article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
every parcel of gas along the line of sight to a distant quasar will selectively absorb certain wavelengths of continuum light of the quasar due to the presence of the various chemical elements in the gas . through the analysis of these quasar absorption lines we can study the spatial distributions , motions , chemical enrichment , and ionization histories of gaseous structures from redshift five until the present .
this includes the gas in galaxies of all morphological types as well as the diffuse gas in the intergalactic medium .
-0.15 in figure 1 illustrates many of the common features of a quasar spectrum . the relatively flat quasar continuum and broad emission features are produced by the quasar itself ( near the black hole and its accretion disk ) . in some cases , gas near the quasar central engine also produces `` intrinsic '' absorption lines , most notably , and relatively high ionization metal transitions such as , , and .
these intrinsic absorption lines can be broad [ thousands or even tens of thousands of in which case the quasar is called a broad absorption line ( bal ) qso ] , or narrow ( tens to hundreds of ) .
however , the vast majority of absorption lines in a typical quasar spectrum are `` intervening '' , produced by gas unrelated to the quasar that is located along the line of sight between the quasar and the earth .
a structure along the line of sight to the quasar can be described by its neutral hydrogen column density , @xmath0 , the number of atoms per .
@xmath0 is given by the product of the density of the material and the pathlength along the line of sight through the gas .
each structure will produce an absorption line in the quasar spectrum at a wavelength of @xmath1 , where @xmath2 is the redshift of the absorbing gas and @xmath3 is the rest wavelength of the transition . since @xmath4 , the redshift of the quasar , these absorption lines form a `` forest '' at wavelengths blueward of the emission .
the region redward of the emission will be populated only by absorption through other chemical transitions with longer @xmath5 .
historically , absorption systems with @xmath6 have been called forest lines , those with @xmath7 are lyman limit systems , and those with @xmath8 are damped systems .
the number of systems per unit redshift increases dramatically with decreasing column density , as illustrated in the schematic diagram in figure 2 .
lyman limit systems are defined by a sharp break in the spectrum due to absorption of photons capable of ionizing , i.e. those with energies greater than @xmath9 .
the optical depth , @xmath10 , of the break is given by the product @xmath11 , where the cross section for ionization of hydrogen , @xmath12 , ( and the flux is reduced by the factor @xmath13 ) .
the energy dependence of @xmath14 leads to a recovery of the lyman limit break at higher energies ( shorter wavelengths ) , unless @xmath15 ( see figure 1 ) .
the curve of growth describes the relationship between the equivalent width of an absorption line , @xmath16 , ( the integral of the normalized profile ) and its column density , @xmath17 . figure 3 shows that for small @xmath0 the number of absorbed photons , and therefore the flux removed , increases in direct proportion to the number of atoms .
this is called the linear part of the curve of growth .
as @xmath17 is increased the line saturates so that photons are only absorbed in the wings of the lines ; in this regime the equivalent width is sensitive to the amount of line broadening ( characterized by the doppler parameter @xmath18 ) , but does not depend very strongly on @xmath0 .
this is the flat part of the curve of growth . finally , at @xmath8 , there are enough atoms that the damping wings of the line become populated and the equivalent width increases as the square root of @xmath0 , and is no longer sensitive to @xmath18 .
in addition to the ( @xmath19 ) and higher order ( @xmath20 ) lyman series lines , quasar spectra also show absorption due to different ionization states of the various species of metals .
figure 1 illustrates that the damped system at @xmath21 that is responsible for the absorption line at @xmath22 and a lyman limit break at @xmath23 also produces absorption at @xmath24 due to the presence of in the absorbing gas at that same redshift . like
many of the strongest metal lines seen in quasar spectra , is a resonant doublet transition due to transitions from @xmath25 energy levels to the @xmath26 and to the @xmath27 energy levels .
( the left superscript `` 2 '' represents the number of orientations of the electron spin , the letter @xmath28 or @xmath29 represents the total orbital angular momentum , @xmath30 , and the right subscript represents the total angular momentum , @xmath31 . )
doublet transitions are easy to identify .
the dichotomy between rest wavelength and redshift is resolved because the observed wavelength separation of the doublet members increases as @xmath32 .
table 1 lists some of the metal lines that are commonly detected for intervening absorption systems .
many of these are only strong enough to be observable for quasar lines of sight that pass through the higher @xmath0 regions of galaxies . 0.3 in 2.5in[]lr + transition & @xmath33 [ ] + & @xmath34 + & 972.537 + & 1025.722 + & 1215.670 + 1393 & 1393.755 + 1402 & 1402.770 + 1548 & 1548.195 + 1550 & 1550.770 + 2382 & 2382.765 + 2600 & 2600.173 + 2796 & 2796.352 + 2803 & 2803.531 + 0.56 in
the history of quasar absorption lines began within a couple of years of the identification of the first quasar in 1963 . in 1965 ,
gunn and peterson considered the detection of flux blueward of the emission line in the quasar 3c 9 , observed by schmidt , and derived a limit on the amount of neutral hydrogen that could be present in intergalactic space . in that same year , bahcall and salpeter predicted that intervening material should produce observable discrete absorption features in quasar spectra .
such features were detected in 1967 in the quasar pks @xmath35 by greenstein and schmidt , and in 1968 in phl @xmath36 by burbidge , lynds , and stockton . by
1969 many intervening systems had been discovered , and bahcall and spitzer proposed that most with metals were produced by the halos of normal galaxies . as more data accumulated , the sheer number of forest lines strongly supported the idea that galactic and intergalactic gas , and not only material intrinsic to the quasar , is the source of most quasar absorption lines . in the 1980
s many more quasar spectra were obtained and many large statistical surveys of the different classes of absorption line systems were published .
the emphasis was to characterize the number of lines per unit redshift , @xmath37 , stronger than some specified equivalent width limit . with 4m class telescopes [ equipped with charge coupled device ( ccd ) detectors ] it was possible to conduct surveys with a spectral resolution of @xmath38 .
the spectral resolution is defined as @xmath39 , so that @xmath40 corresponds to @xmath41 or @xmath42 at @xmath43 .
separate surveys were conducted for lines , doublets , doublets , and also for lyman limit breaks , all as a function of redshift .
the line is observable in the optical part of the spectrum for @xmath44 , for @xmath45 , for @xmath46 , and the lyman limit break for @xmath47 .
however , a break is also easily identified in lower resolution space based uv spectra , which extended lyman limit surveys to lower redshift . in order to consider the cross section of the sky covered by the different populations
, it can be assumed that absorption will be observed for all lines of sight within some radius of every luminous galaxy ( @xmath48 ) .
( @xmath49 represents the schechter luminosity , i.e. the transition between the exponential and the power law forms of the luminosity function , and corresponds to a @xmath50band absolute magnitude of @xmath51 ) . to explain the observed @xmath37 at @xmath52 , this radius would be @xmath53 kpc for strong ( detection sensitivity @xmath54 ) , and @xmath55 kpc for strong ( detection sensitivity @xmath56 ) and also for lyman limit systems , implying that the latter two populations are in fact produced in the same gas .
the higher @xmath0 damped absorbers would be produced within @xmath57 kpc of the center of each galaxy , while the forest lines would require a considerably larger region , hundreds of kpcs around each galaxy to produce a cross section consistent with the observed number of weak lines .
up until the 1990 s , the focus of quasar absorption line work was to separately consider the properties of the individual classes of absorbers ( eg .
forest or absorbers ) . in the 1990 s , however , three different observational advances led to recognition of the direct connections between the different classes of quasar absorption lines , and of direct associations with the population of galaxies : \1 .
deep images of quasar fields could be obtained , and redshifts of the galaxies in the field could be determined from low resolution spectra .
steidel found that whenever absorption with @xmath58 is observed , a luminous galaxy ( @xmath59 ) is found within an impact parameter of @xmath60 with a redshift coincident with that determined from the absorption lines . also , it is rare to find a galaxy within this impact parameter that does not produce absorption .
there appears to be a one to
one correspondence between strong absorption and luminous galaxies .
the absorbing galaxies span a range of morphological types .
. the high resolution spectrograph on the keck i @xmath61-meter telescope made it possible to obtain quasar spectra at a resolution of @xmath62 , which corresponds to @xmath63 .
the previous surveys with resolution of order hundreds of identified absorption due to entire galaxies and their environments . with @xmath64 resolution it became possible to resolve structure within a galaxy : the clouds in its halo , the interstellar medium of its disk , and the satellites and infalling gas clouds in its environment .
figure 4 is a dramatic illustration of this contrast for the absorber at @xmath65 toward the quasar pg @xmath66 .
-3.6 in \3 .
the faint object spectrograph ( fos ) on the hubble space telescope provided resolution @xmath38 in the uv , from @xmath67@xmath68 .
observations of forest clouds could be extended from @xmath69 down to the present epoch .
furthermore , absorption from a given galaxy could be observed in numerous transitions ; if was observed in the optical , the lyman series and could be studied in the uv ( see figure 4 ) . with information on transitions with a range of ionization states , consideration of the degree of ionization ( related to the gas density and the intensity and shape of the ionizing radiation field ) and
the multiple phase structure of galactic gas became possible .
no longer is analysis of absorption lines in quasar spectra an esoteric subject .
it has developed into a powerful tool to be used in the study of galaxy evolution ( eg .
similar to imaging the stellar components of the galaxies ) . at least in principle
, quasar spectra can be used for an unbiased study of the gaseous environments of galaxies from the present back to the highest redshifts at which quasars are observed .
gas structures smaller than @xmath70 @xmath71 can be detected if they are intercepted by the quasar line of sight , irrespective of whether they emit light . through the tool of quasar absorption lines , proto
galactic structures and low surface brightness galaxies can be studied as well as high luminosity galaxies .
with high resolution spectra of quasars , it is possible to consider the physical conditions of the gaseous structures that produce absorption .
however , it is challenging to separate the various effects that `` shape '' the spectral features in the different chemical transitions .
the absorption profiles observed for the different chemical transitions are determined by a combination of the spatial distribution of material along the line of sight , its bulk kinematics , temperature , metallicity , and abundance pattern .
the ionization structure is influenced by gas densities and by the uv radiation field , which is a combination of the extragalactic background radiation due to the accumulated effect of quasars and stellar photons escaped from galaxies ( and corrected for absorption by the intergalactic medium ) .
the shape of an absorption line can be modeled with a voigt profile , which is a combination of the natural , quantum mechanical lorentzian broadening and the gaussian broadening caused by the thermal and turbulent motions in the gas .
several voigt profiles can be blended together to form an overall complex absorption feature ( see figure 4 ) .
the `` width '' of a single voigt profile is characterized by the doppler parameter , @xmath18 ( expressed in velocity units and related to the gaussian @xmath14 by @xmath72 ) .
physically , the doppler parameter is the sum of thermal and turbulent components , @xmath73 , where @xmath74 is the temperature of the gas , and @xmath75 is the mass of an atom .
-3.0 in two of the simplest types of organized kinematics in galaxies are illustrated in figure 5 : clouds distributed in a rotating disk , and radial infall of clouds in a spherical distribution .
here , absorbers are used as an example , but the same kinematic arguments would apply to other transitions . for radial infall , clouds can be distributed over the range of velocities , with a tendency for a `` double peak '' from material that is redshifted and blueshifted but with a considerable amount of variation if there are typically several discrete clouds along the line of sight . a rotating disk with a vertical velocity dispersion characteristic of a spiral galaxy disk ( @xmath61@xmath76 ) will have clouds superimposed in velocity space , and an overall kinematic spread of tens of .
strong absorption has been found to arise along nearly all lines of within @xmath77 kpc of normal galaxies ( i.e. the covering factor is nearly unity within that radius ) .
the large variety of kinematics evident in absorption profiles is , in fact , consistent with a superposition of disk and radial infall ( halo ) motions , and not with just one or the other .
in addition to these simple , toy models , insights can be gleaned by passing lines of sight through the structures in cosmological n body / hydrodynamic simulations . in a few studies ,
metals have been added uniformly throughout the simulation box and photoionization models used to predict the absorption expected from different structures .
this is especially important for establishing the kinematics that would be observed from the process of structure formation at high redshifts .
consider a cloud of material , modeled by a plane parallel slab with a certain total column density of hydrogen , @xmath78 , and with a constant total number density @xmath79 along the line of sight .
the cloud is also characterized by its metallicity , @xmath80 , which is the ratio of fe / h expressed relative to the solar value , @xmath81 , and by an abundance pattern ( the abundance ratios of all other elements to fe ) .
the degree of ionization in the gas depends upon the intensity and shape of the spectrum of ionizing radiation .
the intensity is characterized by the ionization parameter , @xmath82 , which is the ratio of the number density of photons at the lyman edge to the number density of hydrogen ( @xmath83 , where @xmath84 is the total number density of electrons ) .
the larger the value of @xmath85 , the more ionized the gas .
collisional ionization can also be an important process for some absorption systems with gas at high temperatures ( hundreds of thousands of degrees ) .
photoionization equilibrium models typically yield temperatures of tens of thousands of degrees .
once the metallicity , abundance pattern , ionization parameter , and spectral shape are specified the equations of radiative transfer can be solved to find the column densities of all the different ionization states of various chemical elements .
figure 6 illustrates , for @xmath86 and @xmath87 , the dependence of column densities of various transitions on the ionization parameter , @xmath85 . for optically thin gas [ @xmath6 ] ,
the column density ratios of the various metal transitions are not dependent on the overall metallicity , i.e. the curves shift vertically in proportion to @xmath80 . for optically thick gas , ionization structure develops , with an outer ionized layer around a neutral core , and there is no simple scaling relation with metallicity . in practice ,
if we assume that a cloud has a simple , single phase structure , the ratios of the column densities can be used to infer the ionization parameter , which relates to the density of the gas .
however , the abundance pattern can differ from the solar abundance pattern because of differing degrees of depletion onto dust , or because of different processing histories .
most of the so called @xmath88 particle nuclei ( such as mg and si ) are synthesized primarily by type ii supernovae during the early history of a galaxy when most massive stars form and quickly evolve to reach their end states . on the other hand , the fe
group elements are primarily produced by type ia supernovae , and therefore build up over a longer timescale . in the basic picture of galaxy evolution ,
the halo stars are formed early , have been enriched only by type ii supernova , and therefore are @xmath88element enhanced .
younger disk stars have incorporated also the type ia processed material and therefore have relatively larger fe group abundances . ideally , several different ionization states of the same chemical element are observed so that there is no ambiguity between the ionization parameter and the abundance pattern , but this has generally not yet been possible because of limited wavelength coverage at high resolution .
examples of the variation of column density ratios with velocity in two absorption systems are shown in figures [ fig : fig7 ] and [ fig : fig8 ] . in figure 7 , @xmath89 varies by an order of magnitude over the four components in the @xmath90 system toward the quasar q@xmath91 .
this represents a variation of an order of magnitude in the ionization parameter ( @xmath92 ) , or an order of magnitude variation in the abundance pattern .
figure 8 is a very unusual system with two clouds separated by only @xmath76 in velocity , one of which has a silicon to aluminum ratio similar to the milky way ism , and the other which requires a significant enhancement of aluminum .
the gaseous component of the the milky way and nearby galaxies have phase structure ( i.e. spatial locations with different densities and/or temperatures ) .
examples are the disk / halo interface ( galactic coronae ) and the cold , warm , and hot phases of the interstellar medium . from photoionization models , it is not usually possible to generate absorption that is simultaneously consistent with all observed chemical transitions for a given system .
for example , in single cloud systems , figure 6 ( with @xmath86 ) shows that if is detected at a similar column density to , the ionization parameter must be small , and @xmath93 can not be large .
many systems have absorption which exceeds this limit and requires a higher ionization ( lower density ) phase ; generally , this phase must have structure over a large velocity range ( a large `` effective '' doppler parameter ) .
the @xmath65 system toward the quasar pg 1206 + 459 is another case that requires multiphase structure .
the observed profile in figure 4 is much too strong for this absorption to arise in the same clouds that produce the , even if their ionization parameters are pushed to the largest values consistent with the data .
future quasar absorption line studies will combine insights gained from detailed analyses of individual systems with conclusions drawn from the large statistical samples assembled over cosmic time .
evolution of the ensemble of absorption profiles generated by the universal collective of intervening structures is a result of the combined effects of numerous processes .
these include growth of structure , star formation , morphological evolution of galaxies , galaxy mergers , and changes in the extragalactic background radiation . here , we summarize the best present statistical data and likely interpretations for the different classes of absorbers . the number of lines per unit redshift for various populations of absorbers is represented by a power law @xmath94 . for a universe with only the cosmological evolution due to expansion , @xmath95 for deceleration parameter @xmath96 and @xmath97 for @xmath98 .
-0.5 in the forest evolves away dramatically from high to low redshift , as is strikingly clear from the spectra of @xmath99 and @xmath100 quasars in figure 9 .
the evolution of the lines with @xmath101 can be characterized by a double power law with @xmath102 for @xmath103 and @xmath104 for @xmath105 .
help in understanding the physical picture has come from sophisticated n body / hydrodynamic simulations that incorporate the gas physics and consider cosmological expansion of the simulation box .
the dynamical evolution of the gas can be described as outflow from the centers of voids to their surrounding shells , and flows along these sheets toward their intersections where the densest structures form .
this picture is consistent with observational determinations of the `` sizes '' of structures .
it is difficult to obtain direct measurements of sizes except in some special cases to use `` double lines of sight '' , close quasar pairs , either physical or apparent due to gravitational lensing .
if the spectra of the two quasars both have a absorption line at the same wavelength that implies a `` structure '' which covers both lines of sight . from these studies , it is found that `` structures '' are at least hundreds of kpc in extent . at redshifts
@xmath106 to @xmath107 @xmath37 for forest absorption is quite large , but it is declining very rapidly over that range .
this dramatic evolution in the number of forest clouds is mostly due to the expansion of the universe , with a modest contribution from structure growth . at @xmath108 ,
the extragalactic background radiation field is falling , and structures are becoming more neutral .
therefore , the more numerous , smaller @xmath109 structures are observed at a larger @xmath0 and this will counteract the effect of expansion , thus slowing the decline of the forest . the high redshift forest was once thought to be primordial material , but in fact it is observed to have a metallicity of @xmath110% solar , even at @xmath111 . for @xmath112 ,
the expected @xmath113 would be below the detection thresholds of current observations , so truly pristine material still eludes us .
perhaps it does not exist . to spread metals all through the intergalactic medium
may have required a `` pre galactic '' population of stars at @xmath114 that polluted all of intergalactic space .
the @xmath37 of lyman limit systems is consistent with that of strong absorbers [ with @xmath58 ] over the redshift range for which both have been observed , @xmath45 . for @xmath58 , @xmath115 ,
consistent with no evolution . for even stronger systems [ @xmath116 ]
, @xmath37 increases more dramatically with @xmath117 , with @xmath118 .
the number of systems ( equivalent width distribution ) continues to increase down to the sensitivity of the best surveys , @xmath119 , such that @xmath120 at @xmath100 .
the `` weak '' absorbers are therefore more common than the strong systems [ @xmath58 ] known to be associated with luminous galaxies . unlike the strong absorbers , the weak absorbers are
sub - lyman limit systems ( they do not have lyman limit breaks ) , and no galaxies have been identified at the redshift of absorption . yet
, photoionization models indicate that the metallicities of these weak absorbers are at least @xmath61% of the solar value , and in some cases comparable to solar .
they are a varied population : some have relatively strong while others have no detected , and some have strong that requires a separate phase while others have no detected .
those with strong are constrained to be smaller than 10 pc ( the ionization parameter must be small and @xmath84 large as can be seen in figure 6 ) . also ,
since fe is produced primarily by type ia supernovae they must be enriched by a relatively old stellar population .
those with weaker or undetected could be larger ( kpcs or tens of kpcs ) and possibly enriched by type ii supernovae .
candidate environments that could be traced by weak absorption are : remnants of pre
galactic star clusters formed in mini halos at @xmath114 , super star clusters formed in interactions , tidally stripped material , low surface brightness galaxies , and ejected or infalling clouds ( analogous to the milky way high velocity clouds ) . the evolution of @xmath37 for absorbers can be studied in the optical for high redshifts . for @xmath121 and @xmath122 ,
the number decreases with increasing @xmath117 , as @xmath123 . in this same interval
, the number of lyman limit systems is still increasing with redshift , with @xmath124 .
this implies that the dramatic evolution in the number of systems is either due to a change in metallicity or a change in ionization state .
the @xmath37 for systems peaks at intermediate @xmath117 and declines , consistent with no evolution until the present .
combining optical and uv data , and have been compared at @xmath45 .
the fraction of systems with large @xmath125 decreases rapidly with decreasing redshift ; there is a shift toward `` lower ionization systems '' .
it is important to consider that the , , and absorption do not always arise in the same phase .
it is possible that the in many @xmath100 absorption systems arises in a phase similar to the galactic coronae . if the origin of this phase is related to star forming processes in the disk
, then it might be expected to diminish below @xmath126 since the peak star formation rate is passed .
another important trend is the fact that the very strongest absorbers evolve away from @xmath107 until the present .
if we study the kinematic structure of these objects , we find that they commonly have a `` double '' structure , with two separate kinematic regions in the profile .
these objects also have strong which also has separate components around the two regions in the `` double '' structure .
the does not arise primarily in the individual clouds , nor is it in a smooth , `` common halo '' structure that extends in velocity space around the entire profile . as more data are collected on the kinematic structure of various transitions in these `` double '' systems
, it will be interesting to consider the hypothesis that galaxy pairs in the process of merger are responsible . the number of these is thought to have been dramatically larger in the past .
the @xmath8 systems are of particular interest because it is possible to observe many different chemical elements ( such as zn , cr , fe , mn , and ni ) in these objects back to high redshift . metallicities and abundance patterns can be studied and compared to those of old stellar populations in the milky way . back to @xmath111 ,
the metallicity in dlas , as measured by the undepleted element zinc , is about 10% of the solar value , but it may decline at @xmath127 .
the identity of sites responsible for dlas at high @xmath117 remains controversial , but they do contain most of the neutral hydrogen in the universe , from which most of its stars form . the kinematic structure of the absorption profiles of neutral and low ionization species is consistent with the rotation of a thick disk , so that it is possible that these are the @xmath111 progenitors of normal spiral galaxies .
however , this signature is not unique .
it could also be the consequence of directed infall in an hierarchical structure formation scenario .
the higher ionization species show complex kinematics which vary in relation to those of the lower ionization gas ; in some systems they appear to trace relatively similar structure , and in others there are clearly several different phases . at low redshift , many of the galaxies that are responsible for the dla absorption can be directly identified .
these galaxies are a heterogeneous population .
they are not just the most luminous galaxies , but include dwarf and low surface brightness galaxies , and even cases where no galaxy has been identified to sensitive limits . damped absorption does not trace the most luminous objects , but rather it traces the largest neutral gas reservoirs .
an additional selection effect may be important .
the most dust rich galaxies that have the potential to produce dla absorption could produce enough extinction that their background quasars will not be included in quasar surveys . in this way
, the population of dlas that are actually observed could be significantly biased against dusty galaxy hosts .
the next decade will see the synthesis of the various techniques for the study of galaxy evolution , through their stars and through their gas .
higher resolution quasar spectra will be obtained in the ultraviolet ( with the space telescope imaging spectrograph ( stis ) and with the cosmic origins spectrograph ( cos ) on the hst , and later , hopefully , with a larger uv space telescope ) .
it will then be possible to conduct a systematic analysis of the relationships between the different ionization species that trace the different phases of gas in @xmath128 galaxies . in this redshift regime ,
comparisons to the detailed morphological structure and orientations of the absorbing galaxies is possible from hst images .
invaluable insights into the origin of quasar absorption lines have been gleaned from absorption studies of nearby galaxies , for which it is possible to directly observe the processes that are involved .
making more observations of this type will be possible by discoveries of bright quasars that fall behind nearby galaxies .
the discoveries of quasars in large surveys will also include multiple lines of sight behind distant absorption line systems which can be used to produce 3d maps of the structures .
the interstellar medium of the milky way shows structure on sub pc scales , and absorption features can only be resolved with resolution @xmath129 .
such a resolution will soon be available on 8m class telescopes .
this is important for separating blends and for looking for metallicity , ionization , and abundance pattern gradients along the line of sight .
the key low ionization transitions of and are shifted into the near ir region of the spectrum for @xmath130 . very soon , near
ir quasar spectra will be obtained at relatively high resolution ( @xmath131 ) .
also , ir imaging , narrow band techniques , and multi object spectroscopy in the near
ir should provide much more information about absorbing galaxies at higher redshifts .
this will extend evolutionary studies back to an epoch at which formation processes may be contributing significantly to evolution .
articles in review journals :
|
we briefly review , at a level appropriate for graduate students and non - specialists , the field of quasar absorption lines ( qals ) .
emphasis is on the intervening absorbers .
we present the anatomy of a quasar spectrum due to various classes of intervening absorption systems , and a brief historical review of each absorber class ( lyman - alpha forest and lyman limit systems , and metal - line and damped lyman - alpha absorbers ) .
we also provide several heuristic examples on how the physical properties of both the intergalactic medium and the gaseous environments associated with earlier epoch galaxies can be inferred from qals .
the evolution of these environments from z=4 are discussed .
|
cooling of primordial gas is determined by molecular hydrogen , which forms in an expanding universe after recombination ( lepp & shull 1983 , puy et al 1993 , palla , galli & silk 1995 , galli & palla 1998 , stancil , lepp & dalgarno 1998 ) . inside the first virialized dark matter
haloes the fraction of h@xmath3 can reach @xmath4 , which is able to cool gas to 200 k. at lower temperatures energy losses in roto - vibrational h@xmath3 lines become insufficient to cool gas further . on the other hand , owing to a low rotational energy and a large dipole moment hd molecules might be an efficient coolant at @xmath0 k. however , the abundance of hd is fairly sensitive to thermal history of the gas , so that the question of whether hd dominates radiative cooling at low temperatures or its contribution is negligible depends on physical conditions along the evolutionary path .
this circumstance is reflected in conradictory conclusions about the role of hd in thermodynamics of primordial gas : for instance , lepp & shull ( 1983 ) , bromm , coppi & larson ( 2002 ) point out that hd cooling can not dominate in primordial gas when the first objects form , while bougleux & galli ( 1997 ) , puy & signore ( 1997 , 1998 ) , and more recently uehara & inutsuka ( 2000 ) , flower ( 2002 ) , nakamura & umemura ( 2002 ) , flower & pineau des forets ( 2003 ) and machida et al ( 2005 ) find that hd molecules form in significant amount and can play a dominant role in thermal evolution of pregalactic gas .
transformation of h@xmath3 molecules into their isotop analog hd is energetically favoured because of higher binding energy of hd molecules . in equilibrium ( solomon & woolf 1973 , varshalovich & khersonskii 1976 ) = 2n(d)n(h)e^465/t , where factor 2 stems from the difference between the chemical constants of hd and h@xmath3 molecules : @xmath5 ( landau & lifshitz 1969 ) .
it is seen , that at @xmath6 k radiative hd cooling can enhance chemical fractionation of deuterated molecules progressively : a small decrease in temperature results in an increase of @xmath7 , what in turn increases radiative cooling in hd lines ( shchekinov , 1986 ) . in realistic conditions
thermal equilibrium can be reached only in a characteristic time depending on a set of chemical reactions which control the transformation h@xmath8hd . in primordial gas h@xmath3
converges into hd most efficiently in the reaction h@xmath3+d@xmath9
hd+ h@xmath10 ( palla et al . 1995 ) ; other channels are unimportant on the hubble time .
the background h@xmath3 abundance and the frozen - out post - recombination fractional ionization @xmath11 are sufficient to support deuterated h@xmath3 only at the level @xmath12 .
thus , a necessary condition for the h@xmath13 hd convergence to be efficient is that both h@xmath3 and electrons are abundant . in a cosmological pregalactic substrate
such conditions are most naturally fulfiled behind shock waves associated with formation and virialization of dark matter haloes ( tegmark et al .
1997 , barkana & loeb 2001 ) .
when formed in sufficient amount hd molecules can cool gas down to the temperature of the cmb radiation the lowest possible temperature , and thus provide conditions for formation of low - mass stars .
it is clear that this can occur with shock waves whose parameters ( for instance , postshock temperature , total mass involved , and so on ) have reached some critical values . in the hierarchical scenario
dark matter haloes of larger mass form through mergings of smaller objects minihaloes ( see for review barkana & loeb 2001 , ciardi & ferrara 2004 ) . in these conditions
the baryonic component of dark haloes is compressed by shock waves .
apparently , all subsequent evolution of baryons in dark haloes and their condensation into stars is determined by these shock waves .
futhermore , one can consider that the virial state of dark matter haloes and their baryons is reached through formation of shocks .
it is clear , that the physical state of the baryons processed by shocks depends on many factors , such as geometry of a shock front , duration of a compressing flow etc .
however , basics characteristic features can be understood in the framework of a head - on collision of two clouds of equal masses ( suchkov , shchekinov & edelman 1983 , shapiro & kang 1987 ) . in the context of the hierarchical scenario a possible role of shock waves in enhancing h@xmath3 cooling and stimulation of primordial star formation
was studied by yamada & nishi ( 1998 ) , and more recently by cen ( 2005 ) .
radiative cooling in h@xmath3 lines was indeed found extremly efficient in decreasing the postshock temperature to @xmath14 k. in spite of the fact that hd molecules are expected to form at this temperature , hd chemistry was not taken into consideration in this analysis .
the main goal of our study are the conditions when hd molecules can be the dominant cooling agent in a pregalactic gas . in section 2
we describe a simple thermo - chemical model of a shocked gas ; in section 3 the conditions for hd molecules to be the dominant cooling agent are analyzed ; in section 4 we discuss qualitatively a possibility of a cold compressed layer to fragment ; the discussion and summary of the results are given in section 5 . throughout the paper
we assume a @xmath15cdm cosmology with the parameters @xmath16 as inferred from the wilkinson microwave anisotropy probe ( wmap ) , and deuterium abundance @xmath17 ( spergel et al .
for simplicity , we assume a head - on collision of two identical minihaloes , which are virialized at redshift @xmath18 and have density of matter @xmath19 , @xmath20 being the critical density ; the collision is assumed to take place at the same redshift @xmath18 .
the minihaloes move with the relative velocity @xmath21 , where @xmath22 is a 1d velocity dispersion of the larger halo @xmath23 in the center of mass a discontinuity forms at the symmetry plane , and two shocks propagate outwards .
the time for the shock to pass through the entire minihalo is @xmath24 . as it was pointed out by gilden ( 1984 ) for supersonic collisions the rarefaction time in transverse direction is greater than the collision time , which means that a 1d scheme describes the overal picture qualitatively correct .
we assume the post - shock flow to be isobaric ( suchkov et al .
1983 , shapiro & kang 1987 , anninos & norman 1996 ) .
the initial post - shock temperature is @xmath25 where the post - shock gas is assumed to be isothermal , i.e. temperatures of the electrons , ions and neutrals are kept always equal @xmath26 .
chemical kinetics and thermal evolution of the gas are described by the following set of equations @xmath27 @xmath28 where @xmath29 is the boltzmann constant , @xmath30 is the fraction of @xmath31-th species , @xmath32 , @xmath33 are respectively the formation and destruction rates , @xmath34 , @xmath35 are the heating and cooling rates ; @xmath36 includes 14 species : h , h@xmath10 , h@xmath37 , he , he@xmath10 , he@xmath38 , h@xmath3 , h@xmath39 , d , d@xmath10 , d@xmath37 , hd , hd@xmath10 , @xmath40 . in the energy equation ( [ temp ] ) all relevant cooling processes ( such as collisional excitation and ionization , compton cooling , recombination and bremsstrahlung radiation , cooling by hi , h@xmath3 and hd line emission ) are accounted for . for the cooling functions of h@xmath3 and hd we adopted the expressions given by hollenbach & mckee ( 1979 ) , and flower et al .
( 2000 ) , respectively ; note , that the hollenbach & mckee h@xmath3 cooling function practically coincides with the one calculated by le bourlot , pineau des forets & flower ( 1999 ) .
the most recent calculations of the hd cooling function by lipovka , n~ nez - lpez , & avila - reese ( 2005 ) show that in the temperature range of interest ( @xmath41 k ) it coincides with that found by flower et al .
( 2000 ) . at higher temperatures , where lipovka et al .
( 2005 ) predict an order of magnitude enhanced cooling from hd , the abundance of hd molecules is too low to contribute significantly .
the other cooling functions connected with excitation and ionization of h and he are taken from cen ( 1992 ) . in general , radiation from a hot shocked gas can heat the gas , however , for the shock waves with the velocity @xmath42 km s@xmath43 ( shull & mckee 1979 , shull & silk 1979 ) its contribution to ionization and heating is unimportant , therefore in our calculations we assume @xmath44 .
the chemical reaction rates are taken from galli & palla ( 1998 ) .
the thermo - chemical equations ( [ spe ] ) , ( [ temp ] ) are solved in one collisional time @xmath45 ; @xmath45 is much shorter than the age of the universe at any redshift @xmath18 .
fig.1 shows an example of physical characteristics of the postshock gas versus temperature in the collision of two minihaloes with the total masses @xmath46 each at @xmath47 .
the fractional ionization @xmath48 and h@xmath3 abundance behind the shock are taken equal to their background values @xmath49 and @xmath50 , @xmath51 . immediately behind the shock concentration of
h@xmath3 increases very rapidly and reaches its maximum value @xmath52 promoting cooling of gas to @xmath53 k. if h@xmath3 cooling decreases the gas temperature below @xmath54 k the abundance of hd molecules grows fast and hd cooling becomes dominant .
it sets a runaway cooling on . at @xmath55
k almost all the deuterium is converted into molecular form , and in one collisional time gas cools down to @xmath56 k. haiman , rees , & loeb ( 1997 ) showed that the very first stars emit a large amount of photons in lyman - werner band ( lw ) , which produces a hostile uv background for hd molecules .
the characteristic time of photodissociation of hd can be estimated as @xmath57 yr , where @xmath58 is the lw flux in units of @xmath59 ergs s@xmath43 @xmath60 hz@xmath43 sr@xmath43 ( le petit , roueff , & le bourlot , 2002 , johnson & bromm , 2005 ) .
one can assume that the lw flux is by an order of magnitude equal to the ionizig flux from the first stars @xmath61 $ ] in the range @xmath62 ( ciardi et al . , 2000 ) .
on the other hand , the characteristic formation time of hd is @xmath63 yr at @xmath64 .
thus , practically in the whole range of redshifts considered here photodissociation from the very first stars can be neglected : @xmath65 becomes comparable to @xmath66 only at @xmath67 . at lower
@xmath18 self - shielding seems to come into play .
indeed , for the masses of merging haloes @xmath68 the characteristic column density of hd molecules in the shocked layer at final stages can be as high as @xmath69 @xmath60 , which is factor of 3 greater than the column density needed to provide optical depth in lw band of the order of one @xmath70 @xmath60 ( le petit et al . , 2002 ) .
haloes of small masses ( @xmath71 ) merge with slow relative velocities , and the post - shock temperature is so low that fractional ionization remains equal to its post - recombination value @xmath11 . as a result ,
@xmath72 diagrams connecting @xmath73 and @xmath74 along evolutionary paths very weakly depend on halo masses , as seen from fig .
mergings of more massive haloes produce shocks with temperature @xmath75 k , so that fractional ionization can be as high as @xmath76 . immediately behind the shock @xmath48 starts growing from its pre - shock value to the corresponding equilibrium .
however , as the characteristic collisional ionization time is longer than the cooling time , gradual increase of fractional ionization is accompanied by cooling , which in turn weakens the rate of collisional ionization . as a result ,
fractional ionization remains always lower that the equilibrium value determined by collisional ionization and radiative recombination . at temperature @xmath77 k @xmath48 reaches the maximum which depends on the postshock temperature , and then goes down as shown in fig .
2 . this behavior restricts the efficiency of molecule formation : as seen from fig . 1 h@xmath3 saturates at @xmath78 k when fractional ionization begins to decrease sharply .
hd instead monotonously grows until practically all deuterium is converted into molecular form .
the abundance of h@xmath3 molecules at the end of baryon compression increases with the mass of the merging haloes , and the minimum temperature reached at this stage becomes lower .
when it reaches @xmath6 k a rapid fractionation of hd molecules sets on . for larger halo masses
hd cooling dominates and cools baryons down to the temperature of the cmb radiation .
3 connects variations of temperature and molecular concentrations ( h@xmath3 and hd ) along the evolutionary paths for different masses of the haloes merging at @xmath47 .
their contribution to the total cooling is shown in fig .
4 where open circles show the ending stage after one collision time @xmath45 : while immediately after crossing the shock front the contribution from hd ( left curve ) is negligible , it increases in the course of gas cooling , and the transition to the regime with a predominance of hd cooling is clearly seen at @xmath79 .
the critical mass corresponding to the onset of a predominantly hd cooling is @xmath80^{-2}\msun$ ] as shown in fig .
5 . this value is factor of 5 higher than the minimum mass of the first objects found by tegmark et al .
one may conclude therefore , that at the lower end of masses the primeval gravitationally bound baryonic condensations stop cooling at relatively high temperature @xmath53 k due to h@xmath3 radiation , and hd molecules do not form in sufficient amount to cool the baryons further .
however , soon later , at @xmath81 more massive haloes can form and merge with a clear predominance of hd cooling , and in principle with lower masses of possible stars .
note , that the critical mass @xmath82 is fairly unsensitive to the choice of cooling function .
in particular , for the cooling function of galli & palla ( 1998 ) , who included the collisional rates from martin , schwarz & mandy ( 1996 ) and forrey et al .
( 1997 ) , which in the temperature range @xmath83 k is approximately factor of 3 higher than h@xmath3-cooling function of hollenbach & mckee in our calculations , @xmath82 increases only by 1015% in comparison with that shown in fig . 5
when cooled the shocked gas layer can fragment due to gravitational instability ( stone 1970 , elmegreen & elmegreen 1978 , gilden 1984 ) . for
a layer formed in a collision of two clouds the necessary conditions for instability imply that _
i ) _ the characteristic growth time is shorter than the collision time , and _
ii ) _ the critical wavelength is shorter than the initial size of the clouds ( gilden 1984 ) . the dotted line in fig .
5 depicts the critical mass @xmath84 of a merging halo gravitationally unstable in gilden sense : merging haloes with individual masses above this line are unstable againts fragmentation ; @xmath84 is approximated as @xmath85^{-2}\msun$ ] .
it is readily seen that @xmath84 is practically coincident with the mass corresponding to the transition to a predominance of hd cooling @xmath82 .
the explanation of this coincidence can be found in the fact that the shock compressed layers become unstable in gilden sense only when hd molecules cool the compressed gas below @xmath86 k ( vasiliev & shchekinov 2005 ) .
it is obvious therefore that all masses @xmath87 at the latest stages are dominated by hd cooling , and as a consequence the fragments compressed by mutual action of ram pressure and gravitation are able to cool further down to @xmath55 k. in the layer formed in a collision of two haloes with masses @xmath88 the corresponding jeans mass is @xmath89 , which is less than 1% of the baryonic mass involved into collision .
6 shows the dependence of jeans mass in unstable layers on the mass of merging haloes and the redshift @xmath18 when the haloes merge : thick dotted white line delimits the haloes whose compressed layers are gravitationally unstable all masses above this line are unstable .
the thin solid lines are the iso - jeans masses in compressed layers .
when temperature in the compressed layer goes to the cmb value the jeans mass becomes independent on redshift .
clarke & bromm ( 2003 ) have estimated a lower limit of the masses of fragments formed in similar conditions , with baryons confined by ram pressure from shocks and gravitation . however , for the primordial gas they assumed the lower temperature limit @xmath90 k determined by h@xmath3 cooling . as a result , the masses of the fragments are systematically higher compared to our estimates . in order to obtain smaller values of the masses they had to assume that the compressed gas has cooled down to the cmb temperature due to cooling provided by co molecules only after the gas is pre - enriched with metals .
the lower panel of fig .
6 shows the gas temperature at the latest stages of gas compression , i.e. at @xmath91 , for mergings of haloes with masses @xmath92 at redshift @xmath18 .
it is clearly seen that the compressed gas can cool down to the lowest temperature @xmath93 in haloes with masses @xmath94 as their cooling is dominated by hd molecules .
note , that this result does not contradict the conclusions drawn from numerical simulations by abel et al .
( 2000 , 2002 ) and bromm et al .
( 2002 ) , because these simulations have been done for halo masses close to the minimum mass expected for first objects .
fragmentation of a compressed gas layer can be initiated by thermal instability at the stages when h@xmath3 and hd begin to dominate radiative cooling . in fig . 7 ( upper panel ) temperature dependence of the effective ccoling function @xmath95 and its logarithmic derivative @xmath96 are shown in the temperature interval @xmath97 k , where hi radiative energy losses become unimportant and molecular cooling sets on .
the curves corresponding to different masses are plotted in temperature interval from the initial value immediately behind the shock to the lowest value at @xmath91 . in a non - steady cooling medium without external heating
the condition for the isobaric mode to be unstable reads as @xmath98 ( e.g. , shchekinov 1978 ) .
it is evident that this condition is fulfiled at @xmath99 k due to cooling of h@xmath3 molecules , and in the lower temperature end at @xmath100 k where hd cooling dominates .
the characteristic length of thermally unstable perturbations is @xmath101 , where _ r~32kt , is the characteristic radiative cooling time , @xmath102 is the sound speed .
the corresponding masses @xmath103 , estimated for the parameters of a shocked gas , are shown in the lower panel of fig . 7 ; here @xmath104 is the baryon density at the stage when thermal instability begins . for the haloes of large masses
the mass of fragments which can form through thermal instability is lower than the total mass of baryons and jeans mass practically in the whole temperature range of compressed gas layers .
this means that in these conditions thermal instability stimulates subsequent fragmentation of the compressed gas .
the minimum mass of fragments expected from thermal instability @xmath105 can be as small as @xmath106 for halo masses @xmath107 .
however , thermal instability does not work in the haloes of smaller masses when @xmath108 exceeds their baryon mass .
as it seen from fig .
7 this occurs at @xmath109 .
the clouds formed under thermal instability are pressure supported , but not necessarily gravitationally bound
it is seen from fig . 6
that thermal instability forms gravitationally bound clouds of stellar masses only in mergings of massive haloes @xmath110 .
therefore not always thermal instability can give rise directly to star formation , however collisions of clouds formed through it can be a stimulating factor .
the compressed layers gravitationally unstable against fragmentation , @xmath111 , lie in the halo mass range where hd cooling can keep gas isothermal with temperature close to @xmath112 .
therefore , once the fragments are formed , they can afterwards either collapse homogeneously or brake further on smaller masses as prescribed by the hierarchical fragmentation scenario ( hoyle 1953 ) .
the possibility of the fragmentation cascading down to lower scales still remains under debate ( see , e.g. , discussion in coppi et al .
2001 , glover 2005 ) .
an essential condition for such a cascading process is a pressure - free contraction regime . in practice
, gravitational collapse turns out to settle onto a slow quasi - hydrostatic regime , which results in an efficient subsonic damping of perturbations ( abel et al .
2002 , bromm et al .
2002 ) , and therefore in a strong suppress of further fragmentation .
the mass of dense clumps which can be reached in the hierarchical fragmentation scenario gives , however , an estimate of the lower mass limit of protostellar condensations expected at given physical conditions the opacity limit for fragmentation ( low & lynden - bell 1976 , rees 1976 ) .
this mass is determined by a maximum density when the forming fragments become optically thick in the lines providing dominant radiative cooling .
when hd molecules are the main cooling agent , a hierarchical fragmentation becomes optically thick in hd rotational lines , i.e the optical depth of a fragment at line centre becomes @xmath113 , at densities @xmath114 @xmath115 .
even though at these densities most of hydrogen converts to molecular form in three - body collisions ( palla et al . 1983 ) its contribution to the total cooling remains negligible because of low temperatures .
the corresponding jeans mass at this stage is @xmath116 ( vasiliev & shchekinov 2005 ) . for @xmath64
this gives a relatively low mass limit : @xmath117 .
a similar estimate for the mass of fragments formed behind the shock waves with velocities @xmath118 km s@xmath43 from supernovae explosions was obtained by uehara & inutsuka ( 2000 ) .
one should stress though , that the condition @xmath113 implies that photons from hd rotational lines experience in average only one scattering , which seems insufficient to lock radiation inside a contracting cloud and stop further decrease of jeans mass .
from this point of view , the condition @xmath119 equivalent to about 10 scatterings of photons ( ivanov , 1973 ) , gives apparently more accurate estimate of the opacity mass limit : the corresponding jeans mass is @xmath120 . for comparison , the estimates of the opacity mass limit of the fragments determined by h@xmath3 cooling vary from @xmath121 to about @xmath122 ( nakamura & umemura 1999 ) depending on the initial h@xmath3 abundance .
note , that the exact value of the opacity limit depends on dynamical regime of a collapsing cloud : the optical depth of a contracting cloud decreases with the velocity gradient as @xmath123 ( sobolev , 1960 ) . in numerical simulations @xmath124 at late stages of the contraction ( abel et al .
2002 ) , here @xmath125 is the jeans length .
abel et al .
( 2002 ) stop their similation when optical depth at line centre is 10 . with accounting the factor 2 in the velocity gradient
this corresponds to the effective optical depth of 5 nearly at the beginning of the stage when radiation becomes locked inside the contracting fragment .
in this paper we showed that 1 . in the hierarchical scenario of galaxy formation mergings of massive dark matter haloes , @xmath126^{-2}\msun$ ]
develop dense baryon layers , where practically all the deuterium becomes confined into hd molecules , which then provide a very efficient radiative cooling ; 2 .
only slightly higher is the critical mass of the merging haloes when the shocked baryon layers can be unstable against gravitational fragmentation @xmath127^{-2}\msun$ ] .
3 . in these conditions
the shocked baryons can cool down to the lowest temperature @xmath128 k ; 4 .
the fragments formed through instability evolve then practically isothermally with @xmath129 , and can in principle undergo further sequencial fragmentation , which stops when the minimum mass @xmath130 is reached .
this minimum mass is smaller than the one expected when h@xmath3 molecules control radiative cooling ( nakamura & umemura 2002 , ciardi & ferrara 2005 ) .
this means that the first stars can be less massive in galaxies with masses @xmath131^{-2}\msun$ ] , where hd cooling dominates .
|
hd molecules can be an important cooling agent of the primordial gas behind the shock waves originated through mergings of the dark matter haloes at epochs when first luminous objects were to form .
we study the necessary conditions for the hd cooling to switch on in the low temperature range @xmath0 k. we show that these conditions are fulfiled in merging haloes with the total ( dark matter and baryon ) mass in excess of @xmath1^{-2}\msun$ ] .
haloes with masses @xmath2 may be the sites of low - mass star formation .
[ firstpage ] early universe
cosmology : theory shock waves .
|
biological cells use a variety of molecular machines representing enzymes , ion channels or pumps , and motors .
motor proteins are nanometer - size devices generating forces and actively moving or rotating under the supply of chemical energy through atp hydrolysis .
they are crucial for many cell functions and promising for nanotechnology of the future .
although such motors represent single molecules , their operation cycles can not be in detail followed in simulations even on the best modern supercomputers and some approximations need to be employed .
it is often assumed that conformational dynamics of motor proteins is well described within a linear response approximation and corresponds to excitation of normal modes .
we have checked this assumption for two motor proteins , myosin v and kinesin kif1a .
our results show that , while both these biomolecules respond by well - defined motions to energetic excitations , these motions are essentially nonlinear .
the effect is particularly pronounced in kif1a where relaxation proceeds through a sequence of qualitatively different conformational changes , which may facilitate complex functional motions without additional control mechanisms .
protein machines , which may represent enzymes , ion pumps or molecular motors , play a fundamental role in biological cells and understanding of their activity is a major challenge .
operation of these machines is based on slow conformational motions powered by external energy supply , often with ligands ( such as atp ) . in molecular motors , binding of atp and its subsequent hydrolysis induce functional mechanochemical motions , essential for their operation .
these motions , which follow after an energetic activation , are conformational relaxation processes .
large - scale conformational changes may take place in proteins as a result of ligand binding @xcite . despite the large magnitude of such changes ,
they are nonetheless often considered in the framework of the linear response theory @xcite and the normal mode approximation @xcite .
the normal mode analysis is furthermore broadly employed in the elastic - network studies of proteins @xcite .
however , there is no general justification to assume that relaxation processes in proteins are linear and this assumption has to be verified for particular macromolecules .
it is known that relaxation processes in complex dynamical systems may be strongly nonlinear and deviate much from simple exponential relaxation . as an example borrowed from a distant field
, we can mention the belousov - zhabotinsky reaction which exhibits a great variety of spatiotemporal patterns ( pacemakers , rotating spiral waves ) that are however only complicated transients accompanying relaxation to the equilibrium state @xcite .
conformational relaxation in single protein molecules may also be a complicated process , comprising qualitatively different kinds of mechanochemical motions . while partial unfolding and refolding , associated with ligand binding , are known for some protein machines , such as the enzyme adenylate kinase @xcite , usually functional conformational motions in molecular machines and , specifically , in motor proteins are _ elastic_.
this means that the pattern of contacts between the residues in a protein is not changed upon ligand binding and preserved during the relaxation process , as generally assumed in the elastic network modeling ( enm ) . here
, we provide detailed analysis of conformational relaxation processes , associated with ligand binding and hydrolysis , in two motor proteins myosin v @xcite and kinesin kif1a @xcite .
our investigations , performed in the framework of the enm approximation , reveal that nonlinearity is characteristic for both macromolecules and the normal mode description is not really applicable for any of them . for kif1a , a monomeric motor protein from the kinesin superfamily ,
nonlinear effects are found to dominate completely functional mechanochemical motions which turn out to be qualitatively different from the normal mode predictions . despite the nonlinearity , well - defined conformational relaxation paths , robust against perturbations ,
have been found in both motor proteins .
within the coarse - grained enm approach , a protein is modeled as a network of point - like particles , corresponding to residues , which are connected by a set of elastic links @xcite .
a link between two particles is present if the distance between them in the equilibrium conformation of the considered macromolecule is shorter than a cutoff length .
the elastic energy of the network is @xmath0 , where @xmath1 is the stiffness constant of the network links , @xmath2 is the matrix of connections inside the network , @xmath3 is the distance between particles @xmath4 and @xmath5 , and @xmath6 is the respective distance in the equilibrium reference state .
the characteristic time scales of functional mechanochemical motions in motor proteins are in the millisecond range and slow conformational relaxation motions on such timescale should be overdamped @xcite .
neglecting hydrodynamic interactions , relaxation dynamics is then described by equations @xmath7 for the coordinates @xmath8 of the particles , where @xmath9 is their mobility .
relaxation dynamics for elastic networks of proteins has been previously considered @xcite . despite a wide - spread misunderstanding , elastic dynamics
is generally nonlinear .
for example , macroscopic objects , such as ribbons or membranes , can still exhibit pronounced nonlinear effects of spontaneous twisting or buckling , while fully retaining their elastic behavior and not undergoing plastic deformations @xcite .
the energy @xmath10 of an elastic network is quadratic in terms of the distances @xmath3 and the forces acting on the particles are linear in terms of such distances .
however , the distance @xmath11 is itself a _ nonlinear function _ of the coordinates @xmath8 and @xmath12 and this makes the forces also nonlinear functions of dynamical variables .
the presence of nonlinear effects in conformational relaxation of proteins in the enm approximation has been previously demonstrated @xcite .
explicitly , relaxation dynamics of considered proteins is described by equations ( [ relaxation ] ) in the methods section , where further details are also given . to study conformational relaxation , these equations were numerically integrated starting from various initial conditions .
the reference conformation , used to construct the elastic network , was that of the atp(analog ) bound state ( protein data bank ( pdb ) i d code : 1w7j , with mgadp - befx as the atp analog @xcite ) .
as the initial condition , the conformation corresponding to the nucleotide - free state was taken ( pdb i d : 1oe9 @xcite ) .
the elastic network had 855 particles connected by 7261 links .
note that only the residues whose @xmath13-carbon positions are contained in both pdb data sets have been taken to construct the network .
additionally , relaxation processes starting from randomly generated initial conditions ( see methods ) have been considered . for visualization purposes , motions of three particles ( asp122 in chain a , and val22 and ser135 in chain b ) have been traced ( figure [ fig - myo5]a ) .
thus , each relaxation process was characterized by a certain trajectory in the space of distances between the three chosen labels .
figures [ fig - myo5]b,[fig - myo5]c display 100 conformational relaxation trajectories , each starting from a different random initial condition .
although the initial conditions were generated by applying relatively strong deformations ( without unfolding ) to the reference state , almost all of them were leading back to that reference state , with just a few metastable states found .
furthermore , one can observe that the trajectories converge to a well - defined relaxation path . the red trajectory in figures [ fig - myo5]b,[fig - myo5]c is for the relaxation starting from the nucleotide - free conformational state of myosin v ( so that the mechanochemical motion following atp binding is simulated ) .
after a transient , this special trajectory joins the well - defined relaxation path .
this functional trajectory is robust against perturbations , as shown by figure [ fig - myo5]d .
several snapshots of the conformation along this trajectory are shown in figure [ fig - snapshots - myo5 ] ( see also video s1 ) .
the attractive path corresponds to a deep energy valley in the energy landscape of myosin v. once this valley is entered , the conformational relaxation motion becomes effectively one - dimensional and characterized by a single mechanical coordinate .
the profile of the elastic energy along the bottom of such energy valley determines the dependence of the elastic energy on the collective mechanical coordinate ( see methods ) .
figure [ fig - pot]a shows the dependence of the elastic energy along the special attractive relaxation path starting from the nucleotide - free state and leading to the atp - bound state .
markers indicate positions along the trajectory in figure [ fig - myo5]c . for @xmath14 ,
the elastic energy @xmath10 is approximately quadratic in terms of the mechanical coordinate @xmath15 , i.e. @xmath16 . because @xmath17 , this implies that then @xmath18 and the relaxation is exponential . only within such harmonical neighborhood of the reference state , the normal mode description becomes applicable ( see methods for further discussion ) . the dotted blue line in figure [ fig - myo5]d shows the direction of the distance changes corresponding to the slowest normal mode ( see methods ) .
the nucleotide - free state of myosin v lies away from this direction and also outside of the harmonical neighborhood .
the initial stage of the functional mechanochemical motion ( until time @xmath19 ) can not be quantitatively analyzed in terms of the normal modes . in the normal mode description ( eqn .
( [ eigen ] ) ) , normalized to the lowest non - zero eigenvalue @xmath20 , are shown .
there is a significant gap between the lowest and the second lowest modes .
, width=313 ] the reference conformation for kif1a is the adp - bound state ( pdb i d : 1i5s , with mgadp @xcite ) .
relaxation starting from the initial condition , corresponding to the atp(analog)-state ( pdb i d : 1i6i , with mgamppcp as an atp analog @xcite ) and from randomly generated initial conditions was considered .
the elastic network has 320 particles and 2871 links .
only the residues whose @xmath13-carbons are in both pdb data sets have been used .
visualization labels are glu233 , ala286 , and asn211 ( figure [ fig - kif1a]a ) .
100 relaxation trajectories , starting from random initial conditions , are shown in figures [ fig - kif1a]b,[fig - kif1a]c .
the presence of an attractive relaxation path , corresponding to a deep energy valley , can be noticed .
the red lines in figures [ fig - kif1a]b,[fig - kif1a]c display the special relaxation trajectory starting from the atp - bound state .
surprisingly , we find that , in contrast to myosin v , this trajectory is _ different _ from the typical relaxation path . by applying small random initial perturbations to the initial atp - bound state and integrating the dynamical equations , it can be demonstrated that this trajectory is , however , also stable with respect to the perturbations ( figure [ fig - kif1a]d ) . the dotted blue line in figure
[ fig - kif1a]d shows the direction of the distance changes in the slowest normal mode of kif1a .
thus , in kif1a the deep energy valley leading to the reference adp - bound state gets branched at some distance from it .
the path corresponding to the functional mechanochemical motion from the atp - bound state belongs to the side branch . only at the final relaxation stage , in the immediate vicinity of the equilibrium ,
the valleys merge and the functional motion begins to coincide with the typical relaxation motion in this protein .
the branching of the energy valley is already an indication of strong nonlinearity in the relaxation dynamics .
we have also determined the profile of the elastic energy @xmath10 as a function of the mechanical coordinate @xmath15 along the path connecting the atp- and adp - bound states ( figure [ fig - pot]b ) .
the profile becomes quadratic only starting from time @xmath21 , very close to the equilibrium reference state .
figure [ fig - snapshots - kif1a ] shows snapshots of kif1a along the special attractive relaxation path ( see also video s2 ) . at the early relaxation stage ( until @xmath22 ) , the relaxation motion represents a combination of the rotation of the switch ii helix and of the sliding of the switch i loop .
relaxation at the end of such initial stage is apparently hindered , as revealed in the presence of a plateau in the dependence of the elastic energy on the mechanical coordinate in figure [ fig - pot]b near @xmath23 .
only once the sliding is completed , further local structural reorganization , representing a transition from the loop to the @xmath13-helix , becomes possible and is indeed observed approximately after time @xmath19 .
the normal mode description is broadly used in structural studies of proteins .
the analysis of thermal fluctuations and the interpretation of the respective experimental structural data are traditionally performed assuming that fluctuations are linear and , hence , correspond to thermal excitation of various normal modes ( see , e.g. , @xcite ) .
the linear response of a protein macromolecule to structural perturbations , such as ligand binding , is an often used assumption @xcite . to a large extent ,
the elastic - network analysis of ligand - induced macromolecular motions is based on determining normal modes in the elastic networks of the considered proteins ( see , e.g. , @xcite ) .
the patterns of atomic displacements in such normal modes are further compared with the experimentally measured atomic displacements in the same proteins that are induced by a change of the chemical state , such as binding of an atp molecule @xcite .
large overlaps with only a few slowest normal modes are seen as the evidence for the applicability of the elastic - network ansatz , whereas the wide distribution of overlaps is considered as the indication that the elastic network description fails for a particular macromolecule .
specifically , strong overlaps between ligand - induced conformational changes and atomic displacements in the few slowest modes have been found for scallop myosin and @xmath24-atpase , while such overlaps were absent for kinesin kif1a @xcite .
our numerical investigations of elastic conformational motions in two motor proteins ( myosin v and kif1a ) have revealed however that in both of them the nonlinearities play an essential role .
while slow conformational relaxation motions in myosin can still be qualitatively characterized in terms of the normal modes , the normal mode description breaks down _ completely _ for kif1a .
the observed breakdown of the normal mode description does not however mean that conformational motions become irregular .
we have seen that ordered and robust mechanochemical motions are characteristic for both protein motors , even though they can not be described in terms of the linear response .
we want to emphasize that , when the dynamics is nonlinear , neither a single normal mode , nor a combination of many such modes can reproduce the motions . thus , the normal mode description fails completely in this case and the problem is not that many normal modes must be taken into account . actually , as we have shown , even for kif1a , one normal mode would be sufficient to describe long - time relaxation within the harmonic domain however , this domain is restricted to a tiny neighbourhood of the equilibrium state .
thermal fluctuations have not been explicitly included into our dynamical enm simulations . however , such fluctuations are effectively generating random conformational perturbations . in our study ,
relaxation processes starting from random conformational perturbations have indeed been considered . in myosin
v , one well - defined nonlinear conformational relaxation trajectory , leading to the equilibrium state , has been identified .
starting from an arbitrary initial conformation ( but still without unfolding ) , rapid convergence to this special trajectory takes place .
while the motion corresponding to the special attractive trajectory is initially nonlinear , it becomes harmonical later and a substantial part of the ordered conformational relaxation process is within the harmonic domain of the equilibrium state .
similar behavior has been previously noted by us @xcite for scallop myosin and @xmath24-atpase , but its detailed analysis has not yet been performed .
the situation is more complex for the monomeric kinesin kif1a . instead of a unique deep energy valley leading to the reference adp - bound state , two such valleys ,
both leading to the equilibrium state , are present .
these valleys correspond to two kinds of ordered conformational motions possible in the protein .
the first of them is relatively wide and , when thermal conformational fluctuations are excited , they would typically proceed along it .
however , the conformational relaxation motion starting from the atp - bound state follows a different path , which corresponds to the second energy valley branching from the typical fluctuation path already at very small deviations from the equilibrium state .
note that the branching takes place as the change in the distance between the molecular labels glu233 and ala286 is still less than an angstrom , which is much smaller than the intensity of typical thermal fluctuations for such a distance .
thus , the nonlinear effects in kif1a are strong even for the typical thermal fluctuations .
remarkably , such second relaxation path is also stable with respect to perturbations , i.e. structurally robust .
our numerical investigations reveal that motion along this path can be divided into two _ qualitatively different _ stages . at the first of them , sliding of the switch i loop
is observed , whereas at the second stage a transition from the loop to the @xmath13-helix is realized .
structural reorganization , corresponding to this transition , is not possible until the sliding motion is completed , lifting a restriction through the backbone chain .
recent crystallographic studies suggest that the switch i loop / helix plays an important role in control of the motor function through interaction with @xmath25 and switch ii @xcite . thus , in contrast to myosin , a single atp binding event induces in kif1a a complex , but ordered conformational motion characterized by two qualitatively different consequent phases .
as we conjecture , this special dynamical property of kif1a may be needed for the processive motion of this single - headed molecular motor @xcite . in myosin
v , conformational motions driven by random thermal fluctuations are similar in their properties to the relaxation motion from the nucleotide - free state .
this may facilitate exploitation of such fluctuation motions for the motor operation , as suggested by recent single - molecule experiments @xcite . in kif1a , where the energy valley splits into two branches , typical thermal conformational fluctuations are qualitatively different from the relaxation motion starting from the previous atp - bound state .
the latter motion is entropically hindered for thermal fluctuations and can not be reversed through them .
this may turn out to be important for the understanding of the operation of the monomeric kinesin as a molecular motor .
latest experimental techniques permit simultaneous observation of stepping motion and conformational changes of a motor @xcite .
the coarse - grained modeling , including our present study , can contribute further suggestions for the design ( e.g. , by determining positions for fluorescent labeling ) of such experiments .
finally , we note that our study has been based on the elastic network approximation for proteins .
more detailed descriptions , such as , e.g. , g - like models , can also be used to consider conformational relaxation processes @xcite .
we expect that similar behavior will then be observed .
in this study , we employ elastic network models where material points are connected by a set of elastic springs @xcite .
each particle corresponds to a residue in the considered protein .
the equilibrium positions @xmath26 of the particles are determined by the locations of @xmath13-carbon atoms in the reference state of the protein , taken from the pdb database .
two particles in a network are connected by an elastic spring if at equilibrium the distance @xmath27 between them is less than a certain cutoff length @xmath28 .
the natural length of an elastic link is equal to the equilibrium distance @xmath6 .
the cutoff distance @xmath29 has been used in our study .
the elastic forces obey the hooke law and all springs have the same stiffness constant @xmath1 .
elastic torsion effects are not included .
thus , the force acting on particle @xmath4 is @xmath30 where @xmath31 is the total number of particles in the network , @xmath32 is the actual position of the particle @xmath4 and @xmath11 is the actual distance between two particles @xmath4 and @xmath5 .
the adjacency matrix of the network is defined as having @xmath33 if @xmath34 and @xmath35 otherwise .
the total elastic energy of the network is @xmath36 because slow conformational dynamics of proteins in the solvent is considered , the motions are overdamped ( see @xcite ) and the velocity of a particle is proportional to the force acting on it , i.e. @xmath37 where @xmath9 is the mobility .
we assume that the mobilities of all particles are the same .
hydrodynamical effects are neglected ( they can be however incorporated into the elastic network models as shown in ref .
@xcite ) .
explicitly , the relaxation dynamics is described by a set of differential equations : @xmath38 here , time is rescaled and measured in units of @xmath39 .
hence , the relaxation dynamics of a network is completely determined by its pattern of connections ( matrix @xmath2 ) and the equilibrium distances @xmath6 between the particles . equations ( [ relaxation ] ) were integrated to determine conformational relaxation motions . to prepare random initial conditions ,
the following procedure has been employed .
random static forces @xmath40 , acting on all particles in the network have been independently generated with the constraint that @xmath41 .
the equations of motion were integrated in the presence of such static forces for a fixed time @xmath42 .
the conformation which was thus reached has been then used as the initial condition for the relaxation simulation .
the parameters were @xmath43 , @xmath44 for myosin v and @xmath45 , @xmath46 for kif1a . with these parameter values , relatively large overall deformations ( @xmath47 typical )
could be reached , while still avoiding unfolding . in the deformed states
, the lengths of the links did not exceed @xmath48 for myosin v and @xmath49 for kif1a .
when relaxation from specific conformations has been considered , initial positions of all particles were allocated according to the respective pdb structures .
when robustness of a relaxation path starting from a specific conformation was investigated , the initial condition was prepared by randomly shifting the positions of all particles with respect to their locations in that conformation with a certain root - mean - square displacement @xmath50 .
we have chosen @xmath51 for myosin v and @xmath52 for kif1a . to visualize conformational motions , three particles labeled as @xmath53 , @xmath54 and @xmath55 were chosen and the distances @xmath56 and @xmath57 were monitored in the simulations .
thus , the relaxation motion was represented by a trajectory on the plane @xmath58 .
the choice of the visualization labels is essentially arbitrary . in a simulation ,
motions of all residues were traced ( see , e.g. , videos s1 and s2 ) and different residues could be selected for a specific visualization . if a molecule has a low - dimensional attractive relaxation manifold , this is a property of the respective dynamical system and it can not depend on the visualization method . when selecting the labels , one should only pay attention to the fact that the distances between them should significantly vary during the relaxation process . if , by chance , two labels belonging to the same stiff domain in a protein have been taken , the distance between them would remain almost constant , so that such a choice would be inconvenient .
when the normal mode description approximately holds and , furthermore , relaxation is well described by a few slowest modes , one can choose the labels so that the distances between them reveal variations characteristic for the first few normal modes .
such selection was previously made @xcite for scallop myosin and @xmath24-atpase , and it has been adopted in the present study for myosin v. for kif1a , the labels have been chosen in such a way that motions in switch i and switch ii regions are well resolved .
the collective mechanical coordinate @xmath15 along a relaxation path was defined by requiring that its dynamics obeys the equation @xmath59 and that @xmath60 as @xmath61 .
multiplying both parts of this equation by @xmath62 , we find that it is equivalent to the equation @xmath63 equation ( [ coordinate ] ) can be used to determine the coordinate @xmath15 along a given relaxation trajectory and the dependence of the elastic energy @xmath10 on this coordinate . for each point along the trajectory
, the time @xmath64 when it is reached in the relaxation process is known .
moreover , the actual network configuration corresponding to this point is also known from the simulation .
therefore , for each point specified by time @xmath64 the respective elastic energy @xmath65 is determined .
the mechanical coordinate @xmath15 , reached at time @xmath64 , is given by the integral @xmath66 we provide a summary of the results on the normal mode description of conformational relaxation processes .
if deviations @xmath67 from the reference conformation are small for all particles , the nonlinear equations ( [ relaxation ] ) , describing conformational relaxation of an elastic network , can be linearized : @xmath68 \mathbf{u}_{ij}^{(0 ) } , \label{linear}\ ] ] where @xmath69 .
equations ( [ linear ] ) can be written in the matrix form as @xmath70 where @xmath71 is the @xmath72 linearization matrix : @xmath73 where @xmath74 .
the general solution of these linear differential equations is given by a superposition of @xmath75 exponentially decaying normal modes , i.e. @xmath76 here , @xmath77 and @xmath78 are the eigenvalues and the eigenvectors of the linearization matrix , i.e. @xmath79 this matrix has @xmath80 eigenvalues , but 6 of them must be zero , corresponding to free translations and rotations of the entire network . generally , all normal modes are initially present .
as time goes on , first the normal modes with the larger eigenvalues @xmath77 decay . in the long time
limit , relaxation is characterized by the soft modes corresponding to the lowest eigenvalues .
figure [ fig - eigen ] shows the computed eigenvalue spectra of myosin v and kif1a .
the eigenvalues are normalized to the lowest nonzero eigenvalue @xmath20 and the logarithmic representation is chosen .
note that in both motor proteins a significant gap , separating the soft mode from the rest of the spectrum , is present .
this means that , in the linear approximation , long - time relaxation in these proteins is effectively characterized by a single degree of freedom , representing the amplitude of the soft mode .
the pattern of displacements of particles ( i.e. , residues ) from the reference positions is determined by the eigenvector @xmath81 of the soft mode . in the plane @xmath58 of the distances between the labels @xmath53 , @xmath54 and @xmath55 , used by us for
the visualization of conformational motions , the exponential relaxation motion corresponding to the soft mode should proceed along the direction defined by the vector with the components @xmath82 and @xmath83 .
such directions are indicated by dotted blue lines in figure [ fig - myo5]d and figure [ fig - kif1a]d . when relaxation is reduced to a single soft mode
, the elastic potential is quadratic in terms of the mechanical coordinate , i.e. @xmath84 . note that the representation of the relaxation process as a superposition ( [ nmdecay ] ) of normal relaxation modes holds only in the harmonic domain , i.e. when linearization ( [ linear ] ) of full nonlinear relaxation dynamics equations ( [ relaxation ] ) is valid .
if dynamics is nonlinear and the linearization does not hold , relaxation dynamics can not be viewed _ at all _ as a superposition of any normal modes . whether just one normal mode or many of them should be included into a description of long - time relaxation dynamics is determined by the properties of the eigenvalue spectrum and not related to the possible invalidity of the harmonic approximation . as an extension , iterative normal mode analysis has been proposed @xcite .
this method is applied to obtain an optimal sequence of conformational states , transforming an initial given conformation into a target conformation , which may be known with a low resolution or only partially , and thus to reconstruct missing details of that structure .
each next conformation in the sequence is obtained by making a step into the direction maximizing similarity with the target , restricted however to a superposition of a certain number of the lower normal modes . at the next iteration step ,
the previous conformation is chosen as a new reference state and a new set of normal modes is determined .
this prediction method is useful and provides valuable results , e.g. , in the refinement of low - resolution structures from electron microscopy @xcite .
it should be however emphasized that the sequence of conformational states yielded by such a method is generally different from the path along which conformational relaxation from the target to the reference state would proceed . even in the normal mode approximation ,
dynamics of conformational relaxation depends not only on the eigenvectors , but also and very significantly on the eigenvalues of normal modes .
generally , the next iteration state in this method would not be the next conformation along the actual relaxation path .
this difference can be clearly demonstrated by considering the example of kif1a .
the conformational relaxation path transforming the initial atp - bound state into the ( equilibrium ) adp - bound state is non - monotonous ( figure [ fig - kif1a ] ) .
it proceeds via intermediate states ( particularly of the switch i region ) which can not be obtained by gradual interpolation maximizing similarity of the structures along the optimization path .
we thank m. ueda , h. takagi and t. komori for helpful comments .
j - l , beratan dn ( 2004 ) how does protein architecture facilitate the transduction of atp chemical - bond energy into mechanical work ? the cases of nitrogenase and atp binding - cassette proteins .
biophys j 87 : 13691377 .
kitao a , hirata f , g n ( 1991 ) the effects of solvent on the conformation and the collective motions of protein : normal mode analysis and molecular dynamics simulations of melittin in water and in vacuum .
chem phys 158 : 447472 .
tama f , miyashita o , brooks iii cl ( 2004 ) flexible multi - scale fitting of atomic structures into low - resolution electron density maps with elastic network normal mode analysis .
j mol biol 337 : 985999 .
|
the assumption of linear response of protein molecules to thermal noise or structural perturbations , such as ligand binding or detachment , is broadly used in the studies of protein dynamics .
conformational motions in proteins are traditionally analyzed in terms of normal modes and experimental data on thermal fluctuations in such macromolecules is also usually interpreted in terms of the excitation of normal modes .
we have chosen two important protein motors myosin v and kinesin kif1a and performed numerical investigations of their conformational relaxation properties within the coarse - grained elastic network approximation .
we have found that the linearity assumption is deficient for ligand - induced conformational motions and can even be violated for characteristic thermal fluctuations . the deficiency is particularly pronounced in kif1a where the normal mode description fails completely in describing functional mechanochemical motions .
these results indicate that important assumptions of the theory of protein dynamics may need to be reconsidered . neither a single normal mode , nor a superposition of such modes yield an approximation of strongly nonlinear dynamics .
|
inflation , a stage of accelerated expansion in the very early universe , is now widely accepted as part of the standard evolutionary scenario of the universe . on the other hand , many models of inflation have been proposed but we are still far from being able to narrow down the possible models sufficiently . among those many models of inflation ,
much attention has been paid recently to the ones based on string theory @xcite , which is considered to be a promising candidate for the ultimate unified theory . in particular , it is of great interest if the string theory landscape @xcite , in which there are many local minima , or false vacua , and the universe jumps from one minimum to another by quantum tunneling , can be observationally tested @xcite .
quantum tunneling of a scalar field with gravity is usually treated with the coleman - de luccia ( cdl ) instanton method @xcite , where the evolution of a scalar field is described with an o(4)-symmetric instanton , which is a solution of the euclidean equations of motion . motivated by the string landscape , inflation models with multi - scalar fields and/or with tunneling
are now keenly studied .
extension of the cdl instanton method to a multi - scalar field system is already discussed in @xcite . in those studies , however , many inflation models have been proposed , and now it is important to distinguish those models by observation .
observations of the power spectrum of cmb temperature anisotropies have convinced us of the existence of an inflationary stage in the very early universe . on top of that , if a non - gaussian feature such as skewness or bispectrum is detected in the cmb anisotropy , it will have a strong impact on the physics of the early universe @xcite .
in particular , since any single - field slow - roll inflation model with canonical kinetic term predicts almost gaussian fluctuations @xcite , any non - zero non - gaussianity will exclude all these models .
observations of non - gaussianity use templates , such as local type @xcite , equilateral type @xcite , and orthogonal type @xcite , in order to increase the statistical significance
. however , since all of these templates assume statistical isotropy , there may be anisotropic non - gaussianities which may not have been detected by these templates . in this letter
, we study a multi - field model in which a nonlinear interaction between two scalar fields , one of which being responsible for the curvature perturbation of the universe ( that is , for the formation of the large scale structure ) and the other for quantum tunneling via a cdl instanton , induces an anisotropic non - gaussianity . to be specific , we introduce an inflaton field @xmath2 that realizes slow - roll inflation , a tunneling field @xmath1 that governs the tunneling dynamics , and a curvaton field @xmath0 that contributes to the curvature perturbation of the universe @xcite .
the inflaton dominates the energy density of the universe during inflation but rapidly decays to radiation after the end of inflation , on the other hand , the energy density of the curvaton field is negligible during inflation but its decay is delayed after inflation so that it gradually begins to dominate the universe .
approximating the universe during inflation by an exact de sitter spacetime , the inflaton behaves as a cosmic clock and determines an appropriate time - slicing , namely , a spatially flat time - slicing of the de sitter spacetime . in this setup , assuming that the energy scale associated with the tunneling field is much smaller than the energy scale of inflation , the bubble nucleation can be well described by a single - field cdl instanton with no backreaction to the geometry , that is , on the exact de sitter background @xcite .
the curvaton @xmath0 is affected by the background bubble through a coupling with the tunneling field @xmath1 . for simplicity and definiteness
, we consider a potential of the form , @xmath3 . assuming that @xmath4 is non - vanishing only at or inside the bubble wall
, we expect that @xmath0 may have a spatially localized , bubble - shaped non - gaussianity due to the background bubble - shaped configuration of @xmath1 .
this leads to an anisotropic , bubble - shaped skewness of the cmb temperature anisotropy .
this paper is organized as follows .
we first illustrate the background spacetime and the configuration of the bubble .
next , we briefly review a useful formalism for computing the equal - time @xmath5-point functions , the tunneling in - in formalism , and calculate the skewness of the curvaton fluctuations .
then we demonstrate that a sky map of an anisotropic non - gaussian parameter @xmath6 in our model .
finally , we end with a few concluding remarks .
let us start from a brief description of de sitter spacetime @xcite .
a lorentzian 4-dimensional de sitter spacetime is o(4,1)-symmetric , which is obtained by analytical continuation of an o(5)-symmetric 4-dimensional euclidean sphere , as illustrated in fig .
[ fig : penrose_ds ] . among various coordinatization of the spacetime ,
the uniform tunneling field slicing is most appropriate to see o(4)(o(3,1))-symmetry of the cdl instanton , which corresponds to the time - slicing inside the bubble that describes a spatially homogeneous and isotropic open universe . for brevity ,
let us call it open slicing .
open slicing in the c - region in fig .
[ fig : penrose_ds ] is a time - like slicing , an the metric in the c - region may be expressed as @xmath7 where @xmath8 and @xmath9 , and @xmath10 is the metric on the unit 2-sphere .
theses coordinates can be extended to the other regions of the spacetime by analytical continuation : @xmath11 where @xmath12 , @xmath13 and @xmath14 are the coordinates for the r- , l- , and e - regions , respectively , in fig .
[ fig : penrose_ds ] .
the metric for the flat time - slicing is @xmath15 as mentioned in the introduction , the @xmath16 slices are on which the inflaton is uniform , and give the cosmic rest frame of our universe .
let us briefly describe the bubble configuration of @xmath1 under the thin - wall approximation .
the o(4)-symmetric cdl instanton for @xmath1 can be described as a function of @xmath17 , which we denote by @xmath18 hereafter .
we denote the wall radius by @xmath19 .
hence @xmath20 though @xmath21 is homogeneous in the r- or l - regions , the physical bubble radius @xmath22 on flat slices increases as time goes on .
we take the origin @xmath23 of the flat slicing metric ( [ flatslicing ] ) to be the center of the bubble at the time of nucleation .
then the relations between the coordinates in the metrics ( [ openslicing ] ) and ( [ flatslicing ] ) are @xmath24 and @xmath25 , which gives @xmath26 .
as seen from this expression , a bubble once nucleated with radius @xmath27 ( @xmath28 for @xmath29 ) expands to the hubble horizon scale within one or two @xmath30-folds of time , @xmath31 , and then expands comovingly as @xmath32 .
it should be noted that models with multiple nucleation is possible .
if those bubbles do not interact each other , we can take into account the effect of all bubbles by summing up the effect of each bubble .
hereafter , we consider a model with a single bubble for simplicity .
here we calculate the skewness in the curvaton fluctuations on the single bubble background .
we consider the lagrangian for the curvaton @xmath0 as @xmath33 where @xmath34 .
we expand the above by setting @xmath35 , where @xmath36 is a homogeneous classical part and @xmath37 is a quantum fluctuation .
we assume that @xmath36 is approximately constant during inflation , and we concentrate on the evolution of @xmath37 .
further , for simplicity , we assume @xmath38 is non - vanishing only on the wall . hence we approximate it as @xmath39 with @xmath40 .
then the lagrangian for @xmath37 is given as @xmath41 , where @xmath42 is free - part and @xmath43 is interaction - part , @xmath44 where @xmath45 .
to calculate the quantum fluctuations on the instanton background , we extend the in - in formalism to euclidean spacetime , which we call the tunneling in - in formalism .
this formalism is based on the wkb analysis of a tunneling wave function for free theory @xcite to the case with nonlinear interactions .
a detailed derivation will be given elsewhere @xcite .
the tunneling in - in formalism tells us that the @xmath5-point function of @xmath37 is given by @xmath46 where the tunneling in - in path @xmath47 and @xmath16 surfaces @xmath48 are as shown in fig .
[ fig : path_pen ] . the first half of @xmath47 ( the arrowed green line in fig .
[ fig : path_pen ] ) goes from one end of the euclidean region to future infinity in the lorentzian region through the bubble nucleation surface . the second half of @xmath47 ( the arrowed blue line in fig .
[ fig : path_pen ] ) goes back from future infinity through the nucleation surface to the other end of the euclidean region .
the slicing @xmath49 covers the whole euclidean region and the future half of the lorentzian region twice . in the lorentzian region
any @xmath50 is a cauchy surface .
the operator @xmath51 in eq .
is the path - ordering operator . in the lorentzian region
, @xmath51 reduces to the time - ordering and anti - time - ordering operators @xmath52 and @xmath53 on the first and second halves of @xmath47 , respectively .
it should be noted that eq .
is independent of the choice of coordinates .
eq . is evaluated in the same way as in the usual perturbation theory . after expanding the interaction part perturbatively ,
operators are transformed to products of the free correlation function @xmath54 using wick s theorem .
we note that @xmath55 is not the correlation function for the bunch - davies vacuum due to the non - trivial nature of @xmath42 .
it may be obtained by studying the `` evolution '' of the mode functions in the euclidean space @xcite . while @xmath55 is a single - valued function when @xmath56 and @xmath57 are in space - like separation , when they are in time - like separation , @xmath52 or @xmath53 in the expression singles out the feynman or anti - feynman propagator , respectively .
a branch of @xmath55 , when @xmath57 is in the euclidean region , is determined by analyticity on @xmath57 along @xmath47 .
it may be noted that by this way of choosing branches the expression in eq .
is equivalent to that obtained by analytical continuation of euclidean quantum field theory @xcite , and o(4)(o(3,1 ) ) symmetry of the result is guaranteed . , but with all domains of integration given in eq . .
, width=188 ] now we evaluate the skewness @xmath58 by substituting @xmath59 in eq .
to the tunneling in - in formula in eq . .
in the following calculations , we approximate @xmath55 by that of the bunch - davies vacuum , by neglecting corrections due to the non - trivial part of @xmath42 in eq . . this effect may affect the value of the skewness , but since it only induces a statistically homogeneous non - gaussianity , we simply ignore it here .
the free correlation function @xmath55 of the bunch - davies vacuum is given by @xcite @xmath60 , \label{eq:31}\end{aligned}\ ] ] where @xmath61 and @xmath62 , where @xmath63 is the geodesic distance between @xmath56 and @xmath57 .
we have @xmath64 , @xmath65 and @xmath66 , for a timelike ( imaginary @xmath67 ) , null ( @xmath68 ) , and spacelike ( real @xmath67 ) separation , respectively . as mentioned before , a branch of @xmath69 should be specified when @xmath64 and it can be done by adding small imaginary part @xmath70 to @xmath71 as @xmath72 .
we assume @xmath73 to avoid possible strong coupling problems . to leading order in @xmath74
, we obtain @xmath75 where @xmath76 as shown in fig .
[ fig : des_inin ] . among the domains of integral ,
@xmath77 and @xmath78 correspond to the e- and c - regions , respectively .
integration in the r- or l - regions is unnecessary because @xmath43 vanishes there .
a small imaginary part in @xmath79 gives a small imaginary part to @xmath80 to select the feynman or anti - feynman propagator .
evaluation of eq .
is straightforward in principle , but it is significantly simplified if we use the o(4)-symmetry of the @xmath5-point functions in the tunneling in - in formalism .
just for illustration , let us consider the case when @xmath56 is in the r - region . in this case
, computing @xmath81 at @xmath82 , where integration over @xmath17 and @xmath83 is trivial , is enough since the value at any other point in the r - region can be known from the o(4)-symmetry . after integrating over @xmath17 and @xmath83 , the integration over @xmath84 along @xmath47 remains .
since the original path @xmath47 passes near the singularity , it makes the integral difficult to evaluate numerically . to avoid it
, we deform the path @xmath47 to another path @xmath85 without crossing the branch cut or the poles of the integrand as in fig .
[ fig : des_inin ] .
the resulting @xmath81 is obtained in terms of the open - slice coordinates , but the expression in terms of the flat - slice coordinates can be easily found by the coordinate transformation . , while the solid line is a detoured path @xmath85 used for numerical evaluation . ,
width=264 ] fig .
[ fig : phi3_tf0 ] shows the skewness on flat slices at time @xmath86 as a function of the radial coordinate @xmath87 , @xmath88 , for @xmath89 and @xmath90 for several values of @xmath91 . to see the dependence of the skewness on the model parameters , we introduce a normalized skewness @xmath92 at the center of the bubble by @xmath93 the result
is shown in fig .
[ fig : f_mh ] . from this , we see that @xmath94 .
we see that the skewness is large inside the bubble and it decreases as one moves away from the bubble , and its radial dependence is stronger for larger @xmath91 .
these features can be understood from the fact that @xmath55 is roughly given by @xmath95 for @xmath96 and @xmath71 is not very close to 1 .
the apparent divergent behavior at the bubble wall comes from the uv divergence in @xmath55 .
but , this divergence disappears as soon as the finite resolution of observations or renormalization of the theory is taken into account .
normalized by the value at the center as a function of @xmath87 at @xmath89.,width=264 ] .
the solid lines are the result of numerical evaluation and the dashed lines are analytical approximation , @xmath97 . , width=264 ]
now we study signatures of our model in the cmb anisotropy . in our model ,
the evolution of the universe after the inflaton and the tunneling field decay , is exactly the same as the ordinary curvaton scenario @xcite . here
, we estimate the curvature perturbation after the curvaton decay by using the sudden - decay approximation @xcite .
after inflation the curvaton starts to roll down the potential and undergoes damped oscillations when the hubble parameter becomes smaller than @xmath98 .
the curvaton energy density behaves like a pressureless matter during this stage of damped oscillations . at the time @xmath99
when the curvatons decay , the energy density of the universe consists of that of radiation @xmath100 generated from the decay of the inflatons and that of the curvaton @xmath101 .
the contribution of each component to the total curvature perturbation of the universe may be conveniently expressed in terms of @xmath102 and @xmath103 , which are defined as the curvature perturbations on the uniform energy density slices of the radiation and the curvaton , respectively .
it is known that each of them is separately conserved on superhorizon scales at @xmath104 . as for @xmath103
, it may be evaluated by the energy density fluctuations on the uniform inflaton field slice at the end of inflation , @xmath105 , as @xmath106 where @xmath107 is the time at the end of inflation .
after the curvaton decay , the total curvature perturbation on the slices of uniform total energy density , @xmath108 , is conserved on superhorizon scales .
it is given by @xmath109 where @xmath110 .
for the moment , we ignore the contribution from @xmath111 and concentrate on the skewness generated by the nonlinear interaction with the tunneling field .
then we have @xmath112 where @xmath113 is given by eq . .
an important feature is that the skewness of the curvature perturbation depends on the position of the bubble due to the spatial dependence of @xmath113 .
to proceed , we focus only on the large angular scale cmb anisotropy for which the sachs - wolfe effect dominates . in this regime
, the cmb temperature anisotropy can be written in terms of the curvature perturbation as @xmath114 , where @xmath115 is the comoving position of the observer measured from the bubble center , @xmath116 is the directional cosine of the sky seen by the observer , @xmath117 is the comoving distance from the observer to the last scattering surface and @xmath118 is the time at the last scattering surface .
then , defining the non - gaussianity parameter @xmath6 as @xmath119 , we have @xmath120 one explicitly sees that the dependence on the position of the bubble center breaks the statistical isotropy . for completeness ,
let us discuss the additional contributions to @xmath6 which we have ignored .
the contribution of @xmath102 to non - gaussianity is known to be very small as long as the vacuum is in the bunch - davies vacuum @xcite .
a deviation from the bunch - davies vacuum may give rise to a non - zero non - gaussianity .
it is expected to be statistically homogeneous and isotropic but scale - dependent ( in the fourier space ) .
it may be detected by the templates of the equilateral or orthogonal types .
the contribution from the part of @xmath103 quadratic in @xmath121 gives rise to a local type non - gaussianity which is again statistically homogeneous and isotropic , and which may be detected by the squeezed type templates .
this contribution from the nonlinearity of @xmath121 in @xmath103 may be estimated as follows . roughly speaking
, @xmath122 for @xmath123 , and the contribution to @xmath124 is about @xmath125 .
thus , the condition that @xmath6 is dominated by the nonlinear interaction with the bubble is given as @xmath126 this is satisfied in an example we compute below .
for illustration , we plot @xmath127 in the cmb sky in fig .
[ fig : skymap ] .
the parameters are @xmath128 , @xmath90 @xmath129 @xmath130 , @xmath131 , @xmath132 , @xmath133 .
for the variance @xmath134 we simply impose the observational result , @xmath135 @xcite , assuming that it is dominated by @xmath102 .
a curvaton , since it never dominates the curvature perturbations .
nevertheless , we call it a curvaton for notational convenience . ] since @xmath136 when @xmath123 , this assumption is justified if @xmath137 , which is marginally satisfied in the above example
. the typical value of @xmath6 at the center of the bubble when the parameters satisfy the above conditions is estimated as @xmath138 this gives @xmath139 for the parameters given above and agrees with the result in fig .
[ fig : skymap ] within the errors shown in fig .
[ fig : f_mh ] . from this estimation , we find that the resultant @xmath6 is rather sensitive to the values of @xmath91 and @xmath140 .
it is enhanced for smaller @xmath91 and larger @xmath140 . .
, width=264 ]
in this paper , we calculated the skewness in the cmb temperature anisotropy in a model with bubble nucleation during inflation , motivated by the string theory landscape .
we considered bubble nucleation in the curvaton scenario of inflation in which the curvaton vacuum fluctuations are affected by the bubble nucleation through interaction with the tunneling field .
the calculation was done by extending the in - in formalism to the instanton background @xcite .
we found that there can be spatially localized , bubble - shaped skewness which is large inside the bubble .
as far as we know , bubble - shaped non - gaussianities have not been studied carefully yet in observation .
so it seems interesting to look for such a non - gaussianity already in the current observational data .
as pointed out by komatsu and spergel @xcite , the bispectrum , corresponding to the 3-point function , contains much more information than the single value of the skewness .
thus , analysis beyond skewness may improve the observability of bubble - shaped non - gaussianities .
we hope to come back to this issue in a future publication .
since the string theory landscape gives a strong motivation for inflation models with bubble nucleation , studies of such inflation models may be regarded as testing string theory using the universe as a laboratory . in any case , if any signature of bubble nucleation during inflation is found in observation , it gives a huge impact on the physics of the early universe , including string theory .
finally , we note that in the evaluation of the skewness we neglected the effect of deviations from the bunch - davies vacuum .
this may affect details of our result , though generic features are expected to remain the same .
this effect can be evaluated by studying the evolution of mode function on the instanton background @xcite .
we plan to come back to this issue in the near future .
ks thanks t. tanaka , y. korai and k. iwaki for useful discussions and valuable comments .
this work was supported in part by monbukagaku - sho grant - in - aid for the global coe programs , `` the next generation of physics , spun from universality and emergence '' at kyoto university .
this work was also supported in part by jsps grant - in - aid for scientific research ( a ) no . 21244033 , and by grant - in - aid for creative scientific research no .
ks was supported by grant - in - aid for jsps fellows no .
23 - 3437 .
kachru s. , kallosh r. , linde a. d. and trivedi s. p. , _ phys .
rev.d , _ * 68 * ( 2003 ) 046005 ; kachru s. _ et al .
_ , _ jcap , _ * 0310 * ( 2003 ) 013 .
susskind l. , _ arxiv : hep - th/0302219 _ ( 2003 ) ; b. , kleban m. , rodriguez martinez m. and susskind l. , _ jhep , _ * 03 * ( 2006 ) 039 .
yamamoto k. , sasaki m. and tanaka t. , _ phys .
rev.d , _ * 54 * ( 1996 ) 5031 .
yamauchi d. , linde a. , naruko a. , sasaki m. and tanaka t. , _
phys.rev.d , _ * 84 * ( 2011 ) 043513 .
coleman s. r. and de luccia f. , _ phys .
rev.d , _ * 21 * ( 1980 ) 3305 .
sugimura k. , yamauchi d. and sasaki m. , _ jcap , _ * 1201 * ( 2012 ) 027 ; aguirre a. , johnson m.c . and m. , _ phys . rev .
d , _ * 81 * ( 2010 ) 043527 .
komatsu e. _ et al .
_ , _ astrophys
. j. supp . ,
_ * 192 * ( 2011 ) 18 .
maldacena j. m. , _ jhep , _ * 0305 * ( 2003 ) 013 .
komatsu e. and spergel d. n. , _ phys .
rev.d , _ * 63 * ( 2001 ) 063002 .
creminelli p. , nicolis a. , senatore l. , tegmark m. and zaldarriaga m. , _ jcap , _ * 0605 * ( 2006 ) 004 .
senatore l. , smith k. m. and zaldarriaga m. , _ jcap , _ * 1001 * ( 2010 ) 028 .
lyth d. h. , ungarelli c. and wands d. , _
phys.rev.d , _ * 67 * ( 2003 ) 023503 ; enqvist k. and sloth m.s . , _
nucl.phys.b , _ * 626 * ( 2002 ) 395 ; moroi t. and takahashi t. , _ phys.lett.b , _ * 522 * ( 2001 ) 215 ; mazumdar a. and rocher j. , _ phys.rept . , _ * 497 * ( 2011 ) 85 .
sasaki m. , valiviita j. and wands d. , _
phys.rev.d , _ * 74 * ( 2006 ) 103003 .
birrell n. d. and davies p. c. , _ quantum fields in curved space _
( cambridge university press ) 1982 .
|
we point out a possible generation mechanism of non - gaussian bubbles in the sky due to bubble nucleation in the early universe .
we consider a curvaton scenario for inflation and assume that the curvaton field @xmath0 , whose energy density is subdominant during inflation but which is responsible for the curvature perturbation of the universe , is coupled to another field @xmath1 which undergoes false vacuum decay through quantum tunneling . for this model ,
we compute the skewness of the curvaton fluctuations due to its interaction with @xmath1 during tunneling , that is , on the background of an instanton solution that describes false vacuum decay .
we find that the resulting skewness of the curvaton can become large in the spacetime region inside the bubble .
we then compute the corresponding skewness in the statistical distribution of the cosmic microwave background ( cmb ) temperature fluctuations .
we find a non - vanishing skewness in a bubble - shaped region in the sky .
it can be large enough to be detected in the near future , and if detected it will bring us invaluable information about the physics in the early universe .
|
historically the so - called abc effect , which denotes an intriguing low - mass enhancement in the @xmath1 invariant mass spectrum , is known from inclusive measurements of two - pion production in nuclear fusion reactions to the few - body systems @xmath8 , @xmath0he and @xmath9he .
it has been named after the initials of abashian , booth and crowe , who were the first to observe this effect in 1960 by studying the inclusive @xmath10he x reaction @xcite .
its explanation has been a puzzle since then . in subsequent bubble - chamber @xcite and single - arm magnetic spectrometer measurements @xcite this enhancement
was observed also in double - pionic fusion reactions leading to @xmath8 , @xmath0he and @xmath9he , if an isoscalar pion pair was produced .
however , such an enhancement was not observed in fusion reactions leading to deuteron and triton , if an isovector pion pair was produced .
these results led to the conclusion that this effect only appears in reactions , where the participating nucleons fuse to a nuclear bound system in the final state in combination with the production of an isoscalar pion pair . in recent exclusive and kinematically complete measurements of the @xmath11 reaction
it has been demonstrated @xcite that the abc effect in this basic double - pionic fusion reaction is correlated with a narrow structure in the total cross section with quantum numbers @xmath12 , a mass of 2.37 gev and a width of about 70 mev .
the mass is about 90 mev below 2@xmath13 , the mass of a @xmath7 system , and the width is three times narrower than expected from a conventional @xmath6-channel @xmath7 process . on the contrary the basic isovector fusion process @xmath14 exhibits neither an abc effect nor a narrow resonance structure @xcite in agreement with the observations in all other @xmath15 initiated two - pion channels @xcite .
isospin decomposition of all three reactions @xmath16 , @xmath17 and @xmath14 leading to the double - pionic fusion of deuterium ensured that the resonance structure is of purely isoscalar nature @xcite .
also recently published data on the @xmath18 reaction show evidence for the resonance structure , though in this case of an isovector pion pair the abc effect is absent @xcite .
compelling evidence that the isoscalar resonance structure observed in two - pion production processes denotes truly a @xmath19-channel resonance in the @xmath20 system comes from polarized @xmath21 scattering in the energy region of the abc effect @xcite .
inclusion of these data in the said data base with subsequent partial - wave analysis produces a pole at ( @xmath22 ) mev in the coupled @xmath23 partial waves in full agreement with the here discussed resonance hypothesis . since in these latter reactions
the resonance is not associated with any abc effect , it was called no longer abc resonance , but @xmath5 @xcite in historical reference to a predicted @xcite dibaryon with exactly the quantum numbers as we observe it now .
the existence of the abc effect in the double - pionic fusion to @xmath0he and @xmath9he has been confirmed by exclusive and kinematically complete experiments at celsius / wasa @xcite and recently also at anke - cosy @xcite . in measurements at wasa - at - cosy it has been additionally shown that in the @xmath3he@xmath4 reaction the abc effect is again correlated with a resonance structure in the total cross section at @xmath24 2.37 gev + 2@xmath25 @xcite .
however , in comparison to the basic fusion reaction to deuterium the width of the resonance structure appears substantially broadened , which may be attributed to the fermi motion of the nucleons in initial and final nuclei as well as due to collision damping .
so what is left in this scenario is the question , whether also in case of the double - pionic fusion to @xmath0he the abc effect is correlated with a resonance structure in the total cross section .
in an effort to find an experimental answer for this question we have analyzed corresponding two - pion production data , which were obtained with wasa at cosy @xcite primarily for other reasons .
the data sets , which we used , originate from two different runs .
the first run concerns a proton beam of energy @xmath26 = 1.0 gev hitting the deuterium pellet target @xcite .
this allows us to analyze the reaction @xmath10he@xmath4 at @xmath26 = 1.0 gev .
the beam energy corresponds to a center - of - mass ( cm ) energy of @xmath27 = 3.42 gev = 2.48 gev + @xmath25 , _ i.e. _ pertains to the high - energy end of the region , where the abc effect has been observed previously @xcite
. the second data set used for our purposes concerns runs with a deuteron beam of @xmath28 = 1.4 and 1.7 gev , respectively , hitting the deuterium pellet target .
we use these runs to obtain data for the quasifree reaction @xmath29he@xmath30 in the range 3.1 gev @xmath31 3.4 gev ( with respect to the @xmath0he@xmath4 system ) , _ i.e. _ covering just the abc region .
both data sets allow an exclusive and kinematically complete reconstruction of the @xmath0he@xmath4 events with kinematic overconstraints .
the trigger for a valid event was just a single track in the forward detector of wasa with high thresholds in its first scintillation detector layers , in order to suppress fast protons and deuterons . with this trigger condition the data rate of accepted events was at moderate 2 khz .
the selection criteria for the offline analysis were a single he track in the forward detector and four neutral hits in the central detector .
the emerging @xmath0he particles were registered in the forward detector of wasa and identified by the @xmath32e - e technique .
the photons from the @xmath33 decay were detected and identified in the central detector @xcite .
consequently four - momenta were measured for all emitted particles of an event with the exception of the spectator neutron , which appears in the second reaction type only .
together with the condition that two pairs of the detected photons have to fulfill the @xmath33 mass condition , we have 6 overconstraints for the kinematic fit of an event in the first reaction type and 3 overconstraints in the second case . from the three possible combinations to reconstruct the four - momenta of the two pions out of four photon signals the one with the smallest @xmath34 has been selected @xcite .
all particles have been detected over the full solid angle with the exception of those @xmath0he ejectiles , which escaped in the beam - pipe ( polar angles @xmath35 ) .
acceptance and efficiency corrections have been made by use of monte carlo ( mc ) simulations of detector setup and performance . for a self - consistent procedure
the reaction model used in the mc simulations has been iteratively adjusted to the experimental results . with regard to the second data set the momentum spectra of the reconstructed neutron
are shown in fig . 1
, at the top for @xmath28 = 1.4 gev and at the bottom for @xmath28 = 1.7 gev .
the strong enhancement of events at low momenta corresponds to the situation , when the spectator neutron originates from the target deuteron , whereas the enhancement at the high - momentum end corresponds to a spectator neutron stemming from a beam deuteron .
the area in between is covered by non - quasifree processes , so - called coherent processes , where the reconstructed neutron is not just a kinematic spectator , but also plays an active role in the reaction dynamics .
misidentified @xmath9he particles could be eliminated in subsequent analysis steps by the constraint that the reconstructed neutron should not have the same direction and velocity as the detected he particle .
he@xmath36 reaction at @xmath28 = 1.4 gev ( top ) and 1.7 gev ( bottom ) , respectively .
data are given by solid dots .
the dashed ( dotted ) lines show the expected distribution for the quasifree process based on the cd bonn potential @xcite deuteron wavefunction , if the spectator originates from the target ( beam ) deuteron .
the peak near 0.5 gev / c originates from @xmath9he contamination , which has been removed in subsequent analysis steps .
, title="fig : " ] he@xmath36 reaction at @xmath28 = 1.4 gev ( top ) and 1.7 gev ( bottom ) , respectively .
data are given by solid dots .
the dashed ( dotted ) lines show the expected distribution for the quasifree process based on the cd bonn potential @xcite deuteron wavefunction , if the spectator originates from the target ( beam ) deuteron .
the peak near 0.5 gev / c originates from @xmath9he contamination , which has been removed in subsequent analysis steps .
, title="fig : " ] he ejectiles in the @xmath37he@xmath36 reaction at @xmath28 = 1.7 gev versus the @xmath0he lab scattering angle @xmath38 . at the top data are shown and at the bottom the corresponding mc simulation of the processes with either the neutron spectator in the target or in the beam .
the high - density area at small angles and large kinetic energies corresponds to the process with the spectator neutron in the target , whereas the distribution at small energies and large scattering angles belongs to spectator neutrons in the beam .
the area in between corresponds to coherent processes .
, title="fig : " ] he ejectiles in the @xmath37he@xmath36 reaction at @xmath28 = 1.7 gev versus the @xmath0he lab scattering angle @xmath38 . at the top data are shown and at the bottom the corresponding mc simulation of the processes with either the neutron spectator in the target or in the beam .
the high - density area at small angles and large kinetic energies corresponds to the process with the spectator neutron in the target , whereas the distribution at small energies and large scattering angles belongs to spectator neutrons in the beam . the area in between corresponds to coherent processes .
, title="fig : " ] in case of a target spectator neutron ( @xmath39 reaction ) the emitted @xmath0he particles are at very small forward angles due to the lorentz boost , so that the lower limit of @xmath40 cuts severely into the reaction phase space rendering acceptance corrections largely model - dependent . hence we refrain from giving results for this scenario . for the case of the neutron spectator originating from the beam deuteron ( @xmath41 reaction ) the situation is kinematically much more favorable .
unfortunately also here we met difficulties , since the energies of the @xmath0he ejectiles deposited in the segments of the forward detector turned out to be partly below the trigger thresholds , which were increased for the observation of other reaction channels of primary interest in these runs .
as a consequence we had to tune the actual trigger thresholds individually for each of the detector segments very carefully by adjusting the simulations of the detector performance to the observed response of each of the corresponding scintillation detectors . comparing the spectra in fig . 1 , top and bottom
, we see that at higher beam energy the coherent process has much reduced compared to the quasifree process . hence , also in this respect it is more favorable to analyze the 1.7 gev data for the @xmath0he@xmath4 production channel .
2 displays the two - dimensional scatter plot of the kinetic energy of the @xmath0he ejectile versus its polar scattering angle in the lab system .
the @xmath0he particles originating from the quasifree process in the target deuteron ( @xmath39 reaction ) produce a strong enhancement at small angles in combination with large kinetic energies in the scatter plot , whereas the @xmath0he ejectiles from the quasifree process in the beam deuteron ( @xmath41 reaction ) produce a strong enhancement at small energies over a wide region of angles . in the scatterplot these two regions are strongly populated and well separated from the region in between , which covers coherent processes . in order
to get rid of the latter as well as of the target related spectators , we subsequently constrain the polar angle for the reconstructed neutrons to the kinematical spectator limit of @xmath42 for beam related spectators , where the superscript @xmath43 denotes the angle in the @xmath39 center - of - mass system . that way we obtain a momentum spectrum of the spectator neutrons , which is very close to that given by the dotted line in fig . 1 and which is essentially free of background .
the absolute normalization of the data from the single - energy measurement at @xmath26 = 1.0 gev was obtained by a relative normalization to the @xmath10he@xmath33 reaction measured simultaneously with the same trigger .
our results for this reaction in turn have been normalized to those from saclay measurements at neighboring energies @xcite .
though this procedure appears to be straightforward , it contains a number of difficulties .
the saclay data appear to be most reliable at @xmath44 @xcite , where wasa can not measure .
hence we used the full back - angle hemisphere to adjust the wasa results to those of saclay .
however , due to the scarcity of saclay data at finite angles we estimate that the total uncertainty in the absolute normalization could be as large as 30@xmath45 .
for details see ref .
@xcite . the data of the quasi - free run overlap with the single - energy measurement at their high - energy end .
hence , for simplicity they have been normalized to the result of the single - energy measurement .
resulting observables of the normalized as well as acceptance and efficiency corrected data are displayed in figs .
the total cross section data obtained from the analysis of both experiments are shown in fig . 3 , which exhibits the energy dependence of the total cross section for the @xmath0he@xmath4 production .
our result from the run at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) is shown by the filled triangle symbol , whereas the results from the quasifree run are given by the filled circles .
the shaded area denotes the estimated systematic uncertainties , which result dominantly from the efficiency and acceptance corrections .
also uncertainties from rest gas contributions and kinematic fit are contained in this estimate . included in fig .
3 are also the results from previous exclusive measurements at celsius - wasa at @xmath26 = 0.893 gev ( open square ) @xcite and at promice / wasa at @xmath26 = 0.477 gev ( open circle ) @xcite , the latter carried out at celsius , too . in order to avoid systematic discrepancies in the procedure used for the absolute normalization
, the celsius - wasa result has been reanalyzed by subjecting it to exactly the same procedure ( _ i.e. _ , considering the full back - angle hemisphere ) as applied now for the single - energy measurement at @xmath26 = 1.0 gev . as a result
the celsius - wasa value at @xmath26 = 0.893 gev changed from the published value of 2.8(3 ) @xmath46b @xcite to 1.9(3 ) @xmath46b with the latter value being plotted in fig .
the revised value is in good agreement with the new data .
we note that the cosy - anke result for the @xmath10he@xmath47 reaction is also lower by 40@xmath45 @xcite , when compared to the corresponding published value from celsius - wasa in agreement with our finding for the @xmath0he@xmath4 channel .
he with the production of a @xmath4 pair .
data obtained in this work by measurements of the @xmath48he@xmath4 reaction at @xmath26 = 1.0 gev and of the @xmath37he@xmath30 reaction at @xmath28 = 1.7 gev are given by the filled cross and the filled circles , respectively .
they are compared to previous results from promice / wasa @xcite ( open circle ) and celsius / wasa @xcite ( open square ) .
the latter has been renormalized , see text .
the shaded area denotes the estimated systematic uncertainties .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath49 process and the solid line their ( coherent ) sum . ] versus @xmath50 of the data at cm energies of @xmath51 = 3.25 gev , @xmath51 = 3.31 gev , @xmath51 = 3.35 gev and @xmath51 = 3.41 gev ( from top to bottom ) .
, title="fig : " ] versus @xmath50 of the data at cm energies of @xmath51 = 3.25 gev , @xmath51 = 3.31 gev , @xmath51 = 3.35 gev and @xmath51 = 3.41 gev ( from top to bottom ) .
, title="fig : " ] versus @xmath50 of the data at cm energies of @xmath51 = 3.25 gev , @xmath51 = 3.31 gev , @xmath51 = 3.35 gev and @xmath51 = 3.41 gev ( from top to bottom ) .
, title="fig : " ] versus @xmath50 of the data at cm energies of @xmath51 = 3.25 gev , @xmath51 = 3.31 gev , @xmath51 = 3.35 gev and @xmath51 = 3.41 gev ( from top to bottom ) .
, title="fig : " ] [ t ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum . , title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) . filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev )
. data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) .
filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) . data from celsius
/ wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] ( * left * ) and @xmath52 ( * right * ) at @xmath51 = 3.25 , 3.31 , 3.35 and 3.41 gev ( from * top * to * bottom * ) . filled circles denote data from the quasifree runs , open crosses those from the @xmath41 reaction at @xmath26 = 1.0 gev ( @xmath27 = 3.416 gev ) .
data from celsius / wasa at @xmath26 = 0.89 gev ( @xmath27 = 3.35 gev ) @xcite are shown by open squares .
the shaded area denotes the phase - space distribution .
the dotted curve gives the @xmath5 contribution , the dashed line the @xmath6-channel @xmath7 process and the solid line their sum .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] he ( left ) and @xmath33 ( right ) ejectiles in the center - of - mass system .
, title="fig : " ] for sake of completeness we note that there is also a cosy - momo measurement of the @xmath0he@xmath47 channel at 70 mev above threshold , _
i.e. _ at @xmath27 = 3.16 gev @xcite . from the observed @xmath53 distributions
it was concluded that the produced @xmath47 pair is dominantly in relative @xmath54-wave , _ i.e. _ of isovector character @xcite , which is excluded in the @xmath4 system discussed here .
the energy dependence of the total cross section is consistent with some resonance - like structure , though we do not observe a substantial decrease of the cross section at high energies within the measured interval .
the cross section appears to peak at a similar excess energy as was observed in the fusion reactions to deuterium and @xmath9he .
however , as the detailed investigation of the differential cross sections will show , the @xmath5 resonance at @xmath27 = 2.37 gev + @xmath25 shows up in the total cross section only as a shoulder within the ascending slope . in marked difference to the double - pionic fusions to @xmath8 and
@xmath9he the main contribution to the total cross section in the @xmath0he case does not originate from the @xmath5 resonance , but from the conventional @xmath6-channel @xmath7 process , which has a large isovector contribution .
this process peaks at around @xmath55 and has a width of about @xmath56 @xcite .
next we discuss the differential cross sections , which are shown in figs . 4 - 6 and which completely describe the 3-body reaction .
the shape of all differential distributions remains rather stable over the region of the @xmath5 resonance structure , however , starts to change significantly towards the high - energy end of the measured region , where the @xmath6-channel @xmath7 process becomes dominant .
4 shows the variation of the dalitz plots for the invariant masses squared @xmath57 versus @xmath50 over the measured resonance region .
the dalitz plots are similar to those obtained in the basic reaction .
they exhibit an enhancement in horizontal direction , in the region of the @xmath32 excitation , as it prominently shows up in the @xmath58 spectra displayed in fig . 5 .
this feature is consistent with the excitation of a @xmath7 system in the intermediate state as discussed for the basic reaction @xcite .
more prominent and also similar to the situation in the basic reaction we observe here the abc effect as a strong enhancement at the low - mass kinematic limit of @xmath52 .
consequently the dalitz plot is mainly populated along the @xmath1 low - mass border line . in fig . 5 , left ,
the @xmath58 distribution is shown for four selected energies over the measured region .
at all energies this distribution is far from phase - space like ( shaded areas in fig . 5 ) and exhibits a clear signal from @xmath32 excitation . the @xmath52 distribution is shown on the right - hand side of fig . 5 for four different beam energies .
it clearly exhibits the abc effect at the lower three energies . at the highest energy
we see the transition to a two - hump structure with both a low - mass enhancement and a high - mass enhancement .
the latter is the characteristic feature of the @xmath6-channel @xmath7 process as predicted originally by risser and shuster @xcite in search for a plausible explanation of the abc effect . in fig .
6 we show angular distributions at the selected energies . on the left the angular distribution of the @xmath0he ejectiles is depicted and on the right that of the emitted @xmath33 particles both in the center - of - mass system .
since the collision partners are not identical particles , the cm angular distributions do not need to be symmetric about @xmath59 .
however , in case of a @xmath19-channel resonance process they have to be and the data appear to be compatible with this .
the observed @xmath0he angular dependence is similar to the corresponding one in the basic reaction , though significantly more peaked near @xmath60 however , still less curved than in the double - pionic fusion to @xmath9he .
the @xmath33 angular distribution resembles that for p - waves as one would expect from the decay of @xmath32s in the intermediate state .
note that an intermediate @xmath7 system shows up both in the case of @xmath5 excitation and in the case of a @xmath6-channel meson exchange leading to a mutual excitation of the colliding nucleons to their first excited state , the @xmath32 resonance ( @xmath6-channel @xmath7 process ) .
since the features , which we observe here , are very similar to those observed for the basic double - pionic fusion reaction , we adapt the ansatz used there for the description of the @xmath0he case @xcite .
there are only two major differences : * first , the nucleons momenta are smeared due to their fermi motion in initial and final nuclei . in particular
the fermi motion in the appreciably bound @xmath0he nucleus leads to a sizeable smearing of the energy dependence in the total cross section adding nearly 30 mev to the total width .
* second , the reaction process @xmath10he@xmath4 involves also the proton within the target deuteron , which does not participate actively in the formation of the @xmath20 resonance , but finally forms a bound @xmath0he system together with the @xmath20 pair from the decay of the @xmath5 resonance .
the results of this calculation is shown in figs . 3 , 5 - 6 by the solid lines , which provide a reasonable description of the data .
in these calculations it is assumed that both the @xmath5 resonance and the @xmath6-channel @xmath7 process happen on the active @xmath20 pair .
the @xmath7 process is of both isoscalar and isovector character . from isospin
coupling it follows that the latter is more than three times as large @xcite .
since only the isoscalar part of the @xmath7 process interferes with @xmath5 , the interference effect between both processes is small .
the relative size of both processes as well as the width of @xmath5 resonance has been adjusted for best reproduction of the observed @xmath52 distributions .
the resulting effective @xmath5 width of 85 mev means that there is if at all only a small broadening due to collision damping .
it is appreciably smaller than in the @xmath9he case , where the collision broadening was about 50 mev .
the result of the fit to the @xmath52 spectra has been scaled in absolute height to the total cross section data in fig .
we see that the @xmath5 resonance dominates only at low energies in the strongly ascending part of the total cross section .
thereafter the conventional @xmath6-channel @xmath7 process takes over . from fig .
3 we see that the maximum cross section for @xmath5 production in the process @xmath61he@xmath4 is about 0.8 @xmath46b .
this is a factor of 300 less than in the basic reaction @xmath62 , but also about a factor of two less than in the process @xmath63he@xmath4 , where there are twice as many combinatoric possibilities to form @xmath5 in the intermediate state .
this result suggests that the @xmath5 production in still heavier nuclei does not just fade away , but rather could give sizable contributions . in view of the now achieved understanding of the double - pionic fusion it appears historically rather fortunate that more than fifty years ago the berkeley 184-inch synchrocyclotron allowed only a maximum proton beam energy of 743 mev .
that way abashian , booth and crowe were in the position to just enter the energy region , where the @xmath1 low - mass enhancement appears to be largest and thus discover the abc effect .
would they have had access to a beam energy of 1 gev instead , they would then have observed both a low - mass and a high - mass enhancement with the latter being the dominant one .
but such a scenario was readily explained later - on by risser and shuster @xcite to originate naturally from the conventional @xmath6-channel @xmath7 process .
would thus the abc puzzle has escaped possibly its detection without providing later on the trace to the discovery to the @xmath5 resonance ?
in conclusion , our data on the double - pionic fusion process to @xmath0he establish the correlation of a resonance - like energy dependence in the total cross section with the abc effect in very much the same way as shown before for the double - pionic fusion reactions to deuterium and @xmath9he . a calculation based on the @xmath5 resonance
gives a good account of the observed distributions .
the enlarged width of the resonance - structure in the total cross section is explained by the fermi motion of the nucleons in initial and final nuclei , which includes also collision damping . that way the abc effect in the double - pionic fusion to nuclei
is traced back to a @xmath20 resonance , which obviously is strong enough to survive even in the nuclear medium .
it would be very interesting to see , whether also in nuclei heaver than he both abc effect and @xmath5 resonance could be observed .
since the next heavier nuclei are not stable or do not have the proper spin and isospin , the next suitable candidate reaction appears to be @xmath64 or in inverse kinematics @xmath65 .
however , measurements of such reactions necessitate dedicated detector setups and /or accelerators . another great possibility to search for abc effect and @xmath5 resonance
might be given by high - resolution measurements of heavy ion reactions .
a trace that @xmath5 production , indeed , takes place in heavy - ion collisions has recently been found in connection with the so - called dls puzzle , which denotes an unusual enhancement in the @xmath66 production at 0.3 @xmath67 0.7 gev initiated by neutron - proton collisions in vacuum or within heavy - ion systems . in ref .
@xcite it has been shown that a possible solution of this puzzle is presented by accounting for @xmath7 and @xmath5 production .
since a dibaryon resonance has integer spin and thus is of bosonic nature , the survival of such a resonance in a nuclear surrounding may have an important impact on the equation of state .
bosons are not pauli - blocked and as such allow for higher densities under same pressure and energy conditions .
the behavior of matter under extreme conditions is needed , _
e.g. _ , for a better understanding of the evolution of compressed matter in the course of heavy - ion collisions or in compact ( neutron ) stars see , _
e.g _ refs . @xcite .
we acknowledge valuable discussions with c. hanhart , e. oset and c. wilkin on this issue . this work has been supported by bmbf , forschungszentrum jlich ( cosy - ffe ) , dfg , the polish national science centre through the grant no .
2011/01/b / st2/00431 and by the foundation for polish science ( mpd ) . n. e. booth , a. abashian , k. m. crowe , phys .
* 7 * , 35 ( 1961 ) ; * 5 * , 258 ( 1960 ) ; phys . rev . * 132 * , 2296ff ( 1963 )
. i. bar - nir _ et al .
b * 54 * , 17 ( 1973 )
. a. abdivaliev _
et al . _ ,
* 29 * , 7896(1979 ) .
r. j. homer _ et al .
* 9 * , 72 ( 1964 )
. j. h. hall _
et al . _ ,
b * 12 * , 573 ( 1969 ) .
j. banaigs _
et al . _ ,
b * 43 * , 535 ( 1973 ) .
j. banaigs _
et al . _ ,
b * 67 * , 1 ( 1973 ) .
j. banaigs _
et al . _ ,
b * 105 * , 52 ( 1976 ) .
et al . _ ,
a * 302 * , 413 ( 1978 ) .
f. plouin , p. fleury , c. wilkin , phys .
lett . * 65 * , 690 ( 1990 ) .
r. wurzinger _
et al . _ ,
phys . lett .
b * 445 * , 423 ( 1999 ) . for
a review see a. codino and f. plouin , lns / ph/94 - 06 p. adlarson
_ et al . _ ,
b * 721*,229 ( 2013 ) .
p. adlarson _
et al . _ ,
106 * , 242302 ( 2011 ) .
m. bashkanov _
et al . _ ,
lett . * 102 * , 052301 ( 2009 ) .
et al . _ ,
b * 684*,110 ( 2010 ) and b * 702 * , 312 ( 2011 ) ; arxiv:0910.0995v2 [ nucl - ex ] .
h. clement , prog .
. phys . * 67 * , 486 ( 2012 ) .
t. skorodko _
b * 695 * , 115 ( 2011 ) .
p. adlarson _
b * 706 * , 256 ( 2012 ) .
t. skorodko _
et al . _ ,
b*679 * , 30 ( 2009 ) .
t. skorodko _
et al . _ ,
j. a * 47 * , 108 ( 2011 ) .
p. adlarson _
et al . _ , phys .
c * 88 * , 055208 ( 2013 ) .
p. adlarson _
* 112 * , 202301 ( 2014 ) .
p. adlarson _
et al . _ ,
submitted for publication ; arxiv:1408.4928 [ nucl - ex ] .
t. goldman , k. maltman , g. j. stephenson , k. e. schmidt , fan wang , phys .
c * 39 * , 1889 ( 1989 ) .
j. l. ping , h. x. huang , h. r. pang , fan wang , phys .
c * 79 * , 024001 ( 2009 ) .
m. bashkanov _
et al . _ ,
b * 637 * , 223 ( 2006 ) .
s. keleta _
et al . _ ,
a * 825 * , 71 ( 2009 ) .
m. mielke _
j. a * 50 * , 102 ( 2014 ) .
p. adlarson _
et al . _ ,
c * 86 * , 032201 ( 2012 ) rapid communication .
et al . _ ,
inst . meth .
a * 594 * , 339 ( 2008 ) .
h. h. adam et al . , arxiv : nucl - ex/0411038 ( 2004 ) .
r. machleidt , phys .
c * 63 * , 024001 ( 2001 ) . c. kerboul
_ et al . _ ,
b * 181 * , 28 ( 1986 ) .
j. banaigs _
b * 45 * , 394 ( 1973 ) .
e. perez del rio , phd thesis , university of tbingen 2014 , http://hdl.handle.net/10900/53837 .
m. andersson et al . , phys .
b * 485 * , 327 ( 2000 ) .
f. bellemann et al . , phys . rev .
c * 60 * , 061002 ( 1999 ) .
m. i. krivoruchenko _
nucl . * 74 * , 371 ( 2011 ) .
r. m. aguirre and m. schvellinger , phys .
b * 449 * , 161 ( 1999 ) .
a. faessler , a. j. buchmann and m. i. krivoruchenko , phys .
c * 57 * , 1458 ( 1998 ) .
a. w. thomas , d. l. whittenbury , j. d. carroll , k. tsushima and j. r. stone , epj web of conferences * 63 * , 03004 ( 2013 ) .
|
exclusive and kinematically complete measurements of the double pionic fusion to @xmath0he have been performed in the energy region of the so - called abc effect , which denotes a pronounced low - mass enhancement in the @xmath1-invariant mass spectrum .
the experiments were carried out with the wasa detector setup at cosy .
similar to the observations in the basic @xmath2 reaction and in the @xmath3he@xmath4 reaction , the data reveal a correlation between the abc effect and a resonance - like energy dependence in the total cross section .
differential cross sections are well described by the hypothesis of @xmath5 resonance formation during the reaction process in addition to the conventional @xmath6-channel @xmath7 mechanism .
the deduced @xmath5 resonance width can be understood from collision broadening due to fermi motion of the nucleons in initial and final nuclei .
|
blazars are among the most powerful accelerators of relativistic particles in nature , as shown by egret observations of 100 mev - gev emission from over 60 fsrqs and bl lac objects .
consequently , blazars represent a potential source of high - energy neutrinos ( see gaisser et al .
1995 and halzen 2001 for reviews ) .
the bright , highly variable @xmath2 radiation from blazars argues for the existence of emission regions with high photon energy densities and intense nonthermal particle populations , which are needed for efficient photomeson production ( e.g. , mannheim and biermann 1992 ; mannheim et al.2001 ) .
previous papers have considered internal synchrotron photons as targets for high - energy proton interactions ( e.g. , mannheim 1993 ) .
here we show that uv photons from external radiation fields , which have earlier been proposed as a target photon source for nonthermal electrons to produce compton - scattered gamma - rays ( e.g. , dermer and schlickeiser 1993 ; sikora et al .
1994 ) , also provide the most important photon source for photomeson production of @xmath3 tev neutrinos in fsrqs .
we use the measured spectral fluxes and variability time scales of synchrotron and compton radiation from the blazar 3c 279 to determine physical parameters of its gamma - ray emission region .
lower limits to the doppler factor @xmath4 are defined by the condition that the emitting region be transparent to gamma rays .
equipartition arguments and observed fluxes of the synchrotron and compton components are used to determine the magnetic field strength @xmath5 in the plasma blob . assuming efficient acceleration of power - law energetic protons to a maximum energy limited by the magnetic field and size scale of the region
, we calculate spectral emissivities of pion decay neutrinos formed through photomeson production . because the properties of the external radiation field are independent of the bulk lorentz factor @xmath6 of the radiating region , neutrino emissivity calculations using duty cycle estimates give accurate neutrino flux estimates during low @xmath2-ray states .
we predict detectable fluxes of neutrinos from @xmath2-ray loud fsrqs using km@xmath0 neutrino telescope arrays such as icecube .
the comoving synchrotron photon energy density is given by @xmath7 @xmath8@xmath9 , where @xmath10 is the observed bolometric synchrotron luminosity , @xmath11 is the comoving radius of the blob , here assumed spherical , and primes denote comoving quantities .
the ratio of the ssc and synchrotron luminosities @xmath12 , where @xmath13 . for simplicity
, we consider the regime where the quasi - isotropic scattered external radiation field dominates the direct disk radiation field ( dermer and schlickeiser 1994 ) .
if the @xmath2-rays are due to external compton - scattered radiation , then @xmath14 ( sikora 1997 ; dermer , sturner , and schlickeiser 1997 ) . combining these two expressions
gives @xmath15@xmath16 @xmath17 @xmath18
@xmath19 when viewing within the beam of the jet ( @xmath20 ) . consequently @xmath21 @xmath22 @xmath23 , where @xmath24 the energy of the external photons in the comoving frame is @xmath25 .
models taking into account ssc and external compton ( ec ) components show that a complete spectral fit requires synchrotron , ssc and ec components ( bttcher 1999 ; hartman et al .
thus @xmath26 - 1 for bl lac objects and @xmath27 - 10 for fsrqs . the spectral energy density @xmath28 of the photons in the blob frame is given in terms of the @xmath29 flux @xmath30 through the relation @xmath31 where we relate the measured variability time scale @xmath32 to the blob radius @xmath11 through the expression @xmath33 @xmath34
, @xmath35 is the observed variability time scale in days , and @xmath36 is the redshift .
the expression @xmath37 relates the photon energies measured in the source and observer frames .
the magnetic field in the blob can be estimated by noting that the total soft photon energy density @xmath38 , where @xmath39 is the total compton power .
hence @xmath40 .
using the earlier expressions for @xmath41 and @xmath11 gives @xmath42 an independent estimate for @xmath5 can also be made by introducing the equipartition parameter @xmath43 for the ratio of nonthermal electron to magnetic energy densities in the jet .
thus @xmath44 corresponds to equipartition , though with electrons only . for the measured synchrotron flux density @xmath45 with @xmath46 at @xmath47 , this gives @xmath48^{2/7}\;{\nu_{12}^{1/7 } d_{28}^{4/7}[s({\rm jy})]^{2/7}(1+z)^{5/7}\over \eta^{2/7}[t_{\rm var}({\rm d})]^{6/7}\delta^{13/7}}\ ; \label{beq}\ ] ] in gauss , where @xmath49 is the flux density measured at @xmath50 hz . transparency to @xmath51 pair - production attenuation requires that @xmath52 , which can be rewritten as a lower limit on @xmath4 using equation ( [ eq1 ] ) .
energy losses of relativistic protons ( and neutrons ) are calculated on the basis of standard expressions ( see , e.g. , berezinskii and grigoreva 1988 ) for the cooling time of relativistic protons due to photopion production in @xmath53 collisions .
if the ambient photons have spectral density @xmath54 , then @xmath55 where @xmath56 is the proton lorentz factor , @xmath57 is the photon energy in the proton rest frame , @xmath58 is the photopion production cross - section , @xmath59 is the threshold energy for the parent photon in the proton rest frame , and @xmath60 is the inelasticity of the interaction .
the latter increases from @xmath61 at energies not very far above threshold where the single pion production channels dominate , to @xmath62 at larger values of @xmath57 where multi - pion production dominates ( e.g. see berezinskii and grigoreva 1988 , mcke et al .
1999 ) .
a detailed recent study of this photohadronic process has been given by mcke et al .
( 1999 ) . in order to simplify calculations , we approximate the cross - section @xmath63 as a sum of 2 step - functions @xmath64 and @xmath65 for the ( total ) single - pion and multi - pion channels , respectively , with @xmath66 for @xmath67 and @xmath68 outside this region , whereas @xmath69 at @xmath70 . the inelasticity is approximated as @xmath71 and @xmath72 below and above 500 mev .
the results of our calculations in this simplified approach are found to give good agreement with earlier calculations ( berezinskii and grigoreva 1988 ) . in particular
, the time scales for photopion interactions of ultra - relativistic cosmic rays with the cmbr are accurately reproduced ( see also stanev et al .
this approach also works well for a broad power - law distribution of field photons @xmath73 for different photon spectral indices @xmath74 , and readily explains the significant increase in the mean inelasticity of incident protons ( or neutrons ) from @xmath75 for steep photon spectra @xmath76 with @xmath77 , to @xmath78 for hard spectra with @xmath79 ( mcke et al .
1999 ) .
calculations of the spectra of secondary particles and photons ( @xmath80 ) are done in the @xmath4-function approximation , assuming that the probabilities for producing pions of different charges ( @xmath81 , @xmath82 and @xmath83 ) are equal for the multi - pion interaction channel . to correctly apply the @xmath4-function approximation
, one has to properly take into account the different inelasticities of the multi - pion and single - pion production channels , which could explain the statement of mcke et al .
( 1999 ) that the @xmath4-function approximation does not work well in the case of hard photon spectra .
we consider 3c 279 as a prototype fsrq .
the broad - band radiation spectra produced by the relativistic ejecta in the jet of 3c 279 is generally thought to consist of a low energy synchrotron component and high energy compton component . in 3c 279
, the synchrotron component dominates at observed frequencies @xmath84 .
we assume that the @xmath85 hz synchrotron emission is cospatially located with the gamma - ray production region .
this assumption is supported by observations of correlated variability between the synchrotron and compton components in some bl lac objects and fsrqs .
the high - energy compton emission consists of an ssc component and an external compton ( ec ) component due to target photons that originate from outside the jet ; for example from the disk , torus , scattered - disk radiation , reflected synchrotron radiation , etc . for calculations of @xmath86 from 3c 279 ,
we approximate the differential photon flux @xmath87 observed during the powerful flare of 1996 ( wehrle et al .
1998 ) in the form of a continuous broken power - law function , with power - law photon indices @xmath88 , @xmath89 , and @xmath90 at frequencies @xmath91 , @xmath92 , and @xmath93 , respectively . the @xmath29 energy flux @xmath94 of the synchrotron radiation reaches a maximum value of @xmath95 at @xmath96 .
the maximum flux of the compton component is @xmath97 times larger than the peak synchrotron flux during the flare , and is reached at @xmath98 .
we suppose that the ssc component makes a substantial contribution to the hard x - ray emission in 3c 279 ( hartman et al .
2001 ) , and therefore that @xmath99 .
the redshift of 3c 279 is @xmath100 , and its luminosity distance is @xmath101 for an @xmath102 , @xmath103 cosmology with a hubble constant of 65 km s@xmath104 mpc@xmath104 .
the measured @xmath105 hz flux during the 1996 flare implies that @xmath106^{0.19}$ ] from the @xmath51 transparency arguments .
thus @xmath107 corresponds to the minimum doppler - factor required for relativistic blobs in 3c 279 to become transparent to absorption of the observed gev photons in the radiation field inside the blob . expressing @xmath11 via the observed variability timescale @xmath108 1 - 2 days as observed by egret ( wehrle et al .
1998 ) , one finds from equation ( [ eq1 ] ) that @xmath109^{-2 } \ ,
\delta_{5}^{-6 } \ ; { \rm erg~cm}^{-3 } \ , .\ ] ] here @xmath110 @xmath111erg @xmath112 s@xmath104 ) , and @xmath113 . from the observed flux of synchrotron radiation from 3c 279 with photon number index @xmath114 at @xmath115
, we deduce @xmath116^{-6/7 } \ , \delta_{5}^{-13/7}$ ] from equipartition arguments .
the estimation of @xmath5 from equation ( [ b ] ) gives @xmath117 , using @xmath118 ergs s@xmath104 , @xmath119 , and @xmath99 .
the external radiation energy density in the blob rest frame is , using the expression for @xmath5 derived from equipartition arguments , given by @xmath120^{-12/7}\delta_{5}^{-26/7 } \,{\rm erg~cm}^{-3 } \ , .\ ] ] for the calculations we assume @xmath121 and take @xmath122 .
we assume that the spectrum of the external uv radiation field arises from a shakura / sunyaev ( 1973 ) optically - thick accretion disk model that is scattered by broad - line region clouds .
this disk model has flux density @xmath123 up to a maximum photon energy @xmath124 determined by the innermost radius of the blackbody disk and properties of the central engine .
we take @xmath125 ev ( dermer and schlickeiser 1993 ) . the energy density of the external radiation in the comoving plasma blob frame can be determined by noting that @xmath126 corresponds to the ratio @xmath127 of the directional external compton and synchrotron powers as measured along the jet axis . , assuming 3 different doppler - factors for the jet : @xmath128 ( solid curve ) , @xmath129 ( dashed curve ) , and @xmath130 ( dot - dashed curve ) .
the dotted and triple - dot dashed curves are calculated for @xmath128 and @xmath129 , respectively , when @xmath131 interactions with the synchrotron radiation field alone are considered.,width=313 ] fig . 1 shows energy loss timescales ( in the observer frame ) of protons due to photopion production in a jet of 3c 279 calculated for 3 different doppler - factors : @xmath128 ( solid curve ) , @xmath129 ( dashed curve ) , and @xmath132 ( dot - dashed curve ) .
the dotted and triple - dot dashed curves show the photopion timescales corresponding to interactions with the synchrotron radiation only , for the cases @xmath133 and @xmath129 . calculated for the same jet lorentz - factors @xmath134 , 10 , and 15 and assumptions for the external and internal radiation fields as made in fig . 1 , assuming a power - law injection of relativistic protons with number index @xmath135 during @xmath136 weeks with a power @xmath137 ergs s@xmath104 . , width=313 ] fig . 2 shows the energy distributions of relativistic protons @xmath138 in the ejecta / blobs ( for the same cases as in fig . 1 ) , which are calculated assuming power - law injection of relativistic protons with number index @xmath135 on observed timescales @xmath139 weeks .
the total injection power of the protons @xmath140 ergs s@xmath104 .
note that @xmath141 s@xmath104 is the characteristic _ apparent _ luminosity of the flares detected by egret from 3c 279 . in calculations of @xmath142
we take into account the photohadron interaction energy losses , as well as the escape losses of the protons in the bohm diffusion limit .
this limit is more restrictive than the condition that the gyroradius of the highest energies of accelerated protons should not exceed the blob size ( hillas 1984 ) .
3 shows the expected energy fluxes of the neutrinos produced by protons in fig . 2 .
for the fluxes in fig . 3 , the total number of neutrinos that could be detected by a @xmath143 detector during 2 weeks of the flare , using neutrino detection efficiencies given by @xcite is 0.19 , 0.12 and 0.05 for the cases of @xmath144 , 10 and 15 , respectively .
it is important to note that if the external uv field is neglected , these numbers would be significantly reduced .
we calculate the number of neutrinos that would be detected by a km@xmath0 array ranging from @xmath145 to @xmath146 for the fluxes in fig . 3 , which would not leave a realistic prospect for the detection of at least 2 - 3 neutrinos , which is required for a positive detection of a source .
bl lac objects , which have weak broad line regions and , therefore , a weak scattered external radiation field , should consequently be much weaker neutrino sources .
because the properties of the external radiation field are insensitive to the value of the blob lorentz factor , neutrino production in the quiescent state can be accurately estimated on the basis of properties of the external radiation field derived during the flaring state .
for a typical duty factor of the flares @xmath147 1 - 2 months per year , and considering the additional neutrino production during the quiescent phase , we can reasonably expect that icecube or other km@xmath0 array may detect several neutrinos per year from 3c 279-type blazar jets with @xmath148 .
allowing for the possibility that the nonthermal protons may have an overall power larger than that for primary electrons improves the prospects for neutrino detection .
it is important to notice that the external photon field reduces very significantly the characteristic energy of protons for effective photopion production , compared with the case of the internal synchrotron radiation field .
in fact , acceleration of protons only to @xmath149 is required for efficient neutrino production through photomeson interactions , as can be seen from fig.1 .
this effect permits us to predict that a km@xmath0 array will detect high - energy neutrinos from fsrqs without requiring acceleration of protons to ultra - high energies @xmath150 .
( solid curve ) @xmath151 ( dashed curve ) , and @xmath152 ( dot - dashed curve ) .
the dotted and triple - dot dashed curves show the calculated neutrino fluxes if the external photon field in the jet is neglected.,width=313 ] 99 berezinskii , v. s. and grigoreva , s. i. 1988 , astron .
199 , 1 b " ottcher , m. 1999 , astrophys .
j. 515 , l21 dermer , c.d . , and schlickeiser , r. 1993 , astrophys .
j. 416 , 458 ( ds93 ) dermer , c.d . , and schlickeiser , r. 1994 , astrophys .
j. suppl .
90 , 945 dermer , c.d . ,
sturner , s.j . , and schlickeiser , r. 1997 , astrophys .
j. suppl .
109 , 103 gaisser , t.k .
, halzen , f. , and stanev , t. 1995 , phys .
repts . , 258(3 ) , 173 halzen , f. , 2001 , in high energy gamma ray astronomy , ed . f.a .
aharonian and h.j .
vlk ( aip : new york ) , p. 43
hartman , r.c . , et al .
2001 , astrophys .
j. in press ( astro - ph/0102127 ) hillas , a.m. 1984 , ann .
22 , 425 mannheim , k. 1993 , astron .
269 , 67 mannheim , k. and biermann , p.l .
1992 , astronomy and astrophys .
253 , l21 mannheim , k. , protheroe , r.j .
, & rachen , j.p .
2001 , phys .
d , 63 , 023003 mcke , a. , rachen , j.p .
, engel , r. , protheroe , r.j . , & stanev , t. 1999 , pub.astron .
australia , 16 , 160 sikora , m. , begelman , m.c . , and rees , m.j .
1994 , astrophys .
j. 421 , 153 sikora , m. 1997 , in the fourth compton symposium , ( aip : new york ) , p. 494 shakura , n.i . , and sunyaev , r.a .
1973 , a&a , 24 , 337 stanev , t. , engel , r. , mcke , a. , protheroe , r.j . , and rachen , j.p . 2000 , phys .
d , submitted ( astro - ph/0003484 ) wehrle , a.e .
, et al .
1998 , astrophys .
j. 497 , 178
|
the measured spectral energy distribution and variability time scale are used to determine the radiation and magnetic - field energy densities in the relativistic plasma that forms the gamma - ray emitting jet in the blazar 3c 279 . assuming that protons are accelerated as efficiently as electrons to a maximum energy determined by the size and magnetic field of the emitting region , we calculate the emissivity of neutrinos produced by protons that interact with the external radiation field intercepted by the jet .
the external radiation field provides the most important target photons for photomeson production of high - energy neutrinos in flat spectrum radio quasars ( fsrqs ) . because of photomeson interactions with this field , km@xmath0
neutrino telescopes are predicted to detect @xmath1 - 1 neutrinos per year from blazars such as 3c 279 .
bl lac objects are weaker neutrino sources if , as widely thought , their @xmath2-ray emission is due to compton - scattered synchrotron ( ssc ) radiation .
|
the excess large scale polarization signal in the wilkinson microwave anisotropy probe ( wmap ) data has allowed the first measurement of the optical depth to re - scattering by electrons with a value of @xmath2 ( kogut et al . 2003 ) .
if interpreted as a sharp transition to a reionized universe from a neutral one , the redshift at which this transition happens is @xmath3 @xcite .
theoretical studies based on cold dark matter ( cdm ) models suggest that such a high reionization redshift beyond what was previously suggested with gunn - peterson troughs in @xmath4 quasars from the sloan survey ( e.g. , fan et al .
2002 ) generally requires significant star - formation early on , at @xmath5 ( cen 2003 ; haiman & holder 2003 ; sokasian et al .
2003 ; yoshida et al .
2003 ) . while conventional models of reionization based on ordinary stellar populations ( population ii stars ) formed in galaxies alone are still consistent with the measured optical depth ( chiu , fan & ostriker 2003 ; somerville & livio 2003 ) , these models are not capable of explaining the high end of the allowed range for the optical depth .
the best fit value of @xmath6 is only achieved in models where reionization takes place with the maximal efficiency in terms of star - formation and associated ionizing photon production ( fukugita & kawasaki 2003 ; haiman & holder 2003 ; cen 2003 ; sokasian et al .
2003 ; wyithe & loeb 2003 ) . alternatively , with a combination of a high photon escape fraction and a top - heavy imf in early galaxies ,
the resulting total optical depth can be increased to be marginally consistent with the wmap data ( e.g. , ciardi , ferrara & white 2003 ) . although recent theoretical studies consistently indicate that the first stars formed out of a chemically pristine gas were rather massive ( abel , bryan & norman 2002 ; bromm , coppi & larson 2002 ) , hence supporting at least the latter condition , it is unclear whether such conditions are realized in primeval galaxies over a sufficiently long period .
it is therefore intriguing to consider alternative mechanisms that do not require considerably high efficiency in star - formation or in ionizing photon production rate .
interestingly , such a scenario is possible even within the standard cosmological models .
for example , as we show later in the present paper , with @xmath7 equal to or larger than unity , the optical depth to reionization can reach within the range of the wmap result since with a higher normalization , halos collapse slightly at earlier times and are more numerous at a given epoch when compared to models with a lower value for the normalization of the matter power spectrum . in light of this , with currently favored values for cosmological parameters , invoking non - gaussianities in primordial density fluctuations may be an interesting possibility and deserves further study . in the non - gaussian models ,
one finds an increased fraction of high density peaks relative to gaussian models , and these peaks are expected to collapse earlier than in the standard case ( peebles 1997 ; white 1998 ; koyama , soda & taruya 1999 ; robinson & baker 2000 ; willick 2000 ; seto 2001 ; mathis 2002 ) .
if star - formation is triggered in these collapsed halos , photons emitted from them may be able to , at least partially , reionize the universe very early on , causing a large total thomson optical depth .
excess non - gaussian fluctuations generally lead to formation of structures at late times and could overproduce the cluster abundance at low redshifts ( willick 2000 )
. however , the constraint obtained from cluster abundance is rather uncertain , and previous studies based on models of primordial non - gaussianities ( e.g. , verde et al . 2001 ) suggest that , in certain models where non - gaussianities are associated with the density field , even a small primordial non - gaussianity can significantly affect the number counts of objects at high redshifts .
avelino & liddle ( 2003 ) recently studied the evolution of the mass fraction in collapsed halos in non - gaussian models and calculated a simple lower estimate for the redshift of reionization . in this _
letter _ , we use a detailed model for the mass function in non - gaussian models and compute the reionization history for some cases .
to this end we compute the total thomson optical depth as a function of redshift .
we conclude that such non - gaussian models , consistent with current constraints , can easily explain wmap s results .
we also argue that , while these models may eventually be constrained with large scale structure observations , there is still some freedom to obtain the necessary non - gaussianity with models which are strongly scale - dependent .
following haiman & holder ( 2003 ) ( hereafter hh ) , we study star formation and reionization using a simple toy model . throughout the paper
we assume the best fit flat @xmath8cdm model with power law fluctuations for the wmap data : @xmath9 . given these cosmological parameters , we calculate the collapsed fraction of baryons using a spherical halo model .
hh distinguished three different types of halos based essentially on the halo mass , which captures different nature of stars formed in them .
the collapsed halo fractions for each halo type is @xmath10 where @xmath11 and @xmath12 take values based on the gas temperature within the halo such that @xmath13 where @xmath14 is the mean molecular weight : in the case of ionized gas with @xmath15 k , @xmath16 and for neutral gas with @xmath17 k , @xmath18 .
the collapse fractions are divided to three halos between 100 k and 10@xmath19 k ( called type ii where molecular hydrogen cooling dominates ) , between 10@xmath19 k and @xmath20 k ( type ia ) , and above @xmath20 k ( type ib ) .
in both type i halos , the dominant cooling mechanism to form stars involves atomic hydrogen lines . using the collapsed mass fractions
, we can now calculate the ionized hii filling factor , which is the same as the mean electron fraction @xmath21 , by integrating for ionization produced by collapse at all higher redshifts , @xmath22\right.\times\\ & & \left.\left [ \epsilon_{\rm ia } \frac{df_{\rm coll}^{\rm ia}}{dz } ( z^{\prime } ) + \epsilon_{\rm ii } \frac{df_{\rm coll}^{\rm ii}}{dz } ( z^{\prime } ) \right ] \right\ } { \tilde v}_{\rm hii}(z^{\prime},z ) .
\label{eq : filling}\end{aligned}\ ] ] the second term on the right hand side includes the factor @xmath23 to allow for photo - ionization feedback , while @xmath24 is the volume per unit mass and unit efficiency ionized by redshift @xmath25 by a single source that turned on at an earlier redshift of @xmath26 . in eq .
( [ eq : filling ] ) , @xmath27 represents the overall efficiency of star - formation and photon emission , where @xmath28 is the fraction of baryons in the halo that turns into stars , @xmath29 the mean number of ionizing photons produced by an atom cycled through stars ( averaged over initial mass function ) , @xmath30 is the fraction of these photons escape to igm . in the following we use fiducial values of @xmath31 for type ia and ib halos as advocated by hh .
we set @xmath32 , i.e. we do not consider star formation in `` minihalos '' in which the gas can cool via molecular hydrogen cooling .
it is straightforward to include type ii halos in our models , but the ionization history involving them is more complicated . for our purpose
test the effect of non - gaussianity it is unnecessary to introduce such additional complexity .
indeed , one aim is to see if early reionization could be obtained without invoking molecular hydrogen cooling .
the evolution of the ionization front is calculated by solving the following differential equation @xmath33 where @xmath34 @xmath35 s@xmath36 is the case b recombination coefficient .
adopting a fixed value of @xmath37 is equivalent to assuming a constant temperature of the ionized region . strictly speaking ,
this is not true , and one needs consider the evolution of the thermal state of the ionized igm .
however , with atomic cooling the gas in the ionized , photo - heated regions quickly approaches this temperature , and as shown by hh this provides a good approximation .
@xmath38 is the clumping factor , which we take to be at a constant value of 10 .
@xmath39 is the expansion rate , with @xmath40^{1/2}$ ] , and @xmath41 is the rate of ionizing photon emission as a function of time .
we take @xmath41 to be a constant for @xmath42 years with a value of 3.7 @xmath43 10@xmath44 s@xmath36 per solar mass of stellar content and decrease from this constant value as @xmath45 thereafter .
we start integration at a sufficiently high redshift , and solve eq .
