex @xmath97 can be arranged in a linear order @xmath98 in such a way that the subcomplex @xmath99 is a nonempty union of maximal proper faces of @xmath100 for @xmath101 .
such an ordering of facets is called a _
shelling_. let @xmath102 denote the collection of edges in the hasse diagram of a poset @xmath67 . an _ edge labeling _ of @xmath67 is a map @xmath103 where @xmath104 is some poset . for @xmath105 , a maximal chain of @xmath67 , the _ edge label _ of @xmath106 is the sequence of labels @xmath107 .
an edge labeling @xmath108 is called an _ el - labeling ( edge lexicographical labeling ) _
if for every interval @xmath109 $ ] in @xmath67 , 1 .
there is a unique maximal chain @xmath106 in @xmath109 $ ] , such that the labels of @xmath106 form an increasing sequence in @xmath104 .
we call @xmath106 the unique increasing maximal chain in @xmath109 $ ] .
@xmath110 under the lexicographic partial ordering in @xmath104 for all other maximal chains @xmath111 in @xmath109 $ ] .
a graded poset that admits an el - labeling is said to be _ el - shellable ( edge lexicographically shellable)_. we further recall the following fundamental result of bjrner and wachs . @xcite el - shellable posets are shellable .
we now define an edge labeling of the poset of admissible symbols @xmath11 .
[ elshelling ] let @xmath112 take the form @xmath113 where @xmath114 .
[ posetexample ] the labeled hasse diagram for the poset of admissible symbols , @xmath11 , of the stable ideal @xmath115 is recall that given a poset @xmath67 , the _ dual poset _
@xmath116 has an underlying set identical to that of @xmath67 , with @xmath117 in @xmath116 if and only if @xmath118 in @xmath67 .
further , an edge labeling of a poset @xmath67 may also be viewed as an edge labeling of its dual poset and @xmath67 is said to be _ dual shellable _ if @xmath116 is a shellable poset .
[ pnshellable ] the poset @xmath11 is dual el - shellable with @xmath108 defined as above . before turning to the proof of theorem [ pnshellable ] , we discuss some properties of the decomposition map @xmath119 and the edge labeling @xmath108 .
[ stableremarks]@xmath120 1 .
( * ? ? ? * lemma 1.3 ) for any monomial @xmath121 and any monomial @xmath76 , we have @xmath122 and @xmath123 .
we refer to the first property as the _ associativity _ of @xmath119 .
2 . suppose that @xmath124 $ ] is a closed interval in the dual poset @xmath125 .
given a sequence of labels @xmath126 there is at most one maximal chain @xmath106 in the closed interval such that @xmath127 when it exists , this chain must be equal to @xmath128 where @xmath129 , the set @xmath130 and @xmath131 for @xmath132 with @xmath133 and @xmath134 .
suppose that @xmath135 is a pair of comparable admissible symbols .
then @xmath136 for some @xmath137 .
let @xmath138 .
then by associativity and ( * ? ? ?
* lemma 1.2 ) we have @xmath139 . in this way , any representation of @xmath136 may be reduced to @xmath140 under the conditions above .
implicit in all subsequent arguments is the convention that a representation @xmath140 is written in reduced form .
[ uniquerep ] for a reduced representation of @xmath140 the set @xmath141 is the unique subset of minimum cardinality among all @xmath137 for which @xmath136 .
suppose that @xmath141 is not the subset of @xmath142 with smallest cardinality , namely that there exists @xmath143 with @xmath144 and @xmath145 . by definition , @xmath146 and @xmath147 where @xmath79 and @xmath148 .
the assumption of @xmath144 implies that there exists @xmath149 such that @xmath150 .
rearranging and combining the two equations above , we arrive at the equality @xmath151 .
this equality allows us to conclude that @xmath152 divides @xmath153 since it can not divide @xmath154 . by definition , @xmath79 and therefore @xmath155
further , since we assumed that @xmath156 possessed the property that @xmath157 we have @xmath158 so that @xmath159 .
this equality also has implications for @xmath154 and @xmath160 , namely that @xmath161 and @xmath162 so that @xmath163 since @xmath150 .
however , @xmath164 is a contradiction , and our original supposition that such a @xmath165 exists is false .
if @xmath141 and @xmath165 are distinct subsets of @xmath142 with @xmath166 and @xmath167 then there is a @xmath149 and @xmath168 for which @xmath150 and @xmath169 .
as before , we use the equality @xmath151 and now conclude that @xmath152 divides @xmath153 and @xmath170 divides @xmath160 .
therefore , @xmath171 and similarly @xmath172 so that @xmath173 , @xmath174 to prove the dual el - shellability of @xmath11 , recall that for the poset of admissible symbols @xmath11 , we have comparability in the dual poset given by @xmath175 if and only if @xmath176 .
we proceed with the proof by considering the various types of closed intervals that appear in the dual poset @xmath177 .
_ case 1 : _ consider the closed interval @xmath178 $ ] .
write @xmath179 so that @xmath180 , for every @xmath181 .
the maximal chain @xmath182 has the increasing label @xmath183 consider a maximal chain @xmath184 $ ] where @xmath185 .
if each label in the sequence @xmath186 ( except the label of coverings of the form @xmath187 ) is negative , then the sequence @xmath186 can not be increasing , for it must be a permutation of the sequence @xmath188 where the rightmost label @xmath189 is fixed .
if any label within the sequence @xmath186 is positive , then again @xmath186 can not be increasing since every maximal chain contains the labeled subchain @xmath190 for every @xmath191 .
therefore , @xmath106 is the unique rising chain in the interval @xmath178 $ ] .
further , @xmath188 is lexicographically first among all chains in @xmath178 $ ] since @xmath192 .
_ case 2 : _ consider the closed interval @xmath124 $ ] of @xmath177 where @xmath193 and @xmath194 .
again write @xmath195 such that @xmath196 .
every maximal chain @xmath106 in @xmath197 $ ] has a label of the form @xmath198 where @xmath199 is a permutation of the set @xmath200 .
therefore , the label @xmath201 corresponding to the identity permutation is the unique increasing label in @xmath197 $ ] and is lexicographically first among all such labels .
_ case 3 : _ consider the closed interval @xmath124 $ ] of @xmath177 where @xmath193 and @xmath202 . by lemma [ uniquerep ] , @xmath203 for a unique @xmath204 where @xmath205 and the set @xmath141 is of minimum cardinality .
writing the set @xmath206 and @xmath207 where @xmath208 and @xmath209 , it follows that the sequence of edge labels @xmath210 is the increasing label of a maximal chain @xmath106 in @xmath124 $ ] .
turning to uniqueness , suppose that @xmath185 is also a chain which has a rising edge label .
then @xmath211 where @xmath212 and @xmath213 . since @xmath185 , then @xmath214 and in particular , @xmath215 .
if there exists @xmath216 with the property that @xmath217 , we must have @xmath218 so that @xmath219 for some @xmath220 and the label @xmath188 has the form @xmath221 by the definition of @xmath119 , we have the equalities @xmath146 and @xmath222 , which may be combined and simplified to arrive at the equation @xmath223 .
the assumption that @xmath224 and @xmath225 implies that @xmath226 divides @xmath160 so that @xmath227 .
it therefore follows that @xmath228 when @xmath229 so that @xmath230 , which is a contradiction the admissibility of the symbol @xmath231 . if @xmath232 then @xmath233 and therefore @xmath234 which implies that the symbol @xmath235 which would neccesarily precede @xmath236 in the chain is not admissible .
if there exists @xmath237 with @xmath238 then a similar argument again provides a contradiction to admissibility .
we now prove that @xmath188 is lexicographically smallest among all chains .
aiming for a contradiction , suppose that the label @xmath188 is not lexicographically smallest so that there exists a maximal dual chain @xmath239 with @xmath240 .
without loss of generality , we may assume that @xmath188 and @xmath186 differ at their leftmost label @xmath241 , where @xmath242 .
such a @xmath243 must be an element of the set @xmath141 since @xmath244 is inherent in the structure of @xmath188 . by construction
, @xmath149 implies that @xmath157 and utilizing the equations @xmath146 and @xmath222 , to produce @xmath223 , it follows that @xmath152 divides @xmath153 and therefore @xmath245 so that @xmath159 .
this forces the element @xmath246 for otherwise , the chain @xmath106 would contain the subchain @xmath247 where @xmath248 is not an admissible symbol .
the desired contradiction will be obtained within an investigation of each of the three possibilities for the relationship between @xmath249 and @xmath250 .
suppose @xmath251 so that @xmath252 , based upon the structure of the set @xmath253 and the definition of the decomposition map @xmath119 . in this case , the chain @xmath239 can not end in @xmath236 since @xmath241 , the leftmost label of @xmath239 , labels the subchain @xmath254 and the @xmath152 degree of every monomial appearing in the chain @xmath239 may not increase .
if @xmath255 then the unique decomposition @xmath256 implies that @xmath152 divides @xmath160 , for otherwise @xmath149 implies that @xmath252 , a contradiction . the conclusion that @xmath152 divides @xmath160 allows @xmath256 to be simplified to @xmath257 where @xmath258 and @xmath259
this contradicts the condition that @xmath141 is the set of smallest cardinality for which @xmath260 .
lastly , if @xmath261 we turn to the chain @xmath106 , whose rightmost label is @xmath243 .
the subchain with this label is @xmath262 where @xmath263 where @xmath80 does not contain this new factor of @xmath152 .
the monomial @xmath152 therefore divides @xmath153 and we can reduce @xmath263 to @xmath264 where @xmath259 , a contradiction to @xmath265 .
this completes the proof . with theorem [ pnshellable ] established
, we immediately have the following corollary .
[ finshell ] every interval of @xmath11 which is of the form @xmath266 $ ] is finite , dual el - shellable and therefore shellable .
[ propertiesofcpn ] to establish the connection between the poset @xmath11 and the sequence @xmath267 , we recall the definition of _ cw poset _ , due to bjrner @xcite .
@xcite a poset @xmath67 is called a _ cw poset _ if 1 .
@xmath67 has a least element @xmath268 , 2 .
@xmath67 is nontrivial ( has more than one element ) , 3 . for all @xmath269 , the open interval
@xmath270 is homeomorphic to a sphere . after establishing this definition ,
bjrner describes sufficient conditions for a poset to be a cw poset .
* proposition 2.2 ) suppose that @xmath67 is a nontrivial poset such that 1 .
@xmath67 has a least element @xmath268 , 2 . every interval @xmath109 $ ] of length two has cardinality four , 3 . for every @xmath271
the interval @xmath272 $ ] is finite and shellable .
then @xmath67 is a cw poset . with this proposition in hand
, we now may conclude the following about the structure of @xmath11 , the poset of admissible symbols .
[ pncwposet ] the poset of admissible symbols @xmath11 is a cw poset .
the poset @xmath11 has a least element by construction and each of its intervals @xmath266 $ ] is finite and shellable by corollary [ finshell ] .
thus , it remains to show that every closed interval in @xmath11 of length two has cardinality four .
_ case 1 : _ let @xmath273 so that the set @xmath253 is a singleton .
it follows that the only poset elements in the open interior of the interval are @xmath274 and @xmath275 .
_ case 2 : _ let @xmath193 and suppose that @xmath276 $ ] is a closed interval of length two in the poset of admissible symbols , @xmath11 . since the interval is of length two , the set @xmath277 has the form @xmath278 for some @xmath279 .
further , any poset element in the interval must have either @xmath280 or @xmath281 as its first coordinate , for these sets are the only subsets of @xmath253 which contain @xmath278 .
write @xmath282 where @xmath283
. we must now consider each of the possible orderings for the elements of the ( multi ) set @xmath284 to ascertain the choices available for the monomial @xmath80 .
our assumptions of the inequalities @xmath285 and @xmath286 together with the admissibility of the symbol @xmath84 imply that @xmath287 .
hence , determining the number of orderings amounts to producing a count of the number of orderings for elements of the set @xmath288 , of which there are three , since @xmath285 .