( [ eq : filling ] ) up to @xmath46 , at which point we consider reionization is completed and keep @xmath47 fixed at 1 for redshift below this ) if integrated past complete reionization , however the result is approximately correct if the integration is limited to before the completion of reionization . ] .
the optical depth is calculated by integrating the ionized volume fraction weighted by the number density such that @xmath48 the factor 1.08 accounts approximately for the contribution of helium reionization ( see hh ) .
the collapsed fraction can be calculated using a standard mass function , such as the press - schechter theory ( ps ; press & schechter 1974 ) based on gaussian density fluctuations , or the numerical simulated mass function by jenkins et al .
( 2001 ) , which was used by hh . to describe the linear power spectrum , we use the transfer function from eisensetin & hu ( 1998 ) .
we now modify the standard mass function by adding non - gaussian fluctuations to the density field . to compare with various available constraints , we make use of a simple model for the non - gaussian mass function which has been previously studied in the literature ( koyama , soda & taruya 1999 ; willick 2000 ) , though we note that , under specific models of primordial non - gaussianity , more detailed mass functions can be constructed ( e.g. , matarrese , verde & jimenez 2000 ) .
we write the generalized form of the mass function as @xmath49 where the probability distribution function for density fluctuation , @xmath50 , is written in terms of a dimensionless function @xmath51 : @xmath52 where @xmath53 , @xmath54 and @xmath55 . here , @xmath56 is the rms mass function on a scale corresponding to a mass @xmath57 , and @xmath58 is the critical density for collapse ( see , e.g. , kitayama & suto 1996 for details ) .
the standard press - schechter formula assumes a gaussian probability distribution function ( pdf ) such that @xmath59 and @xmath60 . to include a mild non - gaussian tail to the gaussian description , essentially a skewness
, we consider a pdf which is based on the poisson description , but modified appropriately for the continuous variables involved ( see , willick 2000 for details ) .
the function is written as @xmath61 such that @xmath62 , which is a function of redshift , is a free - variable that captures the non - gaussian nature of the pdf ; as @xmath62 is increased , the mass function approaches to the ps value for gaussian distribution of density fluctuations .
@xmath63 is chosen to return the correct normalization : @xmath64 .
note that in our non - gaussian mass function , skewness , @xmath65 , scales as @xmath66 , which explains why non - gaussianity increases with decreasing @xmath62 . in terms of the moments of the primordial density field , one can write skewness as @xmath67 where @xmath68 is the third moment of the density perturbations , scaled to a redshift of @xmath25 , when @xmath31 is the parameter related to the quadratic correction ( matarrese , verde & jimenez 2000 ) . if we ignore for now the mass- , or equivalently scale - dependence , of primordial non - gaussianity , then @xmath69 , and assuming linear growth of fluctuations , @xmath70 where @xmath71 is the growth factor .
the latter scaling can be understood based on the fact that @xmath72 and @xmath73 scale with redshift as @xmath74 and @xmath75 , respectively . since in our mass function , @xmath76
, we expect @xmath62 to behave as @xmath77 as a function of redshift . in fig . 1 , we plot the collapsed fraction of these halos for the standard ps mass function and the non - gaussian mass functions with @xmath78 and @xmath79 as a function of redshift .
we note that these values could be a function of redshift , although we have taken @xmath62 to be a constant , for purposes of the present discussion . as @xmath62 is decreased , or models become strongly non - gaussian , the fraction of mass in collapsed halos increases at a fixed redshift .
this illustrates the basic idea behind our calculation that with increasing non - gaussianity of the density field , there are more high overdensity peaks which collapse earlier .
these early collapsing halos can form stars at an earlier time than in the standard description and reionize the universe such that the total optical depth is higher than previously estimated in the gaussian models . in fig . 2 , we show the ionization fraction and optical depth for the models with the ps mass function and with the non - gaussian mass function with @xmath78 and @xmath79 .
in addition to our standard cosmological values , to illustrate the increase in the optical depth by halos collapsing at an earlier time , we also calculate the optical depth in a model where @xmath80 ( fig.2 , right plot ) . we emphasize that such high values of @xmath7 easily make the optical depth large even within the gaussian model . for a given set of cosmological parameters ,
higher optical depths can be achieved with the non - gaussian mass functions . as shown in fig . 2 , compared with the standard gaussian ps function
, there is only a moderate increase in optical depth for the @xmath78 model ( about 0.01 in @xmath81 ) .
this difference is in fact comparable in size with the difference between the ps and jenkins mass functions : it is well known that compared with the ps mass function , the jenkins et al .
( 2001 ) function predicts an excess of halos at the upper end of the halo mass . our calculated optical depth with the jenkins et al .
( 2001 ) mass function of @xmath82 0.13 is consistent with previous calculations on the optical depth based on pop ii stars alone ( e.g. , hh ; fukugita & kawasaki 2003 ) . to increase the optical depth beyond this value
, one can either add more population iii stars at earlier epochs ( cen 2003 ; wythie & loeb 2003 ; sokasian et al .
2003 ) or , as we have considered , by considering smaller @xmath62 value ( greater non - gaussianity ) while resorting only ordinary stellar populations . a significantly non - gaussian model with @xmath83 increases the optical depth by a large fraction ( about 0.04 in @xmath81 ) .
from eq .
[ eqn : skew ] and the expected scale with redshift , such a non - gaussianity implies a primordial non - gaussianity with @xmath84 .
non - gaussianities at this level can soon be tested with observations of galaxy counts at redshifts of 5 , though low redshift large scale structure observations will not be helpful for this purpose ( verde et al .
2001 ) . at present , according to willick ( 2000 ) ( see his fig .
11 ) , for @xmath85 , the low redshift galaxy cluster abundance is consistent at the 90% confidence level with a non - gaussian mass function of @xmath86 .
the cluster abundance at low - redshift is not a strong discriminator of non - gaussian models , as argued by verde et al .
note that only a constraint on the upper limit of @xmath62 is available from the cluster abundance in willick ( 2000 ) , though in koyama et al .
( 1999 ; also , robinson , gawiser & silk 1998 ) @xmath87 is preferred for @xmath8cdm cosmological models .
again this is not a strong constraint of non - gaussianity at redshifts of order 30 .
apart from low and moderate redshifts observations , non - gaussianity can also be constrained , at a redshift of 1100 , from observations of cosmic microwave background anisotropies ( e.g. , komatsu et al .
current constraints indicate a limit of 134 on @xmath88 , at the 95% confidence level ; this parameter is related to the quadratic corrections to potential field and is different from @xmath31 we have been using to describe primordial non - gaussianity based on corrections to the density field directly . through the poisson function
, one can relate the two parameters such that at large scales of order 100 mpc , roughly , @xmath89 , and the upper limit on @xmath88 of 134 from wmap corresponds to @xmath90 and allows for significant non - gaussianity , rendering a lower limit on @xmath62 at a redshift @xmath82 30 of 0.08 .
while cluster counts from low redshifts have constrained the upper limit and cmb observations have constrained the lower limit , the allowed range is still significant enough to produce substantial non - gaussian fluctuations and to easily reproduce the observed optical depth with early formation of structures and first stars .
note that with completed wmap results and with planck , the limits on non - gaussianity will be improved and , naively , small values for @xmath62 at @xmath91 may be ruled out .
such a limit , however , can not fully constrain non - gaussianities , since one might still circumvent this by considering scale - dependent non - gaussian models . for example
, one can apply the significant non - gaussianity to scales which are related to star - formation at redshifts above 10 and still allow such non - gaussianity to remain today .
it will then affect the number counts of @xmath92 m@xmath93 halos at low redshifts , but cmb anisotropies at large scales will remain nearly gaussian .
the same models can also be tuned such that cluster counts at low redshifts remain gaussian also .
we note that it is not difficult to envision how such models can be constructed with examples including non - standard inflationary models ( wang & kamionkowski 2000 ; bartolo , matarrese & riotto 2002 ; bernardeau & uzan2003 ) and models with primordial black hole formation proposed by bullock & primack ( 1997 ) .
while constructing a concrete model for producing such scale - dependent non - gaussian primordial fluctuations is beyond the scope of the present study , we argue that these models can also be constrained eventually . the best way to probe the level of non - gaussianity suggested above , and at scales related to this problem , may be to study number counts of galaxies at the highest redshifts possible ( verde et al .
another possibility involves a measurement of the non - gaussianity in the surface brightness fluctuations produced by first stars and proto - galaxies in the near - infrared background and potentially detectable with upcoming deep observations from near - ir imaging camera on the astro - f mission ( cooray et al .
2003 ) . studying these constraints on the presence of non - gaussianities at highest redshifts related to formation of first objects
is clearly needed to understand the extent to which the reionization process is affected .
we have examined how the observed high optical depth is achieved in cosmological models with non - gaussian primordial density fluctuations . in these models ,
high over - density peaks due to the non - gaussianity collapse early on and may harbor first stars at redshifts above what is generally expected in the standard gaussian model . for mild non - gaussian fluctuations at redshifts of 30 ,
the increase in optical depth is at the level of 0.01 , and it is unlikely to explain the observed value of optical depth by this mechanism alone . to obtain significant increase in the optical depth , the first collapsing halos associated with the primordial non - gaussianity must contribute to reionization significantly .
a substantial volume fraction of the igm may be reionized if a number of population iii stars are formed in these early halos .
we found that increasing the degree of non - gaussianities can easily reproduce the high optical depth .
such models are still viable within constraints from both cluster number counts at low redshifts and cmb anisotropies at a redshift of 1100
. while constraints from these two redshift ends will improve , models with scale - dependent non - gaussianity are possible to explain the optical depth .
such models will only be constrained by probes of non - gaussianity related to mass scales of 10@xmath94 m@xmath95 at redshifts of 30 and will involve , for example , study of first galaxy counts at very high redshifts . the ongoing operation of wmap will pin down a presice value for the total optical depth . for a longer - term
, post - wmap cmb polarazation experiments such as planck will probe the reionization history ( kaplinghat et al .
2003 ) , and detection of the second - order polarization anisotropies can place additional constraints on details of the reionization process ( liu et al . 2001 ) .
data from these future observations will improve our picture of reionization and will enable us to further distinguish theoretical models related to the first generation of luminous structure formation .
we thank organizers of the us - japan workshop on the sz effect , where this paper was conceived , for a wonderful conference . this work is supported by the sherman fairchild foundation and doe de - fg 03 - 92-er40701 ( at caltech ) and by nsf phy99 - 07949 ( at kitp ) .
ny acknowledges support from japanese society of promotion of science special research fellowship ( grant 2674 ) .
ns is supported by japanese grant - in - aid for science research fund of the ministry of education , no.14340290 .
|
the first - year wilkinson microwave anisotropy probe data suggest a high optical depth for thomson scattering of 0.17 @xmath0 0.04 , implying that the universe was reionized at an early epoch , @xmath1 .
such early reionization is likely to be caused by uv photons from first stars , but it appears that the observed high optical depth can be reconciled within the standard structure formation model only if star - formation in the early universe was extremely efficient . with normal star - formation efficiencies , cosmological models with non - gaussian density fluctuations may circumvent this conflict as high density peaks collapse at an earlier epoch than in models with gaussian fluctuations .
we study cosmic reionization in non - gaussian models and explore to what extent , within available constraints , non - gaussianities affect the reionization history . for mild non - gaussian fluctuations at redshifts of 30 to 50 ,
the increase in optical depth remains at a level of a few percent and appears unlikely to aid significantly in explaining the measured high optical depth . on the other hand , within available observational constraints , increasing the non - gaussian nature of density fluctuations can easily reproduce the optical depth and may remain viable in underlying models of non - gaussianity with a scale - dependence .
|
the discovery of dirac fermions as low - energy quasi - particle excitations in graphene and on the surfaces of topological insulators has drawn significant attention in both fundamental physics research and device applications @xcite .
since these dirac - like fermions propagate as massless relativistic particles they behave differently from the conventional charge carriers in metals , semiconductors , and insulators . recently
, a new form of massless fermions with lifted degeneracy at the nodal point , from four - fold to two - fold , has been proposed to exist in condensed matter systems through a time reserval or inversion symmetry breaking mechanism .
this symmetry breaking operation modifies the quasi - particle s dispersion relation from a dirac to a weyl equation.@xcite many theoretical proposals exist for realizing weyl semimetals that possess interesting physical properties , such as discontinuous fermi arcs and negative magnetoresistance due to the chiral anomaly.@xcite a weyl node with definite chirality is associated with the berry curvature and may be thought of as realizations of magnetic monopoles in momentum space.@xcite recent theoretical works have proposed the realization of the weyl semimetal state in the inversion symmetry breaking taas family.@xcite shortly after the prediction , the first weyl semimetal was experimentally discovered in taas @xcite .
the electronic weyl semimetal state in taas was experimentally observed through photoemission spectroscopy@xcite .
electrical transport experiments have shown that taas has very high mobility @xcite and , therefore , is consistent with the protected nature of the weyl fermions and reported signatures of the chiral anomaly @xcite .
soon after the initial experimental discovery other independent photoemission experimental works have confirmed the weyl semimetal state in taas and , furthermore , the weyl state in other members of the same family , which includes nbas and tap .
@xcite provided the inclusion of spin - orbit coupling , these four non - centrosymmetric compounds are proposed to realize weyl fermions without breaking time - reversal symmetry and , consequently , are easily studied under ambient conditions . for example , no external magnetic field or pressure is required for the realization of the weyl semimetal state in this taas family.@xcite conventional semimetals are materials that have a small overlap in energy between the valence and conduction bands . by contrast , a weyl semimetal has valence and conduction bands that touch at a set of discrete points in the bulk brillouin zone , called the weyl nodes . in the presence of additional doping , fermi pockets , or multiple weyl nodes at different energies , a weyl semimetal will naturally contain electron and hole charge carriers , giving rise to a compensated semimetal .
this is indeed the case for the taas family of weyl materials as reported in photoemission and transport experiments @xcite the fermi surfaces can originate from the conduction or valence band crossing the fermi energy , the weyl cones , and their combinations .
the charge carriers are expected to have small concentrations with a high mobility , at least for the weyl fermions .
the interplay between tunable concentrations of multiple charge carriers , high fermi velocities , and the protected nature of the weyl fermion carriers suggests a new avenue for designing novel electronic devices .
the weyl semimetals are also good candidates for exploring critical phenomena near the fermi energy , due to their small carrier concentrations and linear band touchings . therefore , it is of interest to explore the electronic structure of these four compounds to better understand their similarities and differences .
previous theoretical studies of this family focused on establishing taas as a weyl semimetal and therefore only reported a few key aspects of the electronic structure for that purpose . on the other hand , many systematic details of the band structure , which are crucial for spectroscopy and transport studies , have not been studied . furthermore , a comparative study of the electronic structure of taas with the other members remains lacking . in this study
, we perform a comprehensive first - principles calculation to explore the properties of fermi surfaces for the entire family of weyl semimetals , taas , tap , nbas , and nbp , and , thereby , providing the necessary building block for future investigations of weyl semimetals from a theoretical , experimental and application perspective .
the paper is organized in the following manner : in section [ sec : computational ] we provide the computational details . in section [ sec : bandstructure ] ,
electronic band structures and density of states are studied to improve our basic understanding of these four compounds . in section [ sec : fermisurface ] , we discuss the details of the fermi velocities of weyl fermions and fermi surfaces . finally , in section
[ sec : discussion ] we present a brief discussion covering the important conclusions reached .
first - principles calculations of taas , tap , nbas , and nbp , which have space group @xmath0 ( 109 ) , a body - centered - tetragonal structure , were performed using the openmx code . the openmx code was based on norm - conserving pseudopotentials generated with multi reference energies@xcite and optimized pseudoatomic basis functions.@xcite for each ta atom , three , two , two , and one optimized radial functions were allocated for the @xmath1- , @xmath2- , @xmath3- , and @xmath4-orbitals ( @xmath5 ) , respectively .
for as , nb , and p atoms , @xmath6 , @xmath7 , and @xmath8 were adopted , respectively . a cut - off radius of 7 bohr was adopted for ta , nb , and p basis functions while a 9 bohr cut - off radius was used for the as basis functions .
the spin - orbit coupling was taken into account through @xmath9-dependent pseudopotentials,@xcite and the generalized gradient approximation ( gga ) was adopted for the exchange - correlation energy functional.@xcite the cut - off energy of 1000 ry was used for numerical integrations and for the solution of the poisson equation . for the quality of @xmath10-point sampling , a @xmath11 mesh for the conventional unit cell was adopted .
the experimental lattice parameters were chosen in the calculations and are listed in table [ table : lattice ] . the conventional unit cell and first brillouin zone of the primitive unit cell of the representative taas
is plotted in fig .
[ fig : structure ] . throughout this paper , for brevity , we will label the crystal momentum in units of @xmath12 unless otherwise specified .
.experimental lattice parameters ( @xmath13 fig .
fig : structure ) . [ cols="^,^,^,^,^",options="header " , ] the fermi surfaces of nbp are presented in fig .
[ fig : nbpfermisurface ] . similar to tap and nbas ,
the energy of @xmath14 and @xmath15 in nbas are below and above the fermi energy , respectively .
the two concentric - like hole fermi surfaces split by spin - orbit coupling are also present in nbp as intersecting the the @xmath16=0 plane at @xmath17(0 , 0.2042 , 0 ) , giving rise to four crescent fermi surfaces in the first brillouin zone with each surface containing one more surface inside .
each outer surface encloses one pair of @xmath15 points . in nbp ,
one long crescent electron - like fermi surface resembling the shape of half a ring can be found . the surface becomes thinner while deviating from the @xmath14 and the detailed shape is easier to be observed in fig .
[ fig : nbpfermisurface](c ) .
there are two additional electron - like fermi surface inside this long crescent surface , @xmath18 and @xmath19 .
referring to the band dispersion in fig .
[ fig : w1dispersion ] , we have to carefully identify the concentric electron pockets .
although @xmath18 encloses one w1 point , the associated band does not evolve from the band of weyl nodes and , therefore , is a trivial fermi surface .
the size of the surface @xmath19 is small and touches the limit for the chosen grid size
. the carrier concentration and maximal areas of cross sections for each fermi surface is listed in table [ table : concentration ] .
several interesting properties of the charge carriers can be found even without the consideration of additional conditions , like applying magnetic or strain fields in first - principles calculations .
we first note that the energy difference between @xmath14 and @xmath15 is @xmath20 mev , which is realistic to access by doping in real crystals .
the fermi energy for taas was experimentally found @xcite to be located between @xmath14 and @xmath15 ( @xmath21 and @xmath22 , respectively ) , which is approximately 11.5 mev above @xmath21 and , therefore , different from our calculations shown in table [ table : weylpoints ] .
the energy difference between two weyl nodes is directly related to the minimal area of the fermi surface as a function of binding energy , with a larger separation , @xmath23 , corresponding to a larger minimal fermi surface . referring to table [ table : weylpoints ] ,
the energy differences of @xmath22 and @xmath21 are 0.0132 , 0.0727 , 0.0364 , and 0.0793 ev for taas , tap , nbas , and nbp , respectively .
recall , @xmath24 .
note that while the separation in momentum space of the weyl nodes is determined by the strength of the spin - orbit coupling , the separation in energy of the weyl nodes can be viewed as arising from the dispersion of the spinless line node crossing , observed before spin - orbit coupling is introduced . amongst the four weyl semimetal compounds ,
the relatively small value for taas explains the relatively small carrier concentration , e.g. 0.134@xmath25 for @xmath26 .
similarly , the largest hole pocket can be found in nbp , e.g. 82.16@xmath25 for @xmath26 .
the rich and tunable behavior of charge carriers in these four semimetals strongly suggests that these compounds are a good playground for engineering the electronic properties through doping , strain , electric fields , magnetic fields and/or other external perturbations .
furthermore , we also point out that an abrupt change in the fermi surface volume may occur by tuning , for example , the fermi energy .
since each pair of weyl nodes are located on two sides of the mirror plane , after turning on the spin - orbit coupling , they are , in general , close to each other
. therefore , there exists a critical fermi energy that can make the two fermi surfaces , each enclosing the weyl node(s ) on one side of the plane , touch at one @xmath10 point . by changing the critical fermi energy , either one big surface enclosing both weyl nodes or two small surfaces with each enclosing a single weyl node can be formed .
the fermi surfaces of the @xmath14 or @xmath15 pairs in taas is an example for the latter case while those in tap , nbas , and nbp are examples of the former case .
conyers herring , accidental degeneracy in the energy bands of crystals , _ physical review _ * 52 * , 365 - 373 ( 1937 ) .
a. a. abrikosov , s. d. beneslavskii , some properties of gapless semiconductors of the second kind , _ journal of low temperature physics _ * 5 * , 141 - 154 ( 1971 ) .
holger bech nielsen , masao ninomiya , the adler - bell - jackiw anomaly and weyl fermions in a crystal , _
physics letters b _ , * 130 * , 389 - 396 ( 1983 ) .
xiangang wan , ari m. turner , ashvin vishwanath , sergey y. savrasov , topological semimetal and fermi - arc surface states in the electronic structure of pyrochlore iridates , _ phys .
b _ * 83 * , 205101 ( 2011 ) .
shin - ming huang , su - yang xu , ilya belopolski , chi - cheng lee , guoqing chang , baokai wang , nasser alidoust , guang bian , madhab neupane , chenglong zhang , shuang jia , arun bansil , hsin lin , m. zahid hasan , a weyl fermion semimetal with surface fermi arcs in the transition metal monopnictide taas class , _ nature communications _ * 6 * , 7373 ( 2015 ) .
su - yang xu , ilya belopolski , nasser alidoust , madha neupane , guang bian , chenglong zhang , raman sankar , guoqing chang , zhujun yuan , chi - cheng lee , shin - ming huang , hao zheng , jie ma , daniel s. sanchez , baokai wang , arun bansil , fangcheng chou , pavel p. shibayev , hsin lin , shuang jia , m. zahid hasan , discovery of a weyl fermion semimetal and topological fermi arcs , _ science _ * 349 * , 613 - 617 ( 2015 ) . hongming weng , chen fang , zhong fang , b. andrei bernevig , xi dai , weyl semimetal phase in noncentrosymmetric transition - metal monophosphides , _ phys . rev .
x _ * 5 * , 011029 ( 2015 ) .
b. q. lv , h. m. weng , b. b. fu , x. p. wang , h. miao , j. ma , p. richard , x. c. huang , l. x. zhao , g. f. chen , z. fang , x. dai , t. qian , h. ding , experimental discovery of weyl semimetal taas , _ phys .
x _ * 5 * , 031013 ( 2015 ) .
a. h. castro neto , f. guinea , n. m. r. peres , k. s. novoselov , a. k. geim , the electronic properties of graphene , _ rev .
_ * 81 * , 109 - 162 ( 2009 ) .
s. das sarma , shaffique adam , e. h. hwang , enrico rossi , electronic transport in two - dimensional graphene , _ rev .
phys . _ * 83 * , 407 - 470 ( 2011 ) .
m. z. hasan , c. l. kane , _ colloquium _ : topological insulators , _ rev .
phys . _ * 82 * , 3045 - 3067 ( 2010 ) .
xiao - liang qi , shou - cheng zhang , topological insulators and superconductors , _ rev .
phys . _ * 83 * , 1057 - 1110 ( 2011 ) .
f. d. m. haldane , attachment of surface
fermi arcs to the bulk fermi surface : fermi - level plumbing in topological metals .
http://arxiv.org/abs/1401.0529 ( 2014 ) .
bahadur singh , ashutosh sharma , h. lin , m. z. hasan , r. prasad , a. bansil , topological electronic structure and weyl semimetal in the tlbise@xmath28 class of semiconductors , _ phys .
b _ * 86 * , 115208 ( 2012 ) .
a. m. turner , a. vishwanath , beyond band insulators : topology of semi - metals and interacting phases . http://arxiv.org/abs/1301.0330 ( 2013 ) .
gang xu , hongming weng , zhijun wang , xi dai , zhong fang , chern semimetal and the quantized anomalous hall effect in @xmath29 , _ phys . rev .
lett . _ * 107 * , 186806 ( 2011 ) .
daniel bulmash , chao - xing liu , xiao - liang qi , prediction of a weyl semimetal in @xmath30@xmath31@xmath32te , _ phys .
b _ * 89 * , 081106 ( 2014 )
. jianpeng liu , david vanderbilt , weyl semimetals from noncentrosymmetric topological insulators , _ phys .
b _ * 90 * , 155316 ( 2014 ) .
pavan hosur , s. a. parameswaran , ashvin vishwanath , charge transport in weyl semimetals , _ phys .
lett . _ * 108 * , 046602 ( 2012 ) .
heon - jung kim , ki - seok kim , j .- f .
wang , m. sasaki , n. satoh , a. ohnishi , m. kitaura , m. yang , l. li , dirac versus weyl fermions in topological insulators : adler - bell - jackiw anomaly in transport phenomena , _ phys .
lett . _ * 111 * , 246603 ( 2013 ) .
s. a. parameswaran , t. grover , d. a. abanin , d. a. pesin , a. vishwanath , probing the chiral anomaly with nonlocal transport in three - dimensional topological semimetals , _ phys .
x _ * 4 * , 031035 ( 2014 ) .
pavan hosur , xiaoliang qi , recent developments in transport phenomena in weyl semimetals , _ comptes rendus physique _ * 14 * , 857 - 870 ( 2013 ) . andrew c potter , itamar kimchi , ashvin vishwanath , quantum oscillations from surface fermi arcs in weyl and dirac semimetals , _ nature communications _ * 5 * , 5161 ( 2014 ) .
zhong fang , naoto nagaosa , kei s. takahashi , atsushi asamitsu , roland mathieu , takeshi ogasawara , hiroyuki yamada , masashi kawasaki , yoshinori tokura , kiyoyuki terakura , the anomalous hall effect and magnetic monopoles in momentum space , _ science _ * 302 * , 92 - 95 ( 2003 ) .
gbor b. halsz , leon balents , time - reversal invariant realization of the weyl semimetal phase , _ phys .
b _ * 85 * , 035103 ( 2012 )
. a. a. zyuzin , si wu , a. a. burkov , weyl semimetal with broken time reversal and inversion symmetries , _ phys .
b _ * 85 * , 165110 ( 2012 ) .
tanmoy das , weyl semimetal and superconductor designed in an orbital - selective superlattice , _ phys .
b _ * 88 * , 035444 ( 2013 ) .
teemu ojanen , helical fermi arcs and surface states in time - reversal invariant weyl semimetals , _ phys .
b _ * 87 * , 245112 ( 2013 ) . x. huang , l. zhao , y. long , p. wang , d. chen , z. yang , h. liang , m. xue , h. weng , z. fang , x. dai , g. chen , observation of the chiral anomaly induced negative magneto - resistance in 3d weyl semi - metal taas .
http://arxiv.org/abs/1503.01304 ( 2015 ) .
c. zhang , s .- y .
xu , i. belopolski , z. yuan , z. lin , b. tong , n. alidoust , c .- c .
huang , h. lin , m. neupane , d. s. sanchez , h. zheng , g. bian , j. wang , c. zhang , t. neupert , m. zahid hasan , s. jia , observation of the adler - bell - jackiw chiral anomaly in a weyl semimetal .
http://arxiv.org/abs/1503.02630 ( 2015 ) .
xu , n. alidoust , i. belopolski , c. zhang , g. bian , t .-
chang , h. zheng , v. strokov , d. s. sanchez , g. chang , z. yuan , d. mou , y. wu , l. huang , c .- c .
huang , b. wang , a. bansil , h .-
jeng , t. neupert , a. kaminski , h. lin , s. jia , and m. zahid hasan , discovery of a weyl fermion state with fermi arcs in niobium arsenide .
_ nature physics _ doi:10.1038/nphys3437 ( 2015 ) .
xu , i. belopolski , d. s. sanchez , c. guo , g. chang , c. zhang , g. bian , z. yuan , h. lu , y. feng , t .-
chang , p. p. shibayev , m. l. prokopovych , n. alidoust , h. zheng , c .- c .
huang , r. sankar , f. chou , c .- h .
jeng , a. bansil , t. neupert , v. n. strocov , h. lin , s. jia , m. zahid hasan , experimental discovery of a topological weyl semimetal state in tap .
http://arxiv.org/abs/1508.03102 ( 2015 ) .
b. q. lv , n. xu , h. m. weng , j. z. ma , p. richard , x. c. huang , l. x. zhao , g. f. chen , c. matt , f. bisti , v. strokov , j. mesot , z. fang , x. dai , t. qian , m. shi , h. ding , observation of weyl nodes in taas .
_ nature physics _ doi:10.1038/nphys3426 ( 2015 ) .
n. xu , h. m. weng , b. q. lv , c. matt , j. park , f. bisti , v. n. strocov , d. gawryluk , e. pomjakushina , k. conder , n. c. plumb , m. radovic , g. auts , o. v. yazyev , z. fang , x. dai , g. aeppli , t. qian , j. mesot , h. ding , m. shi , observation of weyl nodes and fermi arcs in tap . http://arxiv.org/abs/1507.03983 ( 2015 ) .
l. yang , z. liu , y , sun , h. peng , h. yang , t. zhang , b. zhou , y. zhang , y. guo , m. rahn , d. prabhakaran , z. hussain , s .- k .
mo , c. felser , b. yan , y. chen , discovery of a weyl semimetal in non - centrosymmetric compound taas , _ nature physics _ doi:10.1038/nphys3425 ( 2015 ) .
nj ghimire , yongkang luo , m neupane , dj williams , ed bauer , f ronning , magnetotransport of single crystalline nbas , _ journal of physics : condensed matter _ * 27 * , 152201 ( 2015 ) .
chandra shekhar , ajaya k. nayak , yan sun , marcus schmidt , michael nicklas , inge leermakers , uli zeitler , yurii skourski , jochen wosnitza , zhongkai liu , yulin chen , walter schnelle , horst borrmann , yuri grin , claudia felser , binghai yan , extremely large magnetoresistance and ultrahigh mobility in the topological weyl semimetal candidate nbp , _ nature physics _ * 11 * , 645 - 649 ( 2015 ) .
y. luo , n. j. ghimire , m. wartenbe , m. neupane , r. d. mcdonald , e. d. bauer , j. d. thompson , f. ronning , a novel electron - hole compensation effect in nbas . http://arxiv.org/abs/1506.01751 ( 2015 ) .
z. wang , y. zheng , z. shen , y. zhou , x. yang , y. li , c. feng , z .- a .
xu , helicity protected ultrahigh mobility weyl fermions in nbp .
http://arxiv.org/abs/1506.00924 ( 2015 ) .
c. shekhar , f. arnold , s .- c .
wu , y. sun , m. schmidt , n. kumar , a. g. grushin , j. h. bardarson , r. donizeth dos reis , m. naumann , m. baenitz , h. borrmann , m. nicklas , e. hassinger , c. felser , b. yan , large and unsaturated negative magnetoresistance induced by the chiral anomaly in the weyl semimetal tap .
http://arxiv.org/abs/1506.06577 ( 2015 ) .
p. j. w. moll , a. c. potter , b. ramshaw , k. modic , s. riggs , b. zeng , n. j. ghimire , e. d. bauer , r. kealhofer , n. nair , f. ronning , j. g. analytis , magnetic torque anomaly in the quantum limit of the weyl semi - metal nbas .
http://arxiv.org/abs/1507.06981 ( 2015 ) .
chenglong zhang , cheng guo , hong lu , xiao zhang , zhujun yuan , ziquan lin , junfeng wang , and shuang jia , large magnetoresistance over an extended temperature regime in monophosphides of tantalum and niobium , _ phys .
b _ * * 9**2 , 041203 ( 2015 ) . c. zhang , z. lin , c. guo , s .- y .
xu , c .- c .
lee , h. lu , s .-
huang , g. chang , c .- h .
hsu , h. lin , l. li , c. zhang , t. neupert , m. zahid hasan , j. wang , s. jia , quantum phase transitions in weyl semimetal tantalum monophosphide . http://arxiv.org/abs/1507.06301 ( 2015 ) .
j. du , h. wang , q. mao , r. khan , b. xu , y. zhou , y. zhang , j. yang , b. chen , c. feng , m. fang , unsaturated both large positive and negative magnetoresistance in weyl semimetal tap .
http://arxiv.org/abs/1507.05246 ( 2015 ) .
i. morrison and d.m .
bylander and l. kleinman , _ phys .
b _ `` 47 '' , 6728 ( 1993 ) . t. ozaki , _ phys
b _ , * 67 * , 155108 ( 2003 ) . t. ozaki _ et .
http://www.openmx - square.org/. gerhard theurich , nicola a. hill , self - consistent treatment of spin - orbit coupling in solids using relativistic fully separable _ ab initio _ pseudopotentials , _ phys .
b _ * 64 * , 073106 ( 2001 ) .
w. kohn and l. j. sham , _ phys .
rev . _ * 140 * , a1133 ( 1965 ) .
j. p. perdew and k. burke and m. ernzerhof , _ phys .
lett . _ * 77 * , 3865 ( 1996 ) .
jian - hui zhou and hua jiang and qian niu and jun - ren shi , topological invariants of metals and the related physical effects , _ chinese physics letters _ * 30 * , 027101 ( 2013 ) .
ching - kai chiu , andreas p schnyder , classification of reflection - symmetry - protected topological semimetals and nodal superconductors , _
physical review b _ * 90 * , 205136 ( 2014 ) .
d. gosalbez - martinez , i. souza , d. vanderbilt , chiral degeneracies and fermi - surface chern numbers in bcc fe .
http://arxiv.org/abs/1505.07727 ( 2015 ) .
|
the family of binary compounds including taas , tap , nbas , and nbp was recently discovered as the first realization of weyl semimetals . in order to develop a comprehensive description of the charge carriers in these weyl semimetals
, we performed a detailed and systematic electronic band structure calculations which reveal the nature of fermi surfaces and their complex interconnectivity in taas , tap , nbas , and nbp .
our work report the first comparative and comprehensive study of fermi surface topology and band structure details of all known members of the weyl semimetal family and provide the necessary building blocks for advancing our understanding of their unique topologically protected low - energy weyl fermion physics .
|
let @xmath2 be an ordered alphabet of size @xmath1 and let a word be made of the random letters @xmath3 ( independently ) drawn from @xmath4 . recall that the robinson schensted
knuth ( rsk ) correspondence associates to a ( random ) word a pair of ( random ) young tableaux of the same shape , having at most @xmath1 rows ( see , e.g. , @xcite or @xcite ) .
it is then well known that the length , @xmath5 , of the top row of these tableaux coincides with the length of the longest ( weakly ) increasing subsequence of @xmath6 .
the behavior of @xmath5 when @xmath0 and/or @xmath1 go to @xmath7 and its connections to various areas of mathematics ( e.g. , random matrices , queueing theory , percolation theory ) have been investigated in numerous papers ( @xcite , @xmath8 ) .
for instance , appropriately renormalized and for uniform draws , @xmath5 converges in law , as @xmath0 goes to infinity and @xmath1 is fixed , to the largest eigenvalue of an @xmath9 matrix from the traceless gaussian unitary ensemble ( gue ) .
more generally ( see @xcite ) , when @xmath10 ( and @xmath1 is fixed ) , the shape of the whole young tableaux associated to a uniform random word converges , after renormalization , to the law of the spectrum of an @xmath9 traceless gue matrix . for different random words , such as non - uniform or markovian ones ,
the situation is more involved ( @xcite ) . for independently and uniformly drawn random words , the following result holds , where , below and in the sequel , `` @xmath11 '' stands for convergence in distribution .
[ theo : hl3 ] let @xmath12 be the sum of the lengths @xmath13 of the first @xmath14 rows of the young tableau .
then @xmath15 where @xmath16 is a multidimensional brownian motion with covariance matrix having diagonal terms equal to 1 and off - diagonal terms equal to @xmath17 , and where @xmath18 is defined by @xmath19 here , and in the sequel , the rows beyond the height of the tableau are considered to be of length zero .
if we let @xmath20 be defined via @xmath21 , @xmath22 , then the shape of the young tableau is given by @xmath23 .
moreover , let @xmath24 be the spectrum , written in non - increasing order , of an @xmath9 traceless element of the gue , where the gue is equipped with the measure @xmath25 and @xmath26 ( see @xcite ) . an important fact
( see @xcite ) asserts that @xmath27 \\[-8pt ] \nonumber & & \quad\stackrel{\mathcal{l}}{= } ( \lambda_{{\mathrm{gue}},m}^{1,0},\lambda_{{\mathrm{gue}},m}^{2,0 } , \ldots , \lambda _ { { \mathrm{gue}},m}^{m,0}).\end{aligned}\ ] ] in fact , if @xmath28 is the ( ordered ) spectrum of an @xmath9 element of the gue , then @xmath29 where @xmath30 is a centered gaussian random variable with variance @xmath31 , independent of the vector @xmath32 , and where @xmath33 ; see @xcite for simple proofs of ( [ eq : spectrum ] ) and ( [ eq : gue0 ] ) . finally , recall that , as @xmath34 , the asymptotic behavior of the spectrum of the gue has been obtained by tracy and widom ( see @xcite , @xcite and also theorem 1.4 in @xcite , with a slight change of the notation ) .
[ theo : johansson ] for each @xmath35 , there is a distribution @xmath36 on @xmath37 such that @xmath38 [ rem : johansson - tw ] the distribution @xmath36 is explicitly known ( see ( 3.48 ) in @xcite ) and its first marginal coincides with the tracy
widom distribution . since @xmath39 as @xmath34 , taking successively the limits in @xmath0 and then in @xmath1 , ( [ eq : hl3])([eq : johansson ] ) entail , for each @xmath35 , that @xmath40 following universality argument in percolation models developed by bodineau and martin ( @xcite ) , we show below that the limits in @xmath0 and @xmath1 in ( [ eq : nm2 ] ) can be explicitly taken simultaneously when the size @xmath1 of the alphabet does not grow too fast with respect to @xmath0 . doing so , we are dealing with growing ordered alphabets and , at each step , the @xmath0 letters @xmath41 , @xmath42 , are redrawn ( and not just the @xmath0th letter as in the case with the model studied in @xcite ) . in a way
, we are thus giving the fluctuations of the shape of the young tableau of a random word when the alphabets are growing and are reshuffled . in the sequel , @xmath1 will be a function @xmath43 of @xmath0 .
however , in order to simplify the notation , we shall still write @xmath1 instead of @xmath43 . a main result of this note is the following .
[ theo : main ] let @xmath1 tend to infinity as @xmath10 in such a way that @xmath44 .
then , for each @xmath35 , @xmath45 remark [ rem : condition ] , below , briefly discusses the growth conditions on @xmath1 . since , again , the length of the first row of the young tableau is the length @xmath5 of the longest increasing subsequence and since the first marginal of @xmath46 is the tracy
widom distribution @xmath47 , we have the following result .
[ corol : tw ] let @xmath1 tend to infinity as @xmath10 in such a way that @xmath44 .
then @xmath48 when the independent random letters are no longer uniformly drawn , a similar asymptotic behavior continues to hold for @xmath5 , as explained next .
let the @xmath41 , @xmath42 , be independently and identically distributed with @xmath49 , let @xmath50 and also let @xmath51 with @xmath52 .
now , from @xcite and as @xmath10 , the behavior of the first row of the young tableau in this non - uniform setting is given by @xmath53 \\[-8pt ] \nonumber & & { } + \mathop{\max_{0=t_0\leq t_1\leq\cdots}}_{\leq t_{k(m)-1}\leq t_{k(m)}=1 } \sum_{l=1}^{k(m)}\bigl(b^l(t_l)-b^l(t_{l-1})\bigr),\end{aligned}\ ] ] where @xmath54 is a standard @xmath55-dimensional brownian motion .
for the limiting behavior in @xmath1 of the right - hand side of ( [ eq : nonunif1 ] ) , as explained next , two cases can arise , depending on the number of most probable letters in @xmath56 .
setting @xmath57 and combining ( [ eq : spectrum ] ) , ( [ eq : gue0 ] ) and ( [ eq : johansson ] ) , as well as remark [ rem : johansson - tw ] , when @xmath58 , and since , clearly , @xmath59 , we have @xmath60 first , let @xmath55 be bounded . eventually extracting a subsequence
, we can assume that @xmath55 is equal to a fixed @xmath61 and since @xmath62 $ ] , we can also assume that @xmath63 . in this case , taking the limit first in @xmath0 and next in @xmath1 yields @xmath64 the limiting distribution on the right - hand side of ( [ eq : lim_2 ] ) depends on @xmath14 . for instance , for @xmath58
, we recover a gaussian distribution , while for @xmath65 and specific choice of the @xmath66 for which @xmath67 , we recover ( [ eq : lim_2 ] ) without the gaussian term .
thus , in general , when @xmath55 is bounded , there is no global asymptotics , but only convergence ( to different distributions ) along subsequences .
next , let @xmath68 . in this case , in ( [ eq : nonunif1 ] ) , the gaussian contribution is negligible .
indeed , since @xmath69 , when @xmath34 , @xmath70 hence , plugging the convergence result ( [ eq : dk ] ) into ( [ eq : nonunif1 ] ) leads to @xmath71 where the limit is first taken as @xmath10 and then as @xmath72 . in this non - uniform setting
, we have the following counterpart to corollary [ corol : tw ] with an additional control on the second largest probability for the letters of @xmath56 .
more precisely , let @xmath73 .
[ theo : nonunif_tw ] let the size @xmath1 of the alphabets vary with @xmath0 and assume that @xmath74 , the number of most probable letters in @xmath56 , goes to infinity when @xmath10 , in such a way that @xmath75 .
assume , moreover , that @xmath76 then @xmath77 let us again stress the fact that in the previous result , @xmath1 is a function of @xmath0 , with the only requirement being that @xmath75 .
note that in the uniform case , @xmath78 and @xmath79 and that , in general , @xmath80 .
let us now put our results in context , relate them to the current literature and also describe the main steps in the arguments developed below .
bodineau and martin @xcite showed that the fluctuations of the last - passage directed percolation model with gaussian i.i.d .
weights actually extend to i.i.d . weights with finite @xmath81th moment , @xmath82 .
their arguments rely , in part , on a kmt approximation which was already used by glynn and whitt @xcite in a related queueing model . here
, we closely follow @xcite and take advantage of the representation ( [ eq : spectrum ] ) of the spectrum of a matrix in the gue . using brownian scaling in those brownian functionals
, we can mix together @xmath0 and @xmath1 in the corresponding limit ( [ eq : johansson ] ) ( see ( [ eq : scaling2 ] ) below ) . then , exhibiting an expression similar to ( [ eq : spectrum ] ) , but with _ dependent _ bernoulli random variables , for the shape of the young tableau ( see ( [ eq : vn ] ) ) , we show via a gaussian approximation that the bernoulli functionals stay close to the brownian functionals ( see ( [ eq : bound1 ] ) ) , so as to share the same asymptotics .
since we apply a gaussian approximation to bernoulli random variables with a strong integrability property , the strong approximation can be made more precise than in @xcite .
however , this is not enough to obtain the fluctuations for @xmath1 of larger order .
actually , the gaussian approximation is responsible for the condition @xmath44 , which falls short of the corresponding polynomial order condition @xmath83 obtained in @xcite .
however , in contrast to @xcite , the stronger integrability property of the bernoulli random variables and the stronger condition on @xmath1 are required to control the constants appearing in the gaussian approximation applied to a triangular scheme of different distributions . using skorokhod embedding ,
baik and suidan @xcite derived , independently of @xcite , similar convergence results ( see theorem 2 in @xcite ) , under the condition @xmath84 .
see also @xcite for related results ( under @xmath85 ) in percolation models using functional methods in the clt .
finally , note that @xcite deal with percolation models with i.i.d .
random variables under enough polynomial integrability . in our setting , the lengths of the rows of the young tableaux associated to random words are expressed in terms of dependent ( exchangeable in the uniform case ) bernoulli random variables .
we are thus working with much more specific random variables , but without complete independence .
the paper is organized as follows : section [ sec : proof_main ] is devoted to the proof of theorem [ theo : main ] , while we sketch the changes needed to prove theorem [ theo : nonunif_tw ] in section [ sec : nonunif_tw ] .
we conclude in section [ sec : remarks ] with some remarks on the convergence of whole shape of young tableaux when the draws are non - uniform .
_ brownian scaling . _
let @xmath86 , @xmath87 , be independent standard brownian motions .
for @xmath88 , @xmath89 and @xmath90 , let @xmath91 where @xmath92 .
for @xmath58 , @xmath93 coincides with the brownian percolation model used in @xcite ; see also @xcite for a related queueing model . for @xmath94
, @xmath95 has the same law as the spectrum of an @xmath9 gue matrix ; see @xcite and @xcite . since @xmath96 is a continuous function of @xmath97 , which are independent
, brownian scaling entails that @xmath98 plugging ( [ eq : scaling1 ] ) into ( [ eq : johansson ] ) yields @xmath99 _ combinatorics .
_ let @xmath100 be bernoulli random variables with parameter @xmath101 and variance @xmath102 .
for a fixed @xmath103 , the @xmath104s are independent and identically distributed , while for @xmath105 , @xmath106 and @xmath107 are identically distributed , but no longer independent .
again , recall that the length of the first row of the young tableau of a random word is the length of the longest ( weakly ) increasing subsequence of @xmath6 .
let @xmath108 be the number of occurrences of @xmath109 among @xmath110 .
an increasing subsequence of @xmath110 consists of successive blocks , each one made of an identical letter , with the sequence of letters representing each block being strictly increasing .
since , for @xmath111 the number of occurrences of @xmath109 among @xmath112 is @xmath113 , it follows that @xmath114,\ ] ] with the convention that @xmath115 .
more involved combinatorial arguments yield the following expression for @xmath116 ( see theorem 5.1 in @xcite ) : @xmath117 where @xmath118 for @xmath119 , set @xmath120=([t_{j , l } ] \dvtx 1\leq j\leq n , 0\leq l\leq m)\in j_{r , m}(n)$ ] and thus @xmath121}^{m , l}-s_{[t_{j , l-1}]}^{m , l}\bigr),\ ] ] which is to be compared with ( [ eq : percolation2 ] ) for brownian functionals . _ centering and reducing .
_ let @xmath122 and @xmath123 , and , replacing @xmath104 by @xmath124 , similarly define @xmath125 .
clearly , @xmath126 , hence , @xmath127 note that @xmath128 , and that @xmath129 , and so the limit under study is the same as that of @xmath130 _ bound .
_ next , and as in @xcite , we bound the difference between @xmath131 and @xmath132 .
this bound holds true for any brownian motions @xmath133 , but it will only be correctly controlled for a special choice of the brownian motions and for copies of the random variables @xmath124 given by a coupling ( using a strong approximation result , see proposition [ prop : kmt ] below ) .
@xmath134}^{m , l}-s_{[t_{j , l-1}]}^{m , l}\bigr ) -\sup_{\mathbf{t}\in i_{k , m}(n)}\sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(b^{l}(t_{j , l})-b^{l}(t_{j , l-1})\bigr)\biggr|\nonumber\\ & & \quad\leq\sup_{\mathbf{t}\in i_{k , m}(n ) } \biggl| \sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(s_{[t_{j , l}]}^{m , l}-s_{[t_{j , l-1}]}^{m , l}\bigr ) -\sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(b^{l}(t_{j , l})-b^{l}(t_{j , l-1 } ) \bigr)\biggr|\nonumber\\ & & \quad=\sup_{\mathbf{t}\in i_{k , m}(n ) } \biggl| \sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(s_{[t_{j , l}]}^{m , l}-b^{l}(t_{j , l})\bigr ) -\sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(s_{[t_{j , l-1}]}^{m , l}-b^{l}(t_{j , l-1})\bigr)\biggr|\nonumber\\ & & \quad=\sup_{\mathbf{t}\in i_{k , m}(n ) } \biggl| \sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(\bigl(s_{[t_{j , l}]}^{m , l}-b^{l}([t_{j , l}])\bigr)+ \bigl(b^{l}([t_{j , l}])-b^{l}(t_{j , l})\bigr)\\ & & \qquad{}\hspace*{80pt } -\bigl(s_{[t_{j , l-1}]}^{m , l}-b^{l}([t_{j , l-1}])\bigr)- \bigl(b^{l}([t_{j , l-1}])-b^{l}(t_{j , l-1})\bigr)\bigr)\biggr|\nonumber\\ & & \quad\leq\sup_{\mathbf{t}\in i_{k , m}(n ) } \biggl\ { \sum_{j=1}^k\sum_{l = j}^{m - k+j } \bigl(\bigl|s_{[t_{j , l}]}^{m , l}-b^{l}([t_{j , l}])\bigr|+|b^{l}([t_{j , l}])-b^{l}(t_{j , l})|\nonumber\\ & & \qquad{}\hspace*{82pt } + \bigl|s_{[t_{j , l-1}]}^{m , l}-b^{l}([t_{j , l-1}])\bigr|+|b^{l}([t_{j , l-1}])-b^{l}(t_{j , l-1})| \bigr)\biggr\}\nonumber\\ & & \quad\leq2k\sum_{l=1}^{m}(y_n^{m , l}+w_n^{l}),\nonumber\end{aligned}\ ] ] where we set latexmath:[\[y_n^{m , l}=\max_{1\leq i\leq n } |s_i^{m , l}-b^{l}(i)| \quad \mbox{and}\quad w_n^{l}=\mathop{\sup _ { 0\leq s , t\leq n}}_{|s - t|\leq1 } that for each @xmath0 and @xmath136 $ ] ( recall that @xmath137 ) , the random variables @xmath138 , @xmath42 , and the brownian motion @xmath139}$ ] appearing in @xmath140 and @xmath141 ( rewritten as @xmath142}$ ] ) , are given by the following result , which is a compilation of strong approximation results of komls , major and tusndy and of sakhanenko , and for which we refer to @xcite ( theorem 2.1 , corollary 3.2 ) and the references therein . in the sequel ,
we write @xmath143 and @xmath144 , instead of @xmath145 and @xmath141 , to insist on the dependence in @xmath1 of the random variables given by the following proposition .
[ prop : kmt ] let @xmath146 be a sequence of i.i.d . random variables with common distribution @xmath147 having finite exponential moments . then , on a common probability space and for every @xmath148 , one can construct a sequence @xmath149 having the same law as @xmath150 , and independent gaussian variables @xmath151 having the same expectations and variances as @xmath150 , such that for every @xmath152 , @xmath153 where @xmath154 and @xmath155 are positive constants ( depending on @xmath147 ) .
moreover , @xmath156 and @xmath157 , where @xmath158 is an absolute constant and @xmath159 is given by @xmath160|^3\exp ( \lambda|x_1-\mathbb{e}[x_1]|)\bigr]\leq\mathbb{e}\bigl[|x_1-\mathbb { e}[x_1]|^2\bigr]\bigr\}.\ ] ] the strong approximation entails the following bound for the tail of @xmath140 : @xmath161 where @xmath162 and @xmath163 .
observe that @xmath164)$ ] and note that @xmath165 $ ] .
indeed , for @xmath166 , @xmath167|^2]&=&\frac1 m \biggl(1-\frac1m\biggr)\\ & \leq & \frac1m\biggl(1-\frac1m\biggr ) \frac{\lambda}2\\ & \leq&\frac1m\biggl(1-\frac1m\biggr ) \biggl(1-\frac2m+\frac2{m^2 } \biggr)\lambda\\ & = & \lambda\mathbb{e}[|\widetilde{x}^m_{1,1}-\mathbb { e}[x^m_{1,1}]|^3]\\ & \leq&\lambda\mathbb{e}[|x^m_{1,1}-\mathbb{e}[x^m_{1,1}]|^3\exp ( \lambda|\widetilde { x}^m_{1,1}-\mathbb{e}[x^m_{1,1}]|)],\end{aligned}\ ] ] while , since @xmath168|\leq1 $ ] , @xmath169|^3\exp \bigl(\tfrac 12|x^m_{1,1}-\mathbb{e}[x^m_{1,1}]|\bigr)\bigr ] & \leq&\tfrac12\exp\bigl(\tfrac12\bigr)\mathbb{e}\bigl[|x^m_{1,1}-\mathbb { e}[x^m_{1,1}]|^2\bigr]\\ & \leq&\mathbb{e}\bigl[|x^m_{1,1}-\mathbb{e}[x^m_{1,1}]|^2\bigr].\end{aligned}\ ] ] thus , @xmath170 and @xmath171 behave like @xmath172 .
also , note that the bound in ( [ eq : kmt ] ) is non - trivial for @xmath173 .
_ approximating sets .
_ let @xmath174 , for some @xmath175 , where @xmath176 is some finite constant .
we have @xmath177 & \leq&\sum_{l\leq m}\mathbb{p}(y_n^{m , l}>a_n)\\[-1pt ] & \leq&m \mathrm{e}^{-c_1(m)a_n}\bigl(1+c_2(m)n^{1/2}\bigr)\\[-1pt ] & \sim & \sqrt{mn}\mathrm{e}^{-c_1(m)a_n}\\ & = & \sqrt{mn}\mathrm{e}^{-(c_3c(\log n)^2)/2 } \to0 , \qquad n\to+\infty.\end{aligned}\ ] ] let @xmath178 , for @xmath179 .
standard estimates ( including reflection principle , brownian scaling and gaussian tail estimates ) lead to @xmath180 & \leq&\sum_{l\leq m}\mathbb{p}(w_n^{m , l}>b_n)\\[-1pt ] & \leq&m\mathbb{p}(w_n^{m,1}>b_n)\\[-1pt ] & = & m\mathbb{p}\bigl(\mathop{\sup _ { 0\leq s , t\leq n}}_{|s - t|\leq1 } @xmath181 & \leq&\sup_{0\leq i\leq n-2}\bigl(\sup_{i\leq t\leq i+2}b_t^{m,1}-\inf _ { i\leq s\leq i+2}b_s^{m,1}\bigr)\vspace*{-1pt}\end{aligned}\ ] ] and so @xmath182 \nonumber & \leq&mn\mathbb{p}\bigl(\sup_{t\in[0,2]}b_t^{m,1}-\inf_{s\in [ 0,2]}b_s^{m,1}>b_n\bigr)\nonumber\\[-1pt ] & \leq&mn\bigl(\mathbb{p}\bigl(\sup_{t\in [ 0,2]}b_t^{m,1}>b_n/2\bigr ) + \mathbb{p}\bigl(\sup_{s\in[0,2]}b_s^{m,1}>b_n/2\bigr)\bigr)\\[-1pt ] \nonumber & \leq&2mn \mathbb{p}(|b_2^{m,1}|>b_n/2)\nonumber\\[-1pt ] & \leq&4mn \exp(-b_n^2/16 ) \to0,\qquad n\to+\infty.\nonumber\end{aligned}\ ] ] _ final bound . _ from ( [ eq : scaling2 ] ) , the approximation of @xmath183 by @xmath184 will imply the theorem if @xmath185 for some @xmath186 since @xmath187 , it is enough to prove that @xmath188 but @xmath189\\ & & \quad\leq\sum_{k=1}^k2rm\mathbb{e}\bigl [ ( y_n^{m,1}+w_n^{m,1 } ) \mathbf{1 } _ { ( a_1^n)^c\cap(a_2^n)^c}\bigr]\\ & & \quad\leq2r^2m(\mathbb{e } [ y_n^{m,1 } \mathbf{1 } _ { y_n^{m,1}\leq a_n } ] + b_n)\\ & & \quad\leq2r^2m(\mathbb{e } [ y_n^{m,1}-\tilde a_n \mathbf{1}_{y_n^{m,1}\leq a_n } ] + \tilde a_n+b_n)\\ & & \quad\leq2r^2m(\mathbb{e } [ y_n^{m,1}-\tilde a_n
\mathbf { 1}_{\tilde a_n \leq y_n^{m,1}\leq a_n } ] + \tilde a_n+b_n)\\ & & \quad\leq2r^2m\biggl(\int_{\tilde a_n}^{a_n } \mathbb{p}(y_n^{m,1}\geq x)\ , \mathrm{d}x + \tilde a_n+b_n\biggr)\\ & & \quad\leq2r^2m\biggl(\int_{\tilde a_n}^{a_n } \mathrm{e}^{-c_1(m)x}\bigl(1+c_2(m)n\bigr)\ , \mathrm{d}x + \tilde a_n+b_n\biggr)\\ & & \quad\leq2r^2m\biggl(\frac{1+c_2(m)n}{c_1(m ) } \mathrm{e}^{-c_1(m)\tilde a_n}+\tilde
a_n+b_n\biggr)\\ & & \quad\leq2r^2m\biggl(\frac1{c_1(m)}+\tilde a_n+b_n\biggr)\\ & & \quad\leq2r^2m^{3/2}\biggl(\frac{2(1+\log ( 1+c_2(m)n^{1/2}))}{c_3}+b_n\biggr).\end{aligned}\ ] ] finally , @xmath190 \\[-8pt ] \nonumber & & \quad\leq\frac{2r^2m^{3/2}}{c_n}\biggl(\frac{2(1+\log ( 1+c_2(m)n^{1/2}))}{c_3}+\log n\biggr)=\mathrm{o}\biggl(\frac{m^{3/2}\log n}{c_n}\biggr).\end{aligned}\ ] ] a choice of @xmath191 ensuring that the bound in ( [ eq : aa2 ] ) goes to zero as @xmath10 and also compatible with ( [ eq : cn ] ) is possible when @xmath192 , that is , when @xmath44 . finally , ( [ eq : but1 ] ) and ( [ eq : aa1 ] ) hold true , completing the proof of theorem [ theo : main ] . [
rem : condition ] * in the above proof , the condition @xmath44 is needed only once , to ensure the compatibility of ( [ eq : cn ] ) with the bound ( [ eq : aa2 ] ) . however
, this is essential to make the gaussian approximation work .
* when @xmath193 $ ] , the growth condition @xmath194 can be rewritten as @xmath195 and this growth condition remains true , in particular , when @xmath1 is of subpolynomial order .
the condition @xmath195 is stronger than its counterpart @xmath196 in @xcite and this seems to be due to the fact that we work with a triangular array of random variables .
* for the top line of the tableau , our result falls short of a result of johansson in @xcite , which asserts the convergence of @xmath197 ( properly scaled and normalized ) toward the tracy widom distribution .
more precisely , setting @xmath198 for @xmath199 , theorem 1.7 in @xcite actually gives , in our notation , for @xmath200 , @xmath201 for @xmath202 , @xmath203 and , for @xmath204 , @xmath205 in the middle limit above , @xcite , theorem 1.7 , requires @xmath206 , while we do not require a lower bound condition on @xmath1 .
besides , our theorem [ theo : main ] applies to the shape of the whole young tableau . in this section ,
we sketch the changes needed in the previous arguments in order to prove theorem [ theo : nonunif_tw ] .
note that in the uniform setting , the representation ( [ eq : combinatoire_v_k ] ) for @xmath116 is a maximum taken over the most probable letters .
this is trivially true since , in this case , all the letters have the same probability . but this property , which appears to be fundamental when we center and normalize the @xmath104 , is no longer true in the non - uniform setting .
however , we shall approximate @xmath5 below by a random variable @xmath207 defined as a maximum taken over only the most probable letters , as in ( [ eq : combinatoire_v_k ] ) ; see ( [ eq : combinatoire_v_1 ] ) .
part of the remaining work is then to show that we can suitably control this approximation , and this is done in lemma [ lemme : control ] .
this control is at the root of the extra condition ( [ eq : condition2 ] ) in theorem [ theo : nonunif_tw ] .
let us revise our notation for the non - uniform setting .
in this section , @xmath41 , @xmath42 , are independently and identically distributed with @xmath49 . set @xmath50 and @xmath51 , with @xmath52 , and
also set @xmath208 .
finally , note that since @xmath209 and @xmath210 , it follows that @xmath211 as @xmath10 . _
brownian scaling .
_ let @xmath86 , @xmath212 , be independent standard brownian motions .
for @xmath88 , @xmath89 and @xmath90 , let @xmath213 where @xmath214 . recall that @xmath215 has the same law as the largest eigenvalue of a @xmath216 gue matrix ( see ( [ eq : spectrum ] ) , ( [ eq : gue0 ] ) , ( [ eq : johansson ] ) and remark [ rem : johansson - tw ] for @xmath58 ) and so @xmath217 by brownian scaling , @xmath218 , so that when @xmath10 , @xmath219 _ combinatorics revisited .
_ let @xmath220 be bernoulli random variables with parameter @xmath221 and variance @xmath222 .
for a fixed @xmath103 , the @xmath104s are independent and identically distributed .
since the expression ( [ eq : combinatoire_v_1 ] ) has a purely combinatorial nature , we still have @xmath223 with the convention that @xmath224 whenever @xmath225 . in fact , for most draws , the maximum in @xmath226 is attained on the sums @xmath227 corresponding to the most probable letters , that is , letting @xmath228 we have , with large probability , @xmath229 .
however , it is not always true that @xmath229 , for instance , in the case where the @xmath0 letters drawn are letters with associated probability strictly less than @xmath66 , @xmath230 while there is an @xmath231 with @xmath232 such that @xmath233 , ensuring that @xmath234 . in the sequel , we prove theorem [ theo : nonunif_tw ] by first showing that the statement of the theorem is true for @xmath207 instead of @xmath5 and then by controlling the error made when @xmath207 is replaced by @xmath5 . _ centering and reducing .
_ let @xmath235 be the corresponding centered and normalized scaled bernoulli random variables and let @xmath236 .
also , let @xmath237}^{m , j_\delta}-\widetilde s_{[t_{j , l-1}]}^{m , j_\delta}\bigr),\end{aligned}\ ] ] which is to be compared to ( [ eq : percolation3 ] ) .
since @xmath238 , we have @xmath239 since @xmath240 and @xmath241 it remains to show that @xmath242 for which we shall use ( [ eq : scaling3 ] ) .
_ sketch of proof of . _ roughly speaking , the proof of ( [ eq : but11 ] ) follows along the same lines as the corresponding proof of the convergence of ( [ eq : v_k ] ) , changing only @xmath1 to @xmath55 .
we show that if @xmath75 , then for some brownian motions given via strong approximation , we have @xmath243 where @xmath244 indeed , setting @xmath245 for some @xmath246 and setting @xmath247 for some @xmath248 , we show that @xmath249 from ( [ eq : scaling3 ] ) , the approximation of @xmath250 by @xmath251 will imply the theorem if @xmath252 for some @xmath253 since @xmath187 and @xmath254 \\[-8pt ] \nonumber & & \quad\leq\frac{2k(m(n))^{3/2}}{c_n}\biggl(\frac{2(1+\log ( 1+c_2(k(m(n)))n^{1/2}))}{c_3}+\log n\biggr),\end{aligned}\ ] ] a choice of @xmath191 , ensuring that the bound in ( [ eq : cn3 ] ) goes to zero and is compatible with ( [ eq : cn2 ] ) , is possible since @xmath75 .
this proves ( [ eq : but11 ] ) and thus the statement ( [ eq : nonunif_tw ] ) of theorem [ theo : nonunif_tw ] , but for @xmath207 instead of @xmath5 .
_ control of the error @xmath255 .
_ clearly , @xmath256 and is , in fact , zero with a large probability , so that we expect @xmath257 $ ] to be small . actually , we show the following .
[ lemme : control ] for some absolute constant @xmath258 , we have @xmath259\leq cnp_{\mathrm{2nd}}^m,\ ] ] where @xmath260 stands for the second largest probability for the letters of @xmath56 . the conclusion in ( [ eq : nonunif_tw ] )
holds true when @xmath261 \times \frac{k(m(n))^{2/3}}{\sqrt{k(m(n))p_{\mathrm{max}}^{m(n)}n}}\biggr)=0.\ ] ] however , with the help of ( [ eq : control ] ) , the conclusion in ( [ eq : control2 ] ) is then valid when @xmath262 and , since @xmath75 , this will follow from ( [ eq : condition2 ] ) .
it remains to prove lemma [ lemme : control ] , that is , to give an explicit bound on @xmath263 $ ] .
to do so , rewrite @xmath264 and @xmath265 , where @xmath266 , @xmath267 and @xmath268 clearly , since @xmath269 , we have @xmath270 .
moreover , since the @xmath271 are bernoulli random variables with parameter @xmath272 and since the @xmath273 s are independent , we have @xmath274 and @xmath275 , where @xmath276 stands for the binomial distribution with parameters @xmath0 and @xmath277 . if @xmath278 , @xmath279 , since , in this case , @xmath280 .
if @xmath281 , we rewrite @xmath282 as @xmath283 where @xmath284 and @xmath285 is an error term .
indeed , let @xmath286 and for @xmath287 , define @xmath288 where @xmath289 and where @xmath290 .
now , @xmath291 define @xmath284 by @xmath292 if @xmath293 and @xmath294 for @xmath295 , where @xmath296 , with the conventions that @xmath297 and that @xmath298 for @xmath299 .
we then have @xmath300 let @xmath301 be the random variables taking the values @xmath302 and @xmath303 with respective probabilities @xmath304 and @xmath272 . independently ,
let @xmath305 be bernoulli random variables with parameter @xmath306 , where @xmath73 , and define @xmath307 note that @xmath308 and that @xmath309 , so that @xmath310 since @xmath311 , we have @xmath312 next , observe that for @xmath278 , @xmath313 .
however , since the event @xmath314 is non - negligible , we can not change @xmath315 into @xmath316 .
we obtain @xmath317 the random variable @xmath318 is the sum of @xmath319 i.i.d .
random variables so that @xmath315 is distributed according to @xmath320 , where @xmath321 are i.i.d .
with @xmath322 we are now interested in bounding @xmath323 $ ] .
let @xmath324 be i.i.d .
bernoulli random variables with parameter @xmath325 and let , independently , @xmath326 be i.i.d .
rademacher random variables with parameter @xmath327 ( i.e. , @xmath328 ) .
then @xmath329 and @xmath330 have the same distribution and we have @xmath331 & = & \mathbb{e}\biggl[\biggl(\max_{1\leq k\leq n}\sum_{i=1}^k\varepsilon _
i^m y_i^m \biggr)^+\biggr]\\ & = & \mathbb{e}\biggl[\mathbb{e}\biggl[\biggl(\max_{1\leq k\leq n}\sum _ { i=1}^k\varepsilon_i^m y_i^m\biggr)^+\big|\mathcal { g}_n\biggr]\biggr],\end{aligned}\ ] ] where @xmath332
. however , since @xmath324 is independent of @xmath326 , we have @xmath333 = \mathbb{e}\biggl[\biggl(\max_{1\leq k\leq\ell}\sum_{i=1}^k y_i^m\biggr)^+ \big|\mathcal { g}_n\biggr],\ ] ] where @xmath334 has a @xmath335 distribution . but @xmath336 & = & \sum_{k=1}^{+\infty } \biggl(\mathbb{p}\biggl(\biggl(\max_{1\leq j\leq\ell } \sum_{i=1}^j y_i^m\biggr)^+\geq k\biggr)\biggr)\\ & = & \sum_{k=0}^{+\infty}\biggl ( 1-\mathbb{p}\biggl(\max_{1\leq j\leq \ell } \sum _ { i=1}^j y_i^m\leq k\biggr)\biggr)\\ & = & \sum_{k=0}^{\ell-1 } \biggl(1-\mathbb{p}\biggl(\max_{1\leq j\leq \ell } \sum _ { i=1}^j y_i^m\leq k\biggr)\biggr)\\ & = & \ell - u_\ell,\end{aligned}\ ] ] where @xmath337 and @xmath338 . with the latest notation
, we are now investigating @xmath339 $ ] .
for simplicity , in the sequel , we set @xmath340 and @xmath341 . the elements of the sequence @xmath342 satisfy the following induction relations : @xmath343 and @xmath344 for @xmath345 . from these
, we derive @xmath346 and , since @xmath347 , @xmath348 . in order to compute @xmath349
, we introduce the hitting time @xmath350 of the random walk @xmath351 .
we then have @xmath352 so that @xmath353 and @xmath354.\end{aligned}\ ] ] next , @xmath355 = \ell(1 - 2q_{*,m})+q_{*,m}\mathbb{e}[\tau_1^m\wedge\ell|\mathcal { g}_n]\ ] ] and we have @xmath356\\ & = & \mathbb{e}\bigl[\ell(1 - 2q_{*,m})+q_{*,m}\mathbb{e}[\tau _
1^m\wedge\ell|\mathcal { g}_n]\bigr]\\ & \leq & ( 1 - 2q_{*,m}+q_{*,m})\mathbb{e}[\ell]\\ & = & n p_{2nd}^m.\end{aligned}\ ] ] this completes the proof of lemma [ lemme : control ] .
a natural question to consider next would be to derive a result similar to theorem [ theo : main ] for non - uniformly distributed letters .
the special case of the longest increasing subsequence ( i.e. , @xmath357 ) is dealt with in theorem [ theo : nonunif_tw ] .
let us investigate what happens for the whole shape of the young tableau .
first , let us slightly expand our notation . in this section , @xmath41 ,
@xmath42 , are independently and identically distributed with @xmath358 . in order to simplify the notation ,
we assume ( without loss of generality ) that the ordered letters @xmath359 have , moreover , non - increasing probabilities ( i.e. , @xmath360 ) .
let @xmath361 be the multiplicity of @xmath362 and let @xmath363 be the number of letters ( strictly ) more probable than @xmath364 .
let @xmath365 be the indices of the letters with the same probability @xmath366 .
we recover our previous notation , @xmath357 , with @xmath367 and @xmath368 .
since the expression ( [ eq : combinatoire_v_k ] ) has a purely combinatorial nature , it still holds true that @xmath369 let @xmath370 . note that , from theorem 5.2 in @xcite , when @xmath1 is fixed and @xmath371 , we have for each @xmath372 that @xmath373 where the limit is given in section 6 of @xcite by @xmath374 , with @xmath375 for @xmath376 , and @xmath377 for @xmath378 note that @xmath379 is a natural generalization of the brownian functional @xmath380 used in section [ sec : nonunif_tw ] ( see also , in a queuing context , @xcite and @xcite ) . in particular
, @xmath379 is equal in distribution to the sum of the @xmath381 largest eigenvalues of an @xmath9 matrix from the gue and theorem [ theo : johansson ] can be rewritten as @xmath382 arguing as in the previous sections , we would like to derive the fluctuations of @xmath383 with respect to @xmath0 and @xmath1 simultaneously from ( [ eq : hl3bis ] ) and ( [ eq : dkm ] ) . however , in the non - uniform case , this is not that transparent since , for each @xmath35 , the behavior of @xmath384 and of @xmath385 , with respect to @xmath1 , is not that clear cut . in particular , @xmath386 may not be stationary and @xmath387 can no longer be used for @xmath388 . besides , the random fluctuations of @xmath389 in @xmath390 are of order @xmath391 , which , in general , does not dominate those of @xmath375 .
thus , for general non - uniform alphabets , we can not infer which part of the law of @xmath392 will drive the fluctuations .
we can imagine that , taking simultaneous limits in @xmath0 and @xmath1 , the fluctuations of @xmath393 , properly centered and normalized , are either gaussian , or driven by @xmath46 ( as in theorem [ theo : main ] ) or given by an interpolation between these distributions , depending on the alphabets considered .
the first author thanks the school of mathematics of the georgia institute of technology where this work was initiated . the second author would like to thank the quipe modalx of the university paris x for its hospitality and support while part of this research was being carried out .
research of the second author was supported in part by nsa grant h98230 - 09 - 1 - 0017 .
bodineau , t. and martin , j. ( 2005 ) . a universality property for last - passage percolation paths close to the axis .
_ electron .
* 10 * 105112 .
bougerol , p. and jeulin , t. ( 2002 ) .
paths in weyl chambers and random matrices .
_ probab . theory related fields _ * 124 * 517543 .
chistyakov , g.p . and gtze , f. ( 2004 ) .
distribution of the shape of markovian random words .
_ probab . theory related fields _ * 129 * 1836 .
its , a. , tracy , c. and widom , h. ( 2001 ) .
random words , toeplitz determinants , and integrable systems i. in _ random matrix models and their applications_. _ math .
_ * 40 * 245258 .
cambridge univ . press .
|
given a random word of size @xmath0 whose letters are drawn independently from an ordered alphabet of size @xmath1 , the fluctuations of the shape of the random rsk young tableaux are investigated , when @xmath0 and @xmath1 converge together to infinity . if @xmath1 does not grow too fast and if the draws are uniform , then the limiting shape is the same as the limiting spectrum of the gue . in the non - uniform case , a control of both highest probabilities will ensure the convergence of the first row of the tableau toward the tracy
widom distribution .
|
the question of possible modifications of hadron properties in the nuclear medium has attracted considerable interest in recent years .
this interest stems from the hope that studies of these effects can clarify the interplay of the competing degrees of freedom in the nucleus .
in particular it may be possible to learn about explicit quark degrees of freedom in the nucleus through such studies .
there have been several theoretical as well as experimental investigations of these effects , and evidence is gathering that indeed such modifications do take place .
several studies have been carried out for the modifications of meson properties in heavy ion and other collisions @xcite .
photonuclear reactions also lend themselves to investigations of this type @xcite . in particular ,
a promising area of investigation is the photoproduction of @xmath0 mesons on nuclei in the energy range corresponding to the second resonance region of the nucleon .
the fact that one resonance , the s@xmath1(1535 ) , dominates the production process at these energies , makes it attractive to look at these reactions for clues on changes in its properties in the nuclear medium .
three pioneering and complementary measurements of the inclusive eta photoproduction on nuclei have been reported in recent years .
the earlier measurements of robig - landau _
et al _ @xcite at mami included cross sections on a number of nuclei from threshold up to a photon energy of 800 mev .
these cross sections displayed the expected rise on the low energy side of the resonance .
few years later measurements of the same reaction were performed at kek @xcite , extending the energy range to 1gev .
this group reported that a broad resonance due to the excitation and decay of the s@xmath1 resonance in the nucleus has been observed for the first time .
in a recent investigation , kinoshita _ et al _
@xcite reported on measurements of this reaction on c and cu targets at lns , for photon energies up to 1.1 gev .
these authors indicated the importance of the contributions from other nucleon resonances in the second resonance region in addition to the dominant s@xmath1 , at energies higher than 900 mev .
the authors of references @xcite performed calculations for their data using an adaptation of the quantum molecular dynamics model .
@xcite performed calculations based on the semi - classical buu transport model .
another approach was that of maruyama and chiba @xcite who carried out relativistic calculations in infinite nuclear matter based on an effective lagrangian and a scaling factor to account for the final state interactions of the @xmath0 mesons . in the present paper we present an analysis of the data based on a relativistic model for the inclusive photoproduction on finite nuclei .
the main ingredients of model are the use of an effective lagrangian approach ( benmerrouche _ et al_@xcite ) and the relativistic mean field approach to nuclear dynamics @xcite .
the final state interactions of the outgoing @xmath0 meson are taken into account .
this fully non - local model has been discussed recently for the cases of exclusive and incoherent photoproduction @xcite .
these non - local calculations , in contrast to earlier local ones @xcite , allow us to include possible density dependent changes in the mass and width of the resonances in the nuclear medium .
the justification for the claim that the inclusive data from kek , mainz and lns may reveal changes in the properties of the s@xmath1 resonance requires some discussion .
it is known that the s@xmath1 is the main contributor to the @xmath0 photoproduction cross section on free nucleons in the 0.6 - 1gev range .
there are contributions from neighboring resonances and from nucleon born terms as well as meson exchange diagrams ; but these are small by comparison .
we will show in the course of our discussion that this situation carries over to the inclusive production on nuclei .
the contributions are only at the few percent level , a situation which allows us to probe whether the existing data does reveal any information regarding the properties of the s@xmath1 in the medium . as we shall show there are remaining uncertainties regarding other aspects of the analysis which are more important compared to the small contributions of other diagrams .
the plan of the paper follows a recent discussion of the non - local relativistic model for the exclusive quasifree photoproduction of @xmath0 mesons .
the reaction amplitude is cast in a general form in section [ form ] where the method of evaluation of the amplitude in its non - local form is outlined .
the results of the model and the discussion are given in section [ results ] .
conclusions are given in section [ conclusion ] .
in the inclusive reaction , a photon is absorbed by the nucleus and as a result an @xmath0 meson is produced .
this meson is the only detected particle and there is no information about the final state of the nuclear system . in the present work it is assumed that the inclusive reaction is based on the exclusive reaction in which a single nucleon is ejected in the continum in the course of the production process @xcite .
the transition amplitude for the exclusive reaction through the s@xmath1 resonance , in its full non - local form , is given by @xcite , @xmath2^\frac{1}{2 } \nonumber\\ \times & & \{\frac { } { } \int d^4 x d^4 y d^4 p \hspace{0.05 in } \bar{\psi}_{sf}(y ) \phi^*_\eta(y ) \frac { e^{-i p(y - x ) } } { { / \hspace{-0.095 in}p } - m_r + i\frac{\gamma}{2 } } \gamma_5 { { / \hspace{-0.105 in } k } _
{ \gamma } { / \hspace{-.09 in } \epsilon } } \hspace{0.05 in } e^{-i k_{\gamma } x } \psi_b(x ) \nonumber \\ & & + \int d^4 x d^4 y d^4 p \hspace{0.05 in } \bar{\psi}_{sf}(y ) \gamma_5 { { / \hspace{-0.105 in } k } _ { \gamma } { / \hspace{-0.09 in } \epsilon } } \hspace{0.05 in } e^{-i k_{\gamma}y } \frac { e^{-i p(y - x ) } } { { / \hspace{-0.095 in}p } - m_r + i\frac{\gamma}{2 } } \phi^*_\eta(x)\psi_b(x)\hspace{0.05 in}\frac{}{}\ } , \nonumber\end{aligned}\ ] ] where @xmath3 , @xmath4 and @xmath5 are the energies of the final nucleon , @xmath0 meson and incident photon , respectively .
@xmath6 is the nucleon mass , @xmath7 and @xmath8 represent the mass and width of resonance . @xmath9 and @xmath10 are spectroscopic and clebsh - gordon coefficients . @xmath11 and @xmath12 are the anomalous magnetic moment of the resonance and its coupling constant with @xmath0 and nucleon .
@xmath13 , @xmath14 and @xmath15 are wave functions describing the final nucleon , initial nucleon and @xmath0 meson , respectively .
@xmath14 is a solution of the dirac equation with appropriate mean field potentials and @xmath16 is a solution of the klein - gordon equation . as is explained in @xcite , the wave function of the final nucleon ( @xmath13 )
is taken to be a plane wave .
calculations without inclusion of the final state interaction of @xmath0 meson revealed that the u - channel contribution ( second term in ( [ eq1 ] ) ) is less by two orders of magnitude than that of s - channel ( first term in ( [ eq1 ] ) ) . as a result
the non - locality and possible medium modifications of the former will have a negligible effect on the calculated cross section .
we thus ignore this very small change and treat the u - channel locally as in @xcite .
the dominant s - channel diagram is treated in the non - local approach .
the medium modifications of the properties of the nucleon resonance are often expressed in terms of changes of its mass and width from their free values in a manner dependent on the nuclear density @xcite .
our non - local approach can accommodate these density dependent parameters for the resonance . after implementing the density dependent mass and width for the resonance in the s - channel part of the above amplitude , we calculate the resulting integral following an approach similar to the one used in ref.@xcite .
we rewrite the integral of the s - channel part in eq([eq1 ] ) as @xmath17 where we use a plane wave function ( see ref .
@xcite ) for the final state of the struck nucleon ; @xmath18 denotes its spin state . the function @xmath19 has the following form @xmath20 acting with the operator @xmath21 ( @xmath22 refers to the density of the nucleus ) from the left on ( [ eq3 ] ) and then carrying the integration over momentum , the following dirac - type linear differential equation is obtained @xmath23 where the source term is @xmath24 equation ( [ eq4 ] ) leads to the following second order differential equation for the the upper component of @xmath25 .
@xmath26 w_{up}({\bf r } ) = { \bf \sigma } \cdot{\bf p } v_d({\bf r})-\beta(r ) v_{up}({\bf r}),\ ] ] where the indices @xmath27 and @xmath28 indicate the upper and lower components of the functions @xmath29 and @xmath30 .
note these functions continue to be spin dependent .
the lower component @xmath31 can be obtained from the upper component as , @xmath32 the functions @xmath33 and @xmath34 are @xmath35 where @xmath36 is the energy of the bound nucleon .
@xmath37 is obtained by solving eq.([eq6 ] ) following the same approach as in ref.@xcite .
substituting the solution of @xmath38 into eq.([eq1 ] ) we get , @xmath39 } ^{1/2 } } ( z^{s } _ { s_f m_b } + z^{u } _ { s_f m_b } ) \nonumber \end{aligned}\ ] ] the transition function for the u - channel @xmath40 is calculated in the local approximation and is the same as eq.(3 ) in @xcite .
the corresponding function for the s - channel is calculated in non - local approach described above and is @xmath41^ * \nonumber\\ & & \times\left ( l , 1/2 ; m - s_f , s_f| j , m\right ) \\ \times\int & d^3 r & \left[{\cal y}^{m}_{l,1/2,j}(\omega)\right]^ * v_{l_\eta}(r)y_{l_\eta}^{m_\eta}(\omega ) ( w_u({\bf r } ) -\frac{\sigma\cdot { \bf k_f}}{e_n + m}w_d({\bf r } ) ) \nonumber\end{aligned}\ ] ] where @xmath42 is the radial wave function of the @xmath0 meson .
@xmath43 and @xmath44 are unit vectors along the directions of the final nucleon and @xmath0 meson , respectively .
the amplitude ( [ eq11 ] ) is used to calculate the cross section for the exclusive reaction ( see eq(5 ) of ref.@xcite ) .
the inclusive cross section is obtained by numerical integration of the exclusive cross section over the phase space of the final nucleon and summing over all energy levels of the target nucleus . in the following section we shall explore whether the available data can be used to clarify the issue of possible modifications of the mass and width of the s@xmath1 resonance . for this purpose
we will adopt phenomenological forms of these parameters suggested by earlier works @xcite .
for mass we use a density dependent form suggested by the quark - meson coupling model for hadrons @xcite .
@xmath45 for width we use a density dependent form given in @xcite which includes the broadening of the resonance width due to its collisions with hadrons in the medium , @xmath46
throughout the calculations presented here we use scalar and vector potentials of woods saxon form to generate the bound state dirac wave functions of the nucleons .
the parameters of these potentials for different targets are listed in table [ table1 ] .
the final state interactions of the eta meson are taken into account through the use of the optical potential dw1 from ref.@xcite . for the coupling to the proton we use the parameters of set 3 given in table 1 of ref.@xcite and in calculating the neutron contributions we use the factor @xmath47 for the ratio of neutron to proton cross sections .
it is well known that the photoproduction process on the nucleon , in the second resonance region , is dominated by the s@xmath1(1535 ) diagram .
we begin our discussion by showing that this situation carries over to the inclusive reactions on nuclei .
the collective contributions from other diagrams are small by comparison .
this is shown in fig.1 .
figure 1(a ) shows the angular distribution of the inclusive cross section of the reaction on @xmath48c at a photon incident energy e@xmath49 = 820 mev .
the solid curve gives the s@xmath1 contribution while the dashed curve includes the additional contributions due to the d@xmath50 , the nucleon pole terms and the meson exchange diagrams .
the total inclusive cross section on @xmath48c for photon energies in the range 0.65 - 1.1 gev is depicted in fig.1(b ) .
the dominance of the s@xmath1 contributions is clearly evident from these comparisons .
it is worth mentioning here that this rather fortunate circumstance arises not only from the relative smallness of the contributions of the other diagrams , but also from strong cancellations among these contributions .
fig.2 shows the inclusive angular distributions for c , al and cu . for each target nucleus
, we show the distributions at two photon energies from among those measured : 740 mev and 980 mev .
the data are those of kek @xcite , lns @xcite and mainz ( only for @xmath48c at 740 mev ) @xcite .
the solid curves represent non - local calculations that assume free propagation for the s@xmath1 resonance .
we take the free mass and width of the resonance to be 1535 mev and 160 mev , respectively . in these and
subsequent figures these curves are labeled free. general trends in the data are reproduced and the agreement is almost quantitative in some cases .
the main exception is the case of al at 980 mev .
there are some disagreements among the measurements from kek and lns in the case of cu and our theoretical calculations show better agreement with the latter measurements . the dashed curves in fig .
2 represent calculations in which , following suggestions of ref.@xcite , the width of the s@xmath1 resonance is assumed to undergo density dependent broadening in the medium of the form given in eq([eq8 ] ) with c = 50 mev .
we note that the broadening leads to reduction in the cross sections for all three nuclei .
it may be argued that the effect leads to a worsening in the accord with the data for c at the lower energy ( 740 mev ) and by contrast to a slight improvement at 980 mev . for al
the quality of agreement with the data remains about the same , whereas for cu , the disagreement between measurements makes it difficult to judge the outcome .
we now consider the effect of a reduction of mass on the angular distributions . in the discussion of ref.@xcite
, the authors suggested that the nucleon mass in the medium is reduced according to eq.([eq7 ] ) with a value of the reduction parameter b = 0.14 .
fig.3 shows the results of the angular distribution for the same targets and photon energies as in fig.2 .
the dash - dotted curves are obtained when the above value is used for the reduction of the mass of the s@xmath1 .
compared to calculations with no medium effects ( solid curves ) , we note a large decrease in the cross section .
this also brings the calculated cross section much below the data .
this clearly indicates that if the mass of the s@xmath1 were reduced in the medium , the scale of reduction would be much smaller than is indicated by the value of b given above .
if the reduction in mass is made weaker the effect becomes smaller .
for example the dashed curves show the calculations for a much smaller value of b ( b = 0.014 ) . here
the effect on the cross section is relatively small .
we note that at the lower energy the effect leads to a slight increase in the magnitude of the cross section while the effect at the higher energy is a slight decrease .
the data for the total inclusive cross section ( obtained by integrating the angular distributions discussed above ) are shown in figs.4 - 6 .
these include the data for the three nuclei c , al and cu from kek , mainz and lns .
fig.4 displays the inclusive data for c. in fig.4(a ) we show the effects of changes in the mass of the s@xmath1 on the behavior of the total inclusive cross section in the energy range from near threshold to 1.1 gev .
the solid curve represents the case of no medium modification of the mass .
we note that these calculations provide reasonable agreement with the data particularly over and beyond the peak , but are slightly lower at the lower energies .
the dashed and dash - dotted curves represent the two cases where the mass modification parameter b of eq.([eq7 ] ) has the values 0.014 and 0.035 , respectively . the latter value , though it improves the accord with the data on the lower energy side of the peak somewhat , still gives a strong suppression on the higher energy side .
calculations with the smaller value of b lie blew the data at energies above 900 mev . in fig.4(b )
we present similar comparisons involving changes to the width of the resonance .
we show two cases for density dependent changes in the width of the form given in eq.([eq8 ] ) .
the dash - dotted curve is for a value of c = 25 mev and the dashed curve is for c = 50 mev .
both these calculations lead to a lowering of the cross section away from the data .
figs.5 shows similar comparisons for the kek inclusive cross section on al and fig.6 compares the calculations with the kek and lns data for cu .
the qualitative features of the results are similar . again
the data rule out any strong reduction ( associated with b = 0.14 ) in the mass of the resonance . a reduction in the mass of s@xmath1 corresponding to a value of b = 0.014 brings the calculations somewhat closer to the data for al , but leads to a reduced cross section at higher energies for cu . in the case of the modification of width
we note that the density dependent width of eq.([eq8 ] ) , with c = 50 mev , leads to a suppression of the cross section away from the data .
we have argued earlier that the dominance of the s@xmath1 contributions is used to limit the forgoing comparisons and discussions to this contribution . in fig 7
we revisit this question . in part
( a ) we show the results for the case of production on c for the s@xmath1 .
we compare this with the results in the lower panel ( b ) based on calculations in which the small contributions from the d@xmath50 , the nucleon pole and the t - channel vector meson diagrams are included .
these contributions , being small , are calculated using the local approximation ( the non - local corrections in this case are expected to be negligible ) .
each of the curves in panel ( a ) has a counterpart in panel ( b ) : the solid curves represent the free calculations , the dashed curves represent calculations in which the mass reduction is affected by a value b = 0.014 .
the dash - dotted curves represent calculations in which we make changes in both mass and width of the resonance .
we assumed the width is reduced to 150 mev ( from 160 mev ) and in the meantime the mass is reduced as above .
it is seen from the figures that the inclusion of the additional contributions , though it modifies the cross section slightly , does not alter the character of the comparison with the data , from those involving only the s@xmath1 contributions .
the preceding comparisons were based on calculations using the optical potential dw1 to describe the final state interaction of the @xmath0 meson .
the calculations of the cross sections are sensitive to the optical potential used .
for example we show in fig.8 a comparison of the cross sections calculated using one of the optical potentials derived by oset _
et al._@xcite .
these calculations of the eta optical potential depend on the magnitude of the real part of s@xmath1 self energy in the medium , and we have used the case for which the real part of the s@xmath1 self - energy is -50 mev .
the cross sections are seen to lie below those calculated using the dw1 potential .
it is clear that this type of uncertainty should be kept in mind before a final verdict is gleaned form the data as to the extent of any medium modifications of the properties of the resonance .
a relativistic non - local model for the inclusive photoproduction of @xmath0 mesons from complex nuclei is introduced .
this model is based on the dominance of the contributions due to the formation of the s@xmath1 resonance .
the inclusion of non - local effects is the main improvement over our earlier work which was applied to the limited data available at the time .
moreover , the non - local approach makes it possible to include mass and width parameters of the resonance that may depend on the nuclear density .
calculations were first carried out using the free mass and width for the resonance ( i.e. no medium effect ) .
comparison of the results of these calculations with the data available for c , al and cu , reproduce the over all shape of the angular distributions as well as the total cross section data up to 1.0 gev .
it is worth noting that our results are in somewhat better agreement with the recent data of lns than with the older data .
the influence of possible medium modifications of the properties of the resonance was also studied .
simple density dependent forms for the mass and width of the resonance were used to test the effect these have on the calculated cross sections and their accord with the data .
comparison with the data showed that reductions in the mass of s@xmath1 on a scale similar to that suggested for the nucleon leads to cross sections that disagree strongly with the data .
the data can accommodate much weaker changes in the mass and some broadening of the width . although one can claim some scattered instances of improvement with these changes , it is not possible to claim that the data show clear evidence of medium modifications of resonance parameters .
the lack of this evidence is due in part to the occasional disagreement between different measurements .
there is also the inherent weakness in our present knowledge of the final state interaction of the eta meson with nuclei ( represented here by the optical potential ) , which adds to the uncertainty . in conclusion
we would estimate that if any medium modifications of the properties of the s@xmath1 resonance existed , they would likely be rather small .
this is in agreement with the general conclusions reached by the authors of refs.@xcite .
we would like to thank t. yorita and t. kinoshita for their helpful communications regarding their data .
we also are grateful to e. sumber , from computing and network services at university of alberta for parallelizing our computer code and to west grid for allowing us to use their facilities to perform our computations .
this work was supported in part by the natural sciences and engineering research council of canada .
brown and m. rho , phys . rev .
lett * 66*(1991)2720 ; phys . rep . 398
( 2004)301 .
bi pin - zhen and j rafelski , nucl - th/0507037 u. mosel , nucl - th/0507050 k. saito , k. tsushima and a.w .
thomas hep - ph/0506314 d. adamvo _ et al .
lett * 91*(2003 ) 042301 .
cbelsa / taps collaboration , phys .
lett * 94*(2005)192303 .
b. krusche , acta phys . hung .
* a 19*(2004)000 - 000 . m. roebig - landau _
et al . _ ,
b373 * ( 1996 ) 45 .
t. yorita _
et al . _ ,
phys . lett .
* 476 * ( 2000 ) 226 ; h. yamazaki _ et al . _ ,
a670 * , ( 2000)202 . t. kinoshita _
et al._,nucl - ex/0509022 .
j. lehr , m. post and u. mosel , phys .
rev . c*68 * ( 2003 ) 044601 : j. lehr and u. mosel , phys .
rev . c*68 * ( 2003)044603 : t. maruyama and s. chiba , prog .
phys . * 111*(2004)229 .
m. benmerrouche , nimai c. mukhopadhyay , and j.f .
zhang , phys .
rev . d*51 * , ( 1995 ) 3237 b.d .
serot and j.d .
walecka , _ advances in nuclear physics _ , edited by j.w .
negele and e. vogt , * 16*(1986)1 .
m. hedayati - poor , s. bayegan and h.s .
sherif , phys .
c*68 * , ( 2003 ) 045205 .
m. hedayati - poor and h.s .
sherif , nucl .
a740 * , ( 2004 ) 309 .
m. hedayati - poor and h.s .
sherif , phys .
c*56 * , ( 1997 ) 1557 . c. bennhold , h. tanabe , nucl
m. hedayati - poor and h.s .
sherif , phys .
c * 58 * ( 1998 ) 326 .
k. saito and a. w. thomas , phys .
rev . c*51 * ( 1995)2757 .
cooper and o.v .
maxwell , nucl .
* a493 * , ( 1989 ) 486 .
lee , l.e .
wright , c. bennhold and l. tiator , nucl .
* a603 * ( 1996 ) 345 .
i.r . blokland and h.s .
sherif , nucl . phys .
* a694 * ( 2001 ) 337 .
chiang , e. oset and l.c .
liu , phys . rev .
* c44 * ( 1991)738 .
table 1 . strengths , reduced radii and diffuseness for the relativistic scalar and vector mean - field potentials , respectively .
1 . differential ( a ) and total ( b ) inclusive cross section for @xmath51 reaction .
solid curves : s@xmath1 resonance contribution , dashed curves : sum of contributions from s@xmath1 resonance , d@xmath50 resonance , nucleon pole and vector meson .
2 . differential inclusive cross section for @xmath52 reaction on c , al and cu nuclei .
the graphs in the left and right panels represent the results for incident photon energies of e@xmath49 = 740 mev and e@xmath49 = 980 mev , respectively .
solid curves : calculations using the values 1535 mev and 160 mev for the mass and width of s@xmath1 resonance , respectively ( labeled free in this and subsequent figures ) , dashed curves : density dependent width of eq([eq8 ] ) with c = 50 mev ( labeled @xmath53 ) .
optical potential dw1 of ref.@xcite is used .
data are ( kek ) from @xcite , ( lns ) from @xcite and ( mainz ) from @xcite . fig .
3 . same as fig.2 ,
but for the effect of mass modifications .
solid curves : free calculations , dashed curves : density dependent mass of eq([eq7 ] ) with b = 0.014 , dash - dotted curves : density dependent mass of eq([eq7 ] ) with b = 0.14 .
4 . total inclusive cross section for the @xmath51 reaction .
data and potentials are from the same references as in fig.2 .
fig.4(a ) ; solid curve : free calculations , dashed curve : density dependent mass of eq([eq7 ] ) with b=0.014 , dash - dotted curve : density dependent mass of eq([eq7 ] ) with b=0.035 .
fig.4(b ) ; solid curve : free calculation , dashed curve : density dependent width of eq([eq8 ] ) with c = 50 mev , dash - dotted curve : density dependent width of eq([eq8 ] ) with c = 25 mev .
same as fig.4 but for al . in fig.5(a ) : solid curve : free calculations , dashed curve : density dependent mass of eq([eq7 ] ) with b=0.014 , dash - dotted curve : density dependent mass of eq([eq7 ] ) with b=0.14 . in fig.5(b ) : solid curve : free calculations , dashed curve : density dependent width of eq([eq8 ] ) with c = 50 mev , data are those of ref.@xcite .
same as fig.4 .
curves in panel ( a ) are contributions from s@xmath1 resonance , solid curve : free calculations , dashed curve : free width of 160 mev and density dependent mass of eq([eq7 ] ) with b=0.014 , dash - dotted curve : free width of 150 mev and density dependent mass of eq([eq7 ] ) with b=0.014 .
the contributions of nucleon poles , d@xmath50 resonance and vector meson diagrams are added to the ones for s@xmath1 in the calculations presented in panel(b ) .
curves are labeled as in panel(a ) .
same as the free calculations in fig.4 ( s@xmath1 resonance contributions only ) .
solid curve : using optical potential dw1 of ref.@xcite , dashed curve : using optical potential of ref.@xcite with the real part of the s@xmath1 self - energy set to -50 mev .
|
a relativistic non - local model for the inclusive photoproduction of @xmath0 mesons from complex nuclei is introduced .
the model is based on the dominance of the s@xmath1(1535 ) resonance .
we compare the results of our calculations with the available data on inclusive cross sections for the nuclei c , al and cu . assuming the resonance propagates freely in the nuclear medium , we find that the calculated angular distribution and energy dependence of the cross sections reproduce the data in a reasonable fashion .
the present non - local model allows the inclusion of density dependent mass and width in the calculations . including these in the calculations
reveals that the presently available data do not show clear preference for the inclusion of such modifications of the properties of the s@xmath1(1535 ) in the nuclear medium .
pacs number(s ) 25.20.lj , 24.10.jv , 14.20.gk , 13.60.le [ intro ]
|
low mass x - ray binaries ( lmxbs ) contain a neutron star ( ns ) or black hole ( bh ) primary onto which material is transferred from a low - mass ( @xmath2 ) late - type , essentially normal main sequence ( ms ) star ( spectral type @xmath0a - m ) , with longer period systems containing a sub - giant .
there are @xmath0190 luminous ( @xmath3 ergs@xmath4 ) lmxbs known in our galaxy ( liu , van paradijs & van den heuvel 2007 ) .
cataclysmic variables ( cvs ) are very similar , but have white dwarf ( wd ) primaries at a @xmath5 factor reduction in luminosity and have orbital periods that range from hours to @xmath6 d. while the flux from lmxbs is predominantly in x - rays which originate from the inner accretion disc and the ns surface ( where applicable ) , the flux from cvs is predominantly in the optical and originates from the entire accretion disc , the hot spot(s ) on the wd surface and the disc - mass - transfer stream impact region .
( for details see e.g. frank , king & raine 1992 ) . long - term , non - orbital , quasi - periodic variations on timescales of @xmath0tens - hundreds of days in lmxbs were discovered early in x - ray astronomy ( see charles et al .
2008 for a recent review ) .
these `` super - orbital '' periods are thought to be related to the properties of the accretion disc , and there are two basic mechanisms actively under consideration : radiation - induced warping ( ogilvie & dubus 2001 ) and precession ( whitehurst & king 1991 ) of the accretion disc .
radiation pressure from the intense x - rays arising near the compact object causes the warping of the accretion disc , while tidal forces in high mass - ratio binaries lead to the precession of the accretion disc ( an effect first observed in cvs , see e.g. warner 1995 ) .
either of these effects can lead to the periodic obscuration of the compact object in x - ray binaries , causing a super - orbital modulation in the x - ray flux ( clarkson et al . 2003 ) .
the stability criteria established by ogilvie & dubus ( 2001 ) suggest that this effect should not arise in most lmxbs , however it is observed , for example , in the long - period lmxb cygx-2 ( clarkson et al . 2003 ) .
the presence of even longer - term quasi - periodic variations ( on the order of decades ) has been predicted in cvs as a consequence of the modulation in the mass transfer - rate due to a solar - type magnetic - activity cycle occurring in the donor star ( applegate & patterson 1987 , warner 1988 ) .
the donor stars in short - period lmxbs and cvs are tidally locked and therefore rotate at the orbital period of the binary system , which is @xmath7 d for most systems .
this rapid rotation is expected to generate much stronger magnetic fields ( schrijver & zwaan 1992 ) .
long - term variability changes in the surface activity of stars ( such as starspot activity ) have been detected on timescales of decades ( baliunas 1985 ) , similar to the @xmath011 yr magnetic - activity cycle of our sun .
applegate ( 1992 ) suggested that magnetically active donors become more oblate as their outer layers are spun up , due to angular momentum distribution changes brought about by their magnetic activity . as a result , the volume of the roche lobe changes during the magnetic cycle , while the volume of the donor remains unchanged .
the magnetic activity cycle therefore governs the roche - lobe volume and the structure of the donor , which will be varying in oblateness as the cycle progresses .
such a variation will modulate the amount by which the donor overfills the roche lobe and therefore the mass transfer rate will also vary on the magnetic - activity time - scale ( as discussed by richman , applegate & patterson 1994 for cvs , hereafter r94 ) .
considering the similarities between lmxbs and cvs , it is not unreasonable to expect similar longer - term modulations in lmxbs for the same reason . online archival datasets now allow for a detailed investigation of long - term variations in the lightcurves of x - ray sources . here
we report on an analysis of the @xmath013-year x - ray history of lmxbs contained in the rxte / asm datasets . shortly before completing this paper , we became aware of durant et al ( 2009 ) .
while they mention similar long - term timescales of the x - ray flux modulations in the 16 brightest persistent lmxbs contained in the rxte asm datasets , we propose here a mechanism which would explain the observed dichotomy between atoll and z sources , and for the origin of these long time - scales .
the all sky monitor ( asm ) aboard the rossi x - ray timing explorer ( rxte ) is provided and operated by the massachusetts institute of technology ( mit ) .
the asm has observed the x - ray sky since early 1996 by scanning approximately 80% of the sky during each orbit of the satellite , providing monitoring of a source every @xmath090 minutes for at least 90 seconds .
the asm contains three rotating scanning shadow cameras ( ssc ) , allowing positional measurement of previously unknown sources to within 3 arcmin precision and providing individual monitoring sessions , which are referred to as `` dwells '' .
data are available in four energy bands : 1.5 - 3 kev ( a ) , 3 - 5 kev ( b ) , 5 - 12 kev ( c ) and a sum - band 1.5 - 12 kev ( a+b+c ) .
data for each energy band are reduced independently by taking background effects into account and made available as dwell - by - dwell or one - day - averages .
data are reduced and compiled weekly by the asm team and made publically available . for full details
see levine et al .
( 1996 ) .
the asm monitoring began on 20 february 1996 , and for this paper we were able to employ datasets spanning more than 13 years .
therefore these archival datasets contain the long - term intensity history of all x - ray sources in the rxte catalogue . the fact that rxte is sensitive in the soft x - ray regime ( 2 - 10 kev ) is particularly useful for the studies of lmxbs , since their largest flux contribution comes from soft x - rays .
the full asm lightcurves of all 44 significantly detected lmxbs ( average flux @xmath8 0.5 counts s@xmath4 ) were considered for further analysis .
transients were excluded ( aql x-1 , groj1655 - 40 , gx339 - 4 , h1743 - 322 , 4u1543 - 47 , 4u1608 - 52 , 4u1630 - 47 , xtej1550 - 564 , xtej1701 - 462 and xtej1859 + 226 ) , since their outburst(s ) are the only reason these sources were detected above the @xmath9 level .
4u1820 - 30 was excluded , since it is almost certainly a triple system ( chou & grindlay 2001 ) , which would have additional dynamics compared to the other lmxbs and is therefore not relevant in our analysis here .
cir x-1 was also excluded , since its highly eccentric orbit ( murdin et al .
1980 ) will introduce large phase - dependent changes in the donor s roche lobe .
asm one - day - average sum - band lightcurve data were plotted on scales appropriate for showing long - term behaviour , and from visual inspection of those lightcurves , it became apparent that some sources display large amplitude , very long - term modulations that appear periodic or quasi - periodic in nature .
they are : gx3 + 1 , gx9 + 1 , gx9 + 9 , gx354 - 0 , 4u1636 - 536 , 4u1708 - 40 , 4u1735 - 444 , 4u1746 - 37 and serx-1 .
these are all classified as atoll sources , with the exception of the x - ray burster 4u1708 - 40 .
re - examining all the atoll sources , there appear to be very long - term modulations in all of them , although to a lesser extent than those sources listed above .
furthermore , all the z sources display remarkably steady lightcurves over the long - term . for the plots , the liu et al .
( 2007 ) classifications for lmxbs were added in square brackets ( see the caption of fig .
[ lc1 ] for a description of these classifications ) .
our subsequent analysis focussed on the z and atoll sources , since they appear to represent the two extremes in long - term periodic behaviour .
therefore , the remaining sources ( 4u1254 - 69 [ b , d,(sb ) ] , 4u1556 - 60 , 4u1624 - 49 [ d ] , her x-1 [ p , d , e ] , 4u1708 - 40 [ b ] , mxb1730 - 335 [ t , g , b , d , r ] , ks1731 - 260 [ t , b,(sb ) ] , 4u1822 - 000 , 2a1822 - 371 [ p , e ] , gs1826 - 238 [ t , b ] , grs1915 + 105 [ t , d , m , r ] , 4u1957 + 11 [ u ] ) were not considered further .
the variability analysis was conducted on asm dwell - by - dwell data , which were rebinned into 10-day bins , using the prescribed filters@xmath10 for constructing asm averages .
these are : @xmath11 ( @xmath12 for scox-1 ) , number of sources in the field of view @xmath13 , earth angle @xmath14 , exposure time @xmath15 seconds , long - axis angle : @xmath16@xmath17 , short - axis angle : @xmath18@xmath19 .
additional filters applied are : background counts @xmath20 , hardness ratio : @xmath21 , flux error @xmath22 , number of datapoints per bin @xmath23 to give the most reliable dataset .
we include data up to 13 august 2009 ( @xmath24 mjd @xmath25 ) .
since the long - term modulations displayed in these lightcurves have timescales that exceed or are comparable to the observational baseline , we can not make use of the usual period analysis tools such as periodograms . in order to estimate the time - scale of these long - term variations ,
single sine - waves were fitted to all the atoll sources . in the case of gx9 + 9 a linear term
was added ( cf .
harris et al .
2009 , who also noted the presence of the @xmath0 1500 d modulation ) .
the results for the 8 atoll sources which display the most significant long - term modulations , are contained in table [ comparea1 ] ( and their lightcurves in fig .
[ lc1 ] ) .
the results for the remaining atoll sources are included in table [ comparea2 ] ( and fig .
[ lc2 ] ) .
although the z sources appear to display remarkably steady lightcurves over the long - term , they can also be fitted with single sine - waves , but with much lower amplitudes .
the results of their fits are contained in table [ comparez ] and fig .
[ lc3 ] .
.asm properties of the 8 significantly modulating atoll sources [ cols="<,^,^,^,^,^ " , ] uncertainties in the values obtained are included in the tables in parentheses .
the binned data flux errors were adjusted by a factor , so as to obtain fits for which @xmath26 .
this gives more sensible errors on the sine - wave parameters , since the errorbars for the binned data are extremely small in comparison to the larger flux variations ( which dominate ) .
the factors applied , to the sources that show the largest scale systematic variations , were : 43.4 for cyg x-2 , 36.8 for 4u 1705 - 44 and 32.9 for sco x-1 .
the factors applied to the rest of the sources range from 1.5 - 14.7 , which are comparable to the factor of 3.1 required for a constant fit to the crab .
the f - statistic was calculated for the 2 models of a constant fit and a single sine - wave fit to the data . for gx9 + 9
the simpler model considered was a linear variation with time .
the calculated f - values are included in tables [ comparea1 ] , [ comparea2 ] & [ comparez ] .
f - values @xmath27 are obtained , indicating that the inclusion of additional terms ( the sine wave ) is highly significant ( @xmath28 confidence ) .
the f - values are highest in those sources where we could detect long - term variation by simple visual inspection .
even the lowest f - values determined @xmath29 are still highly significant , indicating that all the sources considered here were better fitted with a single sine - wave than with a constant value .
flux modulations of @xmath30 of the average flux values are present in the atoll sources , with the exception of the superburster 4u1636 - 536 and the burster gx13 + 1 .
it is possible that the modulation detected in 4u1636 - 536 represents a superburst , rather than the long - term modulations we consider to be present in the other sources ( which may also be the case for ks1731 - 260 ) .
gx13 + 1 is classified as an atoll source , but it shares certain properties with z sources ( liu et al .
indeed , we find that its amplitude for the very long - term modulation rather agrees with those found in the z sources , than with those obtained for the atoll sources .
flux modulations are @xmath3113% for the z sources , with the brighter sources having the lower amplitudes .
both atoll and z sources contain ns primaries , but z sources have fluxes that are @xmath00.5 - 1@xmath32 , whereas atoll sources and bursters have fluxes @xmath00.01 - 0.5@xmath32 ( van der klis 2006 ) .
once at the eddington limit ( @xmath32 ) , z sources are unlikely to show any x - ray flux modulation due to additional changes in the mass transfer rate .
however , atoll sources would be expected to modulate their flux in response to overall changes in the mass transfer rate .
the results show that , in general , atoll sources have larger amplitude in the very long - term modulations than z sources , which occur on very long - term timescales in both types . a mechanism for modulating the mass transfer rate over long timescales has been proposed for cvs ( applegate & patterson 1987 , warner 1988 ) .
given the similarities between cvs and lmxbs , we decided to investigate whether this mechanism might also be applicable to lmxbs .
our current x - ray observational baseline for lmxbs is too short to cover multiple cycles and thereby establish the stability of these variations . however , given the similarity of these timescales to those exhibited by solar - type stars , we use the approach of r94 for cvs , and see if it is also applicable to lmxbs .
the mechanism proposed in r94 to be responsible for long - term variations in cvs is due to magnetic activity cycles in the donors .
they calculate the variation in the mass transfer rate ( @xmath33 ) which is associated with the observed orbital period variation ( @xmath34 ) brought about by this mechanism in cvs .
r94 proposed that changes in the rotation of a thin outer shell ( mass @xmath35 ) of the donor ( mass @xmath36 ) , rotating with angular velocity ( @xmath37 ) , will affect the orbital period .
they calculate that : @xmath38 where @xmath39 .
they noted that the applegate ( 1992 ) variable differential rotation rates follow the hall ( 1990 ) and hall ( 1991 ) differential rotation - orbital period relation , and consequently applied that relation to the orbital periods for cvs to obtain @xmath40 .
they furthermore assume that @xmath41 and calculate a @xmath33 which is consistent with the observed long - term flux variations in cvs , but consider the observed @xmath34 to be the best evidence for long - term cycles with a preferred timescale of decades ( 5 - 30 yrs ) , reminiscent of solar - like magnetic cycles ( warner 1988 ) .
they also find that cvs show quasi - periodic brightness fluctuations over that timescale .
we apply equation ( 1 ) to the lmxb gx9 + 9 , which has @xmath42 hrs ( kong et al .
2006 ) , and for which @xmath43 has been found spectroscopically by cornelisse et al .
we consequently obtain @xmath44 .
cornelisse et al .
( 2008 ) found no significant change in orbital period over the @xmath011 yr rxte / asm baseline .
considering the result for @xmath34 , we would not expect to detect a change in the orbital period in the rxte / asm dwell - by - dwell data , since the @xmath45 implied is @xmath0 60 times smaller than the error in @xmath1 .
such a change in p will cause a corresponding change in the size of the donor s roche lobe ( @xmath46 ) and therefore modulate the mass transfer rate @xmath47 by an amount determined by r94 @xmath48 where @xmath49 is the photospheric scale - height of a main sequence donor .
assuming the standard paczyski ( 1971 ) relation for the size of the donor s roche lobe gives @xmath50 and if the donor in gx9 + 9 also follows the cv secondary relation of @xmath51 , then @xmath52 for @xmath53 ( see r94 ) .
therefore a maximum flux modulation of @xmath54 is expected by virtue of the orbital period variation and we actually observe @xmath0 8 @xmath55 over a period @xmath56 yrs .
in fact , this result regarding the expected maximum flux modulation will apply to all the lmxbs mentioned in this paper , for the whole range of long - term periods determined from the fitted sine - waves .
the flux modulations determined from the fitted sine - waves for the sources considered are @xmath57 .
it is therefore quite plausible that the long - term modulations observed in the rxte asm lightcurves of the lmxbs , originate from magnetic activity cycles in the donor , just as has been proposed for cvs .
long - term variation in the mass transfer rate due to the magnetic activity cycle of the donor , should therefore translate into long - term quasi - periodic modulations of the x - ray flux for sources in which the additional material can be accreted onto the ns without violating the eddington limit , such as atoll sources and bursters .
however , in z sources very little ( if any ) of the additional material will be accreted and much lower amplitude ( if any ) long - term x - ray flux modulations are expected as a result of the magnetic activity cycle of the donor . therefore , we conclude that rxte asm lightcurve data may now provide the first evidence for very long - term quasi - periodic modulations of the x - ray flux as a result of the modulation in the mass transfer - rate due to a solar - type magnetic - activity cycle in the donor star , similar to those proposed for cvs .
asm results were provided by the asm / rxte teams at mit and at the rxte science operations facility and guest observer facility at nasa s goddard space flight center ( gsfc ) .
we also wish to thank brian warner for his support and comments .
applegate j. h. , patterson j. , 1987 , apj , 322 , l99 applegate j. h. , 1992 , apj , 385 , 621 baliunas s. l. , vaughan a. h. , 1985 , araa , 23 , 379 charles p. a. , clarkson w. , cornelisse r. , shih c. , 2008 , new astron .
, 51 , 768 chou y. , grindlay j. e. , 2001 , apj , 563 , 934 clarkson w. i. , charles p. a. , coe m. j. , laycock s. , 2003 , mnras , 343 , 1213 cornelisse r. , steeghs d. , casares j. , charles p. a. , barnes a. d. , hyne r. i. , obrien k. , 2007 , mnras , 380 , 1219 durant m. , cornelisse r. , remillard r. , levine a. , 2009 , astro - ph/0909.0444v1 frank j. , king a. , raine d. , 1992 , in _ accretion power in astrophysics _ , cambridge univ .
press , p.1 hall d. s. , 1990 , in ibanoglu c. , ed , _ active close binaries _ , kluwer , dordrecht , p.95 hall d. s. , 1991 , apj , 380 , 85 harris r. j. , levine a. m. , durant m. , kong a. k. h. , charles p. a. , shahbaz t. , 2009 , apj , 696 , 1987 kong a. k. h. , charles p. a. , homer l. , kuulkers e. , odonoghue d. , 2006 , mnras , 368 , 781 levine a. m. , bradt h. , cui w. , jernigan j. g. , morgan e. h. , remillard r. , shirey r. e. , smith d. a. , 1996 , apj , 469 , 33 liu q. z. , van paradijs j. , van den heuvel e. p. j. , 2007 , a&a , 469 , 807 murdin p. , jauncey d. l. , lerche i. , nicolson g. d. , kaluzienski l. j. , holt s. s. , haynes r. f. , 1980 , a&a , 87 , 292 ogilvie g. i. , dubus g. , 2001 , mnras , 320 , 485 paczyski b. , 1971 , araa , 9 , 183 richman h. r. , applegate j. h. , patterson j. , 1994 , pasp , 106 , 1075 ( r94 ) schrijver c. j. , zwaan c. , 1992 , astron .
pacific conf .
ser . , 26 , 370 van der klis m. , 2006 , in lewin w. , van der klis m. , eds , _ compact stellar x - ray sources _ , cambridge univ .
press , cambridge , p. 46 warner b. , 1988 , nature , 336 , 129 warner b. , 1995 , in _ cataclysmic variable stars _
, cambridge univ . press ,
cambridge whitehurst r. , king a. r. , 1991 , mnras , 249 , 25
|
long - term monitoring of low mass x - ray binaries ( lmxbs ) by the all sky monitor on board the rossi x - ray timing explorer now covers @xmath013 yrs and shows that certain lmxb types display very long - term ( @xmath0 several to tens of years ) quasi - periodic modulations .
these timescales are much longer than any `` super - orbital '' periods reported hitherto and likely have a different origin .
we suggest here that they are due to long - term variations in the mass - transfer rate from the donor , which are a consequence of solar - like magnetic cycles that lead to @xmath1 changes ( as proposed by richman , applegate & patterson 1994 for similar long - term variations in cvs ) .
atoll sources display much larger amplitude modulations than z sources over these timescales , presumably because z sources are eddington limited and hence unable to respond as readily as atoll sources to fluctuations in the mass - transfer rate from the donor .
[ firstpage ]
|
the recent progress in experimental and theoretical studies of bose - einstein condensates ( becs ) of dilute atomic gases @xcite has been tremendous after their experimental realization @xcite .
this has also led to an explosion of interest in the theme of nonlinear matter - waves such as dark @xcite , bright @xcite and gap @xcite solitons .
two - dimensional ( 2d ) nonlinear excitations of becs , such as vortices @xcite and vortex lattices @xcite , were also realized experimentally , while a multitude of other coherent nonlinear structures were also theoretically predicted ; these include , e.g. , faraday waves @xcite , ring dark solitons and vortex necklaces @xcite , stable solitons and localized vortices in attractive becs trapped in periodic potentials @xcite , matter - wave gap vortices @xcite , 2d bright solitons in dipolar becs @xcite , and so on . from the theoretical standpoint ,
the dynamics of such higher - dimensional structures is , generally , difficult to be treated analytically , especially so in time - dependent settings .
this , in turn , raises important questions concerning their `` controllability '' , which would be of particular relevance regarding potential applications .
ideally , such a controllability would allow `` manipulation '' of the condensates at will , e.g. , sustaining condensates which may expand towards complete dispersion , contract towards a complete collapse , or perform stable breathing oscillations .
moreover , such processes could involve a targeted growth of a condensate up to a certain width or shrinkage down a desired size .
our purpose in the present work is to illustrate how one can achieve this goal , by taking advantage of one of the few analytically tractable tools in higher - dimensional condensates , namely of the analysis of their moments @xcite ( see also references therein ) . in particular , we focus on the case of quasi two - dimensional ( 2d ) , so - called , pancake - shaped condensates @xcite , under the effect of time - varying harmonic trapping @xcite and also time - dependent s - wave scattering length ( which controls the inter - particle interactions ) . note that the controllability over the harmonic trapping is straightforwardly realizable under ac - variation of the atom trapping frequencies , while the controllabity over the interatomic interactions can be realized by using the so - called feshbach resonance @xcite , connecting the s - wave scattering length to external magnetic fields .
this latter technique ( which is usually called feshbach resonance management ( frm ) ) has been proposed theoretically as a means of avoiding collapse @xcite , but also as a way of producing robust coherent nonlinear matter - waves @xcite . for the quasi-2d condensates discussed above , upon presenting the moment analysis
, we obtain a reduced dynamical description effectively involving only a variable associated with the width of the condensate wavefunction . this ordinary differential equation ( ode )
can be reduced to one of the ermakov - pinney ( ep ) type @xcite , whose solutions can be obtained _ analytically _ , provided that the solutions of the underlying linear schrdinger equation can be obtained .
we use this feature and the freedom in selecting the time - dependent trapping and interactions of the condensates to illustrate that one can construct analytical solutions to this ode that completely characterize the temporal evolution of the width of the wavefunction . in so doing ,
we fully prescribe the dynamical evolution of the condensate .
we show three prototypical examples in applying this idea : one in which the width grows in time ( leading to condensate expansion ) , one in which it decreases ( leading to focusing ) , and one in which it periodically `` breathes '' between a minimum and a maximum value . in all three cases , we test the analytical prediction against the full numerical simulation of the mean - field partial differential equation ( pde ) model fully describing the condensate .
we find excellent agreement between the two , showcasing the accuracy of our theoretical approach .
the analysis of the ensuing ep equation in this setting is of interest in its own right .
this is because , as we will see , the resulting ep equation has a time - dependent nonlinearity in the right hand side ( contrary to what is the case for the `` standard '' ep framework ; see e.g. @xcite and references therein ) . in this context , and as an aside , we present a second physically relevant example where such ep equations with time - dependent nonlinearities may arise , by studying anisotropic scalar field cosmologies of a particular anisotropic geometry .
this generalizes the example of @xcite parallelizing becs without frm and scalar field cosmologies in the isotropic case .
our presentation will be structured as follows : in section iii , we provide a brief synopsis of the main features of the ep equation . in section iv
, we proceed to an overview of the moment analysis , following the earlier works @xcite . in section iv
, we present the analytical solutions of the ep equation that we develop for each of the above - mentioned three cases .
section v tests these analytical results against full simulations of the pde describing the condensate .
section vi illustrates the second example of time - dependent ep equations in the description of anisotropic scalar field cosmologies . finally ,
in section vii , we summarize our findings and present our conclusions .
the ermakov - pinney ( ep ) equation is a remarkable nonlinear ode of the form : @xmath0 the particularly attractive feature of this nonlinear ode is that its general solution can be obtained , provided that one is able to solve the time - independent linear schrdinger ( ls ) equation @xmath1 . for details on the properties of the ep equation ,
the interested reader is referred to @xcite and references therein .
here we just mention its basic superposition principle property .
namely , if the linearly independent solutions of the ls equation are @xmath2 and @xmath3 , then the most general possible solution of the ep equation is given by @xmath4 where @xmath5 and @xmath6 are constants connected through @xmath7 where @xmath8 is the wronskian of @xmath2 and @xmath3 .
one of the interesting variants of the `` regular '' ep equation of the form ( [ oeq0 ] ) arises in the study of becs , albeit in a somewhat modified form ( see below ) . the relevant mean - field model for studying atomic bose - einstein condensates at zero temperature consists of the so - called gross - pitaevskii equation @xcite of the following dimensionless form : @xmath9 where @xmath10 represents the wavefunction of the condensate , @xmath11 denotes the harmonic trap confining the bosons [ with a time - dependent frequency determined by @xmath12 and @xmath13 is the coefficient of the nonlinear term proportional to the s - wave scattering length , characterizing the interaction between the particles .
we will take advantage of the feshbach resonance @xcite to consider that the latter is time - dependent , as well .
one of the popular approaches to studying eq .
( [ eq1 ] ) is through the use of moment methods @xcite .
the latter allow us to write odes for the moments of the mean - field wavefunction @xmath10 as follows .
we define @xmath14 where subscripts denote partial differentiation ,
@xmath15 denotes complex conjugate and @xmath16 indexes the dimension ( similarly to @xcite ) .
the hamiltonian of eq .
( [ eq1 ] ) @xmath17 r^{d-1}dr , \label{ham}\end{aligned}\ ] ] can then be written as : @xmath18 one can then infer from the dynamics of eq .
( [ eq1 ] ) that : @xmath19 however , as derived in @xcite , @xmath20 we focus on @xmath21 , since this moment is associated with the width of the spatial profile of the wavefunction . from the latter , we can infer a number of useful pieces of information concerning certain asymptotic values ; in particular , if @xmath22 , the condensate collapses , while if @xmath23 , the bec disperses .
furthermore , since the @xmath24 norm of the wavefunction is conserved , the estimate of @xmath21 on the ( square ) width of the wavefunction can be used together with this conservation law to provide information on the amplitude @xmath25 of the wavefunction ( i.e. , approximately @xmath26 ) . combining the two equations ( [ deriv ] ) and ( [ virial ] ) , we obtain a single equation for the time - dependence of @xmath21 as @xmath27 \label{three}\end{aligned}\ ] ] following @xcite
, we consider a quadratic phase for the solution ( an assumption most relevant to dimension @xmath28 as discussed in @xcite , but which we will also consider more generally ) , and obtain ( cf .
( 6d ) of @xcite ) that @xmath29 where the constant @xmath30 is determined by initial conditions .
notice that this is the _ only _ assumption in our calculations herein , whose validity will be examined a posteriori by comparing our analytical results with numerical computations . denoting for simplicity @xmath31 , the resulting ordinary differential equation for @xmath32
can be written as : @xmath33 clearly , from the above exposition , the most straightforward case is the one with @xmath28 , on which we will focus next .
when @xmath28 , the equation can be directly integrated , yielding @xmath34 where @xmath6 is an integration constant that can be computed from the initial conditions as : @xmath35 if we now use the transformation @xmath36 @xcite , then we obtain an ermakov - pinney ( ep ) type equation of the form : @xmath37 hence , eq .
( [ ode3 ] ) is the equation that describes the dynamics of two - dimensional bose - einstein condensates in the presence of a time - dependent trap @xcite , as well as in the scenario of feshbach resonance management @xcite .
one can now examine particular cases of time - dependence of @xmath38 and @xmath13 , using the analytical tractability of the ensuing ep equation , in order to obtain completely analytical solutions for @xmath39 , and hence for @xmath40 .
there are numerous possiblities in the case of a time - dependent @xmath41 and @xmath42 , thus , as explained above , we limited our investigation to the three cases of an expanding waveform , a collapsing waveform and an oscillatory waveform . in order to derive examples for each of these cases , @xmath13 was chosen to be independent of the choice of @xmath38 and @xmath43 .
solutions to eqn .
( [ ode3 ] ) were determined by `` reverse engineering '' @xcite ( based on the desired behavior of the dynamics of the wavefunction ) , and the details of the ep functional form .
below are the three cases examined .
notice that in all three cases , the function @xmath44 used below is given by @xmath45 i.e. , by the numerator of the right hand side of eq .
( [ ode3 ] ) . 1 .
the first case explores the possibility of an expanding width wavefunction . in order for the wavefunction width to increase
, @xmath38 must be decreasing , thus test functions for @xmath38 and @xmath43 were chosen to reflect this .
a similar methodology was also used in the two other cases to determine test functions .
@xmath46 2 . the second case , a focusing wavefunction , required an additional condition ( @xmath47 ) in order to perform the numerical simulations described below .
additionally , in this case and the next , an initially undetermined function @xmath48 was used in @xmath38 to simplify the derivation of @xmath38 and @xmath43 .
after these were found , the exact form of @xmath48 was calculated by plugging @xmath38 and @xmath43 into eqn .
( [ ode3 ] ) and solving for @xmath48 .
this results in @xmath49 \frac{{b^q } } { { \left [ 2(a^2 + t^p ) \right]^{q/2 } } } \label{case2b } \\
z(t ) & = & \frac{{b^5 } } { { ( a^2 + t^p ) } } \left [ { \frac{{b^4 pt^{p - 2 } ( pt^p - 2pa^2 + 2a^2 + 2t^p } } { { 4(a^2
+ t^p ) ^4 } } + \frac{{f(t)(a^2 + t^p ) ^{5/2 } } } { { b^5 } } - f(t ) } \right ] \label{case2c}\end{aligned}\ ] ] 3 . finally ,
in the last case , equations for @xmath38 and @xmath43 were calculated in order to generate a wavefunction with oscillatory width .
the corresponding functions in this case read @xmath50 ^ 2 } } \label{case3b } \\
z(t ) & = & - \frac{{c^2 } } { 4}\sin ^2 ( 2ct ) + c^2 \cos ( 2ct ) \left [ { b + \sin ^2 ( ct ) } \right ] \label{case3c}\end{aligned}\ ] ] the resulting form of the normalized confining frequency @xmath38 for the three different cases is shown in fig .
[ figlambda ] .
we note in passing that such schemes are definitely realizable within the unprecedented control that exists over the magnetic confinement of the condensates . a very recent example illustrating
this point can be found in the recent work of @xcite ; this experiment clearly realizes a form of the third type of confinement ( an oscillatory one ) , by using a periodic modulation of the transverse confinement of the condensate , in order to induce longitudinal oscillations and the emergence of faraday patterns .
this directly shows that temporal modulation of the trapping frequencies is feasible .
what we propose here consists of the three principal types of trapping frequency dependence , i.e. , monotonic decrease , monotonic increase and oscillatory dependence , that allow an explicit analytical handle on the dynamics of the bec , in excellent agreement with our numerical findings ( see below ) .
hence , we expect that such schemes would be directly applicable in experimental settings similar to those of the above experiment . for cases 1 - 3 , i.e. , respectively expanding ( left ) , contracting ( middle ) and oscillatory ( right )
condensates.,title="fig:",width=188,height=188 ] for cases 1 - 3 , i.e. , respectively expanding ( left ) , contracting ( middle ) and oscillatory ( right ) condensates.,title="fig:",width=188,height=188 ] for cases 1 - 3 , i.e. , respectively expanding ( left ) , contracting ( middle ) and oscillatory ( right ) condensates.,title="fig:",width=188,height=188 ]
in order to validate our analytical results based on the predictions made above , the evolution of the condensates under the time - dependent trappings imposed by eqs .
( [ case1b ] ) , ( [ case2b ] ) and ( [ case3b ] ) were tested by numerically solving the original pde model of eq .
( [ eq1 ] ) .
the numerical simulations were run using spectral methods @xcite , with a spatial grid composed of chebychev nodes ( @xmath51 where @xmath52 is the total number of nodes ) . since
the chebychev nodes are contained within the interval @xmath53 $ ] , the spatial variable was normalized in order for the solution @xmath54 to be supported on the interval @xmath53 $ ] from @xmath55 to some evolution horizon @xmath56 .
the temporal integration was implemented using a 4th - order runga - kutta scheme .
it is worth noting here that in all three cases examined , the choice of @xmath13 was independent of the equations for @xmath38 and @xmath43 , thus @xmath57 was chosen to be used for all cases , in consonance with the feshbach resonance management scheme discussed in the introduction ; in that sense , however , notice that the key temporal variation in the schemes presented herein is that of the magnetic trapping frequency .
as an initial condition to eqn .
( [ eq1 ] ) , we used a generic gaussian profile of the form @xmath58 with the values of @xmath25 and @xmath59 subject to the condition @xmath60 . for the cases of the expanding and collapsing width wavefunctions
, the numerical results are in perfect agreement with the analytical calculations as shown in figs .
( [ figcase1 ] ) and ( [ figcase2 ] ) .
notice in the lower left - hand plot for both figures how the lines for the analytical and numerical plots are completely indistinguishable from each other .
for the last case of the oscillatory width wavefunction , there was a very small discrepancy between the analytical calculations and numerical simulation , as seen in the lower left - hand plot of fig .
( [ figcase3 ] ) .
this discrepancy is only seen at the maximum values of the wavefunction width , starting with the second and subsequent maximums . while this small discrepancy may be triggered by the sole approximation of our approach , namely eq .
( [ proportionality ] ) , further numerical experiments ( not shown here ) seem to indicate that it is more likely to be the result of the numerical approximation to the corresponding moment . in any case
, the overall excellent numerical agreement between the analytically obtained moment @xmath40 and its numerically found counterpart illustrate the relevance of this approach and its usefulness in systematically prescribing the wavefunction behavior , in an analytically tractable way . [
cols="^,^,^ " , ]
an interesting analogy has been recently made between bose - einstein condensates and isotropic scalar field cosmologies in the absence of time dependence in the coefficient of the nonlinearity @xcite .
the so - called friedmann - robertson - walker ( frw ) metric , coupled to a scalar field had been shown earlier @xcite to be described by an ep equation and that viewpoint was exploited in @xcite to illustrate an analogy between cosmic dynamics and becs . here , we extend this analogy to ep equations with time - dependent nonlinearities .
in particular , we show that anisotropic cosmic dynamics can be parallelized with feshbach - managed bose condensates .
if we consider the spatially homogeneous , yet anisotropic geometry of bianchi type i , we have a line element of : @xmath61 we can define the tensor : @xmath62 where @xmath63 is the einstein tensor and @xmath64 is the energy momentum tensor . then , the quadratic constraint is the equation @xmath65 , the kinematic equation is given by @xmath66 , while the klein - gordon equation for the scalar field ( that is coupled to gravity ) is given by @xmath67 .
notice additionally , that the two integrals of the motion , namely @xmath68 and @xmath69 yield @xmath70 and @xmath71 . solving the klein - gordon equation for @xmath72 and substituting the result into @xmath73 ( as well as solving @xmath65 for @xmath74 and using the resulting expression in @xmath73 ) , one
is led to a dynamical equation for the remaining scale factor @xmath75 in the form : @xmath76 using now : @xmath77 and a change of variable @xmath78 , we obtain : @xmath79 hence , a choice of time reparametrization according to : @xmath80 ( which leads to @xmath81 , where @xmath82 is a constant of integration ) , results in the form : @xmath83 eq .
( [ gop4 ] ) becomes an ep equation for the choice of @xmath84 .
notice that in eq .
( [ gop4 ] ) , a nontrivial complication is that the time dependent coefficient @xmath85 depends on the reparametrization of time through @xmath86 .
nevertheless , the original equation ( [ gop1 ] ) has been solved in @xcite .
this may , in turn , provide valuable insights in the solution of such time - dependent ep equations ; this would be an interesting direction for future studies in such time - dependent ep settings .
we close this section by noting that one can therefore generalize the analogy of bose condensates and scalar field cosmologies in the feshbach - managed viz .
anisotropic case of the ep equation with time - dependent nonlinearity .
the quantities that now bear the analogy is the scale factor of bianchi type i with the second moment of the condensate wavefunction , the time - dependent magnetic trap strength with the time - dependent scalar field and finally the time dependent nonlinearity coefficient ( i.e. , the scattering length in the condensate dynamics ) is ( indirectly ) connected with the reparametrization of time in the cosmological problem .
in this short communication , we revisited the theme of higher dimensional bose - einstein condensates under the presence of magnetic trapping and feshbach resonance management .
we used the moment method to develop an ermakov - pinney ode for the second moment of the distribution function , which is associated with the width of the condensate .
this ep equation is analytically tractable , in a number of cases . in particular
, one can `` reverse engineer '' magnetic trappings that will induce the expansion , contraction or oscillatory behavior of the condensate at will . for such scenaria , the relevant moment of the wavefunction
can be obtained analytically and is found to be in excellent agreement with our numerical simulations of the full original pde model .
the approach permits a detailed analytical handle on the behavior of higher dimensional condensates which is usually quite difficult to acquire with different methods based on nonlinear pdes .
it may also , in turn , permit to appropriately craft experiments , based on the reshaping `` operation '' that is desirable to perform on the condensate . as an aside example
, we have illustrated the relevance of such time - dependent ermakov - pinney equations in a completely different physical system , namely in anisotropic scalar field cosmologies of bianchi type i. this approach also suggests a number of interesting questions .
it would be , in particular , relevant to examine whether a general solution of the time - dependent ep equation developed herein can be obtained on the basis of its linear schrdinger counterpart . this would bear direct consequences both in the atomic physics and in the cosmological problem , allowing for a general analytical description of their respective time - dependent properties ( for the condensate wavefunction width or the cosmological scale factor , respectively ) .
furthermore , the approach was presented here for two - dimensional settings for reasons that have to do with the simplification / closure of the moment approach in that case .
however , it would be of particular interest to develop similar approaches and potential closure schemes for one - dimensional or three - dimensional settings .
such studies are currently in progress and will be reported in future publications .
et al _ , science * 269 * , 198 ( 1995 ) ; k.b .
davis _ et al _ , phys .
lett . * 75 * , 3969 ( 1995 ) ; c.c .
bradley _ et al _ , phys .
lett . * 75 * , 1687 ( 1995 ) ; see also the reviews e. a. cornell and c. e. weiman , rev .
phys . * 74 * , 875 ( 2002 ) and w. ketterle , rev .
74 * , 1131 ( 2002 ) .
s. inouye _
et al . _ ,
nature * 392 * , 151 ( 1998 ) ; j. l. roberts _ et al .
_ , phys . rev .
* 81 * , 5109 ( 1998 ) ; j. stenger _
et al . _ ,
lett . * 82 * , 2422 ( 1999 ) ; s. l. cornish _ et al .
_ , phys .
lett . * 85 * , 1795 ( 2000 ) .
|
in this work we revisit the topic of two - dimensional bose - einstein condensates under the influence of time - dependent magnetic confinement and time - dependent scattering length .
a moment approach reduces the examination of moments of the wavefunction ( in particular , of its width ) to an ermakov - pinney ( ep ) ordinary differential equation ( ode ) .
we use the well - known structure of the solutions of this nonlinear ode to `` engineer '' trapping and interatomic interaction conditions that lead to condensates dispersing , breathing or even collapsing .
the advantage of the approach is that it is fully tractable analytically , in excellent agreement with our numerical observations . as an aside
, we also discuss how similar time - dependent ep equations may arise in the description of anisotropic scalar field cosmologies .
|
a photon collider has been proposed as a natural extension of the linear collider project tesla @xcite .
the physics potential of a photon collider is very rich and complementary to the physics program of the and hadron - hadron colliders .
it is an ideal place to study the mechanism of the electroweak symmetry breaking ( ewsb ) and the properties of the higgs - boson . in paper
@xcite we performed realistic simulation of sm higgs - boson production at the photon collider for @xmath0 and @xmath2 decay channels , for higgs - boson masses above 150 gev . due to an interference with a large standard model background ,
the process @xmath3 turns out to be sensitive not only to the higgs - boson partial width to , , but also to the phase of the @xmath4 coupling , .
a precise measurement of both and seems to be crucial for determination of the higgs - boson couplings , see also @xcite . in @xcite
we have found that this is extremely important to combine both @xmath0 and @xmath5 channels , as the first one due to a large background is is very sensitive to a phase , while the second - to a partial width . from combined analysis of @xmath0 and @xmath2 decay channels
the @xmath6 partial width can be measured with an accuracy of 3 to 8% and the phase of the amplitude with an accuracy between 30 and 100 mrad . in this paper
we extend this analysis by considering in detail various extensions of the standard model , in particular the sm - like two higgs doublet model 2hdm ii with and without cp - conservation .
we perform the combined analysis of @xmath0 and @xmath5 invariant - mass distributions and extract the corresponding higgs - boson couplings ( @xmath7 ) . for 2hdm model with a cp violation
, we estimate precision of the @xmath8 mixing angle . we also consider here a potential of establishing cp - properties of higgs - bosons in a general cp - violating model from a combined analysis of the angular distributions of the @xmath0 and @xmath5 decay - products . in this approach
a cp properties of the higgs boson can be studied independently on a production mechanism .
we consider a model with generic , cp - violating higgs couplings to vector bosons , leading to different angular distributions for a scalar- and pseudoscalar - type of couplings . from such measurement
the cp - parity of the observed higgs state can be determined .
precision in the determination of couplings can be significantly improved if measurement of angular distributions is combined with the invariant - mass one . however ,
in such case additional model assumptions are needed .
the compton back - scattering of a laser light off high - energy electron beams is considered as a source of high energy , highly polarized photon beams @xcite . according to the current design @xcite , the energy of the laser photons is assumed to be fixed for all considered electron - beam energies . with 100% circular polarization of laser photons and 85% longitudinal polarization of the electron beam the luminosity spectra peaked at high @xmath6 invariant masses
is expected .
the analysis bases on the compaz parametrization @xcite of the realistic luminosity spectra for a photon collider at tesla @xcite .
we assume that the centre - of - mass energy of colliding electron beams @xmath10 , is always optimised for the production of a higgs boson with a given mass .
the event generation according to the cross - section formula for a vector - boson production including the higgs contribution @xcite is done with pythia 6.152 @xcite .
the fast simulation program simdet version 3.01 @xcite is used to model the tesla detector performance .
the selection cuts are applied to select @xmath11 and @xmath12 events ( @xmath13 ) .
all results presented in this paper were obtained for an integrated luminosity corresponding to one year of the photon collider running , as given by @xcite .
the total photon - photon luminosity increases from about 600 fb@xmath14 for @xmath15305 gev ( optimal beam energy choice for @xmath16200 gev ) to about 1000 fb@xmath14 for @xmath15500 gev ( optimal beam energy choice for @xmath16350 gev ) . in this section
we summarise results of @xcite , where the feasibility of measuring standard model higgs - boson production in @xmath0 and @xmath2 decay channels in option of tesla has been studied for a higgs - boson mass above 150 gev .
we have studied the signal , i.e. the higgs - boson decays into the vector bosons , and the background from direct vector - bosons production . for the @xmath17 final - state a direct , i.e. non - resonant @xmath18 process ,
is rare as it occurs via loop only . on contrary ,
the non - resonant @xmath19 production is a tree - level process , and is expected to be large .
therefore , also an interference between the signal of @xmath19 production via the higgs resonance and the background from direct production may be large .
this effect can be used to access an information about the phase of the @xmath20 coupling , . for the higgs - boson masses around 350 gev we found that the amplitude phase is more sensitive to the loop contributions of new , heavy charged particles than the itself .
the invariant - mass resolution obtained from a full simulation of @xmath0 and @xmath2 events based on the pythia and simdet programs , has been parametrized as a function of the centre - of - mass energy , .
the distributions of the reconstructed invariant mass for @xmath21 and @xmath22 events are shown in figs .
[ fig : convol ] , upper and lower panels , respectively .
results from a full event simulation based on pythia and simdet are compared with the distribution obtained from the numerical convolution of the relevant cross - sections with the compaz spectra and the parametrized detector resolution .
based on a parametric description of the expected mass distributions , many experiments were simulated , each corresponding to one year of a photon collider running at tesla at a nominal luminosity .
the `` theoretical '' distributions were then fitted , simultaneously to the observed @xmath0 and @xmath2 mass spectra , with the width and phase considered as the only free parameters .
results of the fits performed for different higgs - boson masses and at different electron - beam energies are shown in figs .
[ fig : smfinal ] .
they indicate that with a proper choice of the electron - beam energy , the @xmath6 partial width can be measured with an accuracy of 3 to 8% , while the phase of the amplitude with an accuracy between 35 and 100 mrad , see fig .
[ fig : smfinal ] .
the measurement opens a new window to a precise determination of the higgs - boson couplings and to search of a `` new physics '' .
it turns out that the phase is constrained predominantly by the @xmath0 invariant - mass distribution , thanks to large interference effects between higgs - boson decay and nonresonant @xmath0 production .
however , two - photon width of the higgs - boson is much better constrained by the measurement of the @xmath2 mass spectra , as the nonresonant background is much smaller here .
a precise determination of both parameters is only possible when both measurements are combined . in our previous analysis
@xcite we have considered possible deviations from the standard model predictions resulting from the loop contributions to the @xmath23 vertex of new heavy charged particles . however , deviations in the two - photon width and phase can also appear if the couplings of the higgs - boson to the other particles are different than those predicted by the standard model .
this possibility is studied in this paper , see also @xcite . as the simplest model with non - standard higgs couplings
we consider standard model - like two higgs doublet model ( 2hdm ii ) @xcite . we study a simple version of the so called solution b , where the lightest higgs - boson @xmath24 couplings to fermions are fixed and they have the same values as in the standard model , except that for up - type fermions the sign is opposite .
the couplings of @xmath24 to the ew gauge - bosons may differ from the corresponding sm predictions .
there is only one parameter in such model , @xmath25 , and all remaining couplings of @xmath24 and all couplings of @xmath26 and @xmath27 can be expressed in terms of it , as presented in the table [ tab:2hdm ] . and
@xmath26 couplings to the charged higgs boson @xmath28 are calculated according to the 2hdm ii potential @xcite assuming @xmath29=0 . ] .
couplings of the neutral higgs - bosons to fermions and vector bosons , relative to the standard model couplings , for the considered solution b of the sm - like two higgs doublet model ( 2hdm ii ) .
[ cols="<,^,^,^",options="header " , ] one expects significant deviations from the standard model predictions for a light higgs - boson @xmath24 , both for the two - photon width and phase , for @xmath30 . as compared to the sm
there is a change of a relative sign of the top - quark and the @xmath31 contribution , therefore the two - photon width in 2hdm ii ( b ) is significantly larger than in the standard model , where these two contributions partly cancel each other .
for @xmath32 the two - photon width decreases , due to the suppressed @xmath31-loop contribution ( @xmath33 ) . finally , for large values of @xmath34 ( @xmath35 ) the two - photon width of the light higgs - boson @xmath24 tend to be very close to the expectations of the standard model .
the only difference is due to the presence of the charged higgs - boson in the loop .
the difference resulting from the opposite , as compared to sm , sign of the down - type fermion contributions is very small and can be neglected .
results for the light higgs - boson @xmath24 with mass @xmath36 gev , from the measurement of the two - photon width ( times the vector - boson branching ratio ) and phase are presented in fig .
[ fig : showel1 ] for various @xmath34 values .
error contours ( 1@xmath37 ) on the expected deviation from the standard - model predictions correspond to one year of photon collider running , i.e. @xmath38 fb@xmath14 . in the combined fit to the invariant - mass distributions for @xmath0 and @xmath1 events
we take the advantage of the fact that the ratio of the corresponding branching ratios @xmath39 is expected to be the same as in the standard model .
charged higgs - boson mass is set to 800 gev .
results presented in fig . [ fig : showel1 ] show that the measurement of the two - photon width and phase for the light higgs - boson @xmath24 decaying into @xmath0 and @xmath1 would allow a precise determination of the @xmath34 value .
the possible ambiguity in the measurement of the two - photon width is resolved by the phase measurement , which clearly distinguishes between low @xmath34 and large @xmath34 solutions . the statistical error on the extracted @xmath34 value
is shown in fig .
[ fig : errli ] for different values of light higgs - boson mass @xmath40 . for all considered higgs - boson masses
the expected error in the @xmath34 determination is smallest for @xmath34 close to 1 .
this is because the higgs - boson coupling to the vector bosons is most sensitive to @xmath34 , for such values .
the precision of @xmath34 measurement , for @xmath32 , deteriorates with an increase of the higgs - boson mass .
it changes from about 1% for mass of 200 gev to about 10% for mass of 350 gev .
for very high and very low @xmath34 , when the relative higgs - boson coupling to vector bosons is close to @xmath41 ( table [ tab:2hdm ] ) , precise measurement of @xmath34 is not possible .
the measurement of the two - photon width and phase has been investigated also for the heavy scalar higgs - boson @xmath26 of the sm - like two higgs doublet model ( solution b ) , with couplings as given in table [ tab:2hdm ] . in this case
we consider only @xmath42 , since for @xmath43 both the top - quark and @xmath31 contributions are strongly suppressed and the precision of the measurement deteriorates fast . for @xmath32 both the two - photon width and phase of the heavy scalar higgs - boson @xmath26 are close to the expectations of the standard model ( for a given @xmath44 ) . for decreasing values of @xmath34
the @xmath31-loop contribution decreases , while the top - quark one increases and as a result , the two - photon width decreases slightly for @xmath45 and then starts to increase with @xmath34 .
finally , for @xmath46 the higgs - boson decays to @xmath47 start to dominate .
the expected number of events with the @xmath0 and @xmath5 decays drops rapidly and the measurement becomes problematic again .
the results of the analysis of the measurement of the two - photon width and phase , for the heavy scalar @xmath26 with mass @xmath36 gev , are presented in fig .