_ subcase 2.1 : _ suppose that @xmath289 . if @xmath194 , then the poset elements which are contained in the open interior of the interval are forced to be @xmath290 and @xmath291 . if @xmath292 and @xmath293 then the symbol @xmath294 is admissible , so that it is in the open interior of the interval along with the admissible symbol @xmath291 .
the symbol @xmath290 is not comparable to @xmath294 due to the absence of the value @xmath295 .
the symbol @xmath296 is also not comparable to @xmath294 , for if it were then either @xmath297 or @xmath298 . the first equality is impossible since lemma [ uniquerep ] guarantees that @xmath299 is the unique set containing one element for which @xmath292 .
the second equality also can not occur since lemma 1.2 of @xcite guarantees monomial equality @xmath300 if and only if @xmath301 , which would contradict the assumption that @xmath302 is an admissible symbol . if @xmath292 and @xmath303 then the symbol @xmath294 is not admissible and is not an element of @xmath11 .
however , we are assuming that the symbol @xmath304 is admissible , so that @xmath305 and via lemma 1.2 of @xcite , we have the monomial equality @xmath306 .
therefore , @xmath307 and the symbols @xmath291 and @xmath296 are each contained in the interval . since @xmath292 , the symbol
@xmath290 is not comparable to @xmath294 . if @xmath308 then the symbol @xmath290 is certainly contained in the closed interval
further , @xmath296 must be admissible for were it not , then the assumption of admissibility for @xmath309 implies that @xmath310 , a contradiction to the initial stipulation that @xmath285 .
the symbol @xmath291 is incomparable to @xmath309 and were @xmath294 comparable to @xmath309 , then either @xmath311 or @xmath298 .
the first equality contradicts lemma [ uniquerep ] and the second may be used to arrive at a contradiction to the admissibility of @xmath296 .
these arguments are similar to those used in when @xmath292 and @xmath293 .
if @xmath312 and @xmath313 then the symbols @xmath294 and @xmath296 are admissible and are contained in the open interior of the interval .
clearly , the symbols @xmath290 and @xmath291 are not comparable to @xmath314 in this instance .
if @xmath312 and @xmath315 , then we are reduced to an already resolved case . for each of these four choices of @xmath80 , the interval has four elements .
_ subcase 2.2 _ we now consider the two remaining orderings @xmath316 and @xmath317 . under each of these orderings
, we have @xmath318 and in light of lemma 1.3 of @xcite if @xmath319 we have @xmath320 and in turn that @xmath321 .
if @xmath194 , then the poset elements which are contained in the open interior of the interval are forced to be @xmath290 and @xmath291 . if @xmath292 then @xmath321 implies that the symbol @xmath294 is not admissible and is not an element of @xmath11 .
however , we are assuming that the symbol @xmath304 is admissible , so that @xmath305 and again using lemma 1.2 of @xcite , we have @xmath322 .
therefore , @xmath307 and the symbols @xmath291 and @xmath296 are each contained in the interval . since @xmath292 , the symbol
@xmath290 is not comparable to @xmath294 . if @xmath308 then the symbol @xmath290 is certainly contained in the closed interval
further , @xmath296 must be admissible for were it not , then the assumption of admissibility for @xmath309 implies that @xmath323 , a contradiction to the initial stipulation that @xmath285 .
the symbol @xmath291 is incomparable to @xmath309 and were @xmath294 comparable to @xmath309 , then either @xmath311 or @xmath298 .
the first equality contradicts lemma [ uniquerep ] and the second may be used to arrive at a contradiction to the admissibility of @xmath296 .
these arguments are similar to those used in the case when @xmath292 and @xmath293 . again
, for each of these three choices of @xmath80 , the interval has four elements .
we now analyze the vector spaces which are present in the sequence @xmath267 at the level of individual poset elements . in order to do so , we recall the following combinatorial results . as is standard , we write @xmath324 .
if a bounded poset @xmath67 is el - shellable , then the lexicographic order of the maximal chains of @xmath67 is a shelling of @xmath97 .
moreover , the corresponding order of the maximal chains of @xmath325 is a shelling of @xmath326 .
[ bjwachselbasis ] suppose that @xmath67 is a poset for which @xmath327 admits an el - labeling
. then @xmath67 has the homotopy type of a wedge of spheres .
furthermore , for any fixed el - labeling : 1 .
@xmath328 , 2 .
bases for @xmath329-dimensional homology ( and cohomology ) are induced by the falling chains of length @xmath330 . in the analysis that follows
, we again examine the dual poset @xmath177 and focus our attention on the collection of closed intervals of the form @xmath331 $ ] , to each of which we apply theorem [ bjwachselbasis ] . indeed , for each admissible symbol @xmath332 where @xmath333 , the open interval @xmath334 is homeomorphic to a sphere of dimension @xmath335 since @xmath11 is a cw poset .
further , the el - labeling of @xmath331 $ ] guarantees that the unique generator of @xmath336 is induced by a unique falling chain of length @xmath337 . in the discussion that follows , we use the el - shelling of definition [ elshelling ] to produce a canonical generator of @xmath336 as a linear combination in which each facet of @xmath338 occurs with coefficient @xmath339 or @xmath340 . to begin ,
consider a maximal chain @xmath341 which is of length @xmath337 and appears in the dual closed interval @xmath178 $ ] and write the label of this chain as @xmath342 we note that @xmath343 and write @xmath344 where @xmath345 is the permutation arranging the sequence @xmath346 in _ increasing _ order .
we endow the corresponding chain @xmath106 in @xmath347 with this sign @xmath348 and refer to it as the _ sign _ of @xmath106 .
the unique maximal chain @xmath239 in @xmath178 $ ] which has a decreasing label is the chain consisting of admissible symbols having at each stage a different monomial as their second coordinate and the sequence of sets @xmath349 as their first coordinate .
the unique falling chain @xmath184 $ ] is therefore @xmath350 where @xmath351 with @xmath352 and for @xmath353 , the set @xmath354 and the monomial @xmath355 .
the label of the chain @xmath239 is therefore @xmath356 and is decreasing .
if there were another such chain with decreasing label , then such a chain would be counted by theorem [ bjwachselbasis ] and @xmath357 would not have the homotopy type of a sphere , a contradiction to the fact that @xmath11 is a cw poset . in the context of the shelling order produced by the el - shelling above , the chain @xmath239 appears lexicographically last among all maximal chains in the dual interval and is therefore the unique homology facet of @xmath338 . for an admissible symbol @xmath84 , set @xmath358 the linear combination of all maximal chains of the open interval @xmath347 with coefficients given by ( [ facetsigns ] )
viewing the maximal chains of @xmath347 as facets in the order complex @xmath338 we now establish the following .
[ cyclelemma ] the sum @xmath359 is a @xmath335-dimensional cycle in @xmath336 which is not the boundary of any @xmath360-dimensional face .
the maximal chains in the open interval @xmath347 are each of length @xmath335 , so that no @xmath360-dimensional faces are present in @xmath338 .
thus , @xmath359 can not be the boundary of a @xmath360-dimensional face of @xmath338 .
we now show that @xmath359 is a @xmath335-dimensional cycle .
suppose that @xmath106 is a maximal chain in @xmath347 and let @xmath236 be an element of said chain .
we exhibit a unique chain @xmath111 which also appears in @xmath359 and differs from @xmath106 only at the element @xmath236 .
indeed , consider the chain @xmath341 along with its subchain @xmath361 . in the proof of theorem [ pncwposet
] , each closed interval of length two was shown to be of cardinality four , and therefore there exists a unique @xmath362 $ ] which is not equal to @xmath236 . defining @xmath111 by removing @xmath236 and replacing it with @xmath363 , we have constructed the desired chain .
we claim that for the chains @xmath106 and @xmath111 , the associated signs @xmath348 and @xmath364 are opposite to one another . if @xmath365 then @xmath366 for some @xmath367 .
thus , @xmath368 and @xmath369 so that the chains @xmath106 and @xmath111 have the same corresponding permutation @xmath370 .
since either @xmath371 or @xmath372 , without loss of generality we assume that @xmath371 so that @xmath373 .
therefore , the subchain @xmath374 has @xmath375 as its label , while @xmath376 has @xmath367 as its label .
this is the only difference in the labels @xmath188 and @xmath377 and @xmath378 is forced .
if @xmath379 then for each case that appears in the classification of intervals of length two described in the proof of theorem [ pncwposet ] , we can compute @xmath378 . when the differential @xmath380 in the reduced chain complex @xmath381 is applied to the sum @xmath359 , each term appears twice with opposite signs , so that @xmath382 making @xmath359 a @xmath335-dimensional cycle in @xmath336 as claimed .
[ prstableproof ] with the choice for the bases of the vector spaces in @xmath267 established , we now turn to the proof that the poset @xmath11 supports the minimal free resolution of @xmath96 .
we first analyze the action of the differential of @xmath267 when it is applied to an arbitrary basis element @xmath359 .
[ differentialaction ] the map @xmath383 sends a basic cycle @xmath359 to the element @xmath384 , where @xmath351 , the relationship @xmath385 holds and @xmath386 write @xmath380 for the simplicial differential in the reduced chain complex @xmath381 .
the open interval @xmath347 may be realized as the union of half - closed intervals @xmath387 , so that the order complex of each half - closed interval is a cone with apex @xmath236 . applying the differential to the sum of all facets contained in the interval produces the boundary of the cone , which in this case is the order complex of @xmath388 .
indeed , when @xmath380 is applied to the sum @xmath389 the faces in which the element @xmath236 remains appear twice and have opposite signs as described in the proof of lemma [ cyclelemma ] .
thus , the only faces that remain in the expansion of @xmath390 are of the form @xmath391 .
precisely , @xmath392\\ & = & \left [ { \displaystyle}\sum_{\sigma\in\left[(j , n),\hat{0}\right ) } \varepsilon_\sigma\cdot\bar{\sigma } \right ] \nonumber \\ & = & \left [ { \displaystyle}\sum_{\bar{\sigma}\in\left((j , n),\hat{0}\right ) } \varepsilon_\sigma\cdot\bar{\sigma } \right].\nonumber\end{aligned}\ ] ] the facet @xmath393 has an associated permutation @xmath394 , and using elementary properties of permutation signs , we have @xmath395 , where @xmath385 . considering the definition of @xmath348 , for each @xmath236 for which @xmath396 we now have @xmath397 since @xmath398 if @xmath319 and @xmath399 if @xmath194 .
therefore , equation ( [ phimapstable ] ) becomes @xmath400\\ & = & \left [ { \displaystyle}\sum_{\bar{\sigma}\in\left((j , n),\hat{0}\right ) } ( -1)^{p+\delta_{m , n}}\cdot\varepsilon_{\bar{\sigma}}\cdot\bar{\sigma } \right]\\ & = & ( -1)^{p+\delta_{m , n } } \cdot\left [ { \displaystyle}\sum_{\bar{\sigma}\in\left((j , n),\hat{0}\right ) } \varepsilon_{\bar{\sigma}}\cdot\bar{\sigma } \right]\\ & = & ( -1)^{p+\delta_{m , n}}\cdot f(j , n)\end{aligned}\ ] ] which proves the lemma . as described in section [ pres ] , the map @xmath401 is defined componentwise on the one - dimensional @xmath1-vectorspace @xmath402 for each poset element @xmath84 . using the conclusion of lemma [ differentialaction ]
, we immediately have @xmath403 where @xmath351 and @xmath404 and @xmath405 .
recall that the poset map @xmath94 is defined as @xmath406 , so that we can homogenize the sequence of vector spaces @xmath267 to produce @xmath407 a sequence of multigraded modules .
more precisely , for @xmath408 and a poset element @xmath409 where @xmath351 and @xmath404 , the differential @xmath410 acts on a basis element @xmath359 of the free module @xmath411 via the formula @xmath412 where @xmath413 takes the same value as in [ phimap ] , so that @xmath414 . it remains to show that @xmath63 is a minimal exact complex , and to do so we identify it as the eliahou - kervaire resolution .
[ ekres ] the _ eliahou - kervaire minimal free resolution _ @xcite of a stable ideal @xmath9 is @xmath415 where @xmath416 is the free module of rank one with basis @xmath417 and for @xmath408 , @xmath418 has as basis the admissible symbols @xmath419 when applied to a basis element , the differential of @xmath420 takes the form @xmath421 where we define @xmath422 when @xmath423 ( i.e. the symbol is inadmissible ) .
we now are in a position to prove the main result of this paper . the eliahou - kervaire admissible symbols index the multigraded free modules in the complexes @xmath420 and @xmath63 and therefore the generators of these modules are in one to one correspondence with one another .
further , comparing definition [ ekres ] and equation [ presmap ] , @xmath424 and @xmath410 have identical behavior on basis elements .
the minimality and exactness of @xmath420 implies the minimality and exactness of @xmath63 so that @xmath63 is a minimal poset resoution of @xmath96 .
[ stablecellular ] in this section we exhibit a minimal cellular resolution for an arbitrary stable monomial ideal .
cellular resolutions of monomial ideals have received considerable attention in the literature and recently mermin @xcite has shown that the eliahou - kervaire resolution @xmath420 is cellular using methods distinct from those depicted here .
the techniques of discrete morse theory have also been used by batzies and welker @xcite to produce a minimal cellular resolution of stable modules , which contain the class of stable ideals .
more generally , methods for determining whether a given monomial ideal admits a minimal cellular ( or cw ) resolution remain an open question , although velasco @xcite has shown that there exist monomial ideals whose minimal free resolutions are not even supported on a cw complex .
the technique described here is an example of a more general approach which interprets cw resolutions of monomial ideals through the theory of poset resolutions .
this approach is described in @xcite , and is distinct from both the method of @xcite concerning stable modules and the method of @xcite which is specific to stable ideals .
we begin by recalling a fundamental result due to bjrner . in the case of the poset of admissible symbols
@xmath11 , we interpret bjrner s proof explicitly to produce the corresponding regular cw complex @xmath12 . on the level of cells , @xmath425 corresponds to the empty cell and each admissible symbol @xmath84 of @xmath11 corresponds to a closed cell @xmath426 of dimension @xmath427 for which @xmath428 $ ] .
taking @xmath429 we have an isomorphism of posets @xmath430 .
the regular cw complex @xmath12 also comes equipped with a @xmath431 grading by realizing the map @xmath94 of theorem [ stabletheorem ] as a map @xmath432 where a cell @xmath433 .
the stable ideal @xmath115 has minimal resolution supported by @xmath12 , the regular cw complex depicted below , which has six 0-cells , eight 1-cells and three 2-cells .
the face poset of this cell complex @xmath434 is isomorphic to the poset of admissible symbols @xmath11 given in example [ posetexample ] . 1 . for all @xmath439 , the free module @xmath440 has as its basis the @xmath441 dimensional cells of @xmath438 .
2 . for a basis element @xmath442 ,
one has @xmath443 , 3 .
the differential @xmath444 of @xmath436 acts on a basis element @xmath445 as @xmath446 where @xmath447 is the coefficient of the cell @xmath448 in the differential of @xmath449 in the cellular chain complex of @xmath438 . conditions 1 and 2 of definition [ cellresdef ] are clear from the structure of @xmath12 , its correspondence to the poset @xmath11 and the construction of the resolution @xmath63 .
it therefore remains to verify that condition 3 is satisfied .
the main result in @xcite provides a canonical isomorphism between the complex @xmath267 and @xmath450 , the cellular chain complex of @xmath12 .
therefore , the differential of @xmath63 satisfies condition 3 .
|
we use the theory of poset resolutions to give an alternate construction for the minimal free resolution of an arbitrary stable monomial ideal in the polynomial ring whose coefficients are from a field .
this resolution is recovered by utilizing a poset of eliahou - kervaire admissible symbols associated to a stable ideal .
the structure of the poset under consideration is quite rich and in related analysis , we exhibit a regular cw complex which supports a minimal cellular resolution of a stable monomial ideal .
[ intheorem]lemma [ intheorem]proposition [ intheorem]corollary [ intheorem]definition [ intheorem]remark [ intheorem]notation [ equation]theorem [ equation]lemma [ equation]proposition [ equation]corollary [ equation]observation [ equation]conjecture [ equation]fact [ equation]example [ equation]definition [ equation]remark [ equation]remarks [ equation]notation
|
several studies have shown that at _ z _ @xmath3 2 a considerable fraction of the massive galaxies ( stellar mass @xmath7 @xmath3 10@xmath8 @xmath9 ) are compact compared to their local counterparts ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
the rarity of compact massive galaxies at the present time implies a considerable size increase in the last 10 billion years ( @xcite ; @xcite ; @xcite ; @xcite ; but see @xcite ; @xcite ; @xcite ; @xcite ) .
recent comprehensive simulations have found that the commonly used methods for measuring the sizes of these galaxies , such as fitting single - component srsic ( 1968 ) function , is reliable ( e.g. , @xcite ; @xcite ; davari et al .
2016 ) , despite the fact that , in many instances , their sizes are comparable to the scale of the _ hubble space telescope ( hst ) _ point - spread function ( psf ) .
the compactness of high-_z _ massive galaxies strongly suggests that their formation process involved strong dissipation on rapid timescales ( e.g. , @xcite ) .
this can be accomplished by gas - rich major mergers ( e.g. , barnes & hernquist 1992 ) , cold gas flows ( dekel et al .
2009 ) , or some combination of the two . in support of such a scenario , the central regions of local massive ellipticals , the likely descendants of high-@xmath5 compact , massive galaxies , are old and have a high @xmath10/fe abundance ratio ( @xcite ) .
this indicates an early episode of violent star formation , which would naturally accompany a gas - rich , dissipative formation event .
although major mergers have long been thought to transform disky galaxies to bulge - dominated systems ( @xcite ; @xcite ) , more recent simulations show that this may not be always the case .
in fact , gas - rich major mergers can leave large - scale disks ( @xcite ; @xcite ) if the gas retains significant angular momentum during the merger ( @xcite ) , especially those that have a high gas fraction ( @xcite ) .
the study of @xcite lends credence to this picture .
these authors show that massive , evolved , compact galaxies at @xmath0 @xmath3 2 the so - called red nuggets are the direct descendants of the submillimeter galaxies ( smgs ; blain et al .
2002 ) at @xmath0 @xmath1 3 .
smgs are among the most luminous , rapidly star - forming galaxies known , with luminosities greater than 10@xmath11 @xmath12 and star formation rates of @xmath13 @xmath14 yr@xmath15 ( e.g. , @xcite ; @xcite ; @xcite ) . indeed ,
@xcite show that the mass - size distribution and the mean stellar mass surface density of these two classes of high - redshift galaxies are similar .
both types are best fit by low srsic indices ( @xmath4 ) . moreover , from a co study of 30 local merger remnants , ueda et al .
( 2014 ) find that the majority of the sources exhibit kinematic signatures of rotating molecular gas disks .
furthermore , @xcite conclude that more than 95% of smgs have pure stellar disks or disk - dominated stellar structures ; the distribution of axial ratios ( their figure 6 ) rejects the possibility that the sample is bulge - dominated . the above arguments strongly suggest that high-@xmath5 massive galaxies should host large - scale stellar disks .
this hypothesis is attested by a number of studies . from the work of van der wel et al .
( 2011 ) , 50% of massive galaxies at _ z _ @xmath1 2 are disk - dominated .
similarly , @xcite find that massive galaxies at _ z _ @xmath1 1 have higher axial ratios than their lower redshift counterparts , broadly consistent with the tendency for galaxies to become noticeably rounder between @xmath16 and 0 @xcite .
now the question remains : how have the red nuggets , which most likely contained a significant disk component at @xmath0 @xmath3 2 turn , into local giant ellipticals like m87 , which demonstrably do _ not _ have a disk ?
we aim to trace this morphological transition .
we do so by performing detailed two - dimensional modeling of the optical light distribution of massive galaxies within 0.5 @xmath17 @xmath0 @xmath17 2.5 .
besides fitting a traditional , simple single - component srsic function , when possible , we perform a bulge+disk decomposition of these massive systems . examining separately the bulge and disk structural properties , plus
the luminosity bulge - to - total ratio ( @xmath6 ) , provides key indicators that can be missed by studying potentially multiple - component galaxies as a single system .
for instance , from the comprehensive morphological analysis by @xcite , massive galaxies appear to transit from disk - dominated to bulge - dominated between @xmath18 and 1 , with elliptical - like systems emerge at lower redshifts . the bulge+disk decomposition carried out by bruce et al .
was done by fixing the srsic index of the bulge to @xmath19 and of the disk to @xmath20 . in other words ,
all bulges were assymed to follow a de vaucouleurs ( 1948 ) light profile .
the simulations of davari et al .
( 2016 ) show that this method can lead to biases in measuring the properties of the bulge and disk , depending on the size , @xmath21 , and redshift of the galaxy .
for instance , fixing bulge @xmath4 can overestimate / underestimate the bulge / disk total brightness , and in general , the uncertainties tend to be greater when the bulge @xmath4 is fixed . besides , by fixing the bulge srsic index , one can not tell how the bulge density and shape evolve , and important information is lost .
our study relaxes the restriction on the bulge profile shape , which results in more robust and informative bulge+disk decompositions ( davari et al .
2016 ) .
the cosmic assembly near - infrared deep extragalactic legacy survey ( candels;@xcite ; @xcite ) , provides an unprecedented chance to investidagate the morphological evolution of galaxies .
in fact , one of the original science goals of candels is to trace the bulge and disk growth in rest - frame optical wavelengths at 1 @xmath17 @xmath5 @xmath17 3 ( @xcite ) .
we take advantage of all wide and deep images taken in five well - known , widely separated fields : goods - south and goods - north ( the great observatories origins deep survey ; @xcite ) , uds ( ukidss ultra - deep survey ; @xcite ) , cosmos ( the cosmic evolution survey ; @xcite ; @xcite ) , and egs ( the extended groth strip ; @xcite ) .
these collectively yield a statistically uniform and robust sample that mitigates cosmic variance .
the most massive galaxies in the local universe are almost all quiscent ( @xcite ) , which is not the case at earlier epochs ( @xcite ) .
this means that massive star - forming galaxies have all quenched over time . quantifying the evolution of both quiescent and star - forming galaxies helps trace back the formation of massive ellipticals and understand the bigger picture .
we address three key questions : \1 ) how does the size and the shape of the light distribution ( srsic index ) of star - forming and quiescent massive galaxies evolve ?
\2 ) do high - redshift massive galaxies have a prominent stellar disk ?
if yes , do their relative bulge fraction evolve significantly over the last 10 billion years ?
\3 ) what does the observed evolution of bulges and disks teach us about the history of massive galaxies ?
our findings show that the massive galaxies were compact and indeed more disk - dominated at higher redshifts and became more bulge - dominated over time , converging to the population of massive ellipticals by today .
only major mergers can effectively destroy large - scale disks .
thus , while minor mergers were largely responsible for the significant size increase of high-_z _ galaxies , our results underscore that major mergers also played an important role in the morphological transformation of massive galaxies .
this paper is organized as follows .
section 2 provides details of the sample definition , whcih uses techniques described in section 3 . the morphological analysis is presented in section 4 .
section 5 discusses the implications of our results , and a summary is given in section 6 . throughout this study
we adopt a standard cosmology ( @xmath22 = 71 @xmath23 , @xmath24 = 0.27 , and @xmath25 = 0.73 ) and ab magnitudes .
we utilize candels images and catalogs . besides their high - quality near - ir photometry taken with _
hst_/wfc3 , the observations are complemented with deep _
hst_/acs optical images , mid - ir photometry from _ spitzer _ , and near - uv observations from the ground .
this provides a reliable dataset for the determination of photometric redshifts and stellar masses .
the photometric redshifts are computed by combining 11 independent measurements ( @xcite ) , each using different combinations of photometric redshift code , template spectral energy distributions , and priors .
the median fraction difference between the photometric and spectroscopic redshifts is less than 0.01 , with an rms scatter of @xmath26 ( @xcite ) . as this study is mostly concerned with broad evolutionary trends between @xmath0 @xmath3 2.5 to 0.5 , precise redshifts for individual objects are not essential to our analysis .
the final quoted stellar mass is the median of estimates from 10 different candels teams , who used the same photometry and redshifts estimates but different fitting codes , assumptions , priors , and parameter grid ( @xcite ; @xcite ) . for massive galaxies ,
there is good agreement between candels and 3d - hst ( @xcite ; @xcite ) .
@xcite perform extensive simulations to quantify the different sources of errors and uncertainties , using 10 ten independent methods and mock galaxy catalogs with a range of redshifts , masses , and spectral energy distributions .
they concluded that different methods have comparable scatter of 0.136 dex , with no significant bias .
we employ the candels @xmath27-band images and the accompanied catalogs to analyze the evolution of the rest - frame optical properties of massive galaxies between @xmath28 and 0.5 .
the observed @xmath27 magnitudes of our sample range from 24 - 17 mag ( measured by single srsic fit ) , corresponding to the @xmath29-band rest - frame of 15.5 - 12 mag .
we estimated the rest - frame magnitudes by constructing seds of individual galaxies , shifting them to _ _
z__=0 , and convolving the resulting rest - frame seds with the @xmath29-band response function . the high resolution ( pixel scale = 0.06@xmath30 ) ,
bright limiting magnitude ( 5 @xmath31 @xmath3 27 mag ) , and wide areal coverage of the candels fields , coupled with the availability of physical parameters ( photometric redshift , stellar mass ) for individual galaxies , make this dataset unique and ideal for our photometric analysis .
color - color diagram is used for distinguishing quiscent galaxies from star - forming galaxies .
the quiescent galaxies populate the top left region of the diagram .