[ fig : showel2 ] .
light higgs - boson mass is set to 120 gev and that of the charged higgs - boson to 800 gev .
error contours ( 1@xmath37 ) on the measured deviation from the standard model predictions are presented .
results show that a precise determination of @xmath34 , in the considered region of parameters , should also be possible for heavy scalar @xmath26 decaying to @xmath0 and @xmath5 . shown in fig .
[ fig : errhv ] is the statistical error on the extracted @xmath34 values .
results are given for four values of heavy scalar higgs - boson mass @xmath44 , from 200 to 350 gev .
the expected error in the @xmath34 determination is small ( 13 % ) for low @xmath34 , @xmath480.25 . for larger values of @xmath34 ,
the precision depends strongly on the higgs - boson mass . for mass of 200
gev it changes between 3 and 8 % , whereas for mass of 350 gev it is between 3 and 20 % . in the general two higgs doublet model @xcite , the mass eigenstates of the neutral higgs - bosons @xmath49 , @xmath50 and @xmath51
do not match cp eigenstates @xmath24 , @xmath26 and @xmath27 .
we consider here a sm - like version of the cp - violating two higgs doublet model , with cp violation through a small mixing between @xmath26 and @xmath27 states .
we consider a simple version of such model , where the couplings of the lightest higgs mass - eigenstate @xmath49 ( with mass 120 gev ) are expected to correspond to the couplings of @xmath24 boson , whereas couplings of @xmath50 and @xmath51 states can be described as the superposition of @xmath26 and @xmath27 couplings ( see table [ tab:2hdm ] ) , as follows : @xmath52 where @xmath53 denotes a fermion or a vector boson , @xmath54 or @xmath55 .
we study the feasibility of the mixing angle @xmath56 determination from the combined measurement of the two - photon width and phase for the higgs mass - eigenstate @xmath50 .
we consider only small cp - violation effects , i.e. @xmath57 .
results of the combined analysis ( ww / zz decay channels ) of the measurement of the two - photon width ( times vector - boson branching ratio ) and phase for the scalar higgs - boson @xmath50 with mass @xmath58 gev , are presented in fig .
[ fig : showel3 ] .
a light higgs - boson has mass @xmath59 gev while @xmath60 gev .
the simultaneous fit to the observed @xmath0 and @xmath5 mass - spectra was performed .
error contours ( 1@xmath37 ) on the measured deviation from the standard model predictions are shown for @xmath61 , i.e. when cp is conserved , and for cp violation with @xmath62 .
even a small cp - violation can significantly influence the measured two - photon width and two - photon phase , and therefore it is possible to determine precisely both the cp - violating mixing angle @xmath56 and @xmath34 .
next , we address a question : how well one can establish the fact that in a model cp is conserved ?
the answer can be read out from fig .
[ fig : errph ] where the statistical error in the determination of the @xmath8 mixing angle @xmath56 , around @xmath63 value , is shown .
the results are presented as a function of @xmath34 for four values of higgs - boson mass @xmath64 , from 200 to 350 gev .
as above , we assume a light higgs - boson mass @xmath65 gev , charged higgs - boson mass of 800 gev . for low value of @xmath34 , @xmath66 0.25 ,
the expected error is of the order of 2025 mrad and does not depend on the heavy higgs - boson mass . for larger values of @xmath34
the precision worsens for small higgs - boson masses . for @xmath67 the expected error changes from 35 mrad for mass of 350 gev to about 90 mrad for mass of 200 gev . to summarize this part , in the 2hdm model the higgs - boson couplings to the vector bosons are always as for a scalar ( say in the sm ) with a modification by some functions , i.e. they are proportional to @xmath68
this is because there is no pseudoscalar coupling to @xmath0 and @xmath5 in this model ( @xmath69 ) .
the measured @xmath0 and @xmath5 mass - spectra are sensitive to the cp - violating @xmath8 mixing angle only via a yukawa coupling , more precisely due of the top - quark loop contribution to the @xmath70 vertex . in the next section
we will consider more general case , with a generic pseudoscalar couplings to the gauge bosons .
in the general case , with a generic pseudoscalar coupling of higgs - boson to vector bosons , not only the invariant - mass distribution of the reconstructed @xmath19 and @xmath5 events , as discussed in sec .
2.3 , but also the angular distributions of @xmath19 and @xmath5 decay products are sensitive to the cp properties of the higgs - boson @xcite .
we use the following form of a coupling to vector boson @xmath55 of a scalar ( @xmath26 ) and a pseudoscalar ( @xmath27 ) : @xmath71 g_{azz } = i g \frac{m_z}{\cos \theta_w } \ ; \left ( \lambda_a \cdot \varepsilon^{\mu \nu \rho \sigma}\ ; \frac{(p_1+p_2)_\rho \ ; ( p_1-p_2)_\sigma}{m_z^2 } \right)\ , \label{ha}\end{aligned}\ ] ] where @xmath72 and @xmath73 are the 4-momenta of the vector bosons . the coupling with @xmath74 corresponds to the sm - like cp - even coupling , whereas the one with @xmath75 is a general cp - odd coupling for the spin-0 boson ( higgs ) . and we use similar expressions for the coupling to @xmath31 , with an obvious change @xmath76 to @xmath77 in the denominator of the pseudoscalar coupling , and a change in normalization factor as in the sm , namely @xmath78 .
we see , that if @xmath79 , the scalar @xmath26 behaves as the sm higgs boson .
angular variables which can be used in the analysis of cp - properties are defined in fig .
[ fig : ang_def ] . to test cp - properties of the higgs - bosons
one can use the distributions of the polar angles @xmath80 and @xmath81 as well as the @xmath82 distribution , where @xmath82 is the angle between two @xmath55- or two @xmath31- decay planes . to simplify the analysis , instead of two - dimensional distribution in ( @xmath83 , @xmath84 ) we consider the distribution in a new variable , defined as @xmath85 the @xmath86-distribution corresponds to the ratio of the angular distributions expected for the decay of a scalar and a pseudoscalar ( in a limit @xmath87 ) @xcite .
the angular distributions in @xmath82 and @xmath86 , expected for a scalar and a pseudoscalar higgs ( eq . [ ha ] ) are compared in fig .
[ fig : ang_exp ] . both distributions clearly distinguish between decays of scalar and pseudoscalar higgs . for the scalar higgs the distributions are almost flat in both @xmath82 and @xmath86 ,
whereas the pseudoscalar coupling ( see eq . [ ha ] ) introduces a significant @xmath82 and @xmath86 dependence . from the measurement of these distributions we can try to establish the cp properties of the higgs boson , even without taking into account the production mechanism .
such measurement should be considered as the most general one , as in principle some unknown heavy charged particles can significantly modify the production vertex , affecting the invariant - mass distributions on which the measurements of the two - photon width and two - photon phase are based . note , that the distribution of the higgs decay - angle @xmath88 is expected to be flat both for scalar and pseudoscalar higgses .
however , the corresponding angular distribution it is not flat for the non - resonant background .
therefore the observed @xmath88 distributions , where obviously a signal , background and possible interference between them contribute , maybe be sensitive to the higgs - boson couplings .
following the analysis described in @xcite we consider a generic higgs model with the following tensor couplings of a higgs boson , denoted below by @xmath89 , to @xmath5 and @xmath19 : @xmath90 the standard model couplings are reproduced for @xmath91 and @xmath92 .
below we will consider only small deviations from the standard model , i.e. @xmath93 and @xmath94 . to simplify the case
, we assume that higgs @xmath95 couplings to fermions are the same as in the standard model .
measurement of the angular distributions has been studied using samples of @xmath5 events generated with pythia and simdet programs , as described in section [ sec : width ] .
the resolutions of the reconstructed decay angles @xmath96 and @xmath97 , for leptonic and hadronic @xmath55 decays , are compared in fig .
[ fig : ang_res ] , for the scalar higgs - boson with mass of 300 gev .
a very good resolution is obtained for all considered distributions , nevertheless the measured angular distributions are strongly affected by the selection cuts used in the analysis . because of the cut on the lepton and jet angles , applied to preserve a good mass resolution ( see @xcite for more details ) , we observe significant loss of the selection efficiency for events with lepton or jet emitted along the beam direction . also the durham jet algorithm used in the event reconstruction imposes constraints on the angular separation between leptons and jets .
both effects result in a highly non - uniform angular acceptance .
selection efficiencies for @xmath98 events , as a function of the angle @xmath82 , are presented in fig .
[ fig : acc_zz_300 ] .
we consider reconstructed @xmath82 values from 0 to @xmath99 , since we are not able to distinguish between quark and antiquark jet .
the efficiencies for @xmath5 events coming from decays of scalar , pseudoscalar higgs and non - resonant background are compared .
the polar - angle distributions differ among these three classes of events , what leads to a significantly different acceptances .
this simulation was performed for primary electron - beam energy of 209 gev optimal for a higgs - boson mass of 300 gev .
an additional cut on the reconstructed @xmath5 invariant - mass has been introduced to optimise the signal measurement from the angular distribution .
we found that for higgs mass of 300 gev the accepted mass range lies between 286 and 312 gev .
this cuts affects mainly the non - resonant background events .
we calculate the expected angular distributions for @xmath5 and @xmath0 events by convoluting the corresponding cross - section formula with the compaz photon - energy spectra , and in addition with the parametrization of the invariant - mass resolution and acceptance function including the angular- and jet - selection cuts . here
we neglect ( small ) effects of the angular resolution . for the measurement of the event distributions in @xmath82 and @xmath86
we introduce the additional cuts on the reconstructed @xmath5 invariant mass ( for @xmath5 events ) or on the reconstructed @xmath0 invariant mass as well as the higgs - decay angle @xmath88 ( for @xmath0 events ) .
the cuts were optimised for the smallest relative error in the signal cross - section measurement .
expected precision in the measurements of the angle @xmath82 and of the variable @xmath86 distributions , for @xmath98 events and @xmath100 events are illustrated in figs .
[ fig : histzz ] and [ fig : histww ] , respectively .
calculations were performed for primary electron - beam energy of 209 gev and the higgs - boson mass of 300 gev .
the results , presented relative to the sm prediction , are confronted in figs .
[ fig : histzz ] ( [ fig : histww ] ) with the expectations of the generic model with @xmath101 and @xmath91 ( @xmath102 and @xmath91 ; @xmath92 and @xmath103 ) , for the reconstructed @xmath5 ( @xmath0 ) events .
we see , that the generic model predicts deviations both in the normalization and in the shape of angular distributions .
therefor we should be able to constrain higgs - boson couplings from the shape of the distributions , event if the overall normalization related to the higgs - boson production mechanizm is not known .
each of the considered angular distributions discussed above can be fitted with a model expectations , given in terms of the parameters @xmath75 and @xmath74 . if we do not use normalisation constraints ( i.e. do not assume a given production mechanism ) there are large correlations between @xmath75 and @xmath74 in fits to a single - variable distributions , i.e. for @xmath82 , @xmath86 and @xmath88 .
therefore we first limit ourselves to @xmath91 case .
statistical errors in the determination of @xmath75 from the fits to the shape of various angular distributions , to be obtained after one year of photon collider running , are presented in fig .
[ fig : resazzww ] , separately for the @xmath104 and @xmath5 channels . in the figure we present also
the results obtained from the simultaneous ( combined ) fit to all three considered distributions .
it turns out that both for @xmath5 and for @xmath0 events the measurement of the @xmath86-distribution gives the best constraints on the @xmath75 parameter , and a precision of @xmath75 measurement from the @xmath86 distribution only is close to the precision obtained from the combined fit . for @xmath105250 gev
the expected accuracy for @xmath75 is about 0.1 both for the @xmath5 and for @xmath0 events .
finally we estimate the expected statistical errors from the combined fit to angular distributions of @xmath5 and @xmath0 events , they are shown in fig . [
fig : resaa ] . by fitting all angular distributions
simultaneously we found the error on @xmath106 of 0.060.08 for @xmath107 200 gev . in case of the combined fit ,
correlations between @xmath106 and @xmath108 parameters are smaller , so both parameters can be fitted simultaneously .
expected statistical errors on @xmath106 and @xmath108 are shown in fig .
[ fig : resha2 ] . for low higgs - boson masses , @xmath109250 gev ,
better constrains are obtained from the measurement of @xmath0 events , whereas for @xmath105300 gev smaller errors are obtained from the measurement of @xmath5 events .
expected error on @xmath106 is below 0.1 for @xmath109200 gev , while the corresponding error on @xmath108 changes from about 0.1 to 0.3 with increasing higgs - boson mass .
significantly better constraints on the model parameters @xmath106 and @xmath108 can be obtained if we assume that there is no `` new physics '' except for the considered anomalous couplings of the higgs - boson to @xmath0 and @xmath5 .
we can then calculate the expected invariant - mass distributions for @xmath5 and @xmath0 events . in fig .
[ fig : histmzz ] and [ fig : histmww ] statistical errors on the measured invariant - mass distributions , expected after one year of photon collider running at a nominal luminosity , are compared with the predictions of the generic model .
we see , that large deviations from the standard model prediction are expected for @xmath110 , mainly due to the interference effects . expected statistical errors in the determination of @xmath106 and @xmath108 from the combined fit to angular distributions and invariant - mass distributions are compared in fig .
[ fig : resahm ] .
a relative normalisation of @xmath5 and @xmath0 samples is imposed in the fit , as predicted from the cross section calculations .
however , overall normalisation , corresponding to the integrated luminosity of the considered data sample , is allowed to vary . for @xmath107 200 gev the error on @xmath106 of
about 0.01 can be achieved from the combined fit to all considered distributions .
the value of @xmath106 is mainly constrained by the measurement of the invariant - mass distribution for @xmath0 events .
expected statistical error on @xmath108 is slightly bigger , between 0.01 and 0.02 , and comes mainly from the measurement of the invariant - mass distribution for @xmath5 events .
an opportunity of measuring of the higgs - boson properties at the the photon collider at tesla has been studied in detail for masses between 180 and 350 gev , using realistic luminosity spectra and detector simulation .
we found that from the combined measurement of the invariant - mass distributions in the @xmath5 and @xmath0 decay - channels , the parameters of the sm - like two higgs doublet model can be precisely determined .
for both light and heavy scalar higgs boson , a statistical precision in the determination of its coupling ( @xmath34 ) is of the order of 10% already after one year of photon collider running . in case of the two higgs doublet model with a cp violation
, the @xmath8 mixing angle can be determined with statistical precision of 20 to 90 mrad ( in a small - mixing approximation ) .
we also considered the combined measurement of the various angular correlations in the @xmath0 and @xmath1-decay products . for a simple generic model , with sm - couplings to fermions
, the parameters describing a cp violation in the higgs couplings to vector bosons can be determined to about 10% .
if in addition the invariant - mass distribution is used to constrain model parameters , 12 % statistical uncertainty in the determination of the higgs - boson couplings can be achieved .
we are especially grateful to david miller for stimulating discussions .
we would also like to thank other colleagues from the ecfa / desy study for useful comments and suggestions .
m.k . acknowledges partial support by polish committee for scientific research , grant 5 p03b 121 20 ( 2003 ) , and by the european community s human potential programme under contract hprn - ct-2000 - 00149 physics at colliders .
i. f. ginzburg , g. l. kotkin , v. g. serbo and v. i. telnov , 34 , 514 ( 1981 ) , 34 , 491 ( 1982 ) ; preprint inp 81 - 50 , novosibirsk , 1981 and nucl .
instrum .
meth . a 205 , 47 ( 1983 ) ; preprint inp 81 - 102 , novosibirsk , 1981 ; + i. f. ginzburg , g. l. kotkin , s. l. panfil , v. g. serbo and v. i. telnov , a219 , 5 ( 1984 ) ; + v. i. telnov , a294 , 72 ( 1990 ) .
g.belanger , f.boudjema , phys .
b288 ( 1992 ) 210 ; d.a.morris , et al .
b323 ( 1994 ) 421 .
t. sjostrand , p. eden , c. friberg , l. lonnblad , g. miu , s. mrenna and e. norrbin , 135 ( 2001 ) 238 . m. pohl , h. j. schreiber , desy-99 - 030 . i. f. ginzburg , m. krawczyk and p. osland , _ two - higgs - doublet models with cp violation _ , presented at international workshop on linear colliders ( lcws 2002 ) , jeju island , korea , august 2002 , hep - ph/0211371
miller , s.y .
choi , b. eberle , m.m .
mhlleitner and p.m. zerwas , phys .
b505 ( 2001 ) 149 [ hep ph/0102023 ] ; s.y .
choi , d.j .
miller , m.m .
mhlleitner and p.m. zerwas , hep
ph/0210077 ; d.j .
miller , _ spin and parity in the hzz vertex _ , presented at the ecfa / desy meeting , prague , november 2002 .
|
higgs - boson production at the photon collider at tesla is studied for masses from 180 to 350 gev , using realistic luminosity spectra and detector simulation .
parity of the sm higgs - boson can be verified from the measurement of the interference effects in the @xmath0 decay channel and of the angular correlations in the decays of @xmath0 and @xmath1 pairs .
sm - like two higgs doublet model ( 2hdm ii ) and model with the generic higgs couplings are considered .
|
the twin - beam state ( twb ) of two modes of radiation can be expressed in the photon number basis as @xmath0 where @xmath1 , @xmath2 being the twb squeezing parameter .
twb is described by a gaussian wigner function @xmath3 with @xmath4 and @xmath5 .
since ( [ twb : wig ] ) is positive - definite , twb are not suitable to test nonlocality through homodyne detection .
indeed , the wigner function itself provides an explicit hidden variable model for homodyne measurements @xcite . on the other hand ,
it has been shown @xcite that twb exhibits a nonlocal character for parity measurements .
this is known as nonlocality in the phase - space since bell inequalities can be written in terms of the wigner function , which in turn describes correlations for the joint measurement of displaced parity operators .
overall , the positivity or the negativity of the wigner function has a rather weak relation to the locality or the nonlocality of quantum correlations . in ref .
@xcite we have suggested a conditional measurement scheme on twb leading to a non gaussian entangled mixed state , which improves fidelity in the teleportation of coherent states .
this process , called inconclusive photon subtraction ( ips ) , is based on mixing each mode of the twb with the vacuum in a unbalanced beam splitter and then performing inconclusive photodetection on both modes , _ i.e. _ revealing the reflected beams without discriminating the number of the detected photons . a single mode version of the ips , mapping squeezed light onto non - gaussian states , has been recently realized experimentally @xcite .
moreover , ips has been suggested as a feasible method to modify twb and test nonlocality using homodyne detection @xcite . in this paper
we address ips as a _
degaussification _ map for twb , give its explicit expression for the density matrix and the wigner function , and investigate the nonlocality of the resulting state in the phase - space .
the paper is structured as follows . in section [ s : bana ]
we review nonlocality in the phase - space , _ i.e. _ wigner function bell s inequality based on measuring the displaced parity operator on two modes of radiation . in section [ s : degauss ] we illustrate the ips process as a degaussification map and calculate the wigner function of the ips state .
the nonlocality of the ips state in the phase - space is then analyzed in section [ s : nonlocal ] , whereas in section [ s : homodyne ] we discuss nonlocality using homodyne detection , extending the analysis of ref .
@xcite . section [ s : remarks ] closes the paper with some concluding remarks .
the displaced parity operator on two modes is defined as @xmath6 where @xmath7 , @xmath8 and @xmath9 are mode operators and @xmath10 and @xmath11 are single - mode displacement operators .
parity is a dichotomic variable and thus can be used to establish bell - like inequalities @xcite .
since the two - mode wigner function @xmath12 can be expressed as @xmath13 @xmath14 being the expectation value of @xmath15 , the violation of these inequalities is also known as nonlocality in the phase - space . the quantity involved in such inequalities
can be written as follows @xmath16 which , for local theories , satisfies the condition @xmath17 .
following ref .
@xcite , one can choose a particular set of displaced parity operators , arriving at the following combination @xmath18 which only depends on the positive parameter @xmath19 , characterizing the magnitude of the displacement .
if we evaluate the quantity ( [ bell : bana ] ) in the case of the twb , we find that it exceeds the upper bound imposed by local theories for a certain region of values @xmath20 , its maximum being @xmath21 @xcite . on the other hand , the choice of the parameters leading to eq .
( [ bell : bana ] ) is not the best one , and the violation of the inequality @xmath17 can be enhanced using a different parameterization @xcite .
a better result is achieved for @xmath22 which , for the twb , gives a maximum @xmath23 , greater than the value @xmath24 obtained in ref .
@xcite . in the following sections we will see that the violation of the inequalities @xmath25 and @xmath26 can be enhanced by degaussification of the twb .
the degaussification of a twb can be achieved by subtracting photons from both modes @xcite . in ref
. @xcite we referred to this process as to inconclusive photon subtraction and showed that the resulting state , the ips state @xmath27 , can be used to enhance the teleportation fidelity of coherent states for a wide range of the experimental parameters .
scheme of the ips process . ]
the ips scheme is sketched in fig .
[ f : ips : scheme ] .
the two modes , @xmath8 and @xmath9 , of the twb are mixed with the vacuum ( modes @xmath28 and @xmath29 , respectively ) at two unbalanced beam splitters ( bs ) with equal transmissivity @xmath30 ; the modes @xmath28 and @xmath29 are then revealed by avalanche photodetectors ( apd ) with equal efficiency @xmath31 .
apd s can only discriminate the presence of radiation from the vacuum .
the positive operator - valued measure ( povm ) @xmath32 of each detector is given by @xmath33 @xmath31 being the quantum efficiency .
overall , the conditional measurement on the modes @xmath28 and @xmath29 , is described by the povm ( we are assuming the same quantum efficiency for both photodetectors ) @xmath34 when the two photodetectors jointly click , the conditioned output state of modes @xmath8 and @xmath9 is given by @xmath35\ : , \label{ptr}\end{aligned}\ ] ] where @xmath36 and @xmath37 are the evolution operators of the beam splitters and @xmath38 the density operator of the two - mode state entering the beam splitters ( in our case latexmath:[$r = \varrho_{\hbox{\tiny twb } } = partial trace on modes @xmath28 and @xmath29 can be explicitly evaluated , thus arriving at the kraus decomposition of the ips map .
we have @xmath40 with @xmath41}{p!}}\,,\ ] ] and @xmath42 and @xmath43\end{aligned}\ ] ] is the probability of a click in both detectors .
now , in order to investigate the nonlocality of the state @xmath44 in the phase - space , we explicitly calculate its wigner function , which , as one may expect , is no longer gaussian and positive - definite .
the state entering the two beam splitters is described by the wigner function @xmath45 where the second factor at the rhs represents the two vacuum states of modes @xmath28 and @xmath29 . the action of the beam splitters on @xmath46 can be summarized by the following change of variables @xmath47 and the output state , after the beam splitters , is then given by @xmath48 where @xmath49 and @xmath50,\\ w & \equiv & w(r,\phi ) = 2 \cos\phi\ , \sin\phi\ , [ ( 1-a){\overline{\beta}}+ b \alpha]\,.\end{aligned}\ ] ] at this stage conditional on / off detection is performed on modes @xmath28 and @xmath29 ( see fig .
[ f : ips : scheme ] ) .
we are interested in the situation when both the detectors click .
the wigner function of the double click element @xmath51 of the povm ( see eq .
( [ povm11 ] ) ) is given by @xcite @xmath52(\zeta,\xi)\nonumber\\ & = & \frac{1}{\pi^2}\ { 1-q_{\eta}(\zeta)-q_{\eta}(\xi)\nonumber\\ & \mbox{}&\hspace{1.5 cm } + q_{\eta}(\zeta ) q_{\eta}(\xi ) \}\,,\end{aligned}\ ] ] with @xmath53 using eq .
( [ ptr ] ) and the phase - space expression of trace ( for each mode ) @xmath54 = \pi\ , \int_{\mathbb{c } } d^2 z\ , w[o_1](z)\ , w[o_2](z)\,,\ ] ] @xmath55 and @xmath56 being two operators and @xmath57(z)$ ] and @xmath58(z)$ ] their wigner functions , respectively , the wigner function of the output state , conditioned to the double click event , is then given by @xmath59 where @xmath60 and @xmath61 is the double - click probability ( [ dc : fock ] ) , which can be written as function of @xmath62 as follows @xmath63 the quantity @xmath64 appearing in eq .
( [ w : ips : informal : f ] ) is @xmath65 and the expressions of @xmath66 , @xmath67 and @xmath68 are given in table [ t : jj ] .
.[t : jj ] [ cols="^,^,^,^",options="header " , ] the mixing with the vacuum in a beam splitter with transmissivity @xmath69 followed by on / off detection with quantum efficiency @xmath31 is equivalent to mixing with an effective transmissivity @xcite @xmath70 followed by an ideal ( _ i.e. _ efficiency equal to 1 ) on / off detection .
therefore , the state ( [ w : ips : informal ] ) can be studied for @xmath71 and replacing @xmath69 with @xmath72 .
thanks to this substitution , after the integrations we have @xmath73 and @xmath74^{-1 } c_j(\eta ) } { ( b - f_j)(b - g_j)-(2 b h_j \tau_{\rm eff})^2 } \,,\ ] ] where we defined @xmath75\,,\\ g_j & \equiv & g_j(r,\phi,\eta ) = n_j \ , [ x_j b^2 + 4 b^2 ( 1-a ) ( 1-\tau_{\rm eff } ) + y_j ( 1-a)^2]\,,\\ h_j & \equiv & h_j(r,\phi,\eta ) = n_j \ , [ ( x_j + y_j ) b ( 1-a ) + 4 b ( b^2 + ( 1-a)^2 ) ( 1-\tau_{\rm eff})]\,,\\ n_j & \equiv & n_j(r,\phi,\eta ) = { \displaystyle \frac{4 \tau_{\rm eff}\ , ( 1-\tau_{\rm eff})}{x_j y_j - 4 b^2 ( 1-\tau_{\rm eff})^2}\ , . } \end{aligned}\ ] ] in this way , the wigner function of the ips state can be rewritten as @xmath76}\,,\end{aligned}\ ] ] where we introduced @xmath77 the state given in eq .
( [ ips : wigner ] ) is no longer a gaussian state .
in this section we investigate the nonlocality of the state ( [ ips : wigner ] ) in phase - space using the quantity @xmath78 given in eq .
( [ bell : general ] ) , referring to both the parameterizations @xmath79 ( see eq .
( [ bell : bana ] ) ) and @xmath80 ( see eq .
( [ bell : ale ] ) ) .
plot of @xmath79 given in eq .
( [ bell : bana ] ) for @xmath81 .
the dashed line is @xmath82 , while the solid lines are @xmath83 for different values of @xmath72 ( see the text ) : from top to bottom @xmath84 and @xmath85 .
when @xmath86 , the maximum of @xmath87 is @xmath88 .
the lower plot is a magnification of the region @xmath89 of the upper one .
notice that for small @xmath2 there is always a region where @xmath90 . ]
plots of @xmath79 given in eq .
( [ bell : bana ] ) as a function of the squeezing parameter @xmath2 for different values of @xmath19 : ( a ) @xmath91 , ( b ) @xmath92 and ( c ) @xmath93 . in all the plots the dashed line is @xmath82 , while the solid lines are @xmath83 for different values of @xmath72 ( see the text ) : from top to bottom @xmath94 and @xmath95 .
notice that there is always a region for small @xmath2 where @xmath90 . when @xmath96 the maximum of @xmath83 is always greater than the one of @xmath82 . ] as for a twb , the violation of the bell s inequality is observed for small @xmath2 @xcite . from now on , we will refer to @xmath79 as @xmath82 when it is evaluated for a twb ( [ twb : wig ] ) , and as @xmath83 when we consider the ips state ( [ ips : wigner ] ) .
we plot @xmath82 and @xmath97 in the figs .
[ f : ips : vs : twb : j ] and [ f : ips : vs : twb ] for different values of the effective transmissivity @xmath72 and of the parameter @xmath19 : for not too big values of the squeezing parameter @xmath2 , one has that @xmath98 . moreover , when @xmath72 approaches unit , _
i.e. _ when at most one photon is subtracted from each mode , the maximum of @xmath99 is always greater than the one obtained using a twb .
a numerical analysis shows that in the limit @xmath100 the maximum is @xmath101 , that is greater than the value @xmath24 obtained for a twb @xcite .
the limit @xmath100 corresponds to the case of one single photon subtracted from each mode @xcite .
notice that increasing @xmath19 reduces the interval of the values of @xmath2 for which one has the violation . for large @xmath2 the best result
is thus obtained with the twb since , as the energy grows , more photons are subtracted from the initial state @xcite . since the relevant parameter for violation of bell inequalities is @xmath72 , we have , from eq .
( [ taueff ] ) , that the ips state is nonlocal also for low quantum efficiency of the ips detector .
plots of @xmath80 given in eq .
( [ bell : ale ] ) as a function of the squeezing parameter @xmath2 for @xmath102 .
in all the plots the dashed line is @xmath103 , while the solid lines are @xmath104 for different values of @xmath72 ( see the text ) : from top to bottom @xmath105 and @xmath106 . when @xmath96 the maximum of @xmath104 is @xmath107 . ] the same conclusions holds when we consider the parameterization of eq .
( [ bell : ale ] ) . in fig .
[ f : ips : ale ] we plot @xmath103 and @xmath108 , _ i.e. _ @xmath80 evaluated for the twb and the ips state , respectively .
the behavior is similar to that of @xmath79 , the maximum violation being now @xmath109 for @xmath96 and @xmath110 .
finally , notice that the maximum violation using ips states is achieved ( for both parameterizations ) when @xmath72 approaches unit and for values of @xmath2 smaller than for twb .
the wigner function @xmath111 given in eq .
( [ ips : wigner ] ) is not positive - definite and thus @xmath112 can be used to test the violation of bell s inequalities by means of homodyne detection , _
i.e. _ measuring the quadratures @xmath113 and @xmath114 of the two ips modes @xmath8 and @xmath9 , respectively , as proposed in ref . @xcite . in this case , if one discretizes the measured quadratures assuming as outcome @xmath115 when @xmath116 , and @xmath117 otherwise , one obtains the following bell parameter @xmath118 where @xmath119 and @xmath120 are the phases of the two homodyne measurements at the modes @xmath8 and @xmath9 , respectively , and @xmath121\ , p(x_{\vartheta_j } , x_{\varphi_k})\,,\ ] ] @xmath122 being the joint probability of obtaining the two outcomes @xmath123 and @xmath124 @xcite . as usual , violation of bell s inequality is achieved when @xmath125 . in fig .
[ f : homo : id ] we plot @xmath126 for @xmath127 , @xmath128 , @xmath129 and @xmath130 : as pointed out in ref .
@xcite , the bell s inequality is violated for a suitable choice of the squeezing parameter @xmath2 . notice that when @xmath72 decreases the maximum of violation shifts toward higher values of @xmath2 .
as one expects , taking into account the efficiency @xmath131 of the homodyne detection furtherly reduces the violation ( see fig . [
f : homo : eta ] ) .
notice that , when @xmath132 , violation occurs for higher values of @xmath2 , although its maximum is actually reduced : in order to have a significative violation one needs a homodyne efficiency greater than 80% ( when @xmath133 ) .
plots of @xmath126 given in eq .
( [ bell : homo ] ) as a function of @xmath134 for different values of @xmath72 and for ideal homodyne detection ( _ i.e. _ with quantum efficiency @xmath135 ) : from top to bottom @xmath136 and @xmath137 . ] plots of @xmath126 given in eq .
( [ bell : homo ] ) as a function of @xmath134 with @xmath138 and for different values of the homodyne detection efficiency @xmath131 : from top to bottom @xmath139 and @xmath140 .
the maximum of the violation decreases and shifts toward higher values of @xmath2 as @xmath131 decreases . for smaller values of @xmath72
the violation is furtherly reduced . ]
in this paper we have shown that ips can be used to produce non - gaussian two - mode states starting from a twb .
we have studied the nonlocality of ips states in phase - space using the wigner function . as for the improvement of ips assisted teleportation @xcite ,
we have found that the nonlocal correlations are enhanced for small energies of the twb ( small squeezing parameter @xmath2 ) .
moreover , nonlocality of @xmath27 is larger than that of twb irrespective of ips quantum efficiency .
since the wigner function of the ips state is not positive definite , we have also analyzed its nonlocality using homodyne detection . in this case violation of bell s inequality
is much less than in the phase - space , and is furtherly reduced for non unit homodyne efficiency @xmath132 .
however , this setup ( ips with homodyning ) is of particular interest , since it can be realized with current technology achieving a loophole - free test of bell s inequality @xcite .
on the other hand , the experimental verification of phase - space nonlocality is challenging , due to the difficulties of measuring the parity , either directly or through the measurement of the photon distribution . on the other hand ,
the recent experimental generation of ips states @xcite is indeed a step toward its implementation .
so would like to express his gratitude to a. r. rossi and a. ferraro for stimulating discussions and for their continuous assistance .
99 k. banaszek and k. wdkiewicz , phys .
a * 58 * , 4345 ( 1998 ) .
r. garca - patrn snchez et al . , quant - ph/0403191 .
s. olivares , m. g. a. paris and r. bonifacio , phys . rev .
a * 67 * , 032314 ( 2003 ) .
j. wenger , r. tualle - brouri and p. grangier , phys .
rev . lett . * 92 * , 153601 ( 2004 ) .
j. f. clauser , m. a. horne , a. shimony and r. a. holt , phys .
* 23 * , 880 ( 1969 ) .
a. ferraro , private communication .
t. opatrn , g. kurizki and d .-
welsch , phys .
a * 61 * , 032302 ( 2000 ) .
p. t. cochrane , t. c. ralph and g. j. milburn , phys .
rev . a * 65 * , 062306 ( 2002 ) .
m. g. a. paris , m. cola and r. bonifacio , phys .
a * 67 * , 042104 ( 2003 ) .
|
we show that inconclusive photon subtraction ( ips ) on twin - beam produces non - gaussian states that violate bell s inequality in the phase - space .
the violation is larger than for the twin - beam itself irrespective of the ips quantum efficiency .
the explicit expression of ips map is given both for the density matrix and the wigner function representations .
|
recently , we discussed local and linear media in classical electrodynamics @xcite . in particular , we investigated possible magnetoelectric effects , which are related to the crossterms between the magnetic ( electric ) field strength and the electric ( magnetic ) excitation . using the premetric formalism of electrodynamics ,
see @xcite , the most general local and linear constitutive relation can be written as @xmath0 here @xmath1 and @xmath2 are the electric and magnetic excitations , respectively , and @xmath3 and @xmath4 the electric and the magnetic field strengths .
the kronecker symbol is denoted by @xmath5 , the totally antisymmetric levi - civita symbol by @xmath6 and @xmath7 , respectively .
we have 36 constitutive functions or moduli : the permittivity matrix @xmath8 ( 6 independent components ) , the impermeability matrix @xmath9 ( 6 components ) , the tracefree principal magnetoelectric matrix @xmath10 ( 8 components ) , the 15 skewon pieces @xmath11 , and , eventually , 1 axion piece @xmath12 . such a local and linear medium with 36 moduli
called sometimes bi - anisotropic has been considered , amongst others , by lindell , sihvola , tretyakov , and collaborators @xcite . in conventional materials
the _ skewon _ and the _ axion _ pieces _ vanish _ and we are left with @xmath13 the principal magnetoelectric cross terms induced by @xmath10 are known to exist in various media , see odell @xcite .
such media are described by @xmath14 constitutive functions .
eqs.([convent1 ] ) and ( [ convent2 ] ) represent the most general material considered by post @xcite , e.g .. the skewon pieces are not considered in this article , see , however , @xcite .
we address here the question on whether the axion piece is permitted thereby possibly extending a material with 20 moduli to one with 21 moduli and if so whether it can be determined experimentally by standard methods . in post
@xcite it was argued that the axion piece has to vanish , i.e. , @xmath15 ( and even @xmath16 ) .
for this reason , lakhtakia @xcite ( and references given there ) called @xmath17 the post constraint and advocated it as a condition each medium has to fulfill .
we quoted in @xcite literature in which materials are described ( cr@xmath18o@xmath19 and fe@xmath18teo@xmath20 ) that carry an axion piece .
moreover , lakhtakia , loc.cit . , pointed out that a constant axion piece @xmath12 should not be measurable since it drops out of the maxwell equations .
however , this is only true if we have an axion piece that is _ globally constant _ at _ all _ spatial points at _ all _ times .
we will show explicitly that a material with a piecewise constant @xmath21 can very well be investigated experimentally and thereby @xmath21 measured uniquely .
thus , neither can the post constraint be uphold @xcite nor poses the measurability of @xmath21 a problem , as we will show below .
accordingly , we discuss here the case of two neighboring homogeneous media with different but constant axion pieces .
they are separated by the plane @xmath22 , as shown in fig.1 .
permittivity and permeability are assumed to be isotropic .
thus , the constitutive relations for the two half - spaces carry 3 constitutive constants @xmath23 , and @xmath21 , respectively , @xmath24 with @xmath25 as the 3-dimensional hodge star operator and @xmath26 and @xmath27 as electric and magnetic constants ( of the vacuum ) .
we use here the calculus of differential forms , see @xcite . as sideremark let us remind ourselves that ( [ con2 ] ) , ( [ con1 ] ) , formulated 4-dimensionally with the excitation 2-form @xmath28 and the field strength 2-form @xmath29 and _ in vacuum , _ reduce to axion ( maxwell - lorentz ) electrodynamics ( see ni @xcite and wilczek @xcite ) with the constitutive relation @xmath30 here @xmath31 is the 4-dimensional hodge star operator defined in terms of the metric of spacetime .
the field equations read @xmath32 it is as if the current 3-form @xmath33 picked up an additional piece depending on the gradient of the axion field . for @xmath34 ,
this corresponds to the pure axion case , that is , to the perfect electromagnetic conductor ( pemc ) of lindell and sihvola @xcite , a structure that is equivalent to the tellegen gyrator @xcite , see also @xcite . the real part of kiehn s _ chiral vacuum _ ( see @xcite and @xcite , pp
. 140/141 ) is a subcase , for @xmath35 , of axion electrodynamics .
thus we see that if the axion piece is globally constant , @xmath36const , it does not contribute to the maxwell equations ( [ maxax ] ) , even though it emerges in the constitutive relation ( [ vac ] ) and in the boundary conditions to be discussed in the next section .
the maxwell equations without charges and currents read @xmath37 the calculus of differential forms is used and the conventions of @xcite . in the absence of the surface charges and currents , the jump conditions on the boundary surface read ( see @xcite , pp .
150,151 ) : @xmath38 here @xmath39 is the 1-form density normal to the surface @xmath40 and @xmath41 , @xmath42 , are the two vectors tangential to @xmath40 .
the constitutive relations were formulated in ( [ con2 ] ) and ( [ con1 ] ) .
let @xmath40 be the plane @xmath22 in fig.1 that divides the two parts of space that are filled with two different homogeneous material media .
for the left half - space @xmath43 , we assume homogeneous matter characterized by constant values @xmath44 .
similarly , for the right half - space @xmath45 , we have the constant values @xmath46 .
somewhat related situations with reflected and scattered waves were discussed by lindell and sihvola @xcite .
consider a plane electromagnetic wave travelling along the @xmath47 axis in the left half - space . at the interface @xmath40 , such an incident wave will be partly reflected and partly refracted into the right half - space .
accordingly , the ansatz for the electromagnetic field configuration in the first medium will be a superposition of the right- and left - moving plane waves .
let @xmath48 and @xmath49 be the components of the incident wave with the argument @xmath50 , whereas @xmath51 and @xmath52 are those of the reflected wave with the argument @xmath53 . then , with the 1-forms @xmath54 , @xmath55 , and @xmath56 , @xmath57,\label{b1}\\{\cal d } & = & \varepsilon_1\varepsilon_0\left[\left(w_y + r_y\right)dz - \left(w_z + r_z\right)dy\right]\wedge dx + \alpha_1\,{\frac { k_1 } \omega}\,dx\wedge\left[\left(w_y - r_y\right)dy + \left(w_z - r_z\right)dz\right],\hspace{20pt}\label{d1}\\{\cal h }
& = & \varepsilon_1\varepsilon_0{\frac \omega { k_1}}\left[\left(w_y - r_y\right)dz - \left(w_z - r_z\right)dy\right ] - \alpha_1 \left[\left(w_y + r_y\right)dy + \left(w_z + r_z\right)dz\right].\label{h1}\end{aligned}\ ] ] a direct check shows that the maxwell equations ( [ max1]),([max2 ] ) , together with the constitutive relations ( [ con2]),([con1 ] ) , are satisfied provided the only nonvanishing component of the wave covector has the value @xmath58 analogously , the electromagnetic field configuration in the second half - space is represented by the right - moving transmitted wave : @xmath59 here the functions @xmath60 and @xmath61 , with the argument @xmath62 , describe the transmitted wave in the second medium .
analogously to ( [ k1 ] ) we have @xmath63
for concreteness , we confine our attention to harmonic waves . then , @xmath64 in order to construct the complete solution in the two regions
, we have to match the configurations ( [ e1])-([h1 ] ) and ( [ e2])-([h2 ] ) on the interface @xmath40 . using the jump conditions ( [ jump1 ] ) and ( [ jump2 ] ) with @xmath65 and @xmath66
, we find @xmath67\,t_z{}\vline\,{\hbox{\raisebox{-1.5ex } { \scriptsize{x=0}}}},\label{match3}\\ \varepsilon_0c\sqrt{\frac { \varepsilon_1}{\mu_1}}\left(w_z - r_z\right ) { } \vline\,{\hbox{\raisebox{-1.5ex}{\scriptsize{x=0 } } } } & = & \varepsilon_0c \sqrt{\frac { \varepsilon_2}{\mu_2}}\,t_z{}\vline\,{\hbox{\raisebox{-1.5ex } { \scriptsize{x=0 } } } } + [ \alpha]\,t_y{}\vline\,{\hbox{\raisebox{-1.5ex } { \scriptsize{x=0}}}}.\label{match4}\end{aligned}\ ] ] here @xmath68:= \alpha_2 - \alpha_1 $ ] is the _ jump _ of the axion field on the interface @xmath40 . the algebraic system ( [ match1 ] ) to ( [ match4 ] ) can be straightforwardly solved .
it yields the coefficients of the reflected ( @xmath69 and @xmath70 ) and transmitted ( @xmath71 and @xmath72 ) waves as combinations of those of the incident wave : @xmath73 ^ 2}{\lambda_0 ^ 2}}\right)a_{1,2 } + 2{\frac { [ \alpha ] } { \lambda_0}}\sqrt{\frac { \varepsilon_1}{\mu_1}}\,b_{1,2}\right]\,,\\ d_{1,2 } & = & \frac 1{\delta_\bot}\left [ -\,2{\frac { [ \alpha]}{\lambda_0 } } \sqrt{\frac { \varepsilon_1}{\mu_1}}\,a_{1,2 } + \left({\frac { \varepsilon_1 } { \mu_1 } } - { \frac { \varepsilon_2}{\mu_2 } } - { \frac { [ \alpha]^2}{\lambda_0 ^ 2 } } \right)\,b_{1,2}\right],\\ p_{1,2 } & = & \frac{2}{\delta_\bot}\sqrt{\frac{\varepsilon_1}{\mu_1}}\left[\left ( \sqrt{\frac { \varepsilon_1 } { \mu_1 } } + \sqrt{\frac { \varepsilon_2}{\mu_2 } } \right ) a_{1,2 } + { \frac { [ \alpha]}{\lambda_0}}\,b_{1,2}\right],\\ q_{1,2 } & = & \frac{2}{\delta_\bot}\sqrt{\frac{\varepsilon_1}{\mu_1}}\left [ -\,{\frac { [ \alpha]}{\lambda_0}}\,a_{1,2 } + \left(\sqrt{\frac { \varepsilon_1 } { \mu_1 } } + \sqrt{\frac { \varepsilon_2}{\mu_2}}\right ) b_{1,2}\right].\end{aligned}\ ] ] here @xmath74 ^ 2/\lambda_0 ^ 2\,$ ] . as a consistency check , we can easily see that the above formulas , for @xmath68 = 0 $ ] ( that is , either the axion is trivial everywhere or it has equal values for both material media ) , reduce to the well known expressions of the corresponding reflection and transmission coefficients for a plane wave , see born and wolf @xcite , sec.1.5 .
the result obtained clearly shows that despite the fact the constant axion drops out from the maxwell field equation , the electromagnetic wave feels " the presence of the axion by experiencing specific reflection and transmission effects .
= 5truecm in order to discuss the measureability of @xmath21 , let us consider the case when the first region of the space is vacuum , with @xmath75 and @xmath76 , whereas the second half - space is occupied by a material substance that has trivial dielectric and magnetic properties , @xmath77 , but has a constant axion piece with @xmath78 , see fig.2 .
let us assume , for simplicity , that the incident wave is linearly polarized with the electric vector directed along the @xmath79 axis , which is achieved by putting @xmath80 : @xmath81 then we find that the reflected wave is also a linearly polarized wave but with the electric vector tilted with respect to the y - axis [ @xmath82 , @xmath83,\ ] ] whereas the transmitted wave is linearly polarized with the electric vector also tilted with respect to both @xmath79 and @xmath84 axes , @xmath85.\ ] ] as we see , when @xmath86 the reflected wave is absent and the incident wave propagates from vacuum into vacuum without being distorted , @xmath87
. however , when @xmath88 the reflection takes place !
its presence is direct observational evidence for the axion piece .
it can be used for the experimental determination of the value of @xmath12 .
a qualitative check of the axionic nature of the substance is provided by the fact that the polarization of the reflected wave should be rotated with respect to the polarization of the incident wave .
furthermore , the quantitative estimate of @xmath12 can then be extracted from the measurement of the intensity and the angle of rotation of the reflected wave that explicitly depend on the value of the axion .
the above analysis can be generalized to the case when the wave is not normally incident on the interface between the two media , see fig.3 .
then , for an arbitrarily moving wave , we have in the first medium ( left half - space ) a superposition of the incident and the reflected waves : @xmath89,\label{b1a}\\ { \cal d } & = & \varepsilon_1\varepsilon_0\left[\left(w_y dz_{(i ) } - w_z dy \right)\wedge dx_{(i ) } + \left(r_y dz_{(r ) } - r_z dy\right)\wedge dx_{(r ) } \right]\nonumber\\ & & + \,\alpha_1\,{\frac { k_1 } \omega}\,\left[dx_{(i)}\wedge \left(w_y dy + w_zdz_{(i)}\right ) - dx_{(r)}\wedge\left(r_y dy + r_zdz_{(r)}\right)\right],\label{d1a}\\ { \cal h } & = & \varepsilon_1\varepsilon_0{\frac \omega { k_1}}\left[w_y dz_{(i ) } - r_ydz_{(r ) } - \left(w_z - r_z\right)dy\right]\nonumber\\ & & -\,\alpha_1 \left[\left(w_y + r_y\right)dy + w_zdz_{(i ) } + r_zdz_{(r ) } \right].\label{h1a}\end{aligned}\ ] ] here @xmath90 , @xmath91 , @xmath92 , @xmath93 describe the local coordinates adapted to the incident and the reflected wave , respectively , whereas @xmath94 with @xmath95 and @xmath96 .
the angles @xmath97 and @xmath98 give , as usual , the angles of the incident and the reflected waves with respect to the normal of the interface @xmath40 .
= 7truecm the transmitted wave in the right half - space has a form similar to ( [ e2])-([h2 ] ) : @xmath99 here we denote @xmath100 , @xmath101 , where @xmath102 describes the refraction angle .
similarly to ( [ t1 ] ) , we have @xmath103 with the argument @xmath104 .
now , the jump conditions ( [ jump1 ] ) and ( [ jump2 ] ) yield @xmath105\,t_z\right)\cos\theta_3{}\vline \,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0}}}},\label{match3a}\\ \varepsilon_0c\sqrt{\frac { \varepsilon_1}{\mu_1}}\left(w_z - r_z\right ) { } \vline\,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0 } } } } & = & \varepsilon_0c \sqrt{\frac { \varepsilon_2}{\mu_2}}\,t_z{}\vline\,{\hbox{\raisebox{-1.5ex } { \scriptsize{x = z=0 } } } } + [ \alpha]\,t_y{}\vline\,{\hbox{\raisebox{-1.5ex } { \scriptsize{x = z=0}}}},\label{match4a}\\ k_1\left(w_y\,\sin\theta_1 + r_y\,\sin\theta_2\right){}\vline \,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0 } } } } & = & k_2 t_y\,\sin\theta_3 \vline\,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0}}}},\label{match5a}\\ \varepsilon_1\varepsilon_0\left(w_z\,\sin\theta_1 - r_z\,\sin\theta_2\right ) { } \vline\,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0 } } } } & = & \left(\varepsilon_1 \varepsilon_0\,t_z + { \frac { [ \alpha]k_2}{\omega}}\,t_y\right)\sin\theta_3 \vline\,{\hbox{\raisebox{-1.5ex}{\scriptsize{x = z=0}}}}.\label{match6a}\end{aligned}\ ] ] the last two equations ( when combined with the rest ) yield the well known result that relates the angles of incidence , reflection , and refraction , namely , @xmath106 ( law of reflection ) and @xmath107 ( law of refraction ) . in addition , from the algebraic system ( [ match1a])-([match4a ] ) , we find the coefficients of the reflected ( @xmath69 and @xmath70 ) and transmitted ( @xmath71 and @xmath72 ) waves as combinations of those of the incident wave : @xmath108 ^ 2}{\lambda_0 ^ 2}}\right)\cos\theta_1 + \sqrt{\frac { \varepsilon_1 \varepsilon_2}{\mu_1\mu_2}}\left(\cos\theta_1 - \cos\theta_3\right)\right ] a_{1,2}\nonumber\\ & & \qquad + \,{\frac { [ \alpha]}{\lambda_0}}\sqrt{\frac { \varepsilon_1}{\mu_1 } } \,(\cos\theta_1 + \cos\theta_2)\,b_{1,2}\bigg\},\\ d_{1,2 } & = & { \frac { \cos\theta_3}{\delta}}\bigg [ -\,{\frac { [ \alpha ] } { \lambda_0}}\sqrt{\frac { \varepsilon_1}{\mu_1}}\,(\cos\theta_1 + \cos\theta_2 ) \,a_{1,2}\nonumber\\ & & \qquad + \,\left[\left({\frac { \varepsilon_1 } { \mu_1 } } - { \frac { \varepsilon_2}{\mu_2 } } - { \frac { [ \alpha]^2}{\lambda_0 ^ 2}}\right ) \cos\theta_1 + \sqrt{\frac { \varepsilon_1\varepsilon_2}{\mu_1\mu_2 } } \left(\cos\theta_3 - \cos\theta_1\right)\right ] b_{1,2}\bigg\},\\ p_{1,2 } & = & \sqrt{\frac { \varepsilon_1}{\mu_1}}{\frac { ( \cos\theta_1 + \cos\theta_2)}{\delta}}\left[\left(\sqrt{\frac { \varepsilon_1 } { \mu_1 } } \,\cos\theta_3 + \sqrt{\frac { \varepsilon_2}{\mu_2}}\,\cos\theta_2\right ) a_{1,2 } + { \frac { [ \alpha]}{\lambda_0}}\,\cos\theta_3\,b_{1,2}\right],\\ q_{1,2 } & = & \sqrt{\frac { \varepsilon_1}{\mu_1}}{\frac { ( \cos\theta_1 + \cos\theta_2)}{\delta}}\left [ -\,{\frac { [ \alpha]}{\lambda_0}}\,\cos\theta_2 \,a_{1,2 } + \left(\sqrt{\frac { \varepsilon_1 } { \mu_1}}\,\cos\theta_2 + \sqrt{\frac { \varepsilon_2}{\mu_2}}\,\cos\theta_3\right ) b_{1,2}\right].\end{aligned}\ ] ] here we denoted @xmath109 ^ 2/\lambda_0 ^ 2 $ ] . for the special case @xmath110 , we recover @xmath111 .
as we saw , we can read off from the coefficients @xmath69 and @xmath70 of the reflected wave the jump @xmath112 $ ] of the axion piece . in principle , one can also measure the axion piece by means of observing the properties of the transmitted wave inside the medium .
however , the study of the reflected wave alone is clearly preferable as it provides a simple and elegant scheme of experimental determining the properties of a substance without destructing the latter .
financial support from the dfg project he-528/20 - 1 is gratefully acknowledged .
hehl , y. itin , and yu.n .
obukhov , _ recent developments in premetric electrodynamics . _ in : _ proceedings of the 3rd _ _ summer school in modern mathematical physics , _ august 2004 , zlatibor , serbia , b. dragovich et al .
( institute of physics : belgrade , 2005 ) , 34 pages ; to be published ( also in arxiv.org ) .
a. lakhtakia , _ ten years past the post .
_ in : proc .
spie ( int .
soc . for optical engineering ) , vol . * 5508 * _ complex mediums v : light and complexity _ , m.w .
mccall , g. dewar ( eds . ) ( august 2004 ) pp .
8594 .
lindell and a.h .
sihvola , _ electromagnetic image theory for the pemc sphere _ , iee proceedings on science , measurement and technology , submitted , 11 pages , preprint helsinki univ . of technology ,
electromagn . lab . rep .
445 ( oct 2004 ) .
ni , _ a non - metric theory of gravity_. dept .
physics , montana state university , bozeman .
preprint december 1973 .
the paper is available via + http://gravity5.phys.nthu.edu.tw/webpage/article4/index.html .
ni , _ equivalence principles and precision experiments . _ in _ precision measurement and fundamental constants ii , _ b.n .
taylor , w.d .
phillips , eds .
( us ) spec .
617 , us government printing office , washington , dc ( 1984 ) .
a. sihvola and i.v .
lindell , _ perfect electromagnetic conductor as building block for complex materials , _ electromagnetics ( special issue for the 2004 bianisotropics meeting in ghent , belgium ) , submitted , 8 pages , preprint helsinki univ . of technology , electromagn . lab .
443 ( oct 2004 ) .
tellegen , _ the gyrator , an electric network element , _ philips technical review * 18 * ( 1956/57 ) 120124 . reprinted in h.b.g .
casimir and s. gradstein ( eds . ) _ an anthology of philips research .
_ philips gloeilampenfabrieken , eindhoven ( 1966 ) pp .
|
in order to settle the problem of the post constraint " in material media , we consider the propagation of a plane electromagnetic wave in a medium with a piecewise constant axion field .
although a constant axion field does not affect the wave propagation in a homogeneous medium , we show that the reflection and transmission of a wave at an interface between the two media is sensitive to the difference of the axion values .
this observation can be used to determine experimentally the axion piece in matter despite the fact that a constant axion value does not contribute to the maxwell equations .
= 10000
|
in the progress of high - energy physics , so - called energy frontier have reached to a few tev region , where discoveries of new phenomena are expected .
in such situation , it is well known that the five unites of electric charge , q = 0 , 1/3 , 2/3 , 1 , 4/3 appear in guts@xcite , where the 4/3 charge is carried by @xmath0 gauge boson , which is considered to be the particle responsible for proton decay .
thus , so far as the charge units are concerned , it seems that guts are too complicated to be the final theory of elementary particles .
especially , in any gauge model such as guts , the @xmath4 symmetry concerning to the conservation of electric charge is absolutely maintained in the final stage of steps of symmetry breaking .
furthermore , it is known that the electric charge remains and is conserved even in a critical condition such as the gravitational collapse of matter , where almost all quantum numbers looses their meanings@xcite .
thus , the electric charge seems to be one of the most fundamental quantities in particle physics .
it is known that the simplest model based on the elementarity of electric charge is the rishon model@xcite of leptons and quarks . in this model ,
the leptons and quarks are given as = vvv , e^- = [ eq : rishon ] where the @xmath2 rishon has the electric charge @xmath5 , while the @xmath3 rishon is purely electrically neutral .
though models of leptons and quarks are , however , in the stage of schematical model , some remarkable minimal dynamical approaches are proposed@xcite . for further steps ,
it is possible to treat schematically possible processes of sub - constituents under minimal dynamical assumptions for composite system .
on the basis of such approach , we will investigate possible existence of new processes , and possible dynamical approach to sub - system of leptons and quarks .
this paper is concerning to theses problems .
we will assume that leptons and quarks are composite system of sub - constituents confined with the hypercolor singlet conditions , as have been assumed in dynamical rishon mode . in our schematical approach ,
it is pointed out that possible existence of new proton decay mode , which is different from that of guts , is suggested .
we will follow the policy of so - called dynamical rishon model@xcite , where it is assumed that the hypercolor singlet states of composite system are realized as ordinary particles such as leptons and quarks . though the rishon is assumed to be dirac particle in the dynamical rishon model , while we have proposed , under the development of majorana particle model of neutrinos , majorana v rishon model@xcite . in this paper , however , we will at first restrict ourselves to dirac particle model , and the case of majorana model will be investigated in later . in the dynamical rishon model@xcite , the exact symmetry realized by the local gauge group , su_3(c ) su_3(h ) [ eq : cxh ] is assumed for rishons , where c and h represents color and hypercolor , respectively .
the t rishon has @xmath6 and transforms like a @xmath7 representation of @xmath8 .
the v rishon has @xmath9 and is assigned to @xmath10 .
the @xmath11 or @xmath12 is in @xmath13 and @xmath14 , respectively .
possible proton decay modes in guts . the proton decay is caused by x and y bosons . in sub - constituent level , the x or y boson corresponds to the configuration @xmath15 , @xmath16 , respectively . ] we will treat for simplicity the simplest case of one generation , and at first treat the case of dirac type v rishon case . some characteristics and problems appearing in possible majorana type v rishon model will be discussed in later . in the dynamical rishon model ,
the sub - constituents t and v are supposed to be confined in leptons and quarks by hypercolor confinement mechanism realized by the hypercolor singlet condition . under such condition ,
the composite system exhibits various rearrangement processes .
the electroweak structure supposed to appear as a result of such mechanism@xcite . in this paper
, we will investigate rearrangements of sub - constituents under the confinement condition , together with possible sub - electroweak structure .
it should be emphasized that in the rearrangement of sub - constituent , possible existence of new mode of proton - decay is suggested , which is different from that of guts case where the proton - decay is mediated by the @xmath17 and @xmath18 gauge bosons with the color index @xmath19 .
the tttvvv configuration belong to @xmath8 singlet , and is identified to @xmath20 boson . in the same way , the guts gauge bosons @xmath17 and @xmath18 with color index @xmath19 corresponds to the configuration ttttvv and @xmath21 , respectively , and they belong to @xmath22 representation of of @xmath8 . in our sub - constituents model , it should be emphasized that the process caused by exchange of the composite state with @xmath23 configuration is also expected .
the @xmath23 belongs to @xmath22 representation , similar to the case of @xmath17 and @xmath18 bosons .
it is well - known that the exchange of @xmath17 and @xmath18 bosons causes possible proton - decay in guts , as is shown in fig.[fig : p - gut_all-1 ] .
then , what processes are caused by the exchange of composite state of @xmath23 ?
it is known that the x boson with electric charge 4/3 and y boson with charge 1/3 mediate the proton decay process in guts such as @xmath24 model .
the boson x or y correspond to ttttvv , @xmath21 configuration , respectively . an elementary rearrangement in sub - system .
, width=283,height=94 ] it should be emphasized , however , the new type of proton decay mode appears , which is mediated by the exchange of @xmath23 pair with charge 1/3 , in a composite model of leptons and quarks . the elementary process mediated by the rearrangement of sub - constituents , given in fig .
[ fig : tv_element-2 ] , v + t + [ eq : elementary ] will cause the processes ( ttv ) ( ttt ) , ( ) ( ) [ eq : induced ] which are represented as u e^+ , d u^c [ eq : induced2 ] possible @xmath25 process through the exchange of @xmath26 pair . ] the explicit rearrangement diagram of the process in eq .
( [ eq : induced2 ] ) is given in fig .
[ fig : tv_f0 - 3 ] .
thus , the transition of composite quarks , which is caused by this rearrangement ( uud ) ( e^+ u^c u ) [ eq : protondecay0 ] is induced .
this is just the proton decay process , p e^+ + _ 0 [ eq : protondecay ] mediated by the exchange of @xmath27 pair .
that is , in our model the proton decay is caused without the x or y gauge bosons , which are characteristic in guts as shown in fig .
[ fig : p - gut_all-1 ] .
the elementary process in eq .
( [ eq : elementary ] ) will be represented as fig .
[ fig : tv_element-2 ] .
this process will also represented as an effective process mediated by the effective boson @xmath28 with color indexes @xmath19 , as shown in fig . [
fig : tv_element-4 ] .
we will introduce an effective interaction of this process , and at the same time , we will investigate possible scenario to identify it as the fundamental one . for simplicity , we will restrict to the one generation case in this paper . assuming the weak - doublet in sub - level as , _
l = t v^c_l [ eq : rishon2 ] elementary process in sub - system caused by the exchange of effective sub - gauge boson @xmath29 .
, width=283,height=88 ] the effective electroweak lagrangian in our system is given as @xmath30 where @xmath31 and @xmath32 are the gauge fields concerning to sub - system .
the electric charge of component of doublet in eq .
( [ eq : rishon2 ] ) is given as q = i_3 + y [ eq : charge_sub ] in the unit of e/3 , where @xmath33 .
after symmetry breaking , the sub - gauge bosons @xmath34 , @xmath35 and @xmath36 bosons appear , where @xmath34 is characterized by the charge unit e/3 .
the ordinary electroweak gauge bosons @xmath37 , @xmath38 and @xmath39 will be symbolically expressed by making use of the sub - gauge bosons @xmath40 , @xmath41 and @xmath42 as , @xmath43 where @xmath37 is @xmath8 singlet , while @xmath44 belong to @xmath45 representation . it should be noted that @xmath46 and @xmath47 are realized as singlet as it is , provided that the correspondences @xmath48 and @xmath49 together with an appropriate mixing are assumed .
( [ eq : effective_ew ] ) is a symbolical representation of composite system based on yet to be known mechanism .
the sub - gauge bosons @xmath50 should be confined in the ordinary gauge bosons belonging to the level of leptons and quarks . the effective electroweak structure in the level of leptons and quarks will be induced on the basis of eq .
( [ eq : effective_ew ] ) . in order to investigate the properties of composite system given by eq .
( [ eq : effective_ew ] ) , let us examine the familiar ( inverse ) @xmath51-decay process of d quark , d + _ e u + e^- [ eq : beta ] in the picture of fundamental constituents , eq .
( [ eq : beta ] ) is rewritten as , ( ) + ( vvv ) ( ttv ) + ( ) [ eq : beta - rishon ] thus , we see that the fundamental interaction in our process is given as = g_w^(s ) |t^(1 + _ 5 ) v^c w_^(s ) [ eq : fundamental_int ] and corresponding current is = |t^(1 + _ 5 ) v^c [ eq : fundamental_current ] which belongs to @xmath45 representation .
the interaction in the process given by eq.([eq : elementary ] ) , v + t + which is caused by only @xmath45 single current , should be extremely suppressed . on the contrary , in our scheme
, the composite " process with 3-fold single elementary processes such as eq.([eq : beta - rishon ] ) should be not so suppressed . the single elementary process in eq.([eq : elementary ] )
is expected to be extremely suppressed in the same way as the 3-body quarks are almost completely confined in hadrons and realization of colored configuration is extremely suppressed .
the process such as eq.([eq : beta ] ) or equivalently eq.([eq : beta - rishon ] ) are enhanced , similar to the realization of the 3-body color - singlet states of quarks are enhanced .
that is , the process given in eq.([eq : beta - rishon ] ) is essentially the non - perturbative process with extreme suppression of non - color - singlet currents .
thus , we will investigate possible scheme based on fundamental interaction .
we have introduced the effective interaction corresponding to the rearrangement picture in sub - system .
we will investigate here the possible case that eq.([eq : ew_sub ] ) is not an effective interaction but the most fundamental one .
that is , it is worth while to investigate the possibility that the fundamental interaction cause various effective interactions such as the standard model itself , under the restriction of exact symmetry @xmath8 given in eq.([eq : cxh ] ) . in appearance of effective interaction , especially , color and hypercolor confinement condition
will carry most important role .
it should be noted that this approach is partially similar to the case of meson theory for nuclear force where various properties of nucleus are derived on the basis of qcd . it should be emphasized that under development along such strategy it may become possible to resolve the hierarchy problem on coupling constants and masses , which are only given as parameters in the standard model or guts .
in progress of neutrino physics , the existence of majorana particle seems to be almost confirmed .
in addition to the neutrinos , are there any majorana particle in nature ?
in such situation , we have proposed in a few years ago a model of leptons and quarks on the basis of possible majorana v theory together with possible existence of its partner in the framework of rishon model@xcite .
we will investigate possible representation of sub - constituent in our majorana v model , under the local gauge group @xmath8 . in the simplest @xmath52 hypercolor model ,
the assignment to realize the hypercolor singlet states is rather complicated@xcite .
the t rishon is assigned to @xmath53 representation of @xmath8 .
the v rishon is assigned to @xmath54 .
the @xmath11 and @xmath12 are , respectively , in @xmath54 and @xmath55 .
the t rishon has @xmath6 and transforms like a @xmath53 representation of @xmath8 .
the v rishon has @xmath9 and is assigned to @xmath54 .
the @xmath12 corresponds to @xmath55 representation . for majorana type @xmath3
, the @xmath11 is equivalent to @xmath3 and also assigned to @xmath54 .
the quarks u and d belong to @xmath56 representation .
it should be noted that the results given in the previous sections will be almost maintained in majorana v model , except the appearance of characteristic picture of majorana particles that the fermion number is not conserved .
furthermore , the restrictions due to color and hypercolor confinement will be rather weakened , then possible existence of some new particles are expected .
in this paper , we have investigated possible existence of new proton decay mechanism on the basis of a rearrangement picture of sub - system of leptons and quarks , in which the sub - constituents are confined in leptons and quarks by the color and hypercolor confinement condition based on @xmath8 group .
we have discussed the possibility that the pre - electroweak structure appears in the rishon level , and a certain kind of confinement of pre - gauge bosons realizes the well - known electroweak structure in the level of leptons and quarks . in our approach , it may be possible to calculate the magnitude of coupling constants and the masses of composite leptons and quarks , which are given as parameters in pure field theoretical approaches .
we have assumed in this paper that the gauge structure of the standard model appears as a result of the dynamics of the sub - levels of leptons and quarks .
is the sub - system of leptons and quarks completely subject to quantum mechanics ? it should be noted that the standard formulation of the rishon model is not yet known , and it is probable that the rishon behaves as if it is partially the objects beyond the ordinary quantum field theory@xcite,@xcite . in order to approach to such system , however
, it will be meaningful to treat it in the framework of conventional field theory with color and hypercolor confinement .
finally , is there really a sub - level below the level of leptons and quarks ? in our opinion , such a sub - level surely exists , and disclosing it will make it possible to predict theoretically quantities such as the higgs coupling constants , the magnitudes of symmetry breaking @xmath57 , the mass spectrum and mixing parameters of all particles , etc .
we conclude this paper with the note that we have restricted ourselves in this paper to the proposal of the framework of possible approach to sub - system , and details will be discussed in near future .
this work was supported by grant - in - aid of japanese ministry of education , science , and culture ( 15540384 ) .
|
the problem of elementarity of electric charge is investigated .
it is known that possible proton decay is caused by @xmath0 and @xmath1 bosons which are characteristic in guts .
the charge structure of guts is , however , not simple , and the five types of charge units appear .
we investigate the problem of elementarity in taking into account of the simplest model of the charge structure known as the rishon model , where the charge units are known as 1/3 ( @xmath2 ) and 0 ( @xmath3 ) . in order to approach to sub - structure together with the problem of simplicity of charge structure ,
possible rearrangements of rishons are investigated on the basis of color and hypercolor confinement condition .
further , possible existence of the simplest electroweak sub - structure is discussed .
our model predicts possible proton decay mode , in which @xmath0 and @xmath1 bosons are not concerned .
|
symmetry occurs in many constraint satisfaction and optimisation problems @xcite . for example , suppose we have a proper coloring of a graph , and we permute the colors then we will obtain another symmetric coloring .
symmetries are problematic as they increase the size of the search space .
we will waste time visiting symmetric solutions .
worse still , we will waste even more time visiting the ( many failing ) parts of the search tree which are symmetric to already visited states . a common and effective method to deal with symmetry is to add constraints which eliminate some , but not all symmetric solutions ( e.g. @xcite ) . in this paper , we survey recent results on symmetry breaking constraints .
we hope that this survey is of wider interest as many of the results are likely to translate to other domains like planning , model checking and heuristic search .
symmetry breaking constraints have a number of good and bad properties . on the positive side
, a few simple constraints can often eliminate most if not all symmetry in a problem quickly and easily .
in addition , propagation between the problem and the symmetry breaking constraints can reduce search considerably . on the negative side , we pick out particular solutions in each symmetry class , and this may conflict with the direction of the branching heuristic .
an alternative to posting static symmetry breaking constraints is a more dynamic approach that modifies the search method to ignore symmetric states @xcite .
the advantage of a dynamic method is that it can reduce the conflict with the branching heuristic .
however , we may also get less propagation . the results reported here on static symmetry breaking constraints typically have implications for dynamic methods as most dynamic methods are equivalent to adding static symmetry breaking constraints `` on the fly '' .
there are several aspects of symmetry breaking that this survey does not cover .
we do not discuss methods to identify symmetry in a problem .
see , for instance , @xcite .
we suppose that the symmetries are known in advance as they often are , and our challenge is merely to deal with them .
the survey also does not cover other active area of research like the intersection of symmetry reasoning and nogood learning .
see , for instance , @xcite . for background material on constraint satisfaction in general , see a text like @xcite .
we consider a combinatorial problem studied in several earlier works on symmetry breaking .
the equidistant frequency permutation array ( efpa ) problem @xcite is a challenging problem in coding theory .
the goal is to find a set of @xmath0 code words , each of length @xmath1 such that each word contains @xmath2 copies of the symbols 1 to @xmath3 , and each pair of code words is hamming distance @xmath4 apart .
for example , for @xmath5 , @xmath6 , @xmath7 , @xmath8 , one solution is : @xmath9 this problem has applications in communication theory , and is related to other combinatorial problems like finding orthogonal latin squares . we can model this problem by a @xmath0 by @xmath1 array of decision variables with domains @xmath10 to @xmath3 . a _ symmetry _ of such a problem is a bijection @xmath11 that maps solutions onto solutions @xcite .
we consider two special classes of symmetries : _ variable symmetries _ which act just on variables , and _ value symmetries _ which act just on values . both types of symmetry can be found in our model of the efpa problem .
this model of the efpa problem has variable symmetry since we can permute the variables in any two rows , or in any two columns and still have a solution .
for example , we can swap the first column with the third without changing the frequency of symbols in each row or the hamming distance between any two rows : @xmath12 the model also has value symmetry since we can interchange values throughout any solution .
for example ,
the value symmetry @xmath13 swaps the value 0 with 2 throughout solution ( [ epfa ] ) , again without changing the frequency of symbols in each row or the hamming distance between any two rows : @xmath14
a simple but effective method to deal with symmetry is to add constraints which eliminate some ( but not all ) symmetric solutions in each symmetry class . to eliminate the value symmetry in the efpa problem
, we can add a precedence constraint @xcite to ensure that the values in the bottom row occur in order .
that is , the value 0 first occurs on the bottom row before 1 , and 1 first occurs before 2 .
the bottom row of solution ( [ epfa ] ) is @xmath15 which satisfies this symmetry breaking constraint .
the value 0 first occurs in the first position , the value 1 later on in the third position , and the value 2 even later on in the fifth position .
suppose we swap 0 with 2 , then the bottom row of solution ( [ epfa ] ) becomes @xmath16 , the bottom row of ( [ epfa3 ] ) .
this violates the precedence symmetry breaking constraint .
the value 1 first occurs in the third position , which is before the first occurrence of the value 0 in the fifth position .
hence , by posting this symmetry breaking constraint , we only find solution ( [ epfa ] ) , and the symmetric solution ( [ epfa3 ] ) is eliminated . two important properties of symmetry breaking constraints are soundness and completeness .
a set of symmetry breaking constraints is _ sound _ iff it leaves at least one solution in each symmetry class , and _ complete _ iff it leaves at most one solution in each symmetry class . the precedence constraint is , for example , sound and complete . where do symmetry breaking constraints like this
come from in general ?
the lex - leader method offers a generic method for deriving a sound and complete set of symmetry breaking constraints for variable symmetries @xcite and value symmetries @xcite .
the method picks out the lexicographically smallest solution in each symmetry class . for every symmetry @xmath11 , it posts a lexicographical ordering constraint : @xmath17 where @xmath18 to @xmath19 is some fixed ordering on the variables in the problem .
consider again the value symmetry , @xmath13 that swaps 0 and 2 .
we can describe this by the mapping @xmath20 .
the lex - leader method eliminates this symmetry with the ordering constraint : @xmath21 where @xmath18 to @xmath22 is some ordering of the 24 decision variables modelling our 4 by 6 instance of the efpa problem .
this simplifies to : @xmath23 suppose we order the variables in the matrix row - wise from top left to bottom right
. then this constraint would accept solution ( [ epfa ] ) , but eliminate solutions ( [ epfa2 ] ) and ( [ epfa3 ] ) .
many static symmetry breaking constraints can be derived from such lex - leader constraints .
for example , precedence constraints to break the symmetry due to interchangeable values can be derived from them @xcite .
efficient algorithms have been developed to propagate lexicographical ordering constraints @xcite .
one way of constructing new symmetry breaking constraints is to use symmetry itself @xcite .
any symmetry acting on a set of symmetry breaking constraints will itself break the symmetry in a problem .
this requires us to define the action of a symmetry on a set of symmetry breaking constraints .
we defined symmetry as acting on assignments , mapping solutions to solutions .
we can lift this definition to constraints .
for example , the action of a variable symmetry on a constraint changes the scope of the constraint ( that is , the variables on which the constraint acts ) .
given a set of symmetries @xmath24 of c , if s is a sound ( complete ) set of symmetry breaking constraints for @xmath24 then @xmath25 for any @xmath26 is also a sound ( complete ) set of symmetry breaking constraints for @xmath24 .
different symmetries pick out different solutions in each symmetry class . in fact , if we have a particular solution in mind , we can pick it out using a suitable symmetry of a set of symmetry breaking constraints .
let @xmath27 be the set of solutions of a set of constraints @xmath28 .
given a set of symmetries @xmath24 of a set of constraints c , a sound set s of symmetry breaking constraints , and any solution a of c , then there is a symmetry @xmath26 such that @xmath29 .
applying symmetry to a set of symmetry breaking constraints does not change the symmetries which are eliminated .
we say that a set of constraints @xmath30 _ breaks _ a symmetry @xmath11 of a problem @xmath28 iff there exists a solution @xmath31 of @xmath32 such that @xmath33 is not a solution of @xmath32 , and
_ completely breaks _ a symmetry @xmath11 iff for each solution @xmath31 of @xmath32 , @xmath33 is not a solution of @xmath32 .
given a symmetry group @xmath24 , we say that a set of constraints ( completely ) breaks @xmath24 iff it breaks every non - identity symmetry in @xmath24 .
given a problem @xmath28 and a symmetry group @xmath24 , if @xmath30 ( completely ) breaks @xmath24 then @xmath25 ( completely ) breaks @xmath24 for any @xmath26 .
we have used these ideas as the theoretical basis for a hybrid symmetry breaking method that tries to have the best of both worlds , posting static symmetry breaking constraints dynamically during search according to the direction of the branching heuristic @xcite .
the symmetry groups seen in practice can be very large . as a result , symmetry breaking can be computationally challenging .
for example , if we have a @xmath3 interchangeable values ( as in our efpa model ) , then we have @xmath34 symmetries . to eliminate each of these symmetries requires a separate lex - leader constraint . as a consequence ,
the lex - leader method is not tractable in general .
an alternative is to break symmetry partially by posting just a subset of the lex - leader constraints ( e.g. @xcite ) .
crawford _ et al . _
( @xcite ) proved that breaking symmetry completely by adding constraints to eliminate symmetric solutions is computationally intractable in general .
more specifically , they prove that , given a matrix model with row and column symmetries , deciding if an assignment is the smallest in its symmetry class is np - hard where we append rows together and compare them lexicographically
. there is , however , nothing special about appending rows together or comparing assignments lexicographically .
we could break symmetry with _ any _ total ordering over assignments .
recently we have shown that , under modest assumptions , breaking symmetry remains intractable if we use a different ordering of variables , or even a different ordering over solutions than the lexicographical ordering .
more precisely , we show that the problem is np - hard for any _ simple _ ordering where , given an assignment , we can compute the position in the ordering in polynomial time and , in the reverse , given a position in the ordering , we can compute the associated assignment in polynomial time .
given any simple ordering , there exists a symmetry group such that deciding if an assignment is smallest in its symmetry class according to this ordering is np - hard .
since breaking symmetry appears intractable in general , a major research direction is to identify special cases which occur in practice where the symmetry group is more tractable to break .
we consider two such cases : row and column symmetry , and value symmetry .
a matrix of decision variables has _ row symmetry _
iff given a solution , any permutation of the rows is also a solution .
similarly , it has _ column symmetry _
iff given a solution , any permutation of the columns is also a solution .
for example , our model of the efpa problem has both row and column symmetry . row and
column symmetry occurs in many models with matrices of decision variables @xcite .
we can break row and and column symmetry using the lex - leader method .
however , this is not practical in general .
a @xmath35 by @xmath36 matrix has @xmath37 row and column symmetries , and each symmetry would require a separate lexicographical ordering constraint . to break all row symmetry , we can post a linear number of constraints that lexicographically order the rows . similarly , to break all column symmetry we can post a linear number of constraints that lexicographically order the columns . when we have both row and column symmetry , we can post a @xmath38 constraint that lexicographical orders both the rows and columns @xcite . in fact , @xmath38 constraints can be derived from a subset of the constraints introduced by the lex - leader method @xcite . as a result
, @xmath38 does not break all row and column symmetry .
nevertheless , it is often highly effective in practice .
consider again solution .
if we order the rows of lexicographically , we get a solution in which rows and columns are ordered lexicographically : @xmath39 similarly if we order the columns of lexicographically , we get a different solution in which both rows and columns are again ordered lexicographically : @xmath40 all three solutions are thus in the same row and column symmetry class . note that both and satisfy the @xmath38 constraint
. therefore @xmath38 can leave multiple solutions in each symmetry class and is not a complete symmetry breaking method .
in fact , @xmath38 is a very incomplete method for breaking symmetry
. it can leave @xmath41 symmetric solutions in an @xmath42 by @xmath42 matrix model .
in addition , it only partially provides tractability . whilst it is polynomial to check a @xmath38 constraint given a complete assignment , it is not tractable to propagate it completely given a partial assignment . that is
, pruning all symmetric values is np - hard .
propagating the constraint is np - hard .
we have identified three tractable cases where we can break all row and column symmetry in polynomial time @xcite : ( 1 ) a matrix with a bounded number of rows ( or columns ) ; ( 2 ) a 0/1 matrix model where every row sum is 1 ; ( 3 ) an all - different matrix where all entries are different .
the first two are perhaps the most interesting and useful cases .
we have , for example , used the first case as the basis of a complete method for breaking row and column symmetry .
we used this to measure how many symmetries are left by in practice .
see table 1 for some results that demonstrate actually leaves very few symmetric solutions in practice despite the worst case result that it can leave a factorial number of such solutions .
more recently , yip and van hentenryck have turned this theoretical result into a complete and efficient method for breaking all row and column symmetry in matrix models with a small number of rows ( or columns ) .
another interesting research direction is to identify other constraints that can be effectively added to to increase the amount of row and column symmetry broken @xcite . .[t
: t1 ] equidistant frequency permutation array problems .
number of solutions left when posting symmetry breaking constraints . [ cols=">,^,^,^,^ " , ]
a second important class of symmetries which are more tractable to break than symmetries in general are value symmetries .
when a problem has many value symmetries , the lex - leader method again introduces too many symmetry breaking constraints to be practical .
puget has proposed an effective alternative that maps value symmetries into variables symmetries @xcite .
his method channels into some auxiliary variables which , as in @xcite , permit the ( symmetry breaking ) constraints to be easily specified .
@xmath43 records the first index using each value @xmath44 with constraints of the form : @xmath45 puget then breaks the variable symmetry on the @xmath43 by posting binary ordering constraints .
for example , if the values 1 to @xmath3 are interchangeable , we can strictly order the @xmath43 : @xmath46 this ensures that the first occurrence of 1 is before that of 2 , that of 2 is before 3 , etc .
this is the condition enforced by the precedence constraint mentioned previously .
puget proved that we can break _ any _ number of value symmetries ( and not just those due to interchangeable values ) with a linear number of ordering constraints on the @xmath43 .
this means that we can break any number of value symmetries in polynomial time .
unfortunately , this decomposition into ordering constraints on @xmath43 hinders propagation , and we may not prune all symmetric values .
this is not surprising as pruning all symmetric values is intractable in general . given a set of value symmetries and a partial assignment , pruning all symmetric values is np - hard .
we have , however , identified a common case where pruning all symmetric values is tractable .
if values partition into a fixed number of sets and values within each set are interchangeable then we can prune all symmetric values in polynomial time using a global regular constraint @xcite or one of its decompositions @xcite .
when there is a single set of interchangeable values , we can prune all symmetric values using a global precedence constraint @xcite or its decomposition @xcite .
another important question is how we deal with with combinations of symmetries .
we have already seen that we can take a symmetry group that is tractable to break completely ( row symmetry ) , combine it with another symmetry group that is tractable to break completely ( column symmetry ) , and end up with a symmetry group ( row and column symmetry ) that is intractable to break and where we lack good methods to eliminate _ all _ symmetric solutions in general .
another challenging combination is when we have both variable and value symmetries .
how do we break both types of symmetry simultaneously ? we can use the lex - leader method but this is not practical when we have many variable and value symmetries .
unfortunately , puget s method for breaking value symmetry on the @xmath43 variables is not compatible in general with breaking variable symmetry via the lex - leader method on the @xmath47 variables .
this corrects theorem 6 and corollary 7 in @xcite which claim that the two methods are compatible if we use the same ordering of variables in each method .
there exist problems on which posting lex - leader constraints to break variable symmetries and applying puget s method to break value symmetries remove all solutions in a symmetry class irrespective of the orderings on variables used by both methods .
there is , however , an important and common case where the two methods for breaking variable and value symmetry are compatible . if values partition into interchangeable sets then it is always consistent to post symmetry breaking constraints using both methods provided we index their variables identically in both cases @xcite .
this is again a case where tractability is limited as pruning all symmetric values is np - hard .
a number of important and challenging questions about symmetry breaking constraints remain including : identifying tractable cases : : : are there other common types of symmetry which occur in practice that are polynomial to break ?
for instance , suppose we are coloring the edges of a graph , and we have a clique of interchangeable vertices .
this induces a particular symmetry on the decision variables representing the colors of the different edges in the graph .
how do we efficiently break this large symmetry group ?
exploring other orderings : : : are there other orderings besides the lexicographical ordering with which we can break symmetry effectively ? we have seen some promise for the gray code ordering .
there are , however , many other orderings we could consider .
for example , frisch _ et al .
_ have proposed the multiset ordering @xcite . even though this is only a partial ordering ,
it has also shown some promise .
in addition , can we use a different ordering in each symmetry class ? reducing branching conflict : : : one of the major issues with symmetry breaking constraints is that they may conflict with the direction of the branching heuristic .
how can we retain the benefits of static symmetry breaking constraints and of dynamic branching heuristics , whilst avoiding such conflict ?
how do we get the best of static and dynamic symmetry breaking methods ? how do we choose static symmetry breaking constraints that align with a dynamic branching heuristic ? studying combinations of symmetries : : : how do we eliminate combinations of symmetries effectively ? there are issues of both soundness ( which symmetry breaking constraints can be combined together safely ? ) , completeness ( how we do eliminate all combinations of symmetry ? ) and tractability ( which symmetry groups that are tractable to break individually combine together to give a symmetry group which is also tractable to break ? ) .
dominance detection and elimination : : : in constraint optimisation problems , the notion of dominance plays a very similar role to symmetry .
one ( partial ) solution dominates another if it is sure to be at least as good in quality .
how do we identify dominating solutions ?
is there a generic method like lex - leader for adding constraints that remove dominated solutions ? what are interesting and useful tractable cases ?
backofen , r. , and will , s. 1999 . excluding symmetries in constraint - based search . in jaffar , j. , ed .
, _ proc . of the 5th international conf . on principles and practice of constraint programming _ , lncs 1713 , 7387 .
springer .
benhamou , b. ; nabhani , t. ; ostrowski , r. ; and sadi , m. 2010 . enhancing clause learning by symmetry in sat solvers . in _ proc . of 22nd
ieee international conf . on tools with artificial intelligence (
ictai 2010 ) _ , volume 1 , 329335 .
ieee computer society .
bessiere , c. ; narodytska , n. ; quimper , c .- g . ; and walsh , t. 2011 . the alldifferent constraint with precedences . in achterberg , t. , and beck , j. c. , eds .
, _ proc . of 8th international conf . on integration of ai and or techniques in constraint programming for combinatorial optimization problems ( cpaior 2011 ) _ ,
_ lncs _ 6697 , 3652 .
springer .
chu , g. ; stuckey , p. ; de la banda , m. ; and mears , c. 2011 .
symmetries and lazy clause generation . in walsh ,
of 22nd international joint conf . on artificial intelligence
( ijcai-2011 ) _ , 516521 .
ijcai / aaai .
crawford , j. ; luks , g. ; ginsberg , m. ; and roy , a. 1996 .
symmetry breaking predicates for search problems . in _ proc . of the 5th international conf . on knowledge representation and reasoning , ( kr 96 )
_ , 148159 .
fahle , t. ; schamberger , s. ; and sellmann , m. 2001 .
symmetry breaking . in walsh , t. , ed .
, _ proc . of 7th international conf . on principles and practice of constraint programming ( cp2001 ) _ , 93107 .
springer .
flener , p. ; frisch , a. ; hnich , b. ; kiziltan , z. ; miguel , i. ; pearson , j. ; and walsh , t. 2001a .
symmetry in matrix models .
technical report apes-30 - 2001 , apes group .
presented at symcon01 ( symmetry in constraints ) , cp2001 post - conference workshop .
flener , p. ; frisch , a. ; hnich , b. ; kiziltan , z. ; miguel , i. ; and walsh , t. 2001b .
matrix modelling .
technical report apes-36 - 2001 , apes group . presented at formul01 workshop on modelling and problem formulation , cp2001 post - conference workshop .
flener , p. ; frisch , a. ; hnich , b. ; kiziltan , z. ; miguel , i. ; pearson , j. ; and walsh , t. 2002 .
breaking row and column symmetry in matrix models . in _
8th international conf . on principles and practices of constraint programming ( cp-2002)_. springer .
flener , p. ; pearson , j. ; sellmann , m. ; and hentenryck , p. v. 2006 .
static and dynamic structural symmetry breaking . in _ proc . of 12th international conf . on principles and practice of constraint programming ( cp2006)_. springer .
frisch , a. ; hnich , b. ; kiziltan , z. ; miguel , i. ; and walsh , t. 2002 . global constraints for lexicographic orderings . in _
8th international conf . on principles and practices of constraint programming ( cp-2002)_. springer .
frisch , a. ; jefferson , c. ; and miguel , i. 2003 .
constraints for breaking more row and column symmetries . in rossi , f. ,
of 9th international conf . on principles and practice of constraint programming ( cp2003)_. springer .
gent , i. , and walsh , t. 1999 . : a benchmark library for constraints . technical report , technical report apes-09 - 1999 .
a shorter version appears in _ proc . of the 5th international conf . on principles and practices of constraint programming ( cp-99)_. grayland , a. ; miguel , i. ; and roney - dougal , c. 2009 .
snake lex : an alternative to double lex . in gent , i.
, _ proc . of 15th international conf . on principles and practice of constraint programming _ , _ lncs _ 5732 , 391399
. springer .
huczynska , s. ; mckay , p. ; miguel , i. ; and nightingale , p. 2009 . modelling equidistant frequency permutation arrays : an application of constraints to mathematics . in gent , i. p. , ed .
, _ proc . of 15th international conf . on principles and practice of constraint programming ( cp 2009 )
_ , _ lncs _ 5732 , 5064 .
springer .
jefferson , c. , and petrie , k. 2011 .
automatic generation of constraints for partial symmetry breaking . in lee , j. h .-
, _ proc . of 17th international conf . on principles and practice of constraint programming ( cp 2011 )
_ , _ lncs _ 6876 , 729743 .
springer .
katsirelos , g. ; narodytska , n. ; and walsh , t. 2009 . combining symmetry breaking and global constraints . in oddi ,
a. ; fages , f. ; and rossi , f. , eds .
, _ recent advances in constraints , 13th annual ercim international workshop on constraint solving and constraint logic programming ( csclp 2008 ) _ , _ lncs _ 5655 , 8498 .
springer .
katsirelos , g. ; narodytska , n. ; and walsh , t. 2010 . on the complexity and completeness of static constraints for breaking row and column symmetry . in cohen , d. , ed . , _ proc . of the 16th international conf . on the principles and practice of constraint programming ( cp 2010 )
_ , _ lncs _ 6308 , 305320 .
springer .
law , y. , and lee , j. 2004 .
global constraints for integer and set value precedence . in _ proc . of 10th international conf . on principles and practice of constraint programming ( cp2004 ) _ , 362376 .
springer .
law , y .- c . ; lee , j. ; walsh , t. ; and yip , j. 2007 .
breaking symmetry of interchangeable variables and values . in _
13th international conf . on principles and practices of constraint programming ( cp-2007)_. springer .
mears , c. ; de la banda , m. ; wallace , m. ; and demoen , b. 2008 .
a novel approach for detecting symmetries in csp models . in perron , l. , and trick , m. a. , eds .
, _ proc . of 5th international conf . on integration of ai and or techniques in constraint programming for combinatorial optimization problems ( cpaior 2008 ) _ , volume
_ lncs _ 5015 , 158172 .
springer .
narodytska , n. , and walsh , t. 2012 .
beyond lex leader : breaking symmetry with other orderings .
under review for _ proc . of the 18th international conf . on the principles and practice of constraint programming ( cp 2012 )
springer .
pesant , g. 2004 . a regular language membership constraint for finite sequences of variables . in wallace , m. , ed .
, _ proc . of 10th international conf . on principles and practice of constraint programming ( cp2004 ) _ , 482295 .
springer .
puget , j .- f . 2005a .
automatic detection of variable and value symmetries . in van beek ,
, _ proc . of 11th international conf . on principles and practice of constraint programming ( cp 2005 )
_ , _ lncs _ 3709 , 475489 . springer .
puget , j .- f . 2005b .
breaking all value symmetries in surjection problems .
in van beek , p. , ed .
, _ proc . of 11th international conf . on principles and practice of constraint programming ( cp2005)_. springer .
walsh , t. 2011 . on the complexity of breaking symmetry . in _ proc . of the 11th international workshop on symmetry in constraint satisfaction problems ( symcon11)_. held alongside the 1th international conf . on the principles and practice of constraint programming ( cp 2011 ) . yip , j. , and hentenryck , p. v. 2011 .
symmetry breaking via lexleader feasibility checkers . in walsh , t. , ed .
, _ proc . of the 22nd international joint conf . on artificial intelligence
( ijcai-2011 ) _ , 687692 .
ijcai / aaai .
|
symmetry is an important problem in many combinatorial problems .
one way of dealing with symmetry is to add constraints that eliminate symmetric solutions .
we survey recent results in this area , focusing especially on two common and useful cases : symmetry breaking constraints for row and column symmetry , and symmetry breaking constraints for eliminating value symmetry .
|
ngc 4631 is one of the best examples of a nearby edge - on galaxy with a star formation driven outflow resulting in a bright and extended gaseous halo .
this is seen in many tracers : h@xmath3 ( rand , kulkarni , & hester 1992 ; hoopes , walterbos , & rand 1999 ) , radio continuum ( e.g. ekers & sancisi 1977 ; hummel & dettmar 1990 ; golla & hummel 1994 ) , and x - rays ( wang et al . 1995 ; vogler & pietsch 1996 ) .
this galaxy is undergoing interactions with two companions , ngc 4627 and ngc 4656 , and shows a complex pattern of tidal features seen in hi ( weliachew , sancisi , & gulin 1978 ; rand 1994 ) , and more recently to a lesser extent in cold dust ( neininger & dumke 1999 ) .
it also shows two large and highly energetic hi supershells ( rand & van der hulst 1993 ) , the origin of which is not clear ( rand & stone 1996 ; loeb & perna 1998 ) .
the disturbances to the disk resulting from the interactions may have caused the rather high level of star formation [ @xmath4 ergs s@xmath5 , uncorrected for internal extinction ( rand , kulkarni & hester 1992 ) ] .
the above - mentioned gaseous halos are particularly bright above the central 4 kpc of the disk , where almost all of the @xmath6co @xmath0 emission is found ( golla & wielebinski 1994 ; hereafter gw ) .
the co observations show a rigidly rotating inner disk with a total molecular mass of @xmath7 m@xmath8 . whether the inner 4-kpc structure might be a bar is a subject of debate ( e.g. roy , wang , & arsenault 1991 ; gw ) .
it is conceivable that the interactions have driven much of the dense ism into the central regions of the galaxy , providing the fuel for the strong star formation activity .
ngc 4631 provides us with an excellent opportunity to study star formation , galaxy interactions , possible radial inflows , and the production of gaseous halos , and thus deserves continued investigation .
therefore , to probe further the structure and kinematics of the central concentration of dense gas , we have observed it in the @xmath6co @xmath0 line with the berkeley - illinois - maryland array ( bima ) .
the data were taken in september 1995 in the c configuration of the bima array , with baselines ranging from 12 to 52 m. a single track was obtained , with an approximate on - source integration time of five hours .
the primary beam of the antennae is 100 at the 2.6 mm wavelength of the @xmath6co @xmath0 transition .
the quasar 3c273 was used for flux calibration .
while observing ngc 4631 , pointing was switched every 50 seconds between two positions with offsets of 25 east and 15 west of a position near the radio nucleus [ r.a .
12@xmath9 [email protected]@xmath11 , dec .
32@xmath12 32 29
( 2000.0 ) ; duric , seaquist , & crane 1982 ) , ensuring that the all of the entire bright emission from the central regions mapped by gw was contained within the primary beams .
calibration , mapping , and cleaning were carried out with the miriad package ( sault et al .
1995 ) , using the mosaic option in the program invert to create a mosaic dirty map of the two fields .
the emission was bright enough to allow self - calibration to be performed . for each pointing a clean map
was initially produced and used as a model in the miriad program selfcal. the self - calibrated @xmath13-data were then used to create a new clean map , which became the new model for the next iteration .
convergence was achieved after four iterations .
the clean beam has dimensions 9.8x6.7
82@xmath12 ) .
the noise in the final channel maps is 130 mjy ( beam)@xmath5 or 0.2 k. a total intensity map was formed from emission in the range @xmath14 km
s@xmath5 , using data above 2@xmath15 in each channel . the systemic velocity of ngc 4631 is @xmath16 km s@xmath5 ( rand 1994 ) , and we adopt the commonly used distance of 7.5 mpc ( hummel , sancisi , & ekers 1984 ) .
figure 1 shows the map of total co intensity , including the fwhm of the primary beam for the two pointings , while figure 2 shows the same map overlaid on h@xmath3 and red - continuum ccd images from rand , kulkarni , & hester ( 1992 ) .
some emission at the western end of the co distribution is beyond the fwhm of the western primary beam and may not be real or accurately mapped .
other isolated features above the plane which reach only the first contour level ( 2@xmath15 ) are not significant enough to be considered real .
position - velocity diagrams parallel to the major axis of the galaxy are shown in figure 3 .
the major axis position angle of 86@xmath17 is the large - scale value derived from hi data ( rand 1994 ) . the basic co structure and kinematics of the central region
have been analyzed in detail by gw ; hence , we only briefly discuss them here , and focus on the small - scale features revealed by the high resolution ( 3.2 ) .
the emission shows a continuous structure with maxima at either end .
the central minimum is at the position of the hi kinematic center ( rand 1994 ) .
the brightening towards the far ends of these features is consistent with previous single - dish observations ( gw ; sofue et al .
1990 ; sofue , handa , & nakai 1989 ) , and suggests a ring or bar - like structure of length @xmath1 4 kpc .
the spectra of gw show that the emission falls off sharply beyond the radius of this structure .
the `` bend '' noted by gw is even more noticeable here : the eastern half is elongated along a position angle @xmath18 while the position angle of the western half is @xmath19 .
in this way , the central region reflects the well - known overall bend of the galaxy .
this bend causes an exaggeration of the e - w asymmetry of the emission in the position - velocity ( @xmath20 ) diagrams : with the northern ( southern ) slices showing brighter emission from the eastern ( western ) end .
the midplane @xmath20 diagram indicates rigid rotation with a slope of 1.9 km s@xmath5 arcsec@xmath5 , in agreement with the slope measured from the co 21 map of gw .
velocities reach about 130 km s@xmath5 from the systemic velocity .
the bright concentration at the eastern end of the emission shows a two - peaked velocity profile in a few cuts in figure 3 .
the concentration at the western end does not show a clear bifurcation , but the velocity width is relatively large , similar to the eastern concentration .
these kinematics could be due to localized events ( see 3.2 ) , or they may indicate non - circular motions . in particular , at the ends of a bar , gas streaming nearly perpendicular to the bar on nearly circular orbits where the spiral pattern begins
may run into gas streaming along the bar ( e.g. regan , vogel , & teuben 1997 ) , creating a shock ( see also roberts , huntley , & van albada 1979 ) or at least orbit crowding .
the co velocity field at the transition between the bar and the spiral arms in m83 has been interpreted in this way ( kenney & lord 1991 ) .
this flow convergence may lead to bright regions of star formation , as observed at the ends of many bars , although more often in sbb galaxies ( phillips 1993 ) .
a high level of star formation at this co peak is indicated by bright h@xmath3 , x - ray and radio continuum emission , as discussed in 3.2 . against the bar hypothesis
is the fact that nothing else in the @xmath20 diagram strongly suggests a bar , such as the expected `` parallelogram '' appearance of the emission due to @xmath21 orbits parallel to the bar potential ( binney et al .
this conclusion was also reached by gw . only for a bar viewed side - on do
the @xmath21 orbits occupy a narrow locus in @xmath20 diagrams ( see figure 8 of garca - burillo & gulin 1995 ) . in support of such a perspective
is that ngc 4631 has one of the most slowly rising co rotation curves of the 14 galaxies studied by sofue et al .
( 1997 ) , as might be expected for a side - on view of a bar .
however , the low slope of the rotation curve could simply be due to the almost complete lack of a visible bulge .
furthermore , there is no indication of gas on @xmath22 orbits perpendicular to the bar , deep in the potential ( cf .
figure 1 of binney et al . 1991 ) .
these are particularly noticeable in side - on views
. a lack of molecular gas in the inner parts of the bar could be responsible .
alternatively , the bar may not have an ilr ( e.g. athanassoula 1992 ) .
for the modeled hi rotation curve of rand ( 1994 ) , this would imply a bar pattern speed of @xmath23 km s@xmath5 kpc@xmath5 .
hence , a bar can not be ruled out but there is little evidence to support anything more complex than a ring structure .
nevertheless , even a ring may represent a resonant response to a bar potential , but again there is little supporting evidence for a bar , particularly in the stellar emission ( see below ) .
there may still be lower - level non - circular motions , however , given that slightly non - axisymmetric bulges are common in spirals ( zaritsky & lo 1986 ) , and given the many other disturbances to this galaxy .
hi data in this central region show a rise in velocity very similar to the co data , but no evidence of a central minimum .
it is more difficult to discern a bar in the hi data because of confusion with the larger , differentially - rotating disk .
there is evidence of broadened velocity profiles in the central 4. the model hi cube of rand ( 1994 ) featured a velocity dispersion dictated by the rate of falloff of emission at the terminal velocity in the differentially rotating component .
this choice produced generally broader profiles than observed in the central 4 , most likely indicating non - circular motions , but whether they are due to an inner bar or to spiral structure can not be determined .
the disturbed optical appearance in the central regions , the lack of a discernable bulge , and the lack of near - ir imaging of this galaxy make it difficult to appeal to the stellar distribution to aid in discerning a bar .
such a structure might be inferred from a boxy or peanut - shaped bulge , as in , for example , ngc 4013 , where the co distribution clearly shows evidence for bar - driven orbits ( garca - burillo , combes , & neri 1999 ) . in ngc 4631
, the central regions in the red image ( figure 2b ) show a very irregular morphology , due to some combination of irregular stellar and dust distributions .
the central regions of ngc 4631 have been recently mapped in submillimeter continuum emission by three different groups ( braine et al .
1995 ; alton , davies , & bianchi 1999 ; neininger & dumke 1995 ) .
the co map in general resembles these maps rather well in that they all show a central structure of about 4 kpc extent with emission enhancements near the ends .
the prominent extraplanar feature seen in the 1.3 mm continuum map of braine et al .
( 1995 ) , extending about 1 southward of the eastern complex , is not seen in co. it is not likely to have been resolved out , since essentially all of its structure is on scales to which the co map is sensitive . using the conversion of 1.3 mm surface brightness to gas mass given by braine et al .
, the typical gas surface density in this filament should be of order 200 m@xmath8 pc@xmath24 .
if all of this gas were molecular and detectable in co , we should expect emission in our map at a typical level of 30 jy ( beam)@xmath5 km s@xmath5 , or 6@xmath15 .
this is clearly not seen .
possibilities are that the dust is associated with hi gas , the gas to dust ratio is much lower in ngc 4631 , any molecular gas in the spur is too cold to emit substantially in the co lines , or the emission is spurious . for the main central 2 feature at least , the agreement of the derived mass with the total hi+h@xmath25 mass is better than a factor of 2 ( braine et al . ) , suggesting that the essential conversion from 1.3 mm emission to total gas mass is reasonably accurate .
it is difficult to extend this conclusion to the high-@xmath26 gas , however .
more relevant is that the feature is not seen in the 1.2 mm map of neininger & dumke ( 1999 ) , which has very similar resolution and sensitivity
. it may well be spurious .
on the other hand , the beginnings of the loop south of the nucleus in braine et al.s map coincide with a co spur extending south from the western half of the emission ; these in turn may be associated with an extraplanar h@xmath3 feature ( figure 2 ) .
the 850 @xmath27 m continuum map of alton et al .
( 1999 ) also shows extraplanar emission just south of this co spur .
the 1.4 ghz map of golla ( 1999 ) at 1.45 resolution shows much agreement with the co distribution on scales of a few arcseconds , including bright concentrations coincident with the two co peaks at each end of the main structure , and a relative deficit of emission between them .
earlier radio maps also show a concentration of synchrotron emission in the central 4 kpc of the disk ( e.g. ekers & sancisi 1977 , duric et al .
1982 ; golla & hummel 1994 ) .
most of the diffuse extraplanar h@xmath3 emission ( figure 2 ; see also hoopes , walterbos , & rand 1999 ) is found above and below the 4-kpc extent of the co emission .
the quasi - vertical double - worm structure reported by rkh does not rise from the center of the co structure but from a position about 700 pc to the east .
diffuse , soft x - ray halo emission generally surrounds the inner disk and is brightest in the approximate radial range of the co emission , as seen in the overlay of the rosat pspc map from vogler & pietsch ( 1996 ) , kindly provided by a. vogler , and the co map in figure 4a ( see also wang et al .
finally , bright polarized emission at 4.88 and 8.46 ghz is found in the halo above this region of the disk , where the inferred intrinsic magnetic field runs largely perpendicular to the major axis ( golla & hummel 1994 ) .
all of these associations suggest active star formation and outflow of gas into the halo .
all three halo tracers are brighter on the north side of the co emission than the south side . assuming a conversion between co brightness and h@xmath25 column density of @xmath28 , the total molecular mass detected is @xmath29 m@xmath8 . within the uncertainties
, this value agrees with the estimate of gw for the central 140 ( using the same value of @xmath30 ) , suggesting that little , if any , emission has been resolved out by our observations .
the most important new result from this observation is the filamentary and shell - like emission found north of the eastern end of the main co feature .
a weaker feature is also seen to the south .
the extraplanar emission is detected up to 750 pc above the midplane of the co distribution .
although the galaxy is not quite edge - on , it is unlikely that the feature is in the plane of the galaxy . at @xmath31 , it would have to extend about 10 kpc along the midplane to explain the apparent extent parallel to the minor axis .
also , kinematically , in a highly inclined , differentially rotating disk , one would expect the feature to show a smooth gradient towards lower observed velocities with increasing distance from the major axis if it were in the disk .
although its kinematics are complex ( see below ) , it does not show such a signature .
the mass of the feature is estimated at roughly 10@xmath32 m@xmath8 , using the above value of @xmath30 .
it splits into a quasi - linear feature on the west side , rising to the north - west , and a more shell - like feature on the east side , suggesting that it may have been produced by two events .
figure 5 shows position - velocity diagrams parallel to the r.a .
axis in the region of this feature , covering its full vertical extent .
the disk emission below the feature is the brightest in the map and kinematically splits into two components , as discussed above , with the lower velocity component deviating somewhat from the general trend of velocity with position . just to the east of the bright disk feature , most of the emission centers around @xmath33 km s@xmath5 ,
while fainter emission is seen at @xmath34 km s@xmath5 ( more apparent in the @xmath35 panel ) .
velocities in the western , filamentary component of the extraplanar feature ( best seen in the 12 , 15 and 18 panels at r.a .
offsets around 12 ) center around @xmath36 km s@xmath5 , a typical velocity for its distance along the major axis , and no clear trend of velocity with position can be identified as a clue to its nature . however , in the eastern extraplanar feature ( around r.a .
offset 12 ) , the two disk components at @xmath33 km s@xmath5 and @xmath34 km s@xmath5 are seen up to the 15 panel ( also to the 6 panel ) .
this velocity splitting may indicate a structure expanding at @xmath37 km s@xmath5 .
given that the eastern feature accounts for slightly more than half the emission , the equivalent kinetic energy of a spherically expanding shell would be enormous : @xmath38 erg , comparable to the two hi supershells .
on the other hand , there are other possible sources of peculiar kinematics at this radius , as discussed above , and the line splitting may have an orbital origin , although the evidence is inconclusive .
another indication of the energy involved in creating such an outflow comes from an estimate of the potential energy of the gas mass at its height above the plane .
this has been estimated for several dust clouds above the disk of ngc 891 by howk & savage ( 1997 ) .
unfortunately , the estimate depends on the rather poorly known total midplane mass density and scale height .
the potential energy is given by : @xmath39\ ] ] where @xmath40 is the total mass scale height , @xmath41 is the total midplane mass density and @xmath42 is the mass of the high-@xmath26 feature . the @xmath43band scale - height has been determined by hummel & dettmar ( 1990 ) to be about 1 kpc .
this may be a poor estimate of the true mass scale height , which itself could vary significantly around this disturbed galaxy
. however , use of this value should still give a rough estimate of the energies involved .
the total mass density is unknown but presumably less than in ngc 891 or the milky way given the low rotation speed of 140 km s@xmath5 ( rand 1994 ) .
using the above molecular gas mass and an average height above the plane of 500 pc , the potential energy is then @xmath44 comparable to the kinetic energy estimate .
of course , this analysis does not account for energy radiated away or converted to thermal or turbulent motions .
the bright disk emission below this feature abuts the western edge of the very bright complex of hii regions cm65 , cm66 , and cm67 ( crillon & monnet 1969 ; roy et al .
1991 ) , suggesting a region of intense star formation .
in fact , the eastern , shell - like extraplanar feature almost appears to surround the peak of h@xmath3 ( and associated red continuum ) emission ( figure 2 ) .
the h@xmath3 luminosity from this complex , measured from the image of rand , kulkarni , & hester ( 1992 ) , is @xmath45 erg s@xmath5 ( uncorrected for extinction ) , about equal to that of 30 doradus in the lmc ( kennicutt & hodge 1986 ) .
roy et al .
( 1991 ) noted a complex velocity structure for these hii regions , with components at @xmath46 = 641 , 687 , and 753 km s@xmath5 .
thus , both co and h@xmath3 emission show a large range of velocities , and there is some correlation of these components between the two tracers .
of course , it must be recalled that the disk emission may have a significant extent along the line of sight ( for instance if it is part of a ring viewed along the tangent point ) , and this may explain , at least in part , why it is so bright in several ism tracers .
there are no well - defined _ extraplanar _
h@xmath3 features which are associated with the co emission .
in particular , the double - worm structure lies between this co structure and the center of the molecular ring .
it may originate from a star - forming complex elsewhere on the ring .
x - rays from this part of the disk provide further evidence for a high rate of star formation .
figure 4b shows the co emission with contours of x - ray emission from the `` adaptive - filtered '' rosat hri ( 0.12.4 kev ) image of vogler & pietsch ( 1996 ) .
the filtering technique smooths the original hri image with a kernel whose size depends on the image intensity , thus enhancing low - level diffuse emission while retaining high resolution for bright , compact features .
no x - ray feature can be specifically associated with the molecular outflow , but the bright concentration of disk co emission beneath it is coincident within the positional uncertainties with the brightest peak of x - ray emission in the inner disk .
this source has a luminosity , corrected for galactic absorption , of @xmath47 erg s@xmath5 ( 0.12.4 kev ) , and is too weak to obtain a spectrum .
this luminosity is a lower limit due to unaccounted for absorption in ngc 4631 , and is comparable to those of the x - ray sources in m101 postulated to be hypernova remnants by wang ( 1999 ) .
coincident with these disk co and x - ray sources is a bright source of radio emission in golla s ( 1999 ) 1.4 ghz map at 1.45 .
this source resolves into five features in a 4.86 ghz map at sub - arcsecond resolution , four of which have thermal spectra .
golla & hummel ( 1994 ) trace one of the radio continuum spurs back down to the cm67 region in the disk .
none of the extraplanar emission from cold dust found by neininger & dumke ( 1999 ) appears to be associated with the co feature .
examination of the hi data cube of ngc 4631 at 12x22 ( rand 1994 ) shows no obvious associated kinematic or morphological peculiarities .
hence , there is much evidence for a very intense region of star formation at in this disk location , and an extraplanar radio continuum filament also seems to be associated with the co feature .
it is difficult to conclude much about the origin of this feature at this point .
if the kinetic energy is of order 10@xmath2 ergs , then it may be difficult to explain it as a result of stellar winds and supernovae , as was the case for the two large hi supershells ( rand & van der hulst 1993 ) .
one would require of order 10,000 supernova progenitors .
alternatively , it may be due to the impact of a rather massive high - velocity cloud , an explanation considered by rand & stone ( 1996 ) for the hi supershells ( there , an impactor of about 10@xmath48 m@xmath8 was required ) .
the impact may have triggered the star formation in the disk at this location . in this case
, the x - ray emission would be due to the resulting supernovae , as the collision itself would not be able to heat gas to x - ray emitting temperatures .
an examination of the hi data of rand ( 1994 ) does not reveal any obvious signs of an impact at this location , such as a trail of gas .
finally , there is the newly recognized possibility that such shells can be produced by a single energetic explosion a hypernova , which may be the cause of gamma - ray bursts ( e.g. loeb & perna 1998 ; efremov , elmegreen , & hodge 1998 ; wang 1999 ) . at best ,
it is difficult to distinguish between putative hypernovae and conventional supernova - driven supershells , and in this case especially so , given the difficulty of disentangling emission along the line of sight , correcting x - ray fluxes for absorption in the intervening gas , and determining the age of the high-@xmath26 structure .
however , it is worth noting that this is now the third problematically energetic event found in this galaxy
. this may provide constraints on the hypernova hypothesis .
whether these events occur because of hypernovae , or multiple supernovae and stellar winds from massive star formation concentrated into unusually large associations , or a few relatively massive high - velocity clouds penetrating the disk , can only be answered by future investigation .
the co emission from the feature reported here is bright enough to allow a higher resolution map to be made .
such a map may provide a clearer delineation of its morphology and kinematics , allowing a more careful comparison with models of its origin .
also chandra , with its higher resolution compared to rosat , may constrain the nature of the x - ray emission by showing whether it remains diffuse on smaller scales or breaks up into multiple components , and also by allowing a measurement of the gas temperature .
in this brief paper we have presented a high - resolution map of the centrally concentrated co emission in ngc 4631 .
the map confirms the ring or bar - like distribution of molecular gas inferred from single - dish data , but there is little direct evidence for a bar . halo emission in several gaseous tracers
is concentrated above , and to a lesser extent , below , this molecular structure , suggesting that the central concentration of molecular gas has provided the fuel for intense star formation which has resulted in a substantial outflow of gas . of great interest
is an extraplanar co feature found above the eastern end of the molecular structure .
it suggests an outflow of about @xmath49 m@xmath8 of molecular gas , although there is only weak evidence for an associated kinematic signature . nevertheless , the energies involved in driving the outflow may be well over 10@xmath2 ergs .
bright hii regions and x - ray emission suggest that the underlying disk location is very active in star formation .
an extraplanar radio continuum filament also appears to have a footprint in this region . like the two highly energetic hi supershells in this galaxy
, this outflow may be driven by multiple supernovae or is perhaps the result of an impact of a high - velocity cloud .
its characteristics also put it in the class of events postulated to be due to hypernova explosions .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
|
we present interferometric observations of co @xmath0 emission in the central regions of the edge - on galaxy ngc 4631 , known for its extended gaseous halo and its tidal interactions .
previous single - dish observations revealed that almost all of the co emission arises from a central ring or bar - like structure of length @xmath1 4 kpc .
we confirm this structure at higher resolution , and find that it is bent at the center , reflecting the overall bend in this galaxy apparent from optical images .
the kinematic evidence favors a rigidly rotating ring over a bar .
the gaseous halo emission in several tracers is concentrated above and below this molecular structure . to the north of an emission peak at the eastern end of
the structure is an extraplanar feature showing filamentary and shell - like properties which we interpret as an energetic molecular outflow .
the energies involved are difficult to estimate , but are probably of order 10@xmath2 ergs or more .
the co concentration in the disk below this structure coincides with a bright hii region complex , a peak of radio emission , and the brightest x - ray feature in the inner disk of the galaxy seen in a rosat hri map , all suggesting intense star formation .
a filament of radio continuum emission may also have a footprint in this region of the disk .
the origin of the outflow is unclear .
|
the cluster galaxy population has evolved rapidly over the last 4gyr ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
clusters at are dominated , particularly at faint magnitudes , by blue spiral galaxies , predominantly irregular or sc sd spirals .
some of these show signs of disturbed morphologies , and many present spectroscopic evidence that they have undergone multiple star - formation events over the last @xcite .
conversely , local clusters are completely dominated by passive early - type galaxies : elliptical and lenticular ( s0 ) galaxies at the brighter end , and dwarf spheroids at fainter magnitudes
. in the standard hierarchical cosmological model , represents the peak infall rate of field galaxies onto the cluster @xcite , and it is the transformation of these infalling field galaxies from star - forming disk - dominated galaxies into passively - evolving spheroids over a period of through their encounter with the cluster environment , that produces the observed changes in cluster galaxy populations .
several physical mechanisms related to the cluster environment have been proposed as producing the observed transformations in galaxies , in which interactions with either other cluster galaxies or the hot intra - cluster medium ( icm ) affect both their structural and star - formation properties . to distinguish between these mechanisms it is necessary to examine both where the transformations occur , and how the star - formation and structural properties of the galaxies are changed ( e.g. * ? ? ?
for example , ram pressure from the passage of the galaxy through the dense icm can effectively remove the cold gas supply and thus rapidly terminate new star - formation , either by stripping the gas directly @xcite , or by inducing a starburst in which all of the gas is consumed @xcite .
the most dramatic icm - galaxy interactions should occur when two clusters merge , as shock fronts created in the icm may trigger starbursts in galaxies over large scales @xcite .
importantly , in terms of their environmental effects , these mechanisms all require a dense icm , and so their evolutionary effects are limited to the cores of clusters .
in contrast , galaxy mergers , which can strongly affect the morphological evolution of disks , can not occur when the encounter velocities are much greater than the internal velocity dispersion of galaxies @xcite , and so while frequent in small groups , are rare in rich clusters @xcite .
alternatively galaxy harassment , whereby repeated close , high - velocity encounters with massive galaxies and the cluster s tidal field cause impulsive gravitational shocks that damage the fragile disk of late - type spirals @xcite , transforming them over a period of several gyr .
galaxy harassment is effective throughout a cluster , including beyond the virial radius , but its effects should be greater for those clusters with higher velocity dispersions . finally , when a galaxy falls into a more massive halo , the diffuse gas in its halo is lost to the icm , thus preventing further cooling and replenishment of the cold gas supply , `` suffocating '' the galaxy ( e.g. * ?
* ; * ? ? ?
star - formation in the galaxy then declines slowly as the remaining cold gas is used up @xcite .
the large datasets provided by the 2dfgrs and sdss have allowed the environmental effects on galaxy properties to be followed statistically over the full range of environments , from the sparse field to the dense cluster cores ( e.g. * ? ? ?
* ; * ? ? ?
* ) , at least for the brightest galaxies ( ) .
they show that star - formation is most closely dependent on local density , and is systematically suppressed above a critical value of density , that is found 34 virial radii from clusters , but also in galaxy groups as poor as .
this suppression is observed to be independent of the richness of the structure to which the galaxy is bound @xcite , indicating that mechanisms such as galaxy harassment or ram - pressure stripping are not important for the evolution of bright galaxies .
instead the strongest candidates for driving their transformation are galaxy suffocation and low - velocity encounters , which are effective in both galaxy groups and cluster infall regions .
however , it is not clear if and how this scenario extends to fainter magnitudes , as there has been observed a strong bimodality in the properties of galaxies about a characteristic stellar mass ( corresponding to ) , with more massive galaxies predominately passive red spheroids , and less massive galaxies tending to be blue star - forming disks @xcite
. this bimodality implies fundamental differences in the formation and evolution of giant and dwarf galaxies , and it has been proposed ( e.g. * ? ? ?
* ; * ? ? ?
* ) that these are driven by thermal processes in the gas inflowing from the halo onto the galaxy , with the characteristic mass scale representing the point at which shocks in the halo become stable , heating up the halo gas , and preventing further cooling
. hence if the formation and evolution of giant and dwarf galaxies are so fundamentally different , then they are likely to be affected differently by mechanisms related to the environment .
for example , galaxy harassment should be most efficient at transforming low - luminosity late - type galaxies .
@xcite find differences in the environmental dependences of faint and bright galaxy populations , and suggest that faint galaxies are affected by mechanisms related to the structure in which the galaxy is found .
to understand the mechanisms underlying the transformation of faint galaxies requires datasets reaching much fainter luminosities .
crucial discriminators between the different transformation mechanisms are the time- and distance - scales involved : while ram - pressure stripping should rapidly terminate star - formation in a galaxy within but requires the dense icm of the cluster core ; a galaxy undergoing suffocation will have its star - formation rate slowly decline over a period of several gyr .
hence the nature of the transition from regions where the majority of galaxies are star - forming , and those dominated by passive galaxies , will depend strongly on the dominant mechanism involved .
one approach is to use galaxy colours , which can be readily obtained to much fainter magnitudes than spectroscopic star - formation rates , and which through the use of models can be directly related to star - formation histories with minimal assumptions ( e.g. * ? ? ?
recent large datasets have shown that the bimodality of galaxies is also manifested through their broadband photometry , in particular a separation can be made on the basis of colour into red and blue galaxy populations @xcite , which correspond roughly to the two broad types previously known from their morphological and spectroscopical characteristics : passively - evolving early - type and star - forming late - type galaxies .
this bimodality has been further quantified , resulting in colour - magnitude ( c - m ) relations and for both the red and blue galaxy populations @xcite , and its evolution observed to @xcite .
@xcite show that the bimodal galaxy colour distribution is strongly dependent on environment , with the fraction of galaxies in the red distribution at a fixed luminosity increasing from 1030% in the lowest density environments , to at the highest densities .
the most dramatic effects of environment on galaxy evolution should occur in superclusters , where the infall and encounter velocities of galaxies are greatest , groups and clusters are still merging , and significant numbers of galaxies will be encountering the dense icm of the cluster environment for the first time . with this in mind
we are undertaking the shapley optical survey ( sos ) , an optical photometric study of the core region of the shapley supercluster @xcite , one the most massive structure in the local universe , containing as many as 25 abell clusters . in this paper
we examine the effect of the supercluster environment on the star - formation histories of galaxies as measured through their galaxy colours .
we present the sos in section [ data ] , and then describe how we quantify the local environment and statistically subtract the field galaxy population in sections [ den ] and [ sub ] .
we present our analysis of the c - m relation in section [ cmrelation ] , which then allows us to separate the red and blue galaxy populations whose spatial distributions are presented in section [ redandblue ] .
we examine the environmental dependencies on galaxy colours in section [ colours ] , and discuss our findings in section [ discussion ] , before presenting our conclusions in section [ conclusions ] . throughout the paper
we adopt a cosmology with , and . according to this cosmology 1
arcmin corresponds to 60kpc at .
the sos comprises wide - field optical imaging of a region covering the whole of the shapley supercluster core ( hereafter ssc ) , which comprises three abell clusters a3556 , a3558 and a3562 and two poor clusters sc1327 - 312 and sc1329 - 314 .
the resultant galaxy catalogues are complete to and , allowing us for the first time to study the effect of the supercluster environment on the photometric properties of galaxies , particularly the dwarf galaxy population where we reach .
the area is covered by extensive redshift surveys ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) which indicate that these clusters form a complex clumpy and highly - elongated structure across , which is in the final stages of collapse , with infall velocities reaching @xcite .
there exists a wealth of multi - wavelength data from radio ( e.g. * ? ? ?
* ) to x - ray ( e.g. * ? ? ?
* ) , which in conjunction with the redshift data , describe a dramatic scenario with ongoing cluster - cluster mergers triggering star - formation in galaxies over @xmath3mpc scales .
it is clear that amid this maelstrom of galaxies , groups and clusters , both galaxy harassment and ram pressure stripping should be at their most effective in transforming the infalling field galaxies .
we aim to use the sos dataset to study the effect of environment on the luminosity distribution , colours and structural properties of galaxies , and through comparison with numerical simulations and theoretical predictions , draw insights regarding which physical processes are most important for the transformation and evolution of galaxies in these environments . in ( * ? ?
* hereafter paper i ) we introduce the survey , and describe the observations , calibrations , and derivation of the galaxy catalogues .
the galaxy luminosity functions are also presented , and significant environmental effects are observed , in the form of a dip at which becomes deeper , and a faint - end slope which becomes steeper , with decreasing density .
we explain these results in terms of the galaxy harassment scenario , in which the late - type spirals that represent the dominant population at are transformed by galaxy harassment into passively - evolving dwarf spheroids , and in the process become magnitudes fainter due to mass loss and an ageing stellar population without new star - formation . in this paper
we examine the effect of the supercluster environment on the star - formation histories of galaxies as measured through their galaxy colours , while in future papers we expect to consider the environmental impact on galaxy structural parameters and compare our observational results with semi - analytical models of galaxy evolution in n - body simulations of a supercluster region .
the data were obtained from the eso science archive ( 68.a-0084 , p.i .
slezak ) , comprising wide - field @xmath4- and @xmath5-band imaging covering a region towards the clusters a3562 , a3558 and a3556 which form the core of the shapley supercluster at .
full details of the observations , data reduction , and the production of the galaxy catalogues are described in paper i , and only a summary is presented below .
the observations were made from march 2002 to april 2003 using the wfi camera , an instrument made up of eight ccds giving a field of view of , and which is set at the cassegrain focus of the 2.2-m mpg / eso telescope at la silla .
the survey is made up of eight contiguous fields , each with total exposure times of 1500s in @xmath4 and 1200s in @xmath5 , and typical fwhms of 0.71.0@xmath6 .
the data are reduced using the alambic pipeline ( version 1.0 , * ? ? ?
* ) , and calibrated to the johnson - kron - cousins photometric system using observations of @xcite standard stars .
the sources are then extracted and classified using sextractor @xcite , resulting in galaxy catalogues which are both complete and reliable ( i.e. free of stars ) to and . for this analysis
we consider only galaxies to where uncertainties in the photometry are less than in both @xmath5 and @xmath2 for galaxies belonging to the supercluster .
to study the effect of the cluster environment on galaxies in the ssc , the local density of galaxies , @xmath7 , is determined across the sos mosaic .
this is achieved using an adaptive kernel estimator @xcite , in which each galaxy @xmath8 is represented by a gaussian kernel , , whose width , @xmath9 is proportional to thus matching the resolution locally to the density of information available . for this study , we consider the surface number density of galaxies , with an additional colour cut applied to reject those galaxies more than 0.2mag redder in than the observed cluster red sequence to minimize background contamination .
as there are no known structures in the foreground of the ssc ( 90% of galaxies have redshifts confirming that they belong to the supercluster ) , any substructure identified in the density map is likely to be real and belonging to the supercluster .
the local density is initially determined using a fixed gaussian kernel of width 2 arcmin , and then iteratively recalculated using adaptive kernels .
the resultant surface density map of the ssc is shown in fig [ densitymap ] , with the three clusters and two groups indicated .
isodensity contours are shown at intervals of , with the thick contours corresponding to 0.5 , 1.0 and , the densities used to separate the three cluster environments described below . the expected density of field galaxies is estimated through analysis of thirteen fields of deep @xmath10 imaging taken from the deep lens survey ( dls ; * ? ? ?
* ) . these data covering in total ,
were taken using the mosaic - ii cameras on the noao kpno and ctio 4-m telescopes , and have 5@xmath11 depths of and typical @xmath5-band fwhms of 0.9@xmath6 , allowing accurate photometric measurements and star - galaxy classifications to be made to at least the depths of our survey . through applying the same colour - magnitude cuts
, we estimate the density of field galaxies to be , and hence the thick contours correspond to overdensity levels of , 200 and respectively .
the entire region covered by the sos can be seen to be overdense with respect to field galaxy counts . for the following analyses on the effect of the cluster environment on its constituent galaxy population we define three regions selected according the local surface number density .
firstly we consider a high - density region with galsarcmin@xmath12 which correspond to the cores of the clusters .
next we consider intermediate- ( ) and low - density ( ) regions which probe the filament connecting the clusters a3562 and a3558 , as well as the wider envelope containing the whole supercluster core complex .
to accurately measure the global photometric properties of the galaxy population in the ssc requires the foreground / background contamination to be estimated efficiently and corrected for .
there exists already a wealth of spectroscopic data in the region , comprising 607 published galaxy redshifts @xcite corresponding to 90% of galaxies , dropping to 50% for . for those galaxies without redshifts , particularly at fainter magnitudes , we estimate the probability that they are supercluster members .
this probability is dependent on the spatial position , the @xmath5-band magnitude and @xmath2 colour .
the dependence on its spatial position , through its local number density , is clear as galaxies towards one of the density peaks will be more likely to be members than those in regions where the surface density is closer to that expected for the field .
the dependence on the galaxies colour and magnitude is complex , with galaxies located on the cluster red sequence most likely to be members , while those galaxies much redder than the sequence most likely to belong to the background population .
it is important that all three parameters are considered simultaneously , as the relationship between the broad - band properties of galaxies and their environment has been shown to be complex @xcite
. we account for this by considering separately the three cluster environments described previously when estimating the probability that galaxies are supercluster members through their @xmath5-band magnitude and @xmath2 colour using the methods described in @xcite . for each of these three cluster environments and the deep lens survey field comparison ,
two - dimensional histograms are built with bins of width 0.4mag in @xmath5 and 0.2mag in @xmath2 to properly map the galaxy c - m distribution .
the histogram of field galaxies is normalised to match the area within each cluster region , and for each galaxy , the probability that galaxy @xmath8 belongs to the supercluster is then defined as : @xmath13 where and are the number densities of galaxies in the supercluster and field environments belonging to the corresponding bin in @xmath5 and @xmath2 . in low - density environments it is possible that for a particular bin , and so in these cases the excess number of field galaxies are redistributed to neighbouring bins with equal weight until for all bins . for those galaxies with redshifts is set to 1 for ( the redshift range of known supercluster members in the sos field ) or 0 otherwise .
hence for each galaxy in the region covered by the sos , we estimate the likelihood of its belonging to the shapley supercluster based upon its @xmath5-band magnitude , @xmath2 colour and local density .
to estimate the location of the colour - magnitude relation for galaxies in the supercluster , we consider those galaxies with local densities , i.e. both intermediate- and high - density environments , in order to avoid _ a priori _ field contamination effects when measuring the c - m slope .
figure [ cm ] shows the resulting c - m diagram for a monte carlo realisation of the galaxy population in these regions , in which those galaxies predicted to belong to the supercluster indicated by the larger symbols , and those belonging to the field indicated by the small points .
a clear red sequence is apparent for .
however at it becomes notably flatter , while at it becomes increasingly difficult to separate it from the remainder of the cluster population .
this flattening at the brightest magnitudes has been noted previously for this region @xcite , and has been observed for a large sample of galaxies from the sloan digital sky survey @xcite , indicating this is a universal phenomenon .
the slope in the c - m relation has been considered due to a progressive increase in metallicity with mass due to the greater ability of more massive galaxies to retain metals dispersed by supernova during their formation , and so the observed flattening of the c - m slope probably represents the same flattening of the mass - metallicity seen for galaxies with @xcite , which in turn may reflect the observation that the brightest galaxies are built up through `` dry mergers '' of galaxies ( e.g. * ? ? ?
* ) , and hence would have the same metallicities and colours as their progenitors .
we perform a fit to the c - m relation over the range , where the red sequence relation appears linear and can be separated from the rest of the cluster population .
although the red sequence appears visually quite distinct , obtaining a robust fit to it is not trivial , particularly given the presence of outliers which are heavily skewed bluewards of the relation , and so simple least - squares methods are unsuitable .
we apply the same method as @xcite , which produced the most robust results by iteratively applying the biweight algorithm @xcite with 3@xmath11 clipping to counteract the heavily skewed distribution .
a range of slope values were searched to determine the best - fitting value that minimizes the biweight scale of the deviations from the median value .
the uncertainties in the slope and intercept are derived using the non - parametric bootstrap method .
the number of cluster galaxies , @xmath14 , is estimated as for the @xmath15 galaxies with , and each bootstrap performed by sampling @xmath14 galaxies with replacement according to the probabilities @xmath16 calculated using eq . [ psc ] .
@xcite have analytically estimated that bootstrap resamplings give a good approximation to the underlying density population .
the resultant best - fitting c - m relation is found to be : @xmath17 where the quoted uncertainty in the intercept assumes a fixed value of slope .
the obtained dispersion about the relation as given by the biweight scale of the deviations is mag . as a robust test of the results obtained using the biweight algorithm
, we plot a histogram of the distribution of , the @xmath2 galaxy colour offset from the c - m relation , and model the distribution as a sum of two gaussian functions , one to represent the red sequence , and another to represent the `` blue cloud '' population .
fig [ cmdist ] shows that a bimodal gaussian distribution describes well the distribution of , but with a mean offset of with respect to the biweight median , and a dispersion of that is significantly smaller than predicted by the biweight scale estimator , and comparable to the dispersion observed for the coma cluster in the same passbands @xcite .
hence despite iteratively clipping galaxies from the median value of the relation , both the intercept and dispersion of the relation obtained using the biweight algorithm have been affected by outliers and the heavily skewed distribution .
we do not expect that photometric uncertainties are a major contributor to the observed dispersion as : firstly the mean @xmath2 uncertainty for galaxies in the red sequence is found to be 0.012mag , only reaching 0.025mag by ; and secondly the observed dispersion does not vary significantly as a function of magnitude over .
it is remarkable that such a small dispersion is obtained , given the large number of galaxies involved over two orders of magnitude in luminosity , and spread across three rich clusters and .
the wide spread of redshifts within the supercluster complex could introduce a significant component to the observed dispersion through the spread of k - corrections required for the red sequence galaxies , as the predicted change in the @xmath2 k - correction of an elliptical galaxy from @xmath18 to @xmath19 is @xcite , as is the observed increase in @xmath2 colour of red sequence galaxies @xcite . to measure the possible effect of k - corrections on the obtained colour dispersion of red sequence galaxies , we consider those red sequence galaxies with known redshifts .
although the redshift data is less than 50% complete beyond , the @xmath2 colour dispersion of the red sequence galaxies with redshifts is consistent with that for the whole sample . after applying suitable k - corrections to each of the galaxies with known redshifts , and the colour dispersion about the red sequence
is recalculated , found to be still consistent with that before the k - correction has been applied .
hence the observed internal dispersion of galaxy colours around the red sequence is not primarily the effect of different k - corrections over the redshift range of the supercluster .
this can be explained by considering that the redshift distribution has a standard deviation of 0.0036 , corresponding to an rms @xmath11 of @xmath20mag in @xmath2 , significantly smaller than the observed @xmath21 .
.gaussian fits to the @xmath2 colour distribution around the c - m relation for galaxies in the different density environments [ cols="^,^,^,^ " , ] to examine the effect of environment on the colour of the c - m relation we fit bimodal gaussian distributions to the distribution in the three environments as parametrised by the local galaxy density .
the slope and zero - point are maintained fixed at the values indicated in eq .
[ cmequation ] , in order to measure the shifts in the overall @xmath2 colour of the red sequence with environment .
we do not make any attempt to determine the slope and intercept in the differing environments using the biweight algorithm , as the effect of the heavily skewed distribution on the determination of these values will be dependent on environment , due to the increasing blue galaxy fraction from high- to low - density regions .
hence the resulting fits would be biased by the changing blue galaxy population , and mask any inherent changes in the red sequence population with environment .
the results of the gaussian fits to the red sequence population are shown in table [ gauss ] . in each case the galaxy colour distribution is well described by a double gaussian function .
the zero - points of the red sequence are consistent for the low- and intermediate - density regions , but the red sequence is found to be redder in the high - density region , a result significant at the level .
the redward shift of the red sequence continues if the density threshold is increased further , reaching for .
the @xmath2 colour dispersion of the red sequence remains constant at , for each of the environments , suggesting that the observed reddening indicates an overall _ shift _ in the red galaxy population , rather than a _ broadening _ of the distribution due to increased numbers of bluer galaxies .
to examine where star - formation has been triggered or inhibited by the supercluster environment , we split the galaxy population into two according to colour .
we identify red galaxies as those within @xmath22 ( after including photometric uncertainties ) of the cluster red sequence , and blue galaxies as those bluer than the red sequence .
hence these blue galaxies are all those _ not _ belonging to the red sequence , due to being either currently having some star - formation , or having had some in the previous .
figure [ bluegals ] shows the spatial distribution of both red and blue galaxies over the supercluster complex , after correcting for the expected background contribution of 0.147 and for the red and blue galaxy subsets respectively , as estimated from the 13 deep lens survey fields .
the blue isodensity contours for the blue galaxies are shown at intervals of , which are filled with increasingly intense blue colours with density .
overlaid are the red isodensity contours at intervals of to represent the local density of the red galaxy population .
note that all the contours indicate overdensities in comparison to the field , and that the spacing of the contours for the red galaxy population is twice that of the those for their blue counterparts .
the centres of the x - ray emission from each of the clusters in the region are indicated .
the spatial distribution of the red and blue galaxy populations differ significantly . whereas the blue galaxies appear reasonably evenly spread across the whole field , the red galaxies are highly concentrated towards the cluster centres and the filamentary structure connecting a3562 and a3558 .
in contrast , none of the three abell or two poor clusters show particular overdensities in the blue galaxy distribution towards their cores , in fact they appear underdense in comparison to their immediate surroundings , as noted by @xcite in the case of a3558
. however , several localised overdensities of blue galaxies are apparent along the filamentary structure connecting a3562 and a3558 , indicative of regions of enhanced star - formation within the supercluster core complex .
what is particularly interesting is their distribution within the filamentary structure with respect to the clusters .
the most overdense regions are located _ between _ each of the clusters a3562 , a3558 and the two poor clusters .
the two most notable overdensities of blue galaxies appear to be the western flank of the poor cluster sc1329 - 313 , and a linear structure bisecting the clusters a3558 and sc1327 - 312 .
both these overdensities contain numerous blue galaxies which are spectroscopically confirmed as belonging to the supercluster .
the region where these overdensities of blue galaxies are located is where the two clusters a3562 and a3558 are both experiencing merging events , which may suggest that this merging has recently triggered star - formation in these blue galaxies , possibly through shock fronts produced when the icms of merging clusters collide .
alternatively these overdensities could reflect infall regions along the filament connecting the clusters a3558 and a3562 .
further insights into the effect of the supercluster on star - formation in its member galaxies are possible through the complementary information provided by the _ fraction _ of supercluster galaxies that are classed as blue .
figure [ bluegalfrac ] shows the fraction of galaxies bluer than the cluster red sequence as a function of spatial position , as obtained by dividing the local surface density of blue galaxies by that for all galaxies after correcting both for background contamination .
the blue galaxy fraction is represented by the coloured contours , with red indicating low blue galaxy fractions and blue indicating high blue galaxy fractions . for comparison
the isodensity contours of the overall galaxy density of fig .
[ densitymap ] are overlaid .
whereas the local blue galaxy density shows little correlation with that of the overall local galaxy density ( see fig .
[ bluegals ] ) , the blue galaxy _ fraction _ appears strongly anti - correlated with the overall local galaxy density .
each of the three abell clusters , and the two poor clusters , are marked by regions of low blue galaxy fractions , while the filamentary structure connecting the clusters a3562 and a3558 has typically blue galaxies , which rises still further for the regions furthest from the clusters .
it is however this avoidance of the cluster cores by the blue galaxies that is the most notable feature of fig .
[ bluegalfrac ] , with of galaxies in these regions located within the narrow locus of the cluster red sequence to or .
it is clear that in the central regions of all the clusters in this region star - formation has been severely truncated in _
almost all if not all _ galaxies .
a crucial insight into what physical process is resulting in this termination of star - formation can be obtained through comparison with complementary xmm x - ray imaging of the region around a3562 and sc1329 - 313 as reported in @xcite .
a comparison of their fig.1 shows that while the centre of the x - ray emission from a3562 is coincident with that obtained from our galaxy surface density , there is an extension to the north - east , where we also find the blue galaxy fraction to be lowest as shown in fig .
[ bluegalfrac ] .
figure [ xmm ] shows how the blue galaxy fraction varies with spatial position in the vicinity of sc1329 - 313 , and whereas in fig .
[ bluegalfrac ] the overlaid black isodensity contours show the local galaxy density distribution , here they represent the x - ray emission from the icm of sc1329 - 313 , taken from the xmm science archive ( oid 10526 , p.i .
aschenbach ) .
the centre of x - ray emission ( as indicated by the arrow in fig .
[ densitymap ] ) does not coincide well with either of the two peaks in the local galaxy density distribution , but appears to lie in between them .
this displacement between the icm and the galaxy distribution may be a dynamical effect from the merger , the result of the collisional gas lagging behind the collisionless dm and galaxies . a comparison of figs .
[ bluegals ] and [ xmm ] suggests that while the region where the blue galaxy fraction is lowest does not coincide with either of the two peaks in the local galaxy distribution , it covers approximately the same region as the observed x - ray emission from the cluster .
for both the clusters a3562 and sc1329 - 313 , regions where the blue galaxy fractions are lowest appear more strongly coincident with the centres and extensions of the x - ray emission , than the corresponding centres and extensions of the clusters as measured by the overall galaxy number density .
the regions of low blue galaxy fractions appear coincident with the peaks in the galaxy surface density for the remaining clusters a3558 , a3556 and sc1327 - 312 , and so it is not possible to separate the effects of galaxy density with those of the icm for these regions .
as galaxy colours can be directly related to star - formation histories , by examining the environmental dependences of galaxy colours it is possible to gain insight into where and how star - formation is affected by the supercluster environment .
figure [ blue_den ] shows the overall blue galaxy fraction as a function of local density for four different @xmath5-band magnitude ranges , quantifying the global anti - correlation with local density .
the highest - density regions are dominated by red sequence galaxies over the entire magnitude range studied .
there are two clear trends : the blue galaxy fraction decreases with increasing density ; and in all environments the blue galaxy fraction increases monotonically with magnitude .
the strength of this second trend appears greatest at : whereas at bright magnitudes or ) the fraction of blue galaxies only increases marginally with decreasing local density , reaching only at the lowest densities studied ; at fainter magnitudes the blue galaxy fraction increases rapidly to in the lowest density bin .
the most luminous members of the supercluster are predominantly red sequence galaxies , and so would be expected to be strongly clustered in the high - density regions . to examine whether luminosity segregation contributes to the observed trends in galaxy colour with density , we plot in fig .
[ blanton ] the average environment as a function of both galaxy colour and magnitude .
the grey - scale and contours indicate mean local density as a function of both the @xmath2 colour offset and @xmath5-band magnitude , smoothed on scales of in @xmath2 and in @xmath5 .
the global trends are very similar to those observed for galaxies in the sdss @xcite , with the predominantly horizontal contours indicating that for all but the brightest galaxies , where environment is strongly luminosity - dependent , local density is a strong function of galaxy colour , while remaining almost independent of luminosity .
red galaxies are typically found in high - density regions , while blue galaxies are on average located in low - density regions , the two populations separated at a density , or .
this separation reflects the trend shown in fig .
[ blue_den ] for the blue galaxy fraction to increase with decreasing local density . in a comparable study of the a901/902 supercluster at , @xcite
use the 17-band combo-17 survey data to precisely isolate supercluster members within a thin photometric redshift slice .
they find a strong segregation of faint quiescent and star - forming galaxies around a local surface density of to a limit of , in excellent agreement with our value . while @xcite find a sharp transition in the galaxy properties at this density , our observed trends in mean galaxy colour and blue galaxy fractions with environment appear more gradual .
this we put down to the fact that whereas @xcite use the 17-band photometry to accurately isolate the supercluster galaxies , we perform a statistical background subtraction based on just the @xmath5-band magnitude , @xmath2 colour and local density , and hence any trends will be blurred out somewhat by interlopers .
we find that the brightest galaxies are the most clustered , which could appear contrary to the results of @xcite who found that galaxies within 1.1mpc of a3558 were significantly less clustered than those with both in projection and redshift space .
however their limit of corresponds to , which is more than a magnitude fainter than the point at which we find luminosity segregation to become important .