[ fig : uvj],width=283 ] we use the rest - frame @xmath32 color - color diagram to separate quiscent galaxies from star - forming galaxies ( see , e.g. , @xcite ; @xcite ; @xcite ; @xcite ) .
we use the selection criteria of @xcite to differentiate between these two types of galaxies ( figure [ fig : uvj ] ) .
quiescent galaxies populate a region defined by @xmath33 where @xmath34 , @xmath29 , and @xmath35 are rest - frame magnitudes , calculated using the eazy photometric redshift code ( @xcite ) and templates from @xcite .
the fraction of star - forming and quiescent galaxies in our sample at different redshifts is shown in figure [ fig : class_fraction ] and table 1 .
it can be seen that at @xmath0 = 2.5 most massive galaxies were star - forming . by @xmath36 ,
the majority of massive galaxies are quenched , in agreement with the findings of @xcite and @xcite .
@xmath0 @xmath17 2.5 .
error bars show our sample proportions standard deviation .
[ fig : class_fraction],width=283 ] we choose massive galaxies based on number density selection rather than a fixed stellar mass limit . for a chosen cumulative number density ,
we rank galaxies according to their stellar mass and chose galaxies of the same rank at different redshifts .
@xcite , using the millennium simulation results ( @xcite ; @xcite ) , show that the former is more reliable for tracing the true evolution of the average stellar mass below _
number density selection , despite its limitations , is more physically motivated ( @xcite ) .
for instance , red nuggets have doubled their stellar masses in the last 10 billion years ( @xcite ) .
in other words , local massive galaxies were less massive in the past and could be left out of a fixed stellar mass selection . to trace the evolution of massive galaxies , it is more sensible to select galaxies at a constant cumulative number density ( @xcite ; @xcite ; @xcite ; @xcite ) .
we use the criteria of patel et al .
( 2013 ) for selecting galaxies at a fixed cumulative number density , @xmath37 .
their figure 2 shows that @xmath38 mpc@xmath39 corresponds to @xmath40 and @xmath41 at @xmath42 and 0.5 , respectively .
local ( @xmath43 ) galaxies with this corresponding stellar mass ( @xmath44 ) are predominantly quiescent ( @xcite ; @xcite ) and have large axial ratios ( @xcite ) , and therefore massive ellipticals .
our sample consists of @xmath2250 massive galaxies , whose properties are summarized in table 1 .
the mass range for each redshift bin takes into account systematic uncertainties in the stellar mass estimate .
) profile of the galaxy , the galfit fit model , and the bulge and disk components ( in cases of bulge+disk decomposition ) .
bottom left panels show the residuals between the model and the observed mean surface brightness .
the error bars are calculated using the rms of the image background and the surface brightness measurement error calculated by ellipse in iraf .
the right panels , from top to bottom , show the observed galaxy , the galfit model , and the residuals , respectively .
the top panel makes it clear that the fitted galaxy has a bulge - line central concentration and spiral arms , and hence a disk ; the model is trying to accommodate both components .
the bottom light profile plots show that the bulge+disk decomposition reproduces the light profile of the galaxy well .
[ fig : spiral_diagnostic],width=536 ] .
this galaxy is basically a single spheroidal , and adding a second component does not improve the model fit significantly .
[ fig : elliptical_diagnostic],width=529 ] we use galfit 3.0 ( @xcite ) as the main modeling tool .
galfit is a powerful , simple - to - use image analysis code to fit the light distribution of galaxies and other objects . of the several available options
, we mainly use the srsic ( 1968 ) function to fit the surface brightness distribution : @xmath45\right \ } } , \label{eq : sersic}\ ] ] where @xmath46 stands for the surface brightness , @xmath47 is the half - light ( effective ) radius , @xmath4 is the srsic index that governs the shape of the profile , and @xmath48 is a variable that depends on @xmath4 ( @xcite ) . the srsic function is a generalization of the special cases of an exponential profile ( _ n _ = 1 ) used to model disks ( @xcite ) and the _ @xmath49 _ law ( _ n _ = 4 ) traditionally used to model elliptical galaxies and bulges ( @xcite ) .
modern studies recognize that ellipticals and bulges have a more varied range of @xmath4 ( e.g. , caon et al .
1993 ; andredakis & sanders 1994 ; @xcite ; @xcite ) .
several inputs are needed to perform fit : a psf model , a `` sigma '' image , and ( sometimes ) a bad pixel mask . for each galaxy ,
we use the @xmath27-band candels hybrid psf corresponding to its field ( e.g. , uds , cosmos , etc . ;
hybrid psfs are built by combining a stacked empirical stellar psf and a synthetic tinytim ( @xcite ) psf .
candels weight maps are used as the input sigma images . as the field around each galaxy
is chosen to be more than 1015 times larger than the size of the galaxy , there are usually several other objects in the field that need to be masked . as in davari
( 2014 ) , we use sextractor ( bertin & arnouts 1996 ) to identify bright field objects and create a bad pixel mask that covers twice the area detected by sextractor . for any given galaxy in the sample ,
our primary goal is to ascertain whether its light distribution , apart from a central bulge , shows evidence for an additional disk component , and if so , to determine its relative light fraction .
we model each galaxy twice , first with a single - component srsic fit , and then with a two - component fit consisting of a bulge and a disk .
the bulge is assigned a srsic function with @xmath4 allowed to vary , and the disk fixed to an exponential .
we then carefully examine the residuals to determine the merits of the two models .
depending on the complexity of the galfit model , the initial parameters ( guesses ) can have a large effect on the fit .
for single - component fits , unless the initial guesses are very far off the actual values , the initial parameters do not have a major effect .
regardless , we use one - dimensional light profiles obtained by iraf / ellipse ( @xcite ) to obtain reliable initial guesses .
we construct a curve - of - growth of the light distribution to estimate the effective radius , total luminosity , axial ratio , and position angle ( for more details , see davari et al .
appropriate initial guesses become much more important for the bulge+disk decompositions . for this type of modeling
, we again use the one - dimensional light profile to obtain initial inputs . assuming that the disk component follows an approximately exponential profile , we look for the part of the profile that traces a straight line in logarithmic space .
depending on the @xmath6 , this region is located between 2 and 5 @xmath47 , where the effect of the bulge is minimal .
a straight line fitted to that section of the light distribution ( in logarithmic space ) provides an estimate for the disk scale length and central surface brightness . the total brightness obtained from a single srsic fit
is used to find the total luminosity of the bulge component , and therefore @xmath6 .
each galaxy is fit numerous times with different initial guesses for bulge @xmath47 and @xmath4 , in addition to all the estimated parameters . while fitting a single srsic function might require only a few iterations
, bulge+disk decompositions can require several fits with different initial guesses .
the diagnostic plots ( explained below ) are imperative for evaluating the goodness of a fit .
lastly , the fitted sky component is left as a free parameter , and its initial values are set to zero .
the simulations performed by @xcite and davari et al .
( 2016 ) show that once the field size of the image is more than 10 times larger than the galaxy , galfit can measure the sky reliably .
figures [ fig : spiral_diagnostic ] and [ fig : elliptical_diagnostic ] give examples of the diagnostic plots used to examine the goodness of a fit and whether or not a second component is needed .
the top left panels show the mean one - dimensional surface brightness ( @xmath50 ) profile of the galaxy , the final galfit model , and the bulge and , if necessary , the disk components .
the bottom left panels show the residuals ( model @xmath51 galaxy ) .
the error bars are calculated using the rms of the image background and the galaxy flux measurement error ( output from ellipse ) .
the right panels , from top to bottom , illustrate the two - dimensional image of the galaxy , model , and residuals .
the top right panel of figure [ fig : spiral_diagnostic ] clearly reveals that the galaxy , in addition to a bright central concentration , has a disk and spiral arms .
the one - dimensional profile in the left panel confirms that the galaxy contains complex structure .
the single - component model is trying to capture the most of the combination of two components , but the fit is clearly inadequate .
the bulge+disk decomposition , by contrast , reproduces well the @xmath50 profile on the bottom left panel . and
not surprisingly , this galaxy is extremely disk - dominated : the best fit yields @xmath6 = 0.05 . at the other extreme
, figure [ fig : elliptical_diagnostic ] showcases a galaxy that is basically a just single big bulge ; adding a second component does not improve the fit significantly .
the high @xmath6 of this galaxy ( 0.92 ) validates this hypothesis . ) , srsic index ( @xmath4 ) , and ellipticity ( @xmath52 ) .
red , blue , and black filled boxes show the median in each redshift bin for quiescent galaxies , star - forming galaxies , and both types combined .
the gray filled diamond shows the median value for a sample from the second data release of gama ( @xcite ; @xcite ) with mass range corresponding to our number density selection criteria .
their morphological parameters are derived from single srsic fits , consistent with our method .
the error bars correspond to the interquartile range of different measurements .
[ fig : singlesersic],width=340 ] ) , bulge magnitude ( @xmath53 ) , and disk magnitude ( @xmath54 ) .
the reported magnitudes are rest - frame magnitude in v - band .
red , blue , and black filled boxes show the median in each redshift bin for quiescent galaxies , star - forming galaxies , and both types combined .
the error bars correspond to the interquartile range of different measurements .
[ fig : bt],width=340 ] ) , bulge srsic index ( @xmath55 ) , and bulge ellipticity ( @xmath56 ) .
red , blue , and black filled boxes show the median in each redshift bin for quiescent galaxies , star - forming galaxies , and both types combined .
the error bars correspond to the interquartile range of different measurements .
[ fig : bulge],width=340 ] ) and disk ellipticity ( @xmath57 ) .
red , blue , and black filled boxes show the median in each redshift bin for quiescent galaxies , star - forming galaxies , and both types combined .
the error bars correspond to the interquartile range of different measurements.[fig : disk],width=340 ] davari et al .
( 2014 , 2016 ) demonstrate that large srsic indices ( @xmath58 ) derived from single - component fits can lead to significant biases , which can be remedied by fixing @xmath4 to 6 , after testing the fit for different initial guesses .
this study follows the same general rule , except that the diagnostic plots are given more weight .
for example , if fixing @xmath4 to 4 or 8 gives a better fit and cleaner residuals than fixing @xmath4 to 6 , then those values are used .
another common symptom of unreliable fits is when the effective radius of the bulge drops below 0.5 pixel . for
many of these cases , changing the initial guesses of @xmath47 and @xmath4 leads to more realistic solutions .
but if the problem persists , we resort to fixing the bulge @xmath47 ( or sometimes @xmath4 or both ) to different initial values , and we rely on visual inspection of the residuals to judge the merit of each model . for about 25% of the galaxies at _ z _ @xmath1 1.5 , the bulge @xmath47 ( mainly to @xmath47=1 ) and/or bulge srsic index ( mainly to @xmath4=1 ) are fixed . in short , the goodness of single and
two component fits are mainly determined by visual inspection of the residual images ( galaxy - model ) , along with the derived sersic profile parameters .
the derived sersic parameters of each component have to be reasonable for a fit ( e.g. , @xmath59/@xmath60 @xmath17 1 , ellipticities @xmath17 0.8 , sizes @xmath1 0.5 pixel , and etc . ) to be considered reliable .
the objective of this study is not fitting two components only when there is an improvement over the single sersic fit residual .
for example , figure 4 shows an example of a bulge - dominated galaxy , where adding a second component does not improve the residual significantly .
however , the derived two component fit parameters not only confirm that this galaxy is bulge dominated ( high @xmath6 ) , but also provides additional information ( e.g. , @xmath59 , @xmath55 , @xmath61 , and etc . ) out of 248 , only 1 and 7 galaxies could not be fit reliably with a single srsic and two components , respectively , and are omitted from the following analysis . in most of these cases ,
the image contains multiple regions with nonuniform and anomalous background values .
single srsic fitting is probably the most widely adopted method in the literature for morphological studies .
this method provides a rather straightforward way for evaluating some key morphological properties of galaxies , namely size ( usually parameterized as the effective radius ; @xmath47 ) , srsic index ( @xmath4 ) , and ellipticity ( @xmath52 ) .
for instance , if a randomly distributed galaxy population has a significant disk component , we expect a wide distribution of @xmath4 and @xmath52 . the simulations of davari et al .