@xcite also reported that , as well the environments becoming denser with increasing luminosity for red galaxies at the brightest magnitudes , the environments of low - luminosity red galaxies become denser with increasing absolute magnitude down to , the limit of their survey .
a similar phenomenon is apparent in our data , and although our survey extends two magnitudes fainter , we find that the trend for low - luminosity red galaxies to be found in dense environments is limited to those with
. however it should be remembered that the uncertainties from the statistical background subtraction we have applied to measure the global supercluster galaxy populations become increasingly large at faint magnitudes .
we have examined the effect of the shapley supercluster environment , as measured in terms of the local surface density of galaxies , on the star - formation histories of galaxies through their colours . to examine the effect of environment on the _ global _ properties of red sequence galaxies , we measure the _ overall _ distribution of @xmath2 colours with respect to the observed c - m relation using a double gaussian fit to describe the global bimodal galaxy distribution .
the use of this approach to measure environmental trends is vital , as the usual method of obtaining the best - fitting c - m relation and measuring the effect of environment on these fits are biased by the strongly skewed distribution and changing contribution of the blue galaxy population .
we show that the _ location _ of the c - m relation is dependent on environment , with the relation redder in @xmath2 in the high - density regions corresponding to the cluster core ( ) , than the low- and intermediate - density regions which represent the cluster periphery ( ) . this is a similar result to that obtained for the rich cluster a209 at @xmath23 where the red sequence was observed to be redder in @xmath2 in the cluster core than the periphery @xcite . by considering a model red sequence galaxy as a 10gyr old ( at ) , , solar - metallicity stellar population @xcite , the observed reddening of the c - m relation in the high - density region of a209 was interpreted as these galaxies being on average 500myr older or 20% more metal - rich than their counterparts in lower density environments .
the observed reddening in the high - density regions of the shapley supercluster is also consistent with the red sequence galaxies being older in the high - density regions than those in lower density environments , the resultant reddening of the @xmath2 colour being 33% lower than for a209 as the galaxies should be on average 2gyr younger in a209 than in the shapley supercluster .
these results are consistent with analysis of the spectra of luminous , red , bulge - dominated galaxies from the sdss , which were found to be marginally older and more metal - rich in high - density regions than their low - density counterparts @xcite , although they were unable to quantify these differences due to their models being unable to vary @xmath24fe abundance ratios . by considering the effects of age , metallicity and @xmath24fe ratios simultaneously on the spectra of 124 local early - type galaxies , @xcite
show that those in high - density environments are older and slightly more metal - rich than their counterparts in field regions .
we should expect much smaller differences in the predicted average ages of red sequence galaxies as even our low - density environments are typically within 1 cluster radii , while gradients in the star - formation histories of cluster galaxy populations have been shown to extend out to 34 virial radii before reaching field values .
these results are understandable in the framework of cosmological structure formation models , in which galaxies form earliest in the highest - density regions corresponding to the cluster cores , and have their star - formation terminated earliest as the cores become filled with shock - heated virialised gas which does not easily cool or collapse , suppressing further formation of stars and galaxies @xcite .
age - gradients remain in clusters to this epoch , as mixing of the galaxy population is incomplete , and the position of the galaxies within the cluster are correlated with the epoch at which they were accreted , and the point at which star - formation within them is suppressed @xcite .
we identify two predominant trends relating galaxy colours , luminosities and their local environment : the anti - correlation between blue galaxy fraction and local density ; and the monotonic increase in the blue galaxy fraction with @xmath5-band magnitude in all environments studied .
these trends mirror those obtained by @xcite relating the red galaxy fraction ( selected by @xmath25 colour ) , @xmath26 and local density for galaxies taken from the sdss first data release , although they indicate that the strength of the trend with environment is independent of luminosity , whereas we find that the effect of environment on the blue galaxy fraction is much greater at fainter magnitudes ( @xmath27 ) than at brighter magnitudes ( @xmath28 ) where red galaxies dominate in all environments .
as all the environments studied here lie within @xmath29 of one of the clusters , this suggests that processes related to the supercluster environment are much more important for the evolution of faint galaxies ( @xmath30 ) than their brighter counterparts .
we find that the @xmath2 colour of a galaxy is dependent its environment , whereas its luminosity is independent of local density , except at the very brightest magnitudes .
red galaxies located on average in regions with a mean local density greater than ( ) , while blue galaxies are on average found in less dense regions .
we find evidence for an increase in clustering of low - luminosity red galaxies as also observed in the sdss @xcite .
the effect appears to be greatest at , with fainter red galaxies found in less dense environments .
this is most likely to be related to the dramatic effects of environment on the luminosity function of faint galaxies observed for this region described in paper i , in which dips are apparent at , which becomes deeper with decreasing local density .
the observed dips are due to a relative lack of red sequence galaxies at these magnitudes in the lower density regions in comparison to the cluster core , and so it is manifested here as an enhanced clustering of red galaxies at .
these observed changes in the galaxy luminosity function with environment were explained in paper i in terms of the galaxy harassment scenario , in which infalling sc
sd spiral galaxies are transformed by repeated high - velocity encounters over a period of several gyr , and in the process become 12mag dimmer due to both mass loss and as a result of their star - formation being terminated @xcite . these galaxies which predominate in clusters at
are thus shifted to fainter magnitudes , producing the observed dip , and hence the red galaxies at are found preferentially in the highest - density environments as they are produced as a direct result of harassment by the supercluster environment . at fainter magnitudes
the harassment becomes too great , and in fact the dwarf galaxies are completely disrupted by tidal forces in the cluster centre @xcite and their stars cannibalised by the cd galaxy .
the cores of each of the clusters are completely dominated by red sequence galaxies , as indicated by regions containing very low fractions of blue galaxies , and red mean galaxy colours . in the cases of a3562 and sc1329 - 313
these regions , which coincide with both the centres of x - ray emission and the location of the brightest cluster galaxy ( bcg ) , are displaced from the peaks in the galaxy surface density , particularly for sc1329 - 313 .
the displacement , at least between the x - ray emission / bcg and the peaks in the galaxy surface density are likely to be related to the dynamics of the system , that is an ongoing cluster merger , which has been proposed from an analysis of the x - ray emission @xcite and the finding of a young radio halo at the centre of a3562 @xcite .
the strong correlation between the regions devoid of blue galaxies and the x - ray emission , as opposed to the density peaks , may be evidence in support of ram pressure stripping having an important role in terminating any remnant star - formation in galaxies that encounter the cluster cores , although equally it could be a consequence of the dynamics of the system displacing the faint galaxy population .
the a3562sc1329 - 313 system appears the most active in terms of star - formation , with the highest densities of blue galaxies across the supercluster survey region found here , in particular the north - western flank of sc1329 - 313 , but also between the two clusters .
these are also the regions where in a vla radio survey of the whole supercluster core complex @xcite found dramatic increases in the fraction of galaxies belonging to the supercluster ( i.e. with redshifts ) which have associated radio emission . from a visual inspection of their morphology
, these radio emitters are found to be spirals / irregulars , and this region marks the greatest concentration of bright spiral galaxies across the supercluster complex .
@xcite interpreted these as young starburst galaxies related to the cluster merger event , and taking this in addition to our results which are sensitive to the dwarf galaxy population , we can say this star - formation activity extends to .
are these galaxies which are undergoing starbursts triggered by the effects of the cluster merger , for example by encountering shocks in the icm @xcite , or are they are a group of normal star - forming galaxies which have just entered the supercluster ? through an analysis of the x - ray emission from a3562 and sc1329 - 313 , @xcite argued that sc1329 - 313 had recently passed westwards through the northern outskirts of a3562 and then been deflected south . this would appear to contradict the idea that sc1329 - 313 is a remnant from the ongoing a3562-a3558 merger , and instead suggests it represents a recent arrival in the supercluster core complex .
however the regions containing the largest concentrations of faint blue galaxies are located _ between _ clusters , in particular along the filament connecting the clusters a3558 and a3562 , which would support the hypothesis that shocks in the icm produced by cluster mergers is triggering starbursts in these galaxies .
the area covered by the sos dataset lies mostly within one virial radius of one of the clusters .
the results presented here and in paper i show that the global properties of faint galaxies change significantly from the cluster cores to the virial radius , both in terms of the luminosity function , and the mean galaxy colours , which indicates that these galaxies are being transformed by processes related to the supercluster environment . as
galaxy mergers should be infrequent in any of the environments covered by the sos , the finding of such large changes in the mean galaxy colour or fraction of faint blue galaxies within the sos , indicates that some process other than merging must be responsible . in paper i , we suggested that galaxy harassment is important for shaping the galaxy luminosity function at magnitudes fainter than , and here we find additional evidence in favour of faint galaxies being transformed by ram - pressure stripping and undergoing starbursts triggered by shocks in the icm produced by cluster mergers .
these results indicate that the effect of environment on faint galaxies is quite different from that observed for bright galaxies @xcite . while bright galaxies appear to be transformed by processes that can take place well outside the virial radius , that is galaxy merging or suffocation , we find here that many faint galaxies are affected by their interaction with the supercluster environment , although we can not rule out that the faint galaxies are also transformed outside the virial radius , beyond the limits of our survey .
it is also possible that the observed environmental trends are partly due to a population of primordial , early - type galaxies that formed preferentially in the high - density regions that later became clusters ( e.g. * ? ? ?
the differences in the environmental trends of bright and faint galaxies are likely to be related to the observed global bimodality in galaxy properties ( e.g. * ? ? ?
* ) about a stellar characteristic stellar mass of .
a possible explanation could lie within the context of the hot and cold flow model of galaxy evolution @xcite . at low masses ,
galaxies are able to merge and still retain their gas supply , and hence mergers would have little effect on star - formation or galaxy colours .
in contrast , once galaxies merge to become more massive than a characteristic mass , shocks in the halo become stable , and preventing further cooling of gas from the halo , bringing about a transformation in the galaxy star - forming properties . in a future article , we plan to compare our observational results with predictions from semi - analytical models of galaxy evolution applied to n - body simulations of a region containing a supercluster constrained to match the dynamical structure of the shapley supercluster region .
we thank the anonymous referee for helpful comments that have improved this paper .
cph , ag and nk acknowledge the financial supports provided through the european community s human potential program , under contract hprn - ct-2002 - 0031 sisco .
am is supported by the regione campania ( l.r .
05/02 ) project _
evolution of normal and active galaxies _ and by the italian ministry of education , university and research ( miur ) grant cofin2004020323 : _ the evolution of stellar systems : a fundamental step towards the scientific exploitation of vst_. nk is partially supported by the italian ministry of education , university and research ( miur ) grant cofin2003020150 : _ evolution of galaxies and cosmic structures after the dark age : observational study_. we thank the deep lens survey team and noao who provided the data used to derive field galaxy counts . 99 aarseth s.j .
, fall s.m . , 1980 , apj , 236 , 43 abadi m.g . ,
moore b. , bower r.g . , 1999 ,
mnras , 308 , 947 babu g.j .
, singh k. , 1983 , ann . stat . , 11 , 999 baldry i.k .
, glazebrook k. , brinkmann j. , ivezi , z. , lupton r.h .
, nichol r.c . ,
szalay a.s . , 2004 ,
apj , 600 , 681 balogh m.l .
, baldry i.k . , nichol r. , miller c. , bower r. , glazebrook k. , 2004 , apjl , 615 , 101 bardelli s. , zucca e. , zamorani g. , vettolani g. , scaramella r. , 1998 , mnras , 296 , 599 bardelli s. , zucca e. , zamorani g. , moscardini l. , scaramella r. , 2000 , mnras , 312 , 540 beers t.c . , flynn k. , gebhardt k. , 1990 , aj , 100 , 32 bell e.f . , wolf c. , meisenheimer k. , et al .
2004 , apj , 608 , 752 bertin e. , arnouts s. , 1996 , a&as , 331 , 439 blanton m. , cen r. , ostriker j.p . ,
strauss m.a . , tegmark m. , 2000 , apj , 531 , 1 blanton m. , hogg d.w . , bahcall n.a .
, et al . , 2003 , apj , 594 , 186 blanton m. , eisenstein d. , hogg d.w . , zehavi i. , 2004 , apj in press ( astro - ph/0411037 ) blanton m. , eisenstein d. , hogg d.w .
, schlegel d.j . ,
brinkmann j. , 2005 , apj , 629 , 143 bruzual g. , charlot s. , 2003 , mnras , 344 , 1000 butcher h. , oemler a. , 1978 , apj , 219 , 18 butcher h. , oemler a. , 1984 , apj , 285 , 426 dekel a. birnboim y. , 2006 , mnras , 368 , 2 diaferio a. , kauffmann g. , balogh m.l . , white s , d.m .
, schade d. , ellingson e. , 2001 , mnras , 232 , 999 dressler a. , oemler a. , butcher h. , gunn j.e . , 1994 , apj , 430 , 107 dressler a. , et al . , 1997 ,
apj , 490 , 577 drinkwater m.j .
, parker q.a . , proust d. , slezak e. quintana h. , 2004 , pasa , 21 89 eisenstein d.j . , hogg d.w .
, fukugita m. , et al . , 2003 , apj , 585 , 694 faber s.m . ,
willmer c.n.a . ,
wolf c. , et al .
, 2005 , apj submitted , ( astro - ph/0506044 ) finoguenov a. , henriksen m.j . ,
briel y g. , de plaa j. , kaastra j.s . , 2004 ,
apj , 611 , 811 fujita y. , nagashima m. , 1999 , apj , 516 , 619 gmez p.l . , nichol r.c .
, miller c.j .
, et al . , 2003 , apj , 584 , 210 ghigna s. , moore b. , governato f. , lake g. , quinn t. , stadel j. , 1998 , mnras , 300 , 146 gray m.e , wolf c. , meisenheimer k. , taylor a. , dye s. , borch a. , kleinheinrich m. , 2004 , mnras , 347 , l73 haines c.p . , mercurio a. , merluzzi p. , la barbera f. , massarotti m. , busarello g. , girardi m. , 2004 , a&a , 425 , 783 hogg d.w . , blanton m.r . , brinchmann j. , et al . , 2004 , apjl , 601 , 29 kauffmann g. , 1995 , mnras , 274 , 153 kauffmann g. , heckman t.m .
, white s.d.m .
, et al . , 2003 , mnras , 341 , 54 kere d. , katz , n. , weinberg , d.h . ,
dav , r. 2005 , mnras , 363 , 2 kodama t. , bower r.g .
, 2001 , mnras , 321 , 18 kodama t. , balogh m.l . , smail i. , bower r.g . , nakata f. , 2004 , mnras , 354 , 1103 landolt a.u . , 1992 ,
aj , 104 , 340 larson r.b . , tinsley b.m . , caldwell c.n . , 1980 , 237 , 692 lewis i. , balogh m. , de propris r. , et al . , 2002 , mnras , 334 , 673 lpez - cruz o. , barkhouse , w.a . ,
yee h.k.c . , 2004 ,
apj , 614,679 mercurio a. , merluzzi p. , haines c.p .
, gargiulo a. , krusanova n. , busarello g. , la barbera f. , 2006 , mnras , 368 , 109 metcalfe n. , godwin j.g . , peach j.v . , 1994 , mnras , 267 , 431 miller n.a . , 2005 , aj , 130 , 2541 moore b. , katz n. , lake g. , dressler a. , oemler a. jr . , 1996 , nature , 379 , 613 moore b. , lake g. , katz n. , 1998 , apj , 495 , 139 pisani a. , 1993 , mnras , 265 , 706 pisani a. , 1996 , mnras , 278 , 697 poggianti b. , 1997 , a&as , 122 , 399 poggianti b. , von der linden , a. , de lucia , g. , et al .
2006 , apj , 642 , 188 quintana h. , carrasco h. , reisenegger a. , 2000 , aj , 120 , 511 reisenegger a. , quintana h. , carrasco h. , maze , j. 2000 , aj , 120 , 523 roettiger k. , burns j.o .
, loken c. , 1996 , apj , 473 , 651 shapley h. , 1930 , bull .
harvard obs . , 874 , 9 strateva i. , ivezc z. , knapp g.r .
et al . , 2001 , aj , 122 , 1861 tanaka m. , goto t. , okamura s. , shimasaku k. , brinkmann j. , 2004 , aj , 128 , 2677 thomas d. , maraston c. , bender r. , mendes de oliveira c. , 2005 , apj , 621 , 673 treu t. , ellis r. , kneib j .-
, dressler a. , smail i. , czoske o. , oemler a. , natarajan r. , 2003 , apj , 591 , 53 tremonti , c.a . , heckman , t.m . , kauffmann g. , et al .
2004 , apj , 613 , 898 vandame , 2004 , phd thesis venturi t. , bardelli s. , morganti r. , hunstead r.w . ,
2000 , mnras , 314 , 594 wittman b.m .
, tyson j.a . , dellantonio i.p .
et al . , 2002 ,
spie , 4836 , 73 zehavi i. , zheng z. , weinberg d.h . , et al .
, 2005 , apj , 630 , 1
|
we present an analysis of the effects of environment on the photometric properties of galaxies in the core of the shapley supercluster at @xmath0 , one of the most massive structures in the local universe .
the shapley optical survey ( sos ) comprises archive wfi optical imaging of a 2.0deg@xmath1 region containing the rich clusters a3556 , a3558 and a3562 which demonstrate a highly complex dynamical situation including ongoing cluster mergers .
the colour - magnitude relation has an intrinsic dispersion of 0.045mag and is mag redder in the highest - density regions , indicative of the red sequence galaxy population being 500myr older in the cluster cores than towards the virial radius .
the @xmath2 colours of galaxies are dependent on their environment , whereas their luminosities are independent of the local density , except for the very brightest galaxies .
the global colours of faint galaxies change from the cluster cores where of galaxies lie along the cluster red sequence to the virial radius , where the fraction has dropped to just this suggests that processes related to the supercluster environment are responsible for transforming faint galaxies , rather than galaxy merging , which should be infrequent in any of the regions studied here .
the largest concentrations of faint blue galaxies are found _ between _ the clusters , coincident with regions containing high fractions of galaxies with radio emission indicating starbursts .
their location suggests star - formation triggered by cluster mergers , in particular the merger of a3562 and the poor cluster sc1329 - 313 , although they may also represent recent arrivals in the supercluster core complex .
the effect of the a3562-sc1329 - 313 merger is also apparent as a displacement in the spatial distribution of the faint galaxy population from both the centres of x - ray emission and the brightest cluster galaxies for both systems .
the cores of each of the clusters / groups are marked by regions that have the lowest blue galaxy fractions and reddest mean galaxy colours over the whole supercluster region , confirming that star - formation rates are lowest in the cluster cores . in the cases of a3562 and sc1329 - 313 ,
these regions coincide with the centres of x - ray emission rather than the peaks in the local surface density , indicating that ram - pressure stripping may have an important role in terminating any remnant star - formation in galaxies that encounter the dense icm of the cluster cores .
[ firstpage ] galaxies : clusters : general galaxies : clusters : individual : shapley supercluster galaxies : photometry galaxies : evolution
|
main problem is shown in figure ( [ schema1 ] ) .
there is viscous flow between two parallel plates .
upper plate is moving in x direction with constans velocity @xmath0 .
lower one is not moving @xmath1 .
we are looking for a solution to describe velocity vector field in the model ( between two plates ) .
most of incompressible fluid mechanics ( dynamics ) problems are described by simple navier - stokes equation for incompressible fluid velocity , which can be written with a form : @xmath2 where @xmath3 is defined is defined as the relation of pressure to density : @xmath4 and @xmath5 is a kinematics viscosity of the fluid .
we will also use a continuity equation , which can be written as follows : @xmath6 of course , in a case of couette incompressible flow we will use several simplifications of ( [ n - s equation ] ) .
incompressible couette problem is not needed to solve full navier - stokes equations .
there is no external force , so first simplification of ( [ n - s equation ] ) will be : @xmath7 in @xcite there can be found easy proof that in couette problem there are no pressure gradients , which means that : @xmath8 we will ignore a convection effects so , equation ( [ n - s equation1 ] ) can be written with a form : @xmath9 now we have simple differential equation for velocity vector field .
that equation is a vector type and can be simplified even more .
let us write continuity equation ( [ continuity equation ] ) in differential form .
let @xmath10 , then continuity equation can be expanded as follows : @xmath11 we know that there is no @xmath12 velocity component gradient along x axis ( symmetry of the problem ) , so : @xmath13 evaluation of taylor series at points @xmath14 and @xmath15 gives us a proof that only one possible and physically correct value for y component of velocity @xmath16 is : @xmath17 because of ( [ yvelocity ] ) equation ( [ n - s equation2 ] ) can be written as follows : @xmath18 our problem is now simplified to mathematical problem of solving equations like ( [ lastequation ] ) .
that is now a governing problem of incompressible couette flow analysis .
analytical solution for velocity profile of steady flow , without time - changes ( steady state ) can be found very easily in an equation : @xmath19 and without any changes of viscosity @xmath5 it can be written in form : @xmath20 after simple two times integration of equation ( [ lastequation - steady state2 ] ) we have analytical solution function of ( [ lastequation - steady state2 ] ) : @xmath21 where @xmath22 and @xmath23 are integration constans .
simple boundary conditions are provided in that problem .
we know that : @xmath24 simple applying it to our solution ( [ analitsolut ] ) gives a more specified one , where @xmath25 and @xmath26 : @xmath27 it means , that a relationship between @xmath12 and @xmath28 is linear . a better idea to write that with mathematical expression is : @xmath29 where @xmath30 is a constans for the problem ( initial @xmath12 velocity vs size of the model ) .
let us define some new non - dimensional variables : @xmath31 now let us place these variables into equation ( [ lastequation ] ) and we have now a non - dimensional equation written as follows : @xmath32 now we replace all the variables to nondimensional , like defined in ( [ dimensionless ] ) : @xmath33 now we will remove all @xmath34 chars from that equation ( only for simplification of notation ) , and the equation becomes to : @xmath35 in that equation reynold s number @xmath36 appears , and is defined as : @xmath37 where @xmath36 is reynold s number that depends on @xmath38 height of couette model .
finally , the last form of the equation for the couette problem can be written as follows : @xmath39 we will try to formulate numerical solution of the equation ( [ lastnumeq ] ) .
in our solution of equation ( [ lastnumeq ] ) we will use crank - nicolson technique , so discrete representation of that equation can be written as : @xmath40 simple grouping of all terms which are placed in time step ( n+1 ) on the left side and rest of them - on right side , gives us an equation which can be written as : @xmath41 where @xmath42 is known and depends only on values @xmath12 at @xmath43 time step : @xmath44 constans @xmath45 and @xmath46 are defined as follows ) . ] : @xmath47 @xmath48
for numerical solution we will use one - dimensional grid points @xmath49 where we will keep calculated @xmath12 velocities .
that means @xmath12 has values from the range @xmath50 .
we know ( from fixed boundary conditions ) , that : @xmath51 and @xmath52 . simple analysis of the equation ( [ systemofequation1 ] ) gives us a system of equations , which can be described by matrix equation : @xmath53 where @xmath54 is tridiagonal @xmath55\cdot [ n-1]$ ] matrix of constant @xmath45 and @xmath46 values : @xmath56\ ] ] @xmath57 vector is a vector of @xmath12 values : @xmath58\ ] ] @xmath59 vector is a vector of constans @xmath42 values : @xmath60\ ] ]
now the problem comes to solving the matrix - vector equation ( [ govequation ] ) .
there are a lot of numerical methods for that ] , and we will try to choose two of them : thomas and gauss method .
both are very similar , and i will start with a description of my implementation with the simple gauss method .
choice of the gauss method for solving system of linear equations is the easiest way .
this simple algorithm is well known , and we can do it very easily by hand on the paper .
however , for big matrices ( big @xmath61 value ) a computer program will provide us with a fast and precise solution .
a very important thing is that time spent on writing ( or implementing , if gauss procedure was written before ) is very short , because of its simplicity .
i used a gauss procedure with partial choice of a / the general element .
that is a well known technique for taking the first element from a column of a matrix for better numerical accuracy .
the whole gauss procedure of solving a system of equations contains three steps .
first , we are look- ing for the general element .
after that , when a general element is in the first row ( we make an exchange of rows matrix , and for @xmath62 too . ] ) we make some simple calculations ( for every value in every row and column of the matrix ) for the simplified matrix to be diagonal ( instead of a tridiagonal one which we have at the beginning ) .
that is all , because after diagonalization i implement a simple procedure ( from the end row to the start row of the matrix ) which calculates the whole vector @xmath57 .
there are my values of @xmath63 velocity in all the model .
thomas method , described in @xcite is simplified version of gauss method , created especially for tridiagonal matrices .
there is one disadvantage of gauss method which disappears when thomas method is implemented .
gauss method is rather slow , and lot of computational time is lost , because of special type of matrix .
tridiagonal matrices contain a lot of free ( zero ) values . in the gauss method
these values joins the calculation , what is useless .
thomas simplification for tridiagonal matrices is to get only values from non - zero tridiagonal part of matrix .
simply applying a thomas equations for our governing matrix equation ( [ govequation ] ) gives us : @xmath64 @xmath65 we know that exact value of @xmath66 is defined as follows : @xmath67 now solution of the system of equations will be rather easy .
we will use recursion like that : @xmath68 that easy recursion provides us a solution for the linear system of equations .
main results are provided as plots of the function : @xmath69 in figure ( [ plot1anal ] ) there is drawn an analytical solution to the problem of couette flow . that is linear function , and we expect that after a several time steps of numerical procedure we will have the same configuration of velocity field . in figure ( [ plot1 ] )
there are results of velocity @xmath12 calculation for several different time steps .
analytical solution is also drawn there . as we can see in the figure ( [ plot1 ] )
- the solution is going to be same as analytical one . beginning state ( known from boundary conditions ) is changing and relaxing . in the figure ( [ plotreynolds ] )
there is plot of numerical calculations for different reynold s numbers .
for example reynold s number depends on i.e. viscosity of the fluid , size of couette model . as it is shown on the plot
there is strong relationship between the speed of the velocity field changes and reynold s number . in a couple of words : when reynolds number increases - frequency of changes also increases . ) and different reynold s numbers . ] in figure ( [ plotgrid ] ) there is an example of calculations of velocity field for different grid density ( @xmath61 number ) .
we see that there is also strong correlation between grid density , and speed of changes on the grid .
also , very interesting case @xmath70 shows , that for low density of the grid changes are very fast , and not accurate . ) , same reynold s numbers ( re=5000 ) and different grid density ( @xmath61 number ) . ]
solving of incompressible couette problem can be good way to check numerical method , because of existing analytical solution . in that report
there were presented two methods of solving system of equations : gauss and thomas method .
system of equations was taken from crank - nicolson implicit scheme .
well known linear relationships were observed .
.... # include < stdlib.h > # include < stdio.h > # include < math.h > # define n ( 40 ) # define nn ( n+1 ) void zamien(double * a , double * b ) { double c ; c=*a ; * a=*b ; * b = c ; } void wypiszmacierz(double a[nn][nn ] , int n ) { int i , j ; for(j=1;j < n;j++ ) { for(i=1;i < n;i++ ) // show matrix { printf("%2.4f " , a[i][j ] ) ; } printf("\n " ) ; } } void gauss(double a[nn][nn ] , double * b , double * x , int n ) { int i , j , k ; double m ; //
gauss elimination for(i=0;i <
n;i++ ) { // step # 1 : change governing element m = fabs(a[i][i ] ) ; k = i ; for(j = i+1;j <
n;j++ ) if(fabs(a[i][j])>m ) { m = fabs(a[i][j ] ) ; k = j ; } if(k!=i ) for(j=0;j <
n;j++ ) { zamien(&a[j][i],&a[j][k ] ) ; zamien(&b[i+1],&b[k+1 ] ) ; } // step # 2 : make it triangle for(j = i+1;j <
n;j++ ) { m = a[i][j]/a[i][i ] ; for(k = i;k < n;k++ ) a[k][j ] = a[k][j ] - m*a[k][i ] ; b[j+1 ] = b[j+1 ] - m*b[i+1 ] ; } } // step#3 : solve now for(i = n-1;i>=1;i-- ) { for(j = i+1;j <
n;j++ ) b[i+1 ] = b[i+1]-a[j][i]*x[j+1 ] ; x[i+1 ] = b[i+1]/a[i][i ] ; } } int main(void ) { double u[n*2 + 2]={0},a[n*2 + 2]={0},b[n*2 + 2]={0},c[n*2 + 2]={0},d[n*2 + 2]={0},y[n*2 + 2]={0 } ; //
initialization double oneovern = 1.0/(double)n ; double re=5000 ; //
reynolds number double ee=1.0 ; // dt parameter double t=0 ; double dt = ee*re*oneovern*2 ; // delta time double aa=-0.5*ee ; double bb=1.0+ee ; int kkend=1122 ; int kkmod=1 ; int kk ; // for a loop int i , j , k ; // for loops too int m ; //
temporary needed variable double gmatrix[nn][nn]={0 } ; // for gauss elimination double test ; y[1]=0 ; //
init // apply boundary conditions for couette problem u[1]=0.0 ; u[nn]=1.0 ; //
initial conditions ( zero as values of vertical velocity inside of the couette model ) for(j=2;j<=n;j++ ) u[j]=0.0 ; a[1]=b[1]=c[1]=d[1]=1.0 ; for(kk=1;kk<=kkend;kk++ ) { for(j=2;j<=n;j++ ) { y[j]=y[j-1]+oneovern ; a[j]=aa ; if(j==n ) a[j]=0.0 ; d[j]=bb ; b[j]=aa ; if(j==2 ) b[j]=0.0 ; c[j]=(1.0-ee)*u[j]+0.5*ee*(u[j+1]+u[j-1 ] ) ; if(j==n ) c[j]=c[j]-aa*u[nn ] ; } // gauss
// c [ ] - free //
a[]b[]d [ ] - for matrix calculation // u [ ] - x //
calculate matrix for gauss elimination gmatrix[0][0]=d[1 ] ; gmatrix[1][0]=a[1 ] ; for(i=1;i < n-1;i++ ) { gmatrix[i-1][i]=b[i+1 ] ; //
gmatrix[1][2]=b[2 ] gmatrix[i][i]=d[i+1 ] ; //
gmatrix[2][2]=d[2 ] gmatrix[i+1][i]=a[i+1 ] ; //
gmatrix[3][2]=a[2 ] } gmatrix[n-2][n-1]=b[n ] ; gmatrix[n-1][n-1]=d[n ] ; gauss(gmatrix , c , u , n ) ; // gauss solving function y[1]=0.0 ; y[nn]=y[n]+oneovern ; t = t+dt ; // time increment test = kk % kkmod ; if(test < 0.01 ) // print the results { printf("kk , time\n " ) ; //
info 1 printf("%d,%f\n",kk , t ) ; printf(",j , y[j],u[j]\n " ) ; //
info 2 for(j=1;j<=nn;j++ ) printf("%d , % f , % f\n",j , u[j],y[j ] ) ; printf("\n \n \n \n " ) ; // for nice view of several datas } } return ( 1 ) ; } ....
.... double oneovern = 1.0/(double)n ; double re=7000 ; // reynolds number double ee=1.0 ; //
dt parameter double t=0 ; double dt = ee*re*oneovern*2 ; // delta time double aa=-0.5*ee ; double bb=1.0+ee ; int kkend=122 ; int kkmod=1 ; int kk ; // for a loop int j , k ; // for loops too int m ; // temporary needed variable printf(",j , y[j],u[j]\n " ) ; //
info 2 for(j=1;j<=nn;j++ ) printf("%d , % f , % f\n",j , u[j],y[j ] ) ; printf("\n \n \n \n " ) ; // for nice view of several datas } } return ( 1 ) ; } .... aa john d. andertson , jr . _
computational fluid dynamics : the basics with applications _ , mcgraw - hill inc , 1995 .
david potter _
metody obliczeniowe fizyki _ , pwn 1982 .
ryszard grybos , _
podstawy mechaniki plynow _ ( tom 1 i 2 ) , pwn 1998 .
|
this project work report provides a full solution of simplified navier stokes equations for the incompressible couette problem .
the well known analytical solution to the problem of incompressible couette is compared with a numerical solution . in that paper
, i will provide a full solution with simple c code instead of matlab or fortran codes , which are known . for discrete problem formulation ,
implicit crank - nicolson method was used .
finally , the system of equation ( tridiagonal ) is solved with both thomas and simple gauss method .
results of both methods are compared .
|
infrared spectroscopy has in the past several decades become one of the premier experimental tools in condensed matter physics @xcite .
thanks to its versatility , it has been successfully applied to essentially all types of condensed matter systems , such as superconductors , topological insulators , graphene , etc . in particular , in high - t@xmath0 cuprate superconductors , infrared based techniques have been extensively used to probe a variety of unusual and yet unresolved issues concerning their unconventional pairing state @xcite . in recent years
a number of attempts has been made to elucidate the properties of quasiparticles and their relaxation in the cuprates @xcite .
these attempts are based on the so - called extended - drude model , which allows both the quasiparticle scattering rate and their effective mass to acquire frequency dependence .
these two quantities can be straightforwardly obtained from the complex optical conductivity @xmath1 as : @xmath2= \frac{\omega_{p}^{2 } } { 4 \pi } \frac{\sigma_1(\omega ) } { \sigma_1 ^ 2(\omega)+\sigma_2 ^ 2(\omega ) } % \label{eq : tau}\ ] ] @xmath3 \frac{1}{\omega } = \frac{\omega_{p}^{2}}{4 \pi } \frac{\sigma_2(\omega ) } { \sigma_1 ^ 2(\omega)+\sigma_2 ^ 2(\omega)}\frac{1}{\omega } \label{eq : mass}\ ] ] where the plasma frequency @xmath4 is usually obtained from the integration of @xmath5 up to the frequency of the onset of interband absorption : @xmath6 equations [ eq : tau ] and [ eq : mass ] are the basis of a so - called _ one - component _ approach @xcite for the interpretation of optical properties , which assumes the existence of a single type of charge carriers in the system .
closely related quantities are the optical self - energy @xmath7(@xmath8 ) @xcite and memory function @xmath9(@xmath8 ) @xcite defined as : @xmath10 as an example , in fig .
[ fig : bisco ] we displays the real and imaginary parts of the memory function @xmath11 , as well as the effective mass @xmath12 for optimally doped bi@xmath13sr@xmath13cacu@xmath13o@xmath14 ( bi2212 ) with t@xmath0= 92 k @xcite , both in the normal and superconducting state .
application of eqs .
[ eq : tau ] , [ eq : mass ] and [ eq : memory ] to the data in the superconducting state is highly problematic , even though it has been routinely done .
namely , in the superconducting state normal fluid coexists with the superfluid , and the one - component assumption is clearly violated .
as was most recently pointed out by homes et al .
@xcite this procedure can not be used to make any reliable statements about quasiparticle properties in the superconducting state . in the superconducting state
the response at microwave and far - infrared frequencies is dominated by the superfluid , causing @xmath15 ( ref . ) and it follows from eq .
[ eq : tau ] that 1/@xmath16 = m@xmath17 @xmath18 acquires small values ( see fig . [
fig : bisco ] ) .
similarly , indiscriminate application of eq .
[ eq : mass ] to the optical conductivity in the superconducting state will result in @xmath19 which decreases when the superfluid forms , and @xmath20 increases ( see fig .
[ fig : bisco ] ) .
= 92 k. ( a ) real part of memory function m@xmath21 obtained from eq .
[ eq : memory ] .
( b ) imaginary part of memory function m@xmath17 from eq . [
eq : memory ] .
( c ) quasiparticle effective mass @xmath12 calculated from eq .
[ eq : mass ] .
the values of plasma frequency @xmath22 used for each temperature are shown in fig .
[ fig : sigma](c).,width=359 ] [ fig : bisco ] in this work we show how to circumvent this problem , and access intrinsic quasiparticle properties in the superconducting state .
the procedure expands the range of applicability of one - component approach , and further extends the power of infrared spectroscopy .
our procedure is based on kramers - kronig transformation of the optical conductivity . in the next section ( sec .
[ sec : kk ] ) we first present the formalism in details . in sec .
[ sec : model ] we apply it to model data , which serves to illustrate the main idea and demonstrate its usefulness .
the procedure is then applied to existing infrared data on optimally doped high - t@xmath0 superconductor bi2212 with t@xmath0= 92 k ( sec .
[ sec : bi2212 ] ) .
we present and discuss the quasiparticle properties obtained for the first time below t@xmath0 . in this section
we also address the issue of a narrow quasiparticle peak that has been observed in microwave spectroscopy , and show how it can be combined with infrared data to obtain a unified and self - consistent picture of electrodynamic properties of bi2212 .
finally , in section [ sec : summary ] we summarize the most important findings made possible by the new procedure .
in this section we present the details of our procedure .
the approach is based on the kramers - kronig transformation of optical conductivity , which we introduced previously to extract the superfluid density from infrared data @xcite . in the superconducting state the real part of optical conductivity
can be written as : @xmath23 where the first term on the right - hand side is the superconducting , and the second the quasiparticle contribution .
the corresponding imaginary part of @xmath24 follows from a kramers - kronig relation : @xmath25 where @xmath26 is the superfluid density or stiffness , and @xmath27 is the superconducting plasma frequency .
the dirac delta function in @xmath24 is not accessible in optical data , which typically start at several mev .
however , the 1/@xmath8 term in @xmath28 is mixed up with @xmath29 and contributes to both 1/@xmath30(@xmath8 ) and m@xmath31(@xmath8)/m@xmath32 ( eqs .
[ eq : tau ] and [ eq : mass ] ) . to determine the intrinsic quasiparticle properties we must separate the two terms in eq .
[ eq : s2 ] . to that end
we employ a kramers - kronig transformation on @xmath33 : @xmath34 we emphasize that this step is completely model - independent ; no a priori assumptions are made about the form of quasiparticle conductivity . once @xmath29 is calculated from eq .
[ eq : s2kk ] , one can calculate the intrinsic scattering rate and effective mass in the superconducting state ( eqs .
[ eq : tau ] and [ eq : mass ] ) using the kramers - kronig - corrected @xmath29 , instead of @xmath28 .
note that @xmath24 does not need to be corrected , as the delta function ( eq . [ eq : s1 ] ) is not accessible to optical experiments .
using the procedure described above , we can also calculate the superfluid density from eq .
[ eq : s2 ] , as was done before @xcite : @xmath35
to test the procedure and to demonstrate its usefulness in this section we perform the calculations on model data .
we adopt a drude model for the normal state , and a bcs model for an s - wave superconductor in the superconducting state .
the model also includes a quasiparticle peak inside the superconducting gap .
the critical temperature was set at t@xmath0= 90 k , and the corresponding t=0 energy gap is 2@xmath36= 220 @xmath37 ( 27.3 mev ) .
real and imaginary parts of @xmath38 are shown with thick lines at several temperatures in fig . [
fig : model](a ) and ( b ) respectively . in the superconducting state a characteristic suppression of @xmath39
is observed below the gap . the spectral weight removed from these frequencies
is transferred to the delta function at zero frequency ( eq .
[ eq : s1 ] ) .
the values of plasma frequency are calculated from eq .
[ eq : plasma ] and displayed in fig .
[ fig : model](c ) with red circles .
note that in eq .
[ eq : plasma ] the integral starts from 0 + , which emphasizes the fact that in the superconducting state only the quasiparticle contribution should be counted towards @xmath22 . on the other hand , @xmath20 is dominated by a characteristic 1/@xmath8 response of the superfluid ( eq . [ eq : s2 ] ) .
we also note that the absolute values of @xmath20 are at least an order of magnitude larger than @xmath39 . of a bcs model with t@xmath0= 90 k. ( b ) imaginary part of the optical conductivity @xmath20 .
thick lines represent the model function , whereas thin lines of the same color represent kramers - kronig - corrected @xmath29 .
( c ) temperature dependence of plasma frequency @xmath22 , superconducting plasma frequency @xmath27 and the total plasma frequency ( @xmath40+@xmath41)@xmath42.,width=359 ] [ fig : model ] we now apply the procedure outlined in the previous section . in fig .
[ fig : model](b ) with thin lines we display @xmath29 calculated from eq .
[ eq : s2kk ] .
we note that these kramers - kronig - corrected curves are not dominated by 1/@xmath8 superfluid response , but instead display a broad peak at finite frequencies , similar to the one seen at 90 k. the removal of superfluid response also reveals pronounced structure at the gap frequency , which is not observable in @xmath28 ( before the correction ) . in fig .
[ fig : modelm ] we display with blue lines the results for the real ( top panels ) and imaginary ( bottom panels ) parts of the memory function @xmath11 from eq .
[ eq : memory ] . in the normal state
both parts display constant values , typical of the drude model . on the other hand , in the superconducting state , m@xmath43(@xmath8 )
is suppressed at higher frequencies , but does not show any characteristic features at the gap , because the response is dominated by the superfluid .
the imaginary part m@xmath13(@xmath8 ) displays characteristic suppression , especially below the gap . in the same figure we also plot with red lines the results for the memory function obtained with kramers - kronig - corrected @xmath29 .
expectedly , in the normal state the memory function is the same as before .
however , in the superconducting state , the removal of the superfluid contribution reveals a very pronounced structure at the gap frequency in m@xmath21 .
the imaginary part m@xmath17 also displays structure at the gap frequency , but more importantly the suppression below the gap is much smaller then before . above the gap m@xmath17 is enhanced compared to the normal state . in fig .
[ fig : model](c ) we display the temperature dependence of the plasma frequency @xmath22 ( red circles ) obtained from eq .
[ eq : plasma ] , and the superfluid density ( blue squares ) obtained from eq . [ eq : rhos ] . as discussed above ,
the spectral weight removed form finite frequencies is transferred to the delta function at zero energy , but the total spectral weight must be conserved .
this is indeed confirmed by fig .
[ fig : model](c ) where the total ( i.e. combined ) plasma frequency ( @xmath40+@xmath41)@xmath42 , is shown with green triangles and is constant within the error bars of numerical calculations .
note that the application of eq .
[ eq : rhos ] to the normal state data results in a small , but finite value of superconducting plasma frequency .
this is due to numerical errors , and does not imply the existence of superfluid in the normal state , above t@xmath0 . and @xmath44 for the model shown in fig .
[ fig : model ] .
the results are shown before ( blue lines ) and after ( red lines ) kramers - kronig correction.,width=359 ] [ fig : modelm ]
before applying the procedure to bi2212 , we must address the issue of a very narrow quasiparticle peak that has been observed in microwave measurements @xcite .
its width is typically a few mev @xcite , which is outside the frequency window of typical infrared measurements .
the existence of this quasiparticle peak is usually ignored during analysis of infrared data , as it does not contain a lot of spectral weight and does not significantly affect the calculation of normal state plasma frequency ( eq . [ eq : plasma ] ) .
however , this narrow peak can produce a significant 1/@xmath8 contribution to @xmath20 , which mimics the superfluid response .
if we want to calculate the london penetration depth or the superfluid stiffness , it must be separated from the superconducting contribution .
we show below that in the case of optimally doped bi2212 this can lead to correction of superconducting plasma frequency @xmath27 by as much as 40 @xmath45 . in order to perform the integration in eq .
[ eq : s2kk ] optical conductivity data must be extended down to zero frequncy .
it was recently shown that because of the kramers - kronig relations between @xmath39 and @xmath20 one can calculate the spectral weight that is located below the lowest measured frequency @xcite .
however , the optical functions themselves ( @xmath39 and @xmath20 ) can not be retrieved without making some model assumptions about the optical spectrum . here
we will make the reasonable assumption that the quasiparticle contribution can be approximated with the drude model , and we combine it with the microwave data on bi2212 .
[ fig : sigma ] displays @xmath39 from infrared , as well as the microwave values at 34.7 ghz @xmath46 1.15 @xmath37 ( the values at 14.4 and 24.6 ghz are similar ) @xcite .
we now fit the complex conductivity @xmath38 _ simultaneously _ with microwave data , imposing the constraint that the total spectral weight in the superconducting state is conserved @xcite . in fig .
[ fig : sigma ] the infrared data is shown with thick lines , and model fits with dotted lines .
the microwave values for the corresponding temperatures are shown with the circles of the same color .
of optimally doped bi2212 with t@xmath0= 92 k. thick lines represent the experimental data , thin dashed lines of the same color represent drude extrapolations , and the circles represent microwave data at 34.7 ghz @xcite .
( b ) imaginary part of the optical conductivity @xmath20 .
thick lines represent the experimental data , thin dashed lines of the same color represent drude extrapolations and thin lines represent kramers - kronig - corrected @xmath29 .
( c ) temperature dependence of plasma frequency @xmath22 , superconducting plasma frequency @xmath27 and the total plasma frequency ( @xmath40+@xmath41)@xmath42 .
open squares represent the values of superconducting plasma frequency @xmath27 obtained without kramers - kronig - correction.,width=359 ] [ fig : sigma ] with optical conductivity extended down to zero frequency we can now apply the procedure to bi2212 and the results are shown in fig .
[ fig : sigma](b ) with thin lines ( only below t@xmath0 ) .
we can see that instead of a characteristic 1/@xmath8 divergence , the spectra display a finite frequency peak , characteristic of quasiparticle response .
the values of normal state plasma frequency @xmath22 and superconducting plasma frequency @xmath27 are shown in fig .
[ fig : sigma](c ) with red circles and blue squares , respectively .
the total plasma frequency ( @xmath40+@xmath41)@xmath42 ( green triangles ) is within 1.5 @xmath45 of the normal state value .
also shown with empty squares is the superconducting plasma frequency calculated using un - corrected @xmath20 , which can be as much as 40 @xmath45 higher than the corrected one .
once the superfluid contribution is removed from @xmath38 , one can calculate intrinsic quasiparticle properties ( eqs .
[ eq : tau ] , [ eq : mass ] and [ eq : memory ] ) . fig .
[ fig : kkbisco ] displays m@xmath43(@xmath8 ) ( top panels ) , m@xmath13(@xmath8)=1/@xmath16 ( middle panels ) and effective mass @xmath12 ( bottom panels ) , both before ( blue lines ) and after kramers - kronig corrections ( red lines ) . several selected temperatures are shown , both in the normal ( 100 k ) and superconducting state ( 80 , 50 and 10 k ) .
expectedly , optical functions in the normal state are the same before and after kramers - kronig correction . in the superconducting state , on the other hand , the corrections are significant and can not be neglected .
we note that even though the plasma frequency @xmath22 decerases in the superconducting state ( fig .
[ fig : sigma](c ) ) , all optical functions are _ enhanced _ compared to their un - corrected values .
in addition , the structure at around 700 @xmath37 is much more pronounced in all corrected spectra .
the reason is the removal of superconducting contribution from @xmath20 , which exposes the true quasiparticle properties .
( top panels ) and @xmath44 ( middle panels ) , as well as quasiparticle effective mass m@xmath31/m@xmath32 ( bottom panels ) for the bi2212 data shown in fig .
[ fig : sigma ] .
the results are shown before ( blue lines ) and after ( red lines ) kramers - kronig correction.,width=359 ] [ fig : kkbisco ] in fig .
[ fig : temp ] we display the temperature dependence of quasiparticle scattering rate and effective mass , both before ( blue circles ) and after ( red circles ) kramers - kronig correction .
the values of scattering rate were extracted as the average values at around 30 @xmath37 . on the other hand the values of the mass were obtained from the linear fits of the low frequency m@xmath43(@xmath8 ) spectra ( eq . [ eq : memory ] ) .
this method has proven to be more reliable then a simple extrapolation , in particular in the normal state .
the temperature dependence of scattering rate is similar before and after the correction , however the suppression of 1/@xmath30(@xmath8 ) below t@xmath0 is much less pronounced after the correction . on the other hand , the effective mass reveals dramatically different behaviour . before kramers - kronig correction the mass decreases in the superconducting , as anticipated above .
however , once the superfluid and the true quasiparticle properties are exposed , we can see that the effective mass actually increases as temperature decreases below t@xmath0 .
this indicates that the correlations are getting stronger in the superconducting state . before and after kramers - kronig correction .
( b ) temperature dependence of effective mass m@xmath31(@xmath8)/m@xmath32 before and after kramers - kronig correction .
these values are extracted as explained in the text .
the effective mass without kramers - kronig correction appears to decrease in the superconducting state .
however , with kramers - kronig correction the effective mass continues to increase below t@xmath0 , which indicates that the quasiparticles are more correlated than in the normal state.,width=359 ] [ fig : temp ]
we presented a way of calculating intrinsic quasiparticle properties , such as memory function and effective mass , in the superconducting state .
the procedure allows access to the quasiparticle properties that were previously inaccessible to infrared spectroscopy .
the method was first tested on model data , and then applied to infrared data on optimally doped bi2212 .
the calculations have revealed that intrinsic quasiparticle scattering rate and effective mass are enhanced in the superconducting state . in particular , the effective mass increases below t@xmath0 compared with the normal state values .
this indicates that thermally excited quasiparticle in the superconducting state are more correlated than in the normal state .
we expect the method described here to be a useful tool for infrared spectroscopy , which will allow quasiparticle properties to be studied for the first time in the superconducting state .
in particular , there are several issues in the cuprates that can be immediately addressed using this new procedure . scaling analysis @xcite
can now be extended below t@xmath0 , and the question of fermi vs non - fermi liquid quasiparticles can be studied .
another important issue is the controversy regarding the doping dependence of the quasiparticle effective mass .
namely it was recently show using quantum oscillations @xcite that the quasiparticle effective mass in yba@xmath13cu@xmath47o@xmath48 in the superconducting state diverges as doping is reduced .
this finding is in apparent contradiction with previous infrared measurements @xcite . using hall data to discriminate between carrier density _
n _ and effective mass m@xmath31 contributions to infrared spectral weight , padilla _
_ @xcite found that the effective mass in both yba@xmath13cu@xmath47o@xmath48 and la@xmath49sr@xmath50cuo@xmath51 was constant across the phase diagram . on the other hand , by fitting a strong - coupling expressions in the normal state , van heumen _
@xcite arrived at a factor of two decrease of the mass enhancement factor when the doping is increased from 0.1 to 0.21 holes per cuo@xmath13 unit , in agreement with the behaviour predicted from dynamical mean field theory @xcite . using the procedure outlined in this paper one can now access quasiparticle effective mass below t@xmath0 and address this important issue in the zero temperature limit .
this work was supported by the swiss national science foundation ( snsf ) through grant 200020 - 140761 .
mirzaei , d. stricker , j.n .
hancock , c. berthod , a. georges , e. van heumen , m.k .
chan , x. zhao , y. li , m. greven , n. barisic , d. van der marel , proceedings of the national academy of sciences , * 110 * 5774 ( 2013 ) .
homes , j.j .
tu , j. li , g.d .
gu and a. akrap , scientific reports * 3 * , 3446 ( 2013 ) .
dordevic , e.j .
singley , d.n .
basov , seiki komiya , yoichi ando , e. bucher , c.c .
homes and m. strongin , physical review b * 65 * , 134511 ( 2002 ) .
shih - fu lee , d. c. morgan , r. j. ormeno , d. m. broun , r. a. doyle , and j. r. waldram and k. kadowaki , physical review letters * 77 * , 735 ( 1996 ) .
d. a. bonn , r. liang , t.m .
riseman , d.j .
baar , d.c .
morgan , k. zhang , p. dosanjh , t.l .
duty , a. macfarlane , g.d .
morris , j.h .
brewer , w.n .
hardy , c. kallin and a. j. berlinsky , physical review b * 47 * , 11314 ( 1993 ) .
it was shown by molegraaf et al .
@xcite that a fraction of spectral weight is transfered from the visible to the infrared spectral range .
note however that this effect is below the error bars of our calculations .
sebastian , n. harrison , m.m .
altarawneh , c.h .
mielke , r. liang , d.a .
bonn , w.n .
hardy and g.g .
lonzarich , proceedings of the national academy of sciences * 107 * , 6175 ( 2010 ) .
padilla , y.s .
lee , m. dumm , g. blumberg , s. ono , k. segawa , s. komiya , y. ando and d.n .
basov , physical review b * 72 * , 060511(r ) ( 2005 ) .
e. van heumen , e. muhlethaler , a.b .
kuzmenko , h. eisaki , w. meevasana , m. greven and d. van der marel , physical review b * 79 * , 184512 ( 2009 ) .
|
quasiparticle properties in the superconducting state are masked by the superfluid and are not directly accessible to infrared spectroscopy .
we show how one can use a kramers kronig transformation to separate the quasiparticle from superfluid response and extract intrinsic quasiparticle properties in the superconducting state .
we also address the issue of a narrow quasiparticle peak observed in microwave measurements , and demonstrate how it can be combined with infrared measurements to obtain unified picture of electrodynamic properties of cuprate superconductors .
|
it has been assumed over the years that the building of galactic disks is driven inside - out by the continuous accretion of in - falling gas ( larson 1976 ) .
evidences for the validity of the inside - out paradigm are still slim however , and are actively searched for . inspection of detailed cdm simulations have shown that this picture may not be universal , and may vary from one galaxy to the other , to the point that some simulations show examples where galaxies can build their disk outside - in , at least partly ( see e.g sommer - larsen et al .
2003 , roberston et al .
2004 ) . in this
respect , radial variations of properties of the galactic disk may provide important observational constraints , but this has been hampered by the lack of tracers for which distance can be measured with the relevant accuracy , explaining why the magnitude of abundance gradients in the disk is still actively debated .
nonetheless , it can be considered that a consensus on the existence of gradients has been reached , and there is little doubt that inner regions are , in the mean , more metal rich than the outer parts of the disk ( davies et al .
2009 ) , and this effect can be seen both in the iron - peak and the @xmath2-elements . also , recent understanding of the local metallicity distribution and kinematics strongly suggest that the solar radius is polluted by wanderers coming from the outer and inner disk , indicating a rather strong radial variation of the metallicity ( haywood 2008 ) .
because radial metallicity gradients in galactic disks may be produced by a number of different processes , the measurement of how these gradients have evolved with time should provide an even stronger constraint ( e. g. , fu et al .
2009 ) , thus the importance of studying metallicity distributions of tracers of different galactic ages . planetary nebulae ( pne ) represent the galactic stellar population with progenitors of turnoff mass m@xmath7 between @xmath01 and 8 m@xmath1 , probing galactic ages between @xmath8 yr and @xmath910 gyr ( maraston 1998 ) .
planetary nebulae are distributed in the galactic disk , bulge , and halo .
their galactic distribution and radial velocities offer a first subdivision into populations , which can be studied separately .
furthermore , markers such as nitrogen abundances and morphological types allow to further discriminate between disk pne of relatively young and old progenitors .
it is then sensible to undergo a study of galactic metallicity for the different pn populations .
this has been done in the past .
galactic disk pne have been studied , among others , by kingsburg & barlow ( 1994 ) and , more recently , by perinotto et al .
( 2004 , p04 ) , while bulge pne are the subject of a focused study by exter et al .
the galactic disk pne have been the subject of many studies on metallicity gradients : @xmath2-element abundances of pne trace the original progenitor composition , thus the chemical evolution of the galactic disk .
perinotto & morbidelli ( 2006 , pm06 ) reviewed all galactic metallicity gradients published since the seventies , and found that pne in the galactic disk trace a one dimensional , negative oxygen gradient @xmath3log(o / h)/@xmath3r@xmath4=-0.07 ( faundez - abans & maciel 1987 ) and -0.03 ( pasquali & perinotto 1993 ) dex kpc@xmath6 . the excellent data revision by perinotto and collaborators provides its own gradient of -0.016 dex kpc@xmath6 ( pm06 ) , which is shallower than previously known .
almost contemporary stanghellini et al .
( 2006 ) found a shallow gradient of -0.01 dex kpc@xmath6 for the oxygen abundances of galactic disk pne , consistent with pm06 s analysis .
the abundance analysis that yield to the gradients published so far might be marginally different , but hardly enough to make the metallicity gradients to differ outside the data uncertainties .
a stronger factor for divergence is certainly the method of galactic pn distance calculation , although pm06 showed that the gradients should result rather flat independent on the distance scale used .
the distance scale of pne used is also a determining factor in the study of the different galactic pn populations , their distance from the galactic plane and their belonging to the thick disk , the bulge , the halo . until a couple of years ago , distances to galactic pne where estimated by means of statistical distance scales , calibrated with a few known distances to galactic pne .
now , with the availability of a wealth of spatially resolved images of magellanic cloud pne from _ hst _ , the calibration has been done in a much sounder way ( stanghellini et al .
2008 , ssv ) .
there is then purpose to reanalyze both the galactic distribution of pne and the metallicity gradients in a homogeneous way , with the distance scales based on the physical parameters of magellanic cloud pne , whose distances are known .
in this paper we use the magellanic cloud - calibrated galactic pn distance scale from ssv to examine the population of pne in our galaxy .
we extend the p04 abundance database with all the abundances published more recently , and calculate the metallicity gradients for the different populations , and homogeneously classify the galactic pn morphology for further insight on population selection . in @xmath102
we describe the database built for this study ; in @xmath103 we discuss the galactic structure based on pne ; in @xmath104 we present the metallicity gradients from our analysis ; sections 5 and 6 present respectively a discussion of our results and the conclusions of this work .
in order to build a database as complete as possible we started from acker et al.s ( 1994 ) catalog of galactic pne and include all confirmed pn therein .
the catalog offers the galactic coordinates and basic data for most nebulae .
the nebular optical diameters , essential for determining the distances , have been taken from cahn et al.s ( 1992 , cks ) study , which takes into account the acquisition aperture , and , if not available therein , from acker et al .
furthermore , for extended northern galactic pne the diameters have been taken from manchado et al .
the nebular flux at 5 ghz has been taken again from cks preferentially , otherwise from acker et al .
if not available , the 5 ghz flux has been approximated by its relation to the h@xmath11 flux ( see cks ) . as mention in the introduction
the major reason for the reanalysis of the metallicity gradients is the advent of the new distance scale by ssv , which has been calibrated on the magellanic cloud pne . in this paper
we preferentially use the best , newly calibrated distance scale , which is the re - calibration of the cks scale ( eq . 8a and 8b in ssv ) , but we also compare to other scales that use the ionized mass , the brightness temperature , and the surface brightness relations to the nebular radii ( eq . 5 , 6 , and 7 in ssv ) to study the impact of different ( but equally well - calibrated ) scales on the metallicity gradients and the galactic distributions .
it is worth noting that the choice of magellanic - cloud calibrated distance scales is based on the fact that these calibrations give the best results when compared to known individual pn distances , and they represent an improvement with respect to the old galactic - calibrated scales ( see table 3 in ssv ) .
planetary nebula morphology has been used successfully as a second parameter to determine the population type ( schwarz et al .
1993 , stanghellini et al .
1993 , stanghellini et al .
the morphological classification available so far does not extend homogeneously to the complete sample of pne used in the present study .
we have then classified all pne homogeneously , based on the available [ o iii ] @xmath125007 images available from manchado et al .
( 1996 ) , schwarz et al .
( 1992 ) , and other images all included in the online image collection by b. balick . knowing the morphological class allow us to exclude bipolar pne from the sample used for gradient determination .
in fact , ssv showed that bipolar pne do not follow the cks calibration , and prefer to use the surface brightness- radius distance scale model for these pne . on the other hand , it is very hard to define a radius for bipolar pne , thus their distance is always very uncertain , and it is just safer to exclude the extremely asymmetric pne from the gradient analysis .
in table 1 ( published electronically ) we give in column ( 1 ) the pn names as in acker et al .
( 1994 ) ; in column ( 2 ) the heliocentric distance calculated from the magellanic cloud re - calibration of the cks scale , except for bipolar pne where we used the magellanic cloud re - calibration of the sb - r@xmath13 relation , and where the uncertainties are assumed to be 20@xmath14 for all pne ; column ( 3 ) gives the galactocentric distances ( see below ) , column ( 4 ) the angular radii , column ( 5 ) the 5ghz flux , and column ( 6 ) the new morphological classification in major classes : round ( r ) , elliptical ( e ) , bipolar core ( bc ) , and bipolar ( b ) . in some cases
the morphology was not defined , due to lack of clarity in the images or to very small nebular size .
galactocentric distances , r@xmath4 , are calculated by de - projecting heliocentric distances onto the galactic plane with the usual relation @xmath15 , where d is the pn heliocentric distance , r@xmath1=8.0 kpc is the solar distance to the galactic center , and b and l are the galactic coordinates of the nebulae . while the statistical distance scale can not provide errors , ssv showed that uncertainties are well within 30@xmath14 of the calibration .
we assign a rather conservative uncertainty of @xmath520@xmath14 to the heliocentric distances , and we then propagate the error to the galactocentric distances , assuming that the galactic coordinates l and b have much lower relative uncertainties than the heliocentric distances . the galactocentric distances and their ( formal ) uncertainties
are given in table 1 , column ( 3 ) .
there are 1143 pne in acker s catalog , whose diameters are measured ( either directly or as an upper limit ) in 1053 cases . we were able to determine the pn distances for @xmath0730 pne .
perinotto et al .
( 2004 ) have carefully reviewed all elemental abundance analysis of galactic pne published since the seventies , and selected a sample that is homogeneous in abundance determination methods and high quality of the observations ( their sample a ) .
we use the a sample from p04 in our study , augmented with the abundances published later than the p04 sample was analyzed , selecting the sets with the same criteria used by p04 .
our updated sample includes abundances from ( 1 ) p04 ; ( 2 ) stanghellini et al .
( 2006 ) ; ( 3 ) the series of papers by pottasch and collaborators , summarized by pottasch & bernard - salas 2006 ; ( 4 ) kwitter & henry ( 2001 ) , milingo et al . ( 2002 ) , and kwitter et al .
( 2003 ) ; and ( 5 ) and costa et al .
it is worth noting that the series of papers by k. kwitter and collaborators are formally dated before the p04 review , but actually the p04 work was completed before the former abundances have became available , thus not included in p04 .
there are several other papers on galactic pn abundances published after 2004 , but with the exception of the ones listed above they are reanalysis of earlier databases , thus already included in p04 , or excluded based on their selection criteria .
finally , it is important to note that stanghellini et al .
( 2006 ) and pottasch & bernard - salas ( 2006 ) abundances were calculated with the same method and ionization correction factors ( icf ) than those of p04 , all utilizing the kingsburg & barlow ( 1994 ) prescription . on the other hand , kwitter & henry ( 2001 ) , milingo et al .
( 2002 ) , kwitter et al .
( 2003 ) , and costa et al .
( 2004 ) used a slightly a different icf for oxygen . in this paper
we correct for the differences in icf(o ) by using the ionic abundances given in the relevant papers . in the case
when more than one reference give abundances for a particular pn we checked that the differences are within the uncertainties , and then chose the abundances preferring , after p04 , all references whose abundances were originally calculated with the icf relations by kingsburg & barlow ( 1994 ) .
we end up with a sample of 206 pne with homogeneously calculated oxygen abundances , while the sample with neon abundances consists of 167 pne . for more than half of the pne of these sample we also have the helium and nitrogen abundance to determine whether belong to the type i class ( peimbert & torres - peimbert 1983 ) .
we use the uncertainties in the abundances as quoted in the original papers , when available .
perinotto et al .
( 2004 ) gives formal errors for all abundances , which are used here directly .
the abundances from stanghellini et al .
( 2006 ) have uncertainties of 0.04 dex for all elements , except for neon where the uncertainty is of the order of 0.1 dex .
kwitter and collaborators assume that the typical uncertainty for the oxygen abundance is @xmath16 . in the cases where the uncertainties are not given ( e. g. , costa et al .
2004 ) we assume an uncertainty of 0.15 dex for a conservative approach . in table 2 ( published electronically ) we list all pn abundances used in this paper .
a pn is in this table if at least one elemental abundances among the ones used in this paper ( helium , nitrogen , oxygen , and neon ) is available .
column ( 1 ) gives the pn name ; columns ( 2 ) through ( 5 ) give the elemental abundances and their uncertainties , derived as explained above , where he / h is given in linear form , while the other abundances are given in the usual form a(x)=log(x
/ h)+12 .
the 1143 pne in acker s catalog belong to the thin disk , thick disk , halo , and bulge populations . in some of the applications below ( i.e. , disk metallicity gradients ) we select a pure disk population .
we use the standard criteria to select bulge population pne within angular distances less than 10 degrees from the galactic center , angular radii smaller than 10 arcsec , and 5 ghz fluxes smaller than 0.1 jy ( e. g. , chiappini et al .
there are 147 pne in our sample that satisfy all these conditions .
next we classify the pne with determined abundances as type i ( with log(n / o)@xmath17 - 0.3 and he / [email protected] ) , type ii ( non - type i pne with peculiar radial velocity v@xmath1960 km s@xmath6 ) , and type iii ( non - type i pne with v@xmath2060 km s@xmath6 ) as defined in p04 .
peculiar velocities have been derived from measured radial velocities by durand et al .
( 1998 ) and galactic rotation model therein , assuming a constant rotation for pne with r@xmath21r@xmath1 .
the type iii pne with altitude on the galactic plane larger than 800 pc are likely to be halo pne . in table 2 ,
column ( 6 ) , we give the disk population designation ( type i , ii , or iii ) for the pne , where possible , based on the abundances and kinematics . for non - disk pne
we indicate whether they belong to the bulge or to the halo .
in figure 1 we show the sample of pne with known distances in the r@xmath4z plane , where z is the altitude on the galactic plane .
we plot the different major morphological types : r ( round ) , e ( elliptical ) , bc ( bipolar core ) , and b ( bipolar ) in different panels , with all distances calculated with the magellanic cloud calibrated cks scale . using the most discordant distance scale , for example the magellanic cloud - calibrated m@xmath22-r@xmath13 scale , gives a very similar distribution for all morphological types , with the exception of bipolar pne , whose statistical distances are not very accurate anyway . from fig .
1 we confirm that the bipolar sample is confined close to the galactic plane ( again the distance bias for this class should be taken into account ) and the bipolar core pne are intermediately distributed between elliptical and bipolar , as found in previous analyses ( stanghellini et al .
1993 , 2002 ) . in figure 2 we plot on the same plane all the pne , and then those of peimbert type i , ii , and iii .
this plot has distances calculated in the same way than in fig . 1 , and includes bipolar pne . in order to assess the type of distance calibration we plot in figure 3 the r@xmath4z locus of all pne with distances calculated with the method above , and then with the t@xmath23r@xmath13 , m@xmath22r@xmath13 , and sb r@xmath13 relations , using in all cases the superior magellanic cloud calibrations ( eq . 5 , 6 , and 7 in ssv ) .
these samples of 728 pne include bipolar pne .
we can see that the different ( magellanic cloud - calibrated ) scales give broad agreement in the pn galactic distribution , but there are differences , in that pne are significantly more clustered around the sun when using magellanic cloud - calibrated cks distances .
the variation of the radially - binned @xmath2-elemental abundances in disk galaxies is generally indicated as the _ metallicity gradient _ , and it is a measure of the composition of the galaxy at the time of formation of the population under study . from the observational viewpoint ,
the metallicity gradient depends on the galaxy type , the population utilized , the method of calculating the @xmath2-element abundances , the distances of the probes from the galactic center , and also on the uncertainties associated to all these variables .
planetary nebulae well encompass several generations of stars in galaxies , thus are essential probes of metallicity gradient : their @xmath2-elemental abundances are relatively easily determined , and the samples are sizable . on the other hand , historically it has been difficult to reach an agreement on the value of the 1d gradient slope from galactic pne , due to two factors : the unreliability of the distance scale , and the selection of the pn population .
it is in fact essential that the gradients do not depend on the distance scale , but also that the disk population selected for this calculation is as complete and homogeneous as possible .
the use of oxygen and neon seems to be adequate in galactic studies , where oxygen production through the agb is marginal , while in low - metallicity galaxies oxygen might not be the ideal choice , since oxygen might be carried at the lims surface by the third dredge - up ( karakas et al .
2002 ) . in figures 4 through 6
we show the oxygen and neon abundances , in the usual form of a(x)=log(x / h)+12 , versus the galactocentric distances of the pne for the three pn types .
the samples have been selected against both bulge and halo pne , following the prescriptions of @xmath102.4 .
bipolar pne have not been included in the gradient plots to minimize the distance uncertainties of the disk pne .
note that the bump around the solar vicinity in figure 5 is due to the historical selection of nearby pne as spectral targets . in table 3
we give the abundance averages and gradient resulting from our analysis with , in column ( 1 ) the @xmath2-element considered , in column ( 2 ) the pn type included in the statistics , in column ( 3 ) the number of pne in the sample ( in parenthesis the sample size excluding bipolar pne , i. e. , the sample size used for gradients and plots ) , in column ( 4 ) the average linear abundance with its dispersion , in column ( 5 ) the dispersion of the y - axis distributions , in dex , in column ( 6 ) the gradient intercepts and their uncertainties , and in column ( 7 ) the gradient slopes ( in dex kpc@xmath6 ) and their uncertainties .
all samples analyzed here exclude halo and bulge pne .
abundance averages include bipolar pne , gradients and dispersions along the y - axis exclude all bipolar pne .
the gradients intercepts and slopes have been obtained with accurate fit .
the statistical distance scales , used when handling a sizable number of pne , do not provide errors for the individual distances .
the impossibility to determine the distance uncertainties , except with ad hoc assumptions , induced us to estimate the fits using the abundance errors , and varying the relative distance errors from 0 up to a reasonable maximum distance uncertainty .
to do so we have used the routines _ fit _ and _ fitexy _ , described in numerical recipes by press et al .
increasing the relative errors in the heliocentric distances from 0 to @xmath030@xmath14 will not change significantly the results of this paper . in figure 7
we show , as an example , the effect of changing the distance uncertainties on the slope of the oxygen gradients for type i ( squares ) and type ii ( circles ) pne .
note how small is the effect on the gradient slope for the large sample of type ii pne , where also the uncertainties in the slopes are much smaller ( yet much larger than the effect the distance uncertainty would have ) . on the other hand , for type
i pne , where the gradient relies on a few pne , the change of slope with augmented uncertainties in the distances has a notable effect , especially since this particular sample of pne has very few objects at large galactocentric distances , especially since the bipolar pne have been excluded .
this plot help us to underline the importance of the sample size as well as its homogeneity .
it is very important noting that for distance uncertainties between 0 and @xmath030@xmath14 the gradient slopes for type i , type ii , and type iii pne are always in a sequence , the former steeper than the latter , both for oxygen and neon , even if , due to the rather large uncertainties in the type i pn gradient slopes , the slopes of type i and type ii pne could be compatible .
the uncertainties to the fitted slopes given in columns ( 7 ) of table 3 have been obtained by minimizing the @xmath24 : we have reiterated the fits by artificially enlarging the data errors until the @xmath24 probability , q(@xmath24 ) , was larger than 0.8 .
note that the actual intercept and slope do not vary with this reiteration , but their uncertainties do .
the metallicity gradients for the whole pn samples are all moderately flat , seemingly excluding gradient flattening with time when compared to gradients from other young stellar populations .
it is worth studying the gradients of pne as a group , indicative of an intermediate - to - old age population , and also fractioned in the different pn populations , marking different evolutionary times of our galaxy .
planetary nebula progenitors have masses from @xmath01 to @xmath08 m@xmath1 .
the lower mass limit is constrained by the dynamical time vs. evolutionary time scale for pn evolution ( stanghellini & renzini 2000 ) , and by observations of pne in old stellar systems ( alloin et al .
1976 ) , while the upper limit , rather uncertain , marks the upper mass limit for a star to develop a degenerate co core , and stellar evolution models find this limit to vary according to model assumptions .
type i pne are nitrogen enriched with respect to the general population . from an observational viewpoint , the minimum progenitor mass of type i pne
is of the order of @xmath02 ( peimbert & serrano 1980 ) to @xmath03 ( kaler et al .
1990 ) m@xmath1 , while evolutionary models predictions again depend on model assumptions , such as extra - mixing or overshooting .
we assume that type i pne have m@xmath252 m@xmath1 , corresponding to an approximate progenitor age t@xmath261gyr derived from the turnoff mass- age relation for low- and intermediate - mass stars ( maraston 1998 ) .
it is worth noting that a mass limit of 3 m@xmath1 , corresponding to progenitors of [email protected] gyr , would not change the impact of our findings .
we then assume type ii , and type iii disk pne have progenitor masses in the approximate ranges of [email protected] , and m@xmath28 m@xmath1 respectively ( p04 ) , with progenitors ages approximately of 1@xmath29t@xmath295 , and t@xmath95 gyr ( maraston 1998 ) .
when we take into account the different pn populations , we see that gradients are much steeper for the young type i than for the old type iii disk pne , consistently for both the oxygen and the neon data sets .
this result goes in the opposite direction to the flattening with time , which maciel et al .
( 2003 ) have proposed based on pne .
we discuss this important result further in section 5 . in the previous section we noted that the gradients derived for non - type i pne are typically not steeper than -0.03 dex kpc@xmath6 , consistent with the most recent determinations ( e. g. , stanghellini et al . 2006 , p06 ) . to check the stability of our results it is worth going through the exercise of calculating the oxygen gradients with different assumptions
first , if we include the few halo pne in our sample the gradients would not change noticeably , even if it would change the meaning of the sample . a different argument
should be applied for bulge pne , since it is hard to separate bulge and disk populations . if we do not exclude bulge pne from our sample , and leave all other selection as for the set of table 3 , we obtain oxygen gradient slopes of [email protected] dex kpc@xmath6 for the complete sample , compared with [email protected] dex kpc@xmath6 obtained if we exclude the bulge population . for this and all other pn samples , the effect of including bulge pne is to increase the ( negative ) gradient slopes .
this could have contributed in part to the steeper gradients slopes of some of the published metallicity gradients from pne , where an accurate selection of pn populations was not in effect .
past metallicity gradient analyses have included bipolar pne . by including bipolar pne in the gradient calculation
we would obtain different gradients only for the type i pne ( @xmath3log(o / h)/@[email protected] dex kpc@xmath6 ) , where the bipolar constitute a very large fraction of the population .
given the large uncertainties in distances to bipolar pne , including them in even type i pn gradients might be of limited significance . including bipolar pne in the sample would not change the global gradient nor that of type ii pne in a significant way .
let us analyze what happens if we change the distance scale used to determine the metallicity gradients .
we run the fit routines for all disk pne and found : \(a ) @xmath31 dex kpc@xmath6 by using the @xmath32 relation ; \(b ) @xmath33 dex kpc@xmath6 for the the @xmath34 relation ; ( c)@xmath35 dex kpc@xmath6 for the sb - r@xmath13 relation , all calibrated with magellanic cloud pne ( ssv ) .
we conclude that including bulge pne steepens the metallicity gradient ; including bipolar pne has an impact but only for type i pne ; different statistical distance scale have a moderate impact on the gradients , if these scales are all calibrated on magellanic cloud pne ; and most importantly , the size of the statistical sample used to derive gradients might have an impact on the slopes .
in figure [ gradage ] we show the oxygen ( left panel ) and neon ( right panel ) metallicity gradient slopes against population age for the type i , type ii , and type iii pn progenitors , for pure galactic disk populations , bipolar excluded . to build this schematic plot we used the average pn population ages for the three types , as described in @xmath104.1 , where the bars represent the age span of each population , and the gradient slopes with their uncertainties as reported in table 3 .
all filled circles represent pn gradients from this paper .
note that the top x - axis gives the turnoff masses corresponding to the progenitor ages in the lower x - axis .
the results clearly indicates that the magnitude of the gradient slopes of both oxygen and neon increase going from the older population pne toward the young ones , indicating a steepening of the @xmath2-element gradient with time in the disk of the milky way .
the age intervals are not sharply defined , given the large uncertainties of data and models defining the different pn types , especially for the young pne , as discussed earlier in the paper .
nonetheless , given the scaling between ages and masses , choosing a lower limiting mass of 3 or 4 even m@xmath1 instead of 2 m@xmath1 for type i pne would not change the implications of fig .
it is also worth noting that the dispersion in the log(o / h ) along the galactic disk varies with pn type , being smaller for type ii pne and larger both for types i and iii ( see table 3 ) .
the gradient evolution with time relies on the results shown in figures 4 through 6 .
while the type ii pn sample is well represented and unambiguously leads to the calculated gradient slope , the other pn types should be discussed further .
first , the type i pn gradient indeed seems to rely on a very few points , especially at high galactocentric distances , with a single pn at 16 kpc skewing the slope . on the other hand , even excluding this type i pn from the gradient calculation we still obtain a steep ( negative ) slope ( although the dispersion is higher , @xmath36 dex kpc@xmath6 ) confirming the steepening with time of gradients from type ii to type i pne .
second , type iii pne represent the very old disk population , thus the question of gradient blurring due to radial migration occurs .
however , in figure 6 we see that no pn seems to stand out as having a particularly high or low abundance for its location , which could be the case if migration had erased or flattened a significant gradient .
moreover , comparison of figures 4 , 5 and 6 shows that most type iii pne at r@xmath37 7kpc have on average lower abundances than those seen at the same radii in the type i and type ii populations .
we consider that it would require a particularly fine tuning of radial migration processes to confine low abundances type iii pn and remove higher abundances in this distance range , particularly since such depletion is not seen for type ii population , which also must be affected by radial migration .
figure 6 shows limited scatter ( but significant dispersion ) ; we are aware that radial migration is an open issue , and it might be even higher than previously estimated ( rokar et al .
2008 ) , thus it is not impossible that type iii pn gradient has been partially flattened by stellar migration , but we consider it could hardly have erased a gradient that would have been higher than the one measured on type ii pn sample , in the case the gradient evolution was one flattening with time .
the pn metallicities and their gradients presented in this paper should be compared with abundances from other galactic components .
the overall comparison is limited for what the @xmath2-elements are concerned , and it is presented in section 5.2 .
the existence of gradients should however also reflects on solar neighborhood stars , because the local abundance distribution is not just the local sampling of stars born at solar galactocentric radius : although they dominate the local pool , the distribution is a complex mix of stars born at different distances from the galactic center . somehow , the amplitude of measured gradients should be compatible with the range of abundances measured at the solar neighborhood .
we address this question in section 5.1 .
the spread of metallicity in the thin disk at the solar radius is mostly due to non - evolutionary effects : the increase in metallicity due to chemical evolution cycling over the last 8 gyr ( the age of the thin disk according to solar neighborhood data , see e.g liu & chaboyer 2000 ) is limited to about [ fe / h]@xmath00.4 dex is intended relative to the sun ] ( from @xmath0 -0.2 to @xmath00.2 dex ) , while the range of measured metallicity is of the order of 1.2 dex ( from @xmath0 - 0.6 to @xmath00.6 dex ) .
how these extreme metallicities are compatible with the gradients measured in the previous section ?
the reason for this spread is to be found in stars that migrate across the disk ( haywood 2008 ) .
suggestions and results by grenon ( 1972 ) , sellwood & binney ( 2002 ) , and haywood ( 2008 ) conclusively point to the fact that local metal - rich ( [ fe / h]@xmath170.2 dex ) and metal - poor ( [ fe / h]@xmath29 - 0.2 dex ) stars of the thin disk have been brought to the solar radius by radial migration .
sellwood & binney ( 2002 ) proposed a mechanism that allows stars to individually exchange their angular momentum , allowing them to swap from one nearly circular orbit to another of different mean orbital radius . at the solar radius ,
this mechanism involves a moderate fraction of stars , but it is important since it gives a glimpse of the inner and outer galactic disk chemistry .
the highest and lowest thin disk iron abundances measured in the solar neighborhood are respectively [ fe / h]@xmath00.6 and @xmath0 - 0.6 dex , corresponding to [ o / h]@xmath00.3 and -0.3 dex ( ramrez et al . 2007 , fig
. 8) , which roughly corresponds to the factor of 1/2 that is often used when scaling iron to oxygen abundances ( e.g martinelli & matteucci 2000 , or more recently , caimmi 2008 ) .
the radial oxygen gradient for intermediate age pne ( type ii ) , @xmath0 - 0.023 dex kpc@xmath6 , would correspond to [ fe / h]@xmath0 - 0.05 dex kpc@xmath6 if using the conversion factor illustrated above .
given that the solar neighborhood mean metallicity is @xmath00.0 dex , such radial gradient implies a mean iron abundance of about 0.25 dex at 5 kpc towards the inner disk ( and -0.25 towards the outer disk ) .
allowing for an intrinsic dispersion similar to what is measured on the local iron distribution ( 0.15 dex ) , we easily reach ( within 1 or 2 sigma ) the highest metallicities that are observed at the solar radius , which means that the oxygen gradient measured on type ii pne is well compatible with abundances measured in the local stellar population . in order to increase the significance of the pn results we explore the literature for object classes
whose @xmath2-element abundances and gradients have been studied .
several studies concerning h ii regions and young stars are available , while the only possible comparison with an older population is that of open clusters . in table 4
we list gradient slopes and their uncertainties ( in column 2 ) and distance ranges ( column 3 ) from different tracers ( column 4 ) , where the authors have measured directly the oxygen abundances of the tracers .
most tracers are very young stars , whose gradients have been plotted in figure 8 with open circles .
these young stellar data seem to fit in well with the young pne of our sample , giving continuity to the plot .
our basic list of open clusters is the one designed by magrini et al . (
2009 ) , to which we added a few more clusters from the most recent literature .
the various parameters are given in table 5 with columns ( 1 ) through ( 3 ) providing the name , age , distance from the galactic center .
iron abundances , referred to solar , are listed in column ( 4 ) .
references to the iron abundances are as in magrini et al .
columns ( 5 ) through ( 8) of table 5 give the oxygen abundance ratios to iron , the oxygen abundances relative to solar , the actual oxygen abundance calculated for the solar value given in the individual papers , and the reference for the oxygen abundances .
if we examine the clusters with nominal ages corresponding to the progenitors of the three pn types , namely , in the age intervals of t@xmath261 , 1@xmath29t@xmath295 , and t@xmath95 gyr , we can compare directly the pn results with those age - appropriate clusters whose oxygen abundances are available .
we have calculated the absolute oxygen abundances for the clusters by using the same solar oxygen ratio used in the original papers . in cases where we could not find the solar oxygen ratio used by the authors
, we do not use the datum in the gradient determination , given that adopted values can differ by large amounts .
oxygen abundances are available for a few very young clusters ( t@xmath261 gyr ) within a very limited galactocentric distance range , thus an estimate of the gradient would not be meaningful ( although the abundance distribution would be compatible , within the uncertainties , with that of type i pne ) .
a better comparison sample is available for the intermediate age clusters , corresponding approximately to the ages of type ii pne progenitors , where the data span a range of galactocentric distances , and a gradient estimate is sensible .
we find that the oxygen gradient in open clusters of ages between 1 and 5 gyr is @xmath3log(o / h)/@xmath3r@xmath4=-0.028 dex kpc@xmath6 . as discussed further in the next section , ngc 6253 and ngc 6791
can be suspected of having an origin in the inner disk , therefore not being truly representative of the metallicity at the radius where they are observed today . only ngc 6253 falls in the age interval considered here .
if it is removed from the sample , the gradient decreases to -0.022 dex kpc@xmath6 .
the two estimates of metallicity gradient for open clusters of ages corresponding to type ii pn progenitors well encompass those derived directly from these pne . as shown in figure 8 , oxygen gradients from open clusters and pne of intermediate ages
are fully consistent with one another .
finally , for clusters older than 5 gyr we have only three possible data points and limited galactocentric distance range , which do not lead to a gradient determination , although the calculated slope would be fully compatible with that of type iii pne . in the section
above we have analyzed oxygen gradients in open clusters , and showed that a quantitative gradient comparison with the pn population could be only achieved with the intermediate age populations , lacking the necessary data for direct comparison of pne and open clusters in the young and old age bins .
for very young and very old clusters , we can recur to a more qualitative analysis , introducing the iron abundances .
figure 9 is a plot of [ fe / h ] vs. galactocentric distances of the clusters in table 5 .
young clusters ( t@xmath261 gyr ) are shown as squares , old ones ( t@xmath95 gyr ) as circles .
intermediate clusters are shown as diamonds .
the very metal rich open clusters ngc 6791 and ngc 6253 are specifically indicated .
the overall distribution has recently been described as testifying a flattening of the metallicity gradient in the outer disk ( at r@xmath3810 kpc ) , or even a plateau - like distribution ( sestito et al .
2008 and references therein ) .
we would like to propose another interpretation , suggested by the behavior of the two ages groups , provided that another parameter is taken into account , namely the possibility that some clusters , being on eccentric orbits , are seen at a radius not representative of their origin ( magrini et al . 2009 offers a similar explanation ) .
this is suggested in particular by the exotic nature of ngc 6791 and ngc 6253 , the most metal - rich clusters near the solar galactocentric radius ( in magrini et al .
ngc 6253 is given an age of 4.3 gyr , thus it is not included in the selection of old clusters , but its case is very similar to ngc 6791 ) .
ngc 6791 and ngc 6253 , with [ fe / h ] 0.2 dex above young metal - rich stars of the solar neighborhood , are possible candidates for being intruders from the inner disk ( see bedin et al .
2006 for the possible origin of ngc 6791 ) . by inspecting fig .
9 we note that ngc 6791 is largely responsible for the impression that old clusters have a steep gradient at r@xmath37 10 - 12 kpc .
once ngc 6791 is removed from the selection of old clusters , fig .
9 illustrates that this population has a shallower gradient than usually invoked .
it is worth noting that cluster collinder 261 , with [ fe / h]=0.13 ( sestito et al . 2008 ) , could also be an intruder .
however , its iron abundance is controversial ( friel et al . 2003 and
carretta et al .
2005 found [ fe / h ] respectively -0.02 and -0.03 dex ; we adopt the latter value in figure 9 ) .
the dichotomy between old and young clusters is then becoming more apparent , and suggests that the gradient slope change noted at r@xmath3910 kpc by several authors could result from the superposition of two different gradients for young and old clusters .
the distribution of intermediate age clusters shows ( expectedly ) a combination of the two behaviors .
we have not tried to make a fit to quantify the gradients within the groups of clusters : the number of objects in each group is small , and some of them are still being affected by large errors , making it difficult to choose the correct values of metallicities and galactocentric distances . moreover , due to the lack of accurate information on the possible orbits of each cluster , it would be difficult not to make an arbitrary selection of which clusters should be included or not in the old sample .
however , for illustrative purpose , we plotted different gradients on figure 9 to see how pn gradients are compatible with the above description .
the oxygen gradient derived for pne with progenitor younger than @xmath01 gyr is -0.035 dex kpc@xmath6 .
the old pn population , with progenitors older than @xmath05 gyr , have flat metallicity gradient ( -0.011 dex kpc@xmath6 ) . allowing for the factor 2 that has been evoked when
converting oxygen to iron abundances , we estimate an iron gradient of the order of -0.07 dex kpc@xmath6 for the young population , and a slope of @xmath0 - 0.02 dex kpc@xmath6 for the old population . these are plotted as continuous lines on fig .
9 . concerning the young population ,
we note good consistency between the gradients derived for type i pne and young clusters .
this is also compatible with the gradient derived from cepheids , which amounts to -0.05 to -0.06 dex kpc@xmath6 ( lemasle et al .
2008 , luck et al .
it is worth noting that the gradients for the old cluster population at the solar galactocentric radius reaches metallicities that correspond well to the metallicities of the oldest thin disk stars , at about @xmath0 -0.2 dex , according to the local age - metallicity relation ( haywood , 2006 ) .
the pn metallicity gradients give a clear indication of gradients that steepen with time . other stellar population sample that can be compared either directly - those with measured oxygen abundances - or indirectly are limited .
they however lead to a qualitatively similar , if tentative , diagnostic : gradients apparently steepen with time , at variance with previous suggestions ( maciel et al .
2003 ) . while a sound comparison of our results with maciel et al.s ( 2003 ) could not be performed until the pn database used by maciel et al .
is appropriately published , there is purpose to explore the consequences of our findings and their implications for galactic evolution .
we note that the change of slope , or the step detected at r@xmath4=10 kpc in the metallicity distribution of open clusters , which we interpret here as the superposition of a shallow gradient due to old clusters and a steeper one due to younger objects , has been detected in a number of other nearby galaxies .
this behavior has been presented in vlaji et al .
( 2009 ) , as an argument for a flattening of gradients with time .
moreover , vlaji et al . suggest that the metal - poor plateau in the outer disks could reflect the presence of an underlying old disk , formed at z@xmath02 . in the case of the milky way the clusters included in the selection at r@xmath38 10 kpc are mostly intermediate age objects ,
having ages between 5 and 10 gyr , thus they could not probe a disk formed at z@xmath02 .
this does not exclude the existence of such a primordial structure though : old open clusters could indeed originate from the slow evolution of a primordial gaseous disk .
the metallicity gradients calculated with pne and other stellar populations should be used to constraint the chemical evolutionary models .
these models can reproduce both steepening and flattening of the radial metallicity gradient with time , depending of the specific radial and time dependence of the star formation law and infall law , for which very little information is available , and sound data are essential to obtain insight on galactic evolution .
gradients of stellar metallicity in the galactic disk may become steeper with time if infall occurs preferentially at larger radii at later times , diluting metals in the outer disk ( i.e. , tosi 2000 ) .
the problem with this scenario is that evidences would be difficult to find , given the lack of possible direct observational clues about the temporal variation of infall .
alternatively , a steepening of the gradients with time could be understood if , starting from a similar metallicity , ( 1 ) the inner and outer parts of the milky way disk have evolved at different rates , or ( 2 ) the inner disk had more time to produce metals . in the latter case , we would expect the outer disk to be , on average , younger than the inner disk .
this is difficult to check with the available data , but such a test could be within the reach of gaia capabilities .
a possible indication however may be found in the local stellar population , since radial migration of stars pollutes it with intruders from the outer disk .
these stars are found in the same age range as stars with other metallicities ( except for the youngest ones , which can not be found locally because of the time needed to migrate from the outer regions of the disk ) , as testified by the local age - metallicity distribution ( haywood , 2006 ) . in other words
, there is no evidence that stars that have migrated from the outer disk can not be as old as stars formed at the solar radii , which would be the case if the outer disk was significantly younger . in this respect , the thin disk star hd 203608 , an example of star migrating from the outer disk , having a v - component space velocity in excess of the local standard of rest by 44 km s@xmath6 , and a metallicity of -0.6 dex , has its age measured by asteroseismic constraints ( mosser et al .
2008 ) at 7.25 gyr .
admittedly however , these facts are only suggestive , and can not be taken as evidence that the outer disk is as old as the inner thin disk , particularly since the time needed for significant migration in the disk introduces a bias favoring older stars and this is difficult to quantify .
the first scenario , where the inner and outer regions of the galactic disk evolve at different rates , might be more appealing .
the disk may have started to form stars at the same metallicity over the whole disk .
this is illustrated for example in the model of schnrich & binney ( 2009 ) .
when looking at their figure 6 , a similar behavior is ( qualitatively at least ) observed : starting at the same initial metallicity , the disk evolves at different rates depending on the distance to the galactic center .
the curves describing the increase of the metallicity at different radius are first close to each other then slowly separate to reach increasingly different metallicity at later times .
hence , a vertical cut through this plot would yield to steeper metallicity gradients for the young than the old stars .
the figure 6 of schnrich & binney ( 2009 ) also shows that the initial increase ( between ages 10 - 12 gyr ) in metallicity could be very steep , making it difficult observationally to distinguish whether the disk has started to form stars from a disk of gas with uniform metallicity as suggested by vlaji et al .
( 2009 ) or if a metallicity gradient pre - dated star formation .
a study on pn metallicity gradients employing the most updated abundances and distance scale shows that gradients of @xmath2-elements across the galactic disk are moderate , and depend on the age of the pn population considered .
the amplitude of these gradients agree with the metallicity dispersion observed in the solar vicinity , which arises mainly from the radial mixing of stars through the disk .
planetary nebulae with young progenitors show steeper gradient slopes than in the old populations , indicating that @xmath2-element gradients are steepening with time .
the results are statistically sound , and do not depend very much on the distance scale , as long as it is magellanic cloud - calibrated .
gradients of oxygen and iron abundance from young ( young stars and h ii regions ) and intermediate - age ( open clusters ) populations agree with those of pne derived here , giving strength to the scenario of gradients steepening with time .
the oxygen gradient is found to increase from near - flat radial abundance distribution in type iii pne , those whose progenitors are in the lowest mass range for agb stars , to -0.035 dex kpc@xmath6 for type i pne , those with the most massive progenitors .
these results contradict recent claim that gradients are flattening with time , and strongly suggest that such constraints should be taken with some caution .
the data on pne are still sparse and the new distance scale , although significantly improved , could still be affected by uncertainties .
in addition , abundance gradients for older populations of pne ( in particular the type iii sample ) might be affected by migration , which tends to flatten such gradients .
however , we find it unlikely that gradients in the old pn population could reach values as high as those found in some recent studies ( e.g maciel & costa 2009 ) , and this is confirmed by the old open cluster population with high resolution spectroscopic abundance determinations .
letizia stanghellini warmly thanks drs .
francoise combes and francois hammer for their hospitality at the observatoire de paris ( paris and meudon ) , where this work has been conceived and completed .
we acknowledge scientific discussions with drs .
bruce balick , laura magrini , karen kwitter , richard henry , angela bragaglia , monica tosi , and mark dickinson .
we are grateful to an anonymous referee for providing important comments to the earlier version of this paper .
manchado , a. , guerrero , m. a. , stanghellini , l. , & serra - ricart , m. 1996 , the iac morphological catalog of northern galactic planetary nebulae , publisher : la laguna , spain : instituto de astrofisica de canarias ( iac ) , 1996 , foreword by stuart r. pottasch , isbn : 8492180609 , 003.9 - 14.9 & [email protected] & [email protected] & 2.0 & 0.030 & e + 004.0 - 03.0 & [email protected] & [email protected] & 2.8 & 0.008 & @xmath40 + 004.0 - 05.8 & [email protected] & [email protected] & 6.0 & 0.003 & @xmath40 + 004.0 - 11.1 & [email protected] & [email protected] & 4.1 & 0.018 & e + 000.3 + 12.2 & 10.99 @xmath5 0.013 & 7.68 @xmath5 0.356 & 8.50 @xmath5 0.059 & 7.82 @xmath5 0.268 & type ii + 002.0 - 13.4 & 11.04 @xmath5 0.150 & 8.18 @xmath5 0.150 & 8.66 @xmath5 0.220 & 8.01 @xmath5 0.150 & type ii + 002.2 - 02.7 & 10.97 @xmath5 0.013 & 7.99 @xmath5 0.373 & 8.39 @xmath5 0.038 & 7.70 @xmath5 0.073 & bulge + o / h & all & 184 ( 145 ) & [email protected] & 0.25 & [email protected] & [email protected] + o / h & i & 23 ( 14 ) & [email protected] & 0.26 & [email protected] & [email protected] + o / h & ii & 103 ( 80 ) & [email protected] & 0.22 & [email protected] & [email protected] + o / h & iii & 29 ( 22 ) & [email protected] & 0.29 & [email protected] & [email protected] + ne / h & all & 154 ( 120 ) & [email protected] & 0.32 & [email protected] & [email protected] + ne / h & i & 16 ( 10 ) & [email protected] & 0.28 & [email protected] & [email protected] + ne / h & ii & 91 ( 70 ) & [email protected] & 0.27 & [email protected] & [email protected] + ne / h & iii & 23 ( 17 ) & [email protected] & 0.31 & [email protected] & [email protected] + cccc deharveng et al . ( 2000 ) & -0.035 @xmath5 0.0049 & 5 - 15 & h ii regions + daflon & cunha ( 2004 ) & -0.031 @xmath5 0.012 & 4.7 - 13.2 & ob stars + rood et al . ( 2007 ) & -0.043 @xmath5 0.007 & 0 - 17 & h ii regions + this paper & -0.028 to -0.022 & 7 - 22 & open clusters , 1 - 5 gyr + lrrrrrrr ic 4665 & 0.025 & 8.18 & -0.03 & 0.35 & 0.32 & 9.19 & s05 + ic 2602 & 0.035 & 8.49 & -0.05 & & & & + ic 2391 & 0.053 & 8.50 & -0.03 & & & & + blanco 1 & 0.1 & 8.50 & 0.04 & 0.02 & 0.06 & & pa08 + pleiades & 0.125 & 8.60 & -0.03 & & -0.01 & 8.85 & s07 + ngc 6475 & 0.220 & 8.22 & 0.14 & & & & + m 11 & 0.250 & 6.86 & 0.17 & & & &
+ m 34 & 0.250 & 8.90 & 0.07 & & 0.04 & 8.80 & sc03 + ngc 1883 & 0.65 & 12.3 & -0.01 & -0.22 & -0.23 & 8.70 & j09 + ngc 3960 & 0.7 & 7.96 & 0.02 & & & & + hyades & 0.7 & 8.50 & 0.13 & -0.01 & 0.12 & 8.81 & sc06 + praesepe & 0.7 & 8.62 & 0.27 & -0.4 & -0.13 & 8.53 & pa08 + ngc 2324 & 0.7 & 11.65 & -0.17 & & & & + ngc 2660 & 0.8 & 9.18 & 0.04 & & & & + mel 71 & 0.8 & 10.0 & -0.32 & -0.2 & -0.50 & 8.465 & b96 + ru 4 & 0.8 & 12.40 & -0.09 & & & & + ru 7 & 0.8 & 13.50 & -0.26 & & & & + ngc 2360 & 0.9 & 9.28 & 0.07 & & & & h00 + ngc 2112 & 1.0 & 9.21 & 0.16 & & & & + ngc 2477 & 1.0 & 8.94 & 0.07 & & & & + ngc 7789 & 1.1 & 9.44 & -0.04 & -0.07 & -0.11 & & t05
+ ngc 6939 & 1.3 & 8.4 & 0.00 & -0.11 & -0.11 & & + ngc 3680 & 1.4 & 8.27 & -0.04 & 0.2 & 0.16 & 8.82 & pa08 + ngc 6134 & 1.6 & 7.68 & 0.15 & & & & + ic 4651 & 1.7 & 7.70 & 0.11 & & & & + ngc 1817 & 1.7 & 10.0 & -0.07 & -0.19 & -0.25 & 8.68 & j09 + be 73 & 1.9 & 17.40 & -0.22 & & & & + ngc 2158 & 1.9 & 12.5 & -0.03 & -0.17 & -0.20 & 8.73 & j09 + ngc 2506 & 2.2 & 10.88 & -0.20 & & & & + ngc 2141 & 2.5 & 12.42 & -0.18 & 0.00 & -0.18 & 8.78 & y05 + ngc 6819 & 2.9 & 8.18 & 0.09 & & & & + be 66 & 3.7 & 12.9 & -0.48 & & & & + be 22 & 4.2 & 11.92 & -0.32 & & & & + m 67 & 4.3 & 9.05 & 0.03 & 0.01 & 0.04 & 8.70 & r06 + be 20 & 4.3 & 16.37 & -0.30 ; -0.49 & 0.18 & -0.31 & 8.55 & s08 ; y05 + be 29 & 4.3 & 22.0 & -0.31 ; -0.54 & 0.23 & -0.31 & 8.55 & s08 ; y05 + ngc 7142 & 4.4 & 9.70 & 0.14 & -0.10 & 0.04 & 8.97 & j08 + ngc 6253 & 4.5 & 7.10 & 0.36 & -0.18 & 0.28 & 9.07 & c07 + ngc 2243 & 4.7 & 10.76 & -0.48 & 0.18 & -0.30 & & gc94 + be 75 & 5.1 & 15.50 & -0.22 & & & & + be 31 & 5.2 & 12 - 16 & -0.57 & 0.24 & -0.33 & & y05 + mel 66 & 5.4 & 10.21 & -0.33 & & & & + be 32 & 6.1 & 11.35 & -0.29 & & & & br08 + be 25 & 6.1 & 18.20 & -0.20 & & & & + ngc 188 & 6.3 & 9.35 & -0.12 & & & & ho90 + saurer 1 & 6.6 & 19.3 & -0.38 & 0.47 & 0.09 & & pa08 + cr 261 & 8.4 & 7.52 & -0.03 ; 0.13 & -0.12 & -0.15 & 8.64 & c05 ; s08 + be 17 & 10.0 & 11.2 & -0.10 & 0.00 & -0.10 & 8.83 & f05 + ngc 6791 & 10.9 & 8.12 & 0.39 & -0.31 & 0.16 & 8.95 & c07 +
|
planetary nebulae ( pne ) derive from the evolution of @xmath018 m@xmath1 mass stars , corresponding to a wide range of progenitor ages , thus are essential probes of the chemical evolution of galaxies , and indispensable to constrain the results from chemical models .
we use an extended and homogeneous data set of galactic pne to study the metallicity gradients and the galactic structure and evolution .
the most up - to - date abundances , distances ( calibrated with magellanic cloud pne ) , and other parameters have been employed , together with a novel homogeneous morphological classification , to characterize the different pn populations . we confirm that morphological classes have a strong correlation with pn peimbert s type , and also with their distribution on the galactic landscape .
we studied the @xmath2-element distribution within the galactic disk , and found that the best selected disk population ( i.e. , excluding bulge and halo component ) , together with the most reliable pn distance scale yields to a radial oxygen gradient of @xmath3log(o / h)/@xmath3r@[email protected] dex kpc@xmath6 for the whole disk sample , and of @xmath3log(o / h)/@xmath3r@xmath4= [email protected] , [email protected] , and [email protected] dex kpc@xmath6 respectively for type i , ii , and iii pne , i.e. , for high- , intermediate- , and low - mass progenitors .
neon gradients for the same pn types confirm the trend .
accurate statistical analysis show moderately high uncertainties in the slopes , but also confirm the trend of steeper gradient for pne with more massive progenitors , indicating a possible steepening with time of the galactic disk metallicity gradient for what the @xmath2-elements are concerned .
we found that the metallicity gradients are almost independent on the distance scale model used , as long as these scales are equally well calibrated with the magellanic clouds .
the pn metallicity gradients presented here are consistent with the local metallicity distribution ; furthermore , oxygen gradients determined with young and intermediate age pne show good consistency with oxygen gradients derived respectively from other young ( ob stars , h ii regions ) and intermediate ( open cluster ) galactic populations .
we also extend the galactic metallicity gradient comparison by revisiting the open cluster [ fe / h ] data from high resolution spectroscopy .
the analysis suggests that they could be compliant with the same general picture of a steepening of gradient with time .
|
the spectral energy distribution ( sed ) of a galaxy contains a wealth of information regarding its star formation history , dust content , and chemical abundance pattern .
these properties provide essential clues to the physical processes governing the formation and evolution of galaxies from high redshift to the present .
it is therefore highly desirable to have a robust method for extracting the physical properties of galaxies from their seds .
the process of translating observed seds into physical properties is , unfortunately , very challenging because it requires 1 ) an accurate understanding of all phases of stellar evolution , 2 ) well calibrated stellar spectral libraries for converting stellar evolution calculations into measurable fluxes , 3 ) an initial mass function ( imf ) , specifying the weight given to each stellar mass , 4 ) detailed knowledge of the star dust geometry in conjunction with an appropriate extinction curve ; i.e. , knowledge of the physical conditions of the interstellar medium ( ism ) .
each of these requirements depend on chemical composition , further compounding the problem . combining these ingredients in order to predict the spectrum of a galaxy
is known as stellar population synthesis ( sps ) , and has an extensive history ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in the past two decades
enormous progress has been made on each of the requirements mentioned above . yet
, substantial uncertainties remain .
a non exhaustive list includes 1 ) the treatment of the core convective boundary in main sequence stars ( i.e. , convective core overshooting ) ; 2 ) metallicity dependent mass loss along the red giant branch ( rgb ) and , relatedly , the morphology of the horizontal branch ( hb ) ; 3 ) the treatment of the thermally pulsating asymptotic giant branch ( tp agb ) phase ; 4 ) the spectral libraries , especially for m giants , tp agb stars , and stars at non solar metallicities ( e.g. , * ? ? ?
* ) ; 4 ) blue straggler ( bs ) stars and their ubiquity ( e.g. , * ? ? ?
* ; * ? ? ?
* ) ; 5 ) the effects of binarism , which might be especially relevant for massive star evolution ( e.g. , * ? ? ?
* ) and bs stars ; 6 ) the importance of rotation on massive star evolution ( e.g. , * ? ? ?
* ) ; 7 ) non solar abundance ratios , which effect not only the stellar evolution calculations but also the stellar spectra ( e.g. , * ? ? ?
* ) ; 8) the unknown dependence of the imf on ism properties such as metallicity and pressure .
these uncertainties can in many cases dramatically impact the ability to convert observables into physical properties and vice versa ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
thankfully , there exists a wide array of data capable of constraining sps models . by far the most common type of object used for comparison
is the star cluster . the approximately uniform age and metallicity of the stars within star clusters and the lack of internal reddening ( except perhaps for very young clusters ) affords direct comparison between them and the most basic ingredients in sps the stellar evolution calculations and stellar spectral libraries .
comparisons to star clusters are either made in cmd space or in an integrated sense ( i.e. , the light from all the cluster stars are added together ) .
the former technique provides a sensitive probe of the main sequence including the turn - off point , sub giant and red giant branches ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , while the latter technique provides a more robust measure of the rarer brightest stars that dominate the integrated light from the cluster .
it is the latter method of comparison that is more relevant if the goal of sps modeling is to understand the integrated light from galaxies .
for this reason , the integrated colors , surface brightness fluctuations ( sbfs ) , and spectra of star clusters have been used extensively to compare and calibrate sps models ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there are two significant sources of concern when attempting to calibrate sps models with star cluster data .
the first concern is that the brightest stars , which dominate the integrated light , are rare , and thus stochastic effects must be carefully modeled .
recently , this issue has been addressed observationally by stacking clusters in bins of age in order to synthesize ` superclusters ' that are relatively unaffected by stochastic effects @xcite .
the second concern is less tractable , even in principle , and arises from the fact that the primary sources of star clusters for which ages and metallicities can be reliably estimated are the milky way ( mw ) , m31 , and the large and small magellanic clouds ( lmc and smc , respectively ) .
this fact imposes severe restrictions on the region of the age metallicity parameter space that can be constrained with star cluster data .
for example , one of the most important regions of this parameter space for studying galaxies old and metal rich contains very few star clusters ( except in the bulge of the mw , although high extinction diminishes the utility of star clusters there ) . in local group star clusters , a bin in age contains clusters of a relatively narrow range in metallicity owing to the simple fact that the metallicity of galaxies tends to increase with age . because of these issues , entire galaxies are often also used to assess the accuracy of sps models ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
galaxies are not subject to the two concerns mentioned above , but using them to constrain sps models can be very difficult because 1 ) galaxies contain stars of a range of ages and metallicities ; and 2 ) starlight in galaxies is attenuated by interstellar dust .
nevertheless , if subsamples of galaxies are carefully constructed with known and regular properties , then galaxies can be used to assess the reliability of sps models .
the present work seeks to provide a comprehensive comparison between sps models and observed star clusters and galaxies .
such a comparison is timely because of an abundance of new , high quality data , including near ir photometry and sbfs of star clusters in the lmc and smc , uv photometry of clusters in m31 and m87 , cmds and integrated spectra of clusters in the mw , and optical and near ir photometry and spectral indices for large samples of low redshift galaxies .
we will consider not only our own sps model @xcite but also the commonly used models of ( * ? ? ?
* bc03 ) and ( * ? ? ?
such a detailed comparison between these popular models has not been undertaken until now , and will thus provide a much needed evaluation of the relative strengths and weaknesses of these models .
in addition , the flexible nature of our own model will be exploited to explore the extent to which various uncertain phases of stellar evolution , including thermally pulsating asymptotic giant branch stars ( tp agb ) , blue hb stars , and post agb stars , can be constrained by the data .
this task is critical if we are to have confidence in the derived physical properties of galaxies , such as stellar masses and star formation rates .
we proceed as follows .
@xmath6[s : data ] contains a detailed description of the star cluster and galaxy data used to constrain the models , which are themselves described in @xmath6[s : model ] . in @xmath6[s : calib ] we present an extensive comparison between models and data .
a discussion and summary are provided in @xmath6[s : disc ] and @xmath6[s : sum ] , respectively .
the zero points of the star cluster ubvrijhk magnitudes are in the vega system ; all other magnitudes are quoted in the @xmath7 system @xcite . where necessary
, we adopt a flat @xmath8cdm cosmology with @xmath9 .
this section describes in detail the sources and treatment of data both for star clusters ( @xmath6[s : scd ] ) and galaxies ( @xmath6[s : gal ] ) .
the calibration of sps models against data is essential because of the many uncertain model ingredients , including late stages of stellar evolution and stellar spectral libraries .
as mentioned in the introduction , the star cluster is the ideal observational datum for sps calibration because it is , at least approximately , a coeval set of stars at a single metallicity . in order to provide comprehensive constraints on sps models
, one would like to have data on star clusters spanning a wide range in age , metallicity , and wavelength .
unfortunately , owing to the details of the chemical history of the mw and its satellites , the assembly of such a dataset from nearby galaxies , including our own , is challenging . in the present work attention
is focused on data from the milky way ( mw ) , m31 , and the magellanic clouds ( mcs ) .
the first two galaxies are sources primarily of old , metal poor clusters , while the mcs provide clusters spanning a wider range of ages and metallicities . before describing the datasets in detail ,
it is worth emphasizing those areas of parameter space that are not well sampled and yet are important for interpreting the integrated light from galaxies .
massive star clusters ( @xmath10 ) at solar or super solar metallicities are not common in the local group .
less massive star clusters of such metallicities are more common , but stochastic effects complicates model comparison .
the star clusters ngc 6791 , ngc 188 , and the bulge star clusters ngc 6553 and ngc 6528 are notable exceptions in that they are relatively old and metal rich .
these clusters will be used herein for calibration .
data on mw star clusters are taken primarily from the globular cluster catalog of @xcite .
this catalog provides @xmath11 photometry , [ fe / h ] metallicities , and @xmath12 reddening values for 150 clusters .
the photometry is corrected for galactic extinction using the reddening values provided in the catalog in conjunction with a @xcite extinction curve .
the @xcite catalog is complemented with @xmath13 photometry from the two micron all sky survey ( 2mass ; * ? ? ?
* ) for 106 clusters as compiled by @xcite . for most of the comparisons between models and data the mw sample
is restricted to clusters with @xmath14 in order to reduce the uncertainties associated with de reddening . when presenting @xmath2 and @xmath1 colors , all clusters with @xmath15 are included since reddening corrections
are smaller in the near ir .
@xcite have measured integrated , flux calibrated , high signal to noise spectra of 40 mw star clusters spanning a range of metallicities and hb morphologies .
the spectra span the wavelength range @xmath16 at a fwhm resolution of 3.1 and sampling of 1.0 .
these spectra have been de - reddened using the @xcite extinction curve with @xmath12 values tabulated by schiavon et al .
metallicities are also provided by schiavon et al . and will be used herein , except for 19 clusters that have updated [ fe / h ] measurements from @xcite .
photometry of star clusters in m31 is available in the revised bologna catalog ( rbc v3.5 ; * ? ? ?
available photometry includes @xmath11 from a variety of sources , @xmath13 from 2mass , and new ultraviolet observations from the _ galaxy evolution explorer _ ( _ galex _ ; * ? ? ?
* ) presented in @xcite .
spectroscopic metallicities of a subset of m31 star clusters are taken from @xcite and @xcite .
galactic @xmath12 reddening estimates are derived from @xcite via the utilities in the ` kcorrect v4.1.4 ` software package @xcite .
photometry is corrected for galactic extinction via @xmath12 and the extinction curve of @xcite .
in addition to galactic reddening , the colors of m31 star clusters will also be reddened by dust within m31 .
unfortunately it is not straightforward to make such a correction , and therefore no additional correction will be applied . in later sections attention
will be focused on the near ir colors of m31 clusters , where dust corrections should be small .
the rbc contains a number of quality flags .
clusters are selected to be definite globular clusters ( classification flag = 1 ) , with a definite confirmation ( confirmation flag = 1 ) , and to not be in the young cluster catalog of @xcite .
this sample is further restricted to clusters with @xmath14 to reduce the uncertainties associated with galactic reddening . the final cluster catalog ,
which includes the aforementioned quality cuts and metallicity measurements , contains 22 clusters , 20 of which include _ galex nuv _ photometry , and 12 of which include _ galex fuv _ photometry .
data on star clusters in the mcs come principally from two sources .
@xcite presented a carefully constructed catalog of 2mass @xmath13 photometry for a sample of 75 star clusters . in @xcite
this near - ir photometry is combined with @xmath17 photometry from the literature .
these authors construct a ` test sample ' of star clusters where high quality data are available to estimate main sequence turn - off ages from cmds and metallicities from either the cmds or individual stellar spectroscopy .
these high quality data are then binned by age to produce average colors and metallicities as a function of cluster age . in the youngest age bin
a less direct method is used for determining cluster ages in the @xcite sample .
the ` @xmath18-parameter ' , introduced by @xcite , provides a simple empirical relation between cluster age and a combination of integrated @xmath19 and @xmath20 colors .
high quality age estimates from hst observations from @xcite were used by @xcite to update the @xmath18-parameter age calibration in order to provide age estimates for the youngest clusters in their sample . the pessev et al .
calibration is substantially different from the commonly used relation of @xcite .
it is important to keep in mind that since ages are based on cmd fitting , they will depend on the stellar evolution calculations used to convert main sequence turn off points into ages . in particular
, the treatment of convective overshooting has a strong impact on the age main sequence turn - off point relation for young and intermediate ages ( @xmath21 yrs ) .
this sensitivity arises because overshooting increases the amount of fuel available for core hydrogen burning .
we will return to these issues when these data are compared to models .
the second significant source of data on mc star clusters comes from @xcite .
these authors analyze 2mass data to produce not only integrated near - ir colors but also surface brightness fluctuations in the 2mass @xmath13 filters .
clusters are stacked according to the classification system devised by ( * ? ? ?
* swb ) , which closely tracks the s - parameter described above .
the relation between swb type , age and metallicity is adopted from @xcite .
we have independently verified that these ages agree with the ages inferred from the @xmath18-parameter calibration discussed in @xcite .
this sample has been augmented with @xmath22 colors @xcite , where @xmath23 magnitudes are from the denis survey @xcite and @xmath24 magnitudes are taken from @xcite .
we point out that the pessev et al . and gonzalez et al .
samples are not disjoint .
comparison of these two catalogs will provide a valuable cross check on the reliability of the data .
the novel feature of these two datasets is that the cluster data have been stacked in bins of age in order to reduce stochastic effects .
a number of authors have discussed the importance of stochastic effects arising from small numbers of very bright stars @xcite . both @xcite and @xcite provide limits on the minimum cluster mass that should be satisfied so that stochastic fluctuations in integrated colors are minimized .
@xcite demonstrate that the masses of their stacked clusters easily satisfy both constraints , ensuring that their stacked results can be robustly compared to sps models .
the masses of the stacked clusters in the @xcite sample are comparable to those in @xcite , and so their results should also be relatively unaffected by stochastic effects .
finally , we include for completeness the @xmath2 and @xmath1 colors of lmc clusters from @xcite , which have been used in many comparisons between observations and sps models .
these @xmath25 and @xmath26 magnitudes are converted from the california institute of technology ( cit ) photometric system to the 2mass system using the transformations provided by @xcite .
ages for these clusters were assigned based on the @xmath18-parameter calibration of @xcite .
@xcite presented a detailed comparison between their near - ir colors and those of @xcite , in some cases noting substantial differences in the sense that colors from persson et al .
were redder than the colors from pessev et al .
they attributed this difference to the fact that persson et al .
centered their photometry on the brightest part of the cluster in order to maximize the measured flux , while pessev et al .
center their apertures using a gaussian smoothing technique . in some cases
the persson et al .
centroids are significantly offset from the true dynamical center of the cluster because of off center ir
bright sources within the cluster .
for this reason , in addition to the fact that the data are stacked to reduce stochastic effects , the results of pessev et al . should be preferred to the older results of persson et al .
we will return to this issue in @xmath6[s : calib ] .
for the reasons mentioned in the previous section , star clusters are the ideal observational data to calibrate sps models .
there are however at least two reasons why it is useful to also consider sps constraints from galaxies : 1 ) galaxies tend to be of higher metallicity than old star clusters , and so sps model predictions for old @xmath27 populations can be assessed with galaxies composed of old stars ; 2 ) galaxies contain many more stars than star clusters , and so stochastic effects are never a source of concern when comparing models to galaxy data .
these two benefits of using galaxies to calibrate sps models are exploited herein by considering two classes of galaxies .
the first class consists of bright red sequence , ` quiescent ' galaxies , which contain old metal rich stars , and will therefore be useful for calibrating old metal
rich sps model predictions . the second class are called ` k+a ' , or ` post starburst ' galaxies , as their spectra are dominated by a mixture of a and k type stars and show no signs of ongoing star formation ( e.g. , * ? ? ?
* ; * ? ? ?
such galaxies are presumed to have had their star formation abruptly truncated .
the presence of a stars in their spectra indicates that the galaxy contains a significant population of @xmath28 gyr old stars ( e.g. , * ? ? ?
. these galaxies will provide valuable constraints on the importance of tp agb stars , as the prominence of this stellar phase peaks for ages @xmath29 gyr @xcite .
since stars spend a small amount of time in this phase , observational constraints from star clusters can be difficult to interpret because of stochastic variations . as mentioned above ,
a stacking technique is employed when comparing to star clusters so that stochastic effects are minimized .
nonetheless , it will be useful to derive independent constraints from galaxies where stochastic effects are known to be negligible . as mentioned in the introduction , the use of galaxies to assess the accuracy of sps models is complicated by at least two effects : the attenuation of starlight by dust and the fact that galaxies contain stars of a range of ages and metallicities .
these complications are addressed where relevant in @xmath6[s : calib ] .
the source of all galaxy positions , optical photometry , and redshifts is the sloan digital sky survey ( sdss ; * ? ? ?
* ) , as made available through the hybrid nyu value added galaxy catalog ( vagc ; * ? ? ?
the sdss photometry is k corrected and corrected for galactic extinction via the ` kcorrect v4.1.4 ` software package @xcite .
the photometry is k corrected to @xmath30 the median redshift of the sample in order to minimize the magnitude of the k correction .
spectroscopic indices are provided by the mpa / jhu database . near ir photometry in @xmath31 filters
is provided by the ukirt infrared deep sky survey ( ukidss ) .
the ukidss project is defined in @xcite .
ukidss uses the ukirt wide field camera ( wfcam ; * ? ? ?
the photometric system is described in @xcite , and the calibration is described in @xcite .
the pipeline processing and science archive are described in irwin et al ( 2009 , in prep ) and @xcite .
we make use of data release 3 , which is publically available .
ukidss photometry is matched to the sdss with a search radius of 5 .
petrosian magnitudes are used for the ukidss photometry .
we do not consider colors that are combinations of sdss and ukidss photometry , so it is not essential that the photometric apertures be precisely matched .
the ukidss survey goes much deeper than 2mass , and is therefore more desirable .
for example , typical 2mass photometric errors for galaxies at @xmath32 are 0.2 mags , compared to @xmath33 for ukidss photometry .
galactic extinction is corrected using @xmath12 reddening estimates @xcite in conjunction with a @xcite extinction curve .
corrections for the ukidss photometry are estimated from the ` kcorrect v4.1.4 ` software package using the 2mass @xmath13 filters since those , and not ukidss filters , are available in the code .
the filter transmission curves for 2mass and ukidss are very similar and so this approximation is justified .
three galaxy samples are considered .
the first are k+a galaxies as identified by @xcite .
this sample contains 1194 galaxies . of these , only 186 have ukidss photometry . when considering near
ir colors of the k+a sample we use only these 186 galaxies , while when sdss colors are considered we use the full sample .
the second sample consists of bright red sequence ( ` quiescent ' ) galaxies .
this sample is defined by @xmath34 , where the superscript indicates that the magnitude has been k corrected to @xmath30 , and @xmath35 which divides galaxies by the valley visible in the color magnitude diagram and effectively isolates red sequence galaxies . applying these cuts and requiring that galaxies have both sdss and ukidss photometry results in a sample of 5340 galaxies at @xmath36 .
the third sample is derived from the spectroscopic indices catalog .
galaxies are selected with @xmath37 and with a median spectroscopic signal to noise of @xmath38 . from this catalog
two classes of galaxies are defined : those with velocity dispersions measured from the stellar continuum of @xmath39 and those with @xmath40 . these samples contain 13214 and 6774 galaxies , respectively .
sps model construction closely follows that of @xcite , to which the reader is referred for details . in brief
, this sps code ( referred to below as fsps , for ` flexible ' sps ) combines stellar evolution calculations with stellar spectral libraries to produce simple stellar populations ( ssps ) , which describe the evolution in time of the spectrum of a mono metallic , coeval set of stars .
the initial stellar mass function ( imf ) specifies the weights given to stars of various masses .
the imf of @xcite is adopted throughout .
fsps also contains a number of phenomenological dust attenuation prescriptions and produces spectra for composite stellar populations .
our sps code is open source and publicly available .
the novel feature of this sps model is that the essential aspects of sps are integrated in a flexible way so that the user may choose , for example , their preferred set of isochrones and stellar spectral libraries .
in addition , the user may alter the weights given to and physical parameters of various advanced stages of stellar evolution ( see below ) , and may specify the imf of their choosing .
the details of these various model choices are now described .
we make use of both the padova @xcite and basti @xcite stellar evolution calculations , both in the versions available as of november 2009 .
these two groups follow stellar evolution from the main sequence through the thermally pulsating asymptotic giant branch ( tp agb ) phase , although the treatment of the tp
agb phase is very different between the two groups .
the padova calculations exist for metallicities in the range @xmath41 , for ages @xmath42 yrs , and for initial masses @xmath43 .
basti outputs cover a somewhat narrow range in ages , @xmath44 yrs , and initial masses @xmath45 but include higher metallicities : @xmath46 .
the basti database provides models for both solar scaled and @xmath47enhanced chemical compositions , for two assumptions regarding convective core overshooting , and for two stellar mass - loss rates .
we use the solar scaled , no overshooting , and @xmath48 mass loss rate parameter model . for both lower and higher masses
the padova calculations are stitched onto the basti grid .
the padova and basti models are compared in figure [ fig : cmd ] . in this figure isochrones
are compared at several ages and metallicities , and in addition the location of the tp agb onset is shown . in order to follow stellar evolution from the end of the tp
agb phase through the post agb phase , the above evolutionary tracks have been supplemented with the post agb models of @xcite .
the contribution of post agb stars to the integrated sed is highly uncertain for a number of reasons , including a lack of calibrating observations and the unknown duration of heavy enshroudment as post agb stars leave their m giant phase .
recent observations of m32 have found that post agb stars are much less frequent than standard models predict @xcite .
nonetheless , the inclusion of post agb tracks may be important for the interpretation of the ultraviolet spectra of old systems ( e.g. , * ? ? ?
the fsps code contains a number of parameters that control the weights given to various uncertain phases of stellar evolution .
for example , the bolometric luminosity and effective temperature of tp agb stars can be arbitrarily modified .
these modifications are applied as overall shifts in log(@xmath49 ) and log(@xmath50 ) for the entire tp
agb phase , and are parameterized by the variables @xmath51 and @xmath52 ( see discussion in * ? ? ?
in addition , an extended blue hb can be included , as can a population of bs stars .
a blue hb population is parameterized as the fraction , @xmath53 of hb stars that are in a blue / extended component .
such a flexible treatment is motivated by the great variety of observed hb morphologies in mw clusters , even at a given metallicity @xcite .
these stars are uniformly populated from the red clump to log@xmath54 .
bs stars are uniformly populated from one half magnitude above the main sequence turn off to two and a half magnitudes brighter than the turn off , along the zero age main sequence @xcite . the weight given to bs stars
is specified as the number of bs stars per unit hb star , @xmath55 .
the weight given to the post agb phase can also be arbitrarily modified .
the effect of these uncertain aspects of stellar evolution will be investigated in later sections .
stellar spectral libraries are required to convert the outputs of stellar evolution calculations surface gravities , @xmath56 , effective temperatures , @xmath50 , and metallicities , @xmath57 , into observable spectral energy distributions .
two spectral libraries are considered herein .
the first is the empirically calibrated theoretical basel3.1 library @xcite , supplemented with empirical , average tp agb spectra from the library of @xcite .
the basel library was constructed from the model atmospheres of kurucz ( 1995 , private communication to r. buser ) , @xcite , @xcite , and @xcite . in the tp agb library , spectra are binned both by near
ir color and by whether the star is carbon rich or oxygen rich .
stars hotter than @xmath58 are assumed to be pure blackbodies .
this combined library covers the full wavelength range ( ultraviolet through infrared ) at a sampling of 10 in the ultraviolet and 20 in the optical .
the second spectral library considered herein is derived from the empirical miles library @xcite .
this atlas contains spectra of @xmath59 stars covering the wavelength range @xmath60 at a sampling of 0.9 per pixel ( fwhm@xmath61 ) .
derived stellar parameters are presented in @xcite .
integrating empirical libraries into sps models is a non
trivial task because the coverage in the stellar parameters @xmath56 , @xmath50 , and @xmath57 is very inhomogeneous .
in fact many regions of this parameter space are not covered by empirical libraries at all , especially at sub solar and super solar metallicities and for spectral types earlier than a. an additional problem with the miles library is that the spectra have been arbitrarily normalized to unity at 5350 .
this is particularly problematic because the wavelength range is not broad enough to compare the integral of the flux to the bolometric luminosity for a star with a given @xmath56 , @xmath62 , and @xmath57 .
therefore , an additional library must be used to provide the absolute flux scale .
the basel library is used for this purpose .
the basel library is also used to provide the relative trends of spectra with stellar parameters for regions of parameter space where the relative trends can not be self consistently determined from miles .
the miles library is binned and , when necessary , linearly interpolated onto the kurucz / basel log@xmath63log@xmath64 grid .
this interpolated library is then interfaced with fsps in a manner identical to the basel library .
five metallicity bins are defined so that roughly similar numbers of stars occupy each bin .
the bin boundaries are : log@xmath65 $ ] .
the resulting average metallicities within these bins are log@xmath66 $ ] . at metallicities less than solar ,
the miles library stars are @xmath47enhanced @xcite . at [ fe / h]@xmath67
the library is on average enhanced in mg by + 0.4 dex .
this will be relevant in our comparison to the spectra of mw star clusters in later sections .
there are several important caveats to these libraries that the reader should keep in mind . in the empirical tp agb library , the metallicities of the stars are _ unknown_. the authors advocate assuming that they are all solar metallicity since a significant fraction are located in the solar neighborhood ( although many are also located in the mcs ) . in order to use the library at non solar metallicities , the authors suggest the use of the theoretical metallicity
dependent color temperature relations in @xcite in order to convert the observed @xmath68 colors into temperatures at other metallicities .
@xcite found that the original theoretical atmosphere calculations used in the basel library could not simultaneously match observed @xmath69 color temperature relations for individual stars and cmds of globular clusters .
the reason for this was thought to be inadequacies in the stellar evolution calculations used in the cmd comparisons .
because of this problem , @xcite presented two normalizations of the basel library , one which matched the color temperature relations , and the other which matched the globular cluster data .
these two normalizations are referred to as the ` wlbc ' and padova2000 libraries , since the padova evolutionary calculations were used in the globular cluster comparisons .
the ` wlbc ' normalization will be used herein because this calibration is not tied to any particular isochrone set . in summary
, we have in hand a flexible sps model from which we can choose two sets of isochrones ( padova or basti ) and two sets of spectral libraries ( basel or miles ) .
in addition , there are a number of free parameters characterizing various uncertain aspects of advanced stages of stellar evolution .
these various options will be exploited in later sections .
for the padova isochrones , the tp
agb phase has been modified in order to generate acceptable agreement with the data .
specifically , the tp
agb phase has been calibrated against both mc star clusters and post starburst galaxies ( @xmath6[s : cmc ] and @xmath6[s : cpsg ] , respectively ) .
these calibrations result in an adjustment of the tp agb parameters @xmath51 and @xmath52 ( quantifying overall shifts in log@xmath70 and log@xmath64 ; see @xmath6[s : fsps ] ) as follows : @xmath71 -{\rm max}\{{\rm min}\{0.4,-{\rm log}(z / z_\sol)\},0.2\ } , & t>9.1 , \end{array } \right.\ ] ] @xmath72 + 0.1-{\rm min}\{(t-9.1)/1.5,0.2\ } , & t>9.1 , \end{array } \right.\ ] ] where @xmath73 is the age of the population measured in log(yrs ) .
we emphasize that the values of @xmath51 and @xmath52 chosen to fit the data are _ not unique _ , and that , unless explicitly stated in the above definitions , they are independent of metallicity . throughout , reference to
the ` modified padova model ' refers to these modifications .
we now describe the salient aspects of the bc03 and m05 sps models .
the bc03 model utilizes the padova isochrones circa 1994 @xcite , supplemented with the @xcite models for tp agb stars and the @xcite models for post agb stars .
bc03 use the models of @xcite to differentiate between carbon rich and oxygen rich tp - agb stars .
there are no blue / extended hb stars in the bc03 models .
their models are constructed with both the theoretical basel library ( described in the previous section ) and the empirical stelib library @xcite .
spectra for carbon stars are taken from model atmosphere calculations @xcite .
the empirical stelib library contains optical spectra for 249 stars at a resolving power of @xmath74 ( fwhm@xmath75 ) .
the effective temperatures in the stelib library are not reliable and so bc03 adopt the color temperature scale from the basel library in order to relate the stellar evolutionary outputs to the empirical library .
there are very few stars at non solar metallicities and so the reliability of models constructed with this library at non solar metallicities is questionable .
indeed , @xcite compared several different sps models and found that the bc03 model constructed with stelib produces unreliable results at non solar metallicities . we make use of the bc03 model that was constructed with the stelib library for the optical spectra . for both ultraviolet and near ir predictions
, the stelib library has been extended with the basel library .
the m05 model is built from the following ingredients .
stellar evolution calculations through the main sequence turnoff are adopted from cassisi et al . @xcite .
the fuel consumption theorem @xcite is then used to incorporate advanced stages of stellar evolution through the tp
agb phase .
the evolutionary calculations are converted to observables with the theoretical basel spectral library for all stars except tp agb stars . for these stars
the empirical library of @xcite is used ( this is the same library used in our fsps code ) .
model magnitudes and spectral indices are computed self consistently from the model spectra .
comparisons between models in later sections are therefore not plagued by potentially different filter transmission curves or index definitions .
we now turn to a comparison between observations and sps models . in @xmath6[s : cmc][s : cmwm31 ] we focus on star clusters , both in the mcs , where constraints are strong for intermediate age , sub solar metallicities , and in the mw and m31 , where constraints are strong for old metal poor systems .
we then turn to galaxies in @xmath6[s : cqg][s : cpsg ] .
two types of galaxies will be considered : massive red sequence galaxies and post starburst galaxies .
the former class will provide a qualitative assessment of the accuracy the sps models in the solar to super solar metallicity , old age regime .
the post starburst galaxies will provide an assessment of the models in the approximately solar metallicity , intermediate age regime .
as discussed in previous sections , it is challenging to use galaxies to assess the accuracy of sps models because galaxies are composed of stars with a range of ages and metallicities and dust reddening can be important .
these complications are discussed extensively in @xmath6[s : cqg][s : cpsg ] .
figure [ fig : mccol1 ] shows the colors of mc star clusters as a function of cluster age . as discussed in @xmath6[s : mcdat ] , individual clusters are binned in age to produce ` superclusters ' that are relatively immune to stochastic fluctuations @xcite . the data from @xcite are also included , and have not been binned in age .
these data are of inferior quality ( see @xmath6[s : mcdat ] ) and are only included for comparison because this data is frequently used to constrain sps models ( e.g. , * ? ? ?
* ) . a variety of sps models
are compared to the mc star cluster data in figure [ fig : mccol1 ] . from our fsps code
we have generated ssp color evolution for both the padova and basti isochrones . in this figure
we make use of the modified padova isochrones ( see @xmath6[s : mod ] ) .
the models of bc03 and m05 are also included .
all model predictions reflect the fact that the average metallicity of the data varies with age . that is , we have interpolated model predictions so that they vary with metallicity in as in the data , where the age
metallicity relation in the mc data is log(@xmath73/yrs)@xmath76 , log@xmath77 @xcite .
the m05 models are significantly redder than the data at intermediate ages , owing to the different tp agb treatment in m05 ( see also * ? ? ?
* who reach similar conclusions ) .
the m05 colors also have an age dependence that does not agree with the data .
the mc cluster ages were estimated via cmd fitting using isochrones that include convective overshooting @xcite , whereas the m05 models do not include overshooting , and so one might expect some mismatch in age between the m05 models and data . however , the differences induced by the treatment of overshooting on the mc ages is @xmath78 dex at ages @xmath79 yrs , whereas the discrepancy in the age dependence between m05 and the data is @xmath80 dex in age .
we therefore regard most of this discrepancy as a shortcoming of the m05 model , with only a fraction being due to different definitions of cluster age .
the persson et al .
data are substantially redder than the newer datasets from pessev et al
. and gonzalez lopezlira et al . as discussed in @xcite , this is probably due to the fact that the persson photometry was probably centered on the brightest star in the cluster , which are often carbon stars , and in any case are very red , and so the photometry will be biased red .
the difference between persson s data and the 2mass based data highlights the care that must be taken to construct datasets that can be meaningfully used for sps model calibration .
the m05 models were calibrated against color color plots of mc star clusters using the persson et al .
data and very wide age bins .
stochastic effects will tend to smear star clusters along an underling color color relation because if a cluster is scattered redder in one color , it will be redder in another color as well .
the m05 model calibration is therefore sensitive to stochastic effects , and is in addition plagued by the systematics in the persson data .
moreover , the wide age bins used in m05 explains why the discrepancies in age seen for the m05 model in figure [ fig : mccol1 ] were not seen in their own calibrations .
the bc03 model and basti isochrones both fare
well except for @xmath81 and @xmath1 colors at intermediate ages ( log@xmath82 ) , for which models are too blue by @xmath83 mags compared to the data .
this disagreement is probably due to the simplistic treatment of tp agb stars in these models
. the basti isochrones also produce redder colors at young ages compared to the padova isochrones , which is probably due a lack of convective core overshooting in the basti isochrones adopted herein ( see figure [ fig : cmd ] ) .
the data do not definitively favor one model over another at log@xmath84 .
the disagreement in @xmath22 between the models and data at intermediate ages is also noteworthy .
the models perform better at old ages , although it is unclear if this is because of the old ages or the lower metallicities of the old star clusters .
we will return to issues with @xmath22 colors in later sections .
fsps model predictions using the padova isochrones are explored further in figure [ fig : mccol2 ] . in this figure
we compare the unmodified padova isochrones to the modified model , and we also show the effect of a variation in metallicity of @xmath85 dex about the modified model . for ages and colors this variation with metallicity does not bracket the modified model because some metallicity color relations in the models are not monotonic and in fact some colors are reddest at approximately the average metallicity of the lmc ( log@xmath86 ) .
it is clear that use of the unmodified padova models results in substantial disagreement with observed star clusters .
this disagreement is driven to a significant extent by the large population of luminous carbon stars predicted in the padova calculations . in our modifications to the padova models we have substantially lowered the bolometric luminosity along the tp
agb phase , which has the effect of down weighting the importance of all tp agb stars , including carbon stars .
it is worth discussing in more detail the substantial disagreement between the unmodified padova calculations and the star cluster data . the padova group provided comparisons between their model and observations in @xcite for @xmath2 and @xmath1 colors , finding acceptable agreement .
the explanation for this difference is twofold .
first , the data used in their comparison is the old persson data , which suffers from the issues discussed above .
second , and perhaps as important , the conversion from @xmath49 and @xmath50 to seds was achieved with the theoretical atmosphere models of @xcite coupled to dust radiative transfer models from @xcite . aside from the fact that the dust models were never demonstrated to fit the @xmath2 colors of agb stars , the @xcite carbon star models are substantially bluer ( by @xmath87mag in @xmath2 ) than observed carbon stars ( r. gautschy , private communication ) .
we believe that our transformation to seds is more robust since we use observed tp
agb spectra .
our qualitative conclusions are echoed by @xcite who compared the population of o rich and carbon stars in the leo ii dwarf galaxy to the padova models and found that the models predict substantially more , and substantially more luminous o rich stars than are seen in that galaxy .
finally , the padova group uses the carbon star luminosity function in the mcs as a constraint on their tp agb model , but this observable can only provide valuable insight if the star formation histories ( sfhs ) of the mcs are known to high precision in the age range relevant for carbon star production ( @xmath88 yrs )
. figures [ fig : sbf1 ] and [ fig : sbf2 ] compare near ir sbfs between several fsps based models and mc data . in figure
[ fig : sbf1 ] the evolution of sbf magnitudes and sbf colors are compared to our fsps model with both the basti and padova isochrones .
we show results both for the unmodified and modified padova isochrones .
the basti isochrones provide an acceptable fit to the @xmath89band sbfs , but produce sbf colors that are somewhat too blue compared to the data .
this implies that the agb and tp agb temperatures are too hot in the basti calculations .
the unmodified padova results show similar discrepancies as seen in figure [ fig : mccol2 ] .
sbf magnitudes and colors are too bright and too red , which can be traced back to the over abundance of carbon stars and the over luminous tp agb evolutionary phase . as discussed earlier in @xmath6[s : mccol ] , these problems are rectified by altering the overall luminosity and temperature of stars in the tp
agb phase .
these modifications result in sbf magnitudes and colors that are in better accord with the data , although @xmath89band sbfs tend to be somewhat lower than the observations at intermediate ages . in figure
[ fig : sbf2 ] we compare models and data in @xmath89band sbfs versus integrated @xmath22 colors .
in addition to mc star cluster data we include data from elliptical galaxies @xcite .
it is clear that the modified padova isochrones span the range of data much better than the unmodified isochrones .
in addition , notice that there is a subsample of ellipticals that have sbfs fainter than predicted by the models .
this has a direct and potentially important impact on the inferred mass to light ratios of these systems and consequently on their inferred stellar masses .
sbfs in the red / near ir will correlate inversely with derived stellar masses , and therefore if sbfs are overestimated in the models ( as is the case here ) , then derived stellar masses will be underestimated .
the models shown in figure [ fig : sbf2 ] assume solar scaled chemical compositions , while many elliptical galaxies are @xmath47enhanced ( e.g. , * ? ? ?
* ; * ? ? ?
however , @xcite have shown that the level of @xmath47enhancement has almost no effect on the model predictions in the @xmath22 vs. @xmath90 sbf plane ( @xmath90 is similar to the @xmath91band ) , and therefore our conclusions should not be effected by these details .
sps model predictions for the metallicity dependent optical and near
ir colors of old populations are compared to observed star clusters in figure [ fig : lz ] .
model predictions are restricted to ages of 12.5 gyr , although results are not sensitive to this particular age .
fsps models with both the basti and modified padova isochrones are included , as are the predictions from m05 and bc03 .
comparisons to star clusters in both the mw and m31 are included .
for all colors except for @xmath1 the mw and m31 data agree ( see below ) , although the scatter in m31 cluster colors is larger .
this is probably due to the effects of internal reddening in m31 that has not been accounted for .
the models reproduce the trends in the observed colors farely well , although there are areas of noteworthy disagreement .
the @xmath92 and @xmath20 colors are somewhat too blue in the models at log@xmath93 .
the fsps model with the basti isochrones perform worst of the models considered .
we can check whether or not this general disagreement between models and data in @xmath94 filters is due to the adopted spectral library by considering the empirical miles stellar spectral library . due to the limited wavelength range of this library
, we can only construct the @xmath20 color .
the @xmath20 colors from the miles empirical library are almost indistinguishable from the colors using the basel library , implying that the stellar atmospheres are not the source of the discrepancy .
a similar test was performed in bc03 , where they compare the basel , stelib , and @xcite libraries and find negligible differences in @xmath20 at @xmath95 .
a second possible explanation for the differences between models and observations in the @xmath92 and @xmath20 colors is the @xmath94 filter transmission curves .
the @xmath96 filter used for these data actually consist of two slightly different filters ( see the discussion in * ? ? ?
it is difficult to determine which filter was used in each observation .
the difference between these two filters varies from 0.02 to 0.04 mags .
the @xmath92 color is constructed by adding the @xmath19 and @xmath20 colors , so this color will also be subject to this source of uncertainty .
the offsets between models and data is of order 0.05 mags the same order as the differences induced by the two @xmath96 filters and so we can not make stronger comparisons between models and data until more reliable optical photometry is available for mw and m31 star clusters .
the morphology of the hb can not explain the observed discrepancy because in all sps models shown in figure [ fig : lz ] adopt a red hb at log@xmath93 .
the hb therefore can not be modified to produce even redder model colors .
the model @xmath22 colors compare favorably with the data at log@xmath93 . at lower metallicities
the models diverge from the data , with the fsps model using basti isochrones performing worst .
all models fail at predicting the observed @xmath2 and @xmath1 colors , both at low and high metallicities for the latter and at higher metallicities for the former .
recall that the near ir data comes from a recent analysis of 2mass data by @xcite .
our results echo those of @xcite who compared mw star cluster @xmath1 colors against several sps models .
in addition to the m05 model , they also compare to the models of @xcite and @xcite and find similar levels of disagreement with these other models .
@xcite speculate that the cause of the discrepancy could be either 1 ) inadequate calibration of the stellar libraries or 2 ) issues with the near ir filter transmission curves ( cohen et al .
used published predictions for johnson @xmath13 photometry from these models ) .
since we compute 2mass magnitudes self consistently from the model spectra , the second explanation can not be the source of the discrepancy .
the problem is not likely to be due to inadequacies in the theoretical stellar spectral libraries , since the colors of giants compare well with observed stars in the near
ir @xcite .
the temperature of the rgb need only be lowered by 200k in order to achieve better agreement in @xmath2 ( although a lower rgb temperature would only exacerbate the disagreement in @xmath1 ) . as explored in @xmath6[s : metalr ] , one need only change the rgb at the brightest @xmath97 mag to in order to produce substantial changes in @xmath2 colors . without definitively identifying the source of this problem ,
we are forced to simply reiterate the conclusions of cohen et al .
that near ir colors of old populations can not be used to constrain the metallicities or ages of either gcs or distant galaxies .
it should also be noted that the near ir photometry presented in cohen et al . , which is based on data from the 2mass survey , differs substantially from previous near ir photometry of mw clusters provided by m. aaronson et al .
( 1977 ; unpublished ) . the differences in @xmath1 are @xmath98 mags in the sense that the latest results are bluer than the older data . since previous models were calibrated against the older data ( e.g. , * ? ? ?
* ) , it is not surprising that the models are all redder than the cohen et al . data .
the most dramatic discrepancy in figure [ fig : lz ] is between the @xmath1 colors in mw and m31 star clusters .
the m31 clusters are approximately 0.2 mags redder in @xmath1 than the mw clusters .
internal reddening in m31 can not be the source of this discrepancy because reddening effects are quite small in @xmath1 , especially in comparison to the colors considered in figure [ fig : lz ] , which do not show such an offset .
@xcite have also presented @xmath13 photometry of old m31 star clusters .
their results are consistent with the m31 results from the rbc catalog shown in figure [ fig : lz ] , suggesting that the discrepancy in @xmath1 is not due to data reduction issues .
we can not provide a compelling explanation for this discrepancy , but note that without an explanation it is impossible to calibrate the old metal
poor models to better than @xmath99 mags in the near ir .
calibrating data at higher metallicities than shown in figure [ fig : lz ] is far more desirable from the standpoint of using sps to model galaxies since galaxies are generally of higher metallicity .
unfortunately , old metal
rich star clusters are rare .
classic metal
rich clusters include ngc 6791 , ngc 188 , ngc 6553 , and ngc 6528 ; the latter two belong to the bulge and thus are along sightlines of high extinction . in figure
[ fig : hz ] we compare multi color cmds of ngc 6791 and ngc 188 to our sps predictions . the consideration of multi color cmds provides a powerful constraint on sps models in part because the reddening value can not be arbitrarily varied to fit a color without potentially worsening the fit at another color .
ngc 188 and 6791 are open clusters in crowded fields and therefore cmd construction is challenging without radial velocities and/or proper motions to isolate true members .
this complication should be kept in mind when considering model comparisons below .
ngc 6791 is assumed to have an age of 8.5 gyr , @xmath100 , [ fe / h]@xmath101 , [ @xmath102/fe]@xmath103 , and @xmath104 @xcite . @xmath105 and @xmath106 photometry are provided by @xcite and @xcite , respectively .
the cluster parameters for ngc 188 are taken to be @xmath107 , [ fe / h]@xmath108 , [ @xmath102/fe]@xmath103 , @xmath109 , and an age of @xmath110 gyr @xcite .
@xmath11 photometry of this cluster is provided by @xcite ; near ir photometry is extracted directly from the 2mass point source catalog .
we use the basti isochrones for the following comparisons because those isochrones have been computed for log@xmath111 , which is comparable to the metallicity of ngc 6791 and is higher than the highest metallicities available from the padova isochrones .
our model cmds are compared to ngc 6791 and ngc 188 in figure [ fig : hz ] for the same reddening values , ages , and distance moduli as quoted above , but for slightly difference metallicities .
metallicities of log@xmath111 for ngc 6791 and log@xmath112 for ngc 188 are adopted based on the available isochrones .
the agreement between data and models is generally good , including the location of the main sequence turn off , the sub giant branch and rgb , and the position of the red hb .
the model fails to reproduce the low mass end of the main sequence in ngc 6791 in the sense that the model isochrone is on the blue side of the data in @xmath20 and @xmath2 and on the red side in @xmath22 .
of course , this disagreement is of little consequence for integrated properties since low mass stars are faint .
the disagreement at low masses is probably due to uncertainties associated with color@xmath50 relations for low mass stars at high metallicity @xcite .
it appears that the model rgb @xmath2 color becomes progressively bluer than the rgb of ngc 6791 at higher luminosities , although it is difficult to make firm statements owing to the paucity of data at the bright end .
notice that this discrepancy is in the same sense as was seen for integrated @xmath2 colors of metal poor mw and m31 clusters ( see @xmath6[s : mwopt ] ) .
the model cmds perform well when compared to ngc 188 .
the only obvious discrepancy is in the @xmath19 color of the rgb .
thankfully , this is of little consequence for integrated colors because the integrated @xmath19 color is determined primarily by the main sequence turn off point and sub giant branch , not the rgb ( see below ) . for the purpose of modeling entire galaxies , we are interested in the ability of sps models to accurately predict the _ integrated _ colors of star clusters .
it is therefore worth considering the extent to which constraints from resolved cmds can constrain integrated colors of clusters . in figure
[ fig : cmdf ] the integrated colors corresponding to the cmd of ngc 188 are explored .
integrated colors are computed by including all stars fainter than a given apparent @xmath113band magnitude .
it is clear from this figure that integrated blue colors such as @xmath19 and @xmath20 are well constrained even at relatively faint magnitudes in @xmath24 .
this occurs because integrated blue colors are determined primarily by the main sequence turn off point and sub giant branch .
in contrast , redder and near ir colors are highly sensitive to the brightest one magnitude ( @xmath114 ) for the simple reason that the rgb and agb dominate the near ir light .
the brightest one magnitude in @xmath24 contributes @xmath115% of the integrated @xmath113band light ( depending on whether or not one includes the red clump hb ) , and the @xmath22 and @xmath2 colors redden dramatically in this brightest magnitude . thus ,
unless the brightest , and rarest stars in clusters can be reliably identified and used for model calibration , model constraints in cmd space will not guarantee accurate model integrated colors .
we now turn to an analysis of the spectral indices of old populations over a range of metallicities .
observed indices are measured from the catalog of optical spectra for 40 mw star clusters from @xcite .
the indices are measured in the exact same way for both the observed spectra and the models , ensuring reliable comparison .
the fsps model spectra are at slightly higher resolution than the mw cluster spectra , so the model spectra have been broadened to match the resolution of the data .
the spectral resolution of the bc03 model is comparable to the observed spectra , and so the bc03 model spectra were not broadened .
figure [ fig : mwlick ] presents the spectral indices d@xmath4 , h@xmath5 , fe5270 , and mgb as a function of metallicity for the schiavon et al . sample and the fsps and bc03 sps models .
the fsps predictions utilize the miles empirical stellar library , while bc03 adopt the stelib empirical library . for these ages
the padova predictions are very similar to the basti isochrones except for h@xmath5 at low metallicity because the basti isochrones have a hotter hb .
this figure also demonstrates the importance of bs stars and an extended / blue hb by including models with @xmath116 and @xmath117 ( see @xmath6[s : fsps ] ) .
there are several important conclusions to draw from this figure .
first , the models are able to reproduce the observed metallicity dependence of the mgb and fe5270 indices over the full observed range . at low metallicities this success
is driven largely by the fact that the bulk of the metal poor stars in the empirical stellar library contains @xmath47enhanced stars @xcite . while the isochrones used in fsps are solar scaledenhanced stellar evolution models .
] , recent work has shown that for old populations the effects of @xmath47enhancement are strong on the stellar libraries and weak on the isochrones @xcite .
the default sps models ( i.e. , without bs or blue hb stars ) predict d@xmath4 in excess of observed values at solar metallicity , and the same models predict h@xmath5 strengths too weak compared to the observations at both very low and solar metallicity .
the disagreement at matching h@xmath5 at very low metallicity is a consequence of the well known fact that most low metallicity clusters have a blue hb .
indeed , the models do a better job of matching the observed h@xmath5 strengths at low metallicity for @xmath118 .
the failure of the models for both d@xmath4 and h@xmath5 at solar metallicity is more striking , and has potentially significant consequences for understanding the spectra of metal rich galaxies .
the two solar metallicity clusters are ngc 6553 and ngc 6528 ; both are in the bulge .
the cmd of ngc 6553 shows an abundance of bs stars @xcite , and so the addition of such stars to the models in figure [ fig : mwlick ] is justified .
indeed , adding bs stars results in much better agreement in h@xmath5 and moderately better agreement for d@xmath4 .
the addition of blue hb stars also alleviates much of the discrepancy between models and data .
@xcite compare their model to the same dataset and find that for solar scaled models the h@xmath5 strengths are too weak compared to the observed @xmath119 clusters , echoing our results regarding our default sps model .
main sequence turn - off stars are the dominant stellar phase controlling the strength of d@xmath4
. the effective temperature of turn - off stars would have to be increased by @xmath120k in order to alleviate the discrepancy at solar metallicity .
such an increase would yield a cmd in gross conflict with observations .
modifications to the standard stellar phases thus can not reconcile the data and model d@xmath4 strengths .
either bs , blue hb stars or both are therefore the most likely candidates for explaining the discrepancy . in the absence of evolved or young stars , the ultraviolet is sensitive to the temperature of main sequence turn off stars .
since the temperature of turn off stars depends on metallicity at fixed age , ultraviolet colors such as @xmath121 and @xmath122 will , in the absence of evolved or young stars , redden substantially with increasing metallicity .
in reality , the interpretation of ultraviolet colors of old populations is complicated by the existence of hot , evolved stars , including post agb , extended / blue hb , and bs stars ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is beyond the scope of this paper to consider in detail the constraints on the relative importance of these various populations afforded by ultraviolet colors . instead , our goal will be to demonstrate that plausible variations in the relative weights assigned to post agb and blue hb stars can explain the ultraviolet properties of observed star clusters .
ultraviolet colors of star clusters in m31 @xcite , the mw ( as compiled by * ? ? ?
* ) , and m87 @xcite are compared to sps models in figure [ fig : uv ] .
the approximate location of elliptical galaxies from the sauron survey @xcite are also included in the figure .
we include model predictions from our fsps code for both the padova and basti isochrones , along with model predictions from bc03 and m05 .
recall that m05 does not include post agb stars and so the fuv photometry , which is sensitive to such hot stars , can not be reliably compared to the data .
m05 does include blue hb stars for metal poor models .
bc03 includes post agb stars but does not include blue hb stars at any metallicity . for the fsps predictions with the padova isochrones , we include several model variations . in particular , we consider models 1 ) without post agb and without blue hb stars ; 2 ) with post agb but without blue hb stars ; 3 ) without post agb stars and with a blue hb fraction of 10% ; 4 ) with post agb stars and a blue hb fraction of 100% ; 5 ) with post agb stars and a blue hb fraction that varies with metallicity as @xmath123 for @xmath124 and zero otherwise . before considering the insights gained from model comparisons , it is worth drawing attention to the considerable variation in observed ultraviolet colors of star clusters at fixed metallicity .
@xmath121 colors vary by @xmath125 mags at fixed @xmath57 , while @xmath122 varies by @xmath87 mag .
this phenomenon is known as the ` second
parameter ' effect , and has bee known for decades ( see the review in * ? ? ? * ) .
parameters that might be responsible for the additional variation at fixed @xmath57 include age , helium content , mass loss , stellar rotation , and magnetic fields .
the particular solution is not as important as the observed variation .
it is difficult to imagine constraining an sps model in the uv given these variations , and it will therefore continue to be very challenging to extract useful information from the uv flux of quiescent galaxies . the locus of ultraviolet colors of elliptical galaxies do not correspond to star clusters of any metallicity in either m31 , the mw , or m87 .
the @xmath121 colors correspond to metal
rich clusters , while the @xmath126 colors are similar to metal poor clusters , and the @xmath122 colors of the elliptical sample are significantly redder than any star cluster
. this conflict may be due in part to the fact that galaxies are not mono metallic .
the ultraviolet is quite sensitive to metallicity , and so accurate modeling of the metallicity distribution function is essential for these objects .
these and related issues regarding stellar population modeling of ellipticals will be discussed in @xmath6[s : cqg][s : ex ] . it should also be noted that the star clusters in m87 are significantly bluer in ultraviolet colors at fixed metallicity compared to clusters from the mw and m31 .
the source of this difference is not known , although differences in cluster age or helium abundance are possible explanations @xcite .
nevertheless , it is clear that even if models can be constructed that match the mw cluster data , such models may not be applicable to other systems .
several conclusions can be drawn from the comparisons between models and data .
first , a 10% blue hb fraction at all metallicities produces ultraviolet colors that are comparable to post agb stars , demonstrating the degeneracy between these two evolutionary phases when considering uv colors .
the basti isochrones and the padova isochrones with variable blue hb fraction are both able to fit a majority of the metallicity dependent ultraviolet data . at metallicities
@xmath127 these models fail to reproduce the observed @xmath128 fluxes , whereas models with 100% blue hb fractions can reproduce this data . when modeling the ultraviolet flux of galaxies , a flexible treatment of hot evolved stars is clearly warranted given our inadequate theoretical understanding of the observed ultraviolet colors of clusters , and given the considerable scatter in observed ultraviolet data at fixed metallicity .
we now consider constraints on sps models provided by massive red sequence , ( i.e. , quiescent ) , galaxies .
such galaxies are thought to be composed primarily of old ( @xmath129 gyr ) , metal rich stars .
figure [ fig : redcol ] compares galaxies to models in color color space .
we include the fsps model predictions with the modified padova isochrones in addition to the bc03 and m05 models .
all models predict similar @xmath3 colors .
one clearly sees the almost perfect degeneracy between age and metallicity in this color color space .
it is also clear that all models predict @xmath3 colors that are substantially redder than the data , assuming that such galaxies are composed of old , metal rich stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the discrepancy between observed and predicted @xmath130 colors is @xmath131 mags .
potential remedies to this discrepancy are explored in @xmath6[s : ex ] .
the models perform reasonbly well in the near ir , although there is a notable disagreement in the @xmath132 color for all models . in the @xmath133 vs. @xmath81 plane
the fsps and m05 models perform well , while bc03 fits the data only for solar or sub solar metallicities .
the bc03 and m05 models predict strong variations in @xmath134 with metallicity , in contrast to fsps .
given the narrow distribution of observed @xmath134 colors , the bc03 models would predict nearly all red sequence galaxies to be solar metallicity , while m05 would predict that all were super solar .
in contrast , fsps allows for a range of metallicities from sub to super
solar , in better agreement with detailed analysis of the spectra of red sequence galaxies ( e.g. , * ? ? ?
perhaps the most important conclusion to be reached from these color color comparisons is that no model is able to reproduce the observed locus of red galaxies for all near ir colors , and in addition no model can match the locus of galaxies in the @xmath135 vs. @xmath130 plane . it should also be noted that within the context of one model , optical and near ir color combinations , in particular @xmath130 vs. @xmath134 , are in principle capable of breaking the age
metallicity degeneracy . in practice
it is not currently possible to use such colors to break this degeneracy for the twofold reason that the models can not predict the observed locus of galaxies in that space and the models do not compare well with each other ( see also * ? ? ?
* ; * ? ? ?
the red and near ir colors of these galaxies are determined primarily by the rgb . as noted in @xmath6[s : metalr ] , resolved cmds do not help constrain model predictions for integrated red and near
ir colors because these colors are determined by the brightest and rarest rgb stars .
homogeneous , integrated near
ir photometry for a sample of old metal
rich clusters would be invaluable for constraining models in this regime .
figure [ fig : lick1 ] compares spectral indices of sdss galaxies to the fsps model constructed with the padova isochrones and the miles empirical stellar library .
the data are split according to their velocity dispersions .
all models in this section are broadened to @xmath136 in order to compare to the more massive data bin .
model predictions are shown for dust
free ssps at @xmath95 and @xmath137 for ages @xmath138 gyr .
the index @xmath139 , which measures the strength of the caii lines and the balmer decrement , is often taken to be sensitive primarily to age , and this is borne out in the figure .
the majority of the other indices are sensitive to a combination of age and metallicity .
the index c4668 , which is a blend of many metallic lines including fe , ti , cr , mg , ni , and c@xmath140 @xcite , is the most sensitive to metallicity of the indices shown , at fixed d@xmath4 .
the index [ mgfe] was defined by @xcite in order to be insensitive to @xmath47enhancement . the g4300 index is sensitive to both ch and fei @xcite .
h@xmath5 is often used to constrain the fraction of recent star formation because of its sensitivity to a stars ( e.g. , * ? ? ?
* ) , and so the reliability of sps models in the h@xmath141 plane is particularly important . for old populations ,
the interpretation of h@xmath5 is complicated by the presence of nearby absorption due to the cn molecule .
this implies that the h@xmath5 index can be sensitive to non solar abundance ratios , depending on the precise definition of the index @xcite .
the trend is such that increasing @xmath47enhancement will result in a weaker measured h@xmath5 index because the flux redward of the line will be depressed due to increased cn absorption .
figure [ fig : lick2 ] compares several sps models to the @xmath40 sdss sample .
the bc03 model predictions are compared to the fsps models using both the modified padova and basti isochrones with the miles empirical stellar library . the m05 model is not included because the spectral library in that model ( the kurucz library ) is not of sufficient resolution to measure these indices .
the basti and padova isochrones have very similar main sequence turnoff points at old ages ( see figure [ fig : cmd ] ) .
the primary difference between these isochrones is that the basti rgb is approximately 100k hotter than the padova rgb .
blueward of @xmath142 the main sequence and sub giant branch dominate the flux , and so one would expect very little differences in spectral indices between these two isochrone sets in the blue , for a fixed set of empirical stellar spectra . indeed , in figure [
fig : lick2 ] there are almost no differences between the two isochrone sets for those indices that are measured at @xmath143 .
the only index that shows appreciable differences is tio@xmath140 , which is measured at @xmath144 , where the rgb contributes more to the integrated flux than the main sequence turnoff point .
the bc03 and fsps models compare favorably for the d@xmath4 , [ mgfe] , g4300 , and tio@xmath140 indices ( except for the @xmath137 models for the latter index ) .
the @xmath137 fsps model predictions for c4668 compare more favorably with the data than the bc03 model of comparable metallicity , while the converse is true for the fe4531 index .
the differences between the bc03 and fsps models are most dramatic in the h@xmath141 plane , where the bc03 h@xmath5 ews are larger than the fsps predictions by @xmath145 , and are in significant discord with the data .
this discrepancy has been noted previously by @xcite , and is most likely due to deficiencies in the stelib empirical stellar library used by bc03 .
one of the most striking results of these comparisons is that all models tend to over predict the d@xmath4 strength and under predict the h@xmath5 strength , especially for @xmath137 . as can be gleaned from the symbols in the upper left panel of figure [ fig : lick1 ] , at @xmath137 the models reproduce the observed d@xmath4 locus of massive galaxies for ages @xmath146 gyr , while for @xmath95 the observations are reproduced for somewhat older ages of @xmath147 gyr . from a detailed analysis utilizing multiple spectral indices , @xcite find most massive ellipticals to have solar to super
solar metallicities and ages @xmath148 gyr ( see also , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in light of these results ,
an explanation is required for the discrepancy between observed and predicted d@xmath4 and h@xmath5 values .
possible explanations are discussed in the following section . before considering possible explanations we point out that the proper inclusion of @xmath47enhancement will only exacerbate the discrepancy between models and data in regards to the h@xmath5 index .
this occurs because @xmath47enhancement will result in an even weaker h@xmath5 index @xcite . in the previous section
two striking disagreements were noted between sps models and a sample of massive red sequence galaxies . assuming that such galaxies are composed of ` normal ' stars of at least solar metallicity and with ages @xmath148 gyr , all sps models predict @xmath3 colors that are too red , d@xmath4 strengths that are too large , and h@xmath5 strengths that are too weak .
there are two possible classes of solutions to this discrepancy . in the first class
, an appeal is made to deficiencies in either the stellar atmospheres or stellar evolution calculations . within this class ,
an appeal to stellar evolution means considering inadequacies in standard phases such as the main sequence and its turn off point , and the rgb .
solutions of the second class abandon one or more of the assumptions regarding the composition of such galaxies . in this
second class appeals are made to minority populations of either young stars or metal poor stars , or exotic aspects of stellar evolution including blue stragglers and mass - loss leading to an extended horizontal branch .
these classes are conceptually different in that the former appeals to a modification of the bulk metal
rich old population while the latter appeals to additions of minority populations .
we consider both of these classes below .
we find that the discrepancies are unlikely to be due to the stellar spectral libraries .
we have considered both the theoretical basel library and the empirical miles library and find similar disagreements in the colors . the empirical stelib library used in bc03 yields similar disagreements between both the @xmath130 color and the d@xmath4 strengths , further suggesting that the libraries are not at fault .
these results are at odds with @xcite who showed that the optical spectra and @xmath130 colors of ssps at @xmath149 are markedly different when comparing models constructed using the basel spectral library and the empirical @xcite library .
the purported discrepancies caused by the spectral libraries can not be reproduced with our fsps code when using the empirical miles library . to test that this was not due to differences between the miles and pickles library
, we also compared spectra generated from the bc03 model using both the basel and pickles library and again find no substantial difference in @xmath130 colors .
we have also compared the spectra of individual stars from the pickles , miles , and basel libraries for the same stellar parameters discussed by maraston et al . and again can not reproduce their results ( i.e. , their figure 4 ) .
we also point out that in the context of the explanation proposed by maraston et al . , the basel library should provide a better fit to the data when the colors are k corrected to @xmath150 ( as opposed to @xmath30 ) , because their discrepancy arises in a region of the spectrum sampled by the @xmath151band redshifted to @xmath30 . however , when we compare the colors of @xmath152 red sequence galaxies we find that the model discrepancies remain .
the discrepancy is also unlikely to be due to unresolved issues with canonical phases of stellar evolution such as the main sequence and rgb , because the aspects of stellar evolution relevant for @xmath3 colors are adequately described by the models for old , metal rich star clusters .
this claim is based on two observations .
first , from figure [ fig : hz ] it is clear that fsps can reproduce the main sequence turn off point and base of the rgb of old , metal rich star clusters .
this is especially the case for @xmath94 colors of the clusters ngc 6791 and ngc 188 . to the extent that there are residual discrepancies , the models tend to be somewhat too blue compared to the data , which is in the opposite sense of the discrepancies noted for galaxies .
second , in figure [ fig : cmdf ] it was shown that for blue colors such as @xmath20 , one need not probe the brightest portions of the rgb in order to achieve convergence in the integrated color ( the results for @xmath130 are very similar to @xmath20 ) . therefore , integrated @xmath130 colors of old metal
rich clusters should be accurate because the main sequence turn off and the base of the rgb are relatively well fit by the model in the cmd . in sum , we believe that the discrepancies between model and data @xmath3 colors and spectral indices are not a reflection of the limitations of either stellar atmospheres or canonical aspects of stellar evolution .
there are two remaining possible solutions to the disagreement between models and data for massive red sequence galaxies , which we lump into one class because they share the common property of being minority additions to the dominant , old metal
rich population .
the first solution consists of modifications to the ssps .
this solution includes the addition of bs or blue hb stars .
the second solution to these disagreements lies not in trying to fix the ssps but rather in relaxing the implicit assumption that these massive red galaxies are composed of mono metallic , coeval stars .
indeed , estimates of the metallicities of stars in our own and nearby galaxies , either from direct spectroscopy or via the colors of giant branch stars , always yield a metallicity distribution function that is approximately lognormal with a full width of @xmath153 dex @xcite .
models for the chemical evolution of galaxies also generically predict that galaxies should contain stars of a range of metallicities , including a non trivial fraction of metal poor stars @xcite .
extreme / blue hbs exist in the solar and super solar metallicity star clusters ngc 188 @xcite and ngc 6791 @xcite , in addition to moderately sub solar clusters @xcite .
resolved uv photometry in the nucleus of m32 has shown that @xmath154% of hb stars are hot @xcite . moreover ,
many metal
rich clusters contain copious bs populations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
these observations suggest that it would be prudent to consider models that allow for blue / extended hbs and bs stars to be associated with metal rich populations . in our treatment
we are agnostic as to which population ( metal rich or metal poor ) these hot stars are associated with .
these various effects are explored in figures [ fig : lick3 ] and [ fig : grcol ] . in these figures
we explore the impact of various modifications to the base dust free ssp models .
modifications include : 1 ) a small fraction ( @xmath155% ) of young ( @xmath156 gyr ) stars ; 2 ) a population of blue hb stars ( @xmath157 ) ; 3 ) a population of bs stars ( @xmath116 ) ; 4 ) a minority population ( 5% ) of metal poor stars .
all of these modifications move the base ssp models in the correct qualitative sense ; that is they result in lower values of d@xmath4 and larger h@xmath5 strengths , and bluer @xmath3 colors .
these trends result from the fact that each modification is adding a population of relatively hot stars , whether in the form of a hotter main sequence turnoff ( as happens when metal poor stars are added ) , hot evolved stars including bs and blue hb stars , or hot young stars .
hotter stars have lower values of d@xmath4 , larger balmer absorption , and bluer continua .
the discrepancies discussed above between ssp predictions and massive elliptical galaxies have been known , in one form or another , for decades ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
our contribution to this issue has been twofold : 1 ) to highlight the connection between the indices and broadband @xmath3 colors , in the sense that whatever hot star population is adopted , it should alleviate the discrepancies seen in both indices and colors ; 2 ) to remind the user of sps models that reliable results concerning the ages and metallicities of massive ellipticals can not be attained unless these minority hot star populations are carefully considered .
many previous studies have considered one or more of the minority populations considered herein as a remedy to the model discrepancies .
@xcite disfavors appeals to blue hb stars based on the consideration of higher order balmer lines in ellipticals .
however , this conclusion rests on the assumption that the temperature distribution along the hb in other galaxies is similar to that seen in the mw . while this may seem to be a relatively benign assumption , recent uv observations of clusters in m87 cast doubt on this @xcite .
if the m87 result holds up to further scrutiny , it will mean that insights from the mw star cluster population can not be readily applied to other galaxies .
this would be a dramatic setback to our attempts at understanding the stellar populations in other galaxies , and would have potentially profound implications for our understanding of stellar evolution .
the addition of metal poor stars is rarely considered without also considering the addition of blue hb stars . in this work
we have separated these two populations , since the observed relation between metallicity and hb morphology in the mw is complex and displays considerable scatter @xcite .
it should be remembered that in every galaxy where metallicity distribution functions have been measured , there exist metal poor stars ( log@xmath158 ) in sufficient numbers ( @xmath159% ; e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) to effect the integrated blue / uv flux .
given this observational fact , metal poor stars _ must _ be included in the models ; the only question that remains is in what proportion .
it is something of a disappointment that the identification of the hot star population in ellipticals remains unresolved more than three decades after it was first discussed . despite enormous advances in both the sps models and the quality of the data
, the fundamental problem remains that these various hot star populations occupy essentially the same ( or similar ) ranges in @xmath50 .
we briefly comment on several possible future directions that might help separate the relative influence of these various populations .
if young stars or metal poor stars are the dominant contributor , then because their metallicities will differ from the bulk population , metallicities measured in the blue / uv will differ from metallicities measured in the red .
models will have to be constructed to determine whether or not this effect is measurable .
if it is then this could be a promising means of discriminating between the various hot star candidates .
if young stars are the culprit , then lookback studies of ellipticals will uncover ellipticals with younger and younger minority populations .
such a study would suffer from the typical difficulties in lookback studies , namely that precise descendant
progenitor relations are difficult to construct . however , with new and upcoming wide
field spectroscopic surveys we believe that the difficulties associated with lookback studies can be mitigated .
lookback studies might also be able isolate the importance of blue hb stars @xcite . a final , somewhat more speculative possibility , is the use of sbfs in the blue to discriminate between these various populations .
while these populations occupy approximately the same range of @xmath62 , they all occupy rather different ranges in @xmath49 , and they will thus produce different sbf magnitudes in the blue .
we are currently constructing models aimed at determining whether or not such an effect is observable with current facilities .
k+a , or post starburst , galaxies offer a unique constraint on intermediate age populations .
the prominent a star spectra in these galaxies without any accompanying h@xmath102 emission ( i.e. , no _ current _ star formation ) , indicates that they are dominated by stars with ages @xmath160 gyr .
such galaxies can therefore provide potentially strong constraints on the importance of tp agb stars , since such stars peak in prominence at ages @xmath160 gyr .
figure [ fig : kacol ] compares the colors of k+a galaxies defined by @xcite to the colors of dust free ssps at ages @xmath161 gyr for metallicities @xmath162 .
model predictions from fsps ( using the padova isochrones ) , m05 , and bc03 are shown .
the fsps model predictions are shown for a @xmath163model sfh ( sfr@xmath164 ) with @xmath165 gyr for cosmetic reasons only .
recall that the tp
agb phase in the padova isochrones was modified in order to produce acceptable agreement with both the mc star clusters and these post starburst galaxies ( although the modifications are slight for intermediate ages at solar metallicity ; see @xmath6[s : mod ] ) . while not shown , fsps predictions using the basti isochrones are similar to the padova isochrones .
a reddening vector is shown to indicate the shift in colors expected for a uniform screen of dust with optical depth @xmath166 and a @xcite attenuation law .
the ratio of a to k stars in the observed spectra decreases with increasing @xmath130 and @xmath167 colors , indicating that the sequence in the left panel is primarily a sequence in age , and not dust nor metallicity .
this observational fact will be a significant discriminator amongst the sps models considered herein .
the fsps model predictions encompass the range of observed @xmath167 , @xmath130 , and @xmath134 colors .
the dominant variable driving the change in model colors is age , in agreement with the data .
the bc03 model predictions are somewhat too blue in @xmath167 , although we believe that the agreement is acceptable given the fact that post starburst galaxies are composite populations whereas we have modeled them as single populations ( see below ) .
in contrast to the fsps and bc03 models , the m05 model performs poorly in @xmath130 and @xmath167 colors .
in particular , model sequences of increasing age produce trends in these colors that are orthogonal to the observed trends with age . in the context of the m05 model the increasing @xmath130 and @xmath167 colors
would be interpreted as a sequence of either increasing dust content or increasing metallicity .
this interpretation would be in contrast with the observed variation in the spectral properties .
it is also difficult to explain the bluer half of the observed k+a galaxies in @xmath168 colors unless unrealistically low metallicities are considered . appealing to younger ages
does not remedy this either , as m05 model predictions at younger ages ( @xmath169 gyr ) continues to diverge from the locus of observed colors . as discussed in previous sections , it is difficult to reliably calibrate sps models against galaxies because galaxies contain stars of a range of metallicities and ages , and the effects of dust can not be ignored .
here we have been considering a dust free , mono metallic , coeval population model . for the comparisons in this section ,
these complications turn out to have minimal consequences regarding the model comparisons .
for example , as shown in figure [ fig : kacol ] , the dust reddening vector lies parallel to the locus of data , and so dust effects can not move the models onto or off of the observed locus , but only along the locus . since our conclusions do not rely on the detailed location of the models along the observed locus , our results are insensitive to dust effects .
we have also considered the effects of adding an old population on top of the intermediate age , post starburst population . even in the extreme case where 90% of the mass is contained in an old population and
only 10% spans the intermediate age range of @xmath161 gyr , the red end of the model predictions only shift slightly toward redder colors .
the primary effect of an addition of an old population is a narrowing of the range in color of the model predictions . in any event
, the addition of old stars has an effect similar to dust attenuation namely , the models shift along the observed locus . we conclude that , in the case of post starburst galaxies and for the colors we have considered , composite population effects do not impact our conclusions .
in the previous section we explored the constraints on several sps models provided by star cluster data in the local group and m87 and by post starburst and massive red sequence galaxies .
we now summarize these results and provide a general evaluation of the models .
star clusters in the mcs provide valuable constraints on models at intermediate age and sub solar metallicities .
the m05 model performs poorly when compared to the extant data , while the bc03 model performs better , although the bc03 model is somewhat too blue in the near ir . in m05 , the colors are too red and the age
dependence is incorrect .
the incorrect age dependence in m05 can not be attributed to the details of mc cluster age estimation .
our own fsps model with padova isochrones is able to achieve excellent agreement with the data only after we modify the tp
agb phase of the input padova isochrones .
the basti isochrones in fsps fare well , although less
so in the near ir , which plausibly reflects their simplistic tp agb treatment , and , at young ages , which may be due to our use of the basti isochrones that suppresses convective overshooting .
similar conclusions are reached regarding the fsps model when considering constraints afforded by sbfs of mc star clusters .
the mw and m31 provide a large sample of old , metal poor star clusters , but only a small handful of old metal rich clusters .
the fsps ( using both the padova and basti isochrones ) , bc03 , and m05 models all make similar predictions for the @xmath170 photometry of old metal poor systems .
the models compare favorably in @xmath11 colors once uncertainties in the filter transmission curves are considered . in the near ir the models perform less well , where for example @xmath2 colors are too blue by @xmath171 mags at log@xmath93 .
the origin of this discrepancy is unclear , but may either be due to an rgb that is too hot ( lowering the rgb by 200k would alleviate this tension ) or stellar libraries that are inadequate in the near ir . at higher
metallicity , multi color cmds in @xmath172 colors are well reproduced by the fsps model in regards to the location of the turn off point , sub giant branch , and the lower portion of the rgb .
we demonstrate that agreement in cmd space between data and models does not imply that the integrated colors are adequately constrained because the integrated colors in the red and near ir are dominated by the brightest stars in the cluster .
these stars are rarely included in cmd studies because they are rare and often saturate the photometry . in any event , even if these stars are considered , owing to their rarity it is challenging to reliably calibrate models in this regime in cmd space .
it is for this reason that we focus our calibrations on integrated cluster properties , rather than resolved cmds .
the spectral indices d@xmath4 and h@xmath5 of mw star clusters are not adequately reproduced by either the fsps or bc03 model at high and low metallicities , respectively ( the m05 model does not contain sufficient spectral resolution to make a robust prediction ) .
these inadequacies are removed if bs or hot hb stars are added . in the ultraviolet ,
the fsps model can reproduce the observed trends because the relevant phases post agb and hb are handled flexibly and so can be tuned to match the data . the bc03 and m05 models generally fare less well , either because of the omission of blue hb stars ( bc03 ) or post agb stars ( m05 ) . turning to galaxies ,
all models predict @xmath3 colors of red sequence galaxies that are too red and the fsps and bc03 models predict d@xmath4 strengths that are too strong ( again , the m05 model does not contain the requisite resolution ) , under the assumption that such galaxies are composed solely of old stars with @xmath173 .
these issues are likely closely connected to the model failures at reproducing the spectral features of metal rich mw star clusters .
these discrepancies can be removed by adding a minority population of hot stars , coming either from metal poor turn off stars , hot hb stars , bs stars , or young ( @xmath97 gyr ) stars .
the fsps and bc03 models are able to reproduce the optical and near ir colors of post starburst galaxies , while the m05 model performs less well .
these galaxies likely have a preponderance of tp agb stars , and so they provide valuable constraints on this phase . in summary ,
our fsps model is able to reproduce the greatest range of observations , due in large part to the flexibility of the model .
bc03 fares somewhat less well , and m05 performs worst of the three . even where calibrating data is available , significant discrepancies remain .
important regions of parameter space do not contain sufficient calibrating data , and so the models are even less reliable there . finally , when modeling predominantly old , metal rich populations , care must be taken to include minority populations of hot stars . in this section
we comment on a number of other sps models in the literature .
the main purpose of the following discussion is to highlight the fact that many sps models rely on the same ingredients as the models considered in detail herein , and therefore conclusions reached in earlier sections should be applicable to other models as well .
the most popular sps models not considered herein are starburst99 @xcite and pegase @xcite . in its most current version
, starburst99 combines the geneva models for massive stars and the padova isochrones for intermediate and low mass stars .
starburst99 employs the tp agb models of @xcite , which are the same models used in bc03 .
the basel stellar library is used to convert the models into observable seds . for hot stars ,
the theoretical models of @xcite are used . for stellar populations
not dominated by massive stars , starburst99 will produce results similar to bc03 .
the treatment of nebular emission and massive stars , including wolf rayet stars , is more sophisticated than either fsps , m05 , or bc03 , and should therefore continue to be the model of choice when emission from massive stars is important .
the recent popstar model @xcite also contains a sophisticated treatment of massive stars , including the theoretical atmosphere models for o and wolf rayet stars of @xcite , and will therefore produce results similar to starburst99 .
pegase utilizes the padova isochrones and couples these to two different stellar spectral libraries : basel , and elodie @xcite .
elodie is an empirical library of @xmath174 stars . in @xcite ,
various spectral indices are computed with pegase and compared to observations .
the comparisons focus on a small sample of indices ( including h@xmath175 , fe5270 , mgb , and fe5335 ) for elliptical galaxies and mw star clusters
. it would be fruitful to consider more comprehensive calibrating comparisons for the pegase model , such as is considered herein . in any event
, the underlying isochrones are similar to those used in bc03 , so the predictions of pegase should be similar to bc03 . the vazdekis models are described in @xcite , @xcite , and vazdekis et al .
( in prep ) .
they utilize the padova isochrones coupled to the empirical miles stellar library . for old populations this model should thus be essentially identical to the fsps predictions when the unmodified padova isochrones are used .
the vazdekis model may thus fail to reproduce the observed near
ir colors of mc star clusters , since , as we demonstrated , the unmodified padova isochrones are a poor match to those data .
significant efforts are currently underway to generate sps models that handle @xmath47enhancement self
consistently , in the sense that both the stellar evolution calculations and the atmosphere models have the same abundance patterns ( e.g. , * ? ? ?
* ; * ? ? ?
. these models will provide valuable constraints on the relative effects of @xmath47enhancement .
however , most models currently only follow stellar evolution through the end of the rgb , and so they can not be used on their own to build model galaxies ( the exception to this is the basti database ; * ? ? ?
* ; * ? ? ?
we anticipate that the treatment of @xmath47enhancement as a variable on par with age and @xmath57 in sps models will likely become commonplace in the near future . from the above discussion
it should be clear that the majority of sps codes use the padova isochrones with the basel stellar library .
our comparisons and evaluations in previous sections , which rely on these ingredients , should therefore be applicable to many other sps models .
we now summarize our principle results . * the latest stellar models from the padova group can not fit the near ir photometry and surface brightness fluctuations of star clusters in the mcs . with our flexible sps ( fsps )
code we have modified the tp
agb phase in the padova isochrones to produce better fits to the data .
the m05 models also fail to reproduce the data , both because the colors are too red and the age
dependence is incorrect .
the bc03 and the fsps model using the basti isochrones fare well , although both models are somewhat too blue in the near ir .
a significant limiting factor in more accurate calibrations is the difficulty in constructing reliable photometry of mc clusters , owing to the fact that tp
agb stars are luminous and rare . * at low metallicities ( log@xmath176 ) and old ages , all models predict similar @xmath170 colors ( the typical variation between models is @xmath177 mags ) .
when compared to mw and m31 star clusters , the models are 0.2 mags too blue in @xmath2 at log@xmath93 , and 0.1 mags too red in @xmath1 at all @xmath57 .
the origin of the discrepancies in the near ir is unclear .
decreasing the temperature of the rgb by @xmath178k alleviates much of the disagreement .
if this modification is isolated to the brightest stars , then such a modification could improve agreement with integrated photometry without compromising agreement in cmd space .
the fsps model with the basti isochrones performs worst at low metallicities and old ages of the models considered ( although the disagreement is not dramatic ) , while the fsps model with the padova isochrones performs well .
+ multi color cmds of two metal rich clusters ,
ngc 6791 and ngc 188 , are well fit by fsps .
spectral indices of mw star clusters are generally well fit by both fsps and bc03 , although the models predict d@xmath4 strengths too large and h@xmath5 strengths too weak compared to two log@xmath179 clusters .
* the ultraviolet photometry of the mw , m87 , and m31 star clusters can be well fit by fsps because fsps contains flexible treatments of the post agb and horizontal branch evolutionary phases .
this flexibility is essential given the substantial scatter in the ultraviolet data at fixed metallicity , and given our inadequate theoretical understanding of these phases .
the m05 and bc03 models perform less well because of their incomplete treatment of these advanced evolutionary phases . *
sps models are compared to @xmath180 photometry of massive red sequence galaxies .
the fsps , bc03 , and m05 models fare well in most respects , with a few exceptions : all models are too blue in @xmath132 ; the m05 model is somewhat too blue in @xmath134 ; all models are far too red in @xmath130 and @xmath135 .
these conclusions hold under the assumption that red sequence galaxies are composed exclusively of old metal rich stars with canonical evolutionary phases .
the disagreement in the @xmath3 colors can be alleviated if some combination of young ( @xmath97 gyr ) stars , metal poor stars , and blue straggler stars are added in small amounts .
* optical spectral indices of massive galaxies are generally well fit by the fsps and bc03 models , with the important exceptions that both models predict d@xmath4 in excess of observations and neither fit the observed locus of galaxies in the d@xmath181h@xmath5 plane .
these conclusions hold for the same assumptions noted above , and , as above , small additions of some combination of young stars , metal poor stars , blue straggler and horizontal branch stars can remedy this disagreement . *
the fsps and bc03 models adequately describe the @xmath182 colors of k+a , or ` post starburst ' galaxies , while the m05 model performs less well .
such galaxies contain a large proportion of intermediate age ( @xmath183 gyr ) stars and thus provide unique constraints on the importance of tp agb stars in galaxies .
we thank david hogg for providing the k+a galaxy catalog , patricia snchez - blzquez for providing the miles library in electronic format , rita gautschy for providing her model results , the basti , padova , basel , and mpa / jhu groups for publically releasing their results and for assistance in their use , and tony sohn for help with interpretation of the ultraviolet star cluster data .
santi cassisi , david hogg , claudia maraston and ricardo schiavon are thanked for comments on an earlier draft .
funding for the sloan digital sky survey ( sdss ) has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions .
the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
|
stellar population synthesis ( sps ) provides the link between the stellar and dust content of galaxies and their observed spectral energy distributions . in the present work
we perform a comprehensive calibration of our own flexible sps ( fsps ) model against a suite of data .
these data include ultraviolet , optical , and near ir photometry , surface brightness fluctuations , and integrated spectra of star clusters in the magellanic clouds ( mcs ) , m87 , m31 , and the milky way ( mw ) , and photometry and spectral indices of both quiescent and post starburst galaxies at @xmath0 .
several public sps models are intercompared , including the models of bruzual & charlot ( bc03 ) , maraston ( m05 ) and fsps .
the relative strengths and weaknesses of these models are evaluated , with the following conclusions : 1 ) the fsps and bc03 models compare favorably with mc data at all ages , whereas m05 colors are too red and the age dependence is incorrect ; 2 ) all models yield similar optical and near ir colors for old metal
poor systems , and yet they all provide poor fits to the integrated @xmath1 and @xmath2 colors of both mw and m31 star clusters ; 3 ) fsps is able to fit all of the ultraviolet data because both the post agb and horizontal branch evolutionary phases are handled flexibly , while the bc03 and m05 models fail in the far uv , and
both far and near uv , respectively ; 4 ) all models predict @xmath3 colors too red , d@xmath4 strengths too strong and h@xmath5 strengths too weak compared to massive red sequence galaxies , under the assumption that such galaxies are composed solely of old metal rich stars ; 5 ) fsps and , to a lesser extent , bc03 can reproduce the optical and near
ir colors of post starburst galaxies , while m05 can not .
reasons for these discrepancies are explored .
the failure at predicting the @xmath3 colors , d@xmath4 , and h@xmath5 strengths can be explained by some combination of a minority population of metal poor stars , young stars , blue straggler and/or blue horizontal branch stars , but not by appealing to inadequacies in either theoretical stellar atmospheres or canonical evolutionary phases ( e.g. , the main sequence turn off ) .
the different model predictions in the near
ir for intermediate age systems are due to different treatments of the tp
agb stellar evolutionary phase .
we emphasize that due to a lack of calibrating star cluster data in regions of the metallicity age plane relevant for galaxies , all of these models continue to suffer from serious uncertainties that are difficult to quantify .
|
in the past decade , observations of x - ray emission from accreting neutron stars and black holes have introduced new possibilities for astrophysical tests of fundamental physics .
recent discoveries made by satellites such as asca , rxte , bepposax , chandra , and xmm - newton provide direct evidence for strong - field gravitational effects in compact binary systems and active galactic nuclei ( agn ) .
these results include doppler - broadened iron k@xmath0 fluorescent emission from microquasars @xcite and millisecond variability of the x - ray flux from black holes in low - mass x - ray binaries @xcite .
these measurements give the exciting prospect for determining a black hole s mass and spin , as well as tests of general relativity in the strong - field regime .
the strong gravitational fields near a black hole ( bh ) introduce significant deviations from newtonian physics , including the existence of an inner - most stable circular orbit ( isco ) , a feature absent in the classical kepler problem .
since accreting gas can efficiently lose energy and angular momentum only outside of the isco , the hydrodynamic and radiative behavior of the inner accretion disk should be strongly dependent on the structure of the space - time metric near the isco .
the famous `` no hair '' theorem states that the only observable features of an electrically neutral black hole are functions of its mass @xmath1 and specific angular momentum @xmath2 . by understanding the behavior of matter near the isco , we can determine the mass and angular momentum , and thus completely describe the black hole .
many authors have approached the problem of accretion in compact binaries with a variety of different methods , including early analytic models by @xcite and @xcite .
some have simplified the hydrodynamics in favor of a flat , thin , steady - state disk and a more detailed treatment of general relativistic effects @xcite . to include dynamic effects , essential for modeling qpos
, others have included magnetohydrodynamics ( mhd ) in a pseudo - newtonian model @xcite or with smoothed particle hydrodynamics @xcite .
a family of perturbative models has given rise to the field of diskoseismology @xcite , where different global modes in the disk oscillate with different frequencies .
@xcite propose a magnetic coupling between the rotating black hole and the accretion disk as a means of producing high - frequency qpos , analogous to the blandford - znajek process @xcite used to describe agn .
recent observations of commensurate relationships in the high - frequency qpos of black hole accretion disks @xcite , as well as the longstanding puzzles of the frequency variability of qpo peaks and their correlations with x - ray flux and energy @xcite , motivate more detailed study of the qpo phenomenon as a means to determining the black hole parameters .
we have developed a model that is a combination of many of the above approaches in which additional physics ingredients can be added incrementally to a framework grounded in general relativity .
the model does not currently include scattering , radiation pressure , magnetic fields , or hydrodynamic forces , instead treating the disk as a collection of cold test particles radiating isotropically in their respective rest frames .
the trajectories of emitted photons are integrated through the metric to a distant observer to construct time - dependent images and spectra of the disk .
the dynamic model uses the geodesic trajectory of a massive particle as a guiding center for a small region of excess emission , a `` hot spot , '' that creates a time - varying x - ray signal , in addition to the steady - state background flux from the disk .
this hot spot model is motivated by the similarity between the qpo frequencies and the black hole ( or neutron star ) coordinate frequencies near the isco @xcite as well as the suggestion of a resonance leading to integer commensurabilities between these coordinate frequencies @xcite .
@xcite investigated primarily the qpo frequency pairs found in lmxbs with a neutron star ( ns ) accretor , but their basic methods can be applied to black hole systems as well .
@xcite have presented a thorough analysis of this hot spot model for a collection of ns binaries for which pairs of qpos have been observed . based on a number of experimental and theoretical arguments ,
they conclude that the geodesic hot spot model is not a physically viable explanation for the observed ns qpos .
for low to moderate eccentricity orbits , the coordinate frequencies simply do not agree with the qpo data .
for highly eccentric geodesics , they argue that the relative power in the different frequency modes are qualitatively at odds with the observations .
furthermore , they show that hydrodynamical considerations place strong constraints on the possible size , luminosity , coherency , and trajectories of the hot spots .
many of these points are addressed in our version of the hot spot model .
also , by including full 3-dimensional ( 3d ) ray - tracing , we can quantitatively predict how much qpo power will be produced by a hot spot of a given size and emissivity moving along a geodesic orbit near the isco .
along with the special relativistic beaming of the emitted radiation , we find that strong gravitational lensing can cause high - amplitude modulations in the light curves , even for relatively small hot spots . the issues of differential rotation and shearing of the emission region are addressed below when we consider the generalization of the hot spot model to include arcs and non - planar geometries .
perhaps the most powerful feature of this hot spot model is the facility with which it can be developed and extended to more general accretion disk geometries .
in addition to providing a possible explanation for the commensurate hfqpos in at least two systems ( xte j1550 - 564 and gro j1655 - 40 ) , the hot spot model with full general relativistic ray - tracing is a useful building block toward any other viable model of a dynamic 3d accretion disk . within the computational framework of the kerr metric
, we can investigate many different emission models and compare their predicted x - ray spectra and light curves with observations .
for example , our ray - tracing code could be used in conjunction with a 3d mhd calculation of the accretion disk to simulate the time - dependent x - ray flux and spectrum from such a disk .
it is with this motivation that we present the initial results of the model .
section 2 describes the relativistic ray - tracing methods and discusses numerical techniques . in section 3 ,
we develop the basic model of a steady - state disk with the potential application to broad iron k@xmath0 emission .
section 4 introduces the basic hot spot model for circular geodesics and shows the effect of binary inclination and black hole spin on the qpo power spectrum . in section 5 the model
is extended to trajectories with non - zero eccentricity and inclination , as well as elongated hot spots or arcs , giving a set of model parameters that best fit the qpo data from a large number of outbursts from xte j1550 - 564 . in section 6
we present our conclusions and a discussion of future work .
we begin by dividing the image plane into regularly spaced `` pixels '' of equal solid angle in the observer s frame , each corresponding to a single ray . following the sample rays backward in time ,
we calculate the original position and direction that a photon emitted from the disk would require in order to arrive at the appropriate position in the detector . the gravitational lensing and magnification of emission from the plane of the accretion disk is performed automatically by the geodesic integration of these evenly spaced photon trajectories , so that high magnification occurs in regions where nearby points in the disk are projected to points with large separation in the image plane . to model the time - varying emission from the disk , each photon path is marked with the time delay along the path from the observer to the emission point in the disk .
to integrate the geodesic trajectories of photons or massive particles , we use a hamiltonian formalism that takes advantage of certain conserved quantities in the dynamics .
the resulting equations of motion do not contain any sign ambiguities from turning points in the orbits , as are introduced by many classical treatments of the geodesic equations in the kerr metric .
we define a hamiltonian function of eight phase space variables @xmath3 and an integration variable ( affine parameter ) @xmath4 along the path length . for a general space - time
metric @xmath5 with inverse @xmath6 , we can define a hamiltonian @xmath7 quadratic in the momenta as @xmath8 where the rest mass @xmath9 is a constant ( @xmath10 for photons , @xmath11 for massive particles ) . applying hamilton s equations from classical mechanics
, we reproduce the geodesic equations : @xmath12 @xmath13 for any metric , the hamiltonian @xmath7 is independent of the affine parameter @xmath4 , allowing us to use one of the coordinates as the integration parameter and reduce the dimensionality of the phase space by two .
we use the coordinate @xmath14 as the time coordinate for the six dimensional phase space @xmath15 .
the corresponding hamiltonian is @xmath16^{1/2}\ ] ] with equations of motion @xmath17 @xmath18 we have thus reduced the phase space to the six - dimensional tangent bundle @xmath15 .
moreover , because the metric is independent of @xmath14 , @xmath19 is also an integral of motion . in boyer - lindquist coordinates @xmath20 , the kerr metric may be written @xmath21 which allows a relatively simple form of the inverse metric @xmath22 for a black hole of mass @xmath1 and specific angular momentum @xmath23 , we define ( in geometrized units with @xmath24 ) @xmath25 \sin^2\theta.\end{aligned}\ ] ] as a check , we see that the metric reduces to the well - known schwarzschild metric in the limit @xmath26 . in the limit @xmath27 , it reduces to flat space - time with hyperbolic - elliptical coordinates . with this form of the metric
, the hamiltonian @xmath28 can be written @xmath29 this new hamiltonian is also independent of @xmath30 ( azimuthally symmetric space - time ) , giving the conjugate momentum @xmath31 as the second integral of motion for @xmath28 .
we are now left with five coupled equations for @xmath32 .
the third integral of motion , carter s constant @xcite @xmath33,\ ] ] is used as an independent check of the accuracy of the numerical integration . as mentioned above , many traditional schemes to calculate trajectories in the kerr metric use this additional integral of motion to further reduce the dimensionality of the problem @xcite and even reduce the problem to one of quadrature integration [ e. g. @xcite ] .
while potentially increasing the speed of the computation , this approach also introduces significant complications in the form of arbitrary signs in the equations of motion corresponding to turning points in @xmath34 and @xmath35 .
another common approach is to define effective potentials in @xmath36 and @xmath37 , which give a set of four parametric equations for @xmath38 and @xmath39 .
these parametric solutions introduce the difficulty of reconstructing the coordinate momenta , which are not evolved along with the coordinate positions . in our ray - tracing algorithm
, we have traded a modest improvement in computational speed for simplicity in implementation .
the initial conditions for the photon or particle geodesics are determined in the local orthonormal frame of a `` zero - angular momentum observer '' ( zamo ) .
the zamo basis is defined such that the spatial axes are aligned with the coordinate axes and then the time axis is determined by orthogonality @xcite .
the zamo tetrad @xmath40 is @xmath41 where @xmath42 is the standard coordinate basis in boyer - lindquist coordinates .
one advantage of the zamo basis is that the basis vector @xmath43 is time - like ( @xmath44 ) everywhere outside of the horizon . for a coordinate stationary observer ,
on the other hand , the time basis vector @xmath45 becomes space - like ( @xmath46 ) inside the ergosphere @xmath47 .
for sufficiently large values of the spin parameter @xmath48 , the inner - most stable circular orbit ( often taken for the inner edge of the accretion disk ) extends within the ergosphere , emphasizing the advantage of using the zamo basis . at a point far away from the black hole , the space - time is nearly flat so euclidean spherical geometry gives the spatial direction of the photon @xmath49 , from which the initial momentum in the coordinate basis is calculated : @xmath50 where the photon energy measured by the distant zamo is @xmath51 .
the photon trajectories are integrated backward in time from the image plane oriented at some inclination angle @xmath52 with respect to the axis of rotation for the black hole , where @xmath53 corresponds to a face - on view of the disk and @xmath54 is an edge - on view .
the accretion disk is confined to a finite region of latitude with angular thickness @xmath55 normal to the rotation axis .
the photons terminate either at the event horizon or pass through the surfaces of colatitude @xmath56 , as shown in figure [ plotone ] .
as trajectories pass through the disk , the photon s position and momentum @xmath57 are recorded for each plane intersection in order to later reconstruct an image of the disk . for an infinitely thin disk @xmath58
, it is easy to show how the image plane maps onto the source plane . taking an evenly spaced grid of initial photon directions , figure [ plottwo ]
shows the positions of intersection with the source plane , in pseudo - cartesian coordinates defined by @xmath59 photons that cross the black hole event horizon before intersecting the plane are not shown .
as the rays are deflected by the black hole , they tend to be focused on the far side , giving a strong magnification by mapping a large area in the image plane onto a small area of the source plane , as seen here by a higher density of lattice grid points . for the flat disk geometry
, rays are not allowed to pass through the plane defined by @xmath60 , so we do not see multiple images of sources `` behind '' the black hole , as is often observed in the strong gravitational lensing of distant quasars by intervening galaxies .
however , for sufficiently high inclinations and spin values , single points in the equatorial plane can be mapped to different regions of the image plane , creating multiple images of certain regions of the disk .
this effect is seen in the folding of the image map onto itself near the horizon in the bottom right of figure [ plottwo ] .
the disk itself is modeled as a collection of mass elements moving along circular geodesics around the black hole , emitting isotropic , monochromatic light with energy @xmath61 in the emitter s rest frame . for each photon with 4-momentum @xmath62 intersecting a particle trajectory with coordinate velocity @xmath63 , the measured redshift is given by @xmath64 where for a distant observer at @xmath65 , we take @xmath66 $ ] . for disk models with finite thickness , the radiation transport equation can be solved as the ray passes through the disk . in a locally flat frame ,
the radiation transport equation is @xmath67 where @xmath68 is the differential path length and @xmath69 , @xmath70 , and @xmath71 are respectively the radiation intensity and the emissivity and opacity of the plasma at a frequency @xmath72 ( scattering terms can be included in both @xmath70 and @xmath71 ) .
additionally , as the photon bundle propagates through the global curvature around the black hole , the spectral intensity at a given frequency evolves as the photons are gravitationally red - shifted , maintaining the lorentz invariance of @xmath73 . for most of the calculations presented in this paper ,
we are primarily concerned with radiation coming from a limited region of the disk , treated as a monochromatic source with @xmath74 and @xmath75 , with @xmath76 the emissivity at space - time coordinate @xmath77 . when calculating the emission from a flat , steady - state disk , the plane defined by @xmath78 is taken to be totally opaque
@xmath79 $ ] so that rays can not curve around and see the `` underside '' of the accretion disk . for each pixel @xmath80 in the image plane ,
an observed photon bundle spectrum @xmath81 is given for each time step by integrating the contribution of the hot spot and the disk through the computational grid .
this collection of incident photons can then be summed to give time - dependent light curves , spectra , or spatially resolved images .
the results of this paper are based primarily on flat , steady state disks and single hot spots , yet with the computational methods described above , arbitrary disk geometries and emissivity / opacity models can be simulated as well . the ray - tracing calculation is carried out by numerically integrating equations ( [ hameq_1 ] ) and ( [ hameq_2 ] ) with a fifth - order runge - kutta algorithm with adaptive time stepping .
this provides high accuracy over a large range of scales as the photon follows a long path through the relatively flat space - time between the observer and the black hole , and then experiences strong curvature over a small region close to the horizon .
we typically maintain an accuracy of one part in @xmath82 , which can be independently confirmed by monitoring @xmath83 , carter s constant .
the images and spectra are formed by ray - tracing a set of photon paths , usually of the order @xmath84 grid points in @xmath80 with @xmath85 latitudinal zones in @xmath35 and spectral resolution of @xmath86 .
a steady - state disk can be made of a collection of massive particles moving in concentric planar circular orbits . for orbits at a radius @xmath34 in a plane orthogonal to the spin axis , a particle s specific energy and angular momentum are given analytically by @xcite : @xmath87 and @xmath88 here the top sign is taken for prograde orbits ( particle angular momentum parallel to black hole angular momentum ) and the bottom sign for retrograde orbits . inside the isco ,
the particles follow plunge trajectories with constant energy and angular momentum determined at the isco . in practice , when calculating emission from a steady - state disk , we take the inner edge of the disk to be the isco radius @xmath89 . for a disk made up of massive particles on circular orbits emitting isotropically from a region between @xmath90 and @xmath91 , the doppler broadening of an emission line ( typically iron k@xmath0 with @xmath92 kev ) may be used to determine the inclination of the disk with respect to the observer .
disks at higher inclination will have an intense blue - shifted segment of the spectrum corresponding to the doppler - boosted photons emitted from gas moving toward the observer .
the higher intensity for the blue - shifted photons is caused by relativistic beaming , determined by the lorentz invariance of @xmath73 along a photon bundle : @xmath93 figure [ plotthree ] shows the integrated spectra from a set of accretion disks with outer radius @xmath94 and inner radius at @xmath95 for a spin parameter @xmath96 , normalized such that @xmath97 .
all spectra are assumed to come from a flat , opaque disk with uniform emission @xmath98 $ ] . repeating this calculation for a range of spin parameters ( and thus a range of @xmath99 ) , we find that the dependence on disk inclination is quite strong , while the dependence on black hole spin is almost insignificant .
this is reasonable because , except for very close to the horizon , the spin has little effect on the orbital velocity for circular orbits : for spin parameter @xmath48 and radius @xmath34 , the observed angular frequency is given by @xmath100 where again we take the upper sign for prograde orbits and the lower sign for retrograde orbits . for the large part of the disk , the orbits have nearly keplerian orbital frequencies , as measured in coordinate time @xmath101 . for larger values of @xmath48 ,
the isco extends in close to the event horizon , increasing the radiative area of the disk .
however , due to the strong gravitational redshift in this inner region , the observed intensity is reduced by a significant factor of @xmath102 , resulting in a weak dependence on spin for disks with uniform emission .
many accretion disk models include an emissivity that scales as a power of the radius .
following @xcite , we apply an emissivity factor proportional to @xmath103 , giving an added weight to the inner , presumably hotter , regions . for the inner section of an accretion disk with @xmath94 , this inverse - square dependence closely approximates the classic shakura - sunyaev steady - state disk model @xcite .
as can be seen in figure 4 , this extra emission from close to the black hole serves to break the otherwise weak dependence on spin . for an inclination of @xmath104 ,
five different spin values are shown : @xmath105 , corresponding to inner disk boundaries at @xmath106 .
since the sign of @xmath48 is defined with respect to the angular momentum of the accretion disk , negative values of @xmath48 imply retrograde orbits that do not survive as close to the black hole , plunging at larger values of @xmath99 .
the disks that extend in closer produce more low - energy red - shifted photons , giving longer tails to the spectra at @xmath107 and smaller relative peaks at @xmath108 .
thus , if we can determine the inclination of the disk independently ( e.g. through spectroscopic observations of the binary companion ) , the spin may be inferred from the broadening of an iron emission line . however , since the plane of the disk tends to align normal to the black hole spin axis close to the isco , the binary inclination may not coincide with the inclination of the inner disk .
the problem of inclination , along with other complications , such as additional emission lines and other causes of scattering and line broadening , motivates us to look more closely at the qpo spectra as a method for determining black hole spin .
given the map from the accretion disk to the image plane , with each photon bundle labeled with a distinct 4-momentum and time delay , we can reconstruct time - dependent images of the disk based on time - varying emission models . the simplest model we consider is a single region of isotropic , monochromatic emission following a geodesic trajectory : the `` hot spot '' or `` blob '' model @xcite .
the hot spot is a small region with finite radius and emissivity chosen to have a gaussian distribution in local cartesian space : @xmath109.\ ] ] the spatial position 3-vector @xmath110 is given in pseudo - cartesian coordinates by the transformation defined by equation ( [ cart_boyer ] ) and @xmath111 . outside a distance of @xmath112 from the guiding geodesic trajectory
, there is no emission .
we typically take @xmath113 , but find the normalized light curves and qpo power spectra to be independent of spot size . because we assume all points in the hot spot have the same 4-velocity as the geodesic guiding trajectory
, one must be careful not to use too large a spot or the point of emission @xmath77 can be spatially far enough away from the center @xmath114 to render the inner product @xmath115 unphysical .
we have also explored a few different hot spot shapes , ranging from spherical to a disk flattened in the @xmath116 direction and similarly find no significant dependence of the spectra on spot shape . after calculating and tabulating the hot spot trajectory as a function of coordinate time @xmath101
, the ray - tracing map between the disk and the observer is used to construct a time - dependent light curve from the emission region . for each photon bundle intersection point
there is a time delay @xmath117 ( where @xmath118 are the coordinate indices in the image plane and @xmath119 is the latitude index in the disk ) so for the observer time @xmath120 , we first determine where the hot spot was at coordinate time @xmath121 . if the spot centroid is close ( within @xmath112 ) to the disk intersection point @xmath122 , then the redshifted emission is added to the pixel spectrum @xmath123 , weighted by equation ( [ i_exp ] ) . in this way
, a movie can be produced that shows the blob orbiting the black hole , including all relativistic effects .
such a movie shows a few immediately apparent special relativistic effects such as the doppler shift and beaming as the spot moves toward and then away from the observer . for a hot spot orbiting in the clock - wise direction as seen from above ( @xmath124 with @xmath125 toward the observer ) ,
the point of maximum blue shift actually occurs at a point where @xmath126 because of the gravitational lensing of the light , beamed in the forward direction of the emitter and then bent toward the observer by the black hole .
gravitational lensing also causes significant magnification of the emission region when it is on the far side of the black hole , spreading the image into an arc , much like distant galaxies are distorted by intervening matter in galaxy clusters . a simulated time - dependent spectrum or _ spectrogram _ for this hot spot model is shown in figure [ plotfive ] .
the horizontal axis measures time in the observer s frame , with @xmath127 corresponding to the time at which the spot center is moving most directly away from the observer @xmath128 . as mentioned above , this is _ not _ the same as the point of maximum redshift , which occurs closer to @xmath129 due to gravitational deflection of the emitted light . the spectrum shown in figure [ plotfive ]
can be integrated in time to give a spectrum similar to those shown in figures [ plotthree ] and [ plotfour ] , corresponding to something like a very narrow circular emitting region with @xmath130 . by integrating over frequency
, we get the total x - ray flux as a function of time , i.e. the light curve @xmath131 .
this time - varying signal can be added to a background intensity coming from the inner regions of a steady state disk described in section 3 . by definition
the hot spot will have a higher temperature or density and thus greater emissivity than the background disk , adding a small modulation to the total flux .
rxte observations find the hfqpo x - ray modulations to have typical amplitudes of 1 - 5% of the mean flux during the outburst @xcite .
@xcite present a first - order argument that a 1% amplitude modulation requires a hot spot with 100% overbrightness extending over an area of 1% of the steady - state region of the disk . for @xmath94 , this requires a hot spot with radius @xmath132 , which they argue is too large to survive the viscous shearing of the disk .
hydrodynamic stability aside , this reasoning ignores the important effects of disk inclination , relativistic beaming and gravitational lensing of the hot spot emission . also , assuming a shakura - sunyaev type disk with steady - state emissivity @xmath133 and a similar scaling for the hot spot emission , we find that hot spots with smaller size and overbrightness are capable of creating x - ray modulations on the order of 1% rms , defined by @xmath134 ^ 2dt } { \int i(t)dt}}.\ ] ] figure [ plotsix ] shows the required overbrightness of a flattened gaussian hot spot orbiting near the isco to produce a modulation with rms amplitude of 1% for a range of inclinations and black hole spin parameters . in the limit of a face - on accretion disk @xmath135 , even an infinitely bright spot on a circular orbit will not produce a time - varying light curve . as the inclination increases , the required overbrightness decreases , as the special relativistic beaming focuses radiation toward the observer from a smaller region of the disk , _ increasing _ the relative contribution from the hot spot . as the spin of the black hole increases , the isco moves in toward the horizon and the velocity of a trajectory near that radius increases , as does the gravitational lensing , magnifying the contribution from the hot spot . understandably , the required overbrightness is inversely proportional to the area of the hot spot so @xmath136*r_{\rm spot}^2 = \rm const$ ] .
for example , from figure [ plotsix ] we see that a black hole binary with inclination @xmath137 and spin @xmath138 would require a spot size of @xmath139 with 67% overbrightness ( e.g. 14% temperature excess ) to produce a 1% rms modulation in the light curve .
this is well within the range of the typical size and magnitude of fluctuations predicted by 3d mhd calculations of 3-dimensional accretion disks @xcite .
the hot spot model is well - suited for simplified calculations of the x - ray emission from these random fluctuations in the accretion disk . by adding the emission from small , coherent hot spots to a steady - state shakura - sunyaev disk
, we can interpret the amplitudes and positions of features in the qpo spectrum in terms of a model for the black hole mass , spin , and inclination .
considering the x - ray flux from the hot spot alone , the frequency - integrated light curves for a variety of inclinations are shown in figure [ plotseven ] .
all light curves are shown for one period of a hot spot orbiting a schwarzschild black hole at the isco .
as the inclination increases , the light curve goes from nearly sinusoidal to being sharply peaked by special relativistic beaming .
thus the shape of a qpo light curve may be used to determine the disk inclination . with current observational capabilities ,
it is not possible even for the brightest sources to get a strong enough x - ray signal over individual periods as short as 3 - 5 ms to be able to differentiate between the light curves in figure [ plotseven ] . instead
, the fourier power spectrum can be used to identify the harmonic features of a periodic or quasi - periodic light curve over many orbits .
disks with higher inclinations will give more power in the higher harmonic frequencies , due to the `` lighthouse '' effect , as the hot spot shoots a high - power beam of photons toward the observer once per orbit , approximating a periodic delta - function in time .
figure [ ploteight]a shows a sample section of such a light curve , including only the x - ray flux from the hot spot , subtracting out the steady - state flux from the disk .
the sharp peaks in the light curve , while unresolvable in the time domain , will give a characteristic amount of power in the higher harmonics , shown in figure [ ploteight]b .
here we have normalized the rms amplitudes to the background flux from the disk with a hot spot overbrightness of 100 % . for a signal @xmath131 with fourier components
@xmath140 : @xmath141 we define the rms amplitude @xmath142 in each mode @xmath143 as @xmath144 with this normalization , the rms defined in equation ( [ rms ] ) can be conveniently written @xmath145 in figure [ ploteight]b , the main peak at @xmath146 hz corresponds to the fundamental azimuthal frequency for an orbit at the isco of a @xmath147 schwarzschild black hole . in the limit where the light curve is a periodic delta - function in time
, there should be an equal amount of power in all harmonic modes , because the fourier transform of a periodic delta - function is a periodic delta - function .
however , even in the limit of edge - on inclination @xmath148 , unless the hot spot is infinitessimally small and ultra - relativistic , the light curve will always be a continuous function with some finite width and non - zero minimum , thus contributing less and less power to the higher harmonics . the harmonic dependence on inclination for a hot spot orbiting a schwarzschild black hole
is shown in figure [ plotnine ] .
predictably , as the inclination increases , we see that both the absolute and relative amplitudes of the higher harmonics increase , almost to the limit of a periodic delta - function when @xmath149 .
interestingly , we find very little dependence of the harmonic structure on hot spot size or shape , as long as the total emission of the spot relative to the disk is constant ( this emphasizes the robustness of the simple hot spot model in interpreting an x - ray power spectrum , without needing to include the detailed physics of the disk perturbations ) . as the spin parameter increases for kerr black holes , the isco moves closer to the horizon , increasing the circular velocities of particles on the isco and thus the doppler shifts , giving broader time - integrated spectra , as seen in figure [ plotfour ] .
the phase lag in time of the peak blueshift with respect to angular phase of the hot spot is also amplified for these smaller values of @xmath99 , giving light curves that are asymmetric in time .
the hot spot orbits around kerr black holes also experience more gravitational lensing as they move through the strongly curved space - time near the horizon .
this lensing becomes more significant for larger inclinations as the black hole and the emitting region approach collinearity with the observer , producing an einstein ring as @xmath149 .
this somewhat counters the special relativistic beaming effect as the light curve peaks over a broader phase of each period due to lensing .
thus the harmonic power dependence on inclination for a kerr light curve is in fact smaller than that for the schwarzschild geometry .
however , in the event that the accretion disk is actually rotating in a retrograde fashion with respect to the black hole spin , the isco will move _ away _ from the horizon @xmath150 $ ] , reversing the above effects and giving a _ stronger _ dependence on inclination .
one of the major unsolved puzzles motivating theoretical models of black hole qpos is the observation of multiple peaks in the high frequency power spectrum @xcite . as discussed above
, any non - sinusoidal light curve will contribute to fourier power in harmonics at integer multiples of the fundamental kepler frequency .
however , for at least two x - ray binary systems ( xte j1550 - 564 and gro j1655 - 40 ; possibly also grs 1915 + 105 ) , peaks are found with rational ( but non - integer ) frequency ratios @xcite . in these particular examples ,
significant power is measured around the frequencies ( 184 , 276 hz ) for xte j1550 - 564 and ( 300 , 450 hz ) for gro j1655 - 40 , almost exactly a 2:3 commensurability in frequencies , while grs 1915 + 105 has peaks at 40 and 67 hz .
following the work of @xcite , we investigate the possibility of these commensurabilities coming from integral combinations of the radial and azimuthal coordinate frequencies of nearly circular geodesics around a kerr black hole . in a newtonian point mass potential ,
the radial , azimuthal , and vertical ( latitudinal ) frequencies @xmath151 and @xmath152 are identical , giving closed planar elliptical orbits . for the schwarzschild metric
the vertical and azimuthal frequencies are identical , giving planar rosette orbits that are closed only for a discrete set of initial conditions .
the kerr metric allows three unique coordinate frequencies , so geodesic orbits in general can fill out a torus - shaped region around the black hole spin axis . when these coordinate frequencies are rational multiples of each other , the trajectories will close after a finite number of orbits .
while there is currently no clear physical explanation for why hot spots may tend toward such trajectories , some recent evidence suggests the possible existence of nonlinear resonances near geodesic orbits with integer commensurabilities @xcite .
another important clue may come from the fact that at these special orbits are closed , perhaps enhancing the stability of non - circular trajectories .
the quasi - periodic nature of the oscillations suggest the continual formation and subsequent destruction of hot spots _ near _ , but not exactly at , the resonant orbits . for the purposes of this paper
, we will take the apparent preference for such orbits as given and concern ourselves primarily with calculating the resulting light curves and power spectra . in geometrized units with @xmath153 ,
coordinate time is measured in units of @xmath154 sec . for example , an orbit with angular frequency @xmath155 around a @xmath147 black hole would have an observed period of 3.1 ms , whereas the analogous orbit around a supermassive black hole with mass @xmath156 would have a period of 8.6 hours . in these units ,
the three fundamental coordinate frequencies for nearly circular orbits are given by equation ( [ omega_phi ] ) and @xcite @xmath157^{1/2}\ ] ] and @xmath158^{1/2},\ ] ] where the upper sign is taken for prograde orbits and the lower sign is taken for retrograde orbits .
the radial frequency approaches zero at @xmath159 , where geodesics can orbit the black hole many times with steadily decreasing @xmath34 . in the limit of zero spin and large @xmath34 ,
the coordinate frequencies reduce to the degenerate keplerian case with @xmath160 . to model the 2:3 frequency commensurability , we begin by looking for perturbed circular planar orbits where the radial frequency @xmath161 is one - third the azimuthal frequency @xmath162 .
since the orbits are nearly circular , the fundamental mode of the light curve should peak at the azimuthal frequency with additional power in beat modes at @xmath163 .
for @xmath161:@xmath162 = 1:3 , the power spectrum should have a triplet of peaks with frequency ratios 2:3:4 .
these commensurate orbits can be found easily from equations ( [ omega_phi ] ) and ( [ omega_r ] ) and solving for @xmath34 : @xmath164 figure [ plotten ] shows the radius ( solid lines ) of these special orbits as a function of spin parameter . also shown ( dashed line ) is the inner - most stable circular orbit for prograde trajectories .
the position of the 1:3 commensurate orbits follows closely outside the isco , suggesting a connection between the high frequency qpos and the black hole isco .
however , other integer commensurabilities such as 1:2 , 2:5 , or 1:4 also closely follow the isco curves for varying @xmath48 , so the proximity to the isco alone is probably not enough to explain the hot spot preference for these specific coordinate frequencies .
it is important to note that any given black hole source will have a constant value of @xmath48 , certainly over the lifetime of our observations .
thus , we may need to observe many more sources like xte j1550 - 564 and gro j1655 - 40 in order to better sample the parameter space of figure [ plotten ] and thus the connection between certain preferred orbits and the black hole isco .
a 1:3 commensurate trajectory moves through three revolutions in azimuth for each radial period , forming a closed rosette of three `` layers . '' for such rosettes , the eccentricity can be defined as @xmath165 the time - dependent light curve for an prograde orbit with eccentricity @xmath166 , spin @xmath138 , and inclination @xmath167 is shown in figure [ ploteleven]a .
the time axis begins at the point when the hot spot is at apocenter , moving away from the observer .
thus the first and third peaks come from the hot spot moving toward the observer at a relatively larger radius , while the second , higher peak is caused by the emitter moving toward the observer through pericenter at a higher velocity , giving a larger blue - shift and thus beaming factor .
the combined doppler beaming and gravitational lensing causes the peak following the pericenter peak to be slightly larger , as the emitter is moving away from the black hole yet the light is focused more toward the observer . the power spectrum for this light curve is shown in figure [ ploteleven]b , with the strongest peaks at the azimuthal frequency of @xmath168 hz and its first harmonic at @xmath169 hz for @xmath170 .
even for this modest deviation from circularity , there is significant power in the frequencies @xmath163 .
the beating of the fundamental @xmath162 with the radial frequency @xmath171 hz gives the set of secondary peaks at @xmath172 and @xmath173 .
additional peaks occur at beats of the harmonic frequencies @xmath174 .
it is interesting to note that there is not significant power in the radial mode at @xmath175 hz , but only in the beats with the fundamental azimuthal frequency and its harmonics .
however , in the limit of a face - on orientation @xmath176 , the radial frequency should dominate the light curve variation as the gravitational and transverse doppler red - shift modulate the intensity as a function of the hot spot s radial coordinate .
the radial mode should also be present in the limit of an edge - on orientation @xmath177 , as gravitational lensing becomes more important , and the hot spot will experience more magnification when closer to the black hole .
figure [ plottwelve ] shows the dependence on inclination of the lower order harmonic and beat modes . at low inclinations ,
the radial frequency @xmath178 contributes significant power , while at higher inclinations , the first harmonic of the azimuthal mode begin to dominate with similar behavior to the circular orbits shown in figure [ plotnine ] .
along with increasing power at @xmath179 , there is also increasing power in the radial beats of the first harmonic at @xmath180 . for a @xmath147 black hole , all these frequencies should be observable within the timing sensitivity of rxte . to further explore the constraints of our model
, we investigated the effect of orbital eccentricity on the qpo power . maintaining a 1:3 commensurability between radial and azimuthal frequencies , we calculated the light curves for a range of eccentricities @xmath181 .
as expected , the beat modes at @xmath182 have more power for more eccentric orbits , as the radial variation of the emitter becomes larger . at the same time ,
the first harmonic at @xmath183 provides relatively less power with increasing eccentricity .
this is best understood as the `` picket fence '' character of the light curve becomes modulated in amplitude and frequency from peak to peak , i.e. for each 3-peak cycle , the time between peaks 1 - 2 , 2 - 3 , and 3 - 1 are not all the same , damping the harmonic overtone .
interestingly , the fourier power in the beat modes appears to saturate at a moderate eccentricity of @xmath184 , suggesting highly eccentric geodesics are not necessary or even helpful for producing power in the beat frequencies . while there is significant evidence for higher frequency harmonic and beat modes in the qpo power spectrum of xte j1550 - 564 , the fourier power
is clearly dominated by the two frequencies @xmath185 and @xmath186 hz @xcite , corresponding to @xmath187 and @xmath162 in our model .
what are the physical mechanisms that could damp out the higher frequency modes ?
one possible explanation is in the geometry of the hot spot .
as explained in section 4 , in order to produce the power observed in qpos , the relative flux coming from the hot spot must be some finite fraction of that of the disk ( typically @xmath188 for a qpo amplitude of @xmath189 ) , so the hot spot must have some minimum size or it would not produce enough emission to be detected above the background disk . yet
if the hot spot is too large , it would be sheared by differential rotation in the accretion disk and not survive long enough to create the coherent x - ray oscillations that are observed . as mentioned above
, we find that the relative qpo power in different modes is not sensitive to the size of the hot spot @xmath190 .
three - dimensional mhd simulations @xcite show a range of density and temperature fluctuations consistent with the hot spot size and overbrightness factor predicted by our model in conjunction with the observations .
it also appears from simulations that as the hot spot is formed in the disk , differential rotation will tend to shear a finite region of gas as it follows a geodesic orbit around the black hole , modifying the shape of the hot spot into an arc in azimuth . in the limit that the emission region could be sheared into a ring of arc length @xmath191
, the fundamental mode and its harmonics would be essentially removed , leaving power only in the radial modulation .
indeed , as shown in figure [ plotthirteen]a , for an arc length of @xmath192 , the higher frequency modes at @xmath193 and @xmath194 are strongly suppressed , while still maintaining a significant amount of power in the fundamental beat modes @xmath195 .
the total qpo power also increases as the area of the emission region increases relative to the circular hot spot geometry .
however , when we allow the arc to be sheared into a ring with @xmath196 , the total qpo power is actually decreased as the differential beaming is essentially eliminated by the extended emission region : there is always some portion of the arc moving toward the observer .
the resulting modulation is then more weighted to the first radial beat mode at @xmath187 , as seen in figure [ plotthirteen]b .
it is not intuitively obvious why the @xmath187 mode is dominant while the @xmath197 mode ( @xmath72 = 368 hz ) is much weaker in the arc geometry .
if anything , this is a strong argument for the necessity of a full ray - tracing calculation of the hot spot light curves when predicting qpo power spectra , as it clearly gives information unavailable to simple analysis of the geodesic coordinate frequencies .
this behavior offers a plausible explanation for the two major types of qpos described in @xcite .
type a outbursts have more total power in the hfqpos , with a major peak at 276 hz and a minor peak at 184 hz .
type b outbursts have most of the qpo power around 184 hz and a smaller peak around 276 hz and less overall power in the high frequency region of the spectrum .
thus we propose that type a outbursts are coming from more localized hot spot regions , while type b outbursts come from a more extended ring geometry .
in addition to the commensurate high - frequency qpos observed in sources like xte j1550 - 564 , there are also strong low - frequency qpos observed at the same time with frequencies in the range @xmath198hz .
there have been suggestions that these low - frequency qpos may be caused by the lens - thirring precession of the disk near the isco , also known as `` frame dragging '' @xcite . for geodesic orbits out of the plane
perpendicular to the black hole spin , the latitudinal frequency @xmath199 of massive particles is not equal to the azimuthal frequency @xmath200 [ see eqs.([omega_theta ] ) and ( [ omega_phi ] ) ] , leading to a precession of the orbital plane with frequency @xmath201 for black holes with the mass and spin used above @xmath202 , the frame - dragging frequency , as calculated at the radius corresponding to the commensurability @xmath161:@xmath162=1:3 , is somewhat higher than that observed in the low - frequency qpos from xte j1550 - 564 .
the type a qpo peaks at 12 and 276 hz appear to be consistent with a black hole mass of @xmath203 and spin parameter of @xmath204 , quite similar to the values used throughout much of this paper . for the bh binary gro j1655 - 40 , we can fit the qpos at 18 and 450hz with a mass of @xmath205 and spin @xmath206 , slightly less than the published mass range of @xmath207 @xcite .
if we relax the requirement of matching the lfqpos and only fit the hfqpos with a 1:3 coordinate frequency commensurability , there remains a 1-dimensional degeneracy in the mass - spin parameter space .
based solely on the hfqpos , for xte j1550 - 564 with @xmath208 , the range of spin parameters would be @xmath209 and for gro j1655 - 40 , the spin would be in the range @xmath210 .
these results are shown in table [ tableone ] .
to get a more quantitative feeling for the effect of lens - thirring precession on the power spectrum , we investigated hot spot orbits with initial trajectories inclined to the plane of the disk : @xmath211 .
this is much like changing the observer s inclination with a period of @xmath212 .
thus we see additional modulation in the hot spot light curve at the `` double - beat '' modes @xmath213 .
we find that , for modestly inclined hot spot orbits @xmath214 , the contribution to the power spectrum at lens - thirring frequencies is quite small ( @xmath215 of total power ) for the basic circular hot spot geometry .
this relative contribution increases with arc length as the spot becomes a ring precessing about the spin axis , consistent with the relative power in lfqpos and hfqpos in the type a ( more high frequency power than low frequency ) and type b ( more low frequency power ) sources described above . under the premise that the hfqpo commensurate frequencies are caused by the geodesic motion of a sheared , overbright region in the disk , in table [ tabletwo ] we show the best fit parameters for the type a and type b qpo spectra from xte j1550 - 564 [ cf .
table 1 in @xcite ] . guided also by the ( somewhat speculative ) assumption that the lfqpos come from the lens - thirring precession of the hot spot orbital plane
, we predict a black hole mass and spin . using a fixed inclination of @xmath216
, we can match the frequencies and amplitudes of the observed hfqpo peaks ( and , equally important , the lack of power at certain frequencies ) for both type a and type b outbursts .
setting constant the eccentricity @xmath217 , geodesic inclination to the disk @xmath218 , and the overbrightness to be a factor of unity , we fit the hot spot radius and arc length to match the observations .
being able to match the qpo rms amplitudes of the peaks ( or lack thereof ) at 92 , 184 , 276 , and 368 hz , for at least two different types of x - ray outburst , shows the robustness of our simple model in explaining these phenomena .
however , even with the sheared arc emission , many observations still show significantly more power in the lfqpos than can be explained solely from the lens - thirring precession of a geodesic near the isco .
coupled with the difficulty in simultaneously fitting the mass and spin to three coordinate frequencies in a manner consistent with spectroscopic mass predictions , it seems likely that the lfqpos may be caused by some other mechanism in the disk that is related only indirectly to the high - frequency hot spot emission .
another possibilty is that the thin , warped disk model breaks down near the isco , allowing more complicated emission geometries and thus amplifying the effects of lens - thirring precession .
we have developed a hot spot model to help understand the high - frequency qpos observed in accreting black holes .
while the model does not consider radiation or hydrodynamic pressures , it is fully relativistic and calculates exact geodesics for both massive particles and photons . treating the hot spot as a monochromatic , isotropic emitter ,
we have calculated the time - dependent spectra of x - ray lines produced as the hot spot orbits the black hole .
the model makes a number of general predictions of the fourier power of the x - ray light curve as a function of inclination and black hole spin , and is also able to explain qpo observations from the black hole binaries xte j1550 - 564 and gro j1655 - 40 . simply by matching the locations of the low - frequency and high - frequency qpos with the coordinate frequencies ( under the condition @xmath219 )
, we can determine the black hole mass and spin .
relaxing the lfqpo constraint , the spin can still be determined uniquely for a given mass , which in turn can often be measured independently with the inclination and radial velocity of the companion star . by matching the amplitudes of various peaks observed in xte j1550 - 564 ,
we have explored the model parameters such as the hot spot size , shape , and the overbrightness relative to a steady - state background disk .
the predicted magnitude of these fluctuations are within the range predicted by 3d mhd calculations of the accretion disk .
future work will investigate the effect of multiple hot spots of various size , emissivity , and lifetime , as guided by mhd calculations .
observations of additional sources with commensurate frequency qpos may help us further constrain the hot spot model and better understand the connection between the lfqpos and hfqpos .
some of the physical problems with the hot spot model have been raised by @xcite , as discussed above .
many of these points are addressed in our model .
first , unlike @xcite , we only attempt to explain a set of qpo data from _ black hole _ binaries , which differ qualitatively from neutron star binaries in many ways , e.g. lacking strong global magnetic fields and thermonuclear activity . and perhaps most significant , black holes have no rotating surface to interact with the accreting matter and provide additional confusion to the qpo power spectrum .
our model produces light curves with power spectra consistent with black hole observations even with low eccentricity hot spot orbits .
because they do not include ray tracing in their calculations , @xcite are unable to model many relativistic effects , including the gravitational lensing of the hot spot source , which can be quite significant for systems with moderate to high inclination angles @xmath220 .
since we calculate the actual x - ray modulation from the orbiting hot spots , we predict both the location and amplitude of every peak in the light curve power spectrum , which can not be done by analyzing the bh coordinate frequencies alone . by introducing a perturbation on circular orbits near the isco ,
additional peaks begin to appear in the power spectrum , caused by beats of the azimuthal and radial frequencies @xmath162 and @xmath161 .
the dependence of the relative power in the different peaks on inclination and spin helps to constrain the details of the hot spot model in explaining the hfqpos , particularly the 2:3 commensurability observed in the power spectra from xte j1550 - 564 and gro j1655 - 40 . as an additional parameter
, we introduce a finite arc length for the emission region , motivated by the shearing of the hot spot by differential rotation in the disk .
the spreading of the hot spot in azimuth leads to suppression of the higher qpo modes , in agreement with observations .
we have also examined the possibility of lens - thirring precession for non - planar orbits as an explanation for the low - frequency qpos that have been observed coincident with the hfqpos yet often with even stronger fourier power .
the predicted power spectra from non - planar precessing arcs are consistent with observations of xte j1550 - 564 if we associate type a outbursts with hot spot arcs of @xmath221 and type b outbursts with hot spot rings of @xmath222 .
however , the difficulty in matching the lfqpos amplitude and frequency with a single hot spot geodesic suggests the low - frequency modulations may be caused by a different mechanism or perhaps our disk geometry is too simplistic .
one major remaining issue with the hot spot model is the preferred location of the geodesic that gives rise to 1:3 coordinate frequencies .
why should the orbital frequencies favor integer ratios , and why should the preferred ratio be 1:3 and not 1:2 or 1:4 ? it is possible that detailed radiation - hydrodynamic calculations with full general relativity will be required to answer this question .
perhaps the non - circular orbits can only survive along closed orbits such as these to somehow avoid destructive intersections . or there may be magnetic interactions with the black hole itself , analogous to the blandford - znajek process , that lock the accreting gas into certain preferred trajectories @xcite . for now
, we are forced to leave this as an open question unanswered by the geodesic hot spot model . a less difficult , yet
still unanswered problem is the explanation of the widths of the qpo peaks .
as it stands , our hot spot model predicts purely periodic light curves and thus power spectra made up of delta functions .
if there _ is _ some physical mechanism that preferentially focuses accreting material onto eccentric orbits at specific radii , then it is likely that these hot spots are forming and then being destroyed as a continual processes .
the superposition of many hot spots around the same orbit , all with slightly different initial trajectories , could explain the quasi - periodic nature of the power spectrum : the phase decoherence of the hot spots would cause a natural broadening of the strictly periodic signal from a single spot . with the computational framework in place , this question can be answered by modeling a whole collection of hot spots and arcs continually forming and evolving in shape and emissivity .
the particular physical parameters for these hot spots can be derived from published mhd calculations such as @xcite . with the basic ray - tracing and radiation transport methods in place
, it should be possible to use our code as a `` post - processor '' to analyze the results of these 3d simulations to simulate x - ray light curves and spectra from a realistic accretion disk .
a final piece of the black hole qpo puzzle is the spectral behavior of the source during outburst . @xcite
have shown there is a large range of x - ray fluxes with different relative contributions from a power - law component and a disk bolometric component .
these relative and absolute fluxes seem to be correlated with the amount of power in both the lfqpos and the hfqpos . future work on 3-dimensional disks and more detailed radiative transfer models should give us important insights into understanding this spectral behavior and its relation to the qpo power .
abramowicz , m. a. , & kulzniak , w. 2001 , a&a , 375 , l19 abramowicz , m. a. , et al .
2002 , preprint ( gr - qc/0202020 ) abramowicz , m. a. , et al .
2003 , pasj , in press ( astro - ph/0302183 ) armitage , p. j. , reynolds , c. s. , & chiang , j. 2001 , , 548 , 868 bao , g. 1992 , astron . & astro . 257 , 594 bardeen , j. m. , press , w. h. , teukolsky , s. a. 1972 , 178 , 347 blandford , r. d. , & znajek , r. l. 1977 , mnras , 179 , 433 bromley , b. c. , chen , k. , & miller , w. a. 1997 , , 475 , 57 carter , b. 1968 , phys .
26 , 331 ford , e. c. 2000 , 535 , l119 ford , e. c. et al .
2000 , 539 , 368 ghosh , p. , & lamb , f. k. 1978 , , 223 , l83 gierlinsky , m. , done , c. , barret , d. 2002 , mnras 331 , 141 hawley , j. f. , & krolik , j. h. 2001 , , 548 , 348 karas , v. , vokrouhlicky , d. , & polnarev , a. g. 1992 , mnras , 259 , 569 kojima , y. 1991 , mnras 250 , 629 kluzniak , w. , & abramowicz , m. a. 2001 , preprint ( astro - ph/0105057 ) kluzniak , w. , & abramowicz , m. a. 2002 , preprint ( astro - ph/0206063 ) lamb , f. k. 2002 , in _ x - ray binaries and gamma - ray bursts _ , ed .
e. p. j. van den heuvel ( san francisco : astronomy society of the pacific ) lanzafame , g. , molteni , d. , & chakrabarti , s. k. 1998 , mnras , 299 , 799 laor , a. 1991 , , 376 , 90 lee , w. h. , & ramirez - ruiz , e. 2002 , , 577 , 893 markovic , d. , & lamb , f. k. 1998 , , 507 , 316 markovic , d. , & lamb , f. k. 2000 , preprint ( astro - ph/0009169 ) merloni , a. , vietri , m. , stella , l. , & bini , d. 1999 , mnras , 304 , 155 miller , j. m. et al .
2002 , , 570 , l69 misner , c. w. , thorne , k. s. , & wheeler , j. a. 1973 , _ gravitation _
( new york : w. h. freeman ) pariev , v. i. , bromley , b. c. , miller , w. a. 2001 , , 547 , 649 rauch , k. p. , & blandford , r. d. 1994 , , 421 , 46 remillard , r. a. et al .
2002 , , 580 , 1030 reynolds , c. s. , & begelman , m. c. 1997 , , 488 , 109 shahbaz , t. , et al . 1999 , mnras , 306 , 89 shakura , n. i. , & sunyaev , r. a. 1973 , astron .
, 24 , 337 shapiro , s. l. , & teukolsky , s. a. 1983 , _ black holes , white dwarfs , and neutron stars _ ( new york : john wiley & sons , inc . ) stella , l. , & vietri , m. 1998 , , 492 , l59 stella , l. , & vietri , m. 1999 , phys .
82 , 17 strohmayer , t. e. 2001 , , 552 , l49 strohmayer , t. e. 2001 , , 554 , l169 wagoner , r. v. , silbergleit , a. s. , & ortega - rodriguez , m. 2001 , , 559 , l25 wang , d. x. 2003 , mnras , in press ( astro - ph/0305478 ) schematic picture of ray - tracing method from distant observer through a disk of angular thickness @xmath223 .
the rays either terminate at the black hole horizon or pass through the disk , with each point of intersection labeled with the photon position and momentum @xmath224 . ]
projection of a uniform cartesian grid in the image plane onto the source plane of the accretion disk @xmath225 .
inclination angles are @xmath53 @xmath226 and @xmath167 @xmath227 and spin parameters are @xmath228 @xmath229 and @xmath230 @xmath231 .
the region inside the horizon is cut out from each picture . ]
normalized spectra of a monochromatic emission line from steady - state accretion disks of varying inclination .
an inclination of @xmath53 corresponds to a face - on view of the disk while @xmath54 would be edge - on .
the emissivity is taken to be uniform between @xmath232 and @xmath233 .
the spin is taken to be @xmath138 but the dependence on @xmath48 is negligible for uniformly emitting disks . ]
normalized spectra of steady - state accretion disks with inclination @xmath234 and varying spin parameter @xmath48 ( negative values of @xmath48 correspond to retrograde disk rotation ) .
the emissivity is taken to be proportional to @xmath103 between @xmath235 and @xmath233 .
black holes with higher values of @xmath48 allow the inner disk to extend in closer to the horizon , giving a greater weight to the high - redshift radiation emitted there . ]
spectrogram of a circular hot spot with radius @xmath236 orbiting a schwarzschild black hole at the isco @xmath237 ) , viewed at an inclination of @xmath238 .
the spot is moving in the @xmath239 direction with the observer at @xmath125 .
the maximum redshift occurs when @xmath240 and the maximum blueshift occurs when @xmath241 . ] overbrightness required of a hot spot on a circular orbit to produce a 1% rms modulation in x - ray flux when added to a steady - state disk with @xmath95 , @xmath94 and emissivity @xmath242 .
an overbrightness of unity means the peak hot spot emissivity is twice that of the steady - state disk with no hot spot .
the spot size @xmath190 is measured in gravitational radii @xmath1 , so for a black hole with @xmath138 and @xmath167 , the required overbrightness for a hot spot with @xmath243 would be 67% . ] frequency - integrated light curves of an orbiting hot spot at the isco of a schwarzschild black hole for different disk inclination angles .
the spot is moving in the @xmath239 direction as in figure [ plotfive ] .
for high inclination angles , the special relativistic beaming causes the light curve to become sharply peaked as the hot spot moves toward the observer . ]
fourier amplitude @xmath142 in higher harmonic frequencies @xmath244 as a function of orbital inclination to the observer , normalized as in equation ( [ arms ] ) .
the hot spot has size @xmath243 , an overbrightness factor of 100% , and is in a circular orbit at @xmath99 around a schwarzschild black hole . ]
radius of prograde orbits with commensurate frequencies @xmath245(1:3 , 1:2 , 2:3 ) ( solid lines ) as a function of dimensionless spin parameter @xmath246 .
the isco ( dashed line ) corresponds to @xmath247 . also shown
are the respective orbital frequencies @xmath162 at these radii for a black hole with mass @xmath147 ( dot - dashed lines ) . ]
power in low - order harmonics and beat modes with frequencies @xmath248 , as a function of disk inclination angle .
the hot spot trajectory is the same as in figure [ ploteleven ] , with @xmath142 normalized as in figure [ plotnine ] .
the curves are labeled by the ratio @xmath249 . ]
|
observations from the rossi x - ray timing explorer have shown the existence of high frequency quasi - periodic oscillations ( hfqpos ) in the x - ray flux from accreting black hole binary systems . in at least two systems ,
these hfqpos come in pairs with a 2:3 frequency commensurability .
we propose a simple `` hot spot '' model to explain the position and amplitude of the hfqpo peaks .
using the exact geodesic equations for the kerr metric , we calculate the trajectories of massive test particles , which are treated as isotropic , monochromatic emitters in their rest frames .
photons are traced from the accretion disk to a distant observer to produce time- and frequency - dependent images of the orbiting hot spot and background disk .
the power spectrum of the x - ray light curve consists of multiple peaks at integral combinations of the black hole coordinate frequencies .
in particular , if the radial frequency is one - third of the azimuthal frequency ( as is the case near the inner - most stable circular orbit ) , beat frequencies appear in the power spectrum at two - thirds and four - thirds of the fundamental azimuthal orbital frequency , in agreement with observations .
we also model the effects of shearing the hot spot in the disk , producing an arc of emission that also follows a geodesic orbit , as well as the effects of non - planar orbits that experience lens - thirring precession around the black hole axis . by varying the arc length
, we are able to explain the different qpo features observed in `` type a '' and `` type b '' x - ray outbursts from xte j1550 - 564 . in the context of this model ,
the observed power spectra allow us to infer values for the black hole mass and angular momentum , and also constrain the parameters of the model , such as the hot spot size and luminosity .
|
quantum states populating open chaotic cavities decay to the outside environment and , thence , have the status of resonances . in spite of the ubiquity of the general setup
open quantum chaos is realized in many of the devices currently explored in mesoscopic physics , quantum optics , and cold atom physics salient features of these resonances are not fully understood . while the deep quantum regime ( the ehrenfest time , @xmath0 , marking the diffractive disintegration of minimal wave packages shorter than classical escape times , @xmath1 ) appears to be under reasonable control @xcite , it is the opposite , semiclassical limit which poses unsettled issues @xcite . broadly speaking , the states populating an open cavity can be grouped into two families : states evolving near classically and escaping deterministically after a classical flight time , and a fraction @xmath2 of quantum resonances , whose probabilistic decay is characterized by a finite imaginary offset @xmath3 to the real resonance energy @xmath4 . the most basic quantity characterizing the statistics of resonances of complex energy @xmath5 is the resonance density @xmath6 .
although the quantitative profile of that quantity is not fully understood , the density appears to be gapped against the real axis , @xmath7 ( the existence of rare midgap states notwithstanding @xcite . )
the integrated number of resonances at a given value of @xmath4 has been found to obey the so - called fractal weyl law , @xmath8 , where @xmath9 is a non - universal fractal exponent .
previous work on the phenomenon includes the formulation of lower bounds on the resonance gap @xcite , semiclassical approaches based on short periodic orbits _ trapped _ in the open system
@xcite , a description in terms of non - unitarily evolving husimi functions @xcite , phenomenology based on a mixture of phase space dynamics and random matrix theory , resp .
@xcite , and numerical analyses @xcite . however , a unified theory of resonance formation in terms of first principle semiclassical dynamics appears to be missing and the formulation of such a theory is the subject of the present work .
specifically , we will explore the quantum dynamics of states concentrated on classical trajectories in terms of phase space wigner functions .
assuming globally hyperbolic classical dynamics we will describe how quantum fluctuations operational on long trajectories convert the deterministic classical escape of short trajectories into probabilistic quantum decay .
our analysis is organized in three conceptual steps .
we first introduce the phase space language used in the rest of the paper on a one - dimensional toy model ( section [ sec1 ] ) .
we then generalize to the more complex setting of a higher dimensional cavity ( section [ sec2 ] ) , and derive effective boundary conditions determining the decay rates of the system .
finally , ( section [ sec3 ] ) we analyze these equations for both short and long trajectories .
we conclude in section [ sec : summary ] .
consider a one - dimensional cavity parameterized by the spatial coordinate @xmath10 $ ] , while coordinates to the right / left of @xmath11/@xmath12 define connecting leads ( see fig .
[ fig:0 ] ) .
we assume free intra - cavity particle dynamics , @xmath13 and , crucially , no backscattering barriers at the cavity / lead interfaces .
, width=245 ] life times and energies of the resonant states supported by the system may be calculated by matching solutions of the cavity schrdinger equation to outgoing boundary conditions @xcite , i.e. by requiring that cavity wave functions @xmath14 and their derivatives smoothly connect to vacuum wave functions @xmath15 at the right / left interface . here ,
@xmath16 , and @xmath17 is a complex wave vector whose real- and imaginary - part define the energy @xmath18 and life time @xmath19 of resonant states , resp . , where @xmath20 .
the divergence of the reference states at spatial infinity @xmath21 , is a formal means @xcite to the fixation of decay rates , as exemplified below . for the intra - cavity wave function
we make an ansatz @xmath22 in terms of left- and right - propagating partial amplitudes where @xmath23 , and the somewhat unconventional denotation @xmath24 and @xmath25 , for the real and imaginary contribution to the phase , resp .
, will be motivated shortly . with this choice ,
the boundary conditions obtained by matching wave functions and their derivatives at the left and right interface reduce to the single algebraic equation , @xmath26 before evaluating this equation , let us translate from the language of wave functions to a phase - space formulation . to this end
, we introduce the wigner function @xmath27 , where @xmath28 . for our specific system , @xmath29
where , @xmath30 labels the wigner transform of the left and right moving components , resp . , and
the ellipses denote rapidly oscillating interference contributions . in discarding the latter , we loose track of the global phase of the wave function , while the information on amplitudes and phase _ derivatives _ necessary to evaluate boundary conditions is retained .
indeed , it is straightforward to check that @xmath31 for the simple @xmath32 system , the linear dependence @xmath33 implies @xmath34 so that reduces to @xmath35 where we have fixed a gauge @xmath36 for the arbitrary phase of the wave function and neglected contributions @xmath37 . for the toy model at hand , @xmath38 , which means that the right hand side of vanishes , and @xmath39 is the only consistent solution .
this reflects the fact that a wave function will decay with probability unity upon passing the reflectionless boundaries of the system .
we next discuss how the situation changes upon generalization to a higher dimensional system with chaotic dynamics .
we consider a two - dimensional cavity with ballistic hamiltonian @xmath40 and chaotic boundary scattering .
the cavity is open such that after an average time @xmath1 , much shorter than any of the relevant quantum time scales , trajectories escape through one or several reflectionless openings .
we define the wigner function of the system s resonance states by obvious generalization of eq .
, i.e. @xmath41 . to obtain the intra cavity evolution equations of @xmath42 ,
one adds and subtracts the schrdinger equations of the resonances @xmath43 and @xmath44 to obtain @xcite @xmath45_+ = 2e\ , w , \quad \left [ h \stackrel{\ast } { , } w\right]_- = - i \gamma\ , w.\end{aligned}\ ] ] here , @xmath46 is the hamilton function and @xmath47_\mp = a\ast b\mp b\ast a$ ] where the moyal product of phase space functions @xmath48 is given by @xcite @xmath49 and @xmath50 is the poisson bracket .
we next consider the vicinity of an exceptionally long trajectory @xmath51 spending time @xmath52 inside the cavity .
for completeness we note that long trajectories in open systems are found with low probability @xmath53 .
they typically form in the phase space neighborhood of strange repellers realized through periodic orbits trapped in the interior of the cavity ( see below for further comments on this aspect ) . assuming global hyperbolicity of the dynamics , we introduce a trajectory coordinate , @xmath54 $ ] , a conjugate momentum @xmath55 transverse to the shell of conserved energy , and a pair @xmath56 of locally unstable and stable coordinates . in the asymptotic neighborhood of @xmath51 , the hamiltonian
can then be approximated as @xmath57 , where @xmath58 is a lyapunov exponent .
the corresponding dynamics is generated by @xmath59_-= i\hbar \{h_0,\;\}\equiv -i\hbar { \cal l}$ ] , where the liouvillian @xmath60 describes propagation in the direction of @xmath51 , and exponential expansion / contraction in the @xmath61 coordinate .
nonlinear corrections to @xmath62 can be described as @xmath63 , where @xmath64 is a polynomial of degree @xmath65 in the variables @xmath56 .
the corresponding modification of the dynamics , @xmath66_-\equiv -i\hbar ( \delta { \cal l}+{\cal q})$ ] , comprises a weak alteration of the classical liouvillian , @xmath67 , and a quantum generator @xmath68 where @xmath69 are coefficient functions whose detailed profile will not be of much importance throughout .
although both contributions are nominally small in @xmath56 , the quantum generator @xmath70 , will be seen to have a regularizing effect on classical singularities @xcite , which will ultimately shape the profile of the resonance density .
close to the trajectory , the first of eqs . , @xmath71_+\simeq \left [ h_0 \stackrel{\ast}{,}\ , w \right]_+\simeq 2 ( p^2/2 m ) w = 2e w$ ] simply describes the on - shell fixation @xmath72 . turning to the second equation ,
@xmath73 we first discuss the linear approximation , @xmath74 , before including the correction terms in a second step .
for @xmath75 , becomes a first order differential equation which is solved in terms of a left- and a right - moving contribution @xmath76 structurally similar to eq . .
here , @xmath77 are effective parameter functions generalizing @xmath78 of the toy model and evolving uniformly along the trajectories @xmath79 piercing the phase - space point @xmath80 , @xmath81 . to solve this ( partial first order differential ) equation , we consider its characteristics , i.e. the trajectory @xmath82 . on @xmath82 , the equation assumes the form @xmath83 , where @xmath84 , @xmath85 , @xmath86 and @xmath87 are starting values of the evolution .
we solve the characteristic equation as @xmath88 , where @xmath89 increases uniformly until @xmath90 hits the effective boundaries of the problem , and @xmath91 is a freely adjustable parameter .
( indicated by a straight line along the upper front corner of the box ) enters / exists the system through phase space interfaces @xmath92 .
spanning the neighborhood of @xmath51 by a stable / unstable coordinate @xmath93/@xmath94 , nearby trajectories contract towards / depart from @xmath51 in the respective coordinate directions .
the exit out of the cavity neighborhood of @xmath51 then is through the system interface @xmath95 ( solid trajectories ) , or through the internal interface , @xmath96 ( dashed trajectories . )
a similar distinction applies to the entry points .
depending on the entry / exit variants , each trajectory neighboring @xmath51 gets assigned entry / exit points @xmath97 , a time like progression parameter @xmath89 , and a phase space parameterization @xmath98 .
[ fig:1a ] , width=321 ] to understand the role of the boundaries , note that @xmath99 will leave the cavity either through a physical interface , @xmath100 along with @xmath101 ( cf .
[ fig:1 ] ) , or it will depart from @xmath51 up to some classical threshold @xmath102 within the cavity ( solid line ) .
we assume that points separated from @xmath51 by scales @xmath103 have become generic and will exit in the classical , and hence negligibly short time @xmath104 .
the union @xmath105 of @xmath100 and the surface @xmath106 then defines the effective ` outgoing interface ' of our problem .
similarly , the union @xmath107 of the left vacuum interface @xmath108 and the surface @xmath109 defines the incoming interface . the traveling phase space point @xmath110 hits the exit interface @xmath111 , at the smaller of two times , @xmath112 , depending on whether @xmath100 or @xmath96 is the terminal .
likewise , @xmath82 has entered the cavity through @xmath113 at a large negative time @xmath114 . fixing the free parameter @xmath91 such that the temporal range of the trajectory is symmetric around zero , @xmath115
, we find that the solution to eq .
( [ 0qcsol ] ) is governed by the function @xmath116 while @xmath117 is the intra cavity flight time of @xmath118 .
notice that for points @xmath119 at the exit interface , @xmath120 , meaning that @xmath121 attains its maximal value . finally , the boundary conditions eq . are generalized by replacing the one - dimensional variable @xmath78 by @xmath77 , and the derivative @xmath122 by @xmath123 , i.e. a derivative acting in the direction of the hamiltonian flow @xcite .
the generalization of eq .
thence reads @xmath124 where @xmath125 , and @xmath119 is on the exit interface .
is a principal result of the present paper . in the following we discuss its implications for different types of trajectories
. collapsed to the two - dimensional sections spanned by the stable and the unstable coordinate , and the trajectory parameter , @xmath126 , resp .
a phase - space point @xmath87 in the vicinity of @xmath51 propagates along a unique classical trajectory , @xmath118 .
it will exit the cavity either through the interface @xmath100 or within the cavity through the surface @xmath96 .
similarly , the union of the left vacuum interface @xmath108 together with the manifold @xmath127 defines the incoming interface .
[ fig:1 ] , width=321 ]
to start with , we consider trajectories @xmath118 which are exceptionally long - lived , @xmath128 , yet short in comparison to the scale @xmath129 @xcite where quantum uncertainty leads to the disintegration of semiclassically propagating wave packets . as long as @xmath130 , the quantum generator @xmath131 does not modify the dynamics in essential ways
a statement to be verified below and the same goes for the generator of weak classical corrections , @xmath132 .
we may thus take the boundary condition at face value , and conclude that due to the homogeneous evolution @xmath133 , the right hand side of the equation vanishes .
as with the toy model , this implies a diverging decay rate , @xmath39 . in this divergence
reflects the fact that wave packages travelling on classically short trajectories leave the cavity with certainty at the exit point .
we now turn to the case of long trajectories , @xmath134 . for
asymptotically long trajectories , @xmath135 , the classical shrinkage @xmath136 would lead to singularities in the function @xmath137 and , ultimately , in the wigner function . within the present formalism
, these singularities are regularized on time scales larger than @xmath0 , where @xmath138 has shrunk down to quantum scales . on these scales , the quantum generator @xmath131 of eq .
is no longer small in comparison to the classical generator @xmath123 ( while the correction @xmath139 to the classical flow continues to be largely irrelevant . )
the ensuing modifications of the dynamics can be described in various ways ( cf .
@xcite for a treatment tailored to the formalism applied here ) , the invariable conclusion being that the shrinkage of classically evolving variables gets cut off by quantum fluctuations .
technically , this conclusion rests on the observation that in the evolution equation for the variable @xmath94 the higher order derivatives @xmath140 present in the quantum generator , build up pressure counteracting the classical contraction .
this is seen in explicit terms in the fourier / laplace representation of the evolution equation , where these derivatives assume the form of algebraic factors , cutting the logarithmic ultraviolet singularities of the classical equation . referring to the appendix for more details
, we note that to leading semiclassical accuracy functions which in the classical theory evolve as @xmath141 get replaced by @xmath142 . here
, @xmath143 is symbolic notation for classical ( @xmath144-independent ) functions over which we have no explicit control , and the substitution @xmath145 becomes effectual in the large negative time asymptotics of a trajectory , where @xmath93 rather than @xmath94 scales to small values . to understand the consequences of this regularization mechanism , consider the trajectory time parameter , @xmath146 at the exit point of @xmath118 .
now notice that @xmath147 , where we used that , @xmath148 .
the crucial observation here is that the regularization effectively truncates the in - time function @xmath149 at values @xmath0 . as a consequence
, the interface derivative @xmath150 reduces to one half of the value before quantum regularization .
substitution of this value into eq .
shows that the quantum theory admits finite values of the decay constant , determined by @xmath151 where @xmath42 is the lambert function and ellipses denote subleading double-`@xmath152 ' contributions .
eq . states the decay rate in terms of the ehrenfest time in combination with non - universal short time cutoff @xmath153 .
however , in the semiclassical limit , @xmath154 , the dependence on @xmath4 drops out , and we are left with the asymptote @xmath155 . before commenting on this result
, we note that the appearance of a finite decay rate within our present formalism follows from the fact that , by heisenberg uncertainty , quantum mechanics is not capable of resolving the phase space fine structures pertaining to the evolution of long trajectories @xmath156 .
each such trajectory should , rather , be thought of as a distribution defined by the union of trajectories with uncertainty @xmath157 in their phase space coordinates . at a given instance of time , a fraction of this distribution escapes , as described by the rate @xmath158 .
our above analysis was oversimplifying in that it treated escape from an isolated long trajectory , @xmath51 , as tantamount to escape into the lead vacuum .
this picture ignores the fact that escaping trajectories may get folded back into the repeller domain supporting @xmath51 , and thence be trapped again .
a statistical theory accounting for the renormalization of the decay of an initial distribution centered around an isolated trajectory by the complex structure of the embedding repeller structure has been developed in refs .
the result of that analysis is an effective renormalization of decay rates as @xmath159 , where the factor @xmath160 effectively measures the fraction of trajectories managing to escape the repeller and @xmath161 is the fractal ( information ) dimension of the latter @xcite .
the ensuing effective rate , @xmath162 is generally identified with the inverse of the classical escape time of the system .
we finally caution that the decay rate will be subject to sources of _ fluctuations _ which are beyond the scope of our analysis .
notably , the lyapunov exponents may vary between trajectories , and along individual trajectories .
the escape from the repeller may introduce additional uncertainty .
our result , thus , yields a characteristic value for the decay rate , where the important role of fluctuations is left unaccounted for .
other effects not captured by our analysis include transient features of the classical dynamics _
outside _ the repeller s area which , as recent work shows @xcite , may have important influence on the resonances of open quantum system .
for completeness we note that a finite quantum mechanical decay rate is attributed to states located in the vicinity of exceptionally long trajectory . due to the exponential scarcity of these trajectories ,
the corresponding phase space measure scales as @xcite @xmath163 , where @xmath164 is the phase space volume of the energy shell ( in units @xmath144 ) , implying that @xmath165 with _ fractal _ dimension @xmath166 @xcite .
we finally note that in the quantum regime , @xmath167 , which is complementary to the semiclassical regime @xmath168 studied here , random matrix scattering theory predicts @xcite @xmath169 , where @xmath170 is some cut - off energy scale .
comparison with shows that the two results match at the boundary @xmath171 .
however , we can not say whether this matching is coincidental or not .
we have formulated a semiclassical theory of quantum escape processes in open chaotic systems .
the most important single contribution of our approach is that it quantitatively describes how deterministic escape after the traversal of generic short trajectories through the system gives way to _ quantum mechanical decay _ on long trajectories .
the latter define the support of resonances whose life times we estimated by imposing effective phase space boundary conditions .
somewhat counterintuitively , it turns out that the ensuing decay rates are classically short @xmath172 , although the relevant escape dynamics takes place on long trajectories @xmath134 .
finally , the escape of individual long trajectories as described in the present paper defines only an initializing stage of the decay of a more complicated repeller structure . as a result ,
the decay rate @xmath173 is subject to renormalization @xmath174 where @xmath161 is the fractal repeller dimension .
qualitatively , the renormalization factor accounts for the probability that a state gets re - captured by the repeller structure after escaping an individual trajectory .
however , a quantitative description of that secondary mechanism is beyond the scope of our approach .
we thank p. w. brouwer , s. nonnenmacher and h. schomerus for discussions and m. novaes for pointing out an inconsistency in the interpretation of our results as given in the first version of this paper .
work supported by sfb / tr 12 of the deutsche forschungsgemeinschaft .
we here discuss how quantum fluctuations regularize the unlimited classical contraction of the stable coordinate @xmath94 in a system with globally hyperbolic dynamics . in the language of eqs .
( 6 ) and ( 7 ) of the main text , the dynamics of the variable @xmath94 is described by a differential equation of the structure @xmath175 where @xmath176 , and in a manner inessential to the present argument the coefficients @xmath177 may depend on the variables @xmath178 .
considering positive starting values , @xmath179 ( the extension to negative values is straightforward ) , we introduce a laplace representation @xmath180 in which takes the form @xcite @xmath181 the general solution of this equation is found by straightforward integration over @xmath182 , and when inserted into gives @xmath183 with an integration constant @xmath184 .
. now illustrates the role played by higher differential operators in . to make the point ,
let us for the moment consider the first order differential equation obtained from by setting all @xmath185 .
the resulting function @xmath186 then displays the singular at small values of @xmath94 plaguing the classical evolution equation of the stable coordinate . in the full solution eq .
the exponential factor @xmath187 cuts the small @xmath94/large @xmath182 singularity at values @xmath188 .
the resulting integral can be estimated by a regularized function @xmath189 finally notice that our argument crucially relies on assumed positivity of the coefficients @xmath177 . while the present construction can not prove this feature , positivity is required on principal grounds to ensure stability of the dynamics .
( otherwise the wigner distribution would cease to exist . ) to actually demonstrate this stability , one has to work harder as in , e.g. , refs .
a discussion tailored to the present formalism is contained in ref .
[ ] .
notice that the appearance of @xmath190 in this context does _ not _ imply a ` classical approximation ' . rather , the first order differential operator @xmath190 describes the projection of the _ quantum _ current onto the outward direction of the dynamical flow at the cavity / lead interface .
|
states supported by chaotic open quantum systems fall into two categories : a majority showing instantaneous ballistic decay , and a set of quantum resonances of classically vanishing support in phase space .
we present a theory describing these structures within a unified semiclassical framework .
emphasis is put on the quantum diffraction mechanism which introduces an element of probability and is crucial for the formation of resonances .
our main result are boundary conditions on the semiclassical propagation along system trajectories . depending on whether the trajectory propagation time is shorter or longer than the ehrenfest time ,
these conditions describe deterministic escape , or probabilistic quantum decay .
|
since the large hadron collider ( lhc ) started its first successful high - energy operations in 2010 , hopes that new physics beyond the standard model ( sm ) would appear any moment were running high . with the large accumulated amount of proton - proton data at center - of - mass energies @xmath0 and 8 tev , more and more sophisticated searches for new physics came along . among the theoretical paradigms tested to a great extent are the recent models with extra spatial dimensions , either flat or curved , that appeared about a decade ago and quickly gained a lot of attention from both theoretical and experimental communities .
these models offer a different solution to the infamous hierarchy problem " that plagues the sm , and promise an exciting possibility of studying quantum gravity at the lhc , including the most mysterious of its manifestations : the black holes . in this review ,
the current status of searches for extra spatial dimensions and quantum gravity with the compact muon solenoid ( cms ) detector at the lhc is presented .
while so far all these searches came empty - handed , several new experimental methods and techniques have been developed , which also apply to other searches for new physics . among these techniques
are an empirical method to predict qcd background in high - multiplicity events from low - multiplicity samples , as well as dedicated methods of reconstructing objects with high lorentz boost , which may be the key signature for discovery of high - mass resonances predicted in a variety of beyond the sm ( bsm ) models , including the extra dimensional paradigm .
the negative results of these searches changed the very way we think about the naturalness " ( i.e. non - fine - tuned solutions to the hierarchy problem ) and paved the way to new strategies of looking for these phenomena , which will become possible once the design lhc energy is reached in the next few years .
several models with extra dimensions have appeared in recent years as a follow - up to the original , nearly century - old idea of kaluza and klein ( kk)@xcite to achieve a unification of electromagnetism and gravity by adding a compact fourth spatial dimension . while the kk model did not quite work , huge progress in string theory in the past quarter of century helped to revive the concept of extra dimensions and resulted in the modern attempts to utilize them to solve the hierarchy problem of the standard model , namely the very large ratio between the electroweak symmetry breaking ( ewsb ) and planck energy scales . among several modern models with extra dimensions , which result in a rich lhc phenomenology
, the following ones have been probed explicitly in the cms experiment : model with large extra dimensions by arkani - hamed , dimopoulos , and dvali @xcite ( add ) . in this model ,
only gravity propagates in @xmath1 flat dimensions assumed to be compactified with a radius @xmath2 either on a sphere or on a torus .
all the sm particles are confined on a 3d membrane ( brane " ) in the ( 4+n)-dimensional space - time .
apparent weakness of gravity ( i.e. , the large value of the planck mass , @xmath3 ) can then be explained due to the volume suppression of the fundamentally strong gravity in 3+n dimensions ( fundamental planck scale being @xmath4 tev ) , once one looks at its effects in 3 dimensions .
typical values of @xmath1 being considered are between 2 and 6 , with the @xmath5 already being significantly constrained by direct gravitaty measurements at short distances and from astrophysical and cosmological observations.@xcite the radius of extra dimensions depends on their number , and the relationship between @xmath6 and @xmath3 can be established via gauss s law:@xcite @xmath7 ( for extra dimensions compactified on a torus ) .
setting @xmath8 tev gives @xmath9 mm ( @xmath10 ) to 1 fm ( @xmath11 ) .
these values are _ large _ compared to the characteristic scales of particle physics interactions , hence the name : _ large extra dimensions_. randall - sundrum ( rs ) model @xcite with a single , warped " extra dimension , which is realized in five - dimensional anti - desitter space - time ( @xmath12 ) .
the metric of the @xmath12 space is given by @xmath13 , where @xmath14 is the coordinate along the compact dimension of radius @xmath2 , @xmath15 is the curvature of the @xmath12 space , often referred to as the warp factor " , @xmath16 are the conventional ( 3 + 1)-space - time coordinates , and @xmath17 is the metric tensor of minkowski space - time . two 3-dimensional branes with
equal and opposite tensions are positioned at the fixed points of the @xmath18 orbifold in the @xmath12 space , at @xmath19 ( sm brane ) and at @xmath20 ( planck brane ) . in this model
, gravity is generated on the planck brane , whereas at least some of the sm particles are confined to the sm brane , separated from the planck brane in the extra dimension .
due to the warping of the extra dimension , operators with the characteristic size of @xmath3 on the planck brane give rise to exponentially suppressed energy scales on the sm brane : @xmath21 , where @xmath22 is the reduced planck scale .
thus the ewsb scale can be connected to the planck scale with a relatively low degree of fine tuning by requiring the product of the warp factor and the radius of the compact dimension : @xmath23 . in this model
@xmath2 could have a natural " value of @xmath24 , thus offering a rigorous solution to the hierarchy problem .
in certain variations of the rs model , some of the sm particles could be displaced w.r.t .
the sm brane ; therefore gravitational coupling may not be universal as it does depend on the geometrical overlap of the graviton and sm particle wave functions in the extra dimension .
universal extra dimensional model @xcite by appelquist , cheng , and dobrescu . in this model
, all sm particles can travel in @xmath1 extra dimensions and standard constraints from atomic physics and low - energy experiments are avoided by introducing a conserved quantum number , kaluza - klein parity @xmath25 @xcite , which prohibits sm particles ( @xmath26 ) to be directly coupled to an excited kk mode ( @xmath27 ) .
this quantum number acts similarly to the r - parity @xcite in supersymmetric models and ensures that kk excitations can only be produced in pairs , thus significantly weakening many of potential constraints .
typical values of @xmath1 being considered are one and two . the tev@xmath28 extra dimensional model of dienes , dudas , and gherghetta @xcite .
in this model , only gauge bosons propagate in an extra dimension with @xmath9 tev@xmath28 , potentially achieving low - energy unification of the strong , electromagnetic , and weak forces .
the phenomenology from the point of view of ( 3 + 1)-dimensional space - time in all these models can be represented by a tower of kaluza - klein ( kk ) excitations of particles propagating in extra dimensions . for compactified extra dimensions ,
such particle could only have quantized values of the momentum projection on the compact dimensions ( cf .
classical particle in a box " problem in quantum mechanics ) . from the point of view of a 3d observer , this quantized momentum in the direction orthogonal to the 3d brane appears as a tower " of massive states , with the @xmath1-th excitation having a mass of @xmath29 , where @xmath30 is the mass of the zeroth kk mode , representing the ground state , or the particle confined to a 3d brane .
, scaledwidth=95.0% ] there are three different classes of signatures that can be used to test the predictions of the add model : _ virtual graviton effects .
_ in the add model , the kk modes of the graviton are very finely spaced in energy , as the size of extra dimensions is large by particle physics standards .
hence , it is experimentally impossible to distinguish between the individual kk states of the graviton , and their spectrum appears to be continuous .
each kk mode couples to the sm particles with the gravitational coupling , @xmath31 , where @xmath32 is the newton s constant ; since there are many kk modes that can be excited when the momentum transfer is large , the effective gravitational coupling becomes strong and the processes normally transmitted by other gauge bosons , e.g. , drell yan fermion pair production , can now be also carried via virtual gravitons ( see fig .
[ fig : dy ] ) .
generally , the graviton - mediated diagrams interfere with their sm counterparts , resulting in a modification of the invariant mass spectrum of the final - state objects , particularly at large masses , for which the amplification of the gravitational interaction due to the a large number of the accessible kk excitations becomes significant .
this modification can be described via an effective field theory ( eft ) approximation , similar to the compositeness operators @xcite , where the compositeness scale " ( i.e. , the eft cutoff ) @xmath33 , is expected to be of order @xmath6 .
the dependence on the number of extra dimensions is predicted to be weak.@xcite _ direct graviton emission . _ since gravity couples to energy - momentum tensor , any sm feynman diagram vertex can be modified by attaching an extra graviton line to it .
thus , one can take any s - channel production vertex , e.g. , @xmath34 or @xmath35 , and add a graviton to it , resulting in a @xmath36 or @xmath37 process , where a gluon jet is recoiling against the graviton , which escapes in the extra dimensions .
just as in the virtual graviton case , when the momentum transfer is large , the corresponding effective coupling increases and the cross section for this reaction could become sizable .
the signature for such a process in a collider detector is a single jet , countered by a significant missing momentum in the direction transverse to the beam , i.e. a _
monojet_. this is a spectacular signature , quite different from usual sm jet pair production .
similarly , other objects can recoil against the graviton , giving rise to , e.g. , monophoton signature .
the probability of such a process is @xmath38 ( see ref . ) , and thus strongly depends both on the fundamental planck scale and on the number of extra dimensions .
_ black hole production .
_ one of the most spectacular signatures predicted @xcite within the add model is a possibility to produce microscopic black holes ( bh ) when the collision energy exceeds the value of the fundamental planck scale @xmath6 .
these bhs may act either as their semi - classical counterparts in general relativity and evaporate via thermal hawking radiation of various particle species ; or they may have more complicated properties of their quantum precursors . lacking the full picture of quantum gravity ,
it is impossible to have exact predictions on how these black holes evaporate , so many different possibilities should be considered .
since the hawking temperature is inversely proportional to the schwarzschild radius of such a bh @xcite , and the bh are tiny , the temperature is very hot and generally speaking the bh is expected to evaporate very fast either thermally ( i.e. , with an emission of a dozen or so particles , each carrying hundreds of gev of energy ) , or via a decay , as a quantum state , into a pair of even more energetic particles . in either case , the signature is quite spectacular and can be also used to probe more general models with low - scale quantum gravity . the terminal stage of bh evaporation is not known .
some models predict that either a stable or unstable remnant with the mass of order of the fundamental planck scale is formed as the evaporating bh mass approaches this planck scale .
in other models it is assumed that thermal evaporation continues to the very end .
virtual graviton effects in cms have been sought in dilepton , diphoton , and dijet data using the invariant mass spectrum of the two produced objects . in both cases ,
the interference with the corresponding sm processes is taken into account .
the dominant background in the mass range of interest is the sm production of dileptons , diphotons , or dijets .
this background , known to next - to - next - to - leading - order ( dileptons ) or next - to - leading - order ( nlo , diphotons , dijets ) precision is reliably estimated from simulation .
no excesses were found in any of the channels . the most stringent limits on @xmath33 come from the dijet data at @xmath39 tev with a sample corresponding to an integrated luminosity of 19.7 fb@xmath28 , and reach 8 tev at a 95% confidence level ( cl ) .
the results of these searches are summarized in table [ table : virtual ] . for the dilepton searches @xcite , a nlo signal @xmath40-factor of 1.3 was assumed,@xcite while for the diphoton search @xcite
a @xmath40-factor of 1.6 was used.@xcite for the dijet @xcite search , leading - order ( lo ) signal cross - section was used . for the dependence of the limits on the number of extra dimensions , the convention of ref .
was used .
given that the values of @xmath33 and @xmath6 , while of the same order of magnitude , can differ from one the other ( and @xmath33 is generally expected to be lower than @xmath6 ) , searches for virtual graviton effects ( sensitive to @xmath33 ) are complementary to the ones for direct graviton emission ( sensitive to @xmath6 ) .
direct graviton emission in cms has been probed in the monophoton and monojet channels . in both cases ,
the dominant background comes from production of a @xmath41 boson in association with a jet or a photon , with the @xmath41 boson decaying invisibly in a pair of neutrinos . in the case of the monojet analysis
, this background is estimated from control samples in data where the @xmath41 boson decays into a pair of muons .
the acceptance correction factor from leptonic to invisible decays comes from simulation . in the monophoton analysis ,
the background is estimated from simulation corrected for nlo effects and cross - checked with the @xmath42 sample , where @xmath43 or @xmath44 .
the second - dominant background originates from @xmath45+jet or @xmath45+photon production , with the @xmath45 boson decaying leptonically ( including the @xmath46 channel ) and the lepton falling either below the selection threshold on the transverse momentum ( @xmath47 ) or outside of the detector acceptance . in the case of the @xmath48-lepton decay
, the events can pass the signal selections either because the @xmath48 lepton decays hadronically , yielding an extra soft jet in the final state , or leptonically , with the lepton falling below the minimum @xmath47 threshold .
for the monojet analysis , this background is estimated from a control sample of @xmath49+jet events , with the acceptance correction factors derived from simulation .
for the monophoton analysis , the estimate is based on simulation adjusted for the nlo @xmath40-factor . in the case of monophotons
, there is an additional sizable background from @xmath50 events with the electron misreconstructed as a photon .
this probability is measured using control sample of @xmath51 events and then applied to the @xmath50 control sample to estimate this background contribution to the signal sample .
data agree well with the sm predictions in both the monojet and monophoton channels , both at @xmath0 and 8 tev .
the most stringent limits on @xmath6 come from the monojet analysis of the @xmath39 tev dataset corresponding to an integrated luminosity of 19.6 fb@xmath28 and reach 5 tev at a 95% cl .
the limits are summarized in table [ table : direct ] . in the monophoton case ,
lo cross section for direct graviton emission was used , as nlo effects have not been calculated for this process . for the monojets , both lo and nlo calculations are available .
only lo - based limits are shown in table [ table : direct ] for easier comparison between the two channels .
since the @xmath52 tev limits are significantly more stringent than the 7 tev ones , only the former are shown in the table . the limits from these two analyses , as well as their comparison with the @xmath0 tev cms results@xcite and earlier measurements are shown in fig . [ fig : direct ] .
at a 95% cl from the monophoton ( left ) and monojet ( right ) analyses . from refs .
[ fig : direct],title="fig:",scaledwidth=51.0% ] at a 95% cl from the monophoton ( left ) and monojet ( right ) analyses . from refs .
[ fig : direct],title="fig:",scaledwidth=47.0% ] the cms experiment has pioneered searches for microscopic black holes with the very first lhc data at @xmath0 tev @xcite . already with this small amount of data ,
corresponding to an integrated luminosity of just 0.035 fb@xmath28 , stringent limits on semiclassical black hole production were set , using a novel method of predicting the dominant background coming from qcd multijet production .
the analysis since then has evolved considerably and its later reincarnations@xcite probed many more additional bh models , including quantum black holes and string balls . here
we summarize the most recent analysis @xcite based on a @xmath39 tev data sample with an integrated luminosity of 12.1 fb@xmath28 .
this analysis was performed in inclusive final states , thus maximizing the sensitivity to bh production and decay .
semi - classical black holes are expected to evaporate in a large ( @xmath53 ) number of energetic particles , emitted nearly isotropically , with the major fraction of the emitted particles being quarks and gluons @xcite , resulting in a multijet final state .
quantum effects and gray - body factors may change the relative fraction of emitted quarks and gluons , but generally it is expected that these particles still appear most often even in the decays of quantum black holes or their precursors , due to a large number of color degrees of freedom that quarks and gluons possess , compared to other sm particles .
the discriminating variable between the signal and the dominating qcd multijet background used in the search is the scalar sum of transverse momenta of all particles in the event , for which the @xmath47 value exceeds 50 gev .
this variable , @xmath54 , was further corrected for any significant in the event by adding the value to the @xmath54 variable , if the former exceeds 50 gev . the choice of @xmath54 as the discriminating variable is very robust and rather insensitive to the exact particle content in the process of bh evaporation , as well to the details of the final , sub - planckian evaporation phase .
the addition of to the definition of @xmath54 further ensures high sensitivity of the search for the case of stable non - interacting remnant with the mass of order of the fundamental planck scale , which may be produced in the terminal stage of the bh evaporation process .
the main challenge of the search is to describe the inclusive multijet background in a robust way , as the bh signal corresponds to a broad enhancement in @xmath54 distribution at high end , rather than a narrow peak .
since the bh signal is expected to correspond to high multiplicity of final - state particles , one has to reliably describe the background for large jet multiplicities , which is quite challenging theoretically , as higher - order calculations that fully describe multijet production simply do not exist .
thus , one can not rely on the monte carlo simulations to reproduce the @xmath54 spectrum correctly . to overcome this problem , the cms collaboration developed and utilized a novel method of predicting the qcd background directly from collision data .
it has been found empirically , first via simulation - based studies , and then from the analysis of data at low jet multiplicities , that the shape of the @xmath54 distribution for the dominant qcd multijet background does not depend on the multiplicity of the final state , above the turn - on threshold .
this observation , motivated by the way parton shower develops via nearly collinear emission , which conserves @xmath54 , allows one to predict @xmath54 spectrum of a multijet final state using low - multiplicity qcd events , e.g. dijets or three - jet events .
this provides a powerful method of predicting the dominant background for bh production by taking the @xmath54 shape from dijet events , for which the signal contamination is expected to be negligible , and normalizing it to the observed spectrum at high multiplicities at the low end of the @xmath54 distribution , where signal contamination is negligible even for large multiplicities of the final - state objects .
the results are shown in fig .
[ fig : cmsbh1 ] ( left ) .
, in the final state with the multiplicity of eight or more particles .
right : model - independent upper limits at a 95% cl on the cross section of a new physics signal decaying in the final state with 10 or more particles , as a function of the minimum @xmath54 requirement . from ref .
.,title="fig:",scaledwidth=45.0% ] , in the final state with the multiplicity of eight or more particles .
right : model - independent upper limits at a 95% cl on the cross section of a new physics signal decaying in the final state with 10 or more particles , as a function of the minimum @xmath54 requirement . from ref .
.,title="fig:",scaledwidth=54.0% ] the cms data with high final - state multiplicities is well fit by the background shape obtained from the dijet events , no excess characteristic of a bh production is observed even for highest multiplicities .
this lack of an apparent signal can be interpreted in a model - independent way by providing a limit on the cross section for any new physics signal for @xmath54 values above a certain cutoff , for any given inclusive final state multiplicity .
an example of such a limit is shown in fig .
[ fig : cmsbh1 ] ( right ) , for the final - state multiplicity @xmath55 . for signals corresponding to large values of @xmath54 (
above 2 tev or so ) the cross section limit reaches @xmath56 fb .
these limits can be compared with the production cross section for black holes in a variety of models and used to set limits on the minimum bh mass ( @xmath57 ) that can be produced in these models .
they are also applicable to other final states , e.g. , from cascade decays of massive supersymmetric particles , thus making this search even more general .
in fact , this very limit has been recently reinterpreted in terms of constraints on supersymmetry in ref . .
for several specific models of semi - classical black holes the limits have been set explicitly via optimization of the analysis for these particular scenarios .
these limits , although not very reliable for the black hole masses approaching the fundamental planck scale , are shown in fig .
[ fig : cmsbh2 ] .
as can be seen from these plots , the excluded minimum bh masses reach 56 tev , which is close to the energy limit of the lhc machine at @xmath39 tev .
therefore , the present searches are completely energy limited and will gain a significant boost once the lhc is restarted at @xmath58 tev this year .
even with early high - energy data , significantly higher bh masses and fundamental planck scale values could be probed .
nevertheless , it is clear that if an excess is found in this searches , we would be dealing with fundamentally quantum objects , given that the fundamental planck scale limits from other searches already reached @xmath59 58 tev .
thus it is very important to also look for quantum black holes , which are expected to decay before they thermalize , resulting in just a few particles in the final state .
the same inclusive @xmath54-based analysis can be also used to set limits on quantum black holes . in this case , the result is based on the @xmath54 spectrum at low final - state object multiplicity ( @xmath60 ) . given that in this case one expects a contamination of the @xmath54 spectrum from signal , we predict qcd background at large @xmath54 by simply extrapolating the fit function from the low-@xmath54 range , which is used to determine the parameters of a smooth background fit and has negligible signal contamination ( see fig .
[ fig : cmsqbh ] ) .
we also check the background prediction obtained in this way by comparing it with the prediction based on the template obtained from the @xmath61 spectrum ( dashed blue line ) , which falls well within the uncertainties of the extrapolated background estimate .
@xmath54 spectrum with the background extrapolated from the low-@xmath54 region .
the dashed blue line shows an alternative background prediction based on @xmath61 @xmath54 spectrum .
a benchmark quantum bh signal is shown with a dotted magenta line .
( right ) limits on the minimum quantum bh mass for various numbers of extra dimensions .
the @xmath62 case corresponds to limits in the randall - sundrum model , in which the black hole formation is also possible ( see section [ sec : rsl ] ) . from ref .
.,title="fig:",scaledwidth=45.0% ] @xmath54 spectrum with the background extrapolated from the low-@xmath54 region .
the dashed blue line shows an alternative background prediction based on @xmath61 @xmath54 spectrum .
a benchmark quantum bh signal is shown with a dotted magenta line .
( right ) limits on the minimum quantum bh mass for various numbers of extra dimensions .
the @xmath62 case corresponds to limits in the randall - sundrum model , in which the black hole formation is also possible ( see section [ sec : rsl ] ) . from ref .
.,title="fig:",scaledwidth=54.0% ] another search for quantum black holes in cms was done using the dijet invariant mass spectrum@xcite obtained in the @xmath39 tev dataset with an integrated luminosity of 19.7 fb@xmath28 . in this case , the spectrum is fit with a sum of a smooth background template and a signal mass template obtained in simulation ( see fig . [
fig : cmsqbh1 ] ( left ) ) .
the analysis is a part of a general program of dijet resonance searches and set similar limits on quantum bh ( see fig .
[ fig : cmsqbh1 ] ( right ) ) as the @xmath54-based analysis @xcite ( see fig . [
fig : cmsqbh ] ( right ) ) .
case corresponds to limits in the randall - sundrum model ( see section [ sec : rsl ] ) , in which the black hole formation is also possible . from ref .
.,title="fig:",scaledwidth=46.0% ] case corresponds to limits in the randall - sundrum model ( see section [ sec : rsl ] ) , in which the black hole formation is also possible . from ref .
.,title="fig:",scaledwidth=52.0% ] while the properties of quantum black holes remain an enigma , one could address a question of a light black hole formation in simple string theory models , which may correctly describe the effects of quantum gravity .
one of the suggestions is that a precursor of a black hole is a highly excited excited string state , randomly folded in a string ball " @xcite .
the properties of such a string ball are expected to be similar to those of a semi - classical black hole , with the exception that its evaporation takes place at a fixed , hagedorn temperature @xcite , which does not depend on the string ball mass , but only on the string scale @xmath63 and string coupling constant @xmath64 .
thus , the semi - classical black hole analysis @xcite can be also used to set limits on string balls , which are shown in fig .
[ fig : sb ] and reach 5.56.0 tev . , string scale @xmath63 , and string coupling constant @xmath64 . from ref .
the lhc phenomenology of the rs model involves high - mass kk excitations of the graviton and , possibly , other bosons of the sm .
there are many flavors " of the rs model , with various particles allowed to propagate in the bulk , so depending on a particular realization , the graviton may not couple to all the particle species in the same way , as was the case in the add model .
consequently , it s important to explore various decay channels of kk graviton and also look for kk excitations of other bosons .
another possible experimental probe for rs model is production of microscopic bh @xcite . in general these black holes
are expected to act similar to the add black holes for the @xmath65 case , however , it has been argued that because of their higher temperature and shorter lifetime , rs black holes do nt have time to thermalize and decay as quantum black holes in a small number of final - state particles .
this signature has been covered in section [ sec : qbh ] of this review .. while the rs model can be described using two parameters : the curvature of the anti - desitter metric @xmath15 and the radius of the compact dimension @xmath2 , phenomenologically , it s more convenient to use a different equivalent set of parameters , namely the mass of the first graviton kk excitation @xmath66 tev and the dimensionless coupling @xmath67 .
the spacing between the kk modes of the graviton is given by the subsequent zeroes @xmath68 of the bessel function @xmath69 ( @xmath70 ) : @xmath71 .
the coupling @xmath72 determines the strength of coupling of the graviton to the sm particles and the width of the kk excitations . generally , it s expected that @xmath73 .
larger values of @xmath72 correspond to stronger coupling and broader kk resonances .
the cms experiment has conducted a large variety of searches for kk excitations of the graviton and gluon in the rs model .
the signatures probed so far include dileptons , pairs of quarks and gluons , including the top - quark pairs , and pairs of vector bosons . in the case of @xmath74 , @xmath75 , and @xmath76 final states ,
an additional complication arises from the fact that for large values of @xmath77 , above @xmath78 tev , the decay products ( top quarks or vector bosons ) are produced with high lorentz boost , so their subsequent decay products are collimated and not always can be resolved , thus appearing as a single jet " in the detector .
special techniques , including jet substructure @xcite are therefore utilized to maintain the sensitivity of these searches for large values of @xmath77 .
these jets of merged objects are referred to as jets " in the list of signatures below .
limits in the dilepton channels come from the analysis focused on a search for @xmath79 bosons@xcite ; limits in the quark - antiquark and @xmath80 final states come from general dijet resonance searches in the inclusive@xcite and @xmath81@xcite final states . searches in the diboson and @xmath74 final state target the rs model specifically . the limits on rs gravitons from these cms searches are summarized in table [ table : rs ] .
the semileptonic@xcite and all - hadronic@xcite @xmath82 analyses also set limits on the bulk gravitons in a modified rs model @xcite . in this scenario , coupling of the graviton to light fermions and photons is suppressed , while the production of gravitons via gluon fusion and their decays into a pair of massive gauge bosons is sizable.the @xmath39 tev analysis focuses mainly on the graviton mass range above 1 tev , where the jet substructure techniques are more efficient than explicitly resolving two jets from the @xmath41-boson decay ; consequently it is not as sensitive at low masses .
the analysis is not yet sensitive to the bulk gravitons with @xmath83 in the entire mass range studied , so only the limits on the cross section are set ( see fig .
[ fig : bulk ] ) . note that cms has conducted an earlier analysis at @xmath0 tev in the @xmath84 final state@xcite focused on the mass range below 1 tev . that analysis claimed an exclusion limit of 0.645 tev on the bulk graviton mass .
however , the mass exclusion was based on the original theoretical calculations @xcite , which have been recently revised down by a factor of four @xcite .
the signal cross section curves in fig .
[ fig : bulk ] do reflect this updated cross section ; hence this figure is not directly comparable with fig .
3 ( right ) of ref . in terms of the signal model . is shown as a red solid curve .
in region i , only the @xmath85 + jet " channel contributes . in region
ii , both @xmath85 + jet " ( where @xmath43 or @xmath44 ) and @xmath86 + jet " channels contribute . in region iii ,
both the semi - leptonic and all - hadronic channels contribute . from ref .
.,scaledwidth=80.0% ] a search for a kk excitation of a gluon in the rs model with gluons in the bulk @xcite has been done in the @xmath74 decay mode , both in semileponic ( resolved and boosted topologies ) and all - hadronic decay modes of the @xmath74 quark pair .
the combined analysis@xcite uses jet substructure techniques to identified a jet originating from a collimated decay @xmath87 jet " , by looking for an object consistent with a hadronic decay of the @xmath45 boson present within the jet , as well as considering the jet mass value .
the analysis set a lower mass limit of 2.5 tev ( 2.4 tev expected ) at a 95% cl .
recently , this result has been superseded by a more advanced jet substructure analysis in the all - hadronic channel , and the dilepton channel has been also analyzed@xcite .
the combination of all there channels improved the earlier limit to 2.8 tev ( 2.7 tev expected ) at a 95% cl ( see fig .
[ fig : gkk ] ) , which is the most stringent limit on kk gluons to date . and
@xmath88 excursions from the expected limits are also shown .
the vertical dashed line indicates the transition between the threshold semileptonic analysis @xcite and combined analysis , chosen based on their relative sensitivity . from ref .
searches for universal extra dimensions in cms have been conducted so far in a few specific ued models .
one of them,@xcite with a single extra dimension and bulk fermions , often referred to as split - ued model " . in this model , only even kk excitations of gauge bosons , e.g. , of the @xmath45 boson , can couple to sm fermions , due to kk parity conservation .
thus , one could search for the decay of the second kk excitation of the @xmath45 boson , w@xmath89 , into a lepton ( e or @xmath44 ) and a neutrino .
this signature is also used to search for additional @xmath45-like gauge bosons , @xmath90 .
the most stringent limits to date come from the cms search@xcite based on the @xmath39 tev sample corresponding to an integrated luminosity of 20 fb@xmath28 .
the limit is set on the inverse size of the compact extra dimension , @xmath91 , as a function of the bulk mass parameter of 5d fermion fields @xmath44.@xcite .
these limits are shown in fig .
[ fig : ued ] and reach @xmath91 of nearly 2 tev for large values of @xmath44 .
and @xmath91 derived from the @xmath90 mass limits taking into account the corresponding width of the @xmath92 .
the limit for the electron ( red dotted line ) and the muon channel ( blue line ) individually along with their combination ( yellow ) . from ref .
.,scaledwidth=70.0% ] another ued model probed at cms is the one where the ued space is embedded in a space with additional large extra dimensions , where only gravity propagates .
the lightest kk particle in this model is the first kk excitation of the photon , @xmath93 , which then decays into a photon and a graviton .
the resulting signature for the pair - produced @xmath93 particles ( either directly or via cascade decay of other kk particles ) is two photons and due to undetected gravitons .
this is a similar final state as expected in supersymmetric scenarios with gauge - mediated supersymmetry breaking ( gmsb ) .
a cms gmsb supersymmetry search@xcite at @xmath0 tev based on a data sample with an integrated luminosity of 4.69 fb@xmath28 has been also interpreted in terms of limits on this ued scenario .
the number of large extra dimensions @xmath1 in this scenario was varied between 2 and 6 . for
@xmath94 decays involving a heavy graviton with mass of order @xmath91 dominate , while for @xmath10 decays involving massless gravitons prevail.@xcite the observed data in this analysis agree with the sm predictions , which allowed to set a limit of @xmath95 tev for @xmath11 at a 95% cl , with an expected limit of 1.35 tev .
for @xmath10 the observed ( expected ) lower limit is reduced to 1.35 ( 1.34 ) tev .
these are the most stringent limits on the probed ued model to date .
in the model@xcite with extra dimensions with a size @xmath96 tev@xmath28 , there is no kk parity conservation , so one could look directly for production of the first excited states of the @xmath45 or @xmath41 bosons .
the search is similar to the standard searches for @xmath90 and @xmath79 bosons with the sm - like couplings , with the caveat that there is strong negative interference between the non - resonant @xmath97 or drell yan production and the first kk excitation at approximately half the resonance mass .
this effect needs to be accounted for in the limit - setting procedure .
while the @xmath90 search @xcite has been reinterpreted in terms of tev@xmath28 extra dimensions , the current limit from direct searches are still significantly weaker than the indirect limits from lep ( see , e.g. , ref . ) , so they are still more in a realm of the next lhc run .
the monophoton analysis @xcite also set limits on branons,@xcite which are excitations of branes in models with extra dimensions inspired by string theory .
these branons , just like string excitations , may be quantum precursor for black hole formation .
limits on branons as a function of the excitation mass @xmath6 and the brane tension @xmath98 are shown in fig .
[ fig : branons ] .
this is the first dedicated search for branons since the lep era.@xcite ( the tevatron limits shown in fig .
[ fig : branons ] are from phenomenological interpretation@xcite of the tevatron monophoton and monojet data . )
the dijet analyses also looked for string resonances , i.e. , excited string states that mainly decay into @xmath99 final states .
these resonances are expected to be strongly produced and also mass - degenerate , with the mass close to the string scale @xmath33.@xcite therefore , the cross section for string resonance production is very large and stringent limits reaching 5.0 tev are set on their existence ( see fig .
[ fig : strings]).@xcite as a function of the brane tension parameter @xmath98 , together with the previous exclusion from the l3 experiment.@xcite from ref .
.,scaledwidth=70.0% ] and @xmath88 levels ( shaded bands ) .
theoretical predictions for string resonance models are shown with the dash - dotted blue line . from ref .
.,scaledwidth=70.0% ] finally , there is also a search for jet extinction in the inclusive jet production , which is another way to looks for black holes or their quantum precursors .
the idea is that jet production may be suppressed above a certain energy threshold where these new phenomena open up , which could be inferred from a deviation of an inclusive jet @xmath47 spectrum from a smooth power - law falling function.@xcite the cms experiment has conducted the first search for this phenomenon to date using a data sample at @xmath39 tev corresponding to an integrated luminosity of 11 fb@xmath28 and set a lower limit on the extinction scale of 3.3 tev at a 95% cl@xcite , see fig .