( 2014 , 2015 ) show that single - component fits of massive galaxies at the redshift range of current interest ( @xmath62 ) can be measured with little to no systematic uncertainty .
figure [ fig : singlesersic ] summarizes the results of single - component fits of our sample , highlighting the redshift evolution of @xmath47 , @xmath4 , and @xmath52 , separately for quiescent and star - forming galaxies . for reference , we overplot the median value for a sample drawn from the second data release of gama ( @xcite ; @xcite ) ; the sample mass range corresponds to our number density selection ( i.e. @xmath63 ) at @xmath64 .
the morphological parameters of the gama survey are derived by single srsic fitting , consistent with our method .
the error bars correspond to the interquartile range of different measurements .
we perform two statistical tests to quantify the significance of the observed evolution of different properties : two - sample kolmogorov - smirnov ( ks ) test and one - way analysis of variance ( anova ) .
the ks test is used to determine whether the star - forming and quiescent samples are drawn from the same parent population . as a non - parametric test , it has the advantage of making no assumption about the distribution of data .
the results of ks test are summarized by the @xmath65-value and the @xmath66-value .
the @xmath65-value shows the maximum difference between the empirical cumulative distribution functions of the two samples , while the @xmath66-value indicates the significance of the difference between two samples .
small @xmath66-values ( @xmath67 ) reject the null hypothesis that the two samples are drawn from the same distribution .
anova tests whether there are any statistically significant differences between the means of sample quantities in our redshift bins .
the @xmath68-value ( i.e. @xmath68 statistics ) quantifies the variance between groups compared to the variance within groups : ( variance of sample means at different redshift bins)/(variance of the whole sample ) .
high @xmath68-values ( i.e. , small @xmath66-values ) reject the null - hypothesis that the mean values are not statistically different at different redshift bins . in this study ,
high @xmath68-values indicate there is evolution over the observed redshift range .
the results of the ks and anova test for the single - component fits are listed in table 2 .
massive galaxies have experienced significant size evolution ( top panel of figure [ fig : singlesersic ] ; table 2 ) , with the size increase being more prominent for quiescent galaxies ( table 3 ) .
quiescent galaxies have increased their sizes by a factor of 3 down to @xmath69 , and more than a factor of 5 by @xmath43 . by contrast
, star - forming galaxies have undergone more modest size growth , by a factor of @xmath70 down to @xmath69 . however , the absolute amount of size increase between @xmath0 = 2.5 and @xmath0 = 0.5 for both star - forming and quiescent galaxies is comparable , about 2 kpc . the slope of the size - mass relation is consistent with the value found in @xcite and @xcite . on average ,
the size interquartile ranges are smaller for quiescent galaxies , which indicates a greater size diversity among star - forming galaxies . in other words ,
the sizes of quiescent galaxies are more homogeneous .
furthermore , the star - forming galaxies at each redshift are larger than their quiescent counterparts , in agreement with previous similar studies ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
star - forming and quiescent galaxies have statistically different size distributions ( table 2 ) .
the global srsic indices of both star - forming and quiescent galaxies in our highest redshift bin cluster around @xmath71 ( middle panel of figure [ fig : singlesersic ] ) , an intermediate value consistent with a composite bulge+disk system . but over time , the two galaxy types diverge ( table 2 ) .
while star - forming galaxies maintain an almost constant @xmath4 , the srsic indices of quiescent galaxies increases significantly and systematically toward lower redshifts , eventually converging to resemble those of local elliptical galaxies ( @xmath72 ) at the lowest redshift bin .
these trends are broadly consistent with the results of @xcite , @xcite , and @xcite
. the trends with regards to ellipticity are less definitive .
if massive galaxies initially host a sizable disk , we expect the eventual disappearance of that component to produce a notable reduction in the typical ellipticity of the population . in practice
, however , the presence of a sizable bulge concentration severely dilutes the expected ellipticity signature of any disk component . indeed , neither the anova nor the ks test indicates any statistically significant redshift evolution of @xmath52 .
( table 2 ) .
however , considering only the quiescent galaxies at @xmath5 @xmath1 1.5 and @xmath5 @xmath17 1.5 , there are two suggestive indicators of ellipticity evolutions .
first , f - test of equality of variances gives a p - value of 0.08 which is a bordering signature of greater range ( variance ) of ellipticities at @xmath5 @xmath1 1.5 .
second , comparing ellipticities of the quiescent galaxies at @xmath5 @xmath1 1.5 and @xmath5 @xmath17 1.5 shows a tentative drop in the overall ellipticity : the two sample one - sided t - test gives a p - value of 0.05 .
these signatures are suggestive and not conclusive .
the following section presents a more detailed analysis which provides an independent gauge for the presence of a disk among massive quiescent galaxies at higher redshifts . to summarize the results and implications of the single - component
fits : the global light distribution of the massive galaxy population evolves significantly from @xmath42 to @xmath69 .
apart from the well - known increase in size , the population as a whole , and in particular the quiescent systems , exhibits systematical evolution toward larger srsic indices and lower ellipticities at lower redshifts , converging to typical values of local ellipticals .
these trends support the thesis that the progenitors of present - day ellipticals were born with a sizable large - scale disk , which over time has been transformed . while single srsic fitting provides a reliable first - order estimate of morphological properties , decomposing a galaxy into its bulge and disk components can reveal a new set of valuable galaxy evolution indicators .
the simulations of davari et al .
( 2016 ) show that gross photometric properties , in particular @xmath6 , of bulge+disk systems usually can be measured accurately , up to @xmath73 , without imposing any constraints on the profile shape of the bulge . however , due to the inherent limitations of resolution , even with _ hst _ , detailed properties of the bulges ( e.g. , @xmath47 or @xmath4 ) can be measured reliable only for galaxies with @xmath6 @xmath74 0.2 .
the disk component , by contrast , can be measured with little difficulty .
figure [ fig : bt ] depicts the overall variation of @xmath6 with redshift for our sample . at higher redshifts ,
quiescent galaxies have intermediate values of @xmath6 ( @xmath20.4 ) , but over time they become more and more bulge - dominated . at the lowest redshift bin , @xmath75 , very close to the median value of local massive elliptical galaxies .
although star - forming galaxies , too , become more bulge - dominated with time , their @xmath6 at all redshift bins are lower than that of their quiescent counterparts ; the two classes have statistically different distributions in @xmath6 ( table 4 ) .
@xcite report that massive galaxies at @xmath76 are mostly disk - dominated and by @xmath77 have increased their @xmath6 to intermediate values , with very few elliptical - like galaxies down to @xmath78 .
they show that disk - dominated galaxies have higher star formation rates , which translates into star - forming galaxies having a lower @xmath6 .
similarly , @xcite also find that massive galaxies increase in @xmath6 between @xmath79 and @xmath80 .
although @xmath27-band images of galaxies at different redshifts capture the flux in different rest - frame bands ( i.e. , approximately @xmath29 to @xmath81 band ) , multi - wavelength studies of nearby galaxies find that @xmath6 does not strongly depend on observed rest - frame wavelength , at least within the standard optical bands ( e.g. , @xcite ; @xcite ) .
the observed variation of @xmath6 between different bands is less than @xmath82 .
the shallow color gradients of quiescent galaxies ( e.g. , wirth 1981 ) further minimizes the impact of rest - frame wavelength on @xmath6 .
while the bulges of both types of galaxies become more luminous over time , the disks component behaves markedly differently : it becomes sub - dominant in quiescent galaxies but brightens for the star - forming group ( middle and bottom panels of figure [ fig : bt ] ; table 4 ) .
meanwhile , the star - forming galaxies disks are becoming brighter at lower redshift bins .
the bulges of quiescent galaxies attain higher luminosities than in star - forming galaxies at all redshifts , and in the lowest redshift bin the luminosities of quiescent galaxies have significantly smaller scatter than in star - forming galaxies .
our bulge+disk decomposition ( figure [ fig : bulge ] and table 3 ) reinforces the size evolution observed in the single - component fits ( figure [ fig : singlesersic ] ) .
the effective radii of the bulges of both classes have grown , and , once more , the evolution is steeper for quiescent galaxies .
the disk scale lengths of both star - forming and quiescent galaxies have increased ( figure [ fig : disk ] ) as well , but their distributions are not distinguishable ( table 4 )
. table 3 indicates that the disk size increase is less significant compared to the bulge component .
as shown in figure [ fig : bulge ] , the srsic indices of quiescent galaxy bulges have increased considerably but have stayed almost the same for the star - forming population . this is similar to the results of single - component analysis ( figure [ fig : singlesersic ] ) . by redshift 0.5 ,
bulges of quiescent galaxies have srsic indices similar to that of typical local ellipticals and classical bulges ( @xcite ) .
the disk ellipticities of both classes have similar distribution and have not changed between _ z _ = 2.5 and 0.5 . on the other hand , the bulges of massive galaxies have become rounder over this period .
quiescent galaxies have lower bulge ellipticities , and by _ z _ = 0.5 , their distribution is similar to that of local massive ellipticals and classical bulges ( @xcite ) . in related studies ,
bruce et al . ( 2014a , 2014b ) analyze the rest - frame optical morphologies of a mass - selected sample of massive ( @xmath83 ) galaxies at @xmath84 in the candels uds and cosmos fields .
similar to our work , they decomposed @xmath85-band images of massive galaxies into their bulge and disk components . in general ,
our results are in agreement with those of these authors .
bruce et al . ( 2014a ) find that from @xmath5 = 3 to @xmath5 = 1 the galaxies transition from disk - dominated to more bulge - dominated ( their figure 6 ) , in accordance with our findings ( our figure 6 ) .
the results of bruce et al .
( 2014b ) show that bulges exhibit a stronger size evolution than disks ( their table 3 ) , with star - forming galaxies having relatively larger disk sizes compared to passive systems ( their figure 8) , in qualitative agreement with our results ( compare with our table 3 and figure 8) . however , with regards to the the bulge components , they show that star - forming galaxies have larger bulges than quiescent galaxies , contrary to the results from our study
. this may be due to the fact that the bulge - disk decomposition of bruce et al .
( 2014a , 2014b ) was done by fixing the srsic index of the bulge and disk components to 4 and 1 , respectively .
as demonstrated in davari et al .
( 2016 ) , fixing the srsic indices can lead to biases and larger uncertainties in measuring bulge and disk properties , depending on their size , redshift , and @xmath21 .
furthermore , assuming a fixed profile for disks and bulges at all redshifts precludes any investigation of the evolution of these parameters with look - back time .
despite the prevalence of spiral structures and bars in the local universe ( e.g. , @xcite ; @xcite ) , these features are not believed to be common among star - forming galaxies at higher redshifts ( _ z _ @xmath1 1.5 ) ( e.g. , @xcite ; @xcite ; @xcite ) , where the disks may be too dynamically hot ( @xcite ; @xcite ; @xcite ) .
our model - subtracted residual images yield an unexpected surprise : a sizable fraction of the sources exhibit spiral structure ( figure [ fig : spiral_images ] ) .
all the cases are star - forming galaxies ; there are no quiescent galaxies with securely detectable spiral structure ( table 5 ) . the case
with the highest redshift is at @xmath86 .
the fraction of star - forming galaxies with spiral structure is @xmath87% at @xmath88 , and by @xmath89 , the spiral fraction reaches nearly 70% .
the fraction of star - forming galaxies with spiral structures drops ( to 30% ) at the lowest redshift bin .
this decline may not be reliable for two reasons .
first , the star - forming sample size at @xmath90 is very small ( only 10 objects ) , and therefore the sample proportions are not statistically significant .
second , the @xmath27-band images for the low-@xmath5 objects are missing the bluer parts of the galaxy flux ( see also @xcite ) .
the conditions necessary for the formation of spiral arms are complex ( see @xcite for a review ) , but one requirement is clear the existence of a disk .
thus , from the point of view of one of the main themes of this paper , the clear detection of spiral features in high-@xmath5 massive galaxies constitutes arguably the strongest , most model - independent evidence for the presence of a substantial disk component in these systems .
we see the spiral features only in the star - forming galaxies and not in quiescent systems , but , by analogy with local s0 and spiral galaxies , this is not surprising .
an edge - on view of a galaxy can reveal another indisputable signature of a prominent stellar disk .
figure [ fig : thindisk ] gives several examples of highly flattened ( @xmath91 ) galaxies in our sample that are consistent with disk structures seen edge - on .
interestingly , most of them are relatively thick
. the fraction of galaxies with a stellar disk can be infered from the frequency of detected edge - on galaxies . for this estimation , we assume a uniform distribution of ellipticity with @xmath92 for a population of bulge+disk systems .
the inferred fraction of galaxies with a disk hovers around 4050% at @xmath93 , both among star - forming and quiescent galaxies , but below @xmath78 , the incidence of edge - on quiescent galaxies drops to zero ( table 5 ) . highly flattened systems ( especially at high-_z _ ) can be hallmarks of merger .
however , considering the fact that the majority of these galaxies are compact , the chance of this degeneracy is low .
the discovery of red nuggets has captured much attention in recent years as it requires a new paradigm for the formation and evolution of massive elliptical galaxies .
the observed compactness of red nuggets at @xmath94 initially raised the question of whether the massive red galaxies have indeed increased their sizes by a factor of roughly @xmath95 while maintaining their passive state , or whether the size measurement might in some way be flawed .
extensive recent simulations ( @xcite ; davari et al .
2014 , 2016 ) , coupled with consistent results from multiple independent studies ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , have minimized skepticism on the fidelity of the size measurements .
this work confirms that massive galaxies at @xmath96 were indeed compact , and over the next 10 billion years their sizes have increased significantly ( figure [ fig : singlesersic ] ) .
the growth occurred inside - out ( @xcite ; @xcite ) .
figure [ fig : inside_out ] shows the median light distribution of massive quiescent and star - forming galaxies in four redshift bins . while the inner few kpc of these galaxies have been in place since @xmath28 , over time more and more material was added to their outskirts .
accretion onto quiescent galaxies continued at least down to @xmath97 , whereas their star - forming counterparts seem to have stopped growing by @xmath98 .
the compactness of the red massive galaxies at higher redshifts and their similarities to smgs implicate the importance of strong gas dissipation during their early formation epochs , which in turn led to the starburst activity and accompanying disk formation ( e.g. , @xcite ; @xcite ) .
this raises some important questions : do red nuggets at @xmath96 have a sizable disk component , and if so , how prevalent was it ? and since red nuggets are widely believed to evolve into present - day elllipticals indeed , our number density selection was specifically chosen to ensure that they do can we trace the redshift evolution of the morphological transformation that must take place ?
table 5 shows the estimated fraction of galaxies with a prominent stellar disk , using three different diagnostics : galaxies with @xmath6 @xmath17 0.5 , visually detectable spiral features ( section 4.3 ) , and inferences from the frequency of detected edge - on ( @xmath91 ; section 4.3 ) systems .
our results suggest that disks may be common among high-_z _ massive galaxies , although it is difficult to obtain a conclusive estimate of their frequency .
they range from an absolute minimum of @xmath99% , as deduced from the incidence of spiral arms among star - forming systems , to as high as @xmath100% for all massive galaxies regardless of star formation activity , according to bulge - disk decomposition .
the three disk diagnostics are not equally reliable and informative . while , visually detectable spiral structures are the most reliable indicator of a disk , they provide the lower limit , as the viewing angle and surface brightness dimming can prevent the detection of these structures .
furthermore , at higher redshifts , disks of massive galaxies may not be favorable to long lived spiral structures .
on the other hand , low @xmath6 at higher redshifts does not necessarily mean a disk resides in a galaxy and provides an upper limit .
the fitted exponential component is not necessarily an indicator of a disk .
lastly , the inferred fraction of disks using the fraction of edge - on disks is probably the best proxy for the detection of disks .
the prevalence of large - scale disks at @xmath101 is further reinforced by the moderate srsic indices and broad distribution of ellipticities derived from the global light distribution , a trend already echoed in other recent investigations . by tracking the population from @xmath28 to 0.5 and performing a consistent analysis of the whole sample
, we witness the gradual transition of the large - scale morphology . by @xmath97
, the red massive galaxies attain a high bulge fraction of @xmath75 , signifying the near disappearance of a dominant disk ( figure [ fig : bt ] ) ; their global ( figure [ fig : singlesersic ] and bulge ( figure [ fig : bulge ] ) srsic indices converge to @xmath102 , values close to that of de vaucouleurs profile ; and their global axial ratios drop to values closely resembling those of local massive galaxies ( e.g. , as measured in the gama survey ) .
all of these indicators strongly support the thesis that high-@xmath5 red nuggets are , in fact , the direct ancestors of today s massive ellipticals .
much attention has been focused recently on the pivotal role that minor , dry mergers play in the evolution of red nuggets into present - day elliptical galaxies .
minor mergers are considered the most plausible mechanism for explaining the dramatic size growth of massive quiescent galaxies ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , their mass increase after being quenched ( @xcite ) , the multiple - component structure of local ellipticals ( huang et al .
2013a , 2013b ) , and the prevalence of tidal features seen in deep imaging of nearby massive galaxies ( @xcite ; @xcite ; @xcite ) .
the simulations of @xcite stress the effectiveness of dry mergers in increasing the sizes of massive compact galaxies .
consistent with @xcite , they find that dry mergers lead to a size - mass relation of the form @xmath47
@xmath103 @xmath104 , with @xmath105 @xmath3 2 .
this is close to the size evolution we measure , @xmath106 , for our quiescent massive galaxies between redshift 2.5 and 0.5 .
interestingly , star - forming galaxies have a considerably smaller value of @xmath107 .
by @xmath108 , more than 80% of the simulated massive galaxies ( @xmath83 ) from @xcite have experienced minor mergers .
the merger rate is expected , on average , to increase monotonically with stellar mass ( @xcite ) , and therefore should be even higher for our sample .
figure [ fig : merger_images ] shows a number of galaxies from our sample with disturbed morphologies and small - scale structure that may be indicative of merging activity .
we estimate , from visual inspection , that at @xmath109 more than 60% of quiescent galaxies have small nearby objects or show merger signatures ( e.g. , distortions , tidal tails , and shells ) ; about 40% of star - forming galaxies show similar features .
by @xmath89 , the fraction of merger candidates , for both classes combined , drops to @xmath110% , presumably because it becomes increasingly difficult to resolve small - scale structure for distant galaxies .
while these morphological indicators are by no means secure estimators of the merger fraction , they at least give the qualitative impression that mergers especially minor mergers
play a part in the morphological transformation of massive galaxies .
the discovery of a significant disk component in massive galaxies at @xmath96 and their eventual disappearance toward lower redshifts brings an important new element to the story .
how were the disks destroyed ? can this be accomplished by minor mergers alone ?
most likely not . breaking a big disk requires hitting it with something hefty , which can only be accomplished with a major merger .
the simulations of @xcite show that major mergers are needed for forming galaxies with high @xmath6 .
another key player in the evolutionary scenario of massive galaxies is , one that has captured less attention , are the compact blue galaxies although star - forming galaxies are larger than quiescent galaxies at each redshift bin , at high redshifts these galaxies are compact , as well ( figure [ fig : singlesersic ] ) .
as local star - forming massive galaxies are rare ( e.g. , @xcite ) , most of high - redshift compact blue galaxies must have also evolved into present - day massive ellipticals .
figure [ fig : class_fraction ] illustrates how over time the star - forming massive galaxies turn into quiescent massive galaxies . at the same time that the blue population quenches star formation , significant morphological transformation must also occur to elevate the relative bulge fraction ( figure [ fig : bt ] ) and increase the srsic index ( both globally and for the bulge alone ; figure [ fig : singlesersic ] and [ fig : bulge ] ) .
the prevalence of prominent stellar disks at higher redshifts raises the possibility that some of these bulge+disk massive galaxies may have survived to the present . where are they ?
the `` superluminous '' spiral galaxies discussed by ogle et al .
( 2016 ) seem to fit the description .
ogle et al .
quote an average number density of 32 gpc@xmath39 at @xmath111 .
interestingly , the fraction of star - forming massive galaxies with spiral arms is 30% at @xmath97 ( table 5 ) . as our overall sample was chosen to satisfy @xmath38 mpc@xmath39 , the observed number density of massive spirals in our lowest redshift bin is @xmath112 , or 42 gpc@xmath39 , very close to the average number density given by ogle et al .
( we note , however , that the sample size of star - forming massive galaxies at the lowest redshift bin is not statistically significant , as discussed in section 4.3 . )
@xcite , using the illustris ( @xcite ; @xcite ) cosmological hydrodynamical simulations , trace the evolution of 35 massive compact galaxies from _ z _ = 2 .
they find that @xmath230% of their galaxies survive undisturbed , while the rest have either experienced inside - out growth or have been destroyed via major mergers .
the discovery of massive compact galaxies at high redshift , specially red nuggets , has offered new insights into galaxy formation and evolution .
these massive galaxies have major differences with their local counterparts , the massive ellipticals .
they are not only compact but , as demonstrated in this study , they also possess a stellar disk . to match the population of present - day ellipticals
, red nuggets must increase significantly in size _ and _ destroy their disks . using a homogeneous and unbiased sample of @xmath2 250 massive galaxies in the candels fields , spanning the redshift range 0.5 @xmath17 @xmath0 @xmath17 2.5 , and selected through the fixed number density technique
, we studied the evolution of morphological parameters as a function of redshift .
further , we classified galaxies into quiescent and star - forming systems using the uvj color - color diagram in order to trace separately their evolutionary histories . * the fraction of quiescent massive galaxies is higher at lower redshifts . * both star - forming and quiescent galaxies have increased their sizes significantly from @xmath0 @xmath3 2.5 to the present time , and the growth has occurred inside - out . *
the global srsic index of quiescent galaxies has increased over time ( from @xmath4 @xmath3 2.5 to @xmath4 @xmath1 4 ) , while that of star - forming galaxies has remained roughly constant ( @xmath4 @xmath3 2.5 ) . *
the distribution of global ellipticities has changed mildly with time , becoming rounder toward lower redshifts . *
the typical value of @xmath6 has increased with decreasing redshift , both for the quiescent and star - forming subsamples . by @xmath5
0.5 , massive quiescent galaxies ( with @xmath6 0.8 ) begin to resemble the local elliptical galaxies .
star - forming galaxies have a lower median @xmath6 at each redshift bin . *
the evolution of sersic index , ellipticity , and @xmath6 suggests that both star - forming and quiescent galaxies have a significant stellar disk at early times , which systematically became less prominent toward lower redshifts . * a considerable fraction of our sample
have visually detectable spiral structures or thin disks observed nearly edge - on , which further confirms that high-_z _ massive galaxies have prominent stellar disks . * while minor dry mergers can explain the inside - out growth of massive galaxies , major mergers are needed to destroy their stellar disks between redshift 2.5 and the present time . *
while the disks of star - forming and quiescent galaxies evolve similarly , their bulges follow different evolutionary trajectories .
the size increase of the bulges of quiescent galaxies is more significant and their srsic indices and axial ratios are , on average , higher than their star - forming counterparts .
rd has been funded by a graduate student fellowship awarded by carnegie observatories .
lch acknowledges support by the chinese academy of science through grant no .
xdb09030102 ( emergence of cosmological structures ) from the strategic priority research program and by the national natural science foundation of china through grant no .
rd thanks heather worthington - davari for providing long - term support .
|
massive galaxies at higher redshifts ( @xmath0 @xmath1 2 ) show different characteristics from their local counterparts : they are compact and most likely have a disk . in this study , we trace the evolution of local massive galaxies by performing a detailed morphological analysis , namely , fitting single srsic profiles and performing bulge+disk decompositions .
we analyze @xmath2 250 massive galaxies selected from all candels fields ( cosmos , uds , egs , goods - south and goods - north ) .
we confirm that both star - forming and quiescent galaxies increase their sizes significantly from @xmath0 @xmath3 2.5 to the present day .
the global srsic index of quiescent galaxies increases over time ( from @xmath4 @xmath3 2.5 to @xmath4 @xmath1 4 ) , while that of star - forming galaxies remains roughly constant ( @xmath4 @xmath3 2.5 ) . by decomposing galaxy profiles into bulge+disk components
, we find that massive galaxies at high redshift have prominent stellar disks , which are also evident from visual inspection of the images .
by @xmath5 @xmath3 0.5 , the majority of the disks disappear and massive quiescent galaxies begin to resemble the local elliptical galaxies .
star - forming galaxies have lower bulge - to - total ratios ( @xmath6 ) than their quiescent counterparts at each redshift bin .
the bulges of star - forming and quiescent galaxies follow different evolutionary histories , while their disks evolve similarly .
based on our morphological analysis and previous cosmological simulations , we argue that major mergers , along with minor mergers , have played a crucial role in the significant size increase of high-_z _ galaxies and the destruction of their massive and large - scale disks .
|
given a bounded domain @xmath1 in @xmath2 , @xmath3 we study the asymptotic behavior as @xmath4 of the spectrum of the following asymmetric elliptic problem @xmath5 either with homogeneous dirichlet or neumann boundary conditions . here , @xmath6 is the @xmath7laplacian with @xmath8 and @xmath9 . the parameters @xmath10 and @xmath11 are reals and depending on @xmath12 .
we assume that the family of weight functions @xmath13 and @xmath14 are positive and uniformly bounded away from zero . for a moment
let us focus problem for fixed @xmath12 with positive weights @xmath15 : @xmath16 with dirichlet or neumann boundary conditions .