[ fig : ext ] . , where @xmath100 is the extinction scale and its translation in a lower limit on @xmath100 .
the green and yellow bands indicate the @xmath101 and @xmath88 variations in the expected limit . from ref .
a large variety of cms analyses based on 20102012 lhc data resulted in stringent limits on the existence of extra dimensions and low - scale quantum gravity , in a number of models . in many cases ,
the limits obtained by cms are the most restrictive to date .
while the possibility to see these phenomena at the lhc is diminishing , the current searches are completely limited by the maximum machine energy of 8 tev reached so far .
there is still a lot of uncovered model parameter space left , which will be explored as early as this year by exploiting the lhc potential close to its design energy of 14 tev .
i m indebted to my many colleagues , who built , commissioned , and ran the cms detector , and produced all the beautiful results included in this review .
these new results made working on the review an exciting and interesting task .
this work is partially supported by the doe award no .
de - sc0010010 - 003376 .
|
the success of the first three years of operations of the lhc at center - of - mass energies of 7 and 8 tev radically changed the landscape of searches for new physics beyond the standard model and our very way of thinking about its possible origin and its hiding place . among the paradigms of new physics that have been probed quite extensively at the lhc , are various models that predict the existence of extra spatial dimensions . in this review ,
the current status of searches for extra dimensions with the cms detector is presented , along with prospects for future searches at the full energy of the lhc , expected to be reached in the next few years . published 4 may 2015
|
it is commonly agreed that at high spacetime curvatures , when the quantum effects of the gravitational field become significant , general relativity needs to be replaced by a quantum theory of gravity . in spite of progress in several directions , finding this new theory has proven a challenging problem for several decades .
current research in the field is paying substantial attention to the numerous indications that gravity may only be emergent , meaning that it is a collective , or thermodynamical , description of microscopic physics in which we do not encounter geometric or gravitational degree of freedom .
an analogy to illustrate this point of view is fluid dynamics and the transition from from thermodynamics to the kinetic theory .
what we currently know is the low energy theory , the analogue of fluid dynamics .
we are looking for the microscopic theory , the analogue of the quantum molecular dynamics . just as there are no waves in the molecular theory , we may not find geometric degree of freedom in the fundamental theory . not surprisingly
, this significant shift in perspective opens up new routes that may take us out of the old problems .
the emergent viewpoint amounts to treating quantum gravity as a problem in statistical physics .
a powerful set of methods in statistical physics involve the use of lattice - based models , such as the ising model for ferromagnetism , the hubbard model for the conductor / insulator transition , etc .
such methods are starting to be introduced in quantum gravity .
examples are g. volovik s work on emergent lorentz invariance at the fermi point @xcite , x .-
wen s work on emergent matter and gravitons from a bosonic spin system @xcite , the emergence of a lorentzian metric and aspects of gravitation such as hawking radiation in analog models of gravity @xcite , as well as long - standing approaches such as matrix models @xcite , and more radical formulations of geometry in terms of information @xcite .
it is natural for the lattice in the lattice system to play the role of ( a primitive form of ) geometry .
now , general relativity is a background independent theory , by which we mean that the geometry of spacetime is fully dynamical . by analogy , we expect that the use of a fixed lattice is inappropriate and one instead needs models on a _ dynamical _ lattice . while for this reason desirable , dynamical lattices raise difficult technical problems that have not been previously addressed in the field of statistical physics . the present article is concerned with exactly this problem and presents a method that deals with dynamical lattices in certain situations . in previous work , we introduced quantum graphity , a background independent model of spacetime in which time is an external parameter and space is described by a relational theory based on graphs @xcite .
the idea is to represent locality by the adjacencies of a dynamical graph on which the diffeomorphism group is replaced , in the high energy phase , by the symmetric group on the complete graph which breaks down to its subgraphs at lower energies .
these kind of models are sometimes called _ _ event - symmetric__@xcite ) .
the principle of event symmetry refers to the replacement , at high energies , of the diffeomorphism group with the group of permutations of events in spacetime . in the context of quantum graphity
, however , the event symmetry is only spatial , in the sense that at high energy the graph is complete and every vertex of the graph is at distance one from each other .
if the dynamics is such that the system settles into a minimum energy subgraph that exhibits geometric symmetries , for instance , a discrete version of flat space in low dimension , we say that geometry emerges in that phase . in @xcite
was shown that desired symmetric lattices , i.e. , discrete 2d frw , are stable local minima under certain choices of parameters in the hamiltonian ( when the effect of the matter on the lattice is neglected ) . following this work , in @xcite
, we used the same concept of locality in terms of a dynamical lattice , but with a new type of matter that interacts non - linearly with the geometry , a precursor of gravity , and initiated a study of the quantum properties of that system . in the present article
, we return to the technical issues of spin systems on dynamical lattices and we introduce a method to deal with a theory of dynamical graphs on @xmath0 vertices . such graphs are subgraphs of @xmath1 , the complete graph on @xmath0 vertices .
we show how , by transforming @xmath1 to its line graph , the theory can be approximated by an ising model on the line graph of a complete graph .
we then use this to study the low energy properties of the quantum graphity model in @xcite . using mean field theory
we calculate the average valence of the graph at low temperature and we evaluate the first corrections due to the presence of 3-loops .
the paper is organized as follows . in section
ii , we review the quantum graphity model @xcite with no matter . in section iii , we define the line graph derived from a generic graph and summarize its properties . in section iv , we show how , in a certain reasonable approximation , the hamiltonian of quantum graphity can be recast as an ising hamiltonian on the line graph of a complete graph . in section
v , we calculate the corrections due to loops at low temperature and describe , in this framework , the behavior of the correlation function in mean field theory .
conclusions follow in section vi .
let us briefly introduce the quantum graphity model @xcite . as the name _
graphity _ implies , quantum graphity is a model for a quantum theory of gravity in which the fundamental microstates are dynamical graphs , postulated to describe relational physics at planckian energies . there is no notion of geometry or quantum geometry at high energy , instead , geometry emerges as the system cools down and away from the planckian regime .
the microstates live in a hilbert space on the complete graph @xmath2 of @xmath0 vertices , given by @xmath3 where @xmath4 represents the edge of the graph @xmath5 between the @xmath6 and @xmath7 vertices , while @xmath8 and @xmath9 are the hilbert spaces associated with edges and vertices respectively . in particular , the hilbert space associated with an edge between vertex @xmath6 and @xmath7
is the two - level state space : @xmath10 the two states 1,0 in ( [ hilberts ] ) are interpreted as the edge being _ on _ or _ off _ respectively .
this choice means that basis states in the hilbert space of the edges represent subgraphs @xmath11 of the complete graph @xmath2 .
a generic state in the hilbert space of the edges is a superposition of such subgraphs : @xmath12 in the full model of @xcite , extra degrees of freedom are assigned to the on states : @xmath13 in @xcite , and in the present work , there are no degrees of freedom associated with the vertices and hence we ignore @xmath14 .
let us focus now on the state space ( [ hilberts ] ) . on the fock space of the edges we can introduce the ladder operators @xmath15 and @xmath16 , with the usual action : @xmath17 dynamics in quantum graphity
is given by a hamiltonian acting on the graph states of the form@xcite : @xmath18 where @xmath19 keeps track of how many _ on _ edges are attached to a single vertex , and @xmath20 counts closed paths in the graph .
the interaction term will not be used in the following , but in a generic model these terms produce alexander moves on the graph . in more detail ,
the term @xmath19 is usually chosen to be of the form : @xmath21 where the indices @xmath22 enumerate vertices , @xmath23 is the usual number operator , and @xmath24 and @xmath25 are free couplings that we assume to be positive .
the purpose of this term is to ensure that at low energies the system has a ( low temperature ) phase in which the average vertex valence ( i.e. _ on _ edges attached to a vertex ) is @xmath26 . later on in the paper we will show that , at least in the mean field theory approximation , this is indeed the case . the term @xmath20 is given by @xmath27 where @xmath28 and @xmath29 are couplings assumed to be positive .
the operator @xmath30 counts the number of non - retracing paths of length l based at the vertex @xmath6 .
this operator is related to the trace of the adjacency matrix in the original model .
we will build this operator in another way in the following . for @xmath31 , so that higher length loops contribute less than short length loops , this term is semi - local .
the @xmath32 comes from the expansion of the exponential of the loop - path operator@xcite .
it was shown in @xcite that @xmath29 determines the length on which loop size is peaked at low energies . in what follows
we introduce a new method to analyze dynamical lattices , by transforming @xmath2 to its _ line graph _ which we define in the next section .
the _ on / off _ edges of @xmath2 will become ising spins on the _ fixed _ line graph , so that standard methods of statistical physics can be used .
we start by defining line graphs .
let @xmath33 denote a graph with vertex set @xmath34 and edge set @xmath35 .
the line graph @xmath36 is the graph of the adjacencies of @xmath37 , containing information on the connectivity of the original graph .
each vertex @xmath38 corresponds to an edge @xmath39 .
two vertices @xmath40 and @xmath41 in @xmath42 are adjacent if and only if the edges in g corresponding to @xmath40 and @xmath41 share a vertex .
the correspondence between @xmath37 and @xmath43 is not one to one . from a given graph @xmath37
we can construct only one @xmath43 but it is not true that any graph is a line graph .
in fact , according to the beineke classification , there are 9 non - minimal graphs that are not line graphs of another graph and each graph containing them is not a line graph@xcite .
the simplest example of a line graph is depicted in fig .
[ beineke ] . given a graph @xmath37
, we can construct its line graph using the following procedure : 1 .
enumerate the vertices of @xmath37 .
2 . enumerate the edges of @xmath37 with a fixed prescription ( see example below ) and put a blob on them .
3 . if two edges share a vertex , draw a bold line between them .
4 . remove @xmath37 and its enumeration .
what is left is the line graph of @xmath37 where the blobs represent its vertices .
let us now introduce some useful quantities : ( _ kirchhoff matrix _ ) let @xmath37 be a generic graph , @xmath44 be the set of vertices of @xmath37 and @xmath45 be the set of edges of @xmath37 .
let us define the matrix @xmath46 of size @xmath47 with entries @xmath48 , where @xmath6 is an integer between @xmath49 and @xmath50 on the set of vertices and @xmath51 is an integer between @xmath49 and @xmath25 on the set of edges , such that : @xmath52 the kirchhoff matrix @xmath53 is the @xmath54 matrix built from @xmath46 , such that : @xmath55 @xmath56 representing the transpose of @xmath46 .
a well - known theorem now gives the incidence matrix of the line graph @xmath43 : let @xmath37 be a graph with p edges and n vertices and let @xmath43 be its line graph . then the matrix : @xmath57 where @xmath58 is the @xmath54 identity matrix , is the incidence matrix of @xmath43 . in the next section
we will show how the graphity hamiltonian can be recast on the line graph using ( [ pij ] ) and ( [ kirchhoff ] ) .
since in the hamiltonian ( [ hamiltonian ] ) we neglect the terms in which vertices are interacting because we assume there are no degrees of freedom on them , one could expect that it can be rewritten only in terms of the connectivity of the graph . to carry out such a reformulation ,
let us expand the first term in ( [ hamiltonian ] ) for small values of the parameter @xmath25 : @xmath59 as we will see later , such an expansion does not modify the properties of the model at low temperature .
the first term in ( [ exp ] ) is an energy shift and , for what is to come , can be neglected
. we should now be able to recognize some particular terms in the expansion ( [ exp ] ) .
the third term is proportional to the operator @xmath60 .
it is the sum over all the edges of the graph , zero or not , of the number operator .
we will change the notation to @xmath61 where @xmath51 , as in the previous section , runs from @xmath49 to @xmath62 and labels the edges of @xmath2 or , equivalently in what follows , the vertices of its line graph . to rewrite the second term in ( [ exp ] ) , we need the matrix @xmath48 of ( [ pij ] ) in this context . this matrix maps the graph to its line graph , as we will see .
we first fix a prescription to label edges .
let @xmath1 be the complete graph of @xmath0 vertices .
let @xmath63 be any enumeration of @xmath64 , @xmath65 .
we identify edges by their endpoint vertices @xmath66 , with @xmath67 .
a _ labeling _
@xmath68 , @xmath69 is an enumeration of @xmath70 , according to the following prescription : @xmath71 using this prescription it is easy to see that the matrix @xmath48 introduced in ( [ pij ] ) , for the complete graph @xmath1 , has the simple ( recursive ) form : @xmath72 where @xmath73 is a row vector of length @xmath74 , @xmath75 is the identity matrix of size @xmath76 and @xmath77 represents a null row vector of length @xmath78 . the indices @xmath79 and @xmath80 run from @xmath49 to @xmath74 and @xmath49 to @xmath81 respectively .
as an example , for the graphs of fig . [ completes ] the @xmath46 matrices are : @xmath82 @xmath83 @xmath84 for ( a ) , ( b ) and ( c ) respectively . )
.,title="fig : " ] ) .,title="fig : " ] ) .,title="fig : " ] it is easy to see that two edges @xmath85 and @xmath51 have a common vertex if and only if we have : @xmath86 where @xmath87 is the transposition operation . by construction
, @xmath88 can take the following values only : @xmath89 in particular , for @xmath90 , @xmath53 is given by : @xmath91 using now the matrix @xmath92 just introduced , we want to construct generic _
n - string matrices _ as composition of @xmath50 edges of the graph , thus a _ path _ on the graph .
the @xmath50-string matrices will be needed both for the non - retracing loop term and the 2-edge interaction term in equation ( [ exp ] ) . the quantity : @xmath93 is the definition of the kirchhoff matrix of equation ( [ kirchhoff ] ) if we sum over the index @xmath6 .
from @xmath94 we can construct strings of @xmath46 s of the form @xmath95 that we call _ string matrices _ of _ _ n__th order .
these string matrices represent paths through vertices @xmath96 and they are zero unless the edges corresponding to @xmath97 are in the correct order , that means , they represent an actual path on the graph . for instance , the number of paths of length 2 on the complete graph is given by @xmath98 or , equivalently , we can use ( [ pij ] ) and rewrite ( [ interacting1 ] ) as @xmath99 the subtraction of twice the identity in ( [ interacting2 ] ) is the same as the subtraction of the self - energy of each edge .
we now clearly see that this matrix is precisely the incidence matrix of the line graph of @xmath1 introduced in ( [ kirchhoff ] ) , with @xmath100 .
so far we have dealt with the complete graph only .
we wish to extend this formalism to a dynamical graph . in order to do that we return to the hilbert space formulation of the graph with _ on / off _ edges .
recall that any graph on @xmath0 vertices is a subgraph of the complete graph @xmath1 , with some edges _
off_. thus , since we can always map a graph on a complete graph , we can count paths on any graph by modifying ( [ interacting2 ] ) so that it counts paths of only _ on _ edges on the corresponding complete graph . to do so ,
we introduce in the sum the number operators @xmath101 in the following way : @xmath102 this term does not contribute if any of the two edges @xmath103 is off .
it is easy to see that this term of the hamiltonian is an ising interaction .
the important difference between these two hamiltonians is that in our case the spin system is on the line graph a complete graph @xmath1 . by extension of the above , we are now able to construct a generic _ path operator _ out of @xmath94 s . we define @xmath104 where the set @xmath105 is @xmath106 it is easy to see that it counts the number of paths of length @xmath50 in the graph , that is why we call the @xmath107 s _ string matrices_. note that @xmath108 can take values @xmath109 and @xmath49 only because it is a product of @xmath109 s and @xmath49 s .
this string matrix is _ not _ the matrix multiplication of the kirchhoff matrices : it only reduces to matrix multiplication for retracing paths where we sum over all possible vertices . in the following
, we will denote the two sets in ( [ qset ] ) as @xmath110 and @xmath111 for the retracing and non retracing cases respectively ; moreover , we may explicitly show the indices on which we are doing the sum as @xmath112 . in order to count loops
, we just need to impose @xmath113 : @xmath114 thus we have discovered that , when there are no degrees of freedom on the vertices of the graph and we neglect the interaction terms , we can recast the quantum graphity hamiltonian on the line graph @xmath115 representation in the weak coupling regime at finite @xmath0 .
we end this section with two properties of the _ _ n__th - order string matrices .
let us define : @xmath116 the following properties of the sum of these string matrices on complete graphs will be required next : + * property 1 : * let @xmath117 .
then , for a loop of @xmath50 edges : @xmath118 and @xmath119 _ proof .
_ these two facts follow trivially if we note that the equations ( [ property1a ] ) and ( [ property1b ] ) count the number of retracing and non - retracing paths of length @xmath120 on the complete graph respectively .
+ + * property 2 : * let @xmath117 .
then , for a loop of @xmath50 edges , and for @xmath121 , we have : @xmath122 while , for @xmath123 : @xmath124 if @xmath125 .
+ + _ proof .
_ note that @xmath126 is the number of _ non - retracing _ loops of length @xmath120 on the complete graph @xmath1 which pass by the edges @xmath127 and @xmath128 .
now , it is easy to see that if the edges @xmath127 and @xmath128 do not share a link this quantity is zero .
also note that , by the symmetry of the complete graph , the number of non - retracing loops based on two neighboring edges must be the same for each pair of edges @xmath129,@xmath130 sharing a node .
since the matrix @xmath131 takes values @xmath49 or @xmath109 depending on whether the edges @xmath129,@xmath130 are neighbors or not , @xmath132 must be proportional to the matrix @xmath133 . in order to evaluate the proportionality constant , let us note that each loop is weighed by a factor of 1 because n - string matrices take values @xmath49 or @xmath109 only .
the combinatorial quantity @xmath134 is then the number of non - retracing loops of length @xmath120 passing from two consecutive fixed edges on the complete graph of @xmath0 vertices , as can be easily checked . the special case ( [ property2b ] ) follows from the fact that if we fix two edges there is only one edge which closes the 3-loop .
+ + note that , for @xmath135 , we have : @xmath136 we can now collect the results of this section to write the hamiltonian ( [ exp ] ) as @xmath137 where @xmath138 and neglecting higher order loop terms .
having rewritten the hamiltonian in an ising fashion , we now can approach the problem of finding a graph observable and its equilibrium distribution using mean field theory . as we will see , the natural graph observable to consider is the average valence of the graph .
we will assume that the system is at equilibrium and we neglect the interaction terms .
in this case , it is straightforward to use mean field theory analysis @xcite . in
what follows , we assume units in which the boltzmann constant @xmath139 .
we start by replacing the number operators @xmath140 with semi - classical analogs , imposing that their expectation value must lie in the interval @xmath141 $ ] : @xmath142 where @xmath46 is a probability measure of the following form : @xmath143 it is easy to see that this probability distribution forces the spin - average to lie in @xmath144 .
recall that in order to obtain the mean field theory distribution we have to extremize the gibbs functional given by @xmath145=h[m]-\frac{1}{\beta}s[m],\ ] ] where @xmath146 , @xmath147 $ ] is the energy and @xmath148 $ ] is the entropy functional .
the latter can be written as : @xmath149=-\sum_{m_\beta=\{0,1\ } } n_i p(m_\beta(i ) ) \log p(m_\beta(i ) ) , \label{entropy}\ ] ] where @xmath150 is the degeneracy of the state . in this subsection
we focus on the case in which the states _ on _ and _ off _ are not degenerate , so that @xmath151 . in the next subsection we will deal with non - degenerate edge states and in particular with 3-degenerate _ on _ states . in the process of extremizing the gibbs functional
we will see how the average valence of the graph naturally emerges .
we impose : @xmath152=0.\ ] ] using @xmath153 = -\log(\frac{m_\beta}{1-m_\beta})\ ] ] and @xmath154=a\sum_{\alpha \in \mathscr b } j_{\alpha \beta } m_\alpha - b -c \sum_{\alpha \gamma \in \mathscr b , \alpha\neq\gamma\neq\beta } \tilde q_{\alpha \beta \gamma } m_\alpha m_\gamma,\ ] ] we find that the distribution for the @xmath155 is @xmath156}}{1+e^{-\beta \partial_{m_\beta } h[m]}}= \frac{1}{1+e^{\beta \partial_{m_\beta } h[m]}}. \label{gdist}\ ] ] the solution of this equation gives the equilibrium value of @xmath157 once the value of the temperature is fixed .
we now want to write ( [ gdist ] ) as a function of an average quantity on the graph .
let us first note that , in the mean field theory approximation , we have @xmath158 where @xmath159 is the mean valence of the graph . the valence @xmath159 is a good graph observable that we can use also as a double check for our procedure since it appears explicitly in the original formulation of the hamiltonian and in the low temperature regime must take the value @xmath26 .
first , we note that : @xmath160 in the first equality , @xmath161 is the number of edges of the graph which are in an _ on _ state . in the second equality ,
the average valence ( the sum over all the local valencies divided by the number of vertices ) is explicitly written as a temperature dependent quantity . in the third equality we used the graph property : @xmath162 the most complicated term in the hamiltonian is the 3-loop one .
the simplest way to deal with it is to use the ansatz dictated by the mean field theory : @xmath163 let us replace @xmath157 with its average value : @xmath164 . using eq .
( [ property1a ] ) for non - retracing paths and assuming @xmath165 we obtain the dependence on @xmath159 .
@xmath166 is a function of order @xmath167 at low temperature , which we assume is dependent on @xmath168 . using these approximations
we can see that @xmath159 is a natural order parameter for our mean field theory since it is easily recognized as implicitly defined in the stable distribution : @xmath169}}. \label{eqdist}\ ] ] again , in order to double check our procedure , we can ask if such an order parameter behaves as expected at low temperature .
we must keep in mind that the starting hamiltonian ( [ hamiltonian ] ) was constructed in such a way that the average valence at zero temperature was a fixed value of the parameter @xmath26 at finite @xmath0
. we can now use ( [ eqdist ] ) to check if this is the case . to do so , we taylor expand both sides and match the zeroth and first order coefficients on the left and right hand side of the equation .
that is , we start with the expansion @xmath170 and , for the approximation to be consistent at @xmath171 , we require analyticity of the order parameter ( this has to be the case for a finite volume system in ordinary statistical mechanics , which is the case for finite @xmath0 ) .
we then require that inside the exponential of equation ( [ eqdist ] ) the temperature independent terms in the numerator cancel out so that at @xmath171 the exponent is well defined .
this gives the second order equation in @xmath85 : @xmath172 now note that , while this equation has two solutions , we need to only look for the one which is analytical in the parameters of the model and tends smoothly to the solution @xmath173 in the @xmath174 limit .
this fixes @xmath85 to the value @xmath175 , given by @xmath176 we can now plug @xmath175 at @xmath171 into ( [ eqdist ] ) : @xmath177 to obtain the value of @xmath178 in ( [ dexp ] ) : @xmath179 it is easy to see that in the limit @xmath180 we have @xmath181 , indicating a second - order phase transition ( a discontinuity in the first derivative of the order parameter ) . in our case , this happens at @xmath171 , meaning that this transition is not possible because there is no way to cool down the system to zero temperature with an external bath .
however , we have to remember that we are just approximating the real system with a semi - classical analog . we then simply interpret the above result as the fact that the system reaches the ground state very quickly when the temperature approaches zero .
it is interesting now to plug in the couplings . inserting equations ( [ couplings ] ) into ( [ cond1 ] )
, we find that at @xmath171 @xmath182 note that , for small values of @xmath29 , when @xmath183 , we have @xmath184 , meaning that at low temperature the mean degree is the one imposed by the degree term of the hamiltonian , as expected .
we can , however , see how the 3-loops term contributes to this quantity by a taylor expansion in @xmath29 : @xmath185 from this expression it is clear that the loop terms are suppressed if @xmath186 .
this is the main result derived in this paper using the line graph representation .
a plot of the function @xmath159 is shown in figure 3 . ) ( @xmath187-axis ) against @xmath168 ( @xmath188-axis ) , for increasing @xmath0 . ]
we now have the tools to calculate the susceptibility function for the theory in the mean field theory approximation .
recall that the susceptibility function tells us how the system reacts to a variation of the external magnetic field . in our case ,
the magnetic field is the combination @xmath189 and we note that the parameter @xmath26 in the hamiltonian appears only here .
we have the following analogy : @xmath26 represents the external magnetic field , while @xmath190 represents the spin - coupling combination . in order to calculate the susceptibility
, we assume that the constant b is site - dependent ( i.e. a field ) .
thus , we have @xmath191=\frac{1}{\beta}\frac{\partial m_\beta}{\partial b_\alpha}.\ ] ] in particular , we are interested in the susceptibility function when @xmath192 . from the study of it
we can gain some information about the low energy behavior of the model .
we expand the equilibrium distribution @xmath193.\ ] ] using the notation @xmath194 , we obtain @xmath195 to invert this equation , we approximate @xmath196 by replacing @xmath197 : @xmath198 we can now can use property ( [ property2b ] ) of the @xmath199 matrices to find that the sum of the @xmath199 s reduces to the incidence matrix of @xmath200 . hence ,
inverting equation ( [ toinvert ] ) , we obtain @xmath201 where @xmath202 and @xmath203 is an effective constant in front of the ising term of the hamiltonian : @xmath204 it is interesting to note that , thanks to property ( [ property2a ] ) , we can sum all the loop terms up to a finite number @xmath205 in the hamiltonian if we assume the mean field theory approximation . inserting the couplings , we find @xmath206 it is interesting to note that in the limit @xmath207 , or @xmath208 ( where @xmath159 tends to a finite number for @xmath165 ) , this effective constant tends to @xmath209 . we interpret this as the fact that at low temperature the loops become less and less important and the model is dominated by the ising term . in particular , since the external `` magnetic '' field is given by @xmath26 and is assumed to be nonzero , it is not surprising that at @xmath171 the average valence , the equivalent `` magnetization '' , approaches this value .
we note that the @xmath210 limit does not behave well unless @xmath123 .
higher loops are highly non - local objects . for a given pair of edges ,
all the @xmath120-loops based on these edges span the whole graph already at @xmath211 , while of course this is not the case for @xmath212-loops . as a result , in formula ( [ effective2 ] )
there is a factor proportional to @xmath213 which is not present at @xmath123 .
the quantum graphity model @xcite allows for degenerate _ on _ states on the edges or the vertices of the graph .
degeneracy of edge states is necessary , for instance , in order to have emergent matter via the string - net condensation mechanism of levin and wen .
degeneracy requires modifying our calculations above and we will address it in this subsection . the first possible generalization of the quantum graphity model is to introduce a hilbert space on the edges of the form ( [ fermhilb ] ) : @xmath214 this changes the degeneracy number in equation ( [ entropy ] ) . with @xmath215 and @xmath216 , we obtain @xmath153=-\log(\frac{m_\beta^3}{1-m_\beta})-2.\ ] ] the equilibrium distribution solves this equation . if we put @xmath217 - 2)]$ ] , we have @xmath218 obtained from the only real solution of the third order polynomial equation @xmath219 . using the same procedure as before ,
it is easy to see that equation ( [ cond1 ] ) remains unchanged : the low energy average valence is the same in both cases .
however , the first derivative , that is , the coefficient of the @xmath168 term in the taylor expansion of the average valence in the temperature , changes , so that @xmath220 ( with the obvious notation for the two coefficients ) .
this phenomenon can be understood using the following argument . at high temperature ,
the two models behave in the same way , forcing the valence to be high .
when the temperature drops , @xmath159 also goes down .
while in the ( 1,1 ) case the phase space of the _ on _ edges is the same as that of the _ off _ edges , in the ( 3,1 ) case the system prefers to stay in the _ on _ state .
thus , when the temperature decreases the system ( 3,1 ) is , at first , slowly converging to the ground state , but at @xmath171 it is forced to go to the ground state . for this reason
the function @xmath159 has a greater derivative near @xmath171 in the ( 3,1 ) case .
in this paper we introduced a technique to map the quantum graphity hamiltonian on the line graph of a complete graph .
this procedure requires the introduction of the kirchhoff matrix of a graph and the @xmath50-string matrices related to these .
this mapping is general and not specific to quantum graphity . using this mapping in a weak coupling approximation of the model , the mean field theory approximation and the low temperature expansion , we studied the properties of the model near zero temperature after having identified the average degree as a order parameter .
we found that the model is dual to an ising model with external nonzero magnetic field if we neglect the interaction terms due to loops .
in particular , we showed that the average valence is naturally a good order parameter for the mean field theory approximation and we found , implicitly , its average distribution using the mapped hamiltonian and the mean field theory approximation for the 3-loop term .
we then studied the susceptibility function and showed how the duality with the ising model can help to interpret the results .
in particular , the parameter @xmath26 plays the role of the external magnetic field . since @xmath26 is assumed to be never zero , the model has no phase transition and at @xmath171 the system goes to the ground state as expected .
in fact , we found that at zero temperature the mean valence is determined by the parameter @xmath26 and we approximated the first order correction , showing the dependence on the coupling constants of the model .
while these results were expected on general grounds , the mapping used here simplified the problem and allowed a quantitative analysis .
what emerged from the study of the average distribution for the valence is that , if the vertex valence term dominates ( @xmath221 ) , the loop term corrections to the average valence of the ground state are suppressed at @xmath171 .
we found the dependence on the coupling constants explicitly .
this result is confirmed by the study of the susceptibility .
thanks to the mean field theory approximation , we found the contribution of all loops to the susceptibility and showed that the susceptibility function tends to the ising one when @xmath168 approaches zero .
we then applied this procedure to the degenerate case , and showed that the degeneracy does not change the average valence at low temperature but only the speed with which this ground state is reached .
as a final remark , we stress that the vertex valence is an important quantity in the model .
in fact , as shown in @xcite using the lieb - robinson bound , the speed with which information can propagate on graphs is bounded by a valence - dependent quantity .
for this reason , as the temperature drops , the speed of the emergent light field must drop with the valence .
we would like to stress that we assumed that the system was at equilibrium with an external bath .
the problem of the required external bath in the model has been studied in @xcite , where additional degrees of freedom ( bosonic particles on vertices ) were introduced to let the system thermalize and reach an equilibrium distribution .
the authors are indebted to cohl furey , piero porta mana , isabeau prmont - schwarz , simone severini , lee smolin , and especially alioscia hamma for reading the manuscript and providing useful advice and comments .
this work was supported by nserc grant rgpin-312738 - 2007 and the humboldt foundation .
research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research & innovation .
g. volovik , _ the universe in a helium droplet _ , oxford university press ( 2009 ) .
see , for instance , michael levin , xiao - gang wen , rev .
77 , 871 - 879 ( 2005 ) , cond - mat/0407140 w. g. unruh , phys . rev .
d 51 , 2827 - 2838 ( 1995 ) , arxiv : gr - qc/9409008 ; m. visser , s. weinfurtner , posqg - ph:042,2007 , arxiv:0712.0427 .
see , for instance , n. seiberg , rapporteur talk at the 23rd solvay conference in physics , december , 2005 , arxiv : hep - th/0601234 .
s. lloyd , arxiv : quant - ph/0501135 .
f. markopoulou , in approaches to quantum gravity - towards a new understanding of space , time and matter , edited by d. oriti , cambridge univ .
press 2009 .
c.rovelli , quantum gravity , cambridge u. press , new york ( 2004 ) ; t.thiemann , `` introduction to modern canonical quantum general relativity , '' arxiv : gr - qc/0110034 . ; a.ashtekar and j.lewandowski , `` background independent quantum gravity : a status report , '' class . quant
. grav .
* 21 * , r53 ( 2004 ) [ arxiv : gr - qc/0404018 ] .
a. perez , `` introduction to loop quantum gravity and spin foams '' , arxiv : gr - qc/0409061 ; d. oriti , `` quantum gravity as a quantum field theory of simplicial geometry '' , arxiv : gr - qc/0512103 ; l. freidel , r. gurau , d. oriti , `` group field theory renormalization - the 3d case : power counting of divergences '' , arxiv:0905.3772 j. ambjorn , j.jurkiewicz and r.loll , `` quantum gravity , or the art of building spacetime , '' arxiv : hep - th/0604212 .
m.levin and x. g. wen , `` fermions , strings , and gauge fields in lattice spin models , '' phys .
b * 67 * , 245316 ( 2003 ) [ arxiv : cond - mat/0302460 ] ; m.a.levin and x.g .
wen , `` string - net condensation : a physical mechanism for topological phases , '' phys .
b * 71 * , 045110 ( 2005 ) [ arxiv : cond - mat/0404617 ] ; m.levin and x.g.wen , `` quantum ether : photons and electrons from a rotor model , '' arxiv : hep - th/0507118 .
a. hamma , f. markopoulou , s. lloyd , f. caravelli , s. severini , k. markstrom , `` a quantum bose - hubbard model with evolving graph as toy model for emergent spacetime '' , arxiv:0911.5075 beineke , l. w. ( 1968 ) , `` derived graphs of digraphs '' , in sachs , h. ; voss , h .- j . ; walter , h .- j . , beitrge zur graphentheorie , leipzig : teubner , pp . 1733
. g. parisi , statistical field theory ( addison - wesley , reading , mass . ) , 1988 i. prmont - schwarz , a. hamma , i. klich , f. markopoulou - kalamara , lieb - robinson bounds for commutator - bounded operators , arxiv:0912.4544v1 [ quant - ph ] ; a. hamma , f. markopoulou , i. premont - schwarz , s. severini , lieb - robinson bounds and the speed of light from topological order , 10.1103/physrevlett.102.017204 , arxiv:0808.2495v2 [ quant - ph ]
|
we present a mapping of dynamical graphs and , in particular , the graphs used in the quantum graphity models for emergent geometry , to an ising hamiltonian on the line graph of a complete graph with a fixed number of vertices .
we use this method to study the properties of quantum graphity models at low temperature in the limit in which the valence coupling constant of the model is much greater than the coupling constants of the loop terms . using mean field theory
we find that an order parameter for the model is the average valence of the graph .
we calculate the equilibrium distribution for the valence as an implicit function of the temperature . in the approximation
in which the temperature is low , we find the first two taylor coefficients of the valence in the temperature expansion .
a discussion of the susceptibility function and a generalization of the model are given in the end .
|
among the known exoplanets , a few special cases stand out as the objects that inaugurated the study of the physics of earth - like exoplanets .
b @xcite is the first super - earth for which the mass and radius have been estimated , and provided the first real constraints on models of the formation , structure , and evolution of small and potentially rocky exoplanets .
kepler-10b has recently been announced as the first rocky planet found by the mission ( batalha et al .
2010 ) , with a mass of @xmath2 and a radius of @xmath3 .
we now also have examples of interesting planets that are intermediate in mass and radius between the earth and neptune , such as gj1214b , with @xmath4 ( charbonneau et al .
2009 ) and @xmath5 @xcite . the first detection of a super - earth was made possible by the successful satellite @xcite .
the discovery was made in the course of observations in the first long run of this mission in the direction of the anti - center of the galaxy ( lra01 ) , which took place from 2007 october to 2008 march .
a small transit - like signal was identified with a depth of 0.35 mmag and duration of 1.3 hr , recurring with a period of 0.8535 days and being consistent with a super - earth size planet orbiting a bright ( @xmath6 , @xmath7 ) g9 dwarf star .
the discovery triggered a series of follow - up observations to clarify the origin of the shallow transit signal .
the detection and the follow - up campaign have been fully described by @xcite .
bruntt et al.(2010 ) subsequently performed a detailed spectroscopic analysis of the star and determined an improved stellar radius of @xmath8 .
this resulted in a revised planet radius of @xmath9 .
@xcite reported on an extensive observational campaign carried out with the harps instrument on the eso 3.6 m telescope at la silla , with the goal of detecting the doppler signal of this small object and measuring its mass .
is an active star , however , with starspot - induced photometric variability at the @xmath102% percent level modulated by the stellar rotation period of about 23 days .
the radial velocity of the star is dominated by an irregular signal with an amplitude several times larger than the sought - after signature of the transiting planet . under these circumstances , measuring the mass accurately is a challenging task that depends strongly on how the activity - induced variability is handled , on assumptions about possible additional non - transiting planets that may be contributing ( up to two have been considered ) , and on the attitude toward systematic errors .
@xcite produced the first mass estimate of @xmath11 .
subsequent authors have reported different values using _ the same harps observations _ or subsets thereof .
these estimates are not always consistent with each other within their formal errors , and vary considerably in significance level : @xmath12 @xcite , @xmath13 @xcite , @xmath14 @xcite , @xmath15 @xcite , and @xmath16 @xcite .
the impact of the discrepancies is not insignificant , as these estimates lead to rather different mean densities and therefore different internal structures for the planet as inferred from current theory . while some of the more recent determinations appear to favor a higher mass for the planet , at the 2@xmath17 level the estimates range all the way from 0 to 10@xmath18 , and for some of the lower values @xcite the statistical significance of the doppler detection of b is considerably less compelling .
similar difficulties are expected to occur for other candidates with shallow transits , in which `` confirmation '' in the usual sense of the word by the detection of a radial - velocity signature that is at the limit of detectability with current instrumentation will be very challenging . with the recent announcement of a large number of shallow transit candidates discovered by the satellite @xcite ,
obtaining assurance that these signals correspond to bona - fide super - earth size planets , as opposed to a false positive , is among the most urgent tasks that lie ahead , and is of primary importance for the statistical interpretation of the results .
motivated by the lingering problems with the b mass determination described above , and the implications for the robustness of the detection , we use this case here to illustrate the application of a powerful technique to help `` validate '' shallow transit candidates _
independently of any radial - velocity information_. it makes use of near - infrared observations with the space telescope , and is based on the premise that true transits are achromatic signals , to first order , so that the transit depth as observed in the near - infrared should be the same as in the passband ( ignoring the effects of limb darkening ) . if the candidate is the result of a blend , however , the depth can be significantly different .
as an example , if the observed transit were due to an object eclipsing a background star of @xmath19 and sun - like age and metallicity , then its transit in the spitzer 4.5 bandpass would be 3.2 times deeper than in the bandpass . with its infrared passband , affords the maximal wavelength separation from the photometry ( passband around 650 nm ) , and places very strong constraints on possible false positive scenarios , as we describe below . even with these constraints , and others available from follow - up observations carried out and reported by the team ,
it is not possible to rule out all possible blend configurations for b , as recognized also by @xcite .
thus , the main goal of this work is to more exhaustively explore the wide variety of false positive scenarios that can mimic the light curve , to obtain a realistic estimate of the blend frequency that may be expected .
we aim to provide an independent assessment of the confidence level that the signal is of planetary nature . rather than focusing solely on the likelihood of a blend ( frequentist approach ) , as in previous studies ,
we adopt a bayesian approach in which we compare the blend frequency with a prior for the likelihood of a planet ( odds ratio ) .
to evaluate the blend frequency we make use of the technique introduced by @xcite , with further developments as described by @xcite .
this methodology has been applied successfully to validate a number of shallow transit signals from the mission including kepler-9d , kepler-10c , kepler-11 g , and kepler-19b @xcite .
we begin by describing our observations ( section [ sec : spitzer ] ) , and then briefly summarizing the data used here along with other follow - up observations relevant to this investigation ( section [ sec : photometry ] ) .
this is followed by the analysis that examines the vast space of parameters for false positives by synthesizing realistic blend light curves and comparing them with the photometry ( section [ sec : blender ] ) .
we next estimate the expected frequency of blends and compare it with the expected frequency of planets ( section [ sec : validation ] ) .
as shown below , this analysis is able to validate b as a planet without relying on any radial - velocity information .
the infrared array camera ( irac ; * ? ? ? * ) of the space telescope @xcite obtains simultaneous images in four bandpasses .
a @xmath20 field of view ( fov ) is imaged in one pair of bandpasses ( 3.6 and 5.8 ) , and a nearly adjacent fov imaged in the second pair ( 4.5 and 8.0 ) .
the two blue channels employ insb detectors , whereas the red channels use si : as ibc detectors .
all four arrays are @xmath21 pixels .
while the present warm mission is restricted to the two shorter wavelengths , the data discussed in this section were obtained just prior to the spacecraft entering that phase in may of 2009 .
we elected to monitor in only one channel pair ( 4.5 and 8.0 ) , as even if the stellar flux in these bandpasses is slightly smaller , we have obtained more precise observations in the past using this pair @xcite .
it is also farther in wavelength from the bandpass , and the expected depth difference if the signal comes from a blend scales with the wavelength difference , as we describe later .
we used irac to observe the primary eclipse of b on ut 2009 april 22 , 23 , 24 and 25 , obtaining data at 4.5 and 8.0 .
we were able to observe it in full array mode in both channels for a total duration of 17.0 hours , including four transits lasting 75 minutes each .
we observed the target in the irac stellar mode , in which the camera gathers simultaneously two 10.4s integrations at 4.5 and 8.0 .
therefore , we gathered respectively 1075 , 1212 , 1212 , and 1212 images both at 4.5 and 8.0 during the four transits we observed .
our goal was to detect the transit signal at 4.5 , but as the 8.0observations are simultaneous and automatic , we present them hereafter for completeness .
we describe below our observations in two sections , as the insb detectors used for irac channels at 4.5 require a different treatment than the si : as detectors of the irac 8.0 channel .
we were mindful that the signal we were looking for was at the limit of what was possible to obtain with a few transits from , as it has mainly been used to look at brighter stars and deeper eclipses .
our expectation in terms of statistical significance for the detection of a single transit was @xmath22 at 4.5 ( @xmath23 when combining the four transits ) , scaled on the detection level we achieved in our previous studies of tres-1 , tres-3 , and tres-4 @xcite .
the contribution of the background to the total flux from is low in the 4.5 irac bandpass , contributing only 0.35% to the total flux in an aperture with a 5-pixel diameter centered on the position of the star .
we obtained the lowest rms time series using an aperture with a radius of 5.0 pixels .
we allowed the position of our aperture to shift with the position of the star in each image .
we estimated the background in each image from an annulus with an inner radius of 12 pixels and an outer radius of 20 pixels centered on the position of the star .
we determined the position of the star in each image by fitting a two - dimensional gaussian to the position of the star .
agol et al .
( 2010 ) have recently completed a comparative analysis of different methods to estimate the stellar centroid in images .
their best results are obtained by fitting a two - dimensional gaussian to the star psf , compared to a flux - weighted centroiding , and parabolic fitting .
this gaussian algorithm uses the @xmath24 pixel sub - array from the image centered on the brightest pixel .
it fits a two - dimensional gaussian to this array , allowing its center , amplitude , and width to vary .
it then uses a non - linear levenberg - marquardt algorithm to optimize these parameters @xcite .
we compared this technique with the position - weighted sum of the flux in a 4-pixel radius disk centered on the approximate position of the star , which we previously used on similar magnitude targets @xcite .
the gaussian fit proved to be better regarding two criteria .
first , the scatter between the position estimates was 1.3 times smaller .
second , the scatter of the differential position between our target star and a nearby reference star ( 2 magnitudes fainter in channel 2 and at 28 ) was smaller and did not show any correlation with the intra - pixel position .
the dominant instrumental systematic effect in the first two irac bandpasses is due to a well - known intra - pixel sensitivity @xcite .
fluxes at these two wavelengths show a strong correlation with the intra - pixel position of the star on the detector , at a level comparable to the expected depth of the eclipse .
we used the following parameters to fit the observed flux as a linear function of the subpixel position : @xmath25\,,\ ] ] where @xmath26 is the original flux from the star , @xmath27 is the measured flux , @xmath28 and @xmath29 denote the location of the gaussian - fit centroid of the star on the array , @xmath30 and @xmath31 are the coordinates of the center of the pixel containing the peak of the star s point spread function , and @xmath32@xmath33 are free parameters in the fit .
we excluded the in - transit measurements , based on the known ephemeris , in order to avoid suppressing the transit depth .
for the @xmath28 and @xmath29 positions above we calculated the centroid of the target in each image and found that the pointing jitter was around 0.12 pixels ( @xmath34 ) over the course of a visit in both @xmath28 and @xmath29 .
the pointing drift of the telescope appears to occur on a longer timescale than the exposures .
therefore , instead of using the actual position estimate at individual exposure times , we smoothed the @xmath28 and @xmath29 curves as a function of time and used the smoothed position instead .
in contrast to previous observations of hd189733 and hd209458 @xcite , we found that adding quadratic or higher - order terms to this equation , or even cross - terms , did not improve the fit significantly , likely due to the lower signal - to - noise ratio of the present observations . after correcting for the intra - pixel sensitivity ,
a decreasing trend was still visible that is likely to be an instrumental effect related to the detector or telescope , and has also been seen in observations of tres-3 and tres-4 @xcite , two stars with similar brightness .
we corrected for this effect by fitting the data in both channels with a linear function of time .
this term was fitted simultaneously with the correction for the intra - pixel sensitivity , so that we can accurately characterize the additional uncertainty in the depth and timing of the eclipse introduced by these corrections .
that is to say , we solved for four parameters including a constant term , a linear function of @xmath28 position , a linear function of @xmath29 position , and a linear function of time .
we also trimmed the first 30 minutes of data that show a larger scatter , as we have done in previous cases .
the fit was performed with a markov chain monte carlo ( mcmc ) method @xcite with @xmath35 steps , where we set the uncertainty of the individual measurements equal to the standard deviation of the out - of - transit data after correction for the various detector effects .
prior to the fit we carried out an initial trimming within our aperture , discarding outliers farther than 3.5@xmath17 from the local median flux ( defined as the median of a 15-minute window centered on the data point ) .
we also removed measurements for which the identified position of the photocenter @xmath28 or @xmath29 deviated by more than 3.5@xmath17 from the same 15-minute median position .
this global trimming excludes 6.6% of the data points in the four visits in the 4.5 bandpass .
we excluded outliers greater than 3.5@xmath17 during each step of the chain , as determined using the residuals from the model light curve , from our evaluation of the @xmath36 function .
we rescaled the value of the @xmath36 function to account for the fact that we are varying the number of pixels included in the fit . after producing the chain
, we searched for the point in the chain where the @xmath36 value first falls below the median of all the @xmath36 values ( i.e. , where the code had first found the best - fit solution ) , and discarded all the steps up to that point .
figure [ four_eclipses ] shows the four individual light curves and the respective fits for the instrumental effects that we removed from these curves before looking for the transit signal itself .
next we carried out a second markov chain fit of the transit signal itself on the trimmed data .
we initially allowed the individual depths and times of the four transits to vary independently , along with the normalized semimajor axis @xmath37 , and the inclination angle of the orbit .
we calculated our transit curves using the formulation by @xcite .
although small at 4.5 , limb darkening was taken into account in our modeling according to the four - parameter law by @xcite , with coefficients taken from the work of @xcite and stellar parameters @xmath38k , @xmath39 , and @xmath40 } = + 0.12 \pm 0.06 $ ] from @xcite .
while the individual depths and transit time parameters were consistent with transits occurring at the expected period from the light curve , the signal - to - noise ratios were too poor to provide a meaningful constraint on the geometric parameters .
we therefore chose to restrict the space of free parameters for the markov chain analysis to the transit depth , @xmath37 , inclination angle , and period , and we held the transit epoch fixed at the value reported by @xcite .
non - random sources of noise in transit and eclipse photometry such as instrumental systematics and stellar variability could dominate the error budget in the derived system parameters .
this is true for ground - based data , and also turns out to be true for space - based data
. the higher stability of space measurements is offset by the fact that smaller effects are being measured , and correspondingly smaller levels of random error are achieved by collecting more signal .
@xcite showed how neglecting this type of noise could lead to an underestimate of the actual uncertainties by a large factor . in the present study ,
a realistic error estimate is important and we therefore attempted to assess the possible impact of non - random noise .
the presence of correlated noise is obvious in our raw photometric sequence .
we corrected for it to first order , but obviously the correction can not be perfect .
we used the single - parameter description of the correlated component of the noise proposed by @xcite , with the further simplification of @xcite adapted to regularly - sampled data , to estimate the impact of the residual systematics after decorrelation .
we repeated the mcmc analysis by using modified uncertainties @xmath41 , where @xmath42 and @xmath43 are the ` white ' and ` red ' components of the noise ( i.e. , random and correlated , respectively ) , and @xmath44 is the number of data points during the eclipse .
we estimated @xmath45 from the dispersion of the flux after decorrelation obtained with slightly different , reasonable decorrelation procedures .
we used @xmath46mag as the dispersion between three different decorrelation techniques ( i.e. , fitting the intra - pixel answer and the transit simultaneously , adding quadratic terms to the intra - pixel answer , and the decorrelation technique previously described ) .
table [ eclipse_depths ] collects our results , and the time series is shown in figure [ folded ] along with the fitted model .
the parameters we derive for the planet are in good agreement with those based on the light curve as observed by the satellite . in particular , the near - infrared and optical depths are consistent within the errors .
also shown in figure [ folded ] is a model based on the optical lightcurve parameters , for comparison with the curve . although we do not reach the precision @xcite obtained for the geometric parameters based on their detection of 153 individual transits , we were able to improve the precision in the period determination of the planet significantly , as our observations were gathered some 640 planetary orbits after the original data .
lcc @xmath47 ( day ) & @xmath48 & @xmath49 + @xmath37 & @xmath50 & @xmath51 + inclination ( deg ) & @xmath52 & @xmath53 + depth ( mmag ) & @xmath54 & @xmath55 + to put our results on a more quantitative basis , we computed the reduced @xmath36 values corresponding to the case where no transit at all is shown in the data , and the case of a transit with a depth corresponding to a planet such as b. we find that the super - earth scenario gives a reduced @xmath36 ( 1.0039 ) very close to the best fit to the data ( 1.0033 ) , whereas the no - transit scenario provides a poor fit ( 1.0276 ) .
since the transit depth we measure in the near infrared is about the same as in the optical , this argues against blends composed of stars of much later spectral type .
importantly , we can use the error bar on the measured depth to rule out blends involving stars of a different temperature that would have produced a significantly different depth in infrared . indeed ,
the depth of a blend in the bandpass compared to the optical depends solely on the infrared contribution of the blending star to the total flux relative to its contribution in the bandpass . for example , if the relative contribution in the infrared of a background eclipsing binary is three times larger in infrared than in the visible , then the observed transit in the infrared would be three times deeper .
the relative infrared - to - visible contribution of a background star is also independent of its distance to the target , to first order ( i.e. , ignoring the effects of interstellar dust ) . as a consequence , for a given target star ,
the observed depth increase or decrease of a blend when observed in the bandpass is only a function of the spectral type of the contaminating star .
using an isochrone of solar age and metallicity representative of the background stars in the galactic plane , it is straightforward to compute the change in the relative infrared - to - visible flux contribution of such a star as a function of its mass , and hence the depth increase or decrease in the 4.5 and 8@xmath56 bandpasses that a blend would display . with the target properties ( mass @xmath57 , age @xmath58gyr ) from @xcite , we find for that all blended stars below @xmath59 would show a transit deeper than the 3-@xmath17 upper limit observed with at 4.5@xmath56 ( 0.77mmag ) .
we make use of this important constraint later in sect .
[ sec : validation ] to eliminate a large fraction of potential false positives for b. for the observations at 8.0 we used the `` preflash '' technique ( knutson et al .
2009 ) , in which we pointed the telescope for 30 minutes towards a bright region before observing .
this was completed in order to reduce the amplitude of the detector `` ramp '' at 8.0 , effectively pre - loading the pixels on which the target star would be pointed .
previous secondary eclipse studies ( e.g. , knutson et al . 2008 ) have shown that psf - fitting can provide a better signal - to - noise ratio at longer wavelengths . at longer wavelengths the flux from the star is smaller and
the zodiacal background is larger ; we find that the background contributes 20% of the total flux in a 3-pixel aperture at 8.0 . at 8.0
, we found that the relative scatter in the time series after model fitting from the psf fits was 15% higher than in the time series from aperture photometry with a 3.0 pixel radius . as a result of this increased scatter ,
which is likely produced by discrepancies between the model psf and the observed psf , we concluded that aperture photometry is also preferable .
we compared the time series using apertures ranging from 3.0 to 4.5 pixels and found consistent results in all cases , but with a scatter that increases with the radius of the photometric aperture .
previous observations ( e.g. , * ? ? ?
* ) have shown that the ramp is well described as following an asymptotic shape , with a steeper rise in the first 30 minutes of observations .
we corrected for this effect by fitting our time series with the following function : @xmath60\,,\ ] ] where @xmath26 is the original flux from the star , @xmath27 is the measured flux , and @xmath61 is the elapsed time in days since the start of the observations .
as the expected statistical significance of the signal at 8.0 was very small ( i.e. , 1.1@xmath17 ) , we chose to fix the transit time and period of the transit to values found in our 4.5 study and verified that the residual signal in the four visits at 8.0 was compatible with the transiting planet scenario .
we carried out a markov chain monte carlo fit to the data as described in sect .
[ sec : short_norm ] , simultaneously fitting eq .
[ eq2 ] and a transit model , with the depth as the only free parameter in the latter .
no significant correlations were found between the variables . as a further check we repeated these fits adding a quadratic term of the form @xmath62 in eq .
[ eq2 ] , and found that the value of @xmath36 for our best - fit solution was similar to the previous value , so we chose not to include this additional term .
figure [ folded8 ] shows the detrended ( from the ramp ) , trimmed , binned , and folded channel 4 light curve .
we estimated the best - fit transit depth using a markov chain monte carlo fit in the same way as we did for the 4.5 data , and found an eclipse depth of @xmath63mmag .
the noise level is too large to say anything about a dip corresponding to the planet eclipse , but it disallows a blend scenario involving a significantly redder star .
specifically , proceeding in the same way as described at the end of sect .
[ sec : short_norm ] , a star of 0.5@xmath64 would be excluded at the 3@xmath17 level .
the photometric observations used in the next section to investigate blends are essentially the same as used by @xcite , with a somewhat different de - trending of the data .
we started with the n2 ( science grade ) light curve , discarding the initial few days , which had a time sampling of 512s , and worked only with the remainder , where the time - sampling was 32s ( see * ? ? ? * for a detailed description of the n2 light curve ) .
we identified and clipped outliers using a running 5-point median filter , and then modeled the intrinsic stellar variability using an iterative non - linear filter @xcite .
this consists in applying a 5-point boxcar filter followed by a median filter with a width of 1 day , @xmath65 clipping the residuals , and iterating until no more points are clipped .
the resulting slowly - varying component was then subtracted from the original light curve , to give the time - series used in this work .
we made use of the white - light data only .
of the extensive follow - up observations carried out by the team as described by @xcite , the most relevant for our study are the high - resolution imaging observations obtained with the naco instrument on the vlt .
these data exclude any blends capable of mimicking the signal outside of 04 from the target down to 6.5 mag fainter than the target . additionally , the spectroscopic observations reported by these authors allowed them to rule out companions that would be bright enough to be visible in the spectrum ( as a second set of lines ) . in particular , their near - infrared ( @xmath66-band ) spectroscopy with crires on the vlt rules out most companions brighter than about 7% of the flux of the target in that band ( @xmath67 ) .
for the analysis below we adopt a more conservative limit of @xmath68 .
blend scenarios satisfying the corot observations were explored with the technique @xcite , by examining the quality of the fit to the photometry of a large array of synthetic model light curves .
these synthetic light curves result from the combined light of three objects composing the blend : the main target , and an unresolved eclipsing pair along the line of sight with the same period as detected in , whose eclipses are attenuated by the light from the target
. the eclipsing system may be physically associated ( hierarchical triple ) , or may be in the background or foreground , and the eclipsing pair may consist of two stars , or a star transited by a larger planet .
uses the detailed shape of the transit light curve to weed out scenarios that lead to the wrong shape for a transit .
the properties of the three objects were taken from model isochrones by @xcite in order to synthesize light curves , and these artificial light curves were compared with the observations in a @xmath36 sense ( i.e. , by computing the sum of the squared residuals normalized by the photometric uncertainties ) . for practical reasons
blend light curves were calculated here in the passband , which is sufficiently similar to the passband of the satellite for our purposes , in both the central wavelength and width .
the properties of the main star ( essentially its intrinsic brightness , which affects the dilution of the eclipses ) were constrained by the spectroscopic observations of ( @xmath69k , @xmath70 , @xmath40 } = + 0.12 \pm 0.06 $ ] ; * ? ? ?
* ; * ? ? ?
* ) , and held fixed . those of the other two objects ( referred to here as the ` secondary ' for the eclipsed star and ` tertiary ' for the eclipsing object ) were varied over a wide range and parametrized by mass . for hierarchical triples
the stellar components were constrained to lie on the same isochrone , while for background / foreground blends the properties of the intruding star were taken from a representative solar - metallicity , 3gyr isochrone .
the distance between the binary ( @xmath71 ) and the target ( @xmath72 ) was expressed for convenience in terms of the distance modulus difference ( @xmath73 $ ] ) , which is independent of interstellar extinction , and varied between @xmath74 and + 10 .
the impact parameter was allowed to vary between 0 and unity , and the mass of the secondary stars ranged from 0.1@xmath64 to 1.4@xmath64 . for stellar tertiaries we explored masses from 0.1@xmath64 up to the mass of the secondary ; for tertiaries that are planets ( assumed to be dark ) we allowed their sizes to be up to 2.0@xmath75 .
we restricted our blend simulations to circular orbits for all eclipsing pairs , as no stellar or planetary systems with periods as short as that of b are known to have eccentric orbits , nor are expected to from theoretical arguments ( see , e.g. , * ? ? ?
synthetic light curves were generated with a detailed eclipsing binary code , including proximity effects ( tidal and rotational distortions ) , limb - darkening , gravity brightening , and contamination from third light .
differential extinction was included for chance alignments .
blends providing poor fits compared to the photometry in a @xmath36 sense were considered to be ruled out .
this enables us to place constraints on the kinds of objects composing the eclipsing pair that yield viable blends ( i.e. , acceptable fits ) , including their size or mass , as well as other properties of the blend such as the overall brightness and color . for further details and applications of to other transiting planet candidates , we refer the reader to the work of @xcite , @xcite , and @xcite .
we consider first the case of blends involving a background or foreground star falling within the aperture , and transited by a larger planet .
because of the very short 0.8535-day orbital period of evolved stars ( giants or subgiants ) are ruled out , so we focus in the following on main sequence stars .
we find that there is a very large range of spectral types ( masses ) permitted for the secondary star , shown in top panel of figure [ fig : bp ] , as well as a wide range of relative distances between the eclipsing pair and the target .
this is indicated by the darker areas in the figure , delimited by the white contour representing light curve fits that differ from the best transiting planet model by a @xmath36 difference corresponding to a 3-@xmath17 confidence level .
however , other constraints available for strongly restrict the number of these false positives .
this is illustrated in the bottom panel of the figure , in which these additional constraints are superposed on the same blend landscape displayed in the top panel .
in particular , by comparing the predicted @xmath76 color of each blend against the measured color of the star from exodat ( @xmath77 ; * ? ? ?
* ) , we find that a large fraction of the secondaries with significantly different masses than the primary are ruled out because the blends would be too red or too blue compared to the known color index of ( by more than 3@xmath17 ) .
these excluded regions are indicated by the blue hatched areas in the bottom panel of figure [ fig : bp ] .
other blends are excluded because the secondary star would be very bright ( within one magnitude of the target ) , and would have been noticed spectroscopically , if unresolved in the high - resolution imaging described earlier .
the section of parameter space excluded by this brightness criterion is indicated by the green hatched area , limited from above by the green diagonal line corresponding to a magnitude difference of 1.0 mag .
note that , as mentioned in sect .
[ sec : photometry ] , this is a very conservative brightness limit as a large fraction of stars within 2 or even 3 magnitudes of would most likely have been detected spectroscopically . as indicated before ,
our observations exclude secondary stars less massive than approximately 0.69@xmath64 , which are common constituents of background blends .
this additional constraint is shown by the shaded gray area leftward of 0.69@xmath64 .
there is some overlap between the gray region and the blue and green hatched areas , indicating that those blends are excluded by more than one observational constraint .
an example of a blend that provides an acceptable fit to the photometry is shown in figure [ fig : blend ] ( top panel ) .
this scenario ( location marked with a cross in figure [ fig : bp ] ) is not excluded either by its color or its brightness , but _ is _ clearly ruled out by our observations because it produces a much deeper transit in the near - infrared ( see bottom panel of figure [ fig : blend ] ) , which would have been easily detected .
to summarize , the combination of the , color , and brightness constraints removes many but not all blends involving a background or foreground star transited by a larger planet .
those that remain reside in the area of figure [ fig : bp ] ( bottom ) labeled `` allowed region '' . for the case of a background eclipsing binary composed of two stars ,
interestingly we find that no combination of relative distance and stellar properties for the eclipsing pair gives an acceptable fit to the light curve .
the reason for this is that all such blend configurations that can potentially reproduce the detailed shape of the transit also lead to out - of - eclipse brightness changes ( ellipsoidal variations ) with an amplitude that is not seen in the data , and that are a consequence of the very short orbital period .
this implies that background blends of this kind can be confidently ruled out , an important conclusion that does not follow from the original @xcite analysis .
figure [ fig : bs ] shows that the blends that yield the best fits to the photometry are excluded at the 8.3@xmath17 level or higher .
this result is significant , as it substantially reduces the overall likelihood of blends for b. we note that white dwarfs are also excluded as potential tertiaries .
although the range of their radii overlaps with those of the planets considered earlier in this section , their considerably larger mass would once again induce significant ellipsoidal variation in the light curve , which is ruled out by the observations . as was the case in the previous section , giant stars do not constitute viable secondaries because of the very short 0.8535-day period of the signal .
finally , for eclipsing binaries ( consisting of two stars ) that are physically associated with the target in a hierarchical triple configuration , we find that the blend light curves invariably have the wrong shape to mimic a true transiting planet signal , for any combination of stellar parameters for the secondary and tertiary . either the depth , duration , or the steepness of the ingress / egress phases of the transits provide a very poor match to the photometry , resulting in @xmath36 differences compared to the best transit model corresponding to several hundred @xmath17 .
these scenarios are therefore all excluded . on the other hand , if we allow the tertiaries to be planets , the blend fits are somewhat better over a wide range of masses for the secondaries when transited by a planet of the appropriate size , but not quite at the 3@xmath17 level or better .
the best @xmath36 for a blend of this type corresponds to about a 3.4@xmath17 departure from a planet model .
although our formal 3-@xmath17 limit is reasonable and consistent with common practice , it is still somewhat arbitrary and one may argue that fits that are only marginally worse might still be tolerable . even accepting this possibility
, figure [ fig : htp ] shows that _ all _ of these blends are excluded by a combination of constraints from , color index , and brightness , even those that are 10@xmath17 or more away from the quality of a planet model . from the simulations described above
, the only viable blend scenarios for b are those involving a larger planet transiting a background or foreground star .
restricts those blends to the area labeled `` allowed region '' in figure [ fig : bp ] .
these configurations involve stars between 1.0 and 5.5 mag fainter than the target within the aperture .
the _ a priori _ frequency of stars in the background or foreground of the target that are orbited by a transiting planet and are capable of mimicking the photometric signal may be estimated from the density of stars in the vicinity of , and the frequency of transiting planets with the appropriate characteristics .
the relevant area around the target is that in which stars of each brightness would go undetected in the high - resolution imaging reported by the team . to obtain the number density ( stars per square degree )
we make use of the galactic structure models of @xcite , and we perform this calculation in half - magnitude bins , as illustrated in table [ tab : blendfreq ] . for each bin we further restrict the star counts using the constraints on the mass of the secondaries supplied by ( see figure [ fig : bp ] ) .
these mass ranges are listed in column 3 , and the resulting densities appear in column 4 .
bins with no entries correspond to brightness ranges excluded by . for the maximum angular separation ( @xmath78 , column 5 ) at which stars of each brightness would escape detection
we have adopted the value 04 , based on the report by @xcite that companions outside this range would have been seen in their vlt / naco image down to a magnitude difference of 6.5 mag compared to the target .
we note that this @xmath78 value is a conservative limit , as stars at closer separations would have been detected if they had smaller magnitude differences , which would reduce the blend frequency . the result for the number of background stars in each magnitude bin is given in column 6 , in units of @xmath79 . to estimate the frequency of transiting planets that might be expected to orbit these stars ( and lead to a false positive ) we rely on the results from @xcite , who reported a total of 1235 planet candidates among the 156,453 targets observed during the first four months of the mission .
these signals have not yet been confirmed to be caused by planets , and therefore remain candidates until they can be thoroughly followed up .
however , the rate of false positives in this sample is expected to be quite small ( @xmath80 or less ; see * ? ? ?
* ) , so our results will not be significantly affected by the assumption that all of the candidates are planets .
we further assume that the census of @xcite is largely complete .
after accounting for the additional constraint on the range of planet sizes for blends of this kind ( tertiaries of 0.241.42@xmath75 ) , we find that the transiting planet frequency is @xmath81 .
multiplying this frequency by the star counts in column 6 of table [ tab : blendfreq ] , we arrive at a total blend frequency ( bf ) listed at the bottom of column 7 , @xmath82 .
this figure represents the _ a priori _ likelihood of a false positive , and we note that it is approximately 3 orders of magnitude smaller than indicated by the calculations of @xcite .
however , we do not consider this to represent the `` false alarm rate '' , as the expected likelihood of a transiting planet of the characteristics implied by the transit signal is also very small .
we adopt here a bayesian approach analogous to that employed to validate previous candidates , in which our confidence in the planetary nature of the signal will depend on how the blend likelihood compares to the _ a priori _ likelihood of a true transiting planet ( pf ) , addressed below .
thus , we seek to estimate the _ odds ratio _ pf / bf .
this is a significant conceptual difference compared to the frequentist approach by @xcite , who considered only the likelihood of a blend ( bf ) .
implicit in the blend frequency calculation above ( @xmath83 ) is the 3@xmath17 criterion on the quality of the light curve fit relative to a transit model fit that we used as a condition for a blend scenario to be acceptable ( see sect .
[ sec : blender ] ) . for a fair comparison ,
we use a similar 3@xmath17 criterion to establish the _ a priori _ transiting planet frequency ( pf ) , for the numerator of our odds ratio .
we estimate pf by counting the candidates in the @xcite sample that have radii within 3@xmath17 of the value determined from the best fit to the data ( @xmath84 ; * ? ? ?
we find 231 candidates within this range , giving a planet frequency @xmath85 .
thus , the likelihood of a planet is more than 3,500 times greater than that of a false positive ( @xmath86 ) , which we consider sufficient to independently validate b as a true planet with a high degree of confidence .
we note that our blend frequency calculation assumes the 1,235 candidates cataloged by @xcite are all true planets . if we were to assume conservatively that as many as 50% are false positives in the radius range specific to the transiting planet case ( an unlikely proposition that is also inconsistent with other evidence ; see * ? ? ?
* ; * ? ? ?
* ) , the planet likelihood would still be nearly 1,800 times greater than the likelihood of a blend , implying a false alarm rate sufficiently small to validate the signal .
additionally , in estimating the _ a priori _ true planet frequency ( pf ) and the frequency of larger planets involved in blends , we have not placed any restriction on the periods of either kind of planet .
one may expect , for example , that limiting the periods of larger planets involved in blends to be similar to that of b , which is quite short , might reduce the blend frequency quite considerably . on the other hand , a similar period limit imposed on
the true planet frequency would decrease pf as well , though these effects may not be equal .
thus , it is possible that the odds ratio would be altered . as it turns out ,
our calculations without any period constraints are the most conservative .
for example , limiting the periods to be shorter than 5 days , the census of @xcite indicates that larger planets are comparatively less common than the smaller ones , resulting in an improvement in the pf / bf odds ratio of a factor of 2.2 .
restricting the periods to be less than 3 days , the odds ratio improves by a factor of 2.8 .
our observation of four transits of b with the space telescope at a wavelength of 4.5 has resulted in the detection of the transit with a depth of @xmath0mmag , which is consistent with the b planet scenario described in @xcite .
although the signal - to - noise ratio of this infrared detection is relatively low , it has been obtained with a now standard treatment of the data with no _ a priori _ knowledge of the transit parameters ( aside from the timing windows in order to plan the four visits ) .
warm is currently the only facility available that has the capability of detecting such shallow transits at wavelengths that are sufficiently separated from the ( or ) passbands to be helpful . in this case
the observations were successful , and the transit at @xmath87 is shown to have virtually the same depth as in the optical .
this places a strong constraint on the color of potential blends , which are restricted to have secondaries of similar spectral type as the primary star . in the case of b
this allows us to rule out most cool stars ( which constitute the majority of background stars ) as potential contaminants .
the detailed analysis of the photometry with combined with constraints from other observations eliminates the vast majority of possible blend scenarios .
this includes all background eclipsing binaries composed of two stars , most of the scenarios involving chance alignments with a star transited by a larger planet , and all hierarchical triple configurations .
the remaining scenarios are much less likely ( by a factor of 3,500 ) than a true transiting planet , thereby validating the planetary nature of the signal with very high confidence .
we point out that this conclusion has been reached with very conservative assumptions regarding some of the observational constraints .
for example , we have ignored the fact that the high - resolution vlt / naco imaging by @xcite permits the detection of background stars closer than our adopted limit of @xmath88 from the target if they are relatively bright . additional imaging by those authors with fastcam on the 2.5 m not telescope provides even tighter constraints that we have not used , reaching down to sensitivities of 4 magnitudes fainter than the target at 018 .
we have also not considered the full potential of spectroscopy to rule out closer companions ; we have assumed only that stars within 1.0 magnitude of the target would have been identified , whereas in reality the sensitivity of those observations ( not only the near - infrared crires spectrum , but also those from harps ) is probably much greater ( @xmath89 23mag ) .
furthermore , we have not made use of other information reported by @xcite , such as the constraints from the red , green , and blue passbands into which the light from the instrument can be split , which can further limit the pool of potential false positives .
incorporating all of these constraints can only reduce the blend frequency , resulting in an even greater confidence level for the validation of b as the first super - earth .
our observations have also significantly improved our knowledge of the ephemeris of the first transiting super - earth .
transits of b have not been observed with any other photometric facility since the end of the lra01 field observations by the satellite .
given the formal uncertainty in the orbital period reported by @xcite , the accumulated error in the predicted times of transit up to the date of our observations is approximately 20min , or about a quarter of the transit duration .
here we have improved the period determination by about a factor of four , to @xmath90 days , ensuring the transits will be recoverable for at least another decade ( provided there are no physical mechanisms operating to change the period ) .
the satellite has recently released a large number of very promising shallow transit candidates @xcite . from the high quality of this photometry
the detailed transit shapes are often very well defined , increasing the power of tools such as that make use of that information to rule out blend scenarios that result in poor fits to the data .
the exquisite relative astrometric precision delivered by the instrument provides important additional constraints from an analysis of the motion of the photocenters of the images in and out of transit ( see , e.g. , * ? ? ?
* ) , which can typically exclude 90% or more of unresolved stars falling within the photometric aperture that can potentially contaminate the flux measurements .
spectroscopy and high - resolution imaging contribute further valuable constraints .
these tools have already been used to validate small signals for which radial - velocity confirmation is currently out of reach , including possibly rocky planets .
examples include the super - earth kepler-9d @xcite , and the neptune - size planets kepler-10c @xcite , kepler-11 g @xcite , and kepler-19b @xcite .
observations provide a very effective way of ruling out a large fraction of the blend scenarios involving a background star with a different effective temperature ( or color ) than the target .
these chance alignments are typically the most serious concern regarding the nature of the transit signals . among
the most interesting candidates expected from the mission are rocky planets in the habitable zone of their parent stars .
the typically longer periods of these objects mean that methods of confirmation relying on the dynamical influence of the planet and/or the quality of the phase - folded photometry ( obtained by summing individual transit events ) will be more problematic , as their efficiency scales down with orbital period .
we anticipate that the telescope in its warm phase will prove critical for validating such objects , as its efficiency does not depend on period , but relies instead on a different intrinsic property of the transits , which is their achromaticity .
we are grateful to the anonymous referee for many very helpful comments and suggestions .
this work is based on observations made with the space telescope , which is operated by the jet propulsion laboratory , california institute of technology , under contract to nasa .
support for this work was provided by nasa through an award issued by jpl / caltech .
this research has made use of the facilities at the nasa advanced supercomputing division ( nasa ames research center ) .
aigrain , s. , & irwin , m. 2004 , , 350 , 331 baglin , a. , auvergne , m. , barge , p. , buey , j .-
, catala , c. , michel , e. , weiss , w. , & corot team 2002 , stellar structure and habitable planet finding , 485 , 17 ballard , s. et al .
2011 , , in press ( arxiv:1109.1561 ) batalha , n. m. et al . 2010 , , 713 , l103 batalha , n. m. et al .
2011 , , 729 , 27 berta , z. k. , charbonneau , d. , bean , j. , irwin , j. , burke , c. j. , dsert , j .- m . , nutzman , p. , & falco , e. e. 2011 , , 736 , 12 boisse , i. , bouchy , f. , hbrard , g. , bonfils , x. , santos , n. , & vauclair , s. 2011 , , 528 , 4 borucki , w. j. et al . 2011b , arxiv:1102.0541
bruntt , h. et al .
2010 , , 519 , a51 charbonneau , d. et al .
2005 , , 626 , 523 charbonneau , d. et al .
2008 , , 686 , 1341 charbonneau , d. et al .
2009 , , 462 , 891 claret , a. 2000 , , 363 , 1081 deming , d. , seager , s. , richardson , l. j. , & harrington , j. 2005 , , 434 , 740 fazio , g. g. et al .
, 2004 , , 154 , 10 ferraz - mello , s. , tadeu dos santos , m. , beaug , c. , michtchenko , t. a. , & rodriguez , a. 2011 , , 531 , 161 ford , e. b. 2005 , , 129 , 1706 fressin , f. , guillot , t. , & nesta , l. 2009 , , 504 , 605 fressin , f. et al .
2011 , , in press girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , , 141 , 371 hatzes , a. p. et al . 2010 , , 520 , a93 hatzes , a. p. et al .
2011 , arxiv:1105.3372 howard , a. w. et al .
2010 , science , 330 , 653 howell , s. b. et al .
2011 , submitted to aj .
jenkins , j. m. et al .
2010 , arxiv:1001.0416 knutson , h. a. , charbonneau , d. , allen , l. e. , burrows , a. , & megeath , s. t. 2008 , , 673 , 526 knutson , h. a. , charbonneau , d. , burrows , a. , et al .
2009 , , 655 , 566 lger , a. et al . 2009 , , 506 , 287 lissauer , j. j. et al .
2011 , , 470 , 53 markwardt , c. b. 2009 , astronomical society of the pacific conference series , 411 , 251 mandel , k. & agol , e. 2002 , , 580 , l171 mazeh , t. 2008 , in tidal effects in stars , planets and disks , eas publications series , eds .
goupil & j .- p .
zahn ( edp sciences ) , vol .
morton , t. d. , & johnson , j. a. 2011 , arxiv:1101.5630 morales - calderon , m. et al .
2006 , , 653 , 1454 pont , f. , zucker , s. , & queloz , d. 2006 , , 373 , 231 pont , f. , aigrain , s. , & zucker , s. 2011 , , 411 , 1953 queloz , d. et al .
2009 , , 506 , 303 reach , w. t. et al .
2005 , pasp , 117 , 978 robin , a. c. , reyl , c. , derrire , s. , & picaud , s. 2003 , , 409 , 523 sing , d. k. 2010 , , 510 , 21 torres , g. , konacki , m. , sasselov , d. d. , & jha , s. 2004 , , 614 , 979 torres , g. et al .
2011 , , 727 , 24 werner , m. w. et al . , 2004 , , 154 , 1 winn , j. n. , holman , m. j. , & fuentes , c. i. 2007 , , 133 , 11 winn , j. n. , johnson , j. a. , albrecht , s. , howard , a. w. , marcy , g. w. , crossfield , i. j. , & holman , m. j. 2009 , , 703 , l99 11.712.2 & 0.5 & & & & & + 12.212.7 & 1.0 & & & & & + 12.713.2 & 1.5 & 0.841.36 & 91 & 0.40 & 3.53 & 0.013 + 13.213.7 & 2.0 & 0.821.33 & 137 & 0.40 & 5.31 & 0.019 + 13.714.2 & 2.5 & 0.771.30 & 207 & 0.40 & 8.03 & 0.029 + 14.214.7 & 3.0 & 0.691.26 & 263 & 0.40 & 10.2 & 0.037 + 14.715.2 & 3.5 & 0.691.22 & 383 & 0.40 & 14.9 & 0.054 + 15.215.7 & 4.0 & 0.691.17 & 515 & 0.40 & 20.0 & 0.072 + 15.716.2 & 4.5 & 0.691.08 & 569 & 0.40 & 22.0 & 0.081 + 16.216.7 & 5.0 & 0.690.99 & 637 & 0.40 & 24.7 & 0.090 + 16.717.2 & 5.5 & 0.690.80 & 186 & 0.40 & 7.21 & 0.026 + 17.217.7 & 6.0 & & & & & + 17.718.2 & 6.5 & & & & & + 18.218.7 & 7.0 & & & & & + 18.719.2 & 7.5 & & & & & + 19.219.7 & 8.0 & & & & & + 19.720.2 & 8.5 & & & & & + & & 2988 & & 115.9 & 0.42 + +
|
the detection and characterization of the first transiting super - earth , b , has required an unprecedented effort in terms of telescope time and analysis .
although the star does display a radial velocity signal at the period of the planet , this has been difficult to disentangle from the intrinsic stellar variability , and pinning down the velocity amplitude has been very challenging . as a result ,
the precise value of the mass of the planet and even the extent to which it can be considered to be confirmed have been debated in the recent literature , with six mass measurements published so far based on the same spectroscopic observations , ranging from about 2 to 8 earth masses . here
we report on an independent validation of the planet discovery , using one of the fundamental properties of a transit signal : its achromaticity .
we observed four transits of b at 4.5 and 8.0 with the infrared array camera ( irac ) onboard the space telescope , in order to determine whether the depth of the transit signal in the near - infrared is consistent with that observed in the bandpass , as expected for a planet .
we detected the transit and found an average depth of @xmath0 mmag at 4.5 , which is in good agreement with the depth of @xmath1 mmag ( ignoring limb darkening ) found by . the observations at 8.0 did not yield a significant detection .
the 4.5 observations place important constraints on the kinds of astrophysical false positives that could mimic the signal . combining this with additional constraints reported earlier
, we performed an exhaustive exploration of possible blends scenarios for b using the technique .
we are able to rule out the vast majority of false positives , and the remaining ones are found to be much less likely than a true transiting planet .
we thus validate b as a bona - fide planet with a very high degree of confidence , independently of any radial - velocity information .
our observations have additionally allowed us to significantly improve the ephemeris of the planet , so that future transits should be recoverable well into the next decade . in its warm phase
is expected to be an essential tool for the validation , along the lines of the present analysis , of transiting planet candidates with shallow signals from as well as from the mission , including potentially rocky planets in the habitable zones of their parent stars .
|
it is well known that the universe could have undergone several phase transitions in the early stages of its history , most of them associated with the spontaneous symmetry breaking of some symmetry .
some examples are the quark - hadron phase transition at the qcd scale , the phase transitions associated to the electroweak @xmath0 symmetry breaking or to grand unified theories , and the peccei - quinn phase transition , related to the axion field and the strong cp problem .
cosmological phase transitions generically produce cosmic relics , such as topological defects , magnetic fields , or baryon number asymmetries , with potentially important cosmological consequences . the mechanisms for generating these relics build on the dynamics of the phase transition .
in a first - order phase transition the dynamics is essentially determined by the nucleation and expansion of bubbles . at zero temperature , when a true vacuum bubble nucleates , it rapidly begins to expand with almost the velocity of light @xcite . on the contrary , at high temperature ,
the bubble expands in a hot plasma , which is perturbed by the motion of the bubble walls .
the plasma thus opposes a resistance to the expansion , that depends on the wall velocity . as a consequence ,
bubble walls feel a friction force , which prevents them to accelerate indefinitely .
then , the velocity quickly reaches a stationary value , determined by the viscosity of the plasma and the pressure difference between the low temperature phase and the supercooled one .
these quantities depend on the model , and it is known that the friction can be large enough to prevent the wall from acquiring relativistic velocities @xcite . furthermore , the release of latent heat at the interfaces of the phase transition reheats the surrounding plasma up to a temperature that in most cases is close to the critical temperature .
consequently , the pressure difference that drives bubble expansion may decrease considerably , causing a drastic slowdown of the phase transition @xcite .
therefore , at the radiation dominated era bubble walls generically undergo non - relativistic motion .
since the heat liberated at the interfaces is taken away by sound waves , the temperature can be assumed to be homogeneous , which simplifies the analysis . in order to have a first - order phase transition
, the free energy must allow the coexistence of two phases .
therefore , we will assume that the free energy density @xmath1 depends on some order parameter ( usually a higgs field ) @xmath2 .
in a certain range of temperatures , the free energy bears two minima separated by a barrier ; one of them at @xmath3 , which corresponds to the symmetric phase , and the other at a nonzero value @xmath4 , corresponding to the broken symmetry phase . the difference in free energy density @xmath5 between some value @xmath6 of the order parameter and @xmath3 ,
is generally given by a finite - temperature effective potential .
the free energy difference between the two minima is thus given by @xmath7 at the critical temperature the two minima are degenerate , @xmath8 .
above the critical temperature the symmetric phase is the stable one , while below @xmath9 it becomes metastable , being @xmath10 the absolute minimum .
finally , at some temperature @xmath11 , the barrier between the minima disappears , and the symmetric phase becomes unstable . at some stage between the temperatures @xmath9 and @xmath12 , bubbles of broken - symmetry phase will be formed in the sea of symmetric phase .
a bubble can be described as a configuration in which the order parameter is non - vanishing inside a spherical region ( see e.g. @xcite ) . after being nucleated
, a bubble will grow with a velocity that depends on the pressure difference at the interface , @xmath13 , and on the viscosity of the hot plasma in which it expands .
in this work we will be concerned with first - order phase transitions in the radiation - dominated epoch .
our aim is to study the development of such phase transitions within a completely analytical approach .
here we concentrate on the determination of the parameters that govern relic formation ( e.g. number density of bubbles , bubble wall velocity , etc . ) .
an analytic study of cosmological consequences will be addressed in @xcite , where the results of the present analysis will be used . in section [ remnants ]
we briefly review the influence of phase transition dynamics on the mechanisms for generating cosmic remnants in first - order phase transitions .
in section [ phaseeq ] we study a phase transition that completely develops at @xmath14 , with the two phases in equilibrium .
this is a good approximation in the case of inhomogeneous nucleation in the presence of impurities . in the case of homogeneous nucleation of bubbles ,
this approximation correctly describes the evolution of the phase transition after some latent heat has been released .
an interesting feature of phase equilibrium is that it is simple enough to solve analytically , including back - reaction on the expansion of the universe .
this is due to the fact that temperature is constant all the way through the phase transition .
we thus can obtain the fraction of volume of the universe that is occupied by the low - temperature phase as a function of time , with no need of any numerical calculations .
section [ superc ] is devoted to the analysis of the phase transition in the case of homogeneous nucleation .
the main difference with the previous case is the initial stage of supercooling and quick reheating back to the critical temperature .
section [ transicion ] contains an analytical study of the phase transition dynamics . in sections [ physquant ] and [ perturb ]
we analyze the relations between the different physical parameters involved in the dynamics of the phase transition , leaving some technical discussions to the appendices .
our conclusions are summarized in section [ conc ] .
several cosmological objects may be formed in a phase transition of the universe .
their abundance and characteristics depend on details of the development of the phase transition . due to the complexity of the mechanisms by which these objects are created and the difficulty of describing the phase transition , several details of the dynamics ( e.g. the variation of the nucleation rate or the wall velocity during the transition ) are often disregarded for the sake of simplicity .
an analytical study of the phase transition is thus important since analytical expressions will help taking into account the dynamics in a more rigorous way . in this section
we review how phase transition dynamics affects the cosmological remnants .
[ [ electroweak - baryogenesis . ] ] electroweak baryogenesis . during the last two decades
there has been much interest in the possibility that the electroweak phase transition could be the framework for the generation of the baryon number asymmetry of the universe ( bau ) .
a first - order electroweak phase transition provides the three sakharov s conditions for the generation of a bau , although physics beyond the minimal standard model ( sm ) is mandatory in order to obtain a quantitatively satisfactory result ( for reviews on electroweak baryogenesis see @xcite ) . due to @xmath15 violating interactions of particles with the bubble walls , an asymmetry between left handed quarks and their antiparticles
is generated in front of the walls of expanding bubbles .
this asymmetry biases the baryon number violating sphaleron processes in the symmetric phase .
the resulting baryon asymmetry is caught by the walls and enter the bubbles , where baryon number violating processes are turned off .
it is important that the sphaleron processes be suppressed in the broken symmetry phase , in order to avoid the washout of the generated bau when equilibrium is established after the phase transition .
this requirement imposes a condition on the value of the higgs field at the temperature of the transition @xcite , @xmath16 .
since @xmath17 is the order parameter , this is a condition on the strength of the first - order phase transition .
the resulting bau depends also on the bubble wall velocity . on one hand ,
if the velocity is too large , the left - handed density perturbation will pass so quickly through a given point in space that sphaleron processes will not have enough time to produce baryons ; thus the resulting bau will be small . on the other hand , for very small velocities thermal equilibrium will be restored and the baryon asymmetry will be erased by sphalerons ; thus the bau will be small again . as a consequence , the generated baryon number will have a peak at a given wall velocity , which is of order @xmath18 @xcite .
both values of @xmath17 and @xmath19 depend on @xmath20 and vary during the phase transition .
the dependence of the wall velocity is more critical , since reheating may cause it to descend two orders of magnitude before the transition completes .
baryogenesis may be either enhanced or suppressed by this effect @xcite , depending on which side of the peak the initial velocity lies .
[ [ baryon - inhomogeneities . ] ] baryon inhomogeneities .
a general feature of cosmological phase transitions is the difference of particle masses between the high- and low - temperature phases .
these mass differences give rise to different number densities in the two phases . at the qcd phase transition ,
for instance , baryons are much heavier in the hadron phase than in the deconfined quark phase .
as the hadron phase expands , baryons are pushed into the quark phase region , leading to localized clumps of high density surrounded by large voids of low baryon density @xcite .
an important consequence is that large amplitude , small scale density fluctuations may survive until the nucleosynthesis epoch , affecting the standard scenario of big bang nucleosynthesis .
therefore , inhomogeneous nucleosynthesis may put constraints on the quark - hadron phase transition ( see e.g. @xcite ) .
moreover , if the quark phase reaches sufficiently high density , its pressure may balance that of the hadron phase .
the quark matter trapped in small regions of space forms quark plasma objects that may survive until the present epoch @xcite .
baryon inhomogeneities may also arise at the electroweak phase transition , since the amount of baryons produced through electroweak baryogenesis depends drastically on the wall velocity , and the latter has a considerable variation during the phase transition @xcite .
the geometry of the electroweak inhomogeneities is in general quite different from the qcd case .
if the bau peaks at a certain wall velocity , then the high density regions will form spherical walls , whose radius depends on the moment in the bubble evolution at which the peak velocity is attained .
furthermore , baryon number densities with the wrong sign may arise in some regions of space , depending on the baryogenesis scenario @xcite .
this gives rise to the interesting possibility of nucleosynthesis in the presence of antibaryons ( see for example @xcite ) .
however , due to baryon diffusion and neutrino inflation " , baryon inhomogeneities generated at the electroweak phase transition hardly survive until the nucleosynthesis time ( see e.g. @xcite ) .
nevertheless , they may survive until the qcd scale .
in that case , electroweak scale inhomogeneities can act as impurities for the quark - hadron phase transition ( see next section ) . in summary
, we can say that the amplitude and scale of baryon inhomogeneities generically depend on the mean nucleation distance and on the variation of the velocity of bubble expansion .
[ [ topological - defects - and - magnetic - fields . ] ] topological defects and magnetic fields .
if a global @xmath21 symmetry is spontaneously broken at a first - order phase transition , the phase angle @xmath22 of the higgs field within each nucleated bubble is essentially constant , but phases in different bubbles are uncorrelated . when bubbles collide , the discontinuity in the higgs phase is smoothed out to become a continuous variation .
the so called `` geodesic rule '' states that ( for energetic reasons ) the shortest path between the two phases is chosen @xcite .
when three bubbles meet , a vortex ( in two spatial dimensions ) or a string ( in 3d ) may be trapped between them .
this mechanism is obviously generalized to higher symmetry groups and other kinds of topological defects . ignoring the dynamics of phase equilibration
, it is easy to see that the number density of defects is proportional to the number density of bubbles .
however , the final number of defects will depend strongly on the velocity of bubble expansion @xcite .
if the latter is much less than the velocity of light , then phase equilibration between two bubbles will have probably completed before they encounter a third bubble , thus reducing the chances of trapping a string .
the above picture is in fact a rough simplification of the defect - formation problem .
one complication is due to dissipation , since the higgs field is coupled to the other fields in the thermal bath .
another complication arises when considering a gauge symmetry , and is caused by the fact that the phase of the higgs field is not a gauge - invariant quantity , so it is convenient to define a gauge - invariant phase difference between two bubbles @xcite .
the phase difference is thus linked to the gauge field . in this case
, dissipation can be taken into account by introducing the conductivity of the plasma .
then one can model the collision of three bubbles and calculate the evolution of the phase difference and gauge field .
one can say that a vortex is formed whenever a quantum of magnetic flux is trapped in the unbroken - symmetry region between the three bubbles
. from the above it is clear that the formation of local vortices is associated with the generation of magnetic fields .
therefore bubble collision constitutes also a mechanism for generating the cosmic magnetic fields ( e.g. , see @xcite ) .
of course , the magnetic field that is formed in this way corresponds to a spontaneously - broken symmetry which can not be the electromagnetic @xmath23 .
nevertheless , this mechanism can take place at the electroweak phase transition , where unstable cosmic strings and hypermagnetic fields may be formed .
the latter are subsequently converted to @xmath23 magnetic fields .
it is interesting to note that the presence of magnetic fields may affect the dynamics of the electroweak phase transition ( see e.g. @xcite ) . calculating the magnitude of the magnetic fields and the density of defects that are left at the end of
the phase transition involves the passage from three - bubble collision simulations to the computation of the full phase transition .
evidently , this is a difficult task .
although some simulations have been made ( e.g. , @xcite ) , several simplifications are generally required , which include forgetting about variations in the nucleation rate and the velocity of expansion of bubbles during the transition .
an analytical investigation of phase transition dynamics may therefore clarify the picture and provide useful tools for the calculation of defect formation and magnetic field generation .
we begin by considering the limiting case in which the first - order phase transition is as slowest as possible , namely , that of coexistence of the high- and low - temperature phases at the critical temperature @xmath9 .
such a cosmic separation of phases has been studied for the qcd phase transition @xcite . at the critical temperature
there is no pressure difference between phases at the bubble walls , so the bubble expansion takes place almost in equilibrium .
assume at @xmath24 there are already regions with low - temperature phase . as the universe expands , the fraction of space occupied by these regions increases , as the high - temperature phase converts to low - temperature phase .
the loss of energy due to the expansion of the universe is thus compensated by the latent heat released at the interfaces , and the temperature remains constant . in this scenario
there is no supercooling .
since at @xmath9 the nucleation rate vanishes , such a first - order phase transition is only possible in the presence of impurities that induce the formation of bubbles . in this case inhomogeneous nucleation theory applies . in a phase transition mediated by impurities
there will still be some supercooling , which we neglect in this section for simplicity .
the role of impurities in the early universe could be played for instance by topological @xcite or non - topological solitons @xcite .
these may exist in the high - temperature phase , containing the low - temperature phase in their core . in this case
their configurations become unstable or metastable below the critical temperature .
when the system cools below @xmath9 these objects begin to expand and convert the universe into the true vacuum .
another example of inhomogeneous nucleation is the case of the qcd phase transition in the presence of baryon number inhomogeneities @xcite .
these may arise as a natural consequence of electroweak baryogenesis @xcite , and can survive until the qcd scale @xcite . since the critical temperature is different in regions with different chemical potential @xcite , bubbles will nucleate first in those regions with a higher @xmath9 @xcite .
if such regions are relatively small , and if they achieve the necessary amount of supercooling while the surrounding background reaches the critical temperature , then the baryon inhomogeneities operate as impurities for inhomogeneous nucleation . even if the phase transition proceeds by homogeneous nucleation of bubbles , phase equilibrium will describe quite well a good part of the transition , whenever the latent heat is at least comparable to the energy density difference between the critical temperature and that at which nucleation effectively begins . in that case , the energy released will reheat the plasma back to a temperature very close to @xmath9 . at that moment bubble nucleation
virtually stops and the two phases remain close to equilibrium until the full latent heat of the transition is eliminated
. we will analyze this case in the next section .
the customary equation for the adiabatic expansion , @xmath25 , tells us how the universe takes energy from the hot plasma . here
@xmath26 is the energy density , @xmath27 is the pressure , and @xmath28 is the expansion rate
. it can be equivalently written in terms of entropy density , @xmath29 , which is just the statement of entropy conservation , @xmath30 .
it will be convenient to use it in the form @xmath31 . at the beginning of the phase transition
the whole universe is in the symmetric phase , while at the end of the transition it is filled with the broken symmetry one , so we can write @xmath32 where @xmath33 is the entropy density of the unbroken ( broken ) symmetry phase , @xmath34 is the cosmic scale factor , and @xmath35 its value at the beginning ( end ) of the transition .
we assume that , since the phase transition occurs very slowly , the latent heat released at the interfaces is quickly distributed throughout space and the temperature is homogeneous .
we also assume that pressure and temperature remain constant during the phase transition .
phase coexistence at @xmath14 means that there are regions of space with different equations of state .
thus the entropy density has different constant values @xmath36 and @xmath37 inside and outside the bubbles of broken symmetry phase respectively . the quantity @xmath38 in eq .
( [ entropy1 ] ) is the average entropy density of the whole system .
the entropy in a comoving volume @xmath39 is the sum of two contributions , @xmath40 .
thus , @xmath41 where @xmath42 is the fraction of space that is already in the broken - symmetry phase .
the entropy , energy and pressure are derived from the free energy . at @xmath14 the pressure @xmath43 is the same in the two phases , @xmath44 .
the latent heat of the phase transition is @xmath45 from eqs .
( [ entropy1 ] ) and ( [ entropy2 ] ) , and using ( [ latent ] ) , we obtain the dependence of @xmath42 on the scale factor @xmath46 the dependence of the scale factor on time is given by the friedman equation @xmath47 where for simplicity we have neglected the term @xmath48 . at constant temperature @xmath49 and pressure @xmath50 ,
the energy density is given by @xmath51 where we have used eq .
( [ entropy1 ] ) in the last equality .
inserting @xmath52 in eq .
( [ friedman ] ) and writing @xmath53 we can easily solve the equation for @xmath54 , @xmath55 , \label{scale}\ ] ] where @xmath56 and the constant phase is determined by the initial condition @xmath57 , @xmath58 . during the phase transition we have two coexisting phases in the radiation dominated era , and we may consider different possibilities for the equations of state .
the simplest one is to assume that the universe is radiation - dominated before the phase transition , i.e. , @xmath59 with @xmath60 where @xmath61 is the number of effectively massless species .
depends on @xmath20 .
we will assume for simplicity that @xmath62 is constant during the phase transition .
we discuss the effect of a variation of @xmath63 in section [ physquant ] .
] in fact , this is not a realistic situation . in the symmetric phase
the higgs vev does not correspond to the true vacuum , so we should add a constant energy density to account for the energy of this state and have a negligible cosmological constant after the phase transition .
it is interesting , however , to consider first this simpler case .
hence , at @xmath14 eqs .
( [ purad ] ) and ( [ rourad ] ) give @xmath64 and @xmath65 , so @xmath66 , @xmath67 , and @xmath68 . before the phase transition @xmath69 , with @xmath70
, so eq . ( [ friedman ] ) gives the familiar relation @xmath71 .
the temperature descends like @xmath72 , and at @xmath14 the phase transition begins , since we are assuming that no supercooling occurs .
hence we can use the relation @xmath73 as an initial condition .
the scale factor thus takes the very simple form @xmath74 which does not depend on any parameter , i.e. , during the transition the expansion of the universe seems to be affected always in the same way , regardless the thermodynamical parameters of the theory .
however , thermodynamics affects the duration of the phase transition and the subsequent evolution of the universe , as we shall see immediately . in the above equations it is apparent that the dynamics of the phase transition depends on thermodynamics only through the ratio @xmath75 the fraction of volume is
thus @xmath76 $ ] , with @xmath77 given by eq .
( [ scalelneg ] ) .
the phase transition concludes when @xmath78 [ equivalently @xcite , when @xmath38 in eq .
( [ entropy1 ] ) equals @xmath79 , so its duration can be determined easily . making use of some trigonometric algebra , @xmath80 notice that eq .
( [ duracionneg ] ) fails to give an answer for @xmath81 .
the problem is that the scale factor given by eq .
( [ scalelneg ] ) reaches a maximum when the argument of the sinus is @xmath82 .
if @xmath83 is small enough , that never occurs between @xmath84 and @xmath85 .
for @xmath86 it occurs at @xmath87 , and for larger @xmath83 it happens before @xmath85 , which means that the universe begins to collapse before the phase transition has completed .
this is not surprising . indeed , if the energy density of the unbroken phase is given by eq .
( [ rourad ] ) , then the energy density of the broken symmetry phase is @xmath88 .
so , there is a negative cosmological constant , which will begin to dominate sooner or later . if @xmath89 ( i.e. , @xmath90 ) , this wo nt happen during the phase transition , but below the critical temperature the universe will collapse . in fig .
[ figequilneg ] we have plotted the expansion rate of the universe from the beginning of the phase transition to the moment at which @xmath28 becomes zero .
the evolution of @xmath28 without a phase transition is represented with a dashed line ( long dashes ) .
we have chosen a relatively large value @xmath91 in order to get a visible departure during the phase transition , which occurs between @xmath84 and @xmath85 ( solid line ) .
after the phase transition ( short dashes ) the expansion slows down , and the rate eventually vanishes at a time @xmath92 . [ ] [ ] @xmath92 [ ] [ r]@xmath93 to consider a more realistic situation we must add a constant term @xmath94 to the initial energy density , so that we do not have a cosmological constant of the order of the scale of the phase transition .
we take for simplicity @xmath95 it will be more convenient to re - express the general solution ( [ scale ] ) in terms of the conditions at @xmath96 , @xmath97 $ ] , with @xmath98 . from eq ( [ roulpos ] ) it follows that @xmath99 , @xmath100 , @xmath101 , and @xmath102 .
therefore we can write @xmath103 \label{scalepos}\ ] ] with @xmath104 where we have defined the time scale @xmath105 . before the phase transition , @xmath26 is given by eq .
( [ roulpos ] ) , so @xmath106 .
this has the form @xmath107 for @xmath108 , and departs from the radiation - domination behavior unless @xmath109 .
the fraction of volume in the broken symmetry phase is @xmath110 , \label{fbeqpos}\ ] ] and the duration of the phase transition is given by @xmath111 it can be easily checked that , for small @xmath83 , this solution coincides with the previous one .
furthermore , for @xmath112 , the duration of the phase transition vanishes , as expected . at first sight
, there seems to be a problem if @xmath113 ( i.e. , for @xmath114 ) .
however , all the previous expressions are still valid and real in the range @xmath115 .
they can be written in the form @xmath116 , \ ] ] with @xmath117 and @xmath118 for @xmath119 the duration of the phase transition becomes infinite because the constant energy density @xmath120 is comparable to the energy density of the radiation , playing the role of a cosmological constant that starts dominating at @xmath121 .
therefore the expansion of the universe becomes too fast and the phase transition never ends .
one expects that some of our initial assumptions will break down near this limit .
for instance , the temperature and expansion rate will not be homogeneous , due to the significant energy density contrast between the two phases and the rapid expansion of the universe . in fig .
[ figequilpos ] we plot the expansion rate as a function of time for @xmath122 . we see that without the phase transition the universe would enter exponential expansion ( @xmath123 constant ) . after the phase transition the evolution returns to the radiation - dominated relation @xmath124 . by the end of the transition
the departure of the expansion rate from its previous evolution becomes appreciable , because in this case @xmath125 .
if the duration of the phase transition is short in comparison with the age of the universe , the back - reaction on @xmath28 can be disregarded . according to eq .
( [ duracionposrpeq ] ) , this happens when the energy released is small in comparison to the energy density of the plasma ( small @xmath83 ) .
in the case of a phase transition mediated by homogeneous nucleation , bubbles start to nucleate at a temperature @xmath126 , when the gain in free energy inside a bubble is enough to compensate the cost of gradient energy at the surface .
we have seen in the previous section that , even if bubbles begin to grow at @xmath14 , the phase transition may not come to an end if the parameter @xmath83 is close to @xmath127 .
the case of homogeneous nucleation is even worse due to the additional supercooling .
as we shall see , a large latent heat is a general feature of strongly first - order phase transitions . in this case
there may be extreme supercooling from which the universe may never recover @xcite .
considerable supercooling and latent heat release may occur for instance in the quark - hadron phase transition @xcite .
notice that in the case of large supercooling there will be an important departure from equilibrium , at least at the beginning of the phase transition .
for the rest of the paper we will be mostly interested in the case @xmath128 .
the nucleation and growth of bubbles in a first order phase transition has been extensively studied in the context of the qcd @xcite and electroweak @xcite phase transitions .
after a bubble is formed , it grows due to the pressure difference at its surface .
there is a very short acceleration stage until the wall reaches a terminal velocity due to the friction of the plasma .
it can be seen that this initial period in the history of the bubble expansion is negligible .
we will assume again that the system remains close to equilibrium , in accordance with the assumption that the velocity of the bubble wall is small . if the wall velocity is less than the speed of sound in the relativistic plasma , @xmath129 , the wall propagates as a deflagration front .
this means that a shock front precedes the wall , with a velocity @xmath130 . for @xmath131 ,
the latent heat is transmitted away from the wall and quickly distributed throughout space .
we can take into account this effect by considering a homogeneous reheating of the plasma during the expansion of bubbles @xcite .
( for detailed treatments of hydrodynamics at different wall velocities see , e.g. , @xcite ) . the thermal tunnelling probability for bubble nucleation per unit volume and time is @xcite @xmath132 the prefactor involves a ratio of determinants associated with the quantum fluctuations around the instanton .
in general it must be evaluated numerically .
it is usually assumed to be roughly of order @xmath133 , since the nucleation rate is dominated by the exponential in ( [ gamma ] ) .
we will consequently assume @xmath134 .
@xmath135 is the three - dimensional instanton action , which coincides with the free energy of a critical bubble ( i.e. , a bubble in unstable equilibrium between expansion and contraction ) , @xmath136 .
\label{s3}\ ] ] the configuration of the nucleated bubble may be obtained by extremizing this action .
hence it obeys the equation @xmath137 for temperatures very close to @xmath9 , the width of the bubble wall at the moment of formation is much smaller than its radius , and a thin wall approximation can be used @xcite , in which @xmath138 is expressed as a function of the critical bubble radius @xmath139 , the free energy difference @xmath140 between the two minima of the potential , and the bubble wall surface energy @xmath141 .
the radius @xmath139 can thus be obtained by finding the maximum of @xmath138 .
a similar approximation can be used to estimate the free energy and radius of a thick - walled bubble when temperature is not so close to @xmath9 @xcite . however , as pointed out in ref .
@xcite , due to the exponential dependence , the tunnelling probability may be strongly overestimated by using approximations to @xmath138 , leading to a sooner completion of the phase transition , so it is a good approximation to replace almost every quantity by its value at @xmath142 .
the important exception are quantities such as @xmath138 , that depend directly on the free energy difference @xmath5 , which varies drastically with @xmath20 at the critical temperature @xcite . ] .
since we do not intend to do numerical calculations in the present work , we will use the thin wall approximation .
this approximation may be reasonable or not , depending on the amount of supercooling . in the previous section we assumed constant temperature , and used eqs .
( [ entropy1 ] ) and ( [ entropy2 ] ) to obtain the fraction of volume @xmath42 in terms of the scale factor @xmath34 .
then the friedman equation determined @xmath143 . in the present case
the temperature is not constant , so we need an extra equation to solve for the three quantities @xmath144 , @xmath34 and @xmath20 .
such an equation arises by considering the nucleation and growth of bubbles @xcite , @xmath145
^{3}\gamma \left ( t^{\prime } \right ) r\left ( t^{\prime } , t\right ) ^{3}dt^{\prime } \right\ } , \label{fb}\ ] ] where @xmath146 is the temperature at @xmath147 , @xmath148 is the time at which the universe reaches the critical temperature , @xmath149 where @xmath150 , and @xmath151 is the radius of a bubble that nucleated at time @xmath152 and expanded until the time @xmath153 , @xmath154 the factors of @xmath34 in eqs .
( [ fb ] ) and ( [ radius ] ) take into account the fact that the number density of nucleated bubbles is diluted , and the radius of a bubble enlarged , due to the expansion of the universe from @xmath152 to @xmath153 ( see e.g. @xcite ) .
we can assume that this effect is negligible if the duration of the phase transition is small in comparison with the age of the universe .
as we have seen in the previous section , this is true when @xmath155 , which is the case we will consider .
( this assumption will not hold , in principle , for the qcd phase transition ) .
the wall velocity @xmath19 is determined by the equilibrium between the pressure difference @xmath156 and the friction force exerted by the plasma .
the latter is proportional to the wall velocity .
the constant of proportionality is the friction coefficient @xmath157 ( see section [ physquant ] ) , so @xmath158 the exponent in eq .
( [ fb ] ) is minus the fraction of volume occupied by bubbles that nucleated between @xmath148 and @xmath153 , if we do not take into account overlapping of bubbles . at the beginning of nucleation
the formula @xmath159 is correct . using again eqs .
( [ entropy1 ] ) and ( [ entropy2 ] ) , we obtain the analogous of eq .
( [ fractiontc ] ) , @xmath160 but since we already have an equation for @xmath42 , namely , eq . ( [ fb ] ) , we use eq .
( [ fractionsuperc ] ) to express @xmath20 in terms of @xmath42 and @xmath34 , @xmath161 eq . ( [ temperature ] ) has come across within an approach that differs from previous works , so it is worthwhile spending a few words discussing its physical meaning .
we may follow for instance ref .
@xcite , and use energy ( non- ) conservation in the following way . on one hand , we may write the total energy in a volume @xmath162 as @xmath163 v_{u}=\rho v_{u}$ ] , where @xmath164 .
if the universe were not expanding , energy conservation during the phase transition would give @xmath165 this gives the rate @xmath166 at which the plasma takes energy from the change of phase .
on the other hand , when it is not undergoing a phase transition , the universe takes energy from the plasma at a rate @xmath167 if we join the two equations , we obtain the total rate of change of energy as @xmath168 from where we get an equation for @xmath169 . however , if the phase transition and the expansion of the universe are taken into account at the same time , additional terms appear both in eqs .
( [ heck1 ] ) and ( [ heck2 ] ) . on one hand , there is a term of the form @xmath170 in @xmath171 , which produces a new term @xmath172 . this just accounts for energy dilution .
on the other hand , the expansion of the universe takes energy not only from radiation , since it is not the only component in the equation of state .
bearing in mind the two coexisting phases during the phase transition , the rate at which energy conservation is violated is given by @xmath173 .
using this to obtain @xmath174 , we finally find @xmath175 where @xmath176 .
since @xmath177 , it can be seen that the first terms of this equation reproduce eq .
( [ heckler ] ) , but there are additional terms . using @xmath178 and rearranging eq .
( [ noheckler ] ) , we see that it is just the equation for entropy conservation , @xmath179 from which we may re - obtain the result ( [ temperature ] ) .
the discrepancies between eq .
( [ heckler ] ) and eq .
( [ noheckler ] ) will not be important as long as the latent heat @xmath120 is not significant and @xmath20 remains close to @xmath9 .
indeed , we can neglect the last two terms inside the parenthesis in eq .
( [ noheckler ] ) , provided that @xmath180 .
the last term in eq .
( [ noheckler ] ) is responsible for the appearance of the entropy difference @xmath181 as the factor of @xmath182 in eq .
( [ entropycons ] ) , instead of the energy difference @xmath183 .
since @xmath184 , the two quantities are related by @xmath185 for @xmath142 .
therefore , eq . ( [ heckler ] ) gives a good approximation in the case @xmath128 .
still , the fact that in eq .
( [ temperature ] ) the temperature is already integrated may constitute an advantage for analytical calculations .
finally , the evolution of the scale factor is given by the friedman equation , @xmath186 .
\label{friedmansuperc}\ ] ] if @xmath180 , then one can use the customary formula @xmath187 for the expansion rate .
in fact , the variation of @xmath188 is of the same order of @xmath189 , so if we neglect @xmath189 , to be consistent we should also set @xmath190 in this approximation .
again , this is reasonable if the duration of the phase transition is short enough . in each particular model ,
( [ fb ] ) , ( [ temperature ] ) and ( [ friedmansuperc ] ) can be used to calculate numerically the evolution of the phase transition . in this paper , dough
, we will make analytical estimates of the parameters that are relevant for the cosmological consequences , such as the temperature and wall velocity in the different stages of the transition .
when bubbles occupy more than 30% of space , they meet and percolate .
this occurs when bubbles have a characteristic radius @xmath191 , where @xmath192 is the number density of bubbles .
bubble coalescence provides a different mechanism of bubble growth , in which the driving force is surface tension instead of pressure difference .
when bubbles collide and percolate , they arrange themselves into a system of fewer , larger bubbles in order to minimize the surface area .
we can estimate the characteristic time of this process as follows @xcite .
assume that when two spherical bubbles of radius @xmath193 meet , they form a single bubble of radius @xmath194 . in this process , a mass of fluid @xmath195 is moved a distance of order @xmath193 .
if this is done in a time @xmath153 , the kinetic energy involved in the process is @xmath196 .
this energy is supplied by the surface of the bubbles .
the surface energy released in the process is @xmath197 , so this occurs in a time @xmath198 .
once bubble coalescence begins , it will be the fastest mechanism of bubble growth if the rate @xmath199 is larger than the rate @xmath200 given by the wall velocity ( [ vw ] ) .
therefore this process will dominate until bubbles reach a characteristic size @xmath201 . if @xmath202 , then it never dominates .
otherwise , during the period in which the radius varies from @xmath203 to @xmath204 bubble coalescence is the fastest process and must be taken into account .
the process of bubble coalescence may end in two different ways .
if @xmath19 is large enough , the radius @xmath204 is very close to @xmath203 , and after a short period bubbles continue to grow with velocity @xmath19 .
if , on the contrary , @xmath19 is very small , then coalescence dominates for a larger period , until the low - temperature phase occupies more than 50% of the total volume . at this moment
the regions of high - temperature phase detach into isolated bubbles and the process stops .
the interface velocity is again determined by pressure difference , friction , and latent heat release , so these bubbles shrink with velocity @xmath19 .
this occurs at a bubble radius @xmath205 .
notice that , although this expression is similar to that for @xmath203 , in fact @xmath206 may be very different from @xmath203 since @xmath192 may decrease significantly during the process due to the expansion of the universe . finally , when bubbles of the high temperature phase are so small that surface energy becomes dominant over volume energy , the shrinking is accelerated until the symmetric - phase bubbles disappear or , eventually , until a topological defect is formed .
in the case of homogeneous nucleation , the universe cools down to a temperature @xmath207 before bubble nucleation becomes appreciable , so the phase transition effectively starts at a time @xmath208 .
then , the phase transition proceeds in two main steps . at the beginning bubbles nucleate with a rate @xmath209 , and expand with a velocity given by the friction coefficient @xmath157 and the pressure difference @xmath210 . after a short time
the energy released by the change of phase reheats the universe up to a temperature @xmath211 close to @xmath9 .
then , a longer period begins , in which the two phases are close to equilibrium .
how long is this period , and how close is @xmath212 to @xmath49 , depends on the latent heat .
the free energy difference in this stage is @xmath213 , so the velocity of bubble expansion decreases and the bubble nucleation rate becomes extremely suppressed .
therefore , this second stage may be very similar to the inhomogeneous nucleation phase transition discussed in section [ phaseeq ] . in order to avoid numerical calculations
, we will need some approximations for the fraction of volume @xmath42 , the temperature @xmath20 , and the expansion rate of the universe @xmath28 [ formulas ( [ fb ] ) , ( [ temperature ] ) and ( [ friedmansuperc ] ) ] . we notice that the latter can not be affected by the phase transition until the energy released becomes appreciable .
furthermore , when this occurs , the plasma has reheated up to @xmath214 , and enters the phase equilibrium stage .
we have seen in section [ phaseeq ] that @xmath28 is not modified significantly if the parameter @xmath83 defined in ( [ r ] ) is small .
so , we do not need to consider back - reaction on the scale factor and we will assume that the evolution of the universe does not depart from the standard relations @xmath215 , @xmath216 .
we will also need an approximation for the nucleation rate and free energy difference .
if the width of the wall is much less than the radius of the bubble , we can neglect the second term in eq .
( [ eqprofile ] ) to obtain the wall profile ( this approximation is exact at the critical temperature , at which @xmath218 , and gives the usual kink profile ) . if we then multiply by @xmath219 and integrate using the boundary conditions @xmath220 and @xmath221 outside the bubble , we find that @xmath222 so , @xmath223 since at the wall @xmath6 falls from @xmath17 to @xmath93 . inserting this in eq .
( [ s3 ] ) we obtain the free energy of the critical bubble in the thin wall approximation , @xmath224 where the free energy difference @xmath156 is defined in eq .
( [ deltav ] ) , and @xmath225 is the surface tension of the bubble wall , @xmath226 maximizing with respect to @xmath139 we get the values of the critical radius and action , @xmath227 since @xmath141 does not change significantly during the phase transition , it can be approximated by @xmath228 . since the thin wall approximation is valid when @xmath20 does not depart significantly from @xmath9 , we can also make a linear approximation for the free energy difference , @xmath229 so , the exponent in the nucleation rate ( [ gamma ] ) becomes @xmath230 and the critical radius is @xmath231 with the linear approximation ( [ linearv ] ) , the wall velocity is given by @xmath232 neglecting the effects of latent heat , the phase transition may be assumed to occur roughly at @xmath233 , since it goes on for a very short interval @xmath234 @xcite . with the inclusion of latent heat
the phase transition evolves during a longer period , that begins at @xmath235 . in this case
, we expect that the first part of the evolution , in which the universe is reheated up to a temperature @xmath211 , has a time scale of the same order of the time interval @xmath236 of the phase transition without latent heat .
this is confirmed by numerical calculations @xcite .
the nucleation rate @xmath237 vanishes at @xmath14 , but it changes very quickly with temperature and becomes of order @xmath133 at @xmath238 , where the barrier between the two minima of the free energy disappears .
this is an extremely large rate , so it is impossible that the universe supercools close to @xmath12 @xcite .
we will assume that the temperature @xmath239 is close enough to @xmath49 that the approximations of the previous subsection can be used . in ref .
@xcite the onset of nucleation was assumed to occur when the probability that a bubble was nucleated inside each causal volume is one , @xmath240 the causal volume is given by @xmath241 , where the horizon size @xmath242 scales like the age of the universe , @xmath243 .
the cosmological scale @xmath153 , however , is in general too large in comparison with the scale of phase transition dynamics , which at @xmath244 is given by @xmath245 .
the scale of phase transition dynamics is roughly determined by the temperature variation during the phase transition , which is bounded by the difference @xmath246 .
consequently , it may be more appropriate to consider a different causal distance @xmath247 , related to the dynamics of the phase transition in the following way .
we may say that bubbles begin to see each other at a time @xmath235 when their mean separation is of the order of the distance travelled by a sound wave since time @xmath148 .
then , the causal distance is given by @xmath248 , where @xmath249 is the velocity of sound in the relativistic fluid .
this defines a causal volume in terms of @xmath250 @xmath251 ^{3 } , \label{causal}\ ] ] we remark that the real improvement in using eq .
( [ causal ] ) instead of @xmath252 does not come from considering the velocity of sound , which is @xmath253 , but from the fact that in many cases the time elapsed from @xmath148 to @xmath254 will be much less than the age of the universe @xmath255 .
the volume @xmath256 is thus suppressed with respect to @xmath252 by a factor @xmath257 ^{3}$ ] .
the nucleation time @xmath235 calculated in this way is larger , since more bubbles need to be nucleated before they are separated by a distance at which they are causally connected to each other .
there will be a bubble in each volume @xmath256 when @xmath258 to evaluate the integral in ( [ eqonset ] ) we make use of the following approximation @xcite ( see also @xcite ) .
the three - dimensional action in eq .
( [ s3thin ] ) can be expanded about any temperature @xmath259 in the form @xmath260 where @xmath261 . since the integrand in eq .
( [ eqonset ] ) is sharply peaked at @xmath250 , we choose @xmath262 and use the expansion ( [ s3expanded ] ) to evaluate the integral .
we find @xmath263 which will be in general dominated by the first term . for temperatures @xmath20 several orders of magnitude below @xmath264 , we have @xmath265 ( it is e.g. @xmath266 for the electroweak scale , and @xmath267 for the qcd scale ) . from eq .
( [ s3thin ] ) we obtain @xmath268 where we have used that @xmath269 . in some of the following estimations we will replace @xmath270 by @xmath246 inside the @xmath271 s , since both differences are generally of the same order . immediately after @xmath233
, the temperature increases at a rate which is given by eq .
( [ noheckler ] ) . under the current approximations
we can write @xmath272 . therefore , the rate of change of energy of the plasma is @xmath273 .
thus , on one hand energy is increased at a rate @xmath274 , where @xmath275 is the fraction of volume converted to the low - t phase during the reheating stage , and @xmath276 is the duration of this stage .
the rate of energy decrease , on the other hand , is given by @xmath277 .
since @xmath276 is much shorter than the age of the universe , the increasing rate is much larger than the latter .
the total change in energy density is thus @xmath278 .
the temperature can not increase beyond @xmath9 , so if @xmath279 , the fraction @xmath275 will be less than one .
this means that before the phase transition completes , a final temperature @xmath211 very close to @xmath9 is reached , and the phase transition proceeds more slowly until @xmath280 .
for @xmath281 we have a variation @xmath282 during the phase transition , which gives us an idea of the reheating that occurs . there may be significant reheating and a considerable variation of the wall velocity , but there will not be a long phase equilibrium stage , since in this case @xmath283 .
the case @xmath284 corresponds to weakly first - order phase transitions , which most likely occur by spinodal decomposition rather than by nucleation and expansion of bubbles @xcite .
we could also have @xmath285 , although we are assuming @xmath286 . in this case
we have @xmath287 , which means that most of the phase transition happens with the two phases near equilibrium , and we can apply the analysis of section [ phaseeq ] . in this section we thus concentrate in the case in which the latent heat is comparable with the energy difference @xmath288 .
the expansion of bubbles is governed by eq .
( [ radius ] ) . at @xmath289 bubbles
nucleate with a radius @xmath139 given by eq .
( [ rc ] ) , and after a time @xmath236 the radius has increased an amount @xmath290 , with @xmath291 given by eq .
( [ vwlinear ] ) .
the ratio of the two distances is @xmath292 for the time scale of the phase transition dynamics , @xmath293 , we have @xmath294 thus , if the phase transition takes place at a temperature sufficiently below the plank scale , the bubbles will grow so rapidly that we can safely neglect the initial radius @xmath139 in eq .
( [ radius ] ) . at the beginning , bubbles expand with constant velocity @xmath295 .
when reheating becomes important , the bubble expansion slows down and we enter the second stage .
consequently , during the first stage @xmath296 can be roughly estimated without taking into account the liberation of latent heat .
we can thus estimate the rate @xmath297 without considering the back - reaction of the reheating on @xmath182 .
it is only at the beginning and at the end of reheating that @xmath298 and @xmath299 .
indeed , soon after @xmath233 , the temperature takes its minimum value @xmath300 ( see fig . [ esquema ] ) , then it increases to @xmath301 .
the value of @xmath302 is important because the nucleation rate turns on at @xmath303 , is maximal at @xmath304 , and turns off again when the temperature has increased back to @xmath305 .
all the process occurs in a time @xmath306 , which is determined by the speed at which temperature changes .
this interval @xmath307 in which @xmath237 is not negligible is much less than @xmath308 , due to the exponential dependence of @xmath237 on @xmath309 .
[ ] [ b ] @xmath84 [ ] [ b ] @xmath85 [ ] [ b ] @xmath310 [ ] [ l]@xmath49 [ ] [ l]@xmath302 [ ] [ b]@xmath308 [ ] [ b]@xmath311 = 5.5 cm accordingly , the temperature @xmath312 occurs when @xmath313 where the time - temperature relation @xmath314 must be used to evaluate the integral .
using the expansion ( [ s3expanded ] ) about @xmath315 , we find @xmath316 ^{-1}e^{-\frac { s_{3}\left ( t_{m}\right ) } { t_{m}}}\simeq \frac{2\pi \sigma ^{6}\eta ^{3}t_{c}^{2}}{9\xi ^{6}l^{8}}\left ( \frac{t_{c}}{m_{p}}\right ) ^{4}\left ( \frac{t_{c}}{t_{c}-t_{m}}\right ) ^{10}. \label{sm}\ ] ] the number density of bubbles is given by @xmath317 since @xmath318 is sharply peaked at @xmath319 , we can estimate @xmath320 as @xmath321 and use again the dependence @xmath322 .
the integral can be calculated again using the expansion ( [ s3expanded ] ) , which gives @xmath323 ^{-1}e^{-s_{3}\left ( t_{m}\right ) /t_{m}}. \label{nb1}\ ] ] we may define the interval @xmath324 by writing @xmath325 .
then we can easily see in eq .
( [ nb1 ] ) that @xmath326 ^{-1}\left ( t_{m}-t_{i}\right ) .\ ] ] therefore , @xmath327 is much less than @xmath328 ( most typically , about two orders of magnitude in this way , rather than subtracting the values given by eqs .
( [ ti ] ) and ( [ sm ] ) . ] ) .
using the result ( [ sm ] ) , with @xmath329 , we finally obtain the density of bubbles , @xmath330 the bubbles thus nucleate during the short time @xmath331 about @xmath332 , and expand with a velocity @xmath333 for a time @xmath334 , until the temperature gets close to @xmath49 . according to eqs .
( [ vwlinear ] ) and ( [ ti ] ) , @xmath335 where @xmath336 is a shorthand for the sum of @xmath271s in ( [ ti ] ) . notice that in eq .
( [ vwlinear ] ) , @xmath337 , but also @xmath338 , so @xmath339 only depends on @xmath120 through the logs in @xmath336 .
if a non - negligible fraction of the volume is taken up by bubbles during the first stage , the interval @xmath276 can be estimated from @xmath340 .
this gives @xmath341 so , apart from a model - dependent factor , we find a general tendency to the relations @xmath342 .
the value of @xmath343 ( and that of @xmath344 ) can be obtained similarly to the case of @xmath250 .
it is interesting to note that @xmath276 has only a logarithmic dependence on the friction coefficient @xmath157 .
this is because the dependence on the wall velocity is twofold . on one hand ,
the lower the wall velocity , the longer the time @xmath345 needed to reheat the plasma .
but on the other hand , the lower the wall velocity , the longer will also be the time @xmath307 in which bubbles are formed , and the larger their number .
this causes a shorter @xmath276 , since there are more bubbles to produce the reheating .
if the formation of bubbles is associated with the presence of impurities , the phase transition occurs at @xmath142 , and the number density of bubbles @xmath346 is an external parameter that depends on the density of impurities . according to eqs .
( [ scalepos])-([fbeqpos ] ) , for small @xmath83 the evolution of the phase transition is given by @xmath347 and the rate at which the phase transition goes on is @xmath348 , i.e. , a factor of @xmath349 larger than the rate of expansion @xmath350 of a comoving volume .
our assumption @xmath128 implies that @xmath351 and @xmath352 . to calculate the velocity of the interfaces
, we assume that all the bubbles begin to expand at @xmath353 .
thus , the fraction of volume occupied by bubbles is @xmath354 , where @xmath355 is the bubble radius . at the midpoint of the transition we have @xmath356 , where @xmath357 is the average radius .
therefore the mean velocity is given by @xmath358 .
we notice , however , that even in this case in which @xmath182 is constant , the wall velocity may change significantly during the transition , since @xmath359 .
the total variation of @xmath339 depends on @xmath320 . in the case of a phase transition with supercooling ,
the situation is very similar after the plasma has reheated up to a temperature @xmath214 .
the transition proceeds at a rate @xmath360 in this stage the nucleation rate has turned off .
we have seen that @xmath237 peaks sharply at a certain time @xmath319 in the previous stage , so we can write eq .
( [ fdot2stage ] ) as @xmath361 , where @xmath362 , @xmath192 is given by eq .
( [ nb2 ] ) , and @xmath363 . in any case , the integral in ( [ fdot2stage ] ) is an average of the squared radius of the bubbles , and for the present estimations we can set @xmath364 , with @xmath365 given by @xmath192 , just as in the case of inhomogeneous nucleation .
since the temperature is almost constant , @xmath366 , i.e. , all the released latent heat is taken away by the expansion of the universe .
thus , again @xmath348 and the velocity coincides with that of the inhomogeneous nucleation case @xmath367 although in eq .
( [ vfinal ] ) it seems that the wall velocity during phase equilibrium does not depend on the friction , in fact it is proportional to @xmath368 due to the dependence of @xmath320 .
since @xmath369 , the duration of this stage is @xmath370 the reheating temperature @xmath211 must be such that the pressure difference is adjusted so as to give the velocity ( [ vfinal ] ) .
using eq .
( [ vwlinear ] ) , we find @xmath371 as expected , the larger the latent heat , the closer will be @xmath211 to @xmath9 .
unlike @xmath311 , the values of @xmath212 and @xmath372 are not easy to determine , since they depend on the number of bubbles that nucleated in the previous stage . using eq .
( [ nb2 ] ) and taking into account that @xmath373 , we find that roughly @xmath374 which confirms that generally , @xmath375 . according to eq .
( [ vwlinear ] ) , the same relation holds for the velocities @xmath376 and @xmath372 . in the range @xmath377 bubble percolation
takes place .
we have seen that this process gives a contribution to the bubble expansion rate , of order @xmath378 . for @xmath379
this rate is @xmath380 to establish the importance of this rate we should compare it with the rates @xmath381 or @xmath382 , corresponding to the two stages we have studied . such a comparison is difficult to carry out without specifying a model , so we will ignore this effect in the subsequent discussions . although coalescence is bounded to occur in the above range of @xmath383 , it could have important consequences if the associated bubble growth rate is significantly larger than those given by the time scales @xmath276 and @xmath384 . in a specific model , comparison of eqs .
( [ deltat1 ] ) and ( [ deltat2 ] ) with ( [ ratecoal ] ) should not be hard to do , once @xmath192 has been evaluated .
we have seen that all the parameters that describe the dynamics of the phase transition ( i.e. @xmath276 , @xmath384 , etc . ) depend on a few thermodynamical parameters , such as the latent heat or the friction coefficient .
the formation of the different cosmological products of a phase transition thus depends on these quantities , and also on other parameters , such as the conductivity of the plasma .
these quantities are physically related , since all of them come from the equilibrium or non - equilibrium thermodynamics of the same underlying theory .
this should be taken into account in phase transition calculations , when ranges of parameters are considered .
unfortunately , it is hard in practice to establish general relations between these physical parameters .
depending on the theory , it may even be impossible to compute some of these quantities .
one can gain some insight on the relations between thermodynamical quantities by conveniently modelling the free energy .
the problem is further simplified by referring to the general form of the perturbative effective potential . in that case
the thermodynamical quantities can be related to the parameters of the microscopic theory .
we dedicate this section and the following to study the aforementioned relations .
we will concentrate only in those parameters which influence directly the dynamics of the phase transition .
an analysis of other parameters that affect the generation of cosmological remnants is considered in @xcite .
we assume the free energy density takes the form @xmath385 where the scalar field @xmath6 is the order parameter , and @xmath386 is the free energy density difference between the symmetric and the broken - symmetry phases .
the parameter @xmath61 is the number of light species of the plasma . in general
, @xmath61 depends on temperature , but it is usually approximated by @xmath387 where the sums are on particles with masses @xmath388 , and @xmath389 is the number of degrees of freedom of species @xmath390 ( see appendix [ ap1 ] ) .
it depends on each particular case wether eqs .
( [ free ] ) and ( [ veff ] ) can be derived from the microscopic theory . at any rate
, they can be regarded just as a simple model for studying the dynamics of the phase transition , being the latter first - order if the coefficient @xmath391 is nonvanishing .
the parameters of @xmath5 can be chosen in such a way that the free energy carries the thermodynamical properties of the theory we wish to study @xcite .
for instance , these parameters determine the values of the critical temperature , latent heat , surface tension , and correlation length .
the thermodynamical parameters could be obtained , e.g. , with lattice simulations ( see for example @xcite ) .
then one can use those values to calculate the parameters @xmath12 , @xmath392 , @xmath391 and @xmath393 .
furthermore , in general the order parameter @xmath6 is a higgs field or a combination of higgs fields , and @xmath394 is the finite - temperature effective potential ( see e.g. @xcite ) .
we will consider this case in the next section .
the effect of viscosity on the propagation of the bubble wall is calculated by considering its equation of motion in the hot plasma , @xmath395 which can be derived by energy conservation considerations @xcite . here
@xmath396 is the zero temperature effective potential , the sum is over all particles that couple to @xmath6 , @xmath397 are the @xmath6-dependent masses ( see the appendices ) , and @xmath398 are the phase space population densities
. this equation can be obtained by thermally averaging the operator equation for @xmath6 .
if we separate @xmath399 into the equilibrium population @xmath400 plus a small deviation @xmath401 , we obtain the equation @xmath402 where @xmath394 is the finite temperature effective potential , given by eq .
( [ veff ] ) .
since the departure from equilibrium is proportional to the velocity of the bubble wall , it is the last term in ( [ eqfiplasma ] ) which gives the friction force of the plasma . a simple approach to the calculation of the wall velocity @xcite consists in replacing the last term in eq .
( [ eqfiplasma ] ) with a typical damping term of the form @xmath403 .
due to lorentz invariance this term must be in fact of the form @xmath404 , where @xmath405 is the four - velocity of the plasma .
( [ eqfiplasma ] ) then may be written as @xmath406 where @xmath407 is a dimensionless damping coefficient that depends on the viscosity of the medium . boosting to a frame that moves with the wall , and assuming stationary and non - relativistic motion in the @xmath408 direction , we have @xmath409 where @xmath410 . multiplying both sides by @xmath411 and integrating over @xmath412 we obtain @xmath413 where @xmath156 is the free energy difference between the two phases , defined in eq .
( [ deltav ] ) , and @xmath141 is the surface tension of the wall , given by eq .
( [ formulasigma ] ) .
we have assumed here that temperature is constant across the wall .
this is right if the wall velocity is small enough , so that the latent heat it releases has time to be uniformly distributed throughout space .
hence , the pressure difference is equilibrated by a friction force proportional to the wall velocity .
the constant of proportionality is the friction coefficient @xmath414 .
since the tension of the wall is related to the wall width @xmath415 by @xmath416 [ see the first integral in eq .
( [ formulasigma ] ) ] , we can also write @xmath417 .
a shortage of modelling the viscosity of the plasma in this way is that @xmath418 is a free parameter .
the correct expression for @xmath157 can be derived from eq .
( [ eqfiplasma ] ) . in appendix [ ap2 ]
we show that particles with a thermal distribution give a friction coefficient @xmath419 while the contribution of infrared gauge bosons is @xmath420 evidently , both formulas agree with the above result if @xmath421 .
this treatment also allows for the evaluation of the coefficients @xmath407 , which depend only on the particle content of the plasma . the free energy given by eqs .
( [ free ] ) and ( [ veff ] ) bears a first - order phase transition , with two minima separated by a barrier .
the critical temperature is related to @xmath12 by @xmath422 at @xmath423 the global minimum of the potential is @xmath3 . at the critical temperature the two minima become degenerate , and below this temperature
the stable minimum is @xmath424 \ . \label{fimin}\ ] ] at @xmath238 the barrier between minima disappears and @xmath3 becomes a maximum of the potential
. therefore the phase transition occurs at some stage in between @xmath9 and @xmath12 .
the value @xmath3 corresponds to the symmetric , high - temperature phase , and @xmath425 corresponds to the broken - symmetry , low - temperature phase .
the jump of the order parameter from the high temperature phase to the low temperature one is thus @xmath426 according to eqs .
( [ free ] ) and ( [ veff ] ) , the free energy density of the symmetric phase is @xmath427 this gives the equation of state of a hot relativistic plasma with a positive cosmological constant @xmath428 the free energy of the broken - symmetry phase is @xmath429 .
the energy density of the broken - symmetry phase is @xmath430 , with @xmath431 the entropy density of the symmetric phase is @xmath432 , and that of the broken - symmetry phase is @xmath433 .
the latent heat of the phase transition is given by @xmath434 ; hence , @xmath435 comparing with eq .
( [ fimtc ] ) , we find the relation @xmath436 between the discontinuity of the order parameter and that of the energy density .
as expected , strongly first - order phase transitions ( i.e. , with large @xmath10 ) have large latent heat .
the surface tension of the bubble wall is given by eq .
( [ formulasigma ] ) in the thin wall approximation . at the critical temperature
the effective potential is given by @xmath437 where @xmath438 .
hence , eq . ( [ formulasigma ] )
is easily integrated and @xmath439 although we havent used it explicitly , in this approximation the field configuration near the wall can be solved analytically with the help of eq .
( [ vtc ] ) , and gives the kink profile , @xmath440 where @xmath441 is the wall width . may change during the bubble expansion due to the friction with the plasma @xcite .
we shall neglect this effect . ] using eqs .
( [ fimtc ] ) , ( [ sigma ] ) and ( [ width ] ) we find the values of the friction coefficients ( [ etath ] ) and ( [ etair ] ) , @xmath442 the two contributions have different parametrical dependence , so each will dominate in different regions of parameter space .
for instance , if @xmath443 the infrared boson contribution may be much larger than that of thermal particles .
the maximum velocity of bubble walls occurs at @xmath444 . according to eqs .
( [ vwn ] ) and ( [ etas ] ) , this velocity is the smallest among @xmath445 to determine which one is the correct , it is necessary to know the relations between the coefficients @xmath391 , @xmath446 , and @xmath407 .
we see that in the opposite limiting cases @xmath443 and @xmath447 , one of the two velocities is @xmath448 , unless @xmath407 is too small . in the case
@xmath449 , the wall velocity will be small if one of the conditions @xmath450 or @xmath451 is fulfilled , which does not seem unlikely in general ( see the next section ) .
this supports the assumption of non - relativistic wall velocities .
it is evident that with the aid of the model ( [ free],[veff ] ) we can get more information about the generalities of phase transition dynamics .
for instance , if we write eq .
( [ ti ] ) as a function of @xmath391 , @xmath452 , and @xmath393 , and compare with eq .
( [ tct0 ] ) , then we can locate the nucleation temperature in the interval @xmath246 , @xmath453 if @xmath391 and @xmath454 are comparable , this gives a value of @xmath270 roughly an order of magnitude less than @xmath246 .
the relations between the different quantities that determine the dynamics of the phase transition are apparent in the above expressions .
specific relations will be of interest for different cosmological consequences . as an example , let us consider the effect of modifying the theory in order to obtain a more strongly first - order phase transition . to do that
, we have to enlarge the value of the order parameter .
assume we accomplish this by increasing the value of the parameter @xmath391 and keeping the other parameters invariant [ see eq.([fimtc ] ) ] .
then , there will be more supercooling , and one expects a larger departure from thermal equilibrium , since the pressure difference at @xmath289 will be larger .
however , according to eqs .
( [ l ] ) and ( [ etas ] ) , @xmath120 and @xmath157 also increase .
this tends to decrease the wall velocity in the two stages of the phase transition , in opposition to the effect of supercooling . in this work
we assume for simplicity that @xmath61 remains constant throughout the phase transition .
in fact , the number of effectively massless degrees of freedom may change during the phase transition .
it is conceivable that some particles acquire large masses and decouple from the thermal bath ; then @xmath455 .
for instance , during the quark - hadron phase transition @xmath61 changes substantially .
it is interesting to note that such a change may affect considerably the dynamics of the phase transition , even in the case @xmath456 .
the effect of a decrease of @xmath61 during the phase transition is twofold . to begin with ,
the free energy of the broken - symmetry phase is larger than in the case of constant @xmath61 , so the critical temperature is lower ( it is given by @xmath457 @xcite ) .
therefore the phase transition is stronger , and the latent heat @xmath458 is larger .
in addition , the entropy released by the decoupling species gives an extra contribution of @xmath459 to the latent heat .
this contribution is comparable to the value of @xmath120 as given by eq .
( [ l ] ) , if @xmath460 . in the case of
a perturbative effective potential , this condition may be easily fulfilled for @xmath461 .
if perturbation theory is applicable , the one - loop effective potential at high temperature has oftentimes the form of eq .
( [ veff ] ) , with parameters generally given by @xmath462 here , @xmath463 are the couplings of the particles with @xmath6 , @xmath464 is the higgs mass , and @xmath465 its zero temperature vev .
the coefficient @xmath391 in general involves only gauge bosons . in appendix [ ap1 ]
we review the derivation of these results and discuss on the general assumptions and approximations that lead to eqs .
( [ veff ] ) , ( [ g * ] ) , and ( [ potparam ] ) . in the discussions that follow
we will sometimes take the electroweak theory as a reference point .
the parameter @xmath12 gives the temperature scale of the phase transition .
its order of magnitude is determined by @xmath464 , so it may be quite less than the scale @xmath465 if @xmath393 is small .
anyway , for the dynamics of the phase transition , the difference @xmath246 is more important than the temperature scale @xmath466 . regarding the viscosity of the plasma , we show in appendix [ ap2 ] that the contribution of thermal particles to the parameter @xmath407 is given by @xmath467 where @xmath468 for fermions and @xmath469 for bosons .
therefore the contributions of bosons to @xmath157 have an enhancement of @xmath470 with respect to fermions with the same yukawa coupling .
for instance , for @xmath471 the boson enhancement is @xmath472 .
this means that friction may be much stronger in supersymmetric theories than in non - supersymmetric ones .
for instance , it was found in ref .
@xcite that a light stop may slow down the electroweak bubble wall in the mssm an order of magnitude with respect to the sm .
the enhancement is larger for lighter particles , but these do not contribute to the friction due to the @xmath473 dependence .
the contribution of infrared gauge bosons is @xmath474 .
\label{etatildeir}\ ] ] here , the sum is only on gauge bosons , but the coefficient @xmath475 also involves a sum over particle species ( see appendix [ ap2 ] ) .
furthermore , the gauge coupling appears only squared , which means less suppression .
the @xmath476 enhancement in this case is @xmath477 .
it is important to compare the value of the parameter @xmath391 with the other parameters , since @xmath391 is responsible for the first - order nature of the phase transition .
we can see in the formulas of the previous section that all the thermodynamical quantities are proportional to some power of @xmath391 , while the parameters @xmath452 and @xmath393 usually appear in the denominators . in the perturbative approach ( [ potparam ] ) , this parameter is generally smaller than the others .
this is because @xmath391 is a sum of gauge couplings to the third power , weighted with gauge boson degrees of freedom , while @xmath452 involves squared couplings , and the sum is over all degrees of freedom .
regarding @xmath393 , it can be comparable to @xmath391 , but this constrains the value of the higgs mass .
the smallness of @xmath391 indicates a tendency of perturbative effective potentials to give weakly first - order phase transitions .
this is apparent in the dependence of the order parameter , @xmath478 , or in the temperature interval in which the first - order phase transition can occur , @xmath479 .
for example , in the case of the electroweak phase transition , we have @xmath480 and @xmath481 for the minimal standard model .
if we take a non - realistic value for @xmath482 to get an order parameter of order @xmath20 , we find a temperature range @xmath483 . in the specific case of the electroweak phase transition ,
a small value of the higgs field @xmath4 is undesirable for electroweak baryogenesis . in general ,
if this parameter is too small , the perturbative approximation breaks down ( see appendix [ ap1 ] ) . in the electroweak case , the way out is to consider extensions of the sm which provide additional bosons that contribute to the parameter @xmath391 @xcite . in the previous section we found that if we increase the parameter @xmath391 while keeping the others constant , then we get a stronger phase transition and larger supercooling , but also larger values of @xmath484 and @xmath120 , which slow down the dynamics .
it is evident that if @xmath391 is augmented by adding particles to the theory , the value of @xmath452 enlarges too , giving an additional increase of the latent heat @xmath120 .
if we add only a boson , the relative change will be more appreciable in @xmath391 , because there are only a few terms in its expression , but if the boson comes together with several new species ( as in the case of supersymmetry ) , then the change of @xmath452 will be much more substantial . according to eqs .
( [ etatildeth ] ) and ( [ etatildeir ] ) , the friction coefficient will also increase significantly when adding bosons to the theory . if @xmath485 , then @xmath486 , and @xmath487 .
it is interesting to compare the value of @xmath120 with that of @xmath488 , to assess the effect of reheating , as discussed in section [ transicion ] . using eqs .
( [ aproxtn ] ) and ( [ tct0 ] ) , we may write @xmath489 therefore , @xmath490 the first factor is likely of order 1 and depends essentially on the energy scale of the transition .
the second factor is determined by the dynamics .
it depends mainly on @xmath452 , and may vary considerably if we change the particle content of the theory . exemplifying again with the electroweak theory , @xmath452 can vary from @xmath491 in the sm to @xmath492 in the mssm , so we pass from little reheating in the first case to large reheating in the latter .
we remark that things may be quite different if eqs .
( [ potparam ] ) are not valid . for instance , in the case of the quark - hadron phase transition @xmath120 and @xmath493 are typically of the order of the energy density @xmath26 .
in this paper we have performed an entirely analytical study of first - order phase transitions in the radiation - dominated era .
we have seen that typically the high - temperature phase is supercooled to a temperature @xmath239 , after which the transition proceeds in two steps , as sketched in fig .
[ esquema ] .
the first stage is complex , and some rough approximations must be made for an analytical treatment .
nevertheless , it can be checked with numerical results , ( e.g. @xcite ) , that the orders of magnitude are correct .
the second stage is much more simple , since bubble nucleation has effectively stopped and bubbles expand very slowly .
this stage develops very close to the critical temperature , with almost zero pressure difference between the two phases .
therefore , this part of the evolution is similar to the case of inhomogeneous nucleation , in which the presence of impurities induces bubble nucleations without need of supercooling .
we have studied the case of a phase transition at phase equilibrium in some detail , taking advantage of the fact that it can be solved analytically for any value of the relevant parameter @xmath494 .
this approach has allowed us to calculate the back - reaction on the expansion rate @xmath28 , which is important for large @xmath83 .
it is well known that supercooling may lead to exponential expansion of the universe @xcite .
this possibility has been considered not only in the context of inflationary models , but also for the quark - hadron phase transition @xcite .
although our approximations break down for @xmath495 , we observe the manifestation of the energy of the false vacuum for large @xmath120 . even if the phase transition begins at @xmath24 , when @xmath120 is comparable to the energy density of the plasma the transition may take a long time to complete due to vacuum energy dominance .
for the more probable case of a phase transition with variation of temperature , we have given a derivation of the integro - differential equations that govern the dynamics .
in particular , we have found a simple algebraic relation between the temperature , the fraction of volume occupied by the low - temperature phase , and the scale factor of the universe , which holds under the usual assumption of adiabatic expansion . using the thin wall approximation and the linearized form of the free energy difference
, we have found analytical formulas for all the quantities that characterize the dynamics of the phase transition and may be relevant for the determination of its cosmological consequences .
these parameters are the durations @xmath308 and @xmath311 of the two stages of the transition , the wall velocities @xmath376 and @xmath372 at each stage , the total number density of bubbles , the time interval @xmath496 in which bubble nucleation is active , etc .
we have expressed these quantities in terms of those that determine the dynamics , namely , thermodynamical parameters like the latent heat @xmath120 , the wall tension @xmath497 , or the friction coefficient @xmath484 .
as expected , for the phase equilibrium stage we have simple expressions , of the sort @xmath498 , with obvious interpretation .
more complex formulas arise instead for the reheating stage . although these parameters must be calculated in each particular case , some relations can be established , that hold quite broadly
. they allowed us to confirm some natural premises for the dynamics , e.g. , that @xmath499 and @xmath500 .
we have studied also the interrelations between the thermodynamical parameters .
when necessary , we made use of a simple model for the free energy . aside from reproducing the desired features of the phase transition , it is well known that this model corresponds to the simplest high - temperature effective potential that arises in perturbation theory .
we have derived general expressions for the parameters of this potential , which is useful in establishing further relations between the thermodynamical parameters .
we have also derived general expressions for the friction on the bubble walls .
this is caused by the perturbation from equilibrium of the particles of the plasma due to the motion of the interfaces .
we have compared the case of thermal particles to that of coherent infrared bosons .
we have found a different parametric dependence of each contribution , which indicates that each of them will dominate in different parameter ranges .
we have argued that probably one of these contributions will cause the wall to move non - relativistically .
this justifies the near - equilibrium approximations that simplify the analysis of the phase transition .
we have thus been able to find some general relations between the parameters that determine the dynamics .
for instance , we have seen that if the first - order phase transition is strengthened , then the supercooling is intensified , but also the friction and latent heat are generically enlarged , giving a slower evolution .
this general feature is easily detected with ad hoc variations of the parameters of the free energy , and further confirmed by the relations that arise for perturbation - theory values of these parameters .
the amount of supercooling is characterized by the difference between the energy density of the plasma at the critical temperature , and that at which nucleation begins , @xmath501 .
the specific relation between @xmath502 and the latent heat is decisive for the phase transition dynamics .
as we have seen , the ratio @xmath503 can either be small or large , depending on the theory .
of course , it is larger for stronger phase transitions .
we have argued that the case of interest corresponds to a latent heat comparable to @xmath502 .
on one hand , a small @xmath120 is related to too weakly first - order phase transitions . on the other hand , for large @xmath504
most of the phase transition occurs close to phase equilibrium , and can be described as a phase transition at constant @xmath505 .
this includes the case @xmath506 if @xmath507 , thus justifying the approximation @xmath128 in the case of supercooling . for the study of the different cosmological consequences
, additional specific relations will be relevant in each case @xcite , which can be obtained from the present analysis .
our analytical approximations will thus prove useful to include details of phase transition dynamics in the calculation of cosmic remnants , particularly with regard to the variation of the pertinent parameters throughout the phase transition .
for example , the importance of the phase equilibrium stage has been already investigated in @xcite in the context of electroweak baryogenesis .
other cosmological consequences are affected as well .
for instance , the fact that the nucleation rate is turned off due to reheating evidently modifies the number of nucleated bubbles , and thus the density of topological defects .
i am very grateful to jaume garriga for useful discussions and suggestions .
i wish to express my gratitude to professor luis masperi , who stimulated my interest on this subject .
this work was partially supported by a postdoctoral fellowship of the ministerio de educacin y cultura , spain .
following @xcite , we will obtain the high - temperature effective potential ( or free energy ) in the one - loop approximation , including leading - order plasma effects .
additional terms appear at higher - loop order .
for instance , potentially important terms of the form @xmath508 arise at two loops @xcite .
however , inclusion of two - loop corrections makes the situation more complicated and lies out of the scope of our general analysis .
we consider a gauge theory which is spontaneously broken by a vev of a scalar field @xmath6 . the tree - level potential is @xmath509 so the vev is given by @xmath510 and the higgs mass by @xmath511 .
at one - loop the effective potential picks up zero - temperature and finite - temperature corrections . with a cut - off regularization and tree level values for @xmath465 and @xmath464 ,
the zero - temperature contribution of a particle species is @xcite @xmath512 + 2m^{2}\left ( v\right ) m^{2}\left ( \phi \right ) \right\ } , \ ] ] where the @xmath513 is for bosons ( fermions ) , @xmath514 is the number of degrees of freedom of the species , and @xmath515 is the mass of the particle in the presence of the background scalar field .
it is in general of the form @xmath516 where @xmath517 is the coupling of the particle with @xmath6 ( i.e. , the yukawa coupling , gauge coupling , etc . ) .
the finite - temperature corrections are of the form @xmath518 where the functions @xmath519 and @xmath520 can be expanded in powers of @xmath521 for @xmath522 , and fall off exponentially for large @xmath521 .
therefore species with @xmath523 decouple from the plasma and we make a high - temperature approximation in which we consider only particles with @xmath524 .
expanding up to @xmath525 we have ( see e.g. @xcite ) @xmath526 where @xmath61 is the effective number of relativistic degrees of freedom , given in eq .
( [ g * ] ) and @xmath527 are the physical masses .
we assume that particles contributing to @xmath528 do not decouple during the phase transition , i.e. , that the condition @xmath529 is preserved in the range of temperatures of interest .
this is a reasonable assumption provided that @xmath530 for temperatures close to @xmath9 , which is consistent with the one - loop approximation .
if some particles decouple from the plasma during the transition , the main effect is a change in @xmath61 .
the @xmath531 term is the contribution of the bosonic zero modes to the one - loop effective potential .
this term is the most important to us ; without it the phase transition would be of second order .
however , for the zero modes the loop expansion has an infrared problem .
the perturbative expansion breaks down at higher - loop order , since higher loops contribute powers of @xmath532 and of @xmath533 @xcite .
the way out is to dress the zero modes with daisy and superdaisy diagrams .
the result of this resummation is a contribution of the form ( to all order in @xmath534 and to @xmath535 ) @xmath536 plus contributions proportional to @xmath537 which are unimportant within the present approximations .
thus the mass @xmath538 gets replaced with its debye mass @xmath539 where @xmath540 is the self - energy of the boson particle . in general
it is a combination of squared coupling constants times @xmath541 .
the exception are the transverse components of the gauge bosons , for which @xmath542 . if we replace the debye mass ( [ debye ] ) in the cubic term of eq .
( [ veffap ] ) , and the masses ( [ mdefi ] ) everywhere else , the resulting terms can be grouped as follows : [ [ constant - terms . ] ] constant terms .
+ + + + + + + + + + + + + + + these are contributions to the cosmological constant .
the total cosmological constant must be set by hand , so that it is almost zero after the phase transition .
[ [ t - dependent - protectphi -- independent - terms . ] ] @xmath20-dependent , @xmath543-independent terms .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + apart from the first term in ( [ veffap ] ) , there are also @xmath541 and logarithmic terms , but these are of order @xmath544 and @xmath545 with respect to the @xmath133 term , so we can neglect them within the approximation @xmath546 .
we notice , however , that we could have a large negative @xmath547 of order @xmath20 , such that @xmath397 is small ( see below ) . in any case , these corrections modify the equations of state of both phases in the same way , and we do not expect them to affect significantly the dynamics of the phase transition .
[ [ protectphi-2-terms . ] ] @xmath548 terms .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the coefficient of @xmath537 is the sum of a term proportional to @xmath541 , a constant term @xmath549 , and other constant and logarithmic terms , which are @xmath550 .
the latter are suppressed unless the higgs mass is too small , so we will disregard them . in any case , these corrections are inconsequential for our purposes .
they contribute to the value of the characteristic temperature @xmath12 of the phase transition .
however , for the dynamics of the phase transition the precise value of @xmath12 is not relevant ; the important parameter is the relative difference between this temperature and the critical one , @xmath551 , which is independent of @xmath12 .
[ [ protectphi-3-terms . ] ] @xmath552 terms .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the @xmath553 term has contributions from all the bosons , proportional to @xmath554 these terms may strongly affect the nature of the phase transition , depending on the value of @xmath555 .
there are two limiting cases , [ [ i ] ] i ) if @xmath556 , ( [ cubicterm ] ) contributes a term of the form @xmath557 to the free energy , which favors a strongly first - order phase transition , and [ [ ii ] ] ii ) if @xmath558 , we can expand ( [ cubicterm ] ) in powers of @xmath6 .
this gives higher - order corrections to the coefficients of @xmath537 and @xmath559 , and no contribution to @xmath560 .
almost all particles fall in the second case , since @xmath561 is of order @xmath562 , and @xmath563 near the phase transition .
the transverse components of the vector bosons , on the contrary , are protected against this thermal screening .
another exception may be a scalar with a negative @xmath564 .
such a particle would fall in the first case or in an intermediate case , and may play a role in determining the character of the phase transition @xcite .
however , such a tuning may induce unwanted minima in the scalar potential @xcite .
we are not going to consider this possibility here .
accordingly , the cubic term in the effective potential is @xmath565 where the sum is only over gauge bosons , and the factor @xmath566 is due to the fact that only two degrees of freedom of the massive vector contribute@xcite . [ [ protectphi-4-terms . ] ] @xmath567 terms .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the corrections to @xmath393 depend logarithmically on @xmath20 , so the effective value of @xmath393 may be regarded to be constant during the phase transition . furthermore , these corrections are of order @xmath568 , so they can be neglected provided that @xmath569 . for simplicity we will assume that this is the case .
putting all these terms together we see that , under the above assumptions and approximations , the free energy density takes the form displayed in eqs .
( [ free],[veff ] ) , with coefficients given by eqs .
( [ potparam ] ) .
in this appendix we make a derivation of the friction exerted by the hot plasma on the bubble walls . for that
, we must calculate the departure from equilibrium of the phase space population functions , @xmath401 in eq .
( [ eqfiplasma ] ) .
the friction on the wall has been extensively studied in the case of the electroweak phase transition @xcite .
our aim here is to discuss the general dependence of the friction on the particle content of a theory , so we will need to use some approximations in order to keep the description as general as possible .
we begin by considering the contribution of particles with @xmath570 ( @xmath571 wall width ) , for which the background field varies slowly and the semiclassical ( wkb ) approximation is valid . since in general @xmath572 , this condition is satisfied for all but the most infrared particles @xcite , which we study below .
we follow refs .
@xcite , but we use a simpler ansatz for the deviations from equilibrium distributions .
this will suffice for our purposes .
we assume that the population density of a particle species in the background of the domain wall ( that moves along the @xmath408-direction ) is governed by the boltzmann equation @xmath573 f =- c\left [ f\right ] , \label{boltzmann}\ ] ] where @xmath574 is the particle energy , @xmath575 is the particle velocity , @xmath576 is the force on the particle , and @xmath577 $ ] is the collision integral .
we use the ansantz @xmath578 , where @xmath579 so the deviation from @xmath580 is @xmath581 .
thus we obtain an equation for @xmath582 by linearizing the boltzmann equation .
keeping only terms of order @xmath583 we have @xmath584 = 0 .
\label{boltzlinear}\ ] ] the mass of the particle is a function of @xmath585 , and so is the perturbation @xmath582 if we assume a stationary state .
thus we make the replacements @xmath586 and @xmath587 , where the prime means derivative with respect to @xmath408 .
we further simplify eq .
( [ boltzlinear ] ) by making the integration @xmath588 .
we obtain @xmath589 where @xmath590 and @xmath591 are defined by the integrals @xmath592 and we have written the collision integral in the form @xcite @xmath593 } { t^{2 } } = t\gamma \delta .
\label{colterm}\ ] ] to lowest order in @xmath521 we have @xmath594 and @xmath595 for fermions , and @xmath596 and @xmath597 for bosons .
for each particle species @xmath390 we have a fluid equation of the form ( [ fluideq ] ) .
these equations are coupled through the collision term ( [ colterm ] ) , and @xmath237 is in principle a matrix with indices running over all particle species is split up into three different perturbations , @xmath598 . in that case
there are three fluid equations for each particle species , with different rates @xmath599 , @xmath600 and @xmath601 , and there arise additional constants @xmath602 and @xmath603 .
our approximation corresponds to considering only the term @xmath604 . ] .
however , only particles with large yukawa couplings are relevant for the friction force , since they have stronger interactions with the bubble wall . in accordance with @xcite ,
in this appendix we call heavy `` these particles with large @xmath463
. notice however that heavy particles with large mass @xmath547 in the unbroken phase may be thermally decoupled and not contribute to the friction at all .
the remaining light '' particles can be treated as a common background perturbation @xmath605 .
the fluid equation for the background is simpler and can be solved to eliminate @xmath605 . moreover
, the heavy particles primarily collide with the light particles , so direct coupling between heavy species can be neglected .
the effect of the light background , though , once eliminated from the equations , is to introduce a weak coupling between heavy particles . as a consequence ,
heavy particles are only weakly coupled through the background , and the non - diagonal terms of @xmath237 are suppressed with respect to the diagonal terms by a factor @xmath606 , where @xmath607 is the number of light species of the background ( it is proportional to the heat capacity of the plasma ) .
we will therefore neglect non - diagonal terms of @xmath237 in our analysis . calculating the rates @xmath608 is well beyond the scope of this work .
they are of the form @xmath609 , where @xmath534 is a gauge coupling @xcite .
we will assume that in general , @xmath610 the r.h.s of eq .
( [ fluideq ] ) is the source term of the equation .
it is localized at the bubble wall , so we expect the same for @xmath582 .
therefore we have @xmath611 .
normally , @xmath612 , so if the wall velocity is small , the first term on the l.h.s of eq .
( [ fluideq ] ) is much less than the second one and can be neglected . with this approximation
the equation has a simple solution , @xmath613 if we now insert @xmath614 in eq . ( [ eqfiplasma ] ) we obtain @xmath615 replacing the value of @xmath616 given by eq .
( [ delta ] ) , multiplying times @xmath617 , then integrating with respect to @xmath408 , we get @xmath618 the l.h.s .
is the pressure difference between the two phases .
it is equilibrated by a friction force of the form @xmath619 .
the friction coefficient is thus @xmath620 the coefficient @xmath590 for bosons depends on @xmath621 , but it is easy to see that its variation with @xmath408 can be neglected , and we can make the approximation @xmath622 to evaluate the integral in ( [ eta ] ) we use the thin wall approximation ( [ thinwallapp ] ) , @xmath623 it is clear that this integral goes like @xmath624 . for the model ( [ veff ] ) , the integral is easily calculated using eq .
( [ vtc ] ) .
it gives @xmath625 , so the friction coefficient is given by @xmath626 where @xmath468 for fermions and @xmath469 for bosons .
according to the arguments above , we will make the assumption that the parenthesis in the denominator of the last equation is roughly @xmath627 .
this gives eqs .
( [ etath ] ) and ( [ etatildeth ] ) .
it has been shown @xcite that coherent gauge fields can have important contributions to the friction .
following ref .
@xcite , we will estimate the contribution of a gauge boson to @xmath157 .
infrared boson excitations must be treated classically @xcite ; furthermore , the dynamics of the soft fields is overdamped by hard particles @xcite . as a consequence , the equation for the population function
is given by @xcite @xmath628 which comes from a similar equation for the amplitude of the field .
here , @xmath629 is the debye mass , given by @xmath630 , where , according to our previous notation , @xmath517 is the gauge coupling , and @xmath475 is roughly proportional to the number of particles that couple to the gauge field . averaging over the noise , we get the restoring term @xmath631 in the rhs of eq .
( [ damp ] ) . since @xmath632 , and @xmath633 for small @xmath19 ,
we can write @xmath634 inserting in ( [ eqfiplasma ] ) , multiplying by @xmath617 , and integrating as before , we obtain @xmath635 since the momentum integral is infrared dominated , we can approximate @xmath636 , so the momentum integral yields @xmath637 . with @xmath638
, we have @xmath639 the last integral can be calculated using again the thin wall approximation , eqs .
( [ thinwallapp ] ) and ( [ vtc ] ) , @xmath640 there is a logarithmic divergence that must be cut off where the approximations used in this derivation break down @xcite .
perturbation theory breaks down when @xmath641 , i.e. , at @xmath642 . the kinetic theory description that leads to eq .
( [ eqfiplasma ] ) breaks down when @xmath643 , i.e. , at @xmath644 .
the latter occurs first , so the @xmath271 is cut off at @xmath645 . in ref .
@xcite it is argued that the contribution of very infrared degrees of freedom is subdominant , since their wavelength can not resolve the thickness of the wall .
hence , the integral in ( [ integral ] ) gives to leading @xmath476 , @xmath646 , and the friction coefficient is @xmath647 which gives eqs .
( [ etair ] ) and ( [ etatildeir ] ) .
s. r. coleman , phys .
d * 15 * , 2929 ( 1977 ) [ erratum - ibid .
d * 16 * , 1248 ( 1977 ) ] ; c. g. callan and s. r. coleman , phys .
d * 16 * , 1762 ( 1977 ) .
g. d. moore and t. prokopec , phys .
d * 52 * , 7182 ( 1995 ) [ arxiv : hep - ph/9506475 ] ; phys .
* 75 * , 777 ( 1995 ) [ arxiv : hep - ph/9503296 ] .
p. john and m. g. schmidt , nucl .
b * 598 * , 291 ( 2001 ) [ erratum - ibid .
b * 648 * , 449 ( 2003 ) ] [ arxiv : hep - ph/0002050 ] .
y. huet , k. kajantie , r. g. leigh , b. h. liu and l. d. mclerran , phys .
d * 48 * , 2477 ( 1993 ) [ arxiv : hep - ph/9212224 ] ; j. ignatius , k. kajantie , h. kurki - suonio and m. laine , phys .
d * 49 * , 3854 ( 1994 ) [ arxiv : astro - ph/9309059 ] ; h. kurki - suonio and m. laine , phys .
d * 51 * , 5431 ( 1995 ) [ arxiv : hep - ph/9501216 ] .
a. f. heckler , phys .
d * 51 * , 405 ( 1995 ) [ arxiv : astro - ph/9407064 ] .
m. quiros , arxiv : hep - ph/9901312 .
a. mgevand , in preparation .
a. g. cohen , d. b. kaplan and a. e. nelson , ann .
nucl . part .
sci . * 43 * , 27 ( 1993 ) [ arxiv : hep - ph/9302210 ] ; a. riotto and m. trodden , ann .
nucl . part .
sci . * 49 * , 35 ( 1999 ) [ arxiv : hep - ph/9901362 ] .
m. e. shaposhnikov , nucl .
b * 287 * , 757 ( 1987 ) ; m. e. shaposhnikov , nucl .
b * 299 * , 797 ( 1988 ) ; a. i. bochkarev and m. e. shaposhnikov , mod .
a * 2 * , 417 ( 1987 ) ; a. i. bochkarev , s. v. kuzmin and m. e. shaposhnikov , phys .
d * 43 * , 369 ( 1991 ) .
b. h. liu , l. d. mclerran and n. turok , phys .
d * 46 * , 2668 ( 1992 ) ; n. turok , phys .
* 68 * , 1803 ( 1992 ) .
a. e. nelson , d. b. kaplan and a. g. cohen , nucl .
b * 373 * , 453 ( 1992 ) .
j. m. cline and k. kainulainen , phys .
85 * , 5519 ( 2000 ) [ arxiv : hep - ph/0002272 ] ; j. m. cline , m. joyce and k. kainulainen , jhep * 0007 * , 018 ( 2000 ) [ arxiv : hep - ph/0006119 ] ; m. carena , j. m. moreno , m. quiros , m. seco and c. e. wagner , nucl .
b * 599 * , 158 ( 2001 ) [ arxiv : hep - ph/0011055 ] .
a. megevand , phys .
d * 64 * , 027303 ( 2001 ) [ arxiv : hep - ph/0011019 ] .
e. witten , phys .
d * 30 * , 272 ( 1984 ) .
g. m. fuller , g. j. mathews and c. r. alcock , phys .
1380 ( 1988 ) . j. h. applegate and c. j. hogan , phys .
d * 31 * , 3037 ( 1985 ) ; h. kurki - suonio , phys .
d * 37 * , 2104 ( 1988 ) .
j. ignatius , k. kajantie , h. kurki - suonio and m. laine , phys .
d * 50 * , 3738 ( 1994 ) [ arxiv : hep - ph/9405336 ] .
j. h. applegate , c. j. hogan and r. j. scherrer , phys .
d * 35 * , 1151 ( 1987 ) .
e. farhi and r. l. jaffe , phys .
d * 30 * , 2379 ( 1984 ) ; m. giovannini and m. e. shaposhnikov , phys . rev .
d * 57 * , 2186 ( 1998 ) [ arxiv : hep - ph/9710234 ] .
j. b. rehm and k. jedamzik , phys .
d * 63 * , 043509 ( 2001 ) [ arxiv : astro - ph/0006381 ] .
r. h. brandenberger , a. c. davis and m. j. rees , phys .
b * 349 * , 329 ( 1995 ) [ arxiv : astro - ph/9501040 ] .
s. sanyal , phys .
d * 67 * , 074009 ( 2003 ) [ arxiv : hep - ph/0211208 ]
. t. w. b. kibble , j. phys . a * 9 * , 1387 ( 1976 ) .
t. w. b. kibble and a. vilenkin , phys .
d * 52 * , 679 ( 1995 ) [ arxiv : hep - ph/9501266 ] .
j. ahonen and k. enqvist , phys .
d * 57 * , 664 ( 1998 ) [ arxiv : hep - ph/9704334 ] ; d. grasso and a. riotto , phys .
b * 418 * , 258 ( 1998 ) [ arxiv : hep - ph/9707265 ] ; e. j. copeland , p. m. saffin and o. tornkvist , phys .
d * 61 * , 105005 ( 2000 ) [ arxiv : hep - ph/9907437 ]
. t. vachaspati , nucl .
b * 397 * , 648 ( 1993 ) ; m. james , l. perivolaropoulos and t. vachaspati , phys .
d * 46 * , 5232 ( 1992 ) ; t. vachaspati and r. watkins , phys .
b * 318 * , 163 ( 1993 ) [ arxiv : hep - ph/9211284 ] ; l. masperi and a. megevand , z. phys .
c * 72 * , 171 ( 1996 ) [ arxiv : hep - ph/9410211 ] .
p. elmfors , k. enqvist and k. kainulainen , phys .
b * 440 * , 269 ( 1998 ) [ arxiv : hep - ph/9806403 ] ; k. kajantie , m. laine , j. peisa , k. rummukainen and m. e. shaposhnikov , nucl .
b * 544 * , 357 ( 1999 ) [ arxiv : hep - lat/9809004 ] ; r. fiore , a. tiesi , l. masperi and a. megevand , mod .
a * 14 * , 407 ( 1999 ) [ arxiv : hep - ph/9804336 ] .
j. borrill , t. w. b. kibble , t. vachaspati and a. vilenkin , phys .
d * 52 * , 1934 ( 1995 ) [ arxiv : hep - ph/9503223 ] ; m. lilley and a. ferrera , phys .
d * 64 * , 023520 ( 2001 ) [ arxiv : hep - ph/0102035 ] .
e. suhonen , phys .
b * 119 * , 81 ( 1982 ) .
k. enqvist , j. ignatius , k. kajantie and k. rummukainen , phys .
d * 45 * , 3415 ( 1992 ) .
p. j. steinhardt , nucl .
b * 190 * , 583 ( 1981 ) .
a. h. guth and e. j. weinberg , phys .
d * 23 * , 876 ( 1981 ) .
r. fiore , d. galeazzi , l. masperi and a. megevand , mod .
lett . a * 9 * , 557 ( 1994 ) [ arxiv : hep - th/9310039 ] .
r. fiore , l. masperi and a. megevand , mod .
a * 11 * , 1847 ( 1996 ) [ arxiv : hep - ph/9510395 ] .
d. spector , phys .
b * 194 * , 103 ( 1987 ) ; a. kusenko , phys .
b * 406 * , 26 ( 1997 ) [ arxiv : hep - ph/9705361 ] ; d. metaxas , phys .
d * 63 * , 083507 ( 2001 ) [ arxiv : hep - ph/0009225 ] .
m. b. christiansen and j. madsen , phys .
d * 53 * , 5446 ( 1996 ) [ arxiv : astro - ph/9602071 ] ; k. jedamzik and g. m. fuller , astrophys . j. * 423 * , 33 ( 1994 ) [ arxiv : astro - ph/9312063 ] .
m. a. halasz , a. d. jackson , r. e. shrock , m. a. stephanov and j. j. verbaarschot , phys .
d * 58 * , 096007 ( 1998 ) [ arxiv : hep - ph/9804290 ] .
z. fodor and s. d. katz , phys .
b * 534 * ( 2002 ) 87 [ arxiv : hep - lat/0104001 ] ; jhep * 0203 * ( 2002 ) 014 [ arxiv : hep - lat/0106002 ] .
a. h. guth and e. j. weinberg , nucl .
b * 212 * , 321 ( 1983 ) . c. j. hogan , phys .
b * 133 * , 172 ( 1983 ) ; t. degrand and k. kajantie , phys .
b * 147 * , 273 ( 1984 ) ; k. kajantie and h. kurki - suonio , phys .
d * 34 * , 1719 ( 1986 ) ; d. chandra and a. goyal , phys .
d * 62 * , 063505 ( 2000 ) [ arxiv : hep - ph/9903466 ] . j. r. espinosa , nucl .
b * 475 * , 273 ( 1996 ) [ arxiv : hep - ph/9604320 ] ; b. de carlos and j. r. espinosa , nucl .
b * 503 * , 24 ( 1997 ) [ arxiv : hep - ph/9703212 ] ; d. bodeker , p. john , m. laine and m. g. schmidt , nucl .
b * 497 * , 387 ( 1997 ) [ arxiv : hep - ph/9612364 ] ; m. carena , m. quiros and c. e. wagner , nucl .
b * 524 * , 3 ( 1998 ) [ arxiv : hep - ph/9710401 ] .
g. w. anderson and l. j. hall , phys .
d * 45 * , 2685 ( 1992 ) .
m. dine , r. g. leigh , p. y. huet , a. d. linde and d. a. linde , phys .
d * 46 * , 550 ( 1992 ) [ arxiv : hep - ph/9203203 ] .
i. affleck , phys .
lett . * 46 * , 388 ( 1981 ) .
a. d. linde , nucl .
b * 216 * , 421 ( 1983 ) [ erratum - ibid .
b * 223 * , 544 ( 1983 ) ] ; phys .
b * 100 * , 37 ( 1981 ) .
n. borghini , w. n. cottingham and r. vinh mau , j. phys .
g * 26 * , 771 ( 2000 ) [ arxiv : hep - ph/0001284 ] .
a. megevand , int .
j. mod .
d * 9 * , 733 ( 2000 ) [ arxiv : hep - ph/0006177 ] .
m. gleiser , phys .
lett . * 73 * , 3495 ( 1994 ) [ arxiv : hep - ph/9403310 ] ; j. borrill and m. gleiser , phys .
d * 51 * , 4111 ( 1995 ) [ arxiv : hep - ph/9410235 ] .
m. laine and k. rummukainen , nucl .
b * 597 * , 23 ( 2001 ) [ arxiv : hep - lat/0009025 ] .
s. y. khlebnikov , phys .
d * 46 * , 3223 ( 1992 ) ; p. arnold , phys .
d * 48 * , 1539 ( 1993 ) [ arxiv : hep - ph/9302258 ] .
m. carena , m. quiros and c. e. wagner , phys .
b * 380 * , 81 ( 1996 ) [ arxiv : hep - ph/9603420 ] .
m. carena and c. e. wagner , nucl .
b * 452 * , 45 ( 1995 ) [ arxiv : hep - ph/9408253 ] .
g. d. moore , jhep * 0003 * , 006 ( 2000 ) [ arxiv : hep - ph/0001274 ]
. g. d. moore and n. turok , phys .
d * 55 * , 6538 ( 1997 ) [ arxiv : hep - ph/9608350 ] .
p. arnold , d. son and l. g. yaffe , phys .
d * 55 * , 6264 ( 1997 ) [ arxiv : hep - ph/9609481 ] .
|
we consider first - order phase transitions of the universe in the radiation - dominated era .
we argue that in general the velocity of interfaces is non - relativistic due to the interaction with the plasma and the release of latent heat .
we study the general evolution of such slow phase transitions , which comprise essentially a short reheating stage and a longer phase equilibrium stage .
we perform a completely analytical description of both stages .
some rough approximations are needed for the first stage , due to the non - trivial relations between the quantities that determine the variation of temperature with time .
the second stage , instead , is considerably simplified by the fact that it develops at a constant temperature , close to the critical one .
indeed , in this case the equations can be solved exactly , including back - reaction on the expansion of the universe .
this treatment also applies to phase transitions mediated by impurities .
we also investigate the relations between the different parameters that govern the characteristics of the phase transition and its cosmological consequences , and discuss the dependence of these parameters with the particle content of the theory .
|
consider a normal distribution with zero mean and an unknown covariance matrix , @xmath2 let denote the eigenvalues of @xmath3 by @xmath4 and eigenvectors matrix by @xmath5 , hence we have the spectral decomposition @xmath6 where @xmath7 means the diagonal matrix with the @xmath8th diagonal element @xmath9 .
it is needless to say that inference on @xmath3 is an important task in many practical situations in such a diversity of fields as engineering , biology , chemistry , finance , psychology etc .
especially we often encounter the cases where the property of interest depends on @xmath3 only through its eigenvalues @xmath10 .
we treat an inference problem on the eigenvalues in this paper .
geometrically viewed , the family of normal distributions @xmath11 is taken as a manifold ( say @xmath12 ) with a single coordinate system @xmath3 .
hence , @xmath12 is identified with the space of symmetric positive definite matrices . geometrically analyzing the space of symmetric positive definite matrices is an interesting topic in a mathematical or engineering point of view .
for example , see smith @xcite , fletcher and joshi , , lenglet et .
al . @xcite . from a statistical point of view
, it could give us some new insights for the inference on @xmath3 . in this paper , we analyze @xmath12 from the standpoint of information geometry while focusing on the inference on the eigenvalues of @xmath13 the two kinds of submanifolds play an important role , that is , a submanifold given by fixed eigenvectors of the covariance matrix and the one given by fixed eigenvalues . based on independent @xmath14 samples @xmath15 from this distribution , we want to make inference on the unknown @xmath10 .
it is well - known that the product - sum matrix @xmath16 is sufficient statistic for both unknown @xmath10 and @xmath5 .
the spectrum decomposition of @xmath17 is given by @xmath18 where @xmath19 are the eigenvalues of @xmath17 , and @xmath20 is the corresponding eigenvectors matrix .
this decomposition gives us two statistics available , i.e. the sample eigenvalues @xmath21 and the sample eigenvectors @xmath20 .
however it is almost customary that we only use the sample eigenvalues , discarding the information contained in @xmath20 . in the past literature on the inference for the population eigenvalues ,
every notable estimator is based simply on the sample eigenvalues .
here we just mention the literature in the statistical field that deals with the estimation of @xmath10 under the situation @xmath22 ; stein @xcite , takemura @xcite , dey and srinivasan , haff @xcite , yang and berger for orthogonally invariant estimators of @xmath3 ; dey @xcite , hydorn and muirhead , jin @xcite , sheena and takemura for a direct estimator of @xmath10 . in a sense
it is natural to implicitly associate the sample eigenvalues to the population eigenvalues , and the sample eigenvectors to the population counterpart .
however the sample eigenvalues is not sufficient for the unknown population eigenvalues .
therefore it is important to evaluate how much information is lost by neglecting the sample eigenvectors .
following amari @xcite , we gain an understanding of the asymptotic information loss with geometric terms such as fisher information metric and embedding curvatures .
another statistically interesting topic is the bias of @xmath21 .
it is well known that @xmath21 is largely biased and the estimators mentioned above are all modification of @xmath21 to correct the bias , that is , `` shrinkage estimators . ''
we show that the bias is closely related to the embedding curvatures .
moreover the geometric structure of @xmath12 naturally leads us to new estimators , which are also shrinkage estimators .
we briefly introduce the order of the contents in this paper . in section [ section : riemannian_manifold_and_metric ] and
[ section : embedding_curvature ] , we treat respectively fisher information metrics and embedding curvatures , and present their specific forms related to the eigenvalues inference . in section [ section : embedding_curvature ] , we refer to the bias of @xmath21 in relation to the curvatures . in section [ section : information_loss ] , we get the explicit form of the asymptotic information loss caused by discarding the sample eigenvectors in the first and second orders w.r.t . the sample size . in the final section ,
we propose the new estimators which is naturally derived in a geometrical view .
all the proofs of the propositions are collected in appendix . for geometrical concepts used in this paper ,
refer to amari @xcite , amari and nagaoka .
the density of the normal distribution @xmath11 is given by @xmath23 if we let @xmath24 and @xmath25 denote the @xmath26 element of respectively @xmath3 and @xmath27 , then the log likelihood equals @xmath28 where @xmath29 and @xmath30 are given by @xmath31 and @xmath32 the summations in the equation is an abbreviation for @xmath33 and @xmath34 , and we will use these kinds of notations hereafter without mention .
the expression gives natural coordinate system @xmath35 of the manifold @xmath12 as an full exponential family .
another coordinate system , so called expectation parameters , are also useful , which are defined as ; @xmath36 for the analysis of the information carried by @xmath21 and @xmath20 , we need to prepare another coordinate system .
the matrix exponential expression of an orthogonal matrix @xmath37 is given by @xmath38 where @xmath39 is the @xmath40-dimensional unit matrix , @xmath41 is a skew - symmetric matrix and parametrized by @xmath42 as @xmath43 the function @xmath44 is diffeomorphic , and @xmath45 gives `` normal coordinate '' for the group of orthogonal matrices .
we can use this coordinate as local system around @xmath39 and construct an atlas for the entire space of p - dimensional orthogonal matrices ( note this space is compact ) ; for each @xmath5 , there exists an open neighborhood and some open ball @xmath46 in @xmath47 around the origin such that these spaces are diffeomorphic by the function @xmath48 on @xmath46 .
we will use @xmath49 as the third coordinate system of @xmath12 and call it `` spectral coordinate ( system ) '' .
notice that this coordinate system is associated with the following submanifolds in @xmath12 .
if we fix @xmath5 in , then we get a submanifold @xmath50 embedded in @xmath12 with a coordinate system @xmath10 .
this is a subfamily in @xmath11 and called curved exponential family .
its log - likelihood is expressed , as we emphasize it as a function of @xmath10 , to be @xmath51 on the contrary , if we fix @xmath10 in , we get another submanifold @xmath52 in @xmath12 , whose coordinate system is given by @xmath45 in a neighborhood of each point of @xmath52 .
its log - likelihood expression is given by @xmath53 first we consider a metric , that is , a field of symmetric , positive definite , bilinear form on @xmath12 .
the statistically most natural metric is fisher information metric .
suppose @xmath54 is a parametric family of probability density functions , whose coordinate as a manifold is given by @xmath55 .
then the @xmath26 component of fisher information metric with respect to @xmath56 is given by @xmath57.\ ] ] for the multivariate normal distribution family , @xmath58 ( @xmath59 , the mean parameter is also included ) , skovgaard @xcite gives a clear form of fisher information metric .
the tangent vector space at a fixed point @xmath3 w.r.t .
@xmath60 coordinate can be identified with the space of symmetric matrices . for any symmetric matrix @xmath61 , @xmath62 , the metric with respect to the @xmath63 coordinate system
is given by @xmath64 we are interested in fisher information metric with respect to the spectral coordinate @xmath65 .
let @xmath66 denote the tangent vectors w.r.t .
the @xmath10 coordinate , @xmath67 denote the tangent vectors w.r.t .
the @xmath45 coordinate .
namely @xmath68 these tangent vectors ( exactly speaking , vector fields ) are invariant with respect to the orthogonal transformation of @xmath3 ; for some orthogonal matrix @xmath37 , the orthogonal transformation @xmath69of @xmath12 is defined as @xmath70 for any @xmath71 @xmath72 [ metric_spectra ] let @xmath73 denote fisher information metric based on @xmath74 , then the components of the metric with respect to @xmath49 is given as follows ; @xmath75 @xmath76 equals one if the logic inside the parenthesis is correct , otherwise zero .
there are two remarkable properties of the metric for the spectral coordinate .
first since the metric components matrix is diagonal , @xmath77 is an orthogonal coordinate system , especially that the submanifolds @xmath50 and @xmath52 are orthogonal to each other for any @xmath10 and @xmath5 .
second it is independent of @xmath5 , hence the metric is invariant with respect to the orthogonal transformation @xmath69 in for any orthogonal matrix @xmath78 ( second property is instantly derived from the expression . ) the fundamental structure of @xmath12 is determined by an affine connection , and it gives birth to curvature , geodesics , etc . as it is seen in amari @xcite , @xcite , amari and nagaoka , @xmath0-connection and @xmath1-connection is important for the analysis of statistical exponential family , but we do nt make further reference here .
all results related to levi - civita connection for @xmath12 can be found in skovgaard @xcite .
for @xmath79-connection for the multivariate normal family with zero mean , refer to eguchi @xcite .
a family of multivariate normal distributions as a riemannian manifold is studied by several authors .
refer to fletcher and joshi , lenglet et .
@xcite and lovri et .
al . @xcite .
we just mention the fact that @xmath12 is @xmath0-flat and @xmath1-flat , and corresponding affine coordinations are given respectively by @xmath80 and @xmath81 .
for the analysis of the asymptotic distribution @xmath82 , it is important to observe the embedding curvatures of @xmath83 and @xmath84 .
( see amari and kumon , kumon and amari . )
especially , asymptotic information loss can be explained in view of these curvatures .
embedding curvature shows how the submanifold @xmath83 or @xmath84 is placed in the manifold @xmath85 we consider the following embedding curvatures ; \1 .
embedding curvature of @xmath83 with respect to @xmath0-connection or @xmath1-connection .
@xmath86 where @xmath87 is the covariant derivative of @xmath88 in the direction of @xmath89 with respect to @xmath0-connection .
@xmath90 is similarly defined .
embedding curvature of @xmath84 with respect to @xmath1-connection @xmath91 where @xmath92 is the covariant derivative of @xmath93 in the direction of @xmath94 with respect to @xmath1-connection . on these curvatures at the point @xmath95 , we have the following results . [ embedding_c ] for @xmath96 @xmath97 for @xmath98 @xmath99 another expression of the embedding curvature of @xmath84
is given by @xmath100 with this notation , the orthogonal projection of the covariant derivative @xmath101 onto the tangent space of @xmath83 is given by @xmath102 from proposition [ metric_spectra ] , [ embedding_c ] , we have @xmath103 hence @xmath104 an embedding curvature has full information about the `` extrinsic curvature '' of the embedded submanifold in any direction .
sometimes it is convenient to compress it into a scalar measure of the curvature .
`` statistical curvature '' by efron ( see efron @xcite , murray and rice ) is such a measure ; for @xmath84 , it is defined by ( see p159 of amari @xcite ) @xmath105 which attains the following value at the point @xmath95 .
@xmath106 from these results , we notice that if @xmath12 is endowed with @xmath1-connection , then 1 ) the embedding curvatures and the statistical curvatures of @xmath84 are independent of @xmath5 , 2 ) any one - parameter curve @xmath107 given by the parameter @xmath108 , while @xmath10 and the other elements of @xmath109 are fixed , is curved in the direction of @xmath110 and contained in a two - dimensional plane composed by @xmath93 and @xmath110 , 3 ) the statistical curvature of @xmath84 could be quite large when @xmath10 are close to each other , while @xmath83 is flat everywhere . here
we introduce another submanifold @xmath111 which is contrasting to @xmath84 in the sense that @xmath111 is flat with respect to @xmath1-connection .
for a point @xmath95 , let @xmath112 we easily notice that @xmath111 is the minimum distance points with respect to kullback - leibler divergence .
that is , @xmath113 where @xmath114 is the kullback - leibler divergence between @xmath11 and @xmath115 , which is specifically given by @xmath116 the minimum distance points with respect to the kullback - leibler distance consists of all the points on the @xmath1-geodesics which pass through the point @xmath95 and are orthogonal to @xmath50 at that point .
( see theorem in a2 of amari [ 1 ] ) .
we can visualize the structure of @xmath12 endowed with @xmath1-connection for the two dimensional case .
see figure [ fig1 ] , where @xmath117 , @xmath118 and @xmath119 are drawn .
when @xmath120 , @xmath83 is a two - dimensional autoparallel submanifold with the affine coordinate @xmath121 , while @xmath84 is a one - dimensional submanifold with an coordinate @xmath122 .
as it is seen in proposition [ metric_spectra ] , all the tangent vectors @xmath123 , @xmath124 , @xmath125 are orthogonal to each other .
@xmath111 is a `` straight '' line which is also orthogonal to @xmath83 .
the arrow on @xmath83 is the line @xmath126 , and the arrow head indicates the direction in which @xmath127 increases .
the statistical curvature turns out to be the increasing function of @xmath128 ; @xmath129 , @xmath84 and @xmath111,width=377 ] we can analyze the bias of @xmath130 from the geometrical structure of @xmath12 .
it is well known that @xmath131\ ( i=1,\ldots , p)$ ] majorizes @xmath132 , that is , @xmath133
\geq \sum_{i=1}^j \lambda_i,\quad 1\leq \forall j \leq p-1 , \qquad \sum_{i=1}^p e[\,\bar{l}_i\ , ] = \sum_{i=1}^p \lambda_i.\ ] ] the bias @xmath131 $ ] is quite large when @xmath14 is small and @xmath9 s are close to each other ( see lawley @xcite , anderson @xcite ) . for the case @xmath120 , @xmath134 \geq \lambda_1,\qquad e[\,\bar{l}_2\ , ] \leq \lambda_2.\ ] ] suppose a sample @xmath135 takes the value at a point @xmath136 let @xmath137 denote the point on @xmath50 designated by the eigenvalues of @xmath138 , namely @xmath139 .
the curve @xmath140 connects @xmath141 and @xmath137 .
if we define @xmath142 as the point on @xmath50 designated by @xmath143 , then @xmath144 connects @xmath141 and @xmath142 .
the three points @xmath141 , @xmath137 and @xmath142 are on the same plane , and if we move from @xmath137 in the direction to @xmath142 , then the statistical curvature of @xmath84 increases ( see figure [ fig2 ] ) .
if we estimate @xmath145 by @xmath146 , then the estimand is the point @xmath137 , while for the unbiased estimator @xmath147 , the estimand is the point @xmath142 .
since the @xmath128-coordinate of @xmath137 is always smaller than that of @xmath142 , the estimator @xmath148 is likely to estimate @xmath149 and @xmath150 too apart , which causes the bias .
it is also seen that the bias gets larger when @xmath128 approaches to one , that is , @xmath149 and @xmath150 get closer to each other . and @xmath111 on @xmath128-coordinate , width=377 ]
now we consider the information loss caused by ignoring @xmath20 for the inference on @xmath10 .
information loss matrix @xmath151 at a fixed point @xmath152 is given by @xmath153=g_{ab}(\bm{s})-g_{ab}({\bm{l}}),\ ] ] where @xmath154 are the components of the metrics w.r.t .
@xmath89 and @xmath88 based on respectively the distributions @xmath17 , @xmath21 and the conditional distribution of @xmath17 given @xmath21 , all of which are measured at the point @xmath152 .
amari @xcite found that the asymptotic information loss can be expressed in terms of the metric and the embedding curvatures ; @xmath155 [ info_loss_result ] @xmath156 where @xmath157 @xmath158 at the point @xmath95 depends only on @xmath159 when the information loss of a statistic has the order @xmath160 , we call the statistic is the @xmath161th order sufficient . consequently the statistic @xmath21 is the first order sufficient , but not the second order sufficient .
@xmath158 , the information loss in the first order term ( @xmath162 ) is negligible when the eigenvalues are separated to some extent . on the contrary ,
it is quite large when the population eigenvalues are close to each other .
note that the information carried by @xmath21 is given by the formula ; @xmath163 since @xmath164 is positive definite , @xmath165 must be bounded in the neighborhood of a point where @xmath166 .
this indicates that the term of order @xmath167 in @xmath158 is also unbounded in such a neighborhood .
hence the expansion of the information loss with respect to @xmath14 is not useful when the population eigenvalues are close to each other . and
@xmath111 on @xmath128-coordinate , width=377 ] in order to check the information loss numerically , we carried out a simple simulation .
consider the null hypothesis @xmath168 the ordinary likelihood ratio test based on @xmath17 ( say the first test ) is given by the rejection region @xmath169 where @xmath170 is chosen so that the size of the test is @xmath79 .
see e.g. theorem 8.4.2 of muirhead @xcite . on the other hand ,
the likelihood ratio test based on the distribution of @xmath21 ( say the second test ) is given by @xmath171 since the density function of @xmath21 is given by @xmath172 where @xmath173 is the uniform probability on the group of @xmath40-dimensional orthogonal matrices , @xmath174 , and @xmath175 is the normalizing constant ( see theorem 3 .
2 . 18 of muirhead @xcite ) .
figure [ fig3 ] shows the simulation result on the powers of the two tests for the case @xmath120 , @xmath176 .
the alternative hypotheses are taken in various directions from the null point @xmath177 ; @xmath178 where @xmath179 ( note that the alternatives are not at the same distance from the null w.r.t .
the metric of proposition [ metric_spectra ] . ) the powers for each alternative are the mean values from @xmath180 times repetition . for the integral calculation in
, we picked up 100 points from @xmath181 in an equidistant manner .
the graph shows that the second test performs as well as the first test , hence it indicates the information loss is not serious at least around the point @xmath177 .
in this section , we apply the geometrical analysis in the previous sections to the estimation of @xmath10 . we consider two cases where the population eigenvectors , @xmath5 , is respectively known and unknown .
first let s consider the case where @xmath5 is known . in this case , statistics
@xmath182 is the second order sufficient , since @xmath183 is orthogonal to @xmath50 , and the @xmath1-embedding curvature of @xmath183 vanishes ( see the equation ) .
therefore the estimator @xmath184 is superior to the ordinary estimator @xmath185 in view of the asymptotic information loss . especially consider the case where we know that @xmath186 , that is , each variate is independent .
then the estimator @xmath187 becomes the ordinary sample variance for the @xmath8th variate .
the lesson is that if each variate is independent , then we should tackle with each variate separately , do nt treat them as a multivariate set . and @xmath188 as @xmath10 changes , width=377 ] we made a numerical evaluation of the superiority of @xmath147 to @xmath189 .
we used kullback - leibler divergence as a loss function ( say kl - loss ) .
namely for the true parameter @xmath190 and the estimators @xmath191 we evaluate the loss of the estimator as @xmath192 we compared the risks @xmath193,\ i=1,2 $ ] for the case @xmath194 under the condition @xmath195 , where @xmath128 varies from one to 0.02 by the increment of 0.02 .
note that both estimators and the loss function are all scale invariant .
figure [ fig4 ] shows the result .
we repeated @xmath180 times risk evaluation for each @xmath128 and took the average .
theoretically the risk of @xmath188 is constant .
the perturbation of the dotted line is due to the simulation error .
we observe that the risk of @xmath147 is reduced by nearly 30 % compared to @xmath189 in the neighborhood of @xmath196 , where the second order term of the information loss of @xmath189 is maximized . and @xmath188 as @xmath5 changes , width=377 ] in a practical situation , it is rare to know the exact population eigenvectors beforehand .
it is more likely that we only have a vague knowledge about the eigenvectors .
for example , we sometimes have a prior knowledge that each variate is almost uncorrelated to each other although they are not perfectly independent .
in this situation , which is better to use @xmath197 or @xmath189 ?
the following simulation suggest that the superiority of @xmath147 to @xmath189 could hold good in a relatively large neighborhood of @xmath198 .
figure [ fig5 ] shows the change of the risks of the two estimators w.r.t .
k - l loss under the condition @xmath199 as @xmath5 varies with @xmath200 as @xmath201 for each @xmath5 , the repetition was done @xmath180 times and we took the average as the risks .
@xmath202 has a constant risk , though the risk line for @xmath202 is perturbed by the simulation error .
the graph shows that @xmath188 still performs better than @xmath202 in spite of the situation @xmath203 as we saw in the previous subsection , if we have information about @xmath5 , the estimator @xmath147 works well .
if we do not have information about @xmath5 beforehand , a naive idea is to estimate it by the sample eigenvectors matrix @xmath20 and substitute it into @xmath147 .
however this only reproduces @xmath189 .
@xmath20 is a point estimator ( m.l.e . ) of @xmath5 , while we could construct a distribution of @xmath5 based on the sample @xmath17 .
if we take the expectation of @xmath204 with this distribution , it could produce a reasonable estimator . first consider the conditional distribution of @xmath20 when @xmath21 is given . since @xmath205 is distributed as wishart matrix @xmath206 , its density w.r.t .
the uniform probability @xmath173 on the group of @xmath40-dimensional orthogonal matrices @xmath174 equals @xmath207 where normalizing constant @xmath208 is given by @xmath209 this conditional distribution depends on @xmath3 . if we substitute @xmath3 with the estimator @xmath138 , a density of @xmath5 on @xmath174 with respect to @xmath210 is given by @xmath211 where @xmath212 take the expectation of @xmath213 w.r.t .
the density , @xmath214 let @xmath215 denote @xmath216 .
because of the invariance of @xmath217 , it turns out that @xmath218 where @xmath219 and @xmath220 as @xmath128 changes , width=377 ] we propose @xmath221 as a new estimator of @xmath10 .
it is easily proved that @xmath222 as @xmath223 , and @xmath224 is a `` shrinkage '' estimator .
the analytic evaluation of this estimator s performance seems difficult even for the large sample case .
instead we show the numerical result comparing @xmath146 and @xmath224 .
we simulated the risks of both estimators w.r.t .
k - l loss for the case @xmath120 .
since they are functions of @xmath21 and scale invariant , it is enough to measure the risks for @xmath225 .
we varied @xmath128 from 0.04 to 1.00 by the increment 0.04 , and for each @xmath128 we repeated the risk evaluation @xmath226 times and took the average .
for the integral calculation of and , we picked up 50 points from @xmath181 in an equidistant manner .
figure [ fig6 ] shows the result .
it seems that the new estimator performs better compared to @xmath189 , especially @xmath10 are close to each other .
as a base for the vector space of real symmetric matrices , we consider @xmath227 which is a @xmath228 matrix defined by @xmath229 where @xmath230 is the @xmath231 matrix whose @xmath232 element equals one , and all the other elements are zero .
the one to one correspondence @xmath233 gives the component expression of @xmath234 since @xmath235 we have the following relations @xmath236 where @xmath237 , @xmath238 for the first order derivative at @xmath239 , we only have to consider @xmath3 up to the term to the first power w.r.t .
@xmath45 , hence we put @xmath240 as @xmath241 therefore we have @xmath242 where @xmath243 which leads to @xmath244 and @xmath245 from and , we have the following results on tangent vectors ; @xmath246 where @xmath247 is the @xmath248th column of @xmath5 , and @xmath249 if we substitute and into , and , we get the results as follows ; @xmath250 @xmath251 note that @xmath252 , hence @xmath253 this means @xmath83 is an affine subspace of @xmath12 w.r.t .
an @xmath35 , which is an affine coordinate system of @xmath12 with @xmath0-connection .
consequently @xmath83 is @xmath0-flat , i.e. @xmath254 .
@xmath255 is similarly proved .
see theorem 1.1 in amari and nagaoka .
now we consider @xmath256 . using ( 4.14 ) in amari @xcite , it is calculated as @xmath257 where @xmath228 matrices @xmath61 , @xmath62 are given by @xmath258 in order to calculate @xmath61 , we only have to consider @xmath3 up to the terms powered by two w.r.t .
@xmath45 ; @xmath259 therefore @xmath24 is expressed as @xmath260 where @xmath261 since @xmath262 truns out to be @xmath263 from this we have @xmath264 where @xmath265 @xmath266 @xmath267 @xmath268
furthermore we have @xmath269 where @xmath270 since @xmath271 we have @xmath272 from and , we have @xmath273 the following equalities hold ; @xmath274 consequently @xmath275 as we will see in the next subsection , @xmath276 combine this with proposition [ metric_spectra ] , we have @xmath277 the term of the order @xmath14 in vanishes since @xmath278 equals zero for @xmath279 .
we consider the term of order @xmath162 . since @xmath280 also vanishes for @xmath281 , @xmath282 , we only have to consider the term @xmath283 because of
, the above term equals @xmath284 if @xmath285 , then equals @xmath286 if @xmath287 , then equals @xmath288
the author really appreciates dr .
m. kumon kindly answering his question on a basic fact of the information loss . c. lenglet , m. rousson , r. deriche and o. faugeras .
statistics on the manifold of multivariate normal distributions : theory and application to diffusion tensor mri processing .
_ journal of mathematical imaging and vision _ , 25 : 423 - 444 , 2006
|
we consider an inference on the eigenvalues of the covariance matrix of a multivariate normal distribution .
the family of multivariate normal distributions with a fixed mean is seen as a riemannian manifold with fisher information metric .
two submanifolds naturally arises ; one is the submanifold given by fixed eigenvectors of the covariance matrix , the other is the one given by fixed eigenvalues .
we analyze the geometrical structures of these manifolds such as metric , embedding curvature under @xmath0-connection or @xmath1-connection .
based on these results , we study 1 ) the bias of the sample eigenvalues , 2)the information loss caused by neglecting the sample eigenvectors , 3)new estimators that are naturally derived from the geometrical view .
msc(2010 ) _ subject classification _ : primary 62h05 ; secondary 62f12 + _ key words and phrases : _ curved exponential family , information loss , fisher information metric , embedding curvature , affine connection , positive definite matrix .
|
mixtures of @xmath0he and @xmath1he atoms exhibit a rich phase diagram , where besides a normal phase , there is a phase where @xmath1he is superfluid , and a phase separated region of superfluid @xmath1he and normal @xmath0he.@xcite in 1971 , blume , emery and griffiths@xcite proposed a model to describe such mixtures .
they simplified the continuous phase of the superfluid order parameter such that it could acquire only two values .
although they made this very rough approximation and modelled the uniform system in a lattice , their results are very interesting .
qualitatively , they reproduced the right phases and the right orders of the phase transitions . furthermore ,
if disorder is introduced by placing the mixture into aerogel , after some modifications,@xcite the model can still yield the experimentally observed phase diagram.@xcite here , we generalize this model to a two - component case in order to describe systems with two order parameters and study the problem numerically , using monte carlo simulations .
the motivation for the model we are proposing is twofold .
firstly , we would like to study condensed matter materials like heavy fermions , high-@xmath2 superconductors , and organic superconductors .
in particular , we want to study the interplay between magnetic and superconducting ordering in these materials .
both order parameters are modelled as an ising spin variable .
concerning the magnetism , we consider the ferro- and the antiferro - magnetic cases , and investigate also the effect of an additional magnetic field .
we find that in the absence of a magnetic field , in the region where the two orders coexist , the system is always phase separated .
when we add a magnetic field , we also find regions with microscopic coexistence of the two phases .
secondly , we want to study mixtures of cold atoms .
cold atoms have emerged in recent years as an ideal simulator of condensed matter systems . because experiments with cold atoms are often carried out in a trap , we add a trapping potential to the model .
this fact qualitatively changes the physics of the problem . for the case of a single component beg model in a trap ,
the results are compared with experimental and theoretical work on imbalanced fermi mixtures .
for the case of the two - component beg model , we make predictions for the phase diagram of boson - boson mixtures .
the outline of this paper is the following : in section [ sec : twobeg ] , we introduce the two - component beg model , and investigate it in the presence and absence of an external magnetic field .
the effect of a trapping potential is described in section [ sec : trap ] . in section [ sec : comp ] , we compare the results with magnetic superconductors and cold atom systems .
our conclusions are presented in section [ sec : conc ] .
the beg model was originally proposed to describe superfluidity.@xcite the phase diagram found by monte carlo simulations exhibits large similarities with the phase diagram of @xmath0he-@xmath1he mixtures measured by experimentalists.@xcite the main idea of studying superfluidity with the beg model relies on the @xmath3 symmetry - breaking of the ground - state wave function . for superconductivity and bose - einstein condensation
we have the same symmetry breaking , hence we can try to model these phenomena in the same way .
several physical systems exhibit two unequal symmetry broken phases simultaneously . a general hamiltonian describing this class of systems reads @xmath4 where @xmath5 can take the values @xmath6 , and @xmath7 .
this choice implies that only one kind of boson can occupy each lattice site .
@xmath8 is an anisotropy field that controls the number of lattice sites with nonzero @xmath9 .
@xmath10 plays the role of an external magnetic field , which may couple only to the order parameter describing a magnetic transition .
the hamiltonian is appropriate for describing phase transitions which require two order parameters , one describing the ordering of the fraction of the system with nonzero @xmath11 , the other one of nonzero @xmath12 .
this yields several possibilities , both fractions can model superfluidity , superconductivity , or ( anti)ferromagnetism .
possible applications could be magnetic superconductors , or two - component bose - einstein condensates . from now on , we will consider the fraction with nonnegative @xmath11 as describing magnetism , and @xmath12 superconductivity ( preformed bosons that can bose - einstein condensate ) .
thus , @xmath9 represents the spin of particle @xmath13 and @xmath14 the discretized phase of the wavefunction .
therefore , @xmath15 can be both positive ( ferromagnetism ) and negative ( antiferromagnetism ) , but @xmath16 has to be positive .
we define the concentration , the ferromagnetic , antiferromagnetic , and superconducting order parameters as @xmath17 note that @xmath18 and @xmath19 can reach a maximum value of @xmath20 , and @xmath21 of @xmath22 .
we define the ratio between the two coupling constants @xmath16 and @xmath15 as @xmath23 we investigate this model by monte carlo simulations . to determine the location of second - order phase transitions
, we performed simulations at constant concentration , in which the elementary moves were flips of @xmath14 and @xmath9 or nonlocal spin exchanges .
the location of the transition is then obtained from the peak location of the magnetic susceptibility .
the locations of first - order phase transitions are obtained from simulations at constant temperature , with as elementary moves local flips of @xmath14 and @xmath9 , as well as same - site replacements of @xmath14 by @xmath9 and vice versa . a jump in the concentration @xmath20 as a function of the anisotropy field @xmath8 is then the signature of the phase transition .
all simulations are performed on lattices with approximately @xmath24 sites . per point in the phase diagram
, simulations were run over @xmath25 to @xmath26 monte carlo steps per site , depending on the correlation times . in the absence of a magnetic field ,
the hamiltonian has ferro - antiferromagnetic symmetry .
first , we consider @xmath28 . in this case ,
@xmath29 , and the shape of the phase diagram must be symmetric under the transformation @xmath30 .
the results of the simulations are plotted in fig .
[ figure : h0k1 ] .
we see that it indeed obeys this symmetry and exhibits four phases : a superconducting phase ( s ) , where the order parameter @xmath21 is nonzero , a ferromagnetic phase ( fm ) , where @xmath18 is nonzero , a phase - separated regime ( ps ) where the spins and the angular phases have formed ordered clusters , and finally the normal phase ( n ) , in which there is neither order nor phase separation .
analogous to the beg model , the transition from the phase - separated regions to other phases are first - order ( dashed line ) , the other ones are second - order ( continuous line ) .
second , we consider the case @xmath31 .
the results of the simulations are plotted in fig .
[ figure : h0k01 ] .
we can understand the results as follows : @xmath16 is much smaller than @xmath15 , hence the spins will not pay much attention to the angular phases , and the part of the phase diagram concerning the spins will be very similar to the beg model . because @xmath16 is so small , the phases will only order at very low temperatures ( at zero concentration , the temperature is ten times lower than the one at which the spins order at a concentration of one ) . if the concentration is slightly raised from zero , the system is already in the phase separated regime .
all the states with a nonzero phase have clustered , and are not diluted by states with nonzero spin .
therefore , the critical temperature in the phase separated region will approximately remain constant . because the temperature at which the angular phases order is lower than the temperature at which phase separation begins
, there is a phase separated region in which the angular phases of the wavefunction are not ordered , which may appear unexpected at first sight .
the transition within the phase separated regime , from the region where the angular phases are not ordered to the phase where they are ordered ( superconductivity ) , is second - order .
this is expected , because in the phase separated regime , all the phases have clustered , and the transition will be comparable with the transition in the ising model , which is also second - order . if we apply a nonnegative uniform magnetic field to the system , the ferro - antiferromagnetic symmetry is broken .
we choose to consider the antiferromagnetic case here , because then there are two competing effects , the magnetic field tends to align the spins , whereas the exchange interaction wants to order the spins antiferromagnetically .
the magnetic field @xmath10 will be measured in units of @xmath15 .
kimel _ et al._@xcite have studied the antiferromagnetic beg model in the presence of a magnetic field , using monte carlo simulations .
their results at zero temperature suggest that the behavior of the system should be separated into three qualitatively distinct regions , namely @xmath32 , h \in[2,4]$ ] and @xmath33 $ ] .
we consider here the cases @xmath28 and @xmath31 for values of @xmath10 within each of these intervals .
first , we considered a magnetic field in the interval @xmath34 $ ] , namely @xmath35 . both for @xmath28 and @xmath31 ,
the results ( not shown ) are qualitatively the same as in the case of @xmath27 .
this behavior was expected from the phase diagram of the single - component beg model at zero temperature .
because the magnetic field tries to align the spins , the antiferromagnetic transition temperature is lower than in the absence of a magnetic field . in the usual beg model , the first - order phase transition disappears in the presence of a magnetic field @xmath36 $ ] . at zero temperature
, there is a second - order phase transition between a state with @xmath37 at every site , and a checkerboard phase , where one sublattice has @xmath37 at every site , and the other one @xmath38 .
there is also a transition between the checkerboard state , and an antiferromagnetic phase , but this transition is absent at nonzero temperature.@xcite for @xmath28 , the behavior of the two - component beg model is still very similar to the case @xmath27 . for @xmath31 ,
the first - order phase transition disappears , and therefore there is no phase - separated region , see fig .
[ figure : h25k01 ] .
we do observe an antiferromagnetic and a superconducting phase , but it is not clear from the figure whether the two phases overlap .
to better understand this low-@xmath39 intermediate regime , we also simulated the problem at a relative coupling strength of @xmath40 . in fig .
[ figure : h25k05 ] , we clearly observe that there is a region where antiferromagnetism and superconductivity coexist , _ without true phase separation _ , since the first - order phase transition has disappeared .
what is also interesting is that at zero temperature this region begins at a nonzero concentration , and ends at a concentration smaller than one .
when there is phase separation , this coexistence region always begins at @xmath41 and ends at @xmath42 . in the original beg model , when the magnetic field is increased to a value higher than @xmath43 at zero temperature , antiferromagnetism totally disappears because the spins tend to align with the magnetic field.@xcite the system is therefore magnetized , but not because of the nearest - neighbor interactions .
therefore , this is not really ferromagnetism , but for the sake of simplicity , we denote it like this . for the case of @xmath28 ,
we observe a phase with ferromagnetic and superconducting ordering , and a ferromagnetic phase ( not shown ) . for @xmath31 , we find another interesting phase , namely a ferromagnetic checkerboard phase , consisting of two sublattices , see fig .
[ figure : h5k01 ] . at the first sublattice , all sites
are randomly occupied by phases with a value of @xmath44 or @xmath45 .
at the second one , all sites are occupied by the spin that is favored by the magnetic field , @xmath38 .
this phase is most likely to occur at a concentration of @xmath46 because in this case a perfect checkerboard is possible .
because experiments with cold atoms are often carried out in a trap , we will add a harmonic potential to the original beg hamiltonian , to describe mixtures of fermions and bosons in a trap .
in general , the potential felt by the bosons is different from the one felt by the fermions , what implies that we must include two terms , @xmath47 here , @xmath48 and @xmath49 are the horizontal and vertical distances of site @xmath13 , measured from the center of the lattice , in lattice units , and @xmath50 and @xmath51 measure how much the bosons ( the states with @xmath52 ) , and the fermions ( the states with @xmath37 ) feel the influence of the trap . if @xmath53 , this term is constant , and the phase diagram is not modified .
we will consider the case @xmath54 , which is the most relevant experimentally .
using the hard core constraint @xmath55 , we can then rewrite this term and add it to the beg hamiltonian , thus obtaining @xmath56 where @xmath57 .
this means that , effectively , the bosons will feel a stronger tendency to go to the center of the trap . in the limit of @xmath58 ,
all the states with @xmath59 will cluster in the center of the trap , and therefore the ordering temperature will be the same as in the ising model .
note that the maximum value of the extra term in the hamiltonian will depend on the size of the lattice .
this way of including the trapping potential is comparable with the work of gygi _
et al._,@xcite where a spatial - dependent chemical potential was added to the bose - hubbard model in order to describe bosonic atoms in an optical lattice .
we simulated the new model using the same procedures as for the beg model and the two - component beg model .
the results for three different strengths of the trapping potential are plotted in fig .
[ figure : trap1beg ] . in the beg model without a trap , there is a second - order phase transition from a normal state to an ordered state , and a first - order phase transition to a phase - separated region@xcite .
for the three values of @xmath60 considered here , we do not find a first - order phase transition any more .
a part of the first - order phase transition line disappears , and a part changes into a second - order one .
we see that for a small difference between the trap potential felt by the bosons and the fermions , @xmath61 , the transition temperatures are very similar to the case without a trap . for a large difference , @xmath62 , the transition temperatures approach the transition temperature of the ising model for almost all concentrations , as expected .
when the states with @xmath59 are ordered , we will not speak of a superfluid state , but of a condensed state , because we now consider bosons in general .
it is important to estimate at which temperature the system starts to feel the influence of the trapping potential . let us assume that a cluster of size @xmath63 feels the potential when the energy difference between the state with this cluster in the center and in the corner of the lattice is of the order @xmath64 . for a lattice of size @xmath65
, this estimation results in @xmath66 in this approximation , a single particle ( @xmath67 ) in a lattice of size @xmath68 will start to feel the potential if @xmath69 . for @xmath62 and @xmath70
this results in @xmath71 and @xmath72 , respectively , in both cases much higher than the temperatures we are interested in , because ordering starts around @xmath73 .
therefore , the single particles will experience the influence of the trap in the entire temperature range of fig .
[ figure : trap1beg ] @xmath74 and @xmath75 . for @xmath61
, a single particle will feel the trapping potential for temperatures lower than @xmath76
. however , for higher temperatures the system already orders , and therefore there are some large clusters that according to eq .
( [ eq : est ] ) will feel the potential already at much higher temperatures .
this reasoning is in agreement with the snapshots in fig .
[ figure : trap1beg ] @xmath77 . for @xmath61 ,
we clearly observe the influence of the trap when the states with @xmath59 have clustered . in the disordered state ,
the influence is less visible . for @xmath62 and @xmath70
, we indeed see the influence of the trap for all temperatures , even in the disordered state .
analogous to the previous subsection , we will also add a trapping potential to the two - component beg model .
in the latter , both the states with @xmath59 and @xmath78 describe bosons , that both can condense .
therefore , this model can be applied to study cold atoms mixtures with two species of bosons .
we will consider the realistic case that the two species feel different trapping potentials .
therefore , we add the extra terms @xmath79 to the hamiltonian . because at every lattice site @xmath55
, we can rewrite this term and add it to the two - component beg hamiltonian , to get @xmath80 where @xmath81 is now the difference between the potentials felt by the two species of bosons .
now , the bosons with @xmath59 have a stronger tendency to go to the center of the lattice .
the results of our simulations are plotted in figs .
[ figure : trapbeg2k01 ] and [ figure : trapbeg2k1 ] .
we considered two different strengths of the trapping potential , and two different ratios of the coupling strengths of the bosons , namely @xmath82 and @xmath31 .
for @xmath31 , the right part of the first - order phase transition disappears and the left one becomes second - order ( see fig .
[ figure : trapbeg2k01 ] ) , whereas for @xmath28 both left and right parts of the first - order phase transition are converted into second - order ( see fig . [ figure : trapbeg2k1 ] ) . + in the limit of @xmath58 , all the sites with @xmath59 will have clustered in the center of the lattice , and all sites with @xmath78 at the corners .
therefore , for all concentrations , the system behaves as two uncoupled ising models . in the case of @xmath28 , we see indeed that the transition temperatures for both species approach the ising transition temperature . for @xmath31 , because @xmath16 is ten times smaller than @xmath15 , one of the species will order at the ising transition temperature , and the other one at one tenth of the ising transition temperature . to find the temperature at which the system starts to feel the presence of the trap
, we can make the same analysis as in subsection [ sec : trapbc ] . also here
, we see in the snapshots of figs .
[ figure : trapbeg2k01 ] and [ figure : trapbeg2k1 ] that for @xmath83 ( not shown ) and @xmath84 , the system always feels the influence of the trap , and for @xmath85 , it does only when the system is ordered . if we inspect fig .
[ figure : trapbeg2k1 ] @xmath77 , we see that there is a phase @xmath86 in which the bosons represented by @xmath78 are ordered , but the bosons represented by @xmath59 are not .
this is somewhat surprising .
a reason for the occurrence of this phase is that when all the bosons that have the tendency to go to the center of the trap have clustered there , automatically also the other bosons have clustered at the edge .
therefore , they can have nearest - neighbor interactions , and they can easily order .
it remains to see whether such a phase indeed occurs in experiments . from the theoretical point of view
, it would be interesting to also allow for states with @xmath87 , to verify the stability of this phase , when we relax the constraint that every lattice site must be occupied by one of the bosons .
note that for small enough concentrations , this phase will always occur , since the bosons @xmath12 are hardly diluted by the bosons @xmath11 .
there are several examples of cerium - based superconductors , for example cecoin@xmath88 and ceirin@xmath88 , as well as antiferromagnets that contain this element , like cerhin@xmath88 , cecocd@xmath88 , cerhcd@xmath88 , and ceircd@xmath88 .
let us consider cecoin@xmath88 and cecocd@xmath88 .
these two materials have two elements in common , ce and co , and differ in the third element . by doping cecoin@xmath88 with cd on the in site , we can change the superconductor cecoin@xmath88 into an antiferromagnet .
there are more of these cerium - based pairs , and therefore , this class of materials is appropriate for studying the interplay between superconductivity and magnetism .
let us consider the heavy fermion superconductor cecoin@xmath88 , with cadmium doping on the in - site .
this material has the highest superconducting transition temperature ( @xmath89 ) of all heavy fermions , and its electronic structure is quasi-2d.@xcite nicklas _
et al._@xcite and pham _
et al._@xcite determined the antiferromagnetic and superconducting onset temperatures of this material as a function of doping by elastic neutron scattering , specific heat , and resistivity measurements .
their results are plotted in fig .
[ figure : ceco ] . for experimental details
we refer the reader to ref . .
the phase diagram of ceco(in@xmath90cd@xmath91)@xmath88 shows three ordered phases : a superconducting phase , a commensurate antiferromagnetic phase , and a region where superconductivity and antiferromagnetism microscopically coexist .
it is interesting to observe that in this material antiferromagnetism suddenly disappears at the point where the onset temperatures for superconductivity and antiferromagnetism are equal .
this feature , however , may change in the presence of an applied magnetic field . in fig .
[ figure : artikel3 ] we see a schematic phase diagram of unconventional superconductors , in temperature - control parameter space . in the case of ceco(in@xmath90cd@xmath91)@xmath88
, the control parameter would be doping .
another example of such a parameter is pressure .
et al._@xcite determined the phase diagram of cerhin@xmath88 in temperature - pressure space with and without a magnetic field . without a magnetic field
, they also found this abrupt disappearance of the incommensurate antiferromagnetic order at @xmath92 .
however , when they applied a field of 33 koe , the line of the magnetic ordering temperature went smoothly down to zero at @xmath93 .
such a phase diagram shows many similarities with fig .
[ figure : h25k05 ] if we identify pressure with inverse concentration in our model .
indeed , for an external magnetic field of @xmath94 and a relative exchange strength @xmath40 ( see fig .
[ figure : h25k05 ] ) , the phase diagram shows the same three ordered phases .
further , the coexisting phase is not phase separated .
finally , we consider the compound ceir(in@xmath90cd@xmath91)@xmath88 , see fig .
[ figure : ceir ] and ref . .
for this material , it is not clear if there is a region where superconductivity and magnetism coexist .
if there is such a region , it is in a small doping interval .
the phase diagram of this material strongly resembles the phase diagram of the two - component beg model with an external magnetic field of @xmath94 and a relative coupling strength of @xmath31 , see fig .
[ figure : h25k01 ] .
although this experiment was also carried out without an external magnetic field , we only find similarities with our model in the presence of a magnetic field . in 2006 ,
two experimental groups , the group of ketterle at mit,@xcite and the group of hulet at rice university,@xcite have performed experiments with imbalanced ultracold @xmath95li atoms in a trap , and obtained contradictory results .
the mit group measured a transition between a normal and a superfluid phase at a polarization of @xmath96 , whereas the group at rice university observed a transition between two superfluid phases at @xmath97 . here
, @xmath98 measures the imbalance between the spin - up and the spin - down atoms , @xmath99 gubbels _ et al._@xcite have set up a theoretical model to describe these imbalanced fermi mixtures and determined a general phase diagram in temperature - polarization space that can explain the observations of both groups .
the topology of their phase diagram shows large similarities with the phase diagram of the beg model .
we can understand this resemblance as follows . in the beg model ,
the concentration @xmath20 is the fraction of lattice sites with @xmath37 , and thus the fraction of the system that can not condense .
the polarization @xmath98 is a measure for the difference of the atoms in the spin - up and the spin - down state , and thus for the number of fermions that remain after the others have paired .
the atoms with spin up and spin down will form pairs , and such a pair can be described as a boson . therefore , the polarization is also a measure for the fraction of the system that can not condense , and the concentration can be mapped onto the polarization .
we can identify the paired atoms , the preformed bosons , with the states @xmath52 , and the remaining fermions with @xmath37 , see fig .
[ figure : beg ] .
the experiments with @xmath95li are carried out in a trap , and the theoretical model of gubbels _ et al .
_ only includes the presence of the trap by using the local density approximation .
now , we would like to compare their phase diagram with our results of the beg model with a trapping potential , in figs .
[ figure : trapbeg2k01 ] and [ figure : trapbeg2k1 ] .
although in the case of imbalanced fermions the frequency @xmath100 of the optical trap felt by the pairs of fermions ( bosons ) and the remaining unpaired fermions is the same , the mass of the bosons is twice as large , and the potential constant @xmath50 is thus larger than @xmath51 .
this means that the comparison must be made with the beg model in a trap . in this model ,
the first - order phase transition , measured by a jump in the concentration as a function of the anisotropy field @xmath8 has disappeared , thus there is no true transition to a phase - separated regime . however ,
if we inspect the snapshots , we see that for low enough temperatures , or large enough trapping potential , there still is a clear separation between the condensed bosons and the fermions , suggesting some kind of effective phase separation .
we note that in experiments , phase separation is measured by inspecting the radii of the clouds of the atoms in the different hyperfine states , and not by a jump in some order parameter.@xcite our results thus suggest that the measured different radii are not per se an evidence of a true thermodynamic phase separation .
further experiments are required to clarify this issue .
although our model describes qualitatively the experimentally observed phases , it can not capture the fine details of recent experimental results .
studies by shin _
et al._@xcite indicate that there is no superfluid phase , or phase - separated phase for polarizations above @xmath101 . by a quantum monte carlo approach , lobo _
et al._@xcite predict a phase transition between a normal and a superfluid state at a polarization of @xmath102 at zero temperature , and gubbels and stoof@xcite recovered this results using a wilsonian renormalization group theory .
we simulated a two - component extension of the beg model without an external magnetic field and determined the phase diagram in the concentration - temperature space . in the region where magnetism and superconductivity coexist ,
the system is always phase separated .
we added a magnetic field to our model , and considered the antiferromagnetic case . in this case
, we also find phase diagrams with true coexistence of two ordered phases .
these diagrams are comparable with the phase diagram of doped heavy fermions in the presence of a magnetic field . in order to describe cold atom systems , we added a trapping potential to the beg model , and our extension of this model .
the added potential changes the phase separation regime conceptually .
we can not speak anymore about true phase separation , but more about a crossover to a phase separated region .
we argue that the beg model with a trapping potential can be used to model imbalanced fermi mixtures .
however , there are still quantitative differences with experiments , which our model is not able to cover .
we also made predictions for the phase diagram of boson - boson mixtures based on our simulations of the two - component beg model with a trapping potential .
although there is no available experimental data on boson - boson mixtures , we hope that our work can motivate further studies in this direction .
we are grateful to k. gubbels , h. van beijeren , a. de vries and r. movshovich for fruitful discussions and to j. de graaf for technical help .
|
we extend the blume - emery - griffiths ( beg ) model to a two - component beg model in order to study 2d systems with two order parameters , such as magnetic superconductors or two - component bose - einstein condensates .
the model is investigated using monte carlo simulations , and the temperature - concentration phase diagram is determined in the presence and absence of an external magnetic field .
this model exhibits a rich phase diagram , including a second - order transition to a phase where superconductivity and magnetism coexist .
results are compared with experiments on cerium - based heavy - fermion superconductors . to study cold atom mixtures
, we also simulate the beg and two - component beg models with a trapping potential . in the beg model with a trap
, there is no longer a first order transition to a true phase - separated regime , but a crossover to a kind of phase - separated region .
the relation with imbalanced fermi - mixtures is discussed .
we present the phase diagram of the two - component beg model with a trap , which can describe boson - boson mixtures of cold atoms .
although there are no experimental results yet for the latter , we hope that our predictions could help to stimulate future experiments in this direction .
|
in recent years , there was remarkable progress in the development of ( semi-)automatized software for the numerical calculation of arbitrary , complicated feynman integrals .
basically , two approaches are advocated .
one relies on sector decomposition . for an introduction given at this conference and for further references
see @xcite .
we will report on the other approach , based on mellin - barnes representations @xcite .
when @xcite appeared in 2005 , several unsolved problems of different complexity existed .
we mention non - planar diagrams , the massive case , multi - loop tensor integrals , minkowskian kinematics . for all of the items ,
progress is reported in @xcite , based on the source - open software ambre / mb @xcite .
an alternative is the direct analytical evaluation of mb - integrals .
this is difficult .
but in view of recent progress in algebraically summing up infinite sums by the linz group s computer algebra algorithms for nested sums and products , one might hope to achieve a breakthrough @xcite ; certainly only if the result leads to appropriate classes of functions .
the idea is to apply the linz group s algorithms ( as well as those of others , e.g. @xcite ) to sums of residues after applying cauchy s theorem @xcite .
a first attempt was reported in @xcite .
the automatized derivation of the multiple sums for a given mb - integral is certainly the easier part of the task , but it is the first step .
we report here on a first version of the mathematica program mbsums @xcite for transforming mb - representations for feynman integrals into multiple sums .
the licence conditions of the source - open package are those formlulated in the cpc non - profit use licence agreement of the computer physics communications program library @xcite .
the authors expect that the potential users read and follow the licence agreement when using this code .
the package ` mbsums ` transforms mellin - barnes integrals into sums , by closing the integration contours and calculating the integrals by the residue theorem , i.e. by constructing sums over all residues inside the contours .
the current version of ` mbsums ` is 1.0 .
the package ` mbsums ` works with wolfram mathematica 7.0 and later . in order to obtain a sum from an mb - integral
the user should use the ` mbinttosum ` function of ` mbsums ` : @xmath0|\ ] ] or @xmath1|,\ ] ] where ` int ` is the mb - integral in the form as it is denoted in the mathematica package ` mb ` @xcite : @xmath2|\ ] ] which corresponds to @xmath3 the integrand f can have the form @xmath4 and each @xmath5 is assumed to be of the form of @xmath6 where @xmath7 , @xmath8 , @xmath9 are free parameters ( usually kinematic parameters ) in ` int ` and @xmath10 is a factor independent of @xmath11-variables .
the @xmath12 where @xmath13 , @xmath14 , @xmath15 are rational numbers and @xmath16 ( ` eps ` ) is an infinitesimal dimensional shift , e.g. arising from @xmath17 .
all the singularities of the integrand of the mb - integral ` f ` are due ( and only due ) to ` gamma ` and ` polygamma ` functions .
the values of ` c1 , c2 , ... , cd ` are converted to rational numbers by ` mbinttosum ` before calculations .
let us now focus on the case when the list ` kinematics ` is empty , i.e. we will consider ( [ examplembgen2 ] ) .
the list ` contours ` has the form @xmath18 the order of the @xmath11-variables defines the order of integrations chosen by the user ( from left to right ) .
the l ( ) means that the contour will be closed to the left ( right ) .
the l/ choice made by the user can be changed if ` kinematics ` is not empty and this will be covered later .
the output of ` mbinttosum ` in ( [ examplembgen2 ] ) is of the form @xmath19 where @xmath20|\ ] ] represents a sum with summand ` sum_coefficient_i ` that has non - negative indices given in the list ` list_i ` , and ` conditions_i ` are conditions on those indices .
the complete answer is the sum of all ` mbsum_i ` in the list . the list `
kinematics ` has the form @xmath21 where ` r_i ` are free parameters ( usually kinematic parameters ) in ` int ` and ` v_i ` are values of ` r_i ` .
if ` kinematics ` is not empty then ` mbinttosum ` will try to change the l/ choice made by the user in the list ` contours ` in order to obtain sums that have good asymptotic behaviour at given values of ` r_i ` . in any case
the user is informed how the contour was closed .
this will be explained later in detail .
the values of ` v_i ` are converted to rational numbers by ` mbinttosum ` before calculations .
the user can turn off all messages printed by ` mbinttosum ` by typing ` mbsumsinfo = false ` and turn them on by typing ` mbsumsinfo = true ` .
in addition we provide function ` doallmbsums[sums , nmax , kinematics ] ` that sums the ` sums ` in the form ( [ sum ] ) .
the ` nmax ` is the maximal value of each index , the minimal value is given by conditions on indices .
the list ` kinematics ` is as above and may be empty .
we used wolfram mathematica function ` parallelmap ` inside ` doallmbsums ` to sum individual sums in the list ` sums ` in parallel .
let us now shortly explain how we obtain the sums .
we point out the most important ingredients in our algorithm .
let us focus on the case when the list ` kinematics ` is empty , i.e. we will consider ( [ examplembgen2 ] ) .
the mb - integral is in the form as it was denoted in ( [ int ] ) .
let us now assume that the user has chosen as first integration the ` z2->l ` .
as a first step we form a list , which we call negargsdoc , of arguments of the ` gamma ` and ` polygamma ` functions in the numerators that give residues for @xmath22 ( see ( [ int ] ) , the remaining contours are seen as a straight lines ) .
we call that list negargsdoc .
next w consider all possible cases : when all ` gamma ` and ` polygamma ` functions that have arguments in negargsdoc contribute to a residue at the same time , and when only some subset of them contributes to a residue at the same time .
we consider all possible subsets of negargsdoc .
additionally we have to be careful when some ` gamma ` functions in the denominator become singular at some points . if we have terms like ` gamma[2 z2 ]
` , then the poles are at @xmath23 and we consider there 2 cases : @xmath24 and @xmath25 , where @xmath26 and @xmath27 are non - negative integers
. similarly we proceed with arbitrary @xmath28 terms , where @xmath29 is some integer value and in general with all @xmath30 terms which appear together with integration variable in the arguments of the ` gamma ` and ` polygamma ` functions .
so we produce a list of cases @xmath31 where @xmath32 are expressions after taking residues of f and @xmath33 are conditions on the index that numerates terms ( residues ) in @xmath32 .
we obtain a list of @xmath34 elements after integrating over z2 .
let us now assume that the user has chosen as second integration variable ` z5->r ` .
then we repeat the whole procedure on each @xmath32 taking into account conditions @xmath33 .
so we produce analogous to ( [ a ] ) a list of cases @xmath35 we repeat the whole procedure for each integration variable .
let us now shortly explain how we obtain the sums if the list ` kinematics ` is not empty .
we follow the order of integration given in the list ` contours ` . our aim is to determine the l/ such that we obtain sums that have good asymptotic behaviour at given values of ` r_i ` in the list `
kinematics ` .
we do it in the following way . at each integration step
@xmath36 we analyse the expressions @xmath37 that are to be integrated over some @xmath38 .
each @xmath39 we decompose as @xmath40 and each @xmath41 is of the form @xmath42 where @xmath9 are the kinematic parameters in the list ` kinematics ` and @xmath43 contains the rest of @xmath44 .
if we integrate over @xmath38 , we consider @xmath45 the value of @xmath46 is calculated . `
mbinttosum ` prints the error message : @xmath47 for each @xmath48 in each @xmath37 when @xmath46 is symbolic ( not a number ) and at the end ` mbinttosum ` prints @xmath49 and returns \{}. the user should complete the list ` kinematics ` .
for each @xmath48 in each @xmath37 it is returned l if @xmath50 or r if @xmath51 indicating how to close the contour or \ { } if @xmath52 . if for each @xmath48 in each @xmath37 it is returneded l ( r ) or \ { } then the contour for @xmath38
will be closed to left ( right ) if it is returned at least one l ( r ) .
if for each @xmath48 in each @xmath39 it is returned \ { } then the choice of user given in the list ` contours ` is taken .
if for some @xmath48 it is returned r and for some @xmath53 it is returned l then ` mbinttosum ` prints the error message : @xmath49 and returns \{}. otherwise we compute the sums as described above .
we repeat the whole procedure for each integration variable .
we stress that this procedure as described above does not always give convergent sums .
there are mb - integrals for which no convergent sums can be found .
one such example is the following mb - integral : @xmath54 here we can apply the first barnes lemma @xcite and obtain @xmath55 , but the reader can check that indeed the infinite series of residues diverge both for @xmath56 and @xmath57 . consider the following integral ( see also @xcite ) : @xmath58 closing the contour to the right ( @xmath56 ) gives the following series @xmath59 convergent for @xmath60 , while closing the contour to the left ( @xmath57 ) gives the following series @xmath61 convergent for @xmath62 .
both @xmath63 and @xmath64 give the same formula after summing up , that is @xmath65 so @xmath66 and @xmath67 .
we also provide function ` simplifymbsums[sums ] ` that takes each ` mbsum [ ] ` in the list ` sums ` in the form ( [ sum ] ) and simplifies it .
the output of ` simplifymbsums ` is of the form ( [ sum ] ) .
if ` mbsum[f , c , l ] ` is the sum to be simplified then ` simplifymbsums ` uses the mathematica function ` reduce ` to simplify the conditions ` c ` .
if ` reduce ` returns after applying mathematica function ` logicalexpand ` to ` c ` for ` c ` an answer in the form ` c_1||c_2|| ... ||c_r ` , then we decompose ` mbsum[f , c , l ] ` into @xmath68,mbsum[f , c_2,l], ... ,mbsum[f , c_r , l]}|\ ] ] .
if some ` mbsum[f , c_i , l ] ` contains conditions ` c_i ` like .... n1 = = c[1 ] & & n2 = = c[1 ] + c[2 ] .... where ` n1 ` , ` n2 ` are summation indices in ` mbsum[f , c_i , l ] `
, then we make the substitutions ` n1 - > c[1 ] ` and ` n2 - > c[1 ] + c[2 ] ` in ` f ` and update ` c_i ` and ` l ` in ` mbsum[f , c_i , l ] ` ; ` c[1 ] ` and ` c[2 ] ` are new summation indices .
we assume that ` c [ ] ` are integers .
next , we unify the names of indices . finally , we combine sums having the same conditions . it can happen e.g. that we obtain 4-dimensional sums from a 3-dimensional sum after applying ` simplifymbsums ` . to illustrate this ,
let us consider the following example .
let ` c ` be .... c = n3 > = 0 & & n2 > = 1 & & n1 > = 0 & & n2 > = 1 + n3 & & 1 + n1
< = n2 .... then .... reduce[c , { n1 , n2 , n3 } , integers ] .... gives .... element[c[1 ] | c[2 ] | c[3 ] | c[4 ] , integers ] & & c[1 ] > = 0 & & c[2 ] > = 0 & & c[3 ] > = 0 & & c[4 ] > = 0 & & n1 = = c[1 ] + c[2 ] & & n2 = = 1 + c[1 ] + c[2 ] + c[3 ] + c[4 ] & & n3 = = c[2 ] + c[3 ] .... in addition we provide function ` mergembsums[sums ] ` that only combines sums in the list ` sums ` ( in the form ( [ sum ] ) ) having the same conditions .
in all the examples we will proceed as follows .
we start with feynman integral @xmath69 in @xmath17 dimensions .
first , the mb - representation for @xmath69 is obtained using the mathematica package ` ambre ` v.1.2 @xcite .
next the analytical continuation @xmath70 as well as the expansion of the integrals around @xmath71 is obtained using the mathematica package ` mb ` v.1.2 @xcite . in some examples
, we also used the mathematica package ` barnesroutines ` v.1.1.1 by david a. kosower @xcite in order to reduce the dimensionality of the mb - integrals .
the first example is the massive qed one - loop box diagram with two photons in the @xmath72-channel ; see the example file ` 1lbox.nb ` @xcite .
the input integral is , with kinematics @xmath73 for @xmath74 and @xmath75 : @xmath76}.\ ] ] using ` ambre ` one obtains the following mb - representation for finite @xmath16 : @xmath77 with @xmath78 , @xmath79 , @xmath80 , @xmath81 .
the mb - representation ( [ box ] ) is the same , up to irrelevant factors , as that in @xcite .
see also the sums there , compare also with the mb - representation in ` example3.nb ` @xcite , described in @xcite .
next one performs the analytical continuation of ( [ box ] ) for @xmath70 and the expansion of the resulting integrals around @xmath71 up to @xmath82 using ` mb ` : .... { mbint[((-x)^z1*y^(-1 - z1)*gamma[-z1]^3*gamma[1 + z1 ] * ( -6 + 3*eps^2*eulergamma^2 + 4*eps^2*pi^2 + 6*eps*log[-t ] - 3*eps^2*log[-t]^2 - 6*eps^2*eulergamma*log[y ] + 3*eps^2*log[y]^2 + 12*eps^2*polygamma[0 , -2*z1]^2 + 12*eps^2*polygamma[0 ,
-z1]^2 + 6*eps^2*eulergamma * polygamma[0 , 1 + z1 ] - 6*eps^2*log[y ] * polygamma[0 , 1 + z1 ] + 3*eps^2*polygamma[0 , 1 + z1]^ 2 - 12*eps^2*polygamma[0 , -z1]*(eulergamma - log[y ] + polygamma[0 , 1 + z1 ] ) + 12*eps^2*polygamma[0 , -2*z1]*(eulergamma - log[y ] - 2*polygamma[0 , -z1 ] + polygamma[0 , 1 + z1 ] ) - 12*eps^2*polygamma[1 , -2*z1 ] + 6*eps^2*polygamma[1 , -z1 ] + 3*eps^2*polygamma[1 , 1 + z1]))/ ( 6*eps*t^2*gamma[-2*z1 ] ) , { { eps - > 0 } , { z1 - > -1/2 } } ] , mbint[(-2*eps*(-x)^z1*y^z2*gamma[-z1 ] * gamma[-1 - z1 - z2]^2*gamma[-z2]*gamma[1 + z2]^2 * gamma[2*(1 + z1 + z2)]*gamma[2 + z1 + z2])/ ( t^2*gamma[2 + 2*z2 ] ) , { { eps - > 0 } , { z1 - > -1/2 , z2 - > -1/4 } } ] } .... the expression agrees numerically with the mb - representations in eq . ( 4.25 ) of @xcite and in eq . ( 4.67 ) of @xcite after analytical continuation and expansion in @xmath16 by ` mb ` .
we would now like to reproduce the representation by sums of the mb - integrals above .
we aim at sums to be convergent at @xmath83 we start with the first , one - dimensional integral on the list , which we denote by ` dim1int `
. then @xmath84|\ ] ] gives .... { mbsum[-(y^n1*n1!^2*(-4 + 3*eps^2*pi^2 + 2*eps^2*harmonicnumber[n1]^2 - 2*eps^2*harmonicnumber[n1 , 2 ] + 4*eps*log[-t ] - 2*eps^2*log[-t]^2 - 4*eps^2*harmonicnumber[n1 ] * log[-x ] + 2*eps^2*log[-x]^2))/(4*(-1)^n1*eps*t^2 * ( -x)^n1*x*(1 + 2*n1 ) ! ) , n1 > = 0 , { n1 } ] } .... together with the message how the contour was closed : @xmath85 .
here , the original choice how to close the contour is unchanged . the user can check that the same result can be obtained with ` dim1sum = mbinttosum[dim1int , { } , { z1 - > l } ] ` .
one can now check the result numerically : @xmath86 //
n|\ ] ] gives .... 0.222878 - 0.0967945/eps + 0.709316 eps .... in agreement with .... { 0.222878 - 0.0967945/eps + 0.709316 eps , 0 .
} .... from ` mbintegrate[{dim1int } , lk]//n ` .
next , we do the two - dimensional integral on the list , which we denote by ` dim2int ` . then ....
dim2sum = mbinttosum[dim2int , lk , { z1 - > l , z2 - > l } ] .... gives .... { mbsum[-(eps*(-x)^(n1 - n2)*y^n2*(-1 + 2*n1)!*n2 ! * ( pi^2 + 6*harmonicnumber[n1]*harmonicnumber [ -1 + n1 - n2 ] - 12*harmonicnumber[-1 + 2*n1 ] * harmonicnumber[-1 + n1 - n2 ] + 3*harmonicnumber[-1 + n1 - n2]^2 + 6*harmonicnumber[n1]*harmonicnumber[n2 ] - 12*harmonicnumber[-1 + 2*n1]*harmonicnumber[n2 ] - 3*harmonicnumber[n2]^2 - 12*harmonicnumber[n1 ] * harmonicnumber[1 + 2*n2 ] + 24*harmonicnumber[-1 + 2*n1]*harmonicnumber [ 1 + 2*n2 ] + 12*harmonicnumber[n2 ] * harmonicnumber[1 + 2*n2 ] - 12*harmonicnumber[1 + 2*n2]^2 + 3*harmonicnumber[-1 + n1 - n2 , 2 ] + 3*harmonicnumber[n2 , 2 ] - 12*harmonicnumber[1 + 2*n2 , 2 ] - 6*harmonicnumber[n1]*log[-x ] + 12*harmonicnumber[-1 + 2*n1]*log[-x ] - 6*harmonicnumber[-1 + n1 - n2]*log[-x ] + 3*log[-x]^2 + 6*harmonicnumber[n1]*log[y ]
- 12*harmonicnumber[-1 + 2*n1]*log[y ] - 6*harmonicnumber[n2]*log[y ] + 12*harmonicnumber[1 + 2*n2]*log[y ] - 3*log[y]^2))/ ( 3*(-1)^(3*n1)*t^2*x*n1!*(-1 + n1 - n2)!*(1 + 2*n2 ) ! ) , 2*n1 > = 1 & & n2 > = 0 & & n1 > = 1 + n2 , { n1 , n2 } ] , mbsum[(2*(-1)^(-2*n1 - n2)*eps*(-x)^(n1 - n2)*y^n2 * ( -1 + 2*n1)!*n2!*(-n1 + n2)!*(harmonicnumber[n1 ] - 2*harmonicnumber[-1 + 2*n1 ] + harmonicnumber [ -n1 + n2 ] - log[-x]))/(t^2*x*n1!*(1 + 2*n2 ) ! ) , 2*n1 > = 1 & & n2 > = 0 & & n1 < = n2 , { n1 , n2 } ] , mbsum[(eps*(-x)^n1*y^(-1 - n1 + n2)*(-1 - n1 + n2 ) ! * ( -1 + 2*n2)!*(pi^2 + harmonicnumber[n2]^2 - 2*harmonicnumber[n2]*harmonicnumber[-1 - n1 + n2 ] + harmonicnumber[-1 - n1 + n2]^2 - 4*harmonicnumber[n2 ] * harmonicnumber[-1 + 2*n2 ] + 4*harmonicnumber[-1 - n1 + n2]*harmonicnumber [ -1 + 2*n2 ] + 4*harmonicnumber[-1 + 2*n2]^2 + 4*harmonicnumber[n2]*harmonicnumber [ -1 - 2*n1 + 2*n2 ] - 4*harmonicnumber[-1 - n1 + n2 ] * harmonicnumber[-1 - 2*n1 + 2*n2 ] - 8*harmonicnumber[-1 + 2*n2]*harmonicnumber [ -1 - 2*n1 + 2*n2 ] + 4*harmonicnumber[-1 - 2*n1 + 2*n2]^2 + harmonicnumber[n2 , 2 ] - harmonicnumber[-1 - n1 + n2 , 2 ] - 4*harmonicnumber[-1 + 2*n2 , 2 ] + 4*harmonicnumber[-1 - 2*n1 + 2*n2 , 2 ] - 2*harmonicnumber[n2]*log[y ] + 2*harmonicnumber[-1 - n1 + n2]*log[y ] + 4*harmonicnumber[-1 + 2*n2]*log[y ] - 4*harmonicnumber[-1 - 2*n1 + 2*n2]*log[y ] + log[y]^2))/ ( ( -1)^(3*n2)*t^2*n1!*n2!*(-1 - 2*n1 + 2*n2 ) ! ) , n1 > = 0 & & 2*n2 > = 1 & & 1 + n1 < = n2 , { n1 , n2 } ] } .... together with the message how the contour was closed : .... z1->r ( re z1 > -1/2 ) z2->r ( re z2 > -1/4 ) .... here the original choice how to close the contours was changed .
we can now check the result numerically : @xmath87 //
n|\ ] ] gives .... -0.0917188 eps .... in agreement with .... { -0.0917189 eps , { 8.79914 * 10 ^ -6 eps , 0 } } .... from ` mbintegrate[{dim2int } , lk]//n ` .
further , we give examples of two potential errors .
the .... dim2sum = mbinttosum[dim2int , lk , { z2 - > l , z1 - > l } ] .... prints ....
z2->r ( re z2 > -1/4 ) unable to found correct contour for z1 . .... and returns ` { } ` , while .... dim2sum = mbinttosum[dim2int , { t - > -10 } , { z1 - > l , z2 - > l } ] .... prints .... found c = -x ( not a number ) : please complete kinematic 's list .
unable to find correct contour for z1 . .... and returns ` { } ` .
the next example is the massive on - shell planar double box diagram ; see the example file ` 2lbox.nb ` at @xcite .
the integral was studied in @xcite .
our kinematics is @xmath73 for @xmath74 , @xmath75 ) : @xmath88 } \nonumber\\ & & \times \frac{1}{(k_{1}-k_{2})^2(k_{2}^2-m^2 ) [ ( k_{2}+p_{1}+p_{2})^2-m^2 ] ( k_{1}-p_{3})^2 } .\end{aligned}\ ] ] the mb - representation of ( [ i2l ] ) agrees with @xcite .
further , we will use the mb - representation ( [ i2l ] ) in ` example7.nb ` from @xcite in variable ` fin ` without the @xmath89 factor , which will be added later .
next the analytical continuation of ` fin ` in @xmath70 as well as the expansion of the resulting integrals around @xmath71 up to @xmath90 is obtained with ` mb ` v.1.2 .
: .... int1 = mbint[((-x)^(z1 + z6)*gamma[-z1]^3*gamma[1 + z1 ] * gamma[-z6]^3*gamma[1 + z6])/(2*eps^2*s^2*t*gamma[-2*z1 ] * gamma[-2*z6 ] ) , { { eps - > 0 } , { z1 - > -1/2 , z6 - > -1/2 } } ] .... with @xmath91 , @xmath80 , @xmath81 .
we now produce sums from the mb - integral above , that are convergent at @xmath92 then @xmath93|\ ] ] gives .... { mbsum[((-1)^(-n1 - n2)*(-x)^(-n1 - n2)*n1!^2*n2!^2)/ ( 2*eps^2*s^2*t*x^2*(1 + 2*n1)!*(1 + 2*n2 ) ! ) ,
n1 > = 0 & & n2 > = 0 , { n1 , n2 } ] } .... together with the message how the contour was closed ( the original choice is unchanged ) : .... z1->l ( re z1 <
-1/2 ) z6->l ( re z6 <
-1/2 ) .... the numerical check : @xmath94 //
n|\ ] ] gives .... -4.68459/eps^2 .... in agreement with .... { -4.6846/eps^2 , { 0.000453913/eps^2 , 0 } } .... from ` mbintegrate[{int1 } , lk ] ` and with the analytical result in @xcite .
the two - dimensional sum in ` s1 ` can be written as a squared one - dimensional and can be summed - up with mathematica 9.0 : .... ( 8 arcsin[1/(2 sqrt[x])]^2)/(eps^2 s^2 t ( -1 + 4 x ) ) .... in numerical agreement with the above .
next we consider the on - shell massless planar double box ( [ i2l ] ) with @xmath95 ; see @xcite .
our kinematics is @xmath96 for @xmath74 and @xmath97 : @xmath98 where @xmath99 , @xmath80 , @xmath81 ; see the example file 2lbox-m0.nb at @xcite .
the mb - representation for ( [ i2lm0 ] ) is obtained using ` ambre ` v.1.2 , based on the derivation of the mb - representation in ` example7.nb ` ( available at @xcite ) . from the four - dimensional mb - representation stored in variable ` fin ` we removed the @xmath100 factor and obtained @xmath101 , stored in variable ` kfin ` .
compare this with the five - dimensional mb - representation for @xmath101 in @xcite .
next , the analytical continuation of ` kfin ` at @xmath70 and the expansion of the resulting integrals around @xmath71 up to @xmath102 are obtained with ` mb ` v.1.2 .
finally , with ` barnesroutines ` v.1.1.1 @xcite one may reduce the dimensionality of the mb - integrals .
one obtains several one - dimensional mb - integrals .
next , these mb - integrals are transformed into one - dimensional sums .
we checked both the mb - integrals and sums to be in good numerical agreement at @xmath103 with the analytic result for @xmath101 in @xcite .
there were no sums ( because zero - dimensional mb - integrals ) to be done for the @xmath90-term , and we obtained the analytic result already .
the sums for the @xmath104-term and the @xmath102-term could be done with advanced tools by j. blmlein and c. schneider .
they agree analytically with @xmath101 in @xcite .
we would like to thank j. blmlein , i. dubovyk , j. gluza and c. schneider for numerous helpful discussions .
the research project is partly supported by the polish national science centre ( ncn ) under the grant agreement no .
dec-2013/11/b / st2/0402 .
v. a. smirnov , analytical result for dimensionally regularized massless on shell double box , phys .
b460 ( 1999 ) 397404 . http://arxiv.org/abs/hep-ph/9905323 [ ] , http://dx.doi.org/10.1016/s0370-2693(99)00777-7 [ ] .
j. tausk , nonplanar massless two loop feynman diagrams with four on - shell legs , phys .
b469 ( 1999 ) 225234 . http://arxiv.org/abs/hep-ph/9909506 [ ] , http://dx.doi.org/10.1016/s0370-2693(99)01277-0 [ ] .
a. smirnov , v. smirnov , on the resolution of singularities of multiple mellin - barnes integrals , eur .
j. c62 ( 2009 ) 445449 . http://arxiv.org/abs/0901.0386 [ ] , http://dx.doi.org/10.1140/epjc/s10052-009-1039-6 [ ] .
j. gluza , k. kajda , t. riemann , ambre - a mathematica package for the construction of mellin - barnes representations for feynman integrals , comput .
commun . 177
( 2007 ) 879893 .
http://arxiv.org/abs/0704.2423 [ ] , http://dx.doi.org/10.1016/j.cpc.2007.07.001 [ ] .
j. gluza , k. kajda , t. riemann , v. yundin , numerical evaluation of tensor feynman integrals in euclidean kinematics , eur .
j. c71 ( 2011 ) 1516 .
[ ] , http://dx.doi.org/10.1140/epjc/s10052-010-1516-y [ ] .
i. dubovyk , j. gluza , t. riemann , non - planar feynman diagrams and mellin - barnes representations with ambre v.3.0 , j. phys .
608 ( 1 ) ( 2015 ) 012070 . http://dx.doi.org/10.1088/1742-6596/608/1/012070 [ ] .
s. moch , p. uwer , xsummer : transcendental functions and symbolic summation in form , comput .
174 ( 2006 ) 759770 . http://arxiv.org/abs/math-ph/0508008 [ ] , http://dx.doi.org/10.1016/j.cpc.2005.12.014 [ ] .
j. gluza and t. riemann , `` simple feynman diagrams and simple sums '' , talk held at the risc - desy workshop on advanced summation techniques and their applications in quantum field theory , on the occasion of the 5th year jubilee of the risc - desy cooperation , may 7 - 8 , 2012 , risc institute , castle of hagenberg , linz , austria .
conference link : http://www.risc.jku.at/conferences/riscdesy12/ , talk link : http://www-zeuthen.desy.de/~riemann/talks/riemann-linz-2012-05.pdf [ http://www-zeuthen.desy.de/@xmath105riemann/talks/riemann-linz-2012-05.pdf ] .
j. blmlein , i. dubovyk , j. gluza , m. ochman , c. g. raab , t. riemann , c. schneider , non - planar feynman integrals , mellin - barnes representations , multiple sums , pos ll2014 ( 2014 ) 052 . http://arxiv.org/abs/1407.7832 [ ] .
j. gluza and t. riemann , _ simple feynman diagrams and simple sums _ , talk held at risc - desy _ workshop on advanced summation techniques and their applications in quantum field theory _ on the occasion of the 5th year jubilee of the risc - desy cooperation , may 7 - 8 , 2012 , risc institute , castle of hagenberg , linz , austria .
conference link : http://www.risc.jku.at/conferences/riscdesy12/ , talk link : http://www-zeuthen.desy.de/~riemann/talks/riemann-linz-2012-05.pdf [ http://www-zeuthen.desy.de/@xmath105riemann/talks/riemann-linz-2012-05.pdf ] .
j. fleischer , j. gluza , a. lorca , t. riemann , first order radiative corrections to bhabha scattering in d dimensions , eur .
c48 ( 2006 ) 3552 . http://arxiv.org/abs/hep-ph/0606210 [ ] , http://dx.doi.org/10.1140/epjc/s10052-006-0008-6 [ ] .
|
feynman integrals may be represented by the mathematica packages ` ambre ` and ` mb ` as multiple mellin - barnes integrals . with the mathematica package ` mbsums ` we transform these mellin - barnes integrals into multiple sums .
|
here we provide a proof of eq .
( [ 2nd - projection - th ] ) and an asymptotic ( large @xmath34 ) estimate of the constant @xmath148 let @xmath149 be the spectral projection of @xmath150 associated with the zero eigenvalue of @xmath13 . since , because of the lindblad structure , there is no nilpotent term associated with the zero eigenvalue , the perturbation theory reads , as shown in in t. kato , _ perturbation theory for linear operators _ , for small @xmath151 i.e. , large @xmath152 @xmath153 from the first equation it now follows ( for sufficiently large @xmath34 ) @xmath154 where @xmath155 is a suitable constant ( notice that @xmath156 ) . on the other hand , using @xmath157 and the definition @xmath158 for the dimensionful effective generator from the last equation in ( [ kato - pert ] ) it follows @xmath159 whence ( for small @xmath160 ) @xmath161 since @xmath162 one can write @xmath163 using @xmath164 with @xmath165 and @xmath166 and the bounds above it also follows that , for @xmath167 , @xmath168 where @xmath169 is a constant of that , for dimensional reasons , is @xmath170 i.e. , @xmath171 from ( [ triangle1 ] ) using @xmath172 and standard operator norm inequalities one finds @xmath173 notice that @xmath174 is the quantity showing up in the lhs of ( [ 2nd - projection - th ] ) namely is the quantity whose upper bound over @xmath175 $ ] we desire to show is @xmath105 now using the bounds ( [ ineq0 ] ) , ( [ ineq1]),([ineq2 ] ) and @xmath176 one finds @xmath177 where @xmath178 by moving to the dimensionless hamiltonian such that @xmath179 the inequality ( [ ineq3 ] ) becomes @xmath180 notice that the requirement of @xmath160 being sufficiently small used repeatedly in the above translate now in the adiabatic criterion " of @xmath34 being sufficiently large . finally by taking the supremum for @xmath181 $ ] one obtains @xmath182 setting @xmath183 completes the proof of ( [ 2nd - projection - th ] ) .
we directly compute the second order effective generator @xmath46 by acting on some state @xmath184 , such that @xmath185 .
one has @xmath186 where we have introduced notation @xmath187 , which only acts non - trivially on system @xmath63 for this set - up , i.e. @xmath188 ( following from @xmath189 , see @xcite ) .
acting with @xmath190 on this we can see that : @xmath191 where @xmath192 , and @xmath193 . in passing we notice that one can rewrite the system @xmath4 part of these equations in a more familiar form , using that without loss of generality @xmath194 and @xmath195 .
just observe that since @xmath76 is a hermitian - preserving map , we have @xmath196 . eq .
( [ 2nd - order - general-2 ] ) then follows as required .
before concluding , we would like to show how the construction leading to the effective dynamics ( [ 2nd - projection - th ] ) can in principle be iterated over a sequence of exponentially longer time - scales .
let us set the error parameter @xmath197 .
the effective relaxation time of ( [ 2nd - projection - th ] ) can be roughly estimated as @xmath198 , where the last inequality stems from the condition @xmath199 suppose that the dynamics generated by @xmath200 admits itself a high - dimensional sss ( let @xmath201 denote the associated projection ) and that one can switch on an _ extra _ hamiltonian @xmath202 such that @xmath203 one can now apply the projection theorem ( [ 2nd - projection - th ] ) to @xmath204 and @xmath202 and argue that the effective dynamics in the sss of @xmath204 is ruled by @xmath205 if even this effective hamiltonian vanishes then one can _ iterate _ the projection procedure assuming as a starting state - space the sss of @xmath206 in general , if at the @xmath207-th level one finds @xmath208 and @xmath209 is high - dimensional then one can move to the next level where @xmath210 and a new hamiltonian @xmath211 such that @xmath212 is introduced .
reasoning as in the above one can show that at each iteration the relaxation time scale ( hamiltonian norm ) gets stretched ( compressed ) by a factor @xmath213 ( @xmath214 ) .
from the point of view of potential applications the interest in exploring this projection hierarchy rests on the possibility that the first non - vanishing effective hamiltonian has some desired property e.g. , higher - locality @xcite .
|
it is by now well understood that quantum dissipative processes can be harnessed and turned into a resource for quantum - information processing tasks . in this paper
we demonstrate yet another way in which this is true by providing a dissipation - assisted protocol for the simulation of general markovian dynamics .
more precisely , we show how a suitable coherent coupling of a quantum system to a set of markovian dissipating qubits allows one to enact an effective liouvillian generator of any lindbladian form .
this effective dynamical generator arises from high - order virtual - dissipative processes and governs the system dynamics exactly in the limit of infinitely fast dissipation .
applications to the simulation of collective decoherence are discussed as an illustration .
quantum decoherence and dissipation have been regarded until recently purely detrimental to the aim of quantum information processing ( qip ) @xcite .
interactions with the environment in fact inevitably lead to entanglement between the quantum computing system and uncontrollable degrees of freedom .
this unwanted entanglement in turn results in a system subdynamics that is in general incoherent and irreversible : unitarity is quickly lost and with it the quantum information processing advantages e.g. , computational speed - ups , one was seeking for
. this state of affairs triggered a spectacular theoretical effort that led to the discovery of a host of techniques to tame decoherence @xcite as quantum error correction @xcite , decoherence - free subspaces @xcite , noiseless subsystems @xcite and holonomic quantum computation @xcite .
it is therefore a conceptually remarkable shift the recent realization that by reservoir - engineering dissipation can be harnessed and turned into a useful practical resource for qip ( see @xcite for an early pioneering insight ) .
for example one can dissipatively achieve quantum state preparation @xcite , quantum simulation @xcite , holonomic quantum computation @xcite and even universal computation @xcite .
simulation of highly non - trivial properties of matter as topological order @xcite and non - abelian synthetic gauge fields @xcite can also be accomplished by dissipative means . finally , all forms of qip that encode information in the ground state of a time - dependent hamiltonian , e.g. , open system adiabatic quantum computation and quantum annealing , also benefit from dissipation and relaxation to negate thermally driven errors @xcite . in particular in @xcite it has been shown that quantum information can be encoded in the set of steady states ( sss ) of a sufficiently symmetric strongly dissipative system and manipulated coherently by an effective dissipation - projected hamiltonian .
the latter is of geometric nature and is robust against some types of hamiltonian and dissipative perturbations @xcite .
the key idea of ref .
@xcite is a simple one : once the system is prepared in the sss the fast dissipative processes adiabatically decouple non steady - states away while at the same time strongly renormalize the system hamiltonian in such a way that the sss remains invariant under this projected dynamics .
this phenomenon can be thought of as a sort of environment - induced quantum zeno effect @xcite at the superoperator space level @xcite .
( blue ball ) is coupled with coupling strengths @xmath0 to @xmath1 qubits ( yellow balls ) .
each of these qubits is subject to amplitude damping with rates @xmath2 .
proposition 2 shows that in the limit of small @xmath3 the qubits can be adiabatically decoupled and the effective dynamics of @xmath4 is described by @xmath1 lindblad operators of strength @xmath5 [ 2nd - order],width=302 ] in this paper we extend the ideas of @xcite to higher order . in the case
in which the dissipation - projected hamiltonian is vanishing , higher order virtual dissipative processes give rise , in a suitable limit , to an effective liouvillian generator that leaves the sss invariant .
however , at variance with the case studied in @xcite this effective generator is no longer hamiltonian : a slow irreversible process unfolds within the sss .
we will show how this mechanism can be exploited to the end of the simulation of _ any _ markovian dynamics .
more precisely , we will show that by suitably coupling a quantum system to a structured reservoir comprising multiple qubits undergoing fast amplitude damping one can implement an effective liouvillian generator in general lindblad form @xcite .
we will illustrate our results by analyzing the dissipative simulation of qubits subject to collective decoherence . _
preliminaries. _ let @xmath6}$ ] denote the hilbert space of the system and @xmath7 the algebra of linear operators over it .
a time - independent liouvillian super - operator @xmath8 acting on l@xmath9 is given .
the sss of @xmath8 consists of all the quantum states @xmath10 ( @xmath11 ) contained in the kernel @xmath12 of @xmath13 .
we shall denote by @xmath14 ( @xmath15 ) the spectral projection over ker@xmath16 ( the complementary subspace of ker@xmath16 . ) as in @xcite the liouvillian @xmath8 is also assumed to be such that : * a ) * @xmath17 defines a semi - group of trace - preserving positive maps with @xmath18 , * b ) * the non - zero eigenvalues @xmath19 of @xmath8 have negative real parts , i.e. , the sss is attractive . in this case @xmath20 and @xmath21 we also denote by @xmath22 the reduced resolvent of @xmath13 at ( @xmath23 ) and by @xmath24 the latter provides a natural time - scale associated with the relaxation processes described by @xmath25 the energy scale @xmath26 is of the order of the dissipative gap of @xmath13 i.e. , the smallest modulus of a non - zero eigenvalue of @xmath25 the dimensionless ( and normalized ) resolvent is defined by @xmath27 we now add an hamiltonian term @xmath28 $ ] where @xmath29 such that @xmath30 .
we set @xmath31 in such a way that @xmath32 is dimensionless and @xmath33 the time - scale @xmath34 is our scaling parameter and has to be thought of as large or even infinite in the spirit of the adiabatic theorem .
we first establish * proposition 1 * : if @xmath35 then for sufficiently large @xmath34 one has that @xmath36}\| ( e^{t { \cal l}_t}-e^{\frac{t}{t}\tilde{{\cal l}}_{{\textrm{\textrm{eff}}}}}){\cal p}_{0}\|\le c_\theta \sqrt{\frac{\tau_r}{t } } \label{2nd - projection - th}\ ] ] where @xmath37 @xmath38 @xmath39 and @xmath40 depends on @xmath41 and @xmath25 _ proof. _ is provided in the appendix .
@xmath42 this result provides the starting point of this paper . in particular from eq .
( [ 2nd - projection - th ] ) it follows that @xmath43 in words : if the system is prepared at time @xmath44 inside the sss and then evolves for finite fraction @xmath41 of @xmath34 , in the large @xmath34 limit the time - evolution leaves the sss invariant and it is governed by the effective ( dimensionless ) generator @xmath45 .
[ it is also sometimes convenient to introduce the effective dimensionful generator @xmath46 whose norm is @xmath47 in terms of @xmath48 the second term in the norm of eq .
( [ 2nd - projection - th ] ) reads @xmath49 _ remarks : _ * 0 ) * the stronger the dissipation outside the sss i.e. , the shorter @xmath50 the _ weaker _ the effective one inside * i ) * since , by construction @xmath51 the ( dimensionless ) action associated to the effective propagator @xmath52 is @xmath53 for @xmath54 * ii ) * the rhs of eq . (
[ 2nd - projection - th ] ) represents an error bound , if we fix it at @xmath55 we see that one needs that @xmath56 * iii ) * if @xmath57 where @xmath58 and @xmath59 then ( [ 2nd - projection - th ] ) holds with @xmath60 and a different constant @xmath40 @xcite the effective liouvillian generator @xmath61 is clearly reminiscent of the second - order effective hamiltonians routinely used e.g. , in quantum optics , and obtained by some sort of adiabatic decoupling technique @xcite .
however , this dynamics , at variance with that case as well as with the situation considered in @xcite is not unitary but of general liouvillian type .
the key point is that this effective non - unitary dynamics depends on @xmath62 and on its non - trivial interplay with the bare dissipation generated by @xmath25 this opens up the possibility of using it to engineer dissipative systems with a desired liouvillian generator .
_ universal lindbladian simulation. _ let us consider a system @xmath4 coupled to a system @xmath63 via the general hamiltonian @xmath64 where the tensor ordering follows that of the total hilbert space , @xmath65 and , without loss of generality , we assume @xmath66 .
we also assume that the dissipative term is of the form @xmath67 , such that @xmath68 where @xmath69 is by assumption the unique steady state of @xmath70 .
the sss of @xmath13 is given by all the states of the form @xmath71 and it is isomorphic to the full - state space of @xmath72 in this case one has @xmath73 and @xmath74 $ ] with @xmath75@xcite .
let @xmath76 be the the projected resolvent of @xmath70 at @xmath77 * proposition 2 : * if @xmath78 then @xmath79 @xmath80 + \sum_{i , j=1}^m2\gamma_{ij}(l_i\rho l_j - \frac{1}{2}\{l_jl_i\,,\rho \ } ) \label{2nd - order - general-2}\ ] ] where @xmath81 , @xmath82 and @xmath83 _ proof. _ is provided in the appendix .
@xmath42 notice that @xmath84 and that eq .
( [ 2nd - order - general-2 ] ) describes a truly lindbladian dynamics iff @xmath85 our main result now follows as a particular case of proposition 2 above .
let us consider a @xmath86-dimensional system @xmath4 coupled to a system @xmath63 comprising @xmath1 qubits , by the hamiltonian @xmath87 where the @xmath88 s are given operators acting on the system state - space only .
let us also suppose @xmath89 where each of the @xmath1 qubits independently dissipates according to the local liouvillian @xmath90 the unique steady state of @xmath70 is @xmath91 and since @xmath92 ) one has @xmath93 * proposition 3 : * @xmath94 where @xmath95 _ proof. _ to obtain eq .
( [ l - simul ] ) from eq .
( [ 2nd - order - general-2 ] ) , re - write the @xmath96 in ( [ k - simul ] ) such that @xmath97 .
remembering that @xmath98 , and @xmath70 as in eq . , we recover eq . as required by direct evaluation of the matrix @xmath99 in prop .
2 @xmath42 + notice now that , in view of remark * iii ) * after prop . 2
, one can add any hamiltonian @xmath100 [ @xmath58 ] acting on the system @xmath4 only ( @xmath102 ) .
this will result in @xmath103 therefore we see that prop .
2 shows that in principle _ any _ liouvillian in the lindblad form @xcite
i.e. , the most general generator of semi - groups of markovian cp maps , can be obtained given the availability of @xmath1 auxiliary qubits ( one for each lindblad operator ) subject to an amplitude damping channel _ and _ the ability to enact the hamiltonian @xmath104 .
dissipation turns into a resource that allows one to _ simulate _ a general lindbladian evolution .
we would like to make a couple of remarks : _ 1 ) _ one might think of obtaining the lindbladian dynamics eq .
( [ l - simul ] ) _ directly _ coupling the system @xmath4 to some reservoir with an interaction hamiltonian of type ( [ k - simul ] ) and then using the standard born markov approximation @xcite .
the point is that the latter involves _ uncontrolled _ approximations ( markov ) whereas eq .
( [ 2nd - projection - th ] ) has a uniform and _ controlled _ error @xmath105 this means that the effective dynamics of @xmath4 becomes _ exactly _ lindbladian , with generator ( [ l - simul ] ) , for @xmath54 of course this is true as long as the auxiliary qubits are _ exactly _ described by the lindbladian in eq .
( [ qubit - damping ] ) i.e. , their genuine markovianity is a key resource in our universal simulation protocol along with the ability of switching on the the hamiltonian in eq .
( [ k - simul ] ) .
_ 2 ) _ in view of physical applications , we stress that the effective dynamics in eq .
( [ l - simul ] ) still holds if the @xmath1 qubits are replaced by @xmath1 bosonic modes subject to amplitude damping i.e. , the @xmath106 in eq .
( [ qubit - damping ] ) are replaced by annihilation operators @xmath107 . in the next section
we will discuss , for the sake of illustration of our general results , the simulation of different types of collective decoherence when @xmath4 is itself a set of multiple qubits .
_ simulating collective amplitude damping. _ here we use our general result eq .
( [ l - simul ] ) to simulate a qubit subject to collective damping .
this type of symmetric noise is interesting as it admits decoherence - free subspaces @xcite and can be used to dissipatively prepare entangled states .
let us consider a system of @xmath108 qubits coupled to a bosonic mode e.g. , @xmath108 atoms coupled to a cavity em mode , via a ( collective ) jaynes - cummings hamiltonian @xmath109 moreover we assume that the system dissipates according to the liouvillian @xmath110 where @xmath111 + \tau_r^{-1}(a\rho a^\dagger -\frac{1}{2}\{a^\dagger a,\,\rho\}).$ ] using eq .
( [ l - simul ] ) with @xmath112 one finds the effective generator @xmath113 where @xmath10 is just the qubit state as the generator is trivial in the bosonic degrees of freedom ( frozen at @xmath114 + proposition 2 shows that one can consider for the auxiliary qubits a liouvillian that is more general than eq .
( [ qubit - damping ] ) ( as long as its steady state is unique ) .
we illustrate this fact by considering the thermalization of an auxiliary qubit at non - zero temperature .
namely , we add to eq . an excitation liouvillian , such that now @xmath115 by explicit computation of the @xmath116 matrix in prop . 2
one can check that the new effective generator ( in the system @xmath4 sector ) is @xmath117 where @xmath118 . a numerical check of the validity of eq .
( [ 2nd - projection - th ] ) is shown in fig .
[ fig : linear - thermal ] .
note that one can add a hamiltonian term of the form @xmath119 in such a way that the unperturbed lindbladian becomes a thermalizing davies - type generator , for which the temperature is fixed by @xmath120 @xcite . in a similar fashion ,
using the remark * iii ) * above , we can add a properly rescaled hamiltonian term @xmath121 in such a way that @xmath122 , and @xmath123 is again of davies type . in other words , also the effective generator can be made thermal and so it defines a temperature according to @xmath124 for some energy scale @xmath125 . from @xmath126
it follows @xmath127 namely one has cooling or heating according to whether @xmath128 or vice - versa . ) and effective one with liouvillian ( [ thermal-2nd - order ] ) , as a function of @xmath129 .
@xmath130 , and @xmath131 ( where the relaxation time is @xmath132 .
the linear fit is obtained using the least squares fitting on all of the data points , and the norm is the maximum singular value of the maps realized as matrices.,width=302 ] _ simulating collective dephasing. _ following a similar set - up as the previous subsection but with a hamiltonian of the form @xmath133 , the effective generator becomes that of collective dephasing along the @xmath134-direction @xmath135 ) and effective one with liovillian eq .
( [ bit - flip ] ) , as a function of @xmath136 .
@xmath137 , and @xmath138 .
note that for the dashed line we have extended @xmath139 past @xmath34 , purely for convenience .
the norm is the maximum singular value of the maps realized as matrices.,width=302 ] in fig .
[ fig : eff - ideal ] we plot the distance between the actual and the effective evolution as a function of @xmath139 for different time - scales @xmath34 . according to eq .
( [ 2nd - projection - th ] ) by changing @xmath140 ( @xmath141 ) , we expect the distance to fall by a factor of @xmath142 ( cf . in fig .
[ fig : eff - ideal ] the maximum error falls from the dash to solid line by a factor of @xmath143 ) . in the limit of @xmath144 ,
the exact evolution becomes identical to that of the effective one for all times @xmath139 ( the actual evolution ` adiabatically follows ' the effective one ) . _
conclusions. _ there is increasing evidence that dissipative and quantum incoherent processes can be used to enact quantum information processing primitives , see e.g. @xcite . in this paper
we have shown how a suitable coherent coupling between a quantum system @xmath4 and an environment comprising multiple qubits subject to strong markovian dissipation , can be used to simulate universal lindbladian dynamics over @xmath72 more precisely , by using high - order virtual dissipative processes , one can build an effective liouvillian generator in arbitrary lindblad form @xcite that governs the dynamics of @xmath4 _ exactly _ in the limit of infinitely fast dissipation .
this approach has to be contrasted with the standard one in which lindbladian evolution arises from a weak hamiltonian coupling to a unitarily evolving environment and involves uncontrolled markovian approximations e.g. , born - markov factorization of the joint density matrix @xcite .
we illustrated our results by numerical simulations of concrete physical models .
our findings show that markovianity itself can be seen as resource in that it allows for universal simulation of an important class of quantum irreversible processes . _
acknowledgements. _ this work was partially supported by the aro muri grant no .
w911nf-11 - 1 - 0268
. _ note added .
_ after the completion of this work we became aware of ref .
@xcite where a different approach to simulation of markovian systems is pursued .
10 w. g. unruh .
maintaining coherence in quantum computers .
, 51:992 , 1995 .
d. aharonov and m. ben - or . .
in _ proceedings of focs _ , p. 46 , los alamitos , ca , 1996 .
ieee comput .
soc . press .
d.a .
lidar and t.a .
brun ( eds . ) . .
cambridge university press , 2013 .
p. w. shor .
scheme for reducing decoherence in quantum computer memory .
, 52:r2493 , 1995 .
d. gottesman .
class of quantum error - correcting codes saturating the quantum hamming bound .
, 54:1862 , 1996 .
p. zanardi and m. rasetti .
noiseless quantum codes . , 79:3306 , 1997 .
d. a. lidar , i. l. chuang , and k. b. whaley .
decoherence - free subspaces for quantum computation . , 81:2594 , 1998 . .
, 291:1013 , 2001 .
e. knill , r. laflamme , and l. viola .
theory of quantum error correction for general noise .
, 84:2525 , 2000 .
p. zanardi .
stabilizing quantum information . , 63:012301 , 2000 .
j. kempe , d. bacon , d. a. lidar , and k. b. whaley .
theory of decoherence - free fault - tolerant universal quantum computation .
, 63:042307 , 2001 .
p. zanardi and s. lloyd .
topological protection and quantum noiseless subsystems .
, 90:067902 , 2003 .
l. viola _
et al_. experimental realization of noiseless subsystems for quantum information processing .
, 293:2059 , 2001 .
p. zanardi and m. rasetti .
holonomic quantum computation .
, 264:94 , 1999 .
l. m. duan , j. i. cirac , and p. zoller .
geometric manipulation of trapped ions for quantum computation .
, 292:1695 , 2001 .
a. beige , d. braun , b. tregenna , and p. l. knight , _ quantum computing using dissipation to remain in a decoherence - free subspace _ , phys .
rev .
lett . * 85 * , 1762 ( 2000 ) .
b. kraus _ et al_. preparation of entangled states by quantum markov processes .
, 78(4):042307 , 10 2008 .
m. j. kastoryano , f. reiter , and a. s. srensen .
dissipative preparation of entanglement in optical cavities .
, 106:090502 , 2011 .
j.t .
barreiro _
et al_. an open - system quantum simulator with trapped ions .
, 470:486 , 2011 . victor v. albert , stefan krastanov , chao shen , ren - bao liu , robert j. schoelkopf , mazyar mirrahimi , michel h. devoret , liang jiang holonomic quantum computing with cat - codes arxiv:1503.00194 f. verstraete , m. m. wolf , and j. i. cirac .
quantum computation and quantum - state engineering driven by dissipation .
, 5:633 , 2009 .
c.e .
_ et al_. topology by dissipation .
, 15:085001 , 2013 .
k. stannigel _ et al_. constrained dynamics via the zeno effect in quantum simulation : implementing non - abelian lattice gauge theories with cold atoms .
, 112:120406 , 2014 .
a.m. childs , e. farhi , and j. preskill .
robustness of adiabatic quantum computation .
, 65:012322 , 2001 .
m. s. sarandy and d. a. lidar .
adiabatic quantum computation in open systems .
, 95:250503 , 2005 .
k. l pudenz , t. albash , and d. a lidar .
error - corrected quantum annealing with hundreds of qubits . , 5:02 , 2014 .
p. zanardi and l. campos venuti .
coherent quantum dynamics in steady - state manifolds of strongly dissipative systems .
phys .
rev .
lett . * 113 * , 240406 ( 2014 ) .
p. zanardi , l. campos venuti , _ geometry , robustness , and emerging unitarity in dissipation - projected dynamics _ ,
phys .
rev . a * 91 * , 052324 ( 2015 ) p. facchi and s. pascazio .
quantum zeno subspaces . , 89:080401 , 2002 . g. a. paz - silva , a. t. rezakhani , j. m. dominy , and d. a. lidar .
zeno effect for quantum computation and control .
, 108:080501 , 2012 .
g. lindblad , _ on the generators of quantum dynamical semigroups _ , commun .
math .
phys . * 48 * , 119 ( 1976 ) from the proof of prop . 1 reported in the appendix it easily follows that only the term @xmath145 is @xmath146 whereas all the other terms involving @xmath121 are at most @xmath147 r. alicki , _ quantum dynamical semigroups and applications _ , ( 2007 ) .
springer science & business media .
gardiner , p. zoller _ quantum noise _ , springer ryan sweke , ilya sinayskiy , denis bernard , francesco petruccione , universal simulation of markovian open quantum systems , arxiv:1503.05028 ( 2015 ) .
|
spin dependent transport in nanostructures has attracted a lot of interest @xcite recently . a familiar example are magnetoresistance effects in hybrid systems @xcite consisting of magnetic and nonmagnetics parts , which have important applications @xcite in present data storage technology . in such magnetoelectronics devices ,
spin dependent conductances arise due to the interplay between ferromagnetic exchange
field splitting and the pauli principle . a finite spin polarization of electric current _ in the normal parts of a hybrid system
_ is a prerequisite @xcite for magnetoresistance to occur .
strong efforts are currently directed towards doing magnetoelectronics using the recently discovered diluted magnetic semiconductor materials @xcite .
parallel to the pursuit of a semiconductor magnetoelectronics paradigm , recent studies have focused on finding out how the quantum nature of spin can affect current flow , in particular , in _
semiconductor nanostructures .
the fundamentally relativistic coupling between charge carriers spin and orbital degrees of freedom turns out to give rise to a host of interesting , and sometimes counterintuitive , spin
dependent transport effects . of special appeal
is the rashba spin splitting @xcite arising in semiconductor heterostructures due to structural inversion asymmetry @xcite .
the possibility to tune its strength by external gate voltages was demonstrated experimentally @xcite and forms the basis for a spin dependent field effect
transistor ( spinfet ) design @xcite
. early studies @xcite discussed magnetoelectric effects in two dimensional electron systems with rashba spin splitting .
( see also recent related work @xcite . )
the possibility to induce spin accumulation , or a nonequilibrium magnetization , by applying an electric field only is very interesting from a basic science point of view , and would certainly be of great importance for spintronics applications .
in this review , we show how the interplay between quantum confinement and spin orbit coupling in quasi onedimensional ( 1d ) systems leads to just such a situation @xcite .
we are focusing on the electronic structure , transport properties , and interaction effects in ballistic quantum wires with rashba spin splitting present . in the following section ,
we formally introduce the theoretical model under consideration .
the limits of weak and strong spin
orbit coupling will be distinguished by comparison of the two fundamental length scales involved , namely the wire width @xmath0 and the rashba spin precession length @xmath1 . in sec .
[ spin_sec ] , the spin properties of electronic states in quasi1d subbands are revealed .
subband hybridization for the case of strong spin
orbit coupling turns out to result in an unusual spin structure where the direction of motion for electrons at the fermi wave number basically fixes their spin state .
this is in stark contrast to conventional wires where states for both spin species exist for either propagation direction .
quantum wires in the limit of strong spin orbit coupling exhibit therefore particularly intriguing transport properties , discussed in detail in sec .
[ trans_sec ] based on the scattering theory formalism of mesoscopic electron transport @xcite .
we find , e.g. , that application of an external voltage can lead to a spin accumulation and concomitant spin polarized current flow .
possibilities for experimental confirmation of our predictions will be elucidated .
the quasi1d subband structure obtained in sec . [ bands_sec ] will form the basis for a description of interacting rashba split quantum wires in sec .
[ int_sec ] , based on generalizations of the tomonaga luttinger model @xcite .
again , the interesting case is the one with strong spin
orbit coupling where the unusual spin properties of subband states result in peculiar properties of spin sensitive correlation functions .
we formulate our conclusions and give a brief outlook in the final sec .
[ concl_sec ] .
the text book example of a two dimensional ( 2d ) quantum well is now routinely realized by appropriate band gap engineering in semiconductor heterostructures @xcite . for low enough electron densities and temperatures ,
it is possible to describe the motion of electrons in the well using the hamiltonian @xmath2 of quasi
free particles in the lowest 2d subband .
( we take the growth direction of the heterostructure to be the @xmath3 axis in our spatial coordinate system . )
corrections to this hamiltonian which lead to a coupling of spin state and motion in real space arise , as in the three dimensional bulk material , in the absence of inversion symmetry . in the quantum
well , there exists an additional possibility for breaking inversion symmetry : creating an asymmetric band bending . for conduction band states , the spin
orbit hamiltonian resulting from this _ structural inversion asymmetry _
@xcite has the form @xcite @xmath4 here @xmath5 denotes the vector of pauli matrices , and the wave number @xmath6 is a direct measure of the rashba spin orbit coupling strength .
we use the latter as a phenomenological input parameter , which has to be determined experimentally @xcite or from spin dependent electronic structure calculations @xcite .
the single particle schrdinger equation for the hamiltonian @xmath7 describing the 2d electronic motion can be solved straightforwardly @xcite .
electronic eigenstates are labeled by a 2d wave vector @xmath8 and the quantum number @xmath9 of spin projection in the direction perpendicular to both @xmath8 and the growth direction .
the energy eigenvalues for states having the same @xmath10 but opposite @xmath11 turn out to differ by a zero field spin splitting @xmath12 . unlike zeeman splitting ,
to which it is often compared , rashba spin splitting does not result in a finite global magnetization of the 2d electron system , as time
reversal symmetry is preserved by the hamiltonian @xmath13 .
no common spin quantization axis can be found for its eigenstates .
furthermore , a cut through the energy dispersions reveals that spin bands are shifted not by a fixed energy , as is the case for zeeman splitting , but rather in wave vector direction .
this is illustrated in fig .
[ compsplit ] . .
degenerate parabolic curves for both spin states get to be shifted in _
energy direction _ by an applied magnetic field . quite differently , finite rashba spin splitting results in _ wave vector shifted _ dispersion curves .
[ compsplit],width=453 ] in the following , we are focusing exclusively on the situation where the motion of electrons is further confined to one spatial dimension by an external potential @xmath14 . to be specific ,
we assume a parabolic confinement @xmath15 as we can expect the qualitative features of spin splitting in quantum wires to be independent of the actual shape of the confining potential . in the absence of spin
orbit coupling , the single electron spectrum is split into quasi1d subbands having quadratic dispersion in the 1d wave vector @xmath16 that labels its eigenstates .
the characteristic energy scale for subband bottoms is related to the width @xmath0 of the quantum wire . for finite @xmath6
, energy eigenstates are still plane waves in wire direction , but the linear dependence of @xmath17 on the momentum @xmath18 in confinement direction introduces a coupling between the quasi1d subbands . using the 1d plane wave representation , we can write the quantum wire hamiltonian in the form @xmath19 , where @xmath20 the importance of quantum wire subband coupling can be quantified by the ratio @xmath21 of the matrix elements of @xmath22 between eigenstates of @xmath23 ( which describes a hypothetical quantum wire having only @xmath24dependent spin splitting ) and the difference of the corresponding eigen energies @xcite . estimating the parameter @xmath21 in terms of the wire width @xmath0 , we find @xmath25 where @xmath1 is the spin precession length familiar from the proposed spinfet @xcite .
hence two limits can be distinguished .
( i ) quantum wire with weak rashba spin splitting , realized for @xmath26 : eigenstates are plane wave spinors that are eigenstates of @xmath27 , i.e. , have their spin polarized in the direction perpendicular to ( and in the plane of ) the wire .
the energy dispersion is given by two parabolas , distinguished by the spin quantum number , that are shifted in wave vector direction by an amount @xmath28 .
this is illustrated in fig .
[ displimit]a , which is essentially identical to the cut through a 2d rashba split dispersion shown in fig .
[ compsplit ] .
( ii ) quantum wire with strong rashba spin splitting , realized when @xmath29 : hybridization of quasi1d subbands results in a nonparabolic energy dispersion @xcite .
see fig .
[ displimit]b . in this case
, no common spin quantum number can be assigned to states within a spin split quantum
wire subband @xcite .
their intriguing spin properties will be discussed in the following section . .
( b ) in the opposite case of strong spin splitting , hybridization between adjacent subbands with opposite spin results in a nonparabolic dispersion .
we show data for the case @xmath30 .
[ displimit],title="fig:",height=188 ] .
( b ) in the opposite case of strong spin splitting , hybridization between adjacent subbands with opposite spin results in a nonparabolic dispersion .
we show data for the case @xmath30 .
[ displimit],title="fig:",height=188 ] a useful analytical result can be obtained for energy eigenvalues of states in a parabolic quantum wire having @xmath31 . as @xmath32 vanishes in this case , these are eigenspinors of @xmath33 whose components are eigenfunctions of harmonic oscillators with a spin dependent boost .
the energy eigenvalue is degenerate in the good quantum number of spin projection in the wire direction . with @xmath34 denoting the oscillator length scale introduced by the parabolic wire confinement , it is explicitly given by @xmath35 \quad .\ ] ] the spin
orbit correction to the energy of these states arises due to their finite quantized motion in the direction perpendicular to the wire .
this exact result is an important benchmark to judge the accuracy of numerical methods that are necessary to obtain the full dispersion curves for all wave vectors @xcite .
while it is not possible to find a common spin quantization axis for eigenstates in a 2d system with rashba spin splitting , restriction to one spatial propagation direction restores a global spin projection axis _ for the case of weak spin
orbit coupling_. in this limit , it is possible to neglect @xmath22 , and the single
electron states in the wire are given , to a good approximation , by eigenstates of spin projection parallel to the wire confinement , i.e. , the @xmath36 direction .
at any given energy , there exist two such eigenstates with opposite spin , which have wave vectors differing by @xmath28 .
the effect of finite @xmath22 is to modify the local spin density of eigenstates across the wire , introducing a small zero average tilt out of the wire plane @xcite .
this inhomogeneity of spin density as a function of the coordinate perpendicular to the wire becomes more important for strong spin
orbit coupling @xcite
. still , direct experimental observation of this texture like spin structure would be quite challenging .
the mixing term @xmath22 couples eigenstates obtained from diagonalizing @xmath37 which have opposite spin . in the special case of a parabolic confinement
, its matrix elements turn out to be finite only between those opposite
spin states whose oscillator band indices differ by one @xcite .
hence , wherever the dispersion relations for such energetically adjacent subbands with opposite spin cross , @xmath22 can not be neglected , as it will induce an anticrossing . another way to define the limit of weak spin
orbit coupling is then to say that such anticrossings occur only at energies that are much higher than the typical quasi1d subband splitting and , hence , can be neglected in the case when only a few wire subbands are occupied . as the parameter @xmath21 defined in eq .
( [ weakstrong ] ) gets close to unity , anticrossings occur at energies comparable to those of low lying wire subbands and affect its properties in an important way . following the picture of anticrossings
, it becomes immediately apparent that a peculiar spin structure emerges for states in hybridized subbands .
in particular , their eigen spin direction becomes a function of wave vector and is not uniform anymore within each band .
are parabolas shifted , due to rashba spin splitting , in wave vector direction , as shown in the right panel .
corresponding eigenstates have spin polarized in the direction perpendicular to the wire .
the mixing hamiltonian @xmath22 induces subband hybridization near crossing points , indicated on the right panel by circles . as a result ,
states in the lowest two spin split subbands have approximately parallel spin at large wave vectors .
[ anticross],height=188 ] consider the hybridization of the lowest two spin split subbands , as illustrated in fig .
[ anticross ] . the resulting nonparabolic dispersion for a particular parameter @xmath38
is given in fig .
[ displimit]b .
note that states in the lowest pair of subbands exist with wave vector larger than that for the crossing point but energies smaller than the bottoms of the next subband pair .
these states originated from the first
excited spin down subband but have become part of the lowest subband after hybridization .
their wave vector distance to the crossing point is large enough such that the spin up admixture in the eigenspinors is quite small .
hence we find a situation where right moving states at still comfortably low energies in the lowest spin
split subbands have basically parallel spin ! for the case depicted , there are basically only spin down right movers at large enough energies , and similarly only spin up left movers .
this direct association of spin state with the direction of motion is peculiar to the strong rashba split situation . in an ordinary quantum wire ,
even with zeeman splitting or weak rashba splitting present , there exist right moving states for both spin up and spin down electrons _ at any energy_. it can be envisioned that , for suitable parameter ranges , there exists a finite energy window for which even more than two sets of low lying subbands have right moving ( and left moving ) states with almost parallel spin .
this turns out to indeed be the case for @xmath30 . in fig .
[ spinpic ] , we show the expectation value of the spin projection perpendicular to the wire for eigenstates in the lowest two pairs of spin split quantum wire subbands . .
( see the corresponding dispersions in fig .
[ displimit ] . ) in both cases , states at large wave vector have approximately parallel spin .
[ spinpic],title="fig:",height=188 ] .
( see the corresponding dispersions in fig .
[ displimit ] . ) in both cases , states at large wave vector have approximately parallel spin .
[ spinpic],title="fig:",height=188 ] in both sets of subbands , right moving states at large wave vector exist only for spin down electrons , and left moving ones only for spin up
@xcite .
exactly at the anticrossing , eigenstates of @xmath39 are the symmetric and antisymmetric superpositions of spin up and spin down states from the two intersecting quasi1d subbands for @xmath37 . due to this peculiar mixing of spin and subband wave functions , expectation values of spin projected in any direction yields zero for quantum wire eigenstates at the anticrossing .
their special spin properties turn out to give rise to an additional spin rotation of incoming electrons with energies within the anticrossing gap @xcite , which would enable enhanced performance of a suitably designed spinfet @xcite .
the key theoretical tool for studying linear transport in mesoscopic systems is the landauer
bttiker formalism @xcite , which is valid when electron
electron interaction is negligible .
the central result of this approach is that the conductance of a mesoscopic sample can be related to the scattering matrix , @xmath40 , of the sample . in the spin
independent case , the scattering matrix is diagonal in spin space , and the problem can be mapped to scattering of spinless particles .
spin is then taken into account only as a factor @xmath41 , and the linear conductance at zero temperature simply reads @xmath42 , where @xmath43 is the scattering amplitude ( transmission coefficient ) between the mode @xmath44 in terminal @xmath45 and the mode @xmath46 in terminal @xmath41 .
in this situation ( spin independent case ) , time reversal symmetry implies @xmath47 , i.e. , @xmath48 , where , in the generic multi - terminal case , @xmath46 and @xmath44 are collective indices labeling both terminals and transverse modes . in the presence of spin orbit coupling
, there is no common spin quantization axis , and the full spin structure of @xmath40 should be retained @xcite .
the total two - terminal conductance reads then @xmath49 denote the spin projection along a common quantization axis .
the spin@xmath11 current in the output contact @xmath41 is given by @xmath50 , where @xmath51 describes the fact that the injecting reservoir @xmath45 is not spin polarized .
in this case of a nondiagonal scattering matrix in spin space , the condition satisfied by the scattering matrix due to time reversal symmetry reads @xmath52 where @xmath11 and @xmath53 denote the spin projection along a common quantization axis and can take the values @xmath54 .
the condition eq .
( [ smat_cond ] ) does not forbid the creation of spin - polarized currents in linear transport .
current @xmath55 ( where the quantization axis is along @xmath36 ) is plotted as a function of the coordinate along the wire . in panel
( a ) there is only one interface between a semi infinite lead ( @xmath56 ) and a semi infinite wire with spin - orbit coupling . in panel
( b ) the wire is contacted by two semi infinite leads .
the two interfaces with the leads are located at @xmath57 and @xmath58 , respectively . in both cases ,
reservoirs inject spin - unpolarized electrons .
the fermi energy is chosen such that only one subband is occupied .
the oscillatory behavior seen in panel ( b ) is due to quantum interference caused by multiple scattering between the interfaces .
the parameters used in the simulation are @xmath59 and @xmath60 .
the calculation has been performed within the two band model described in ref .
[ curr],title="fig:",height=188 ] current @xmath55 ( where the quantization axis is along @xmath36 ) is plotted as a function of the coordinate along the wire . in panel
( a ) there is only one interface between a semi infinite lead ( @xmath56 ) and a semi infinite wire with spin - orbit coupling . in panel ( b ) the wire is contacted by two semi infinite leads .
the two interfaces with the leads are located at @xmath57 and @xmath58 , respectively . in both cases ,
reservoirs inject spin - unpolarized electrons .
the fermi energy is chosen such that only one subband is occupied .
the oscillatory behavior seen in panel ( b ) is due to quantum interference caused by multiple scattering between the interfaces .
the parameters used in the simulation are @xmath59 and @xmath60 .
the calculation has been performed within the two band model described in ref .
[ curr],title="fig:",height=188 ] we now consider a wire with rashba spin - orbit coupling with two nonmagnetic contacts .
these are described by semi - infinite leads with no spin orbit coupling . in the landauer - bttiker approach ,
movers are populated by the left reservoirs ( with chemical potential @xmath61 ) and left
movers by the right reservoirs ( with chemical potential @xmath62 ) .
this consideration , together with the observation that we can realize situations in which the chirality ( direction of motion ) fixes the spin state ( as discussed in the previous section ) , leads to the expectation of spin accumulation in the wire when a transport voltage is applied . ,
title="fig:",height=188],title="fig:",height=188 ] to obtain linear transport currents , the scattering problem is solved by standard mode matching techniques .
the usual condition for the probability current conservation needs to be modified because the velocity operator in the presence of rashba spin - orbit coupling reads @xcite @xmath63= \hbar ( k_y + k_{\mathrm{so}}\sigma_x)/m \quad .\ ] ] to have a clear understanding of what happens at the interface between a lead without spin orbit coupling and the wire with spin orbit coupling , we study first the simple situation of a single interface between a semi
infinite lead and a semi infinite wire with spin orbit coupling .
the results for the total spin - up(-down ) current as a function of the distance from the interface are shown in fig .
the current flowing in the lead is unpolarized , while in the wire , distant enough from the interface , the current becomes spin polarized in agreement with the spin properties of the eigenstates shown in fig .
[ spinpic]a .
hence , the rashba spin splitting leads to _ spin accumulation _ in the wire without ferromagnetic contacts . the other important phenomenon which is apparent from fig .
[ curr ] is the process of _ current conversion _ occurring at the interface .
this process is mediated by scattering into evanescent modes of the wire , and is possible thanks to the anomalous form of the velocity operator ( [ velop ] ) .
[ spinpic]b shows results for the experimentally relevant situation of a finite length wire with spin
orbit coupling between two leads .
also in this case , we have both spin accumulation and current conversion , plus some finite - size oscillation due to quantum interference @xcite .
finally we suggest two possible schemes to enable experimental verification of spin accumulation ( see fig .
[ experiment ] ) .
one possibility would be to weakly couple a ferromagnetic voltage probe to the wire , measuring the chemical potential of the majority spins .
changing the direction of the magnetization of the ferromagnetic contact enables the detection of the difference in chemical potentials of the two spin species .
another possibility could be to use injection of a spin
polarized current from a ferromagnet to probe the chirality of propagating spins in the wire .
the direction of current flow should then depend on the direction of the magnetization of the ferromagnet @xcite .
interactions between electrons turn out to have more dramatic consequences in 1d as compared to higher dimensional systems . instead of following the familiar @xcite fermi
liquid paradigm , 1d conductors form their own new class of interacting metals dubbed _ luttinger liquids _ @xcite .
power law behavior of electronic correlation functions , such as the tunneling density of states or the momentum space occupation number , near
the fermi points are a signature of non fermi liquid properties exhibited by interacting 1d systems .
previous work @xcite has shown that weak rashba spin splitting leaves such power laws essentially unaffected .
however , the interplay between luttinger liquid behavior and spin orbit coupling turns out to be nontrivial in quantum wires with _
strong _ rashba spin splitting and deserves a more detailed discussion .
the current understanding of interaction effects in 1d systems is based on the possibility to map real systems onto the exactly soluble tomonaga
luttinger model @xcite .
the starting point of such a mapping is linearization of the single particle dispersion relation near the fermi points .
furthermore , two chiral electron flavors ( right movers and left
movers ) are distinguished according to their velocity direction . in the present case of a strong rashba split quantum wire , the spin degeneracy of these chiral electron branches is broken . in the most general case ,
we can distinguish two right moving branches having different velocities . as discussed in sec .
[ spin_sec ] above , their spin properties are peculiar because they result from hybridization of quasi1d subbands with opposite spin .
in particular , when fermi points are sufficiently far away from anticrossings , both right moving branches can be modeled as having approximately parallel spin .
the same is then true for left
movers which , however , have their spin opposite to that of right movers . to be most general
, we do not _ a priori _ assign spin labels to the chiral electron branches . instead , we distinguish them by their different velocities .
see fig .