consider the fucik spectrum defined as the set @xmath17 let us observe that when @xmath18 and @xmath19 , equation becomes @xmath20 with dirichlet or neumann boundary conditions , which is the eigenvalue problem for the @xmath7laplacian . these has been widely studied .
see for instance @xcite for more information .
it follows immediately that @xmath21 contain the lines @xmath22 and @xmath23 . for this reason ,
we denote by @xmath24 the set @xmath21 without these trivial lines .
observe that if @xmath25 with @xmath26 and @xmath27 then @xmath28 and @xmath29 .
the study of problem with dirichlet boundary conditions have a long history that we briefly describe below .
the one - dimensional case with positive constant coefficients ( i.e. , @xmath30 and @xmath31 ) was studied in the 1970s by fucik @xcite and dancer @xcite in connection with jumping nonlinearities .
properties and descriptions of the first non - trivial curve on the spectrum of on @xmath2 for the general case ( @xmath32 ) without weights can be found in cuesta , de figueiredo and gossez @xcite , dancer and perera @xcite , drbek and robinson @xcite , perera @xcite .
the case with positive weights @xmath33 and @xmath34 was recently studied , see for instance rynne and walter @xcite , arias and campos @xcite , drabek @xcite , reichel and walter @xcite . for indefinite weights @xmath33 and @xmath34 see alif and gossez@xcite , leadi and marcos @xcite .
the main problem one address is to obtain a description as accurate as possible of the set @xmath35 . in the one - dimensional case , @xmath31 , without weights
this description is obtained in a precise manner : the spectrum is made of a sequence of hyperbolic like curves in @xmath36 , see for instance @xcite .
when @xmath33 and @xmath34 are non - constants weights , in @xcite it is proved a characterization of the spectrum in terms of the so - called zeroes - functions . in @xmath2 with @xmath37 and
dirichlet boundary conditions , only a full description of the first nontrivial curve of @xmath21 is known , which we will denote by @xmath38 .
assuming that the weight functions @xmath39 are positive and uniformly bounded , in @xcite ( see theorem 33 ) is proved that @xmath40 can be characterized by @xmath41 where @xmath42 and @xmath43 are continuous functions defined by @xmath44 and @xmath45 is given by @xmath46 where @xmath47 $ ] . here , the functionals @xmath48 and @xmath49 are given by @xmath50 with @xmath51,w^{1,p}_0(\omega)):\gamma(-1)\geq 0 \textrm { and } \gamma(1 ) \leq 0\}.\ ] ] in @xcite ( see proposition 34 ) some important properties of the functions @xmath42 and @xmath43 are proved .
namely , both @xmath42 and @xmath43 are continuous , @xmath42 is strictly decreasing and @xmath43 is strictly increasing .
one also has that @xmath52 if @xmath53 and @xmath54 is @xmath55 .
having defined these previous concepts and definitions , let us back to problem .
homogenization of the spectrum of elliptic operators was extensively studied in the last years .
the case of the eigenvalues of the weighted @xmath7laplacian operator in periodic settings , i.e. , @xmath56 with dirichlet boundary conditions and @xmath57 , @xmath58 being a @xmath59-periodic function with @xmath59 the unit cube in @xmath2 , together with a family of more general problems it was studied for instance by @xcite,@xcite,@xcite,@xcite,@xcite in the linear case ( @xmath31 ) and by @xcite,@xcite,@xcite in the non - linear case ( @xmath32 ) .
up to our knowledge , no investigation was made in the homogenization and rates of convergence of the fucik spectrum .
we are interested in studying the behavior as @xmath4 of problem when @xmath60 and @xmath61 are general functions depending on @xmath0 , and in the special case of rapidly oscillating periodic functions , i.e. , @xmath62 and @xmath63 for two @xmath64periodic functions @xmath39 uniformly bounded away from zero ( see assumptions ) , @xmath59 being the unit cube of @xmath2 .
our main aim is to study the limit as @xmath4 of the first nontrivial curve in the spectrum @xmath65 , say @xmath66 . we wonder :
there exists a limit curve @xmath67 such that @xmath68 can this limit curve be characterized like a curve of a limit problem ?
we will see that the answer is positive .
therefore , a natural question arises : can the rate of convergence of @xmath69 be estimated ?
i.e. , can we give an estimate of the remainders @xmath70 we give positive answers to these questions in the periodic setting .
in theorem [ teo_2 ] we obtain the bounds @xmath71 where @xmath72 is a constant fully determined which is independent of @xmath73 and @xmath0 , and @xmath74 is a explicit function depending only of @xmath73 ( see ) . particularly , for the limit values of the coordinates ,
we get @xmath75 where @xmath76 , @xmath77 , @xmath78 , @xmath79 and @xmath72 is independent of @xmath73 and @xmath0 .
let @xmath80 be a bounded domain and @xmath0 a real positive number .
we consider functions @xmath81 such that for constants @xmath82 @xmath83 also , we assume that there exist functions @xmath84 and @xmath85 satisfying such that , as @xmath4 , @xmath86 first , we address the problem with dirichlet boundary conditions . when @xmath4 the natural limit problem for is the following @xmath87 where @xmath88 and @xmath89 are given in .
the main result is the following : [ teo_2.gral ] let @xmath81 satisfying and .
then the first non - trivial curve of problem @xmath90 converges to the first non - trivial curve of the limit problem @xmath91 as @xmath4 in the sense that @xmath92 and @xmath93 @xmath94 .
[ remconv ] let us consider the weighted @xmath7laplacian problem @xmath95 where @xmath96 is a function such that @xmath97 weakly * in @xmath98 as @xmath0 tends to zero .
it is well - known that the first eigenvalue of converges to the first eigenvalue of the @xmath7laplacian equation with weight @xmath99 , see for instance @xcite .
the fact that the trivial lines of @xmath100 are defined by @xmath101 and @xmath102 it allows us to affirm the convergence of the trivial lines to those of the limit problem .
using the variational characterization of the second ( variational ) eigenvalue of @xcite , theorem [ teo_2.gral ] implies the convergence of the second ( variational ) eigenvalue of to those of the limit problem , which recover result recently proved in @xcite for the case of the weighted @xmath7laplacian .
however , the results in @xcite consider a more general class of quasilenar operators and @xmath0-dependence on the operator as well . in the important case of periodic homogenization , i.e. , when @xmath62 and @xmath63 where @xmath103 and @xmath104 are @xmath64periodic functions , @xmath59 being the unit cube in @xmath2 , we have that @xmath105 and @xmath106 are real numbers given by the averages of @xmath103 and @xmath104 over @xmath59 , respectively .
consequently , the limit problem becomes @xmath107 in this case , besides the convergence of the curves given in theorem [ teo_2.gral ] and remark [ remconv ] , we obtain the convergence rates .
first , by using the variational characterization of the first eigenvalue of we analyze the trivial lines of @xmath100 : [ teo_1 ] let @xmath81 weights satisfying and given in terms of @xmath64periodic functions @xmath39 in the form @xmath108 and @xmath109 .
let us denote by @xmath110 , @xmath111 , @xmath112 and @xmath113 to the first eigenvalue of equation with weight @xmath13 , @xmath14 , @xmath114 and @xmath115 , respectively .
then @xmath116 with @xmath117 given by @xmath118 where @xmath119 is the first eigenvalue of the dirichlet @xmath7laplacian and @xmath120 . from theorem [ teo_1 ]
it follows the convergence rates of the trivial lines of @xmath100 : if @xmath121 , we get @xmath122 where @xmath123 belongs to the line @xmath124 .
analogously for @xmath125 .
related to the first nontrivial curve of @xmath100 we obtain : [ teo_2 ] under the same considerations of theorem [ teo_2.gral ] , if the weights @xmath13 and @xmath14 are given in terms of @xmath64periodic functions @xmath39 in the form @xmath108 and @xmath109 , for each @xmath126 , we have the following estimates @xmath127 is given explicitly by @xmath128 where @xmath129 and @xmath130 are the poincar s constant in @xmath131 and @xmath132 , respectively , @xmath133 is the second dirichlet @xmath7laplacian eigenvalue in @xmath1 and @xmath74 is defined by @xmath134 according to proposition 34 and proposition 35 in @xcite , when @xmath135 the limits of @xmath136 as @xmath137 and @xmath138 as @xmath53 can be characterized in terms of the first eigenvalues of weighted @xmath7laplacian problems .
moreover , @xmath139 and @xmath140 .
similarly for @xmath141 and @xmath142 .
consequently , by using the estimates obtained in theorem [ teo_1 ] , it is easy to compute the convergence rates in the limit cases when the periodic case is considered , namely @xmath143 now we focus our attention on the neumann boundary conditions case , i.e. , we study the homogenization of the spectrum of the fucik problem @xmath144 where @xmath145 denotes the unit exterior normal . in section [ sec.neu ]
we study the limit problem associated to and the homogenization of the first non - trivial curve of its spectrum .
we obtain similar results to the dirichlet case : in theorem [ teo_2.gral.neu ] we study the convergence in general settings ; in theorem [ teo_2.neu ] we deal with the periodic case , obtaining convergence rates similar to those of theorem [ teo_2 ] .
we begin with the proof of the theorem [ teo_1 ] . for that , we will use a technical result proved in @xcite that is essential to estimate the rate convergence of the eigenvalues since allows us to replace an integral involving a rapidly oscillating function with one that involves its average in the unit cube .
[ theorem 3.4 in @xcite . ] [ teo_n_dim ] let @xmath1 be a bounded domain in @xmath2 , @xmath3 .
let @xmath146 be a @xmath64periodic function , being @xmath147^n$ ] the unit cube in @xmath2 , such that @xmath148 for @xmath149 constants .
then @xmath150 for every @xmath151 where @xmath152 , @xmath80 bounded domain and @xmath153 . here , @xmath129 is the optimal constant in poincar s inequality in @xmath131 which satisfies @xmath120 .
[ teo_n_dim_rem ] sometimes it will be useful to use an inequality involving only the gradient . by using poincar s inequality
we can bound @xmath154 . with the same assumptions of theorem [ teo_n_dim ]
, it allow us to rewrite inequality as @xmath155 where @xmath156 .
@xmath112 can be characterized variationally as @xmath157 for some @xmath158 .
we can bound @xmath159 by using theorem [ teo_n_dim ] , and it follows that @xmath160 where @xmath161 . by replacing and , in we get @xmath162 in a similar way , interchanging the roles of @xmath110 and @xmath112 we obtain @xmath163 by using immediately it follows that @xmath164 where @xmath119 is the first eigenvalue of the dirichlet @xmath7laplacian . from equations , and it follows the result . in the next lemma we obtain upper bounds for the coordinates of the first curve of @xmath165 .
[ lema.2 ] let @xmath39 satisfying @xmath166 and let @xmath167 .
then for each @xmath168 , @xmath169 with @xmath74 defined by @xmath170 where @xmath171 are given by and @xmath133 is the second eigenvalue of the @xmath7laplacian equation in @xmath1 without weights and dirichlet boundary conditions .
let @xmath168 .
when the parameter @xmath172 we can bound @xmath173 let @xmath174 be the second eigenvalue of the problem with weight @xmath33 .
it satisfies that @xmath175 . by using the assumptions over @xmath33 , we can bound @xmath174 by @xmath176 , where @xmath133 is the second eigenvalue of the @xmath7laplacian equation with dirichlet boundary conditions on @xmath1 .
analogously for @xmath177 .
we get @xmath178 when @xmath179 the following bound holds for the second coordinate of @xmath180 @xmath181 by multiplying by @xmath182 and by using that @xmath183 we have @xmath184 being @xmath185 , it follows that @xmath186 by using , and the relation @xmath187 the conclusions of the lemma follows .
the following proposition gives the monotonicity of @xmath45 : [ propo ] if @xmath188 and @xmath189 a.e .
, then @xmath190 where @xmath45 is defined by .
in the next lemma we obtain lower bounds for the coordinates of the first curve of @xmath165 .