[ lineardisp ] for an illustration . ,
title="fig:",height=162],title="fig:",height=162 ] the single
electron part of the tomonaga
luttinger hamiltonian of a quantum wire with strong rashba spin splitting can then be written as @xmath64 we have omitted uniform shifts of the chemical potential and total energy of the system . introducing the familiar chiral bosonic phase fields @xcite @xmath65 that are related to the normal ordered electron density for each branch via @xmath66 , with @xmath67 and @xmath68 ,
we can write the single particle and interaction parts of the tomonaga luttinger hamiltonian @xmath69 as @xmath70 here we have only included the long wave length forward scattering component @xmath71 of the screened coulomb interaction in the wire . changing representation to the nonchiral conjugate phase fields @xmath72 and @xmath73)/\sqrt{8\pi}$ ] , and introducing the abbreviations @xmath74 , @xmath75 , @xmath76 , and @xmath77 , we find after some algebra @xmath78 , where @xmath79 comparison with the tomonaga
luttinger model hamiltonian of an ordinary quantum wire @xcite shows that @xmath80 describes plasmon excitations , i.e. , fluctuations in the total electron density of the wire , where @xmath81 is the usual luttinger parameter .
the term @xmath82 represents long wave length excitations in the density difference between electrons ( both right moving and left moving ) in the b and a branches , which is unaffected by coulomb interactions .
note that @xmath83 and its conjugate field @xmath84 _ do not _ represent the spin density in the strongly rashba split quantum wire not even in the limit where electrons close to the fermi points have approximately parallel spin , as depicted in fig .
[ lineardisp]b . finally , @xmath85 arises because of the velocity difference for electrons from the a and b branches .
its existence leads to the emergence of four different chiral normal modes that diagonalize @xmath86 , in contrast to the case of an ordinary quantum wire where pairs of left moving and right moving normal modes with equal velocities exist that can be combined to two nonchiral normal modes .
( the latter represent total charge and total spin density waves , respectively . ) the peculiar spin structure of single
electron states in strongly rashba
split subbands gives rise to an unusual expression for the zeeman energy in an external magnetic field @xmath87 .
focusing on the case depicted in fig .
[ lineardisp]b , a straightforward calculation yields @xmath88 with electron g
factor @xmath89 and bohr magneton @xmath90 .
the above expressions for @xmath86 and @xmath91 are quadratic in the bosonic phase fields @xmath92 and their conjugates @xmath93 . together with the bosonization identities
@xcite for electrons from each of the chiral branches , it is possible to obtain exact results for electronic correlation functions , which are omitted here because of limited space . in the absence of an external magnetic field ,
spin independent correlation functions are identical to those of a two component luttinger liquid discussed , e.g. , in refs .
our study of rashba spin splitting in low dimensions has given us new insight into the interplay between quantum confinement and spin orbit coupling .
intriguing spin
dependent transport effects arise in quantum wires having a width that is comparable to the spin precession length .
these could form the basis for realizing spintronics devices without involving any magnetic parts .
possibilities for experimental verification of our predictions have been suggested , which could be realized using present nanofabrication techniques .
it will be interesting to investigate , both experimentally and theoretically , spin orbit effects in related quantum confined structures such as quantum point contacts and non semiconductor based systems .
this work was supported in part by the center for functional nanostructures at the university of karlsruhe , germany .
we benefitted from several useful discussions with f. e. meijer and a. f. morpurgo .
we suspect that the shift of quasi1d rashba
split subbands to lower energies , as apparent from eq .
( [ specstate ] ) , actually prevents the appearance of ` bumps ' in the dispersion relation that were claimed to exist in ref .
see fig .
5c in that paper which shows their numerical results for @xmath94 .
for the purpose of direct comparison , we give our numerical results for the same parameter in our fig .
[ displimit]b .
we find no bumps .
surprisingly , the energies for states at @xmath31 obtained numerically in ref .
@xcite are unshifted from the subband
bottom energy at zero @xmath95 , in disagreement with the exact analytical result of our eq .
( [ specstate ] ) .
the spin properties of right movers and left movers in strong
rashba split quantum wires were presented incorrectly in previous works by moroz and barnes @xcite , which has mislead subsequent related studies @xcite .
the results shown in fig . [ curr ] differ from those presented in fig . 4 of ref .
@xcite because , when calculating the latter , the confinement in the leads was taken to be different from that in the wire .
( for data shown in fig . 4 of ref .
@xcite , the first two subband energies in the leads were @xmath96 and @xmath97 , respectively . ) data shown here are for a uniform confinement potential in the entire structure consisting of the leads and the rashba spit wire region .
as it turns out , the phenomena of spin accumulation and current conversion are exhibited for either choice of parameters .
|
this article presents an overview of results pertaining to electronic structure , transport properties , and interaction effects in ballistic quantum wires with rashba spin splitting .
limits of weak and strong spin orbit coupling are distinguished , and spin properties of the electronic states elucidated . the case of strong rashba spin splitting
where the spin precession length is comparable to the wire width turns out to be particularly interesting .
hybridization of spin split quantum
wire subbands leads to an unusual spin structure where the direction of motion for electrons can fix their spin state .
this peculiar property has important ramifications for linear transport in the quantum wire , giving rise to spin accumulation without magnetic fields or ferromagnetic contacts .
a description for interacting rashba split quantum wires is developed , which is based on a generalization of the tomonaga luttinger model .
quasi1d spin
split subbands , spin
dependent transport , two band luttinger model 85.75.-d , 73.23.ad , 73.63.nm , 71.70.ej
|
phylogenetic trees have proved an important tool for representing evolutionary relationships . for a set @xmath5 of species ( or , more generally , taxa ) these
are formally defined as leaf - labelled trees whose leaves are bijectively labelled by the elements of @xmath5 .
advances in dna sequencing have resulted in ever more data on which such trees may be based .
computational limitations however combined with the need to understand species evolution have left biologists with the following fundamental problem which we will refer to as _ amalgamation problem _ : given a collection @xmath0 of phylogenetic trees , how can these trees be amalgamated into an overall parental structure that preserves the phylogenetic relationships supported by the trees in @xmath0 ?
the hope is that such a structure might help shed light on the evolution of the underlying genomes ( and thus the species ) . in the ideal case
that all trees in @xmath0 support the same phylogenetic relationships ( as is the case for trees @xmath6 and @xmath7 depicted in fig .
[ pete - trees ] ) this structure is known to be a phylogenetic tree and a supertree method @xcite may be used to reconstruct it . for the above example the outcome @xmath8 of such a method
is @xmath6 with species _ cru _ ( see fig . [ pete - trees ] for full species names ) attached via a pendant edge to the vertex labelled @xmath9 .
it should be noted that @xmath8 supports the same phylogenetic relationships as @xmath6 and @xmath7 in the following sense : for a finite set @xmath5 , call a bipartition @xmath10 of some subset @xmath11 a _ partial split _ on @xmath5 , or a _ partial ( x)-split _ for short , and denote it by @xmath12 or , equivalently , by @xmath13 where @xmath14 . in particular , call @xmath15 a _ ( full ) split _ of @xmath5 if @xmath16 .
furthermore , say that a partial @xmath5-split @xmath17 _ extends _ a partial @xmath5-split @xmath18 if either @xmath19 and @xmath20 or @xmath21 and @xmath22 .
finally , say that a phylogenetic tree @xmath23 _ displays _ a split @xmath17 if @xmath15 is a partial split on the leaf set @xmath24 of @xmath23 induced by deleting an edge of @xmath23 .
then `` supports the same phylogenetic relationships '' means that for every split @xmath15 displayed by @xmath6 or @xmath25 there exists a split on @xmath26 that extends @xmath15 and is displayed by @xmath8 .
due to complex evolutionary mechanisms such as incomplete lineage sorting , recombination ( in the case of viruses ) , or lateral gene transfer ( in case of bacteria ) the trees in @xmath0 may however support not the same but conflicting phylogenetic relationships . a phylogenetic network in the form of a split network ( see @xcite for overviews ) rather than
a phylogenetic tree is therefore the structure of choice if one wishes to simultaneously represent all phylogenetic relationships supported by the trees in @xmath0 .
an example in point is the split network pictured in fig .
[ pete - network ] which appeared as a weighted network in @xcite . with replacing `` edge '' in the definition of displaying by `` band of parallel edges '' and `` @xmath24 '' by `` set of network vertices of degree 1 '' to obtain a definition for when a split network displays a split , it is straight forward to check that the network in fig .
[ pete - network ] displays all splits displayed by the 3 trees pictured in fig .
[ pete - trees ] .
it should be noted that phylogenetic networks such as the one depicted in fig .
[ pete - network ] ( see e.g. @xcite for recently introduced other types of phylogenetic networks ) provide a means to visualize the complexity of a data set and should not be thought of as an explicit model of evolution .
awareness of this complexity does not only allow the exploration of a data set but , as is the case of e.g. hybridization networks @xcite , can also serve as starting point for obtaining an explicit model of evolution ( see @xcite for more on this ) .
a circular phylogenetic network that represents all phylogenetic relationships supported by the trees depicted in fig .
[ pete - trees ] ( see that figure for full species names ) .
, width=226,height=151 ] apart from displaying all splits induced by the 3 trees depicted in fig . [ pete - trees ] , the network depicted in fig .
[ pete - network ] has a further interesting feature .
it is _ circular_. in other words , if @xmath5 denotes the set of the 7 plant species under consideration , then the elements of @xmath5 can be arranged around a circle @xmath27 so that every split @xmath17 of @xmath5 displayed by the network can be obtained by intersecting @xmath27 with a straight line so that the label set of one of the resulting 2 connected components is @xmath28 and the label set of the other is @xmath29 .
although seemingly a very special type of phylogenetic network , circular phylogenetic networks are a frequently used structure in phylogenetics ( see e.g.@xcite ) as they do not only naturally generalize the concept of a phylogenetic tree but are also guaranteed to be representable in the plane ; a fact that greatly facilitates drawing and thus analyzing them . however , although recently first steps have been made with regards to finding a solution to the amalgamation problem in terms of a phylogenetic network leading to the attractive z - closure @xcite and @xmath2-imputation @xcite approaches , very little is known about a solution of this problem in terms of a circular phylogenetic network .
intrigued by this and motivated by the fact that , from a combinatorial point of view , phylogenetic trees and networks are _ split systems _ ( i.e.collections of full splits ) and that therefore the amalgamation problem boils down to the problem of how to extend partial splits on some set @xmath5 to splits on @xmath5 , we wondered whether closure rules for partial splits could not be of help .
essentially mechanisms for splits enlargement , such rules have proved useful for supertree construction and also underpin the above mentioned @xmath1-closure super - network approach .
as it turns out , this is indeed the case . as an immediate consequence of our main result ( corollary [ cyclosure ] ) ,
we obtain that for a collection of partial splits that can be `` displayed '' by a circular phylogenetic network @xmath30 , the collection of ( full ) splits generated by the closure rules in the centre of this paper is guaranteed to be displayable by @xmath30 and also independent of the order in which the rules are applied . in a study aimed at shedding light into the origin of eukaryotes , rivera et al .
@xcite put forward the idea of a `` ring of life '' with the eukaryotic genome being the result of a fusion of two diverse procaryotic genomes ( see also @xcite ) . a natural and interesting question in this context is how dependent rivera et al.s ring of life is on the fact that all underpinning trees are on the same taxa set . in the last section of this paper ,
we provide a partial answer by presenting an example of a collection of induced partial trees from which the ring of life can be reconstructed using the @xmath4- and @xmath3-rule .
the paper is organized as follows . in section [ closure - rules ] , we first introduce some more terminology and then restate one of meacham s closure rules ( our @xmath4-rule ) and introduce the novel @xmath3-rule . in section [ relship ] , we study the relationship between the @xmath4- and @xmath3-rule and the closure rule that underpins the aforementioned @xmath1-closure super - network approach . in section [ properties ] , we introduce the concept of a circular collection of partial splits and show that both the @xmath3- and @xmath4-rule preserve circularity ( proposition [ circ ] ) . in section [ spl - sequence ] , we introduce the concept of a split closure and show that for certain collections of partial splits this closure is independent of the order in which the @xmath3-rule and/or @xmath4-rule are / is applied ( theorem [ split - closure ] ) .
this result lies at the heart of corollary [ cyclosure ] . in section [ rivera - exmple ]
, we explore the dependency of rivera et al.s ring of life on the fact that the underpinning trees are all on the same data set throughout the paper , @xmath5 denotes a finite set and the terminology and notation largely follows @xcite .
we start this section by introducing some additional terminology and notation .
subsequent to this , we first restate meacham s rule ( which we call the @xmath4-rule ) and then introduce a novel closure rule which we call the @xmath3-rule .
let @xmath31 denote the collection of all partial splits of @xmath5 and suppose @xmath32 .
then a partial split @xmath33 that can be extended by a partial split @xmath34 is called _
redundant_. the set obtained by removing redundant elements from @xmath35 is denoted by @xmath36 .
if @xmath37 then @xmath35 is called _ irreducible _ and the set of all irreducible subsets in @xmath31 is denoted by @xmath38 .
note that the relation `` @xmath39 '' defined for any two ( partial ) split collections @xmath40 by putting @xmath41 if every partial split in @xmath35 is extended by a partial split in @xmath42 is a partial order on @xmath38 @xcite .
suppose for the following that @xmath43 is a closure rule , that is , a replacement rule that replaces a collection @xmath44 of partial splits that satisfy some condition @xmath45 by a collection @xmath46 whose elements are generated in some systematic way from the partial splits in @xmath47 ( see e.g. the @xmath4- and the @xmath3-closure rules presented below for two such systematic ways ) .
suppose @xmath40 are two irreducible collections of partial splits and @xmath48 is the set of all subsets of @xmath49 that satisfy @xmath45 .
if there exists some subset @xmath50 such that @xmath51 then we say that @xmath42 is _ obtained from @xmath35 via a single application of @xmath43_. finally , if for every subset @xmath50 we have @xmath52 then we call an application of @xmath43 to @xmath35 _ trivial _ and say that @xmath35 is _ closed _ with respect to @xmath43 .
we are now in the position to present the 2 closure rules we are mostly concerned with in this paper : the @xmath4-rule which is originally due to meacham @xcite and the novel @xmath3-rule .
we start with meacham s rule .
suppose @xmath53 are two distinct partial splits of @xmath5 .
then the _ @xmath4-rule _
@xmath54 is as follows :
if there exists some @xmath55 , @xmath56 such that @xmath57 then replace @xmath58 by the set @xmath59 which comprises of @xmath47 and , in addition , also the partial splits @xmath60 in case the partial splits @xmath61 and @xmath62 are such that there is no ambiguity with regards to the identity of the sets @xmath63 and @xmath64 in the statement of the @xmath4-rule or they are irrelevant to the discussion , we will simplify @xmath65 to @xmath66 .
clearly , such ambiguity can not arise if @xmath61 and @xmath62 are _ compatible _ , that is , there exist subsets @xmath67 , @xmath56 such that @xmath68 .
however if @xmath61 and @xmath62 are _ incompatible _ , that is , not compatible then caution is required .
note that if @xmath63 and @xmath64 as in the statement of the @xmath4-rule are such that @xmath69 and @xmath70 , then it is easy to verify that @xmath71 applies trivially to @xmath47 . also note that for any @xmath72 and any two distinct partial splits @xmath73 , we have @xmath74 finally , note that any phylogenetic tree on @xmath5 that displays the partial splits in some set @xmath72 also displays the partial splits in @xmath75 , @xmath73 .
suppose @xmath76 , @xmath77 , are three distinct partial splits of @xmath5 .
then the _ @xmath3-rule _
@xmath78 is as follows :
if there exists some @xmath55 , @xmath77 such that @xmath79 ( see fig . [ c_sk](a ) for a graphical interpretation ) , then replace @xmath80 by the set @xmath81 which comprises of the partial splits @xmath82 although the condition in ( [ sk - rule ] ) might look quite strange at first sight , the class of triplets of partial splits that satisfy it is very rich .
for example , suppose that @xmath83 , @xmath77 are splits of @xmath5 that can be arranged in the plane as indicated in fig .
[ c_sk](b ) where each bold , straight line represents one of @xmath84 , @xmath77 and the dots represent non - empty triplewise intersections of the parts of @xmath84 , @xmath77 , in which they lie .
for example , the dot in the bottom wedge represents the intersection @xmath85 .
the shaded regions correspond to the 3 non - empty intersections mentioned in the statement of the @xmath3-rule . the partial splits @xmath86 @xmath77 obtained by restricting @xmath61 , @xmath62 , and @xmath87 to different subsets of @xmath5 so that the shaded regions remain non - empty form a triplet of partial splits that satisfy ( [ sk - rule ] ) .
as the example of set @xmath47 comprising the three partial splits @xmath88 , @xmath89 , and @xmath90 shows different choices of the sets @xmath91 , @xmath77 lead to different sets @xmath81 .
for example , if @xmath92 , @xmath93 , and @xmath94 then ( [ sk - rule ] ) is satisfied and @xmath95 .
if however @xmath63 and @xmath64 are as before and @xmath96 , then ( [ sk - rule ] ) is also satisfied and @xmath81 is the set @xmath97 .
following our practise for the @xmath4-rule , for @xmath80 we simplify @xmath98 to @xmath99 if the partial splits @xmath84 , @xmath77 are such that there is no ambiguity with regards to the identity of the sets @xmath91 , @xmath77 , in the statement of the @xmath3-rule or they are irrelevant to the discussion .
note that if @xmath91 , @xmath77 as in the statement of the @xmath3-rule are such that , in addition , @xmath100 , @xmath101 , @xmath102 , and @xmath103 it is easy to see that @xmath78 applies trivially to @xmath47 .
also note that for any @xmath72 and any 3 partial splits @xmath104 of @xmath5 , we have @xmath105
in this section we first restate the _
@xmath1-(closure ) rule _ which was used in @xcite in the context of a supernetwork construction approach and then investigate the relationship between the @xmath3- , @xmath4- , and @xmath1-rule . also originally due to meacham @xcite , the @xmath1-rule @xmath106 can be restated as follows : suppose @xmath53 are two distinct partial splits of @xmath5 . 1 . if there exists some @xmath55 , @xmath56 such that @xmath107 then replace @xmath58 by the set @xmath99 which comprises of the partial splits @xmath108 and @xmath109 . note that any two compatible partial splits of @xmath5 satisfy the condition in ( [ condition1 ] ) . with this third closure rule at hand
we are now in the position to present a first easy to verify result .
suppose @xmath61 , @xmath62 , and @xmath87 are 3 distinct partial splits of @xmath5 such that there exist parts @xmath55 , @xmath77 as in the statement of the @xmath3-rule .
if , in addition , @xmath110 and @xmath111 , then the partial split @xmath112 generated by @xmath78 is also generated by first applying @xmath54 to @xmath62 and @xmath87 ( with regards to @xmath113 ) and then applying @xmath54 to the resulting partial split @xmath114 and @xmath61 .
in addition , we have the following result whose straight forward proof we leave to the reader .
suppose @xmath61 is a full split of @xmath5
. then the following statements hold . 1 .
if @xmath62 is a partial @xmath5-split and @xmath106 applies to @xmath115 , then @xmath116 2 .
if @xmath62 and @xmath87 are partial @xmath5-splits so that @xmath78 applies to @xmath117 and @xmath106 applies to @xmath118 and @xmath119
. then @xmath120
in this section we introduce the notion of a weakly compatible collection of partial splits and study properties of the @xmath3- and @xmath4-rules regarding such collections .
a particular focus lies on the study of circular collections of partial splits which we also introduce .
as we will see , they form a very rich subclass of such collections of partial splits .
we start this section with a definition that generalizes the concept of weak compatibility for ( full ) splits of @xmath5 @xcite to partial splits of @xmath5 .
suppose @xmath121 , @xmath77 , are three partial @xmath5-splits .
then we call @xmath122 _ weakly compatible _ if at least one of the four intersections @xmath123 is empty , @xmath62 and @xmath87 are full splits and the condition in ( [ thinly ] ) is the same ( see @xcite ) . ] .
since the roles of @xmath91 and @xmath124 in @xmath84 , @xmath77 , can be interchanged without changing @xmath122 we have that @xmath122 are weakly compatible if and only if at least one of the four intersections @xmath125 is empty .
more generally , we call a collection @xmath32 of partial @xmath5-splits _ weakly compatible _ if every three partial splits in @xmath35 are weakly compatible . to give an example , the partial splits @xmath126 , @xmath127 , and @xmath128 are weakly compatible whereas the partial splits @xmath87 , @xmath129 , and @xmath130 are not .
thus , @xmath131 is not weakly compatible .
note that , like in the case of ( full ) splits , it is easy to see that any collection of pairwise compatible partial splits is also weakly compatible .
clearly any three partial splits @xmath121 , @xmath77 , for which precisely one of the four intersections in ( [ thinly ] ) is empty also satisfies condition ( [ sk - rule ] ) .
thus @xmath78 may be applied to @xmath122 . however , as the example of the set @xmath132 shows , application of @xmath78 to a weakly compatible collection of partial splits does not , in general , yield a weakly compatible collection of partial splits .
also it should be noted that @xmath54 applied to a weakly compatible collection of partial splits does not always yield a weakly compatible collection of partial splits
. however , the next result whose proof is straight forward holds .
[ pink ] suppose @xmath133 .
if @xmath42 is weakly compatible and @xmath41 , then @xmath35 must also be weakly compatible .
we now turn our attention to the study of a special class of weakly compatible collections of partial splits called circular collections of partial splits . to be able to state their definition
, we require some more terminology which we introduce next . a _ cycle _ @xmath27 is a connected graph with @xmath134 and every vertex has degree 2 .
we call @xmath27 an _ @xmath5-cycle _ if the vertex set of @xmath27 is @xmath5 .
for @xmath135 ( @xmath136 ) and @xmath27 an @xmath5-cycle , we call @xmath137 a _ vertex ordering ( of @xmath27 ) _ if the edge set of @xmath27 coincides with the set of all @xmath138-sets @xmath139 of @xmath5 , @xmath140 . for a graph @xmath141 and some subset @xmath142 of @xmath143 , we denote by @xmath144 the graph obtained from @xmath145 by deleting the edges in @xmath142 .
we say that a partial @xmath5-split @xmath12 is _ displayed _ by an @xmath5-cycle @xmath27 if there exist two distinct edges @xmath146 and @xmath147 in @xmath27 such that the vertex set of one of the two components of @xmath148 contains @xmath28 and the other one contains @xmath29 .
more generally , we say that a set @xmath32 of partial splits is _ displayed _ by an @xmath5-cycle @xmath27 if every partial split in @xmath35 is displayed by @xmath27 .
finally , we say that a collection @xmath32 is _ circular _ if there exists some @xmath5-cycle @xmath27 such that every partial split in @xmath35 is displayed by @xmath27 .
note that every split collection in @xmath31 displayed by a circular phylogenetic network is circular . as is well - known , every circular split system is in particular weakly compatible .
the next result shows that an analogous result holds for collections of partial splits .
[ circular - twc ] suppose @xmath32 .
if @xmath35 is circular then @xmath35 is also weakly compatible .
_ proof : _ suppose @xmath27 is an @xmath5-cycle that displays @xmath35 but there exist three partial splits @xmath149 such that with @xmath55 , @xmath77 , playing the role of their namesakes in ( [ thinly ] ) none of the four intersections in ( [ thinly ] ) is empty .
then @xmath61 and @xmath62 are incompatible and , since @xmath61 and @xmath62 are displayed by @xmath27 , there must exist edges @xmath150 such that , for all @xmath151 distinct , the vertex set of one component of @xmath152 contains @xmath153 and the other contains @xmath154 . since @xmath87 is displayed by @xmath27 and neither @xmath85 nor @xmath155 , nor @xmath156 is empty , it follows that @xmath157 , which is impossible . + as in the case of full splits , the converse of the above lemma is not true in general .
for example , the set @xmath35 comprising the partial splits @xmath158 , @xmath159 , @xmath160 and @xmath161 is weakly compatible since the sets @xmath162 and @xmath163 are pairwise compatible . yet , as can be easily checked , @xmath35 is not circular .
corresponding to lemma [ pink ] , we have : [ purple ] suppose @xmath133 .
if @xmath42 is displayed by an @xmath5-cycle @xmath27 and @xmath41 , then @xmath35 is also displayed by @xmath27 . as was noted earlier , neither the @xmath3-rule nor the @xmath4-rule preserve weak compatibility in general .
as the next result shows , the situation is different for the special case of circular collections of partial splits .
[ circ ] suppose @xmath40 and @xmath27 is an @xmath5-cycle .
if @xmath42 is obtained from @xmath35 by a single application of either @xmath78 or @xmath54 then @xmath35 is displayed by @xmath27 if and only if @xmath42 is displayed by @xmath27 .
_ proof : _ suppose @xmath40 and @xmath27 is an @xmath5-cycle .
we start the proof with noting that , regardless of whether @xmath42 is obtained from a single application of either @xmath78 or @xmath54 to @xmath35 , @xmath35 is displayed by @xmath27 whenever @xmath42 is displayed by @xmath27 in view of lemma [ purple ] .
conversely , suppose that @xmath35 is displayed by @xmath27 .
assume first that @xmath42 is obtained from @xmath35 by a single application of @xmath78 .
let @xmath164 be the set to which @xmath78 is applied . with @xmath55 , @xmath77 , playing the role of their namesakes in the statement of ( [ sk - rule ] ) , we may assume without loss of generality that none of the three intersections @xmath165 , @xmath166 , and @xmath167 is empty but that @xmath168 .
it suffices to show that the partial split @xmath169 is displayed by @xmath27 . clearly ,
if @xmath170 then @xmath171 and , therefore , @xmath15 is displayed by @xmath27 .
so assume @xmath172 .
then since by assumption @xmath173 , @xmath77 , and @xmath61 and @xmath62 are displayed by @xmath27 , there must exist four distinct edges @xmath150 such that , for all @xmath151 distinct , one component of @xmath152 contains @xmath91 in its vertex set and @xmath174 and the other contains @xmath124 in its vertex set and @xmath175 .
without loss of generality , we may assume that @xmath176 , @xmath177 , that @xmath178 is a vertex ordering of @xmath27 , and that @xmath179 .
furthermore , we may also assume without loss of generality that the component of @xmath180 that contains @xmath181 in its vertex set also contains @xmath63 .
since @xmath182 , and @xmath87 is displayed by @xmath27 there must exist distinct paths @xmath183 and @xmath184 in @xmath27 such that either @xmath185 or @xmath186 ( see figure [ proof ] ) . if @xmath187 then @xmath188 which is impossible .
thus @xmath186 must hold .
suppose @xmath189 are such that when starting at @xmath181 and traversing @xmath27 clockwise @xmath190 is contained in @xmath191 and the next vertex @xmath192 on @xmath27 with @xmath193 is contained in @xmath194 whereas @xmath195 and the next vertex @xmath196 on @xmath27 with @xmath197 is contained in @xmath191 .
let @xmath198 denote the path from @xmath196 to @xmath190 ( taken clockwise ) .
then @xmath147 and @xmath199 are edges on @xmath198 and so @xmath200 .
the choice of @xmath190 and @xmath196 implies @xmath201 and @xmath202 .
hence , the split @xmath203 which is displayed by @xmath27 extends the partial split @xmath15 .
thus @xmath27 displays @xmath15 .
this concludes the proof in case the applied closure rule applied is @xmath78 . to conclude the proof of the proposition suppose @xmath42 is obtained from @xmath35 by a single application of @xmath54 .
let @xmath204 be the set to which @xmath54 is applied . with @xmath55 , @xmath56
, we may assume without loss of generality that @xmath205 and @xmath206 .
if @xmath54 applies trivially to @xmath35 then @xmath207 and so @xmath42 must be displayed by @xmath27 .
if @xmath54 does not apply trivially to @xmath35 it suffices to show that @xmath27 displays @xmath208 .
since @xmath61 and @xmath62 are displayed by @xmath27 there must exist edges @xmath209 such that the vertex set of one of the two components @xmath210 of @xmath152 contains @xmath91 and the other contains @xmath124 , @xmath56 .
put @xmath211 and note that @xmath212 . without loss of generality , we may assume @xmath213 and @xmath214 , @xmath56 .
then , @xmath215 .
since @xmath216 is the vertex set of one of the @xmath217 components of @xmath27 with the edges @xmath218 , @xmath56 removed , it follows that there must exist two distinct edges @xmath219 among the edges @xmath220 so that the vertex set of one of the two components of @xmath221 is @xmath216 .
since @xmath222 is the vertex set of the other component of @xmath221 and @xmath214 , @xmath56 , it follows that @xmath27 displays @xmath208 . this concludes the proof in case @xmath42
is obtained from @xmath35 by a single application of @xmath54 and thus the proof of the proposition .
+ interestingly , the @xmath1-rule does not preserve circularity in general .
an example in point is the @xmath5-cycle @xmath27 with @xmath223 and the natural ordering of the elements of @xmath5 as vertex ordering . then the partial splits @xmath224 and
@xmath225 are clearly displayed by @xmath27 .
yet the @xmath1-rule applied to @xmath118 generates the partial splits @xmath226 and @xmath227 which can not be displayed by @xmath27 .
in this section , we associate to a set @xmath35 of partial splits a split closure sequence and define the last element of such a sequence to be a split closure of @xmath35 .
we also establish a key result for this paper which shows that under certain circumstances a split closure is unique .
suppose @xmath72 is a collection of partial splits that satisfies some partial splits property @xmath228 such as , for example , weak compatibility and @xmath43 is one of the closure rules considered in this paper .
following @xcite , we associate a split closure sequence @xmath229 and a split closure to @xmath35 as follows .
@xmath230 is a strictly increasing ( with respect to @xmath39 ) sequence of sets in @xmath38 so that @xmath231 and , for all @xmath232 , @xmath233 is obtained by one non - trivial application of @xmath43 to @xmath234 whenever @xmath234 satisfies @xmath228 .
note that since @xmath5 is finite , there must exist a last element @xmath235 in @xmath229 such that @xmath235 either satisfies @xmath228 and is closed under @xmath43 or @xmath235 does not satisfy @xmath228 . in the latter case we reset @xmath235 to be a new element @xmath236 .
we refer to @xmath229 as a _ split closure sequence _ for @xmath35 and call @xmath237 the length of @xmath229 .
in addition , we call the last element of @xmath229 a _ split closure _ of @xmath35 . note that in case @xmath238 , @xmath43 applies only trivially to @xmath235
. the following combinations of @xmath228 and @xmath43 are of interest to us : 1 .
@xmath228 is the property that @xmath35 is weakly compatible and @xmath43 is the @xmath3-rule .
@xmath228 is unspecified and @xmath43 is the @xmath4-rule .
@xmath228 is the property that @xmath35 is weakly compatible and @xmath43 is the @xmath239-combination closure rule @xmath240 which applies @xmath54 or @xmath78 to @xmath35 . to elucidate the notion of a split closure sequence and a split closure associated to a set in @xmath38 we next present an example for the assignments of @xmath228 and @xmath43 specified in ( a ) .
consider the set @xmath241 together with the collection @xmath35 comprising of the partial @xmath5-splits @xmath242 , @xmath243 , @xmath244 , and @xmath245 .
clearly , @xmath35 is displayed by an @xmath5-cycle @xmath27 with vertex ordering @xmath246 .
thus @xmath35 is circular and so , by lemma [ circular - twc ] , @xmath35 is weakly compatible
. now @xmath78 applied to @xmath247 generates the split @xmath248 , @xmath78 applied to @xmath249 generates the split @xmath250 and @xmath78 applied to @xmath251 generates the split @xmath252 .
since every subset of @xmath253 of size three contains two pairwise compatible full splits , @xmath78 can only be applied trivially to @xmath42 .
hence , the sequence @xmath254 , @xmath255 , @xmath256 , @xmath42 is a split closure sequence for @xmath35 of length @xmath257 and @xmath42 is a split closure for @xmath35 . regarding ( c ) ,
it should be noted that even if for some @xmath72 two distinct split closure sequences have the same length and terminate in the same element @xmath258 one of them might utilise fewer applications of @xmath78 ( and thus more applications of @xmath54 ! ) than the other . for the previous example ,
one way to construct two such sequences is to exploit the following relationship between the @xmath3-rule and the @xmath4-rule for @xmath259 .
[ relationship ] suppose @xmath260 is such that @xmath261 and @xmath262 . if the @xmath3-rule applies to @xmath35 then @xmath263 _ proof : _ assume that @xmath35 and @xmath84 and @xmath91 , @xmath77 , are such that the assumptions of the proposition are satisfied
. then @xmath170 .
combined with the assumption that @xmath78 applies to @xmath35 , it follows that either ( [ sk - rule ] ) is satisfied with @xmath91 , @xmath77 playing the roles of their namesakes in the statement of ( [ sk - rule ] ) or ( [ sk - rule ] ) is satisfied with @xmath191 playing the role of @xmath194 and @xmath91 playing the role of @xmath91 , @xmath56 , in that statement .
but the latter alternative can not hold since this implies @xmath264 whereas the assumption @xmath265 implies @xmath266 hence , ( [ sk - rule ] ) is satisfied with @xmath91 , @xmath77 playing the roles of their namesakes in the statement of ( [ sk - rule ] ) .
let @xmath267 , @xmath77 be as in the statement of the @xmath3-rule
. then @xmath268 implies @xmath269 . and
since @xmath261 , we have @xmath270 and @xmath170 which in turn implies @xmath271 . consequently , @xmath272 to observe the remaining set equality , note that since ( [ sk - rule ] ) is satisfied with @xmath91 , @xmath77 playing the roles of their namesakes in the statement of ( [ sk - rule ] ) neither @xmath273 nor @xmath274 can be empty .
let @xmath275 be as in the statement of the @xmath4-rule .
then @xmath261 implies @xmath269 and @xmath276 .
this implies the sought after set equality and thus proves the proposition .
+ clearly independent of which one of the rules @xmath78 , @xmath54 or @xmath277 is applied , a split closure sequence must always be finite since @xmath5 is finite .
in addition and by applying the same arguments as semple and steel in @xcite one can show that , for the assignments of @xmath228 and @xmath43 as described in ( a ) , the length of a split closure sequence for a weakly compatible set @xmath72 is bounded from above by @xmath278 .
we start with a lemma that is crucial for showing that the split closure of some collection @xmath72 is unique in any of the three combinations for ( @xmath183 ) and @xmath43 stated in ( a ) ( c ) .
[ key ] suppose @xmath72 , @xmath279 is a split closure of @xmath35 and @xmath280 and @xmath281 are two consecutive elements in a split closure sequence for @xmath35 . 1 .
if @xmath280 is weakly compatible , @xmath282 and @xmath281 is obtained from @xmath280 by one application of @xmath78 , then @xmath281 is weakly compatible and @xmath283 .
2 . if @xmath281 is obtained from @xmath280 by one application of @xmath54 and @xmath284 , then @xmath283 . _
proof : _ suppose @xmath285 are as in the statement of the lemma .
\(i ) assume @xmath281 is obtained from @xmath280 by applying @xmath78 to some set @xmath247 contained in @xmath286 . for @xmath77 and with @xmath287 playing the role of their namesakes in the statement of ( [ sk - rule ] )
, we obtain @xmath288 it follows that @xmath289 since @xmath284 , there exist partial splits @xmath290 with @xmath291 extending @xmath84 , @xmath77 . without loss of generality we may assume for all @xmath292 that @xmath293 and @xmath294 .
since @xmath295 is weakly compatible , ( [ sk - rule ] ) is satisfied by @xmath296 with @xmath297 . since
, by assumption , @xmath298 and so @xmath78 applies trivially to @xmath295 , we must have @xmath299 , @xmath300 , and @xmath301 .
it follows that , for all @xmath77 , @xmath267 is extended by @xmath291 which in turn implies @xmath302 .
since @xmath279 and so @xmath295 is weakly compatible , lemma [ pink ] implies that @xmath281 is weakly compatible .
( ii ) : suppose @xmath281 is obtained from @xmath280 by applying @xmath54 to some set @xmath303 .
put @xmath304 , @xmath56 , and assume without loss of generality , that @xmath205 and @xmath305 .
then @xmath306 and so @xmath307 by assumption , @xmath284 and so there exist partial splits @xmath308 , @xmath309 in @xmath295 with @xmath291 extending @xmath84 , @xmath56 . without loss of generality we may assume @xmath293 and @xmath294 , @xmath56 .
then @xmath310 and @xmath311 .
since , by assumption , @xmath279 and so @xmath54 only applies trivially to @xmath295 we have @xmath312 . hence , there exist partial splits @xmath313 , @xmath314 , so that @xmath315 is extended by @xmath87 and @xmath316 is extended by @xmath317 . without loss of generality , we may assume that @xmath318 and @xmath319 and that @xmath320 and @xmath321 .
then @xmath322 and so @xmath87 extends @xmath275 .
similarly , it follows that @xmath317 extends @xmath323 .
thus , @xmath302 .
+ with this result in hand , we are now in the position to present a key result .
[ split - closure ] suppose @xmath72 .
then any two split closures for @xmath35 are the same if 1 .
@xmath35 is weakly compatible and solely the @xmath3-rule is used to obtain a split closure for @xmath35 , 2 .
solely the @xmath4-rule is used to obtain a split closure for @xmath35 , or 3 .
@xmath35 is weakly compatible and solely the @xmath239-rule is used to obtain a split closure for @xmath35 . _
proof : _ suppose @xmath72 .
we start with remarking that we prove statements ( i ) , ( ii ) and ( iii ) collectively as the proof of all three statements relies on an inductive argument on the length of a split closure sequence for @xmath35
. however , since the arguments for the inductive step differ under the assumptions made in ( i ) , ( ii ) and ( iii ) , we discuss each inductive step separately .
suppose that the assumptions made in ( i ) or in ( ii ) or in ( iii ) hold .
if every split closure of @xmath35 is @xmath324 then the theorem holds trivially .
so we may assume that there exists a split closure @xmath295 of @xmath35 with @xmath279 .
we proceed by showing that every other split closure of @xmath35 must equal @xmath295 .
suppose that @xmath325 is a split closure sequence of @xmath35 .
we now use induction on @xmath237 to show that if @xmath35 satisfies the assumptions made : 1 . in ( i ) then , for all @xmath326 , @xmath327 2 . in
( ii ) then , for all @xmath326 , @xmath328 3 . in ( iii )
then , for all @xmath326 , @xmath329 we start with assuming that @xmath35 satisfies the assumptions made in ( i ) , that is , @xmath35 is weakly compatible and solely the @xmath3-rule is used to generate the elements of @xmath229 .
if @xmath330 then ( [ closure - claim1 ] ) obviously holds since then @xmath331 and @xmath332 satisfies the properties stated in ( [ closure - claim1 ] ) .
now suppose that ( [ closure - claim1 ] ) holds for some @xmath333 .
then , by lemma [ key](i ) , @xmath233 is weakly compatible and @xmath334 .
this completes the induction step and thereby establishes ( [ closure - claim1 ] ) .
next , assume that only the @xmath4-rule is used to generate the elements in @xmath35 .
if @xmath330 , then ( [ closure - claim2 ] ) holds since then @xmath331 and @xmath332 satisfies ( [ closure - claim2 ] ) .
assume that ( [ closure - claim2 ] ) holds for some @xmath333 .
then lemma [ key](ii ) implies @xmath335 which completes the induction step and thereby establishes ( [ closure - claim2 ] ) .
finally , assume that @xmath35 satisfies the assumptions made in ( iii ) , that is , @xmath280 is weakly compatible and only the @xmath239-rule is used to generate the elements in @xmath229 . if @xmath330 , then ( [ closure - claim3 ] ) obviously holds since then @xmath331 and @xmath332 is closed under @xmath277 .
now suppose that ( [ closure - claim3 ] ) holds for some @xmath333 .
then @xmath234 is weakly compatible , @xmath284 , and one of the following two cases must hold . either
( a ) @xmath281 is obtained from @xmath280 by applying @xmath78 or ( b ) @xmath281 is obtained from @xmath280 by applying @xmath54 . if case ( a ) holds , the proof of the inductive step in ( i ) implies that @xmath233 is weakly compatible and @xmath336 .
if case ( b ) holds , @xmath336 follows from the proof of the inductive step in ( ii ) . that @xmath233 is weakly compatible follows from lemma [ key](ii ) and the fact that @xmath295 is weakly compatible .
this completes the induction step and thereby establishes ( [ closure - claim3 ] ) .
we conclude with noting that for @xmath337 , we obtain @xmath338 regardless of whether we are assuming ( i ) or ( ii ) or ( iii ) to hold . in case of ( i ) holding and applying ( [ closure - claim1 ] ) to @xmath337 or ( iii ) holding and applying ( [ closure - claim3 ] ) to @xmath337
, we see that @xmath235 is weakly compatible . by interchanging the roles of @xmath235 and @xmath295
, we deduce @xmath339 .
thus , under the assumptions made in ( i ) or ( ii ) or ( iii ) , we have @xmath340 which concludes the proof of the theorem . + extending in the case of ( @xmath183 )
denoting the condition `` @xmath35 is weakly compatible '' and @xmath43 denoting either the @xmath3-rule or the @xmath239-rule , the definition of the split closure to non weakly compatible sets in @xmath341 by defining the split closure of such sets to be @xmath324 , we obtain [ summary ] suppose @xmath43 is either the @xmath3- or @xmath4- or @xmath239-rule . then for any @xmath72 ,
any two split closures for @xmath35 obtained via @xmath43 are the same . bearing in mind corollary [ summary ] , we denote for @xmath342 the split closure of a set @xmath72 by @xmath343 , note that @xmath344 but that neither @xmath345 nor @xmath346 have to hold . also note that , if we denote the collection @xmath347 by @xmath348 , define @xmath349 for some closure rule @xmath43 , and put @xmath350 for all @xmath72 , then the split closure with respect to @xmath351 satisfies the usual properties of a closure operation .
more precisely , for all @xmath352 we have @xmath353 , if @xmath354 then @xmath355 , and @xmath356 . as an immediate consequence of lemma [ circular - twc ] , proposition [ circ ] , and theorem [ split - closure ] ,
we obtain our main result which we state next .
[ cyclosure ] suppose @xmath72 and @xmath27 is an @xmath5-cycle
. then @xmath35 is displayed by @xmath27 if and only if @xmath357 is displayed by @xmath27 . in that case @xmath358 and @xmath359 are also displayed by @xmath27 .
we conclude this section with remarking that , in general , not all elements in @xmath360 need to be full splits on @xmath5 , @xmath72 circular .
however , it is reasonable to assume that those that have been extended to full splits on @xmath5 contain phylogenetically relevant information and programs such as e.g.splitstree4 @xcite may be employed to produce a circular phylogenetic network that displays them . for the following we refer to the combination of the @xmath239-rule with a phylogenetic network generation package such as splitstree4 as the _ @xmath361-closure approach_.
although a detailed analysis of this approach is beyond the scope of this paper and will be presented elsewhere , we note that the my - closure approach can not be polynomial in the worst case since if the collection of partial splits comprises of all @xmath362 partial @xmath5-splits @xmath363 with @xmath364 and @xmath365 then @xmath277 will generate all @xmath366 splits of @xmath5
one of the most fiercely debated questions amongst biologists is the origin of eukaryotes ( essentially cells that have a nucleus and organelles ) @xcite .
the main reason for this is that eukaryotes have eubakteria - like genes as well as archaebacteria - like genes making it very difficult to establish the evolutionary relationships between eukaryotes and prokaryotes ( essentially cells that lack nucleus and organelles ) which is the collective name for eubakteria and archaebacteria . to help shed light into this question , rivera et al .
@xcite analysed 10 bacterial genomes .
the 5 most probable phylogenetic trees resulting from their analysis are presented in fig . 1 of that paper .
for the convenience of the reader , we depict them in slightly different form in fig .
note that the collection of splits displayed by these trees is circular and also that , when ignoring the fact that the leaves are marked with different symbols , the last 2 trees are the same . using a technique called _ conditioned reconstruction _
@xcite , rivera et al . constructed the phylogenetic network depicted in fig .
[ lake - network ] with the degree 5 interior vertex plus all its incident edges removed and all resulting degree 2 vertices suppressed .
the resulting structure they then interpreted as lending support to the idea that , in its early stages , evolution was not tree - like but rather more like a ring ( hence the term `` ring of life '' ) with the eukaryotic genome being the result of a fusion of 2 diverse procaryotic genomes @xcite . to find out how dependent rivera et al.s ring of life is on the fact that all 5 trees are on the same leaf set , we randomly removed pairs of leaves plus their incident edges ( suppressing resulting degree 2 vertices and always ensuring that there were no 2 trees from which the same pair of leaves was removed ) resulting in 5 trees @xmath367 on 5 leaves . perhaps not surprisingly , we found that , in general , removal of pairs of leaves did not allow us to recover rivera et al.s ring of life .
the exception being the trees depicted in fig . [ ring ] with the leaves marked by a filled - in square removed .
for these 5 trees the associated phylogenetic network @xmath368 produced by the @xmath361-closure approach is depicted in fig .
[ lake - network](left ) .
in addition and with the exception of one instance where one split in @xmath369 , that is , the set of all splits displayed by @xmath367 , was not extended to a full split by our closure rules and thus was not displayed by @xmath368 our rules always generated a minimum collection of splits so that @xmath368 displayed all the splits in @xmath369 .
interestingly , both the @xmath1-closure super - network and q - imputation approach seemed to struggle with this example with , in the case of @xmath1-closure super - network , either yielding a very complex network @xmath368 in which numerous extensions of one and the same split in @xmath369 was displayed ( see fig .
[ lake - network](right ) ) or @xmath368 displayed only a subset of splits in @xmath369 ( q - imputation ) . + * acknowledgment * the authors thank the referees for their helpful comments .
fitzpatrick , d. , creevey , c. , mcinerney , j. , 2006 .
genome phylogenies indicate a meaningful alpha - proteobacterial phylogeny and support a grouping of the mitochondria with the rickettsiales , mol .
evol . , 23 , 74 - 85 . meacham , c.a . , the recovery of trees from measures of dissimilarity , in : mathematics in the archaeological and historical sciences , f.r .
hodson , d. g. kendall , and p. tatu ( eds ) , edinburgh university press ( 1971 ) , 387- 395 .
|
a contemporary and fundamental problem faced by many evolutionary biologists is how to puzzle together a collection @xmath0 of partial trees ( leaf - labelled trees whose leaves are bijectively labelled by species or , more generally , taxa , each supported by e.g.a gene ) into an overall parental structure that displays all trees in @xmath0 .
this already difficult problem is complicated by the fact that the trees in @xmath0 regularly support conflicting phylogenetic relationships and are not on the same but only overlapping taxa sets .
a desirable requirement on the sought after parental structure therefore is that it can accommodate the observed conflicts .
phylogenetic networks are a popular tool capable of doing precisely this .
however , not much is known about how to construct such networks from partial trees , a notable exception being the @xmath1-closure super - network approach and the recently introduced @xmath2-imputation approach . here
, we propose the usage of closure rules to obtain such a network .
in particular , we introduce the novel @xmath3-closure rule and show that this rule on its own or in combination with one of meacham s closure rules ( which we call the @xmath4-rule ) has some very desirable theoretical properties .
in addition , we use the @xmath4- and @xmath3-rule to explore the dependency of rivera et al.s `` ring of life '' on the fact that the underpinning phylogenetic trees are all on the same data set .
our analysis culminates in the presentation of a collection of induced subtrees from which this ring can be reconstructed .
k.t.huber + school of computing sciences , + university of east anglia , + norwich , nr5 7tj , + united kingdom .
+ email : [email protected] + fax : + 44 ( 0 ) 1603 593345 + we would prefer to submit the final manuscript in latex .
|
nowadays carbon fiber is a ubiquitous structural material used in a wide range of applications due to its high modulus and low mass .
it offers great flexibility in terms of tuning thermal and mechanical properties through the orientation and number of lay - ups of the fibers .
carbon fiber support structures have also become common elements of detector designs for high energy physics experiments and are especially prevalent in the design of silicon vertex detectors .
early examples of the use of carbon fiber in external support structures are the delphi silicon detector @xcite , and the space frames for the babar @xcite and cdf isl silicon detectors @xcite . in current designs ,
carbon fiber is integrated in the design of silicon detector modules , like the readout modules for the silicon tracker for the cms experiment @xcite .
an example of a detector with a fully integrated carbon fiber support structure is the layer 00 detector for the cdf collaboration at the fermilab tevatron collider @xcite .
while the use of carbon fiber solves a variety of mechanical problems , it presents a challenging set of electrical concerns .
highly conductive carbon fiber surfaces produce an undesirable capacitance relative to sensors , electronics and cables and can compromise the detector performance due to significant coherent noise pickup @xcite . well - designed coupling and
grounding schemes are essential for producing low - noise environments for these detectors . here
we describe the electrical characteristics of carbon fiber as applicable to a silicon tracker at small radius with an integrated carbon fiber support structure .
measurements are performed on a prototype of a detector that has been proposed for the d experiment at the fermilab tevatron collider , called the layer zero ( l ) detector @xcite . in the next section
the electrical characteristics of carbon fiber will be described .
the carbon fiber test pieces were made using carbon fiber epoxy resin prepregs using k139 , obtained from bryte @xcite , and k13c , obtained from yla @xcite .
the carbon fiber in both products was manufactured by mitsubishi .
k139 uses 110 msi modulus fiber with a thermal conductivity of 210 w / mk , with a fiber aerial weight ( faw ) of 55 g / m@xmath0 .
k13c has a modulus of 130 msi , with a significantly higher thermal conductivity of 620 w / mk and a faw of 69 g / m@xmath0 @xcite . for both prepregs ,
the resin fraction is about 35% by weight , or roughly 50% by volume before curing .
based on the results of the electrical tests of the carbon fiber , methods will be outlined that address the electrical requirements for the construction of support structures for silicon detectors .
the design addresses all potential noise sources .
tests aimed at minimizing coherent noise contributions with a full - scale silicon detector , populated with a limited number of readout modules will be described .
the paper concludes with a suggested set of design rules which need to be respected in the construction of silicon detectors with integrated carbon fiber supports .
carbon fiber with ultra - high modulus ( @xmath11000 gpa ) also has low resistivity ( 100 @xmath2 cm ) @xcite .
these materials are characterized by particularly high conductivity at high frequency with associated design issues .
a series of tests were carried out to verify the conductivity of carbon fiber .
the first study performed was a measurement of capacitor impedance .
two parallel plate capacitors were built using @xmath3 bare fr4 cores , with @xmath4 , approximately @xmath5 thick .
the first capacitor , used for baseline measurements , had two tinned copper electrodes laminated to either side of the fr4 dielectric .
each electrode is approximately @xmath6 thick .
the second capacitor has a single tinned copper electrode attached to one side of the fr4 dielectric and a single carbon fiber electrode attached to the other side .
the carbon fiber electrode was four plies of k139 fiber with a 4-layer @xmath7_{s}$ ] lay - up , approximately @xmath8 thick .
contact with the two electrodes of the capacitor under measurement was established using two 20awg stranded wires connected to the two contacts of a bnc connector .
for the capacitor electrode side , the strands were separated and splayed to maximize contact and taped to the electrode with a specific area of copper tape .
prior to the attachment of the probe , the contact area on the electrode was burnished with mildly abrasive polishing material ( scotch - brite ) and cleaned with ethyl alcohol .
after taping the probe to the electrode , the tape was worked with a blunt tool to maximize contact to the strands of the probe and to force air from under the copper tape to the outside edge to maximize contact to the electrode .
past experience has shown that not taking care at this stage in the test preparation would result in inconsistent measurements .
the size of the copper tape used to make contact with the carbon fiber electrode was varied to measure the effect of the fraction of contact area required to obtain a given impedance response .
impedance measurements were then made using an hp4193a vector impedance meter at specific frequencies over a range of 0.4 to 110 mhz .
the measurements were performed in a temperature and humidity controlled environment .
the measured impedance and phase shift is shown in fig .
[ fig : imp ] for six different capacitors : the reference capacitor with two copper electrodes , and the capacitors with one copper electrode and one carbon fiber electrode with a copper tape contact area of 1 , 2 , 4 , 9 and 36 square inches .
the inset shows a blow - up of the region around the minimum . for low frequencies , the behavior is the expected @xmath9 dependence . in this frequency range , from 0.4 to 50 mhz , the response of all capacitor configurations is virtually identical . with increasing frequency the impedance decreases rapidly and reaches a minimum with vanishing phase shift . at this point
the impedance is real and purely resistive .
the reference capacitor with two copper electrodes reaches the lowest impedance of all configurations .
the capacitor configurations with one carbon fiber electrode with a copper area of 4 , 9 and 36 @xmath10 are virtually indistinguishable . plotting the value of the minimum of the impedance from fig .
[ fig : imp ] as function of fractional copper contact area of the carbon fiber gives the result shown in fig .
[ fig : imp_area ] .
the frequency at which the impedance minimum occurred was at 84@xmath111 mhz for all capacitor configurations tested .
the impedance of the reference capacitor , with dual copper electrodes , is 1.02 @xmath2 and is indicated by the solid line .
the results show that at high frequencies , the frequencies of interest for silicon vertex detectors , carbon fiber is to be regarded as a conductor .
furthermore , a 15 - 20% area coverage of the carbon fiber with a good conductor should provide adequate electrical coupling to the carbon fiber .
the measurements were also performed by placing the capacitors in a bandstop circuit and measuring the power transfer function with an hp3577a network analyzer as function of frequency .
the results are in complete agreement with the results quoted above . to verify that the carbon fiber was not acting as part of the dielectric
, another measurement was taken with the carbon fiber completely removed and with the 1 by 2 inches copper tape directly attached to the fr4 sheet .
the frequency response of the impedance for this configuration , together with the response of the reference capacitor with two copper electrodes , is shown in the upper half of fig .
[ fig : imp_nocf ] .
the upper impedance curve corresponds to the capacitor with the carbon fiber removed and replaced with copper tape with an area of 2 @xmath10 .
the difference between the measured impedance for the two configurations , compared to the frequency response with the carbon fiber sheet ( see fig . [
fig : imp ] ) is striking and clearly shows that the carbon fiber does not act as part of the dielectric
. the lower half of fig .
[ fig : imp_nocf ] shows the capacitance for the two configurations as function of frequency , corrected for the contribution from the test setup .
given an area ratio of 18 for the two configurations ( 36 @xmath10 versus 2 @xmath10 ) the ratio of the two capacitance values is naively expected to be a factor of 18 .
a ratio of 11.9 is measured , however . to understand the measurement , a full ansys calculation
was performed @xcite .
using ansys , the capacitance of the 6 by 6 inch capacitor with the copper electrodes was calculated to solve for the dielectric constant of fr4 .
based on the measured value of @xmath12 pf , the ansys calculation gives a value for the dielectric constant of @xmath13 .
using the simple analytic formula for the capacitance of area over separation , a value of 3.9 is obtained .
ansys predicts a capacitance of 8.0 pf for the capacitor with the 2 @xmath10 copper electrode , whereas the measured value , corrected for capacitance of the leads , is @xmath14 pf . the contributions to the error estimate are a 10% error on the measurement itself , a 10% error on the correction due to the test fixture , and a 5% error on the extraction of the dielectric constant , added in quadrature . the measured ratio of capacitances of 11.9 @xmath11 1.9 is in reasonable agreement with the predicted ratio of 14.6 , validating our conclusion that carbon fiber does not act as part of the dielectric .
these studies show that carbon fiber should be regarded as a conductor for the frequencies above one mhz .
using highly conductive carbon fiber as a mechanical support thus has the potential to exhibit strong capacitive coupling to the silicon sensors it supports , which can lead to very troublesome sources of noise , especially in close - packed structures .
it is therefore imperative that all the carbon fiber in the detector is effectively shorted to the hybrid or bias filter grounds to prevent noise transmission through capacitive coupling to the sensor readout .
it has been demonstrated that copper in contact with the carbon fiber with more than 10 - 15% area coverage provides adequate electrical coupling . in the next section methods to apply our findings to the construction of silicon detector modules will be explored .
in the design of the silicon detector it is envisioned that the sensors will be mounted directly on the support structure and that a low - pass filter card for the bias high voltage and bias return is glued directly onto the sensor .
the ground plane of the sensor and the carbon fiber support structure then form the plates of a parallel plate capacitor .
this puts the sensors between two plates of a capacitor , and any power transfer through the capacitor caused by unequal potential at its plates will generate noise on the sensor .
the challenge is to devise a low mass short between the carbon fiber and ground plane which is highly efficient over a frequency range of 5 khz - 50 mhz . to determine the feasibility of shorting the carbon fiber support cylinder , on which the silicon sensors are mounted , to the hybrid or filter ground plane ,
a series of tests were carried out with a mockup of the real design .
a mock - up of the carbon fiber support / sensor / ground plane configuration was built ( see fig . [
fig : mockup ] ) .
two pieces of 0.5 mil aluminum ( indicated by ` sensor ( al ) ' in fig . [
fig : mockup ] ) representing two sensors , were attached to a strip of kapton , that was mounted directly onto a carbon fiber support structure .
this carbon fiber support piece has a 6-layer @xmath15_{s}$ ] lay - up of k13c high modulus carbon fiber .
it should be noted that k13c is even more conductive than k139 .
a piece of plastic , mimicking the filter card or hybrid , was laid on top of the ` sensors ' . a thin 0.5 mil aluminum foil
was then attached to the plastic representing the ground plane ( indicated by ` hybrid ( al ) ' in fig .
[ fig : mockup ] ) .
the carbon fiber support was driven with the source power from the network analyzer .
the network analyzer input was wired to the sensor aluminum layer and both the source and input grounds were connected to the filter plane ( ` hybrid ( al ) ' in fig . [
fig : mockup ] ) .
power transfer functions were measured with different configurations for shorting the carbon fiber to the filter plane .
one to four @xmath16 wide strips of 1 mil thick aluminized mylar , 0.5 mil aluminum , or 1 mil copper grounding strips were attached between the kapton and the carbon fiber .
the aluminized mylar had a 250 - 300 thick layer of pure aluminum .
short tabs of the same material were attached to the grounding strips and these strips were wrapped around the ` sensors ' and attached to the filter / hybrid ground plane ( see fig .
[ fig : mockup ] ) .
first , a reference power transfer function was measured with no grounding strips attached .
next , the grounding strips were attached one by one , varying the area of the carbon fiber covered with grounding strips , and the power transfer was measured again .
it was found that aluminized mylar strips were minimally effective in reducing power to the sensor .
aluminum and copper strips performed equally well .
the power reduction was independent of the number and size of strips , but proportional to the area used to couple the strips to the carbon fiber .
we achieved a maximum reduction in excess of 40 db at 1 mhz with copper tape covering about 11% of the carbon fiber , identical to the result shown in the previous section ( see fig .
[ fig : imp_area ] ) .
because copper tape coupling would not be feasible in the detector due to mass constraints , the next coupling test was carried out using carbon fiber pieces with embedded aluminum foil and aluminized mylar .
the aluminum or mylar was attached to the carbon fiber by co - curing it with the carbon fiber .
a new sensor / hybrid mockup was built and mounted on different pieces of k13c carbon fiber which had various amounts of surface area covered with embedded aluminum or aluminized mylar .
the embedded grounding material , as before , extends beyond the sensors so that grounding tabs can be attached to it and to the hybrid / filter plane , as sketched in fig .
[ fig : mockup ] .
it was noticed that aluminized mylar grounding produced significantly less attenuation than aluminum ; about 10 db less below 20 mhz .
this test also verified that the attenuation is not affected by the width of the grounding strips .
measurements of transfer functions were taken with different areas of embedded aluminum and equal sized grounding strips .
those data confirmed that the attenuation increases with contact area to the carbon fiber .
the maximum area of embedded aluminum tested was about 4 in@xmath0 ( again , about 11% of the area ) , giving a power decrease of at least 30 db for all frequencies below 50 mhz .
although coupling to carbon fiber through embedded aluminum has thus been shown to be an effective grounding technique , the use of aluminum does not represent a robust solution .
the aluminum is very susceptible to oxidation and the likely choice for contacts to the ground plane would be silver epoxy , which historically has not been particularly reliable .
this has led us to the design which employs a kapton flex circuit with a copper mesh , embedded in the carbon fiber surface . to obtain a more robust design we opted for a configuration where the kapton covers the surface area completely , with the embedded copper mesh having only a limited area coverage .
the design of the l detector calls for the silicon sensors to be mounted directly on a full length carbon fiber support structure at a radius of about 17 mm . sensors with @xmath17-implants on an @xmath18-type substrate with 50 @xmath19 readout pitch , 8 cm long , produced by elma @xcite , were used in the study .
the sensors have 256 readout strips .
a low - pass filter card of 150 @xmath20 m thick g-10 for the bias high voltage and bias return is glued directly onto the sensor , close to the wire bonding pads .
often the hybrid with the front - end readout electronics is mounted directly on the sensor . in this design , due to space constraints and heat dissipation ,
the signals from the silicon sensors are transmitted to the hybrid containing the readout chips by flexible circuits up to 435 mm long . since these circuits carry analogue signals they are referred to as ` analogue cables ' .
although this solution is very attractive because of reduced material and heat generation in the sensitive volume , it represents a considerable technical challenge .
the addition of the analogue cable reduces the signal to noise of the silicon sensors due to the added capacitive load . to minimize the capacitance between the traces of the analogue cable , and thus optimize the s / n ratio , our design uses a stack of two cables with constant 91 @xmath20 m pitch , laminated with a lateral shift of 45 @xmath20 m .
the pitch of the cable stack thus matches the pitch of both the readout strips on the sensor and of the input bond pads on the front - end readout chip .
it should be noted that with this configuration adjacent readout channels alternate between the top and bottom cable in the stack .
stated in a different way , all odd readout channels are connected to one cable in the stack and all even channels to the other cable . for our configuration ,
the first channel is channel 1 , and all odd channels ( 1 , 3 , 5 , @xmath21 ) correspond to the top cable in the stack . in order to reduce the capacitance contribution from the adjacent cable within the stack of two cables , a spacer consisting of 150 @xmath20 m thick polypropylene mesh with about 25% volume occupancy
is placed between the two cables . after a series of prototypes ,
flawless cables were produced by dyconex @xcite .
the capacitance as measured for this arrangement is 0.36 pf / cm in good agreement with ansys calculations . in our setup
we used 42 cm long prototype cables .
each cable has 129 signal traces with 91 @xmath20 m pitch .
each trace is 16 @xmath20 m wide copper on a 50 @xmath20 m thick kapton substrate .
one trace is reserved as a spare trace in case of an open .
two additional , wider traces ( 100 @xmath20 m ) are provided on the cable for the bias voltage and its return .
the analogue cable is wirebonded to a ceramic hybrid containing two svx4 readout chips @xcite with 128 input channels each .
the chip operates at 53 mhz .
the equivalent noise charge ( enc ) of the svx4 chip is @xmath22 , at a fixed risetime of 69 ns .
a full - size prototype crenellated two - part support structure was made using k13c fiber .
it consists of a twelve - sided inner shell and a six - sided outer shell .
the inner shell has a 4-layer @xmath7_{s}$ ] lay - up of k13c high modulus carbon fiber .
the outer shell has a 6-layer @xmath15_{s}$ ] lay - up of the same high modulus fiber .
the inner shell extends beyond the length of the outer shell and forms the support structure for the hybrids .
the outer shell only supports the sensors ( see fig .
[ fig : solidmodel ] ) .
a 0.025 mm kapton sheet with an embedded copper mesh was co - bonded to the outer shell for grounding connections .
the copper mesh , 5 @xmath20 m thick , which is in direct contact with the carbon fiber , has an area coverage of 30% .
round copper pads , one for each sensor , are provided on the top side of the mesh .
the pads are nickel - gold plated , with a thickness of 1.3 @xmath19 and 0.3 @xmath19 , respectively .
four vias , with 0.889 mm diameter holes establish the electrical connection between the pad and the copper mesh .
four vias were chosen to lower the inductance and for redundancy .
the top drawing in fig .
[ fig : cable ] shows one end of the copper - mesh cable , produced by compunetics @xcite .
an insulating two - layer kapton flex circuit , shown in the lower drawing in fig .
[ fig : cable ] , is glued to the backside of the sensors , covering the backside completely .
this circuit , also produced by compunetics , has one tab on either side that wraps around the sensor and is attached to the filter card mounted on the sensor .
the right angles in the drawing indicate where the kapton is trimmed , so that the full backside of the sensor is covered by kapton . after the backside lamination
, the sensor is mounted on the support structure with the embedded mesh .
the tab connected to the round pad in the bottom drawing in fig .
[ fig : cable ] picks up the bias voltage from the filter card and directly applies it to the backside of the sensor .
the small rectangular pad connected to the larger rectangular pad is the ground connection .
the smaller rectangular pad has three 0.889 mm vias which establish the connection to the co - bonded copper mesh of the outer shell of the support structure .
[ fig : flex ] shows a schematic cross section of a sensor mounted on the support structure to aid in the understanding how all electrical connections are established .
the twelve - sided inner shell of the support structure is longer than the crenellated outer shell and forms the support structure for the hybrids .
a separate kapton - copper mesh , with the same specifications as the circuit for the sensor support , was embedded in the carbon fiber to provide the ground for the hybrids . at the transition between the sensor and hybrid support regions the two copper mesh circuits
can be connected electrically through a small jumper mesh cable .
it should be noted that in our initial tests , described below , the two ground mesh circuits were not connected .
[ fig : solidmodel ] shows a detail of the solid model of the transition region between sensor and hybrid support regions .
the silicon sensors are mounted on the outer shell with the embedded grounding copper - kapton mesh using epoxy .
the @xmath17-side of the sensor was grounded to the carbon fiber support structure through the embedded copper - clad kapton flex circuit as described above .
the rectangular ground tab on the ` wrap - around ' flex circuit was attached to the biasing filter card , that was glued on top of the sensor , and the round pad on the backside sensor lamination picks up the sensor bias voltage ( see fig .
[ fig : flex ] ) .
the sensor bias line was wirebonded to the ground pad on the bias filter card .
the two - layer analogue cable was wirebonded to the sensor at one end and to the ceramic hybrid with two svx4 chips at the other end .
the hybrid was mounted on the hybrid support structure .
a photograph of a detail of the silicon module on the support structure is shown in fig .
[ fig : photo ] . although the hybrid can accommodate two readout chips , only the first readout chip was wirebonded .
the first 64 channels of the readout chip were only connected to the analogue cable and the second 64 channels were also connected to the sensor . of this second set of 64 channels ,
the last 30 channels were inactive .
the dc resistance of the 47 cm long sensor support structure was 12 @xmath2 and the resistance of the ground trace on the analogue cable was 20 @xmath2 .
the whole assembly was mounted inside an aluminum faraday cage .
the faraday cage was placed on copper - clad g10 which was connected to the ground of the readout system and of the hv power supply .
the electrical connection between the copper - clad g10 and the faraday cage was quite good ; the dc resistance was measured to be less than 1 @xmath2 .
first , the test results of the l module before installing it on the prototype support structure are given .
this result serves as a reference for the effect of adding the support structure .
we define the noise on the output of a given channel as the standard deviation @xmath23 of the digital output in adc counts for that channel : @xmath24 in practice , the quantity @xmath25 is estimated by computing the root mean square value of the digital output of the channel under consideration over a large number of events ( typically 1000 ) . because we are interested in the susceptibility of the system to external noise sources , we also define the `` differential '' noise of a given channel as @xmath26 of the standard deviation of the difference between the output signal @xmath27 for that channel and the digital output signal @xmath28 of a neighboring channel : @xmath29 again , in practice the quantity @xmath30 is estimated by computing the root mean square value of @xmath31 over a large number of events . assuming that @xmath32 , i.e. that the channel under consideration and its neighbor have the same noise , we have @xmath33 where @xmath34 is the correlation coefficient between @xmath27 and @xmath28 , @xmath35 .
the differential noise is thus equal to the total noise if there is no correlation between the output signals of neighboring channels .
a positive correlation can be due to a common component of noise induced by external pickup and is the object of study .
due to capacitive coupling between neighboring strips in the silicon sensor a negative correlation could be present as well at the few percent level @xcite .
this effect has been neglected here . when the total noise of the device is equal to the differential noise and is the same as the value obtained on the test bench in a faraday cage , the performance of the system is considered to be optimal .
the silicon module was tested on the bench inside a faraday cage .
the total noise before installation on the carbon fiber support structure was measured to be 2.7 adc counts for channels connected to both the sensor and the analog cable with a vanishing correlation coefficient .
that is , there was no significant difference between total and differential noise .
the bandwidth setting of the svx4 was such that the 10 - 90% risetime was measured to be 35 ns and 65 ns for 10 pf and 33 pf capacitive load , respectively .
the integration time of the preamplifier was 132 ns .
the resulting risetime is fast enough to collect all charge for the capacitive loads used in the measurements .
all results are quoted in adc counts .
for the parameter settings used for our tests the gain of the svx4 chip is approximately 700 electrons per adc count .
the measured noise of 2.7 adc counts , or 1890 @xmath36 , is in reasonable agreement with the expected noise , given a strip load capacitance of 1.4 pf / cm .
the module was then installed on the support structure and the whole assembly was placed in a faraday cage .
the ground connection between the hybrid and the support structure was established by soldering a wire from the hybrid ground pad to the copper strip on the embedded copper - kapton mesh circuit , embedded in the support structure .
the support structure itself was connected to the faraday cage with copper tape .
one end of a copper strip was taped to the chassis of the faraday cage , and the other end to the gold pads on the kapton flex embedded in the support structure .
the measured noise level was about 14 adc counts , considerably higher than the performance on the bench .
this indicates a persisting high impedance ground connection .
both resistance and inductance contribute to the impedance , but for high frequencies the inductance often dominates .
the ground connection at the hybrid end was then modified by adding wires connected in parallel between the hybrid and the support structure .
additional wires can be viewed as adding inductances in parallel , thus reducing the overall inductance .
[ fig : noise_nwires ] shows the total noise as a function of the number of wires used in grounding the hybrid to the support ground . because the resistance of each wire is quite small , much less than the other connections in the ground path , the dependence of the noise on the configuration of the wires is due to the inductance of the wires .
shortening the wire length from 4 cm to approximately 1.5 cm suppresses the noise by about 10% .
the setup was then carefully improved by enlarging the ground connection to the support structure and minimizing its inductance using copper braid .
in addition , one end of a thin strip was soldered directly to the ground on the hybrid and the other end taped to the faraday cage using copper tape .
this resulted in the same noise level as before the module was installed on the support structure .
next , two modifications were made to this single - point ground configuration .
first , the sensor was grounded to the sensor support structure , which in turn was connected to the faraday cage .
this was done by connecting the ground on the high voltage filter card , mounted on top of the sensor , to the sensor support structure , which in turn was connected to the faraday cage with copper tape .
one end of a copper strip was taped to the chassis of the faraday cage , and the other end to the gold pads on the kapton flex embedded in the support structure .
secondly , the grounds of the sensor support structure and the hybrid support structure were connected .
one end of a copper ground mesh circuit was attached with silver epoxy to the gold pad on the kapton flex circuit that was co - cured onto the sensor support structure and the other end to the gold pad on the hybrid support structure ( see fig .
[ fig : solidmodel ] ) .
this minimizes the area of the ground loop .
the ground return is now a continuous , low inductance mesh circuit in close proximity to the analogue cable and the hybrid . the equivalent circuit diagram is shown in fig .
[ fig : equiv_ground ] . the measured noise distribution is shown in fig .
[ fig : noise_support ] and is identical to the behavior of the module on the bench .
both the total noise ( solid circles ) and differential noise ( open squares ) is plotted .
an increase in noise is observed at channel 64 .
recall that the first 64 channels were only connected to the analogue cable , whereas the following 34 channels were also connected to the sensor .
the increase in noise from about 1.8 to 2.7 adc counts is consistent with the added capacitive load of 11.2 pf of the silicon strip .
the last 30 channels were inactive .
note that , due to the double stack configuration of the analogue cable , a small odd - even effect in the pedestal distribution persists : the readout channels in the bottom layer of the analogue cable are more susceptible to external noise and have a larger pedestal rms value .
the assumption which lies at the basis of equation [ eq : noise ] , namely that the rms of neighboring channels is the same , is not completely valid in this situation .
odd and even channels have different proximity to the ground plane and thus an intrinsic different noise behavior .
we have developed a method of integrating grounding circuits in a carbon fiber support structure and have demonstrated that embedding copper - kapton mesh circuits in the various support elements , with the copper covering more than 15% of the area , is a flexible way of providing ground connections that is extremely robust .
two grounding configurations were studied .
first , a single - point ground at the hybrids was studied .
theoretically , a single - point ground provides the best grounding configuration . in practice , however , it is difficult to achieve a single - point ground especially for systems like the close - packed l structure , because of the tight space constraints .
for example , the support structures of nearby detector elements can easily couple to l capacitively , resulting in a breakdown of the single - point grounding scheme .
because of the inherent uncertainties when installing a detector as part of a complete system and operating all elements of the system concurrently , it is most prudent to implement a multi - point ground .
the challenge then is to achieve the same noise performance with a multi - point ground .
our test results show that the same noise performance as with the single - point ground , with the sensors floating , is obtained . since there is no improvement using a single - point ground and
because such a system is difficult to implement in practice , the safer approach of carefully grounding everything together and minimizing any potential ground loops was adopted .
our test results indicate that in this configuration two key requirements for low noise performance are to be respected : the ground loop must be minimized , and there must be a low impedance ground connection between the active detector and the front - end readout electronics . for the l silicon detector at small radius
being proposed for the d experiment , these general grounding principles have been implemented and excellent results were obtained .
a copper - kapton mesh circuit was embedded in the sensor support structure with the kapton covering the total surface area and the copper having an area coverage of 30% . with this ground mesh embedded in the surface of the support structure
it was straightforward to make low inductance ground connections to the sensor using small copper - kapton jumper cables attached to precisely positioned gold plated pads in the ground mesh using silver epoxy . to minimize the inductance of the ground path ,
a uniform continuous ground plane between the sensor and hybrid is needed .
another copper - kapton mesh circuit , with the same specifications , was embedded in the hybrid support structure .
small copper - kapton jumper cables attached to gold - plated pads at the end of either support structure provided the electrical connection between them . all hybrids and all sensors are thus tied to one uniform ground mesh .
grounding the hybrid was a bit more difficult due to space constraints .
based on the rule of low inductance connections , the shortest path from the hybrid to the inner support cylinder is needed .
this has led to a design where the hybrid has a ground strip at the bottom which makes direct contact with the support structure .
in this paper a pragmatic approach has been taken in the design of electrically robust detectors using carbon fiber .
it is still an intriguing question why carbon fiber is so highly conductive at high frequencies .
we speculate that this is due to the large skin depth .
carbon fiber consists of strands of conductor embedded in a non - conductive epoxy .
this configuration nearly eliminates eddy currents , which otherwise would reduce conductivity limiting it to the surface .
it is similar to the use of plates in conventional motors or transformers to get rid of eddy currents . without eddy currents ,
the area available for conduction in carbon fiber is much thicker than for a similar amount of copper .
this conjecture , however , needs to be experimentally verified .
a low inductance grounding scheme with the smallest possible ground loop is essential to achieve low noise performance .
this implies the use of short and wide electrical connections , and a careful mechanical design to minimize or eliminate ground loops .
we have shown that for frequencies relevant for operation of silicon detectors in the tevatron environment , carbon fiber is a good conductor and can not be distinguished from copper .
we have demonstrated that the use of copper - kapton flexible mesh circuits , with the copper covering more than 15% of the surface , provides adequate electrical coupling to carbon fiber support elements .
co - curing these circuits with the carbon fiber during the lay - up process provides an electrically integrated support , with great flexibility to adapt to the specific electrical requirements for that application . in a future paper the results of extensive tests on a small radius silicon detector , following the design rules outlined above , will be presented .
we wish to thank kurt krempetz , mike hrycyk and frank lehner for valuable advice , bert gonzalez for the assembly of the modules and all our d silicon colleagues for helpful discussions .
|
carbon fiber support structures have become common elements of detector designs for high energy physics experiments .
carbon fiber has many mechanical advantages but it is also characterized by high conductivity , particularly at high frequency , with associated design issues .
this paper discusses the elements required for sound electrical performance of silicon detectors employing carbon fiber support elements .
tests on carbon fiber structures are presented indicating that carbon fiber must be regarded as a conductor for the frequency region of 10 to 100 mhz .
the general principles of grounding configurations involving carbon fiber structures will be discussed . to illustrate the design requirements , measurements performed with a silicon detector on a carbon fiber support structure at small radius
are presented .
a grounding scheme employing copper - kapton mesh circuits is described and shown to provide adequate and robust detector performance .
carbon fiber , silicon vertex detector , support structure , grounding , tevatron , fermilab , dzero 29.40.gx , 29.40.wk , 29.90.+r , 72.80.tm , 81.05.uw
|
proposed wide - field imaging surveys will be able to discover and build light curves of thousands to hundreds of thousands of high - redshift type ia supernovae ( sne ia ) .
sne ia are established as excellent distance indicators having been used for both for the measurement of the hubble constant @xcite and for the discovery of the accelerated expansion of the universe @xcite .
there is thus interest in exploring how supernovae in new surveys can be used to improve the measurement of the expansion history of the universe and provide further insight into the physical cause of its acceleration .
although planned facilities and surveys provide straightforward harvesting of light curves for large numbers of supernovae , the corresponding spectroscopic observations used for redshift determination , supernova typing , and diagnostics are expensive .
it is unclear whether there will be available spectroscopic resources commensurate to the production of the imaging .
in addition , those surveys that target spectroscopy for a specific redshift range will still accumulate supernova light curves at other redshifts @xcite .
there is therefore interest in the possibility of performing supernova cosmology analysis with photometric data only . in this scenario ,
light curves not only fill their traditional role in measuring distances but are also responsible for supernova typing and redshift determination . in this paper , we explore the feasibility of using survey photometry to simultaneously estimate distance and redshift using fits to a light - curve template .
we do not incorporate typing and assume that the supernova is already known to be type ia .
we use an effective distance - modulus uncertainty as the metric of interest .
barris and tonry @xcite describe an alternative approach to the same problem using bayesian statistics and marginalizing over redshift , obtaining distance precisions close to the intrinsic corrected sn ia magnitude dispersion .
other papers have considered other combinations of information derivable from photometric data only : johnson and crotts @xcite and kuznetsova and connolly @xcite for photometric supernova typing ; sullivan et al .
@xcite for simultaneous typing and redshift determination with early light curves ; and wang @xcite for supernova photometric redshifts .
the effect of supernova redshift uncertainties on the determination of dark - energy parameters is examined in huterer et al . .
this paper is organized as follows : [ approach : sec ] describes our approach towards estimating distance modulus and redshift uncertainties for photometric survey data through light curve fitting . in [
spec : sec ] we specify the properties of the six surveys that we consider : all share a four - day cadence but have differing depths that correspond to obtaining a signal - to - noise ( s / n ) of either 25 or 50 for a fiducial @xmath6 , 0.731 , or 1.11 supernova at peak brightness in each band .
interesting features of our results are discussed in [ discussion : sec ] and conclusions are presented in [ conclusions : sec ] .
our objective is to determine how well the distance and redshift of a single supernova can be determined from a set of photometric data .
we use a parameterized description of the time - evolving spectral energy distributions ( seds ) for sne ia incident to the observer .
an individual supernova is characterized by the date of explosion and redshift in addition to the parameters of the sed model .
the survey is described by the observing cadence , photometric noise , and bands of observation .
model - parameter uncertainties from light - curve fitting are estimated using the fisher information matrix .
these parameters are in turn propagated into a covariance matrix for the distance modulus and redshift .
the covariance matrix is distilled into an effective distance modulus uncertainty to aid in the interpretation of the results .
the fisher analysis provides a firm lower limit on the errors one can obtain : this limit is an excellent estimate of errors when the data uncertainties are small and gaussian distributed .
salt2 @xcite provides an empirical model for the time evolving sed of sne ia .
the model is constructed from a training set of spectroscopic and photometric measurements of both low- and high - redshift supernovae .
salt2 models the sed as a function of phase @xmath7 and wavelength @xmath8 by @xmath9 \times e^{c cl(\lambda ) } \label{saltmodel : eqn}\ ] ] where the model parameters @xmath10 , @xmath11 , and @xmath12 correspond ( approximately ) to the peak luminosity , light - curve shape , and observed color .
the color parameterization simultaneously accounts for intrinsic supernova color variation and foreground dust extinction .
the functions @xmath13 , @xmath14 , and @xmath15 are constructed according to the training set .
the model covers supernova phases from @xmath16 $ ] days and the wavelength range @xmath17 $ ] .
we work with photon rather than energy fluxes since they , combined with transmission functions , describe counter detectors and observer magnitude systems @xcite .
the use of densities in @xmath18 rather than @xmath8 simplifies the description of redshifted spectra .
the residual dispersion of the training set data from the model gives a quantitative estimate of how well salt2 represents supernovae .
salt2 treats the dispersion of the template light curves of an average @xmath11=0 , @xmath19 supernova as @xmath20 where @xmath8 is the effective wavelength of the band in the sn - frame and the functions for variance , @xmath21 , and dispersion scaling , @xmath22 , are provided with the software distribution .
correlations between the residuals within a light curve , though non - zero , are not quantified .
dispersion in colors are given @xcite by a wavelength - dependent , phase - independent magnitude dispersion @xmath23 the survey data are described by the band of observation and photometric uncertainty . each supernova is observed with an undisrupted cadence in all bands .
observations are tuned to provide a fixed signal - to - noise at peak brightness in each band for an average @xmath24 , @xmath19 supernova , and assume a sky dominated background for calculating the noise off - peak .
this noise model approximately describes data from a ground - based rolling supernova search ( e.g. astier et al .
@xcite ) .
the supernova light curves are modeled using the salt2 sed model with the addition of parameters for the date of explosion @xmath25 and redshift @xmath26 @xmath27 where @xmath28 is the transmission of band @xmath29 .
the complete parameter set for the light - curve model is @xmath30 .
parameter uncertainties from light - curve fitting are determined through the fisher information matrix @xmath31 @xcite .
every data point in all useful bands are included in the calculation .
note that for the surveys we consider , photometric noise is independent of the light - curve parameters .
any independent redshift prior uncertainties are added to the @xmath32 element .
the covariance matrix of the parameters is the inverse of the fisher matrix @xmath33 . in salt2
, light - curve parameters are used to calculate the distance modulus through the relation @xmath34 with @xmath35 and @xmath36 .
the covariance matrix @xmath37 of the parameters @xmath38 is given by @xmath39 cosmological parameter fitting can proceed with the likelihood function dependent on the @xmath40 covariance matrix for each supernova .
however , to avoid analysis of this non - gaussian likelihood and provide intuitive insight , we consider the limiting case where redshift uncertainties are small .
taylor expanding the theoretical prediction of the distance modulus @xmath41 around the observed redshift , the probability distribution function to first order is gaussian with a @xmath42 for the one parameter @xmath43 with an effective variance @xmath44 ( see [ sec : app ] for a sketch of the derivation . ) in this fashion , uncertainties in @xmath26 and their correlation with @xmath43 are incorporated into an effective uncertainty in distance modulus .
we use the distance modulus for a photon - counting magnitude system that is related to the standard distance modulus by @xmath45 . as @xmath46 for most
all cosmologies of interest , positive correlations in @xmath43 and @xmath26 increase the effective distance modulus uncertainty .
we adopt a baseline description of salt2 and the supernova survey .
the covariance matrix of the salt2 templates between points @xmath47 and @xmath48 ( with phases @xmath49 and @xmath50 observed in bands @xmath51 and @xmath52 ) is based on the residuals given by equations [ residual : eqn ] and [ residualcolor : eqn ] : @xmath53 we have modeled the day - to - day correlations in the residuals with the @xmath54 function with a time scale of 5 days ; this keeps the intrinsic light curves smooth and provides both positive and negative correlations .
residuals in different bands are uncorrelated .
operationally , this covariance matrix of the model dispersion is added to that of the photometric data .
we consider only data at supernova phases covered by the salt2 model ; observations around peak brightness provide most of the leverage in light - curve fits with our sky - limited background .
observations are modeled as being made in the megacam @xmath3-bands of the canada - france - hawaii telescope with the transmissions provided with the salt2 software distribution .
as salt2 is not well trained at wavelengths smaller than the @xmath55-band and as sne ia have little emission at these low wavelengths , we do not use observations in this regime .
this naturally defines three specific redshifts to consider , @xmath56 , where the effective wavelengths of the @xmath57-bands in turn match that of the sn - frame @xmath55 .
the sn - frame wavelengths that correspond to each observer band are given in table [ effz : tab ] .
the number of data points for each supernova is @xmath58 . at @xmath6
there are a total 92 data points in @xmath3-bands used in the analysis . at @xmath59
there are 90 points in @xmath60-bands and at @xmath61 74 points in @xmath62-bands .
when the @xmath26-band observes the sn @xmath55 , there is no second band for a color measurement and the fit is not constrained .
.supernova - frame effective wavelengths in microns for observer @xmath3-bands .
the three redshifts in the table correspond , respectively , to matches of observer @xmath57 to the restframe @xmath55.[effz : tab ] [ cols="^,^,^,^,^",options="header " , ] all results presented in this paper are calculated for the case where observations are phased to begin on the date of explosion . in the case of @xmath63 and
no redshift priors , results vary by @xmath64% depending on the phase of observations with respect to the underlying light curve .
there are several interesting features of our results that we discuss here in further detail .
distance modulus and redshift measurements have significant positive correlation when there are no or weak redshift priors .
the correlation is dominated by the contribution of @xmath65 .
the positive @xmath66 is due to the supernova - frame @xmath55-band , where a strongly negative @xmath67 arises from the band coverage over the transition between the uv - flux dropout to the bright @xmath68 emission ubiquitous to sne ia . on the other hand ,
salt2 predicts brighter uv flux for larger @xmath11 .
the coverage over ca h&k wavelengths is therefore important for distinguishing supernova redshift and light - curve shape .
the effective distance modulus uncertainty is more sensitive to redshift uncertainties at low redshift .
for example , in the case of no redshift prior , @xmath6 and @xmath63 , the large value of @xmath69 makes it such that @xmath70 comes almost entirely from the @xmath71 term . the approximation derived in [ sec : app ] to get equation [ effmu : eqn ] breaks down as @xmath72 gets close to or larger than @xmath73 , which is certainly the case at low redshifts .
by @xmath59 the effective distance uncertainty is dominated by the @xmath74 contribution and our approximation holds .
the @xmath75 term does give the effective distance modulus uncertainty a dependence on the cosmological parameters .
a fit could proceed using the full @xmath43@xmath26 covariance matrix , or iteratively with the values of the cosmological parameters updated using the fit from the preceding iteration . without redshift priors ,
the redshift uncertainties derived from supernova light curves range from 0.0110.025 for the case of @xmath63 .
significant improvement in the cosmological utility of each supernova is possible with a comparable or better redshift prior .
this is illustrated by the case of @xmath61 in table [ sn25:tab ] . without priors ,
the uncertainty in redshift is @xmath76 .
when the redshift is constrained by a prior with 0.01 uncertainty , the resulting uncertainty in @xmath43 drops precipitously from @xmath77 to @xmath78 mag : @xmath79 ] .
concern about redshift uncertainties are superfluous when spectrograph - quality redshifts @xmath80 are available .
this shows why the propagation of redshift uncertainties is of little concern for experiments with spectroscopic redshifts .
it may seem odd that the covariance matrix @xmath37 gives smaller redshift uncertainties for the @xmath59 survey than for the @xmath6 survey , particularly since the latter has observations in four bands rather than three . by construction ,
the two surveys give the same @xmath5 for their respective target redshift and both use comparable numbers of data points due to the extra time dilation experienced by the more distant object .
qualitative comparison of figures [ partial0.325:fig ] and [ partial0.731:fig ] shows that @xmath81 distinguishes @xmath6 from 0.731 supernovae . at the lower redshift , @xmath81 and @xmath82
have almost identical shapes whereas at @xmath59 they are visibly different in the third band , allowing for tighter constraints in @xmath26 and @xmath12 .
the fisher matrix calculation gives for @xmath6 : @xmath83 , @xmath84 and @xmath85 and for @xmath59 : @xmath86 , @xmath87 and @xmath88 .
differences in @xmath81 between the two redshifts are due to differing contributions from @xmath89 where @xmath90 .
this term reflects changes in the predicted observed light curves with redshifting of the supernova sed at fixed phase .
( the phase dependent contribution to @xmath67 strongly resembles @xmath91 . )
the different shapes and uneven @xmath18-spacing of the @xmath3-bands give redshift - dependent observed supernova - frame wavelengths .
equation [ redshifting : eqn ] is close to zero when the band covers the wavelength region with peak emission ; at shorter wavelengths it is negative and at longer wavelengths it is positive .
the pronounced peaks in the @xmath47-band @xmath81 at @xmath6 and @xmath59 are attributable to their sn - frame wavelengths and the phase - dependent @xmath92 .
we calculate @xmath70 for a series of redshifts from @xmath93 1.1 in steps of 0.05 for the cases of @xmath63 and 50 and no redshift prior , and show the results in figure [ z - dmueff : fig ] .
looking at the low - frequency behavior , at low redshift there is a dramatic decrease in @xmath70 with increasing redshift as @xmath94 falls . at redshifts greater than @xmath95
, @xmath70 degrades as the number of usable data points and band coverage decreases .
the high - frequency oscillatory behavior is due to differing coverage of the supernova sed by the observer bands and the resulting change in sensitivity to the light - curve parameters .
comparison of tables [ sn25:tab ] and [ sn25nonoise : tab ] indicates that photometry uncertainty in the @xmath63 scenario dominates over the uncertainty from the salt2 light - curve templates .
the covariance matrices for the case of no redshift prior differ by a factor of two between @xmath63 and @xmath2 , rather than the factor of four expected if there were no salt2 uncertainties .
selecting cadences of two , four , and eight days while keeping the peak signal - to - noise proportional to the square root of the cadence gives the same results within the second significant digit .
degradation is expected at higher cadences as supernova phases with leverage ( large @xmath96 ) can be missed .
extending the observer bands to include the @xmath97-band can help significantly for objects at high redshift . for a @xmath61 supernova with @xmath63 at peak
, @xmath70 improves from 0.39 to 0.11 mag . for the cases in this paper where @xmath98
, we expect that a full monte - carlo - based fit would deviate from the fisher estimate by something in the order of 10 - 20% ; the broad wiggles in the supernova spectrum produce non - linear effects over this large redshift range .
in assessing whether a photometric survey produces adequate @xmath43@xmath26 uncertainty , we set as a target achieving a @xmath70 comparable to the intrinsic sn ia magnitude dispersion , @xmath99 mag . at this point , it is more advantageous to spend time observing more supernovae rather than improving the statistical precision of each individual object .
if relying on photometric information only , the observer can tune the depth of the survey to obtain the desired distance uncertainties .
measurements that give better than @xmath2 at peak brightness for an average @xmath24 , @xmath19 supernova in each band with a four - day observing cadence would be necessary to meet the target @xmath70 out to @xmath4 .
survey changes that hold fixed @xmath100 have almost no effect on @xmath43@xmath26 uncertainty for the finely sampled light curves considered in this paper .
improvements in photometric quality give declining yields when photometric uncertainties are smaller than those of the sn model . for our simulated data ,
sn model uncertainties begin to dominate between data qualities corresponding to our @xmath63 and @xmath2 surveys .
work on supernova modeling is ongoing and improvements are expected as data with better wavelength and temporal coverage are included in template building
. a judicious selection of observer bands and/or targeting of `` magic '' redshifts can give data with stronger leverage in minimizing the @xmath43@xmath26 covariance matrix , as seen in figure [ z - dmueff : fig ] .
these occur when there are strong distinctive light - curve gradients with respect to the redshift of the spectra , allowing the breaking of the degeneracy between redshift and other parameters .
shallower surveys can use the subset of supernovae with redshift prior uncertainties of @xmath101 or better .
we expect that the patient or well - equipped observer can obtain spectroscopic redshifts of supernovae discovered out to @xmath6 .
complete spectra for an unbiased sample of supernova hosts at higher redshifts will be more of a challenge ; for these , photometric redshifts can be used .
photometric redshift uncertainties of @xmath102 for a galaxy subsample have been obtained with the cfht filter set @xcite and better performance can be obtained with the addition of a redder filter .
a broader and non - gaussian dispersion is expected , however , for the diverse population of supernova host galaxies . in this paper
, we do not consider biases in the salt2 model .
biases in the distance and redshift determination leave irreducible uncertainties that impact cosmological parameter measurements .
the size of the training set necessary for the salt2 model construction and in testing its performance is subject for further study .
as supernova cosmologists concentrate on reducing the systematic uncertainties in using sne ia as distance indicators , focus is being placed on using light curves to distinguish variation in observed colors that are intrinsic to supernovae against those due to dust , and to determine the extinction properties of the host - galaxy dust itself .
other intrinsic supernova parameters may be encoded in the light curves .
in addition , we have not considered the impact of assigning the light - curve fitter the additional responsibility of determining supernova type . constructing a model for such a generalized fitter
is difficult given the heterogeneity of core - collapse supernovae and the dearth of ibc light curves , although bayesian methods have been applied towards this problem @xcite .
the formalism presented in this paper can be applied to future sophisticated light - curve models with an expanded parameter set . with photometric information diverted towards the measurement of additional supernova features , the uncertainties in @xmath43 and @xmath26
can only degrade compared to the 3-parameter salt2 model .
we acknowledge helpful discussions with julien guy , eric linder , lifan wang , and yun wang .
thanks also to the aspen center for physics where the idea for this paper and much of the work took place .
we thank the referee for constructive comments .
this work has been supported in part by the director , office of science , department of energy under grant de - ac02 - 05ch11231 .
when attempting to determine the cosmological parameters using the information contained in the distance modulii and redshifts of a set of sne ia , the joint probability density function ( pdf ) @xmath103 for the distance modulus ( @xmath43 ) and redshift ( @xmath26 ) of each supernova is needed . here , the superindex @xmath104 denotes observed quantities , while @xmath105 stands for the cosmological , and possibly nuisance , parameters .
the pdf is the probability density for observing a @xmath106 pair given @xmath105 .
it can be computed as @xmath107 where we integrate over all possible true values for @xmath43 and @xmath26 .
the first pdf in the right - hand side of ( [ eq:1 ] ) is just a two - dimensional resolution gaussian relating the observed and the true values of @xmath43 and @xmath26 .
the second pdf can be written as @xmath108 where @xmath109 is the distance modulus that results from applying the hubble relationship to the true redshift , considering the values of the parameters @xmath105 . in writing the last proportionality relationship we have taken into account that @xmath110 does not depend on @xmath105 and , therefore ,
can be safely neglected in the following .
therefore , we have @xmath111 \delta(\mu - \overline{\mu}(z,\vec{\theta}))\nonumber \\ & = & \int_{-\infty}^{+\infty } dz \ ; \exp\left[-\frac12\left(\mu^o-\overline{\mu}(z,\vec{\theta}),z^o - z\right)^t \mathbf{u}^{-1 } \left(\mu^o-\overline{\mu}(z,\vec{\theta}),z^o - z\right)\right ] \
.\end{aligned}\ ] ] in the last equation we have dropped the @xmath112 superindices for simplicity , and we have used the delta function to perform the integral over @xmath43 .
@xmath37 is the @xmath113 covariance matrix for the observation of @xmath114 and @xmath115 . in order to simplify the calculations that follow
, we will assume for the moment that @xmath37 is a diagonal matrix with elements @xmath116 , @xmath117 .
in this case , equation ( [ eq:3 ] ) simplifies to : @xmath118 \exp\left[-\frac12\left(\frac{z^o - z}{\sigma_z}\right)^2\right ] \ .\ ] ] although the integral extends to all possible values of @xmath26 , it is clear from the second exponential that only values of @xmath26 sufficiently close to @xmath115 will actually contribute .
let us , then , expand @xmath119 about @xmath115 as @xmath120 introducing equation ( [ eq:5 ] ) into equation ( [ eq:4 ] ) leads to @xmath121 \exp\left[-\frac12\left(\frac{z^o - z}{\sigma_z}\right)^2\right ] \ , \ ] ] which is just a gaussian integral in @xmath26 .
it can be easily computed to give : @xmath122 \ .\ ] ] so finally the desired pdf is a gaussian distribution centered at the expected distance modulus for the measured redshift and with variance @xmath123 , where the derivative is to be taken at @xmath124 .
if the covariance matrix @xmath37 between @xmath114 and @xmath115 is non - diagonal , the resulting pdf is still a gaussian with the same mean but with variance @xmath125 , where @xmath126 is the correlation coefficient between @xmath114 and @xmath115 .
|
the distance and redshift of a type ia supernova can be determined simultaneously through its multi - band light curves .
this fact may be used for imaging surveys that discover and obtain photometry for large numbers of supernovae ; so many that it would be difficult to obtain a spectroscopic redshift for each . using available supernova - analysis tools we find that there are several conditions in which a viable distance redshift can be determined .
uncertainties in the effective distance at @xmath0 are dominated by redshift uncertainties coupled with the steepness of the hubble law . by @xmath1
the hubble law flattens out and distance - modulus uncertainties dominate .
observations that give @xmath2 at peak brightness and a four - day observer cadence in each of @xmath3-bands are necessary to match the intrinsic supernova magnitude dispersion out to @xmath4 .
lower @xmath5 can be tolerated with the addition of redshift priors ( e.g. from a host - galaxy photometric redshift ) , observations in an additional redder band , or by focusing on supernova redshifts that have particular leverage for this measurement .
more stringent @xmath5 requirements are anticipated as improved systematics control over intrinsic color , metallicity , and dust is attempted to be drawn from light curves .
cosmology : distance scale , supernovae : general
|
the lyman alpha ( ly@xmath1 ) forest is an absorption phenomenon which is observed in the spectra of many high redshift quasars .
it arises due to absorption by galactic and intergalactic neutral hydrogen between a quasar and an observer [ see for a recent review ] .
as the universe expands , photons from the quasar are redshifted , and therefore absorption by clouds at different cosmological distances cause absorption at different wavelengths in the spectrum .
the resulting spectrum contains many absorption lines blueward of the ly@xmath1 emission line of the quasar , and the structure of the absorbing material along the line of sight is imprinted on the spectrum ( e.g. * ? ? ?
numerical simulations of cosmological structure formation predict the large scale structure of the gas which can then be compared with observations of quasar spectra @xcite .
the conclusions of these studies are that the ly@xmath1 clouds " are mostly intergalactic structures , with the denser regions corresponding to the locations where galaxies form .
the absorbing gas along the line of sight to a single quasar can be probed by comparing the relative positions and strengths of the absorption lines in the ly@xmath1 forest in the spectrum of the quasar @xcite .
the structure across the line of sight is harder to measure , although if two quasars that are nearby on the sky are observed , the fraction of ly@xmath1 forest lines that are common to the two quasars can provide a measure of the extent of the clouds across the sky @xcite .
traditionally , the large scale structure of the ly@xmath1 forest has been the main focus of research , because of its cosmological significance ( e.g. * ? ? ?
recently , the small scale structure ( on kiloparsec scales ) has also been investigated by several methods @xcite ; potentially revealing structure in and about the interstellar medium in young galaxies . gravitational lensing has previously been used to probe the size of absorbing clouds [ note that if the mass associated with a ly@xmath1 cloud is sufficiently large , it too may result in gravitational lensing effects , see @xcite ] .
when a background quasar ( regarded as a point source ) is lensed by a foreground galaxy , multiple images of the quasar may be obtained .
the light rays responsible for the two different images have traversed slightly different regions of space in their journey to the observer .
hence , a measure of the size of the ly@xmath1 absorbing clouds ( down to kiloparsec scales ) can be made by cross - correlating the the absorption lines in two images of the same quasar .
this method is analogous to using a true pair of quasars , however the lensing supplies a `` pair '' of quasars that are close together on the sky ( separated by @xmath2 arc seconds ) , allowing smaller scales to be investigated .
@xcite used this method in the multiply imaged quasar q1422 + 231 to trace mgii and civ absorption clouds between the lens and the quasar down to scales of @xmath2 26 pc , which is the smallest scale so far considered .
the very small scale ( @xmath3 1 parsec ) structure of the ly@xmath1 clouds is completely unknown ( due to limits in observational techniques , and the fact that the resolution of ly@xmath1 forest formation simulations is too coarse ) , and would be useful for deciding what the local equivalents of the ly@xmath1 clouds are .
some ( but not all ) low redshift ly@xmath1 clouds have been found to be associated with galaxies @xcite , suggesting that the structure of the ly@xmath1clouds might mimic that of the interstellar medium .
this is known to have structure on many scales , exhibiting fractal structure due to turbulence @xcite . in this paper
we present simulations demonstrating how the very small , sub parsec scales of ly@xmath1 clouds can be measured using a quasar that is subject to gravitational microlensing by a foreground galaxy .
in gravitational microlensing , the lensing effects of individual stars in the lensing galaxy are considered [ see @xcite for a review of gravitational lensing , including microlensing of quasars ] .
the cluster of ` micro - images ' that result are too small to be resolved , with the quasar appearing point - like , however brightness fluctuations can result as the relative positions of the source , observer and lensing stars change ; this is the observational consequence of microlensing .
since the rays joining the observer to the quasar traverse different regions of space at different times , we anticipate that small scale structure in the absorption clouds will lead to time variations in the strength of the absorption line relative to the continuum ( see figure [ diagram ] ) . throughout this paper ,
we assume a cosmology with @xmath4 , @xmath5 and @xmath6 km s@xmath7 mpc@xmath7 .
we made use of the microlensing code of @xcite , modified to output the positions of the rays in the lens plane and the corresponding position in the source plane .
the code uses the backward ray tracing method , where a rectangular grid of rays is fired into the lens plane , and the corresponding position in the source plane is calculated by summing the microlensing effects of many stars in the lensing galaxy . using these positions , magnification maps ( plots of the magnification of a point source as a function of position in the source plane )
can be produced by binning the positions of the rays in the source plane ( ie areas which collect a lot of rays have high magnification ) .
for the simulations presented in this paper , typically @xmath8 rays were traced between the observer and the source , ensuring that the resulting magnification maps , consisting of 1024x1024 binned pixels , did not suffer significant noise from poisson sampling .
a third plane , the cloud plane " was introduced , positioned between the lens plane and the source plane . for any ray starting at a position @xmath9 in the lens plane and landing at a position @xmath10 in the source plane , the position @xmath11 at which that ray intersects the cloud plane
is simply found by linear interpolation : @xmath12 where @xmath13 is a parameter which determines the location of the cloud plane .
if @xmath14 , the cloud plane coincides with the lens plane , and can be used to model absorption clouds within the lensing galaxy ( e.g. * ? ? ?
alternatively , @xmath15 corresponds to the case where the absorption clouds are located at the source .
this situation has been used to investigate the effects of microlensing on broad absorption line quasars @xcite .
note that @xmath16 etc are scaled distances , but can be interpreted as the angular position of a point measured in units of the angular einstein radius @xmath17 hence , if a point in the cloud plane has scaled coordinates @xmath18 , the physical coordinates ( eg .
measured in parsecs ) are @xmath19 , where @xmath20 is the angular diameter distance from the observer to the cloud plane .
the clouds were described by an absorption function @xmath21 defined over the cloud plane , which is the fraction of photons that are absorbed as a function of where the ray intersects the cloud plane .
if a ray passes through a region where @xmath22 , it is unaffected , but if a ray passes through a region where @xmath23 say , then the ray is attenuated by 50% .
gaussian clouds of the form @xmath24 were used , distributed randomly over the cloud plane . to investigate the effect of the cloud size @xmath25 , we ran simulations with three different cloud sizes for each value of @xmath13 .
the cloud sizes used were @xmath26 pc , @xmath27 pc and @xmath28 pc ( for typical cloud configurations , see figure [ clouds ] ) .
the number of clouds in the cloud plane was chosen so that the average absorption per unit ( physical ) area of the cloud plane was approximately the same for all of the simulations ( this was not exact due to a ` clipping ' effect , since the absorption can never be greater than 1 where several clouds coincide and their absorption values add ) .
the mean value of the absorption was @xmath2 0.5 . to achieve computational efficiency ,
the absorption as a function of cloud plane position was stored as a 1024x1024 pixel image , and @xmath29 was evaluated for each ray by looking up the appropriate pixel of the image .
the physical size of the cloud plane was chosen to be large enough to ensure that all of the rays intersect it and can be affected by absorption .
for the purposes of this study , the microlensing parameters adopted were those of image a of q2237 + 0305 , with a dimensionless surface mass density ( all in stars of mass 1 @xmath30 ) of @xmath31 and external shear @xmath32 @xcite .
the region of the source plane that was considered was a square with side length 20 einstein radii , ie a physical length 20@xmath33 , however this ray bundle is only square at the source plane .
it is approximately rectangular for the rest of the journey , with side lengths as a function of redshift shown in figure [ size ] , which was calculated assuming that the lensing galaxy was composed of smoothly distributed matter ( see section [ distances ] ) .
the redshift of the lens is @xmath34 and the redshift of the source is @xmath35 , which are the redshifts for q2237 + 0305 and its lensing galaxy @xcite . for these redshifts ,
the physical size of an einstein radius in the source plane is 5.85x10@xmath36 pc and the speed of the quasar across the source plane was assumed to be 600@xmath37 km s@xmath7 , or approximately one einstein radius per decade .
we ignored the individual motions of stars in the lensing galaxy which , for a transverse velocity of 600 km s@xmath7 would cause a 10 per cent increase in the frequency of high magnification events @xcite .
ignoring this effect saves considerable computational effort .
the magnification maps provide a measure of the intensity of the source as a function of its position in the source plane .
however , for a non - point source ,
the total observed intensity is the source profile multiplied by the magnification map , integrated over the source plane .
therefore the intensity as a function of source plane position for an extended source is given by the convolution of the magnification map with the source profile .
the physical size of the continuum source in q2237 + 0305 has been measured to be @xmath3 7.8x10@xmath38 pc @xcite , which is @xmath39 pixels wide in the magnification maps .
therefore , we ignored finite source effects , taking a single pixel to represent a point - like source . for each simulation , a pair of magnification maps was produced .
one was the usual magnification map calculated using the final locations of the rays , neglecting absorption , representing the magnification of the unabsorbed ( continuum ) wavelengths of the quasar spectrum .
an absorbed magnification map was also produced , using the attenuated values of the rays .
this magnification map applies to the absorbed part of the spectrum , so measures the flux at the bottom of the absorption line as a function of source position .
a third map was produced , a `` line strength map '' , which measures the flux at the bottom of the continuum relative to the flux of the continuum as a function of source position .
it is obtained by dividing the absorbed magnification map by the unabsorbed map ( see figure [ examples ] ) . a sample light curve showing the flux in the continuum and the absorption line , as well as the strength of the absorption line as functions of time is shown in figure [ lightcurve ] .
this light curve was taken from the same magnification maps shown in figure [ examples ] by taking a one dimensional horizontal cut across the middle of the magnification map .
note that some ( but not all ) of the sudden changes in continuum brightness ( due to the source crossing a caustic in the source plane ) are accompanied by sudden changes in the line strength .
also , for times when the continuum magnification is approximately constant ( eg from times @xmath40 in figure [ lightcurve ] ) , the line strength time series may still vary , reflecting the structure in the absorption clouds .
the scaled lens mapping equations ( which relate a position in the lens plane , or the image , to a position in the source plane ) are @xmath41 where @xmath42 and @xmath43 are the scaled deflection angles .
substituting these equations into equation [ interp ] , converting back to unscaled coordinates and using the unscaled lens equation @xcite , it can be shown that @xmath13 is given by @xmath44 where @xmath45 , @xmath46 , @xmath20 and @xmath47 are the angular diameter distances from the observer to source plane , lens plane to cloud plane , observer to cloud plane and lens plane to source plane respectively .
the redshifts were converted to angular diameter distances using the assumed cosmology .
the resulting plot of @xmath13 vs z is shown in figure [ z_dist ] .
note that the flatness of this relationship over a wide redshift range implies that the results for @xmath48 are approximately applicable for clouds over a very broad redshift range .
hence , the very small scales of the ly@xmath1 forest could be probed over a wide redshift range ( @xmath49 to @xmath50 ) by comparing observations with the high @xmath13 simulations .
also , the physical size @xmath51 of any feature in the cloud plane with scaled size @xmath52 is simply @xmath53 in this paper , we investigate the cases @xmath54 and @xmath48 . for these values of @xmath13 , the corresponding redshifts for the clouds
are @xmath55 and 0.53 ( although the latter result would also work well for clouds at higher redshifts ) .
absorption by neutral hydrogen at these redshifts would produce an absorption line in the q2237 + 0305 spectrum at wavelengths of 127.5 nm , 130.9 nm and 186.0 nm respectively . while the light from the quasar may suffer significant absorption below the lyman limit [ corresponding to z @xmath31
( @xmath56 245.0 nm ) ] which will influence the study of the ly@xmath1 lines , the analysis presented in this paper is equally applicable to the ubiquitous metal lines that are also seen in quasar spectra and can arise at lower redshifts @xcite .
if the absorbing material was uniform over the small scales probed by microlensing , then the flux in the absorption line and the flux in the continuum would rise and fall together as the source moved across the source plane . in this case
, the line strength map would be uniform , and a time series of measured values of the absorption line strength relative to the continuum would also be flat . also , the distribution of magnifications [ probability density function ( pdf ) for the magnification of a source placed at a random position in the source plane , see @xcite ] for the flux in the line would be the same as that for the continuum , but dimmed by a constant absorption value . figures [ panel1 ] , [ panel2 ] and [ panel3 ] show line strength maps for all of the cases investigated in this paper , together with magnification distributions for the continuum and the flux in the absorption line . the magnification distributions for the absorbed part of the spectrum are all essentially the same , with the most likely value having the flux in the absorption line about 1 magnitude fainter than the flux in the continuum .
the discrepancies in the faint end for the larger cloud sizes are likely to be due to sampling error , with only a small number of clouds affecting the final magnification map . for the @xmath57 and @xmath58 cases , the structure in the line strength map is complex .
the caustic network from the microlensing is evident , so rapid brightness changes due to the source crossing a caustic will usually be accompanied by sudden changes in line strength , albeit not as drastic as the sudden changes in the continuum brightness . since the rays which landed in the relevant area of the source plane occupy a smaller region of space for lower @xmath13 , only a small region of the cloud plane gets imprinted on the line strength map .
therefore , for a particular cloud size , faster variations in the line strength occur for clouds close to the source .
the effect of the external shear in stretching both the magnification map and the cloud related structures in the line strength map in the vertical direction is clearly visible .
when the absorption clouds are at @xmath48 ( close to the source ) , the imprint of the caustic structure becomes much less obvious in the line strength map .
the line strength map in this case approximately resembles the cloud structure , so if @xmath13 is sufficiently high , the absorption can be approximated as being at the source plane , and line strength variations directly probe the structure absorbing material . as figure [ z_dist ] demonstrates , this approximation is applicable over a wide range of redshifts . for the low values of @xmath13 , the rays which land in the region of interest of the source plane
have only crossed a small area of the cloud plane , so the features in the line strength map are quite large . in contrast , the @xmath59 line strength map shows that fast variations would occur in this case . for the large 0.1 pc clouds ( figure [ panel3 ] ) ,
the line strength maps contain features that are typically several einstein radii across , corresponding to time scales of several decades for any changes to be observed .
hence , this cloud size is approaching an upper limit of the scales that could be probed by this method .
the caustic structure is still present , but the same structure is also present in the line strength maps for the other cloud sizes , so this can not be used to distinguish different cloud sizes .
the fact that the same caustic network is present in the magnification map for the continuum and the line strength map suggests that a correlation may be present .
for example , areas where the line strength would be strong may be associated with areas where the quasar has a high magnification , or vice versa . to investigate this ,
histograms were produced , plotting the line strength versus continuum magnification ( one point for each pixel in the maps ) .
these are shown in figure [ corr ] .
an observation of a single light curve would sample regions from one of these distributions , possibly making different cloud sizes distinguishable . in each case
the magnification probability distribution for the unobscured quasar source is the same .
clearly , figure [ corr ] reveals that there is no strong correlation between the continuum magnification and the line strength relative to the continuum . even though the imprint of the caustic network on the line strength maps also implies that rapid changes in the quasar brightness will be accompanied by significant changes in the line strength
, there is no tendency for the changes to be correlated .
therefore , while spectral monitoring of the quasar during a high magnification event may make line strength variations measurable , they do not provide much information on the nature of the absorption clouds .
longer term spectral monitoring would be necessary to achieve this . the usual way to quantify
the time scale for microlensing variations is by computing the autocorrelation functions of a sample of light curves . in this case , it is the line strength variations which are being investigated , so the line strength map was used instead of the magnification map .
we used a slightly different approach to measure the typical time scale for line strength variations as a function of @xmath13 and the size of the absorbing clouds . for a given line strength map ,
a sample of pairs of points a fixed distance apart was selected at random , and the mean difference @xmath60 in the line strength values between the two points was calculated .
for example , if the pairs of points are close together relative to the size of the variations , the line strength is similar for those two points , so the mean difference is small . in this way , a plot of mean difference versus separation was produced .
the separation at which the mean difference is half of its maximum value was taken to be the measure of the length scale of features in the line strength map .
.time scales for line strength variation as a function of cloud size and cloud position .
the units are einstein radii , which take approximately 9.5 years to cross.[timescales ] [ cols="<,<,>,>,>,>,<,>,<,>",options="header " , ] in an unlensed quasar , the transverse velocities of the quasar and the ly@xmath61 clouds would also produce variations in the strength of the absorption lines . for a transverse velocity of 600 km s@xmath7 , the time taken to traverse 1 pc of the cloud plane ( with clouds at redshifts @xmath55 and 0.53 ) are 1700 , 1750 and 2500 years respectively . clearly , these are much longer than the corresponding time scales for the lensed system .
for example , even for the small clouds of size @xmath2 0.005 pc , the time scale is @xmath62 8.5 years , approximately a factor of 10 slower than the microlensed case .
thus , microlensing provides the opportunity to study the small scale structure on a reasonable time scale .
for the large clouds , the effect is even more drastic .
the crossing time of a cloud is @xmath2 2000 years , roughly 3 orders of magnitude longer than in the microlensed case .
the cosmological distribution of neutral hydrogen gas leaves an imprint on the spectrum of background quasars , in the form of a series of absorption lines .
this phenomenon is thought to be related to the large scale flow of gas and formation of galaxies and clusters .
while strong constraints have been placed on the larger scale properties of the ly@xmath1 clouds , virtually nothing is known on smaller scales , where gas may possess similar turbulent properties to the local interstellar medium , or be collapsing to form stars . in this paper , we have examined the influence of gravitational microlensing on the paths taken by light rays impinging on a distribution of ly@xmath1 clouds lying between the lens and the source . for a single macro - image , the population of micro - images samples differing absorption regions , with the net absorption seen ( in a single line ) being the average of that experienced by each ray . as the stars move , the configuration of micro - images changes , sampling different regions of the absorption clouds and causing temporal variability of the absorption line strength . in this study we considered clouds of three different sizes and at three different locations between the lens and the source .
quantitatively , it was shown that the fastest variations in the absorption line strength occur for the smallest clouds , and for clouds close to the source .
we also found that , due to the geometry of the universe , the variations in the absorption lines in the spectrum of q2237 + 0305 due to clouds at redshifts @xmath63 would have approximately the same effect as identical clouds which are just in front of the source .
at all redshifts , the caustic structure evident in the microlensing magnification map is also present in the absorption line strength map .
these discontinuities can result in very rapid changes in the absorption line strength , with a corresponding change in the continuum brightness .
however , a more detailed examination of the of the relationship between the continuum magnification and the line strength shows no apparent correlation between the two .
an analysis of the variability timescales reveals that significant changes in line strength due to the effects of gravitational microlensing would be visible on time scales ranging from years to decades , dependent on the size on the size of the absorption clouds and their location behind the lens ; this is at least an order of magnitude faster than clouds simply drifting across the line of sight to a distant quasar .
hence , in order to measure this effect , long term spectroscopic monitoring of the individual images of microlensed quasars is required .
such observations have already been proposed for studies of quasar structure @xcite .
this study has only considered ly@xmath1 clouds in image a of q2237 + 0305 , which is a positive parity image .
other images of q2237 + 0305 have negative parity , corresponding to one of the curves in figure [ size ] passing through zero .
furthermore , this study has assumed that the absorbing clouds possess no kinematic substructure , and hence the resultant variability applies to the total strength of the line ; for this a simple equivalent width measure in relatively low resulution spectra would suffice .
nevertheless , any real distribution of gas will exhibit some kinematic substructure in the form of rotation , infall , or simple clumping .
this would result in not only variability in the total strength of an absorption line , but also in its profile [ see @xcite ] . to detect line profile variability , higher resolution spectra would be required .
we will discuss these issues in future contributions .
this work was undertaken as part of an honours project at the university of sydney .
bjb was supported by a scholarship from the school of physics and cadbury - schweppes , and would like to thank all the regulars of the sssf ( http://www2b.abc.net.au/science/k2/stn/ ) for not telling me off when i was supposed to be doing work .
gfl would like to thank the rolling stones for pointing out that you ca nt always get what you want .
|
we present the results of microlens ray - tracing simulations showing the effect of absorbing material between a source quasar and a lensing galaxy in a gravitational lens system .
we find that , in addition to brightness fluctuations due to microlensing , the strength of the absorption line relative to the continuum varies with time , with the properties of the variations depending on the structure of the absorbing material .
we conclude that such variations will be measurable via uv spectroscopy of image a of the gravitationally lensed quasar q2237 + 0305 if the lyman alpha clouds between the quasar and the lensing galaxy possess structure on scales smaller than @xmath0 pc .
the time scale for the variations is on the order of order years to decades , although very short term variability can occur . while the lyman alpha lines may not be accessible at all wavelengths , this approach is applicable to any absorption system , including metal lines .
[ firstpage ] gravitational lensing - quasars : individual : q2237 + 0305 - quasars : absorption lines - intergalactic medium
|
active galactic nuclei ( agns ) were once thought to be a rare phenomenon whose major cosmological impact was in their ability to be seen out to great distances .
however , the discovery of supermassive black holes in the centres of nearby galaxies has led to the view that all galaxies went through an active phase .
furthermore , calculations have shown that the energy produced during this phase is sufficient to drive material out of the host galaxy ( silk & rees 1998 ; fabian 1999 ) .
feedback such as this can explain the tightness of the correlation between the masses of the black hole and host galaxy bulge ( e.g. , mclure & dunlop 2002 ) , and it has been suggested that such a process may perform a key role in shaping the galaxy luminosity function ( benson et al .
2003 ) .
semi - analytic models of galaxy formation ( e.g. , galform ; cole et al .
2000 ) are now being adapted to investigate the role of agn - driven feedback , and determine whether a self - consistent picture of both agn and galaxy formation and evolution can be developed .
it is to be hoped that progress in this field may provide important insights into some of the major outstanding issues in agn research and , in particular , the question of agn triggering .
early attempts at this ( e.g. , kauffmann & haehnelt 2000 ) have appeared encouraging in their ability to fit both the galaxy and qso luminosity functions but , by failing to make the distinction between the _ qso _ and _ agn _ luminosity functions , they fail to account for the possibly significant population of `` type 2 qsos '' which have been postulated to explain the shape of the cosmic x - ray background ( cxb ; e.g. , comastri et al .
1995 ) .
the cosmological importance of these objects is only exceeded by their elusiveness , since their optical continua closely resemble those of normal galaxies and therefore they can not be identified by the colour - selection techniques used for qsos ( e.g. , schmidt & green 1983 ) .
radio selection has allowed the radio - loud population of type 2 qsos ( i.e. , radio galaxies ) to be identified , but radio sources comprise only a small fraction of the total agn population .
hard x - ray selection has failed to locate the luminous narrow - line agn in the numbers predicted by model fits to the cxb but current x - ray telescopes are unable to find distant agn with compton - thick absorption , and such objects are known to exist in abundance locally ( e.g. , risaliti , maiolino & salvati 1999 ) .
large spectroscopic surveys such as the sloan digital sky survey ( sdss ; york et al .
2000 ) make it possible to identify agn in an unbiased manner and therefore determine for the first time the fraction of type 2 agn . in the simple unified model for agn , type 1 and type 2 objects ( i.e. , those which display broad permitted lines , and those which show narrow lines only , respectively ) differ only in terms of the angle which the observer s line of sight makes with the axis of a dusty torus ( see antonucci 1993 for more details )
. if this angle is larger than some critical angle , @xmath0 , the line of sight to the central regions ( which includes the broad line region ) is blocked by the torus and a type 2 agn is seen .
qsos are therefore only a subset of the agn population and no cosmological distinction should be made .
if @xmath0 has no dependence on luminosity , then the qso and agn luminosity functions differ only in terms of their space density normalization . however , the almost identical seds of qsos just longward of 1@xmath1 m ( e.g. , kobayashi et al . 1993 ; elvis et al .
1994 ) imply that the location of the inner wall of the torus is set by the distance from the nucleus at which dust reaches its sublimation temperature .
this ides is supported by the observed time delays between variations in the optical and near - infrared continua ( e.g. , fairall 9 ; clavel , wamsteker & glass 1989 ) . in more luminous objects
this distance is further away and , if the height of the torus remains constant , the critical angle @xmath0 must increase , thereby leading to a luminosity dependence in the observed type 1 agn fraction .
this is known as the ` receding torus ' model and was first suggested by lawrence ( 1991 ) . of course
, the height of the torus need not be the same in all objects but , since the luminosity dependence of its inner radius is the result of basic physics , it would be even more surprising if its height varied in such a way as to produce a constant opening angle .
a variation in the type 1 agn fraction with luminosity is therefore to be expected , although the size of any such variation may be small .
if it is large , however , then direct comparisons between the agn luminosity function produced by models and the qso luminosity function derived from observations ( such as that undertaken by kauffmann & haehnelt 2000 ) are seriously compromised .
a number of authors have investigated possible evidence for the receding torus model , frequently using radio - selected agn samples which are unbiased in terms of spectral type .
these studies ( e.g. , hill , goodrich & depoy 1996 ; simpson , rawlings & lacy 1999 ; willott et al .
2000 ; simpson & rawlings 2000 ; grimes , rawlings & willott 2004 ) have all supported the receding torus model , while simpson ( 1998 ) showed how it could explain apparent differences in supposedly isotropic properties between radio galaxies and radio - loud quasars with the same radio luminosity .
recent deep hard x - ray surveys have enabled similar analyses to be performed on samples of radio - quiet agn , and the same trend is seen ( ueda et al .
2003 ; hasinger 2004 ) .
the case for the receding torus model appears strong , although most of the samples studies so far are relatively small .
alternative models for unification exist and include warped obscuring discs , broken shells of material , or an evolutionary scenario where the black hole is at first smothered but powers a wind which is eventually able to blow away material and allow the qso to be directly observed .
however , even if the receding torus model is incorrect , the number of type 2 agn need to be determined in order for agn to fulfill their cosmological potential .
in this paper , we construct separate [ o iii ] emission line luminosity functions for broad and narrow - line agn from the second data release ( dr2 ) of the sdss which we use to investigate how the broad - line fraction varies with luminosity .
the format of the paper is as follows . in section 2
we describe how we have produced a sample of emission - line galaxies from the sdss dr2 catalogue , and in section 3 we describe the steps taken to produce luminosity functions for both type 1 and type 2 agn . in section 4
we derive the type 1 agn fraction and discuss the implications of our findings . finally in section 5 we present a summary . throughout this paper
we adopt a flat cosmology with @xmath2 and @xmath3 .
the sdss dr2 spectroscopic catalogue covers 2627deg@xmath4 down to a petrosian magnitude of @xmath5 ( in addition to colour - selected qso candidates which we exclude by means of this magnitude cut ) .
spectra are taken through 3-arcsecond diameter fibres , and so we must impose distance ( i.e. , redshift ) limits on our sample .
the fibres will not encompass the entire emission line region for nearby agn , since this can be @xmath6 in size ( e.g. , wilson & tsvetanov 1994 ) , while the dilution by starlight in distant galaxies will make the detection of emission lines more difficult .
kauffmann et al . ( 2004 ) show that a seyfert 2 sample with a luminosity limit @xmath7}>10 ^ 7\,l_\odot$ ] and redshift limits @xmath8 does not suffer from aperture biases , and we therefore adopt the same constraints for our sample .
this produces a catalogue of 4304 objects .
we do not apply any reddening corrections to the line luminosities .
this is partly for practical reasons , since there is no reliable way to determine the reddening to the narrow - line region in broad - line objects .
this is always a problem for studies which aim to compare the emission line properties of type 1 and type 2 agn , although the fact that [ o iii ] emission is seen to be isotropic when no reddening correction is applied ( mulchaey et al . 1994 ) suggests that differences between the mean reddening to the two types of galaxy must be small .
dahari & de robertis ( 1988 ) estimate that the median difference is @xmath9 , corresponding to a factor of @xmath10 at 5007 .
we will address whether this could have an impact on our results later .
we also do not correct our emission line fluxes for underlying stellar absorption .
since all our galaxies have very luminous [ o iii ] emission lines , they will also have luminous h@xmath11 emission which is expected to dominate over any absorption .
wake et al .
( 2004 ) find that the observed agn fraction is not sensitive to whether this correction is made .
the sdss spectroscopic pipeline produces a classification for all objects , based on principal component analysis ( schlegel et al .
, in preparation ) . for our sample of objects ,
this classification is either `` qso '' ( 890 objects ) or `` galaxy '' ( 3414 objects ) , which might be expected to correspond to agn types 1 and 2 , respectively .
however , while all objects classified as qsos are expected to display broad emission lines , the galaxy classification will include starburst galaxies as well as objects classified as type 1.8 or 1.9 ( where broad h@xmath12 emission is seen , but not broad h@xmath11 ) and possibly some low - luminosity seyfert 1s .
since broad emission components will preclude the use of a line - ratio diagram to discriminate between starburst and agn galaxies ( baldwin , phillips & terlevich 1981 ) , we elect to classify all 4304 galaxies by eye .
it is essential that the classification we choose is not dependent on the emission line luminosity .
since the broad and narrow emission line luminosities are strongly correlated , this requires that reddened objects where the broad emission component is near the limit of detection be included with the narrow - line objects .
objects which display broad _ wings _ on the h@xmath12 line ( ` type 1.x ' galaxies ) are therefore grouped with the type 2 galaxies , while those which have clear broad h@xmath12 and h@xmath11 emission _
lines _ are classified as type 1s .
one object was discovered to have a large noise spike at the expected location of [ o iii ] @xmath135007 and was removed from the sample , since the low luminosities of the visible emission lines and the absence of the @xmath134959 line indicate that its true line luminosity would be far below our lower limit .
we find that 63 objects classified as qsos ( 7.1per cent ) are not qsos at all , but display only narrow emission lines and appear to be type 2 galaxies .
we therefore reclassify them as such .
we also find that 63 objects classified as galaxies ( 1.8per cent ) are genuinely qsos , with clear broad line emission from both h@xmath12 and h@xmath11 , and we reclassify these objects .
a further 102 ( 3.0per cent ) show quite pronounced broad wings on the h@xmath12 line only , with a similar number showing weaker broad wings whose detectability would be a function of the signal - to - noise ratio in the spectrum .
a subsample of 100 objects was classified by dr c. done , and it was only for objects in this last category where disagreement was found .
the total fraction of sdss - classified ` galaxies ' which we believe display broad line emission is therefore consistent with the figure of 8per cent quoted by kauffmann et al .
( 2004 ) . while the reason for the misclassification of type 2 galaxies as qsos is unclear , the type 1s which are classified as galaxies typically have either weak broad line emission , or have very narrow permitted lines ( @xmath14 ) .
our interpretation of these as _ bona fide _ broad - line objects is based on the presence of fe ii pseudo - continuum emission between h@xmath15 and h@xmath11 , the lack of a strong 4000- break or balmer jump , and a continually rising spectrum at short wavelengths .
we wish to construct comparable type 1 and type 2 samples , so that their luminosity functions ( lfs ) are properly representative of the population as a whole .
unfortunately , there is no straightforward way to do this .
the objects we have classified as type 1s all contain an agn component , since the broad emission lines we observe can not be produced by stellar processes , but they may also possess a significant contribution to their emission line luminosities from star formation activity .
objects where the line emission is almost entirely powered by star formation may not be included in this sample , however , even if we have a direct view of the broad line region , since the broad line emission will be weak .
we therefore need to remove pure starburst galaxies from the sample of narrow - line objects , but also objects where the agn contribution is so low that the objects would not have been classified as type 1s _ if we had been able to see the blr . _ and @xmath11 axes represent the ridge lines from the intersection of the agn and starburst loci .
the dotted and dashed lines are the classification criteria suggested by kewley et al .
( 2001 ) and kauffmann et al .
( 2004 ) , respectively.[fig : bpt2 ] ] kauffmann et al .
( 2004 ) use an emission - line ratio diagram of the type pioneered by baldwin et al .
( 1981 ) to separate seyferts from starburst galaxies , and we propose a similar method . since the appropriate diagram makes use of the narrow balmer line fluxes , it can not be used for type 1 agn , where these data are unavailable . by plotting the type 1.x galaxies on the same diagram , however
, we can estimate what regions the type 1 population would occupy .
[ fig : bpt2 ] displays these data .
since our ` type 1.x ' classification requires that no broad h@xmath11 emission is seen , the location of these points along the ordinate should be representative of the type 1 population .
many of the type 1.x galaxies lie in the same region as the type 2s , because the broad h@xmath12 emission is weak and the sdss measurement therefore only includes the narrow - line flux . however , the tendency for these some of objects to have lower [ n ii]/h@xmath12 ratios than the type 2 population is clear , as the h@xmath12 flux measurement for these includes the broad line emission .
what is also clear is that some galaxies containing an agn have low [ o iii]/h@xmath11 ratios , and therefore any cut which excludes starburst galaxies will also exclude some _ bona fide _ agn .
we attempt to find the appropriate criteria for separating type 2 galaxies from pure starbursts by defining two ( non - orthogonal ) axes as shown in fig .
[ fig : bpt2 ] .
the axes meet where the loci of the starburst and agn sequences meet .
the ` @xmath12 ' axis follows the ridge line of the agn sequence , while the ` @xmath11 ' axis follows the ridge line of the starburst sequence . in this system ,
the demarcation criteria proposed by kewley et al .
( 2001 ) and kauffmann et al.(2004 ) correspond to @xmath16 and @xmath17 , respectively .
we also consider the widths of the narrow emission lines ( specifically , [ o iii ] @xmath135007 ) to help the differentiation between seyfert and starburst galaxies , since it is known that the narrow lines are broader in seyfert galaxies than in starbursts ( shuder & osterbrock 1981 ) .
this is demonstrated to be the case in fig .
[ fig : fwhm ] , where we have plotted the fwhm of the [ o iii ] line against @xmath12 as determined from fig .
[ fig : bpt2 ] . for the narrow - line galaxies
, we have taken the value of @xmath12 from their location in the ( @xmath12,@xmath11 ) coordinate system . since the [ n ii]/h@xmath12 ratio for the type 1.x galaxies is affected by broad line emission , we compute values of @xmath12 for these galaxies by determining where a galaxy with the same [ o iii]/h@xmath11 ratio would lie on the @xmath12-axis . the overall distributions of the type 2 and type 1.x galaxies are inconsistent at the @xmath18% level , using the two - dimensional kolmogorov smirnov test of peacock ( 1983 ) . however , fig .
[ fig : fwhm ] clearly shows two populations of narrow - line galaxies a population with broad ( @xmath19 ) emission lines extending to large values of @xmath12 which appears to be well traced by the type 1.x galaxies ; and a population of galaxies with narrower emission lines and @xmath17 .
these two populations can be associated with seyfert and starburst galaxies , respectively , and it is necessary to find some way to discriminate between the two classes . since the two distributions obviously differ greatly in the number of galaxies in the lower left corner of fig .
[ fig : fwhm ] , we investigate discrimination criteria which remove these objects from the narrow - line sample .
we construct a family of four - parameter selection criteria which require that @xmath12 and the emission line fwhm are both greater than certain values , and also that the points satisfy the relation @xmath20 where @xmath21 .
we then compare the samples of narrow - line galaxies selected by each of these criteria with the complete sample of type 1.x galaxies in the parameter space of fig .
[ fig : fwhm ] , again using a two - dimensional k s test .
we find that the two distributions can be made consistent at better than 90per cent confidence by making a simple cut with @xmath22 .
no improvement is made by imposing an additional cut based on line width , and only a very marginal gain results from adding a further selection criterion which is a function of both parameters .
this selection lies approximately halfway between the criteria suggested by kewley et al .
( 2001 ) and kauffmann et al .
( 2004 ) , as expected .
the implication is that the region between their two sets of criteria represents a mixing line , with our limit being the point at which the agn contribution becomes significant ( i.e. , when it is visible in the spectrum even if lightly - reddened ) . in the sample of 4303 emission - line galaxies
there are 851 type 1 agn , 141 type 1.x agn , 2184 type 2 agn , and 1127 starburst galaxies ( using the @xmath12 criterion to classify the latter two classes ) . since the sdss dr2 spectroscopic survey is a complete magnitude - limited survey , type 1 galaxies will appear in relatively greater numbers because of the additional flux from the nuclear non - stellar continuum
. we can however account for this by constructing lfs using the @xmath23 estimator ( schmidt 1968 ; rowan - robinson 1968 ) .
we then compare the type 1 lf with the lfs of three subsamples of type 2 galaxies , constructed by applying different selection criteria to the entire sample in an attempt to exclude the starburst galaxies .
two of these subsamples , named type 2-ka and type 2-ke , are produced by applying the line ratio selection criteria of kauffmann et al .
( 2004 ) and kewley et al .
( 2001 ) , respectively , while the third , named type 2-@xmath12 , is constructed using the @xmath24 criterion derived above .
the three type 2 lfs are compared with the type 1 lf in fig .
[ fig : lf ] .
it is clear that , although the numbers of each type of seyfert are similar at @xmath25 } \ga 10^{34.5}\rm\,w$ ] , type 2s outnumber the type 1s at lower luminosities by a factor of @xmath26 .
although there are differences between the numbers of lower luminosity type 2s in the three samples , these differences are small compared to the difference between these samples and the number of type 1s . the different shapes of the lfs can not be explained by a difference in the mean reddening of the type 1 and type 2 samples , since a reddening correction is simply a horizontal shift .
the different shapes also can not be due to problems with the seyfert / starburst separation .
the criterion defined by kewley et al.(2001 ) is conservative in the sense that it should exclude _ all _ starburst galaxies , so the excess of type 2 galaxies over type 1 galaxies at lower [ o iii ] luminosities must be genuine , since even the type 2-ke sample exceeds the type 1 sample .
conversely , the difference also can not be due to an artificial deficit of high luminosity type 2 galaxies , since all three samples have effectively the same lfs at @xmath25 } \ga 10^{34.5}$]w .
a final possibility is perhaps that a significant fraction of the lower luminosity type 1 galaxies have been misclassified , presumably because the broad lines are weak and below the detection threshold .
this seems unlikely since the broad lines are much stronger than the narrow lines . in our type 1 sample , the median [ o iii]/h@xmath12 ratio is 0.15 , while 90per cent of objects have a ratio less than 0.5 .
we therefore expect all but a small fraction of type 1 galaxies to have readily identifiable broad h@xmath12 emission ( although the [ o iii ] lines in our sample only need to be detected at @xmath27 , most are detected at much higher significance all but 20 are detected at @xmath28 , and over half have @xmath29 ) .
furthermore , we also display the lf of the type 1.x galaxies in fig .
[ fig : lf ] , and there is clearly no bias against lower luminosity objects .
we therefore conclude that the luminosity functions of type 1 and type 2 galaxies are genuinely different . for the remainder of this paper , we use only one type 2 luminosity function .
we adopt the type 2-@xmath12 sample , and add in quadrature to the poisson uncertainty the standard deviation of the three samples .
the different luminosity functions for our samples of type 1 and type 2 galaxies can also be described as a luminosity dependence of the type 1 agn fraction . in the context of the receding torus model ,
such a change is described by @xmath30^{-0.5 } \label{eq : rt}\ ] ] ( e.g. , simpson 1998 ) , where @xmath31 is the luminosity at which the numbers of type 1 and type 2 agn are equal ( i.e. , an opening angle of @xmath32 ) .
the ` luminosity ' of concern here is the ultraviolet optical radiation which heats the dust , which is difficult to measure directly , but the constant equivalent width of the [ o iii ] emission line in qsos ( e.g. miller et al .
1992 ; see also simpson 1998 ) suggests that it should be a good tracer . since the receding torus model is able to explain a wide range of observed agn properties ( simpson 2003 and references therein ; see also grimes et al .
2004 ) , we investigate whether it is also able to explain the results of the previous section .
we complement our data from similar analyses in the literature , which typically extend to higher luminosities .
both ueda et al . ( 2003 ) and hasinger ( 2004 ) measure the type 1 agn fraction in hard x - ray selected samples , which should be relatively free from orientation biases except for compton - thick agn , whose existence will case the type 1 fraction to be systematically overestimated .
about half of nearby low luminosity type 2 agn are compton - thick ( risaliti et al .
1999 ) and , if this fraction is true at all luminosities then the type 1 fraction will be overestimated by about 50per cent .
we do not make any correction for compton - thick objects but note that the type 1 fractions derived from x - ray - selected samples should be considered as upper limits to the true fractions .
we convert the 210kev luminosity ( corrected for intrinsic absorption ) used by these authors into [ o iii ] luminosity using the mean ratio for seyfert galaxies from mulchaey et al .
( 1994 ) , i.e. , @xmath33 } = 0.015 \times l_{\rm2 - 10kev } \ , .\ ] ] we also add the data of grimes et al .
( 2004 ) from the 3crr , 6ce , and 7c flux - limited samples of radio sources .
the results are shown in fig .
[ fig : qsofrac ] , where a clear and highly significant ( @xmath34per cent confidence ) correlation between the type 1 fraction and luminosity is seen for which the receding torus model provides an attractive , but not unique , explanation for this correlation .
we note that the points from grimes et al .
( 2004 ) tend to be slightly lower than those from the other works and this might reflect a difference in the torus properties of radio - loud and radio - quiet agn , although the discrepancy is not highly significant .
although the four samples plotted in fig . [ fig : qsofrac ] all show the same trend , we concentrate for now on the data from our analysis alone since there are less well understood selection effects in the x - ray samples ( see treister et al .
these cover more than two orders of magnitude in @xmath35}$ ] but are not well - fit by the standard receding torus model , with the best fit having @xmath36 for @xmath37 .
we therefore suggest a modification to the standard receding torus model , where the height of the torus depends on the luminosity of the agn . such a dependence could plausibly be modulated via the mass of the black hole and the strength of its associated gravitational potential .
we parametrize this dependence as @xmath38 , and equation [ eq : rt ] therefore becomes @xmath39^{-0.5 } \ , .
\label{eq : rth}\ ] ] this produces a more acceptable fit ( @xmath40 ) for @xmath41w and @xmath42 , which we also show in fig .
[ fig : qsofrac ] .
arshakian ( 2004 ) made a similar parametrization when studying the luminosity - dependent variations of qso fraction and projected linear size in a radio source sample .
he defined a parameter @xmath43 such that @xmath44 ( therefore @xmath45 ) and found @xmath46 and @xmath47 when considering the linear sizes and qso fractions , respectively .
the latter value is in excellent agreement with the best - fitting value of @xmath48 we determine here .
an alternative explanation for this modification is that @xmath25}$ ] is not a good substitute for @xmath49 , the radiative luminosity which heats the dust in the torus . if @xmath35 } \propto l_{\rm rad}^{1 - 2\xi}$ ] , then the standard receding torus model would fit the data .
however , such an extreme disproportionality can be excluded based on a number of arguments such as the near - constancy of [ o iii ] equivalent width in qsos , ( miller et al .
1992 ) , the proportionalities between [ o iii ] luminosity and other quantities ( e.g. , rawlings & saunders 1991 ; mulchaey et al . 1994 ; grimes et al .
2004 ) , and the dependence of [ o iii ] emission on ionization parameter ( simpson 1998 ) . a more subtle variation in the relationship between @xmath49 and @xmath35}$ ] may provide an explanation . in the standard unified model for agn
, only a fraction @xmath50 of the ionizing radiation can escape to illuminate the narrow - line region on scales @xmath51pc .
this explains the ( bi)conical emission - line structures seen in some seyfert galaxies ( e.g.pogge 1989 ; haniff , wilson & ward 1991 ; simpson et al .
1997 ) , and suggests that @xmath35 } \propto l_{\rm rad } f_1 $ ] . in such a scenario , @xmath52}/(l_0 f_1)]^{-0.5 } \ , , \label{eq : rtf}\ ] ] which can be rearranged to give a cubic equation in @xmath53 whose roots are functions of @xmath35}$ ] .
this model is also shown in fig .
[ fig : qsofrac ] and provides an acceptable fit ( i.e. can not be rejected at 95per cent confidence ) to our data ( @xmath54 for @xmath55w ) .
[ fig : qsofrac ] suggests that we should favour the torus described by eqn [ eq : rth ] over that from eqn [ eq : rtf ] by virtue of the quality of fit to the x - ray datapoints . however , we caution against leaping to such a conclusion and suggest that some of the most luminous qsos may have been misclassified as type 2 objects .
while this seems counterintuitive , such objects will inevitably be at the highest redshifts and therefore the optical spectroscopy performed on them will sample the rest - frame ultraviolet . even relatively small amounts of dust could be sufficient to extinguish the broad emission lines and give such objects the appearance of a type 2 agn in the uv . while there should be very few luminous agn which are lightly reddened because the observer s line of sight just grazes the torus ( see hill et al .
1996 ) , the required extinctions may arise from dust in the host galaxy which is not associated with the active nucleus .
the repeated reclassifications of ` type 2 qsos ' after near - infrared spectroscopy reveals broad h@xmath12 emission ( e.g. , halpern , eracleous & forster 1998 ; halpern , turner & george 1999 ; akiyama , ueda & ohta 2002 ) suggests that there may be some truth in this . on the other hand ,
the qso fraction may be overestimated if there exists a significant population of luminous agn with compton - thick tori which are not detected in hard x - ray surveys .
what is required is a systematic near - infrared study of either spectroscopy of imaging , similar to that which simpson et al .
( 1999 ) performed for @xmath56 powerful radio sources .
this should be possible in the very near future with the combination of _ chandra _ or _ xmm newton _ and _ spitzer _ imaging in a number of fields .
having suggested that both modifications to the receding torus model fit the data equally well , we return to the issue of whether breaking the proportionality between the ionizing and [ o iii ] emission - line luminosities is in conflict with observations .
first , this does not affect the isotropy of the [ o iii ] emission , since the emission - line luminosities of type 1 and type 2 agn with similar ionizing luminosities should still be the same .
however , it predicts that the observed equivalent width of [ o iii ] should be proportional to @xmath53 ( i.e. , increase with increasing agn luminosity ) , whereas observations show it to be constant with luminosity ( miller et al .
however , over the range of luminosities studied by miller et al .
( approximately two orders of magnitude ) , @xmath53 is predicted to vary by a factor of about 2 , which is equal to the @xmath57 scatter in the rest - frame equivalent width .
furthermore , this effect will be lessened by those objects seen at small polar angles where the continuum , but not the emission - line , luminosity is enhanced .
a quantitative analysis of this effect depends on the assumed angular and wavelength dependence of the ionizing continuum and is therefore rather model - dependent and beyond the scope of this paper .
however , we do predict that the scatter in the equivalent width of the [ o iii ] line should increase with luminosity .
studies of agn evolution have so far primarily relied on samples of qsos since these objects are readily visible out to large distances .
obviously such samples contain only a fraction of the agn at any redshift or luminosity since type 2 objects do not meet the colour and/or morphological selection criteria ( while hard x - ray surveys are less biased against type 2 agn , they presently lack the combination of depth and area coverage to properly sample the agn luminosity function in an unbiased manner ) .
we therefore wish to determine how the luminosity - dependent qso fraction might affect the conclusions drawn from studies of qso evolution .
since there is a significant stellar component to the broad - band magnitudes of our type 1 galaxies which is not the case for observed samples of qsos , we estimate the non - stellar continuum luminosities from our [ o iii ] luminosities.we assume a mean rest - frame equivalent width ( relative to the non - stellar continuum ) for the @xmath135007 line of @xmath58 ( miller et al . 1992 ) , and a spectral shape @xmath59 ( vanden berk et al .
2001 ; the result is extremely insensitive to the shape of the spectrum due to the small wavelength difference between the [ o iii ] line and the @xmath60 band ) in the rest - frame optical .
it then follows that @xmath61}/10^{35}\,{\rm w } ) + 2.5 \log ( \epsilon/24{\rm \,\aa } ) \label{eq : mb}\ ] ] and @xmath62 for the canonical qso spectrum ( blair & gilmore 1982 ) . our type 1 lf can be well - fit with the two power - law parametrization of the 2qz team , @xmath63 ( e.g. , croom et al .
2004 ; note that the normalization factor of 2 is missing from papers by this team ) . a good fit ( @xmath64 )
is found with the parameters @xmath65 , @xmath66 , @xmath67 , and @xmath68 .
the power law indices are similar to those derived by croom et al .
( 2004 ) , who found @xmath69 and @xmath70
. however , a better comparison would be to the combined ( i.e. , types 1 and 2 ) seyfert lf since , while the croom et al .
( 2004 ) lfs are relatively insensitive to the lack of type 2 objects , our type 1 lf is a poor representation of the overall agn lf .
the fit to the combined seyfert luminosity function is improved ( @xmath71 ) , primarily due to the larger error bars resulting from the uncertain classification of narrow - line objects , with @xmath72 , @xmath73 , @xmath74 , and @xmath75 .
there is moderately good agreement between the values of @xmath