[ lema.3 ] let @xmath39 satisfying @xmath166 and let @xmath191 .
then for each @xmath168 , @xmath192 with @xmath193 defined by @xmath194 where @xmath195 is a positive constant depending only of the bounds given in .
let @xmath168 . when the parameter @xmath172 we can bound bellow @xmath196 using the relation @xmath187 we obtain @xmath197 similarly , when @xmath179 we have @xmath198 and again , by the relation between @xmath42 and @xmath43 we get @xmath199 using and proposition [ propo ] , we can bound bellow @xmath200 and the result follows .
now we are able to prove theorem [ teo_2 ] . for each fixed value of @xmath12 , by together with the monotonicity of @xmath45 provided by proposition [ propo ]
, we can assert that there exist two curves @xmath201 and @xmath202 such that delimit above and below to the curve @xmath203 .
it follows that for each fixed value of @xmath73 , @xmath204 and @xmath205 are bounded .
let @xmath206 be a point belonging to the curve @xmath203 and let @xmath207 be the point obtained when @xmath4 .
let us see that it belongs to @xmath208 .
fixed a value of @xmath12 and by using , the inverse of @xmath209 can be written as @xmath210 } b_{m_{\varepsilon},n_{\varepsilon}}(u)\ ] ] where @xmath211 for @xmath47 $ ] and @xmath212 @xmath48 and @xmath49 being the functionals defined in . by
and we have the following characterization for the inverse of @xmath204 @xmath213 similarly , we can consider an equation analog to for the representation of the inverse of @xmath214 .
let @xmath215 and @xmath216 such that @xmath217 in order to find a bound for @xmath218 we use @xmath219 , which is admissible in its variational characterization , @xmath220 as @xmath151 , it follows that @xmath221 and @xmath222 belong to @xmath223 .
this allows us to estimate the error by replacing the oscillating weights by their averages by using remark [ teo_n_dim_rem ] . for each fixed function
@xmath224 we bound @xmath225 where @xmath226 and @xmath226 are the constants given in remark [ teo_n_dim_rem ] .
as @xmath227 we have @xmath228 so , from and , taking @xmath229 we get @xmath230 taking infimum over the functions @xmath231 in @xmath232 together with and we obtain @xmath233 letting @xmath234 we get @xmath235 in a similar way , interchanging the roles of @xmath10 and @xmath141 we obtain the inequality @xmath236 from equations and it follows that @xmath237 we can bound the expression as @xmath238 where @xmath239 is given by and @xmath133 is the second eigenvalue of the dirichlet @xmath7laplacian . from the convergence of @xmath10 together with it follows the convergence of @xmath11 and of the whole curve .
the proof of theorem [ teo_2.gral ] , where general weights are considered , is analogous to that of theorem [ teo_2 ] but we need a result similar to theorem [ teo_n_dim ] that works without assuming periodicity .
it is found in the following theorem .
[ teo_n_dim.gral ] let @xmath80 be a bounded domain .
let @xmath240 be a function such that @xmath241 for @xmath149 constants and @xmath242 weakly * in @xmath98 .
then for every @xmath243 , @xmath244 where @xmath152 .
the weak * convergence of @xmath240 in @xmath98 says that @xmath245 for all @xmath246 .
particularly , @xmath243 implies that @xmath247 , it follows that @xmath248 and the result is proved .
the argument follows exactly as in the proof of theorem [ teo_2 ] using the theorem [ teo_n_dim.gral ] instead of the theorem [ teo_n_dim ] .
let @xmath1 be a bounded domain in @xmath2 , @xmath3 with lipschitz boundary and let @xmath39 be two weights satisfying .
we consider the following asymmetric elliptic problem with neumann boundary conditions @xmath249 where @xmath250 denotes the unit exterior normal .
let @xmath99 be a weight satisfying .
now , @xmath251 is a principal eigenvalue of @xmath252 with constants as eigenfunctions .
moreover , the positivity of @xmath253 guaranties that @xmath251 is the unique nonnegative principal eigenvalue , see @xcite .
consequently , the fucik spectrum @xmath254 clearly contains the lines @xmath255 and @xmath256 .
we denote by @xmath24 the set @xmath257 without these two lines . in this case , when @xmath37 only a full description of the first nontrivial curve of @xmath21 , which we will denote by @xmath38 .
moreover , in @xcite ( see theorem 6.1 ) a characterization similar to the dirichlet case is given : @xmath258 where @xmath42 and @xmath43 are continuous functions defined by @xmath259 , @xmath187 and @xmath45 is given by @xmath260 with @xmath261 $ ] , the functionals @xmath48 and @xmath49 given by , and @xmath262 let @xmath13 and @xmath14 be two functions such that satisfy and .
we consider the following problem depending on @xmath12 @xmath263 as we made with the dirichlet equation , we want to study the behavior of the first non - trivial curve in the spectrum of as @xmath4 .
when @xmath0 tends to zero in , according to we obtain the following limit equation @xmath264 analogously to theorem [ teo_2.gral ] , we obtain the following result of convergence : [ teo_2.gral.neu ] let @xmath81 satisfying , and .
then the first non - trivial curve of problem @xmath265 converges to the first non - trivial curve of the limit problem @xmath266 as @xmath4 in the sense that @xmath92 , @xmath93 @xmath94 .
when the case of periodic homogenization is considered , i.e. , @xmath62 and @xmath63 with @xmath103 and @xmath104 @xmath64periodic functions , @xmath59 being the unit cube in @xmath2 , the limit functions @xmath88 , @xmath89 given in are equal to the averages of @xmath103 and @xmath104 over @xmath59 , respectively , i.e. , @xmath105 and @xmath106 .
now , like in the dirichlet case , in addition to the convergence of the first non - trivial curve , we obtain the convergence rates : [ teo_2.neu ] let @xmath267 , @xmath37 be a bounded domain with @xmath268 boundary . under the same considerations of theorem [ teo_2.gral.neu ] , if the weights @xmath13 and @xmath14 are given in terms of @xmath64periodic functions @xmath39 in the form @xmath108 and @xmath109 , for each @xmath126 we have the following estimate @xmath269 is a constant independent of @xmath0 and @xmath73 , and @xmath74 is given by . to prove theorem [ teo_2.neu ] arguments used in the dirichlet case fail .
this is due to the fact that now the functions space is @xmath270 but theorem [ teo_n_dim ] holds only for functions in @xmath271 .
the fact of enlarge the set of test functions is reflected in the need for more regularity of the domain @xmath1 .
we will prove the following result which works with functions belonging to @xmath270 .
[ teo.neu ] let @xmath80 be a bounded domain with @xmath268 boundary and denote by @xmath59 to the unit cube in @xmath2 .
let @xmath272 be a @xmath64periodic bounded function .
then for every @xmath243 there exists a constant @xmath72 independent of @xmath0 such that @xmath273 where @xmath274 and @xmath275 . unlike to theorem [ teo_n_dim ]
, we are not able to compute explicitly the constant @xmath72 in in theorem [ teo.neu ] .
we begin this section by proving some auxiliary results that are essential to prove theorem [ teo.neu ] .
the next lemma is a generalization for @xmath276 of oleinik s lemma @xcite .
[ lemaaux ] let @xmath80 be a bounded domain with @xmath268 boundary and , for @xmath215 , let @xmath277 be a tubular neighborhood of @xmath278 , i.e. @xmath279
. then there exists @xmath280 such that for every @xmath281 and every @xmath282 we have @xmath283 where @xmath72 is a constant independent of @xmath284 and @xmath285 .
let @xmath279 , it follows that @xmath286 are uniformly smooth surfaces . by the sobolev trace theorem we have @xmath287,\ ] ] where @xmath288 is a constant independent of @xmath284 .
integrating this inequality with respect to @xmath284 we get @xmath289 and the lemma is proved .
the next theorem is essential to estimate the rate of convergence of the eigenvalues since it allows us to replace an integral involving a rapidly oscillating function with one that involves its average in the unit cube .
first , we need an easy lemma that computes the poincar constant on the cube of side @xmath0 in terms of the poincar constant of the unit cube .
although this result is well known and its proof follows directly by a change of variables , we choose to include it for the sake of completeness .
[ poincare ] let @xmath59 be the unit cube in @xmath2 and let @xmath130 be the poincar constant in the unit cube in @xmath290 , i.e. @xmath291 where @xmath292 is the average of @xmath231 in @xmath59 .
then , for every @xmath293 we have @xmath294 where @xmath295 .
let @xmath293 .
we can assume that @xmath296 .
now , if we denote @xmath297 , we have that @xmath298 and by the change of variables formula , we get @xmath299 the proof is now complete .
[ teo_n_dim_v2 ] let @xmath80 be a bounded domain with smooth boundary and denote by @xmath59 the unit cube in @xmath2 .
let @xmath272 be a @xmath59-periodic bounded function such that @xmath300 .
then the inequality @xmath301 holds for every @xmath243 and @xmath302 , where @xmath72 is a constant independent of @xmath0 , @xmath231 , @xmath285 and @xmath303 are conjugate exponents .
denote by @xmath304 the set of all @xmath305 such that @xmath306 .
set @xmath307 and @xmath308 .
let us consider the functions @xmath309 and @xmath310 given by the formulas @xmath311 for @xmath312 .
then we have @xmath313 the set @xmath314 is a @xmath284-neighborhood of @xmath315 with @xmath316 for some constant @xmath72 , and therefore according to lemma [ lemaaux ] we have @xmath317 as @xmath272 is bounded , we get @xmath318 now , by lema [ poincare ] we get @xmath319 analogously @xmath320 by the definition of @xmath321 we get @xmath322 finally , since @xmath323 and since @xmath272 is @xmath64periodic , we get @xmath324 now , combining , , , and we can bound by @xmath325 this finishes the proof .
now we are ready to proof theorem [ teo.neu ] : the result follows applying theorem [ teo_n_dim_v2 ] to @xmath326 and taking @xmath327 .
[ remar ] let us observe that @xmath243 is solution of equation if and only if @xmath231 is solution of equation @xmath328 with neumann boundary conditions , where @xmath329 and @xmath330 .
the main advantage between consider equations and is the fact that in the second one the functional @xmath331 defined in becomes in @xmath332 which involves both @xmath333 and the function @xmath231 .
the proof is similar to that of theorem [ teo_2 ] for the dirichlet case .
according to remark [ remar ] we consider equation .
let @xmath334 be a point belonging to the curve @xmath203 and let @xmath335 be the point obtained when @xmath4 .
it follows that @xmath335 belongs to the spectrum of the limit equation .
let us see that it belongs to @xmath336 .
the main difference is that in the characterization of @xmath209 , now we are considering @xmath337 with @xmath261 $ ] .
fixed a value of @xmath12 we write @xmath338 by and we have the following characterization of @xmath339 @xmath340 similarly , we can consider an equation analog to for the representation of @xmath341 .
let @xmath215 and @xmath342 such that @xmath343 in order to find a bound for @xmath344 we use @xmath345 , which is admissible in its variational characterization , @xmath346 to bound @xmath347 we look for bounds of the two quotients in .
since @xmath243 , by theorem [ teo.neu ] we obtain that @xmath348 for every function @xmath349 we have that @xmath350 by using young inequality , for each @xmath282 @xmath351 from it follows that @xmath352 and similarly @xmath353 to bound the second quotient in , we use again theorem [ teo.neu ] and to obtain @xmath354 and similarly @xmath355 now , from equations , together with and we get @xmath356 then combining , , and we find that @xmath357 letting @xmath234 we get @xmath358 in a similar way , interchanging the roles of @xmath359 and @xmath347 , we obtain @xmath360 from and we arrive at @xmath361 now , using lemma [ lema.2 ] , latexmath:[\[\begin{aligned } where @xmath72 is a constant independent of @xmath0 and @xmath73 , and @xmath239 is given by . here
, lemma [ lema.2 ] holds in the neumann case , but now we have @xmath169 with @xmath133 the second eigenvalue of the @xmath7laplacian equation on @xmath1 with neumann boundary conditions . from the convergence of @xmath363 and it follows the convergence of @xmath364 and of the whole curve .
as theorem [ teo_n_dim.gral ] holds for functions belonging to @xmath270 with @xmath1 any bounded domain in @xmath2 , this proof is analogous to those of theorem [ teo_2.gral ] .
i want to thank julin fernndez bonder for valuable help and many stimulating discussions on the subject of the paper .
siham el habib and najib tsouli , _ on the spectrum of the @xmath365-laplacian operator for neumann eigenvalue problems with weights _ ,
proceedings of the 2005 oujda international conference on nonlinear analysis ( san marcos , tx ) , electron . j. differ .
14 , southwest texas state univ . , 2006 , pp .
181190 ( electronic ) .
carlos e. kenig , fanghua lin , and zhongwei shen , _ convergence rates in @xmath366 for elliptic homogenization problems , to appear in archive for rational mechanics and analysis _ , arxiv:1103.0023v1 ( 2011 ) .
o. a. olenik , a. s. shamaev , and g. a. yosifian , _ mathematical problems in elasticity and homogenization _ , studies in mathematics and its applications , vol . 26 , north - holland publishing co. , amsterdam , 1992 .
|
in this work we consider the fucik problem for a family of weights depending on @xmath0 with dirichlet and neumann boundary conditions .
we study the homogenization of the spectrum .
we also deal with the special case of periodic homogenization and we obtain the rate of convergence of the first non - trivial curve of the spectrum .
|
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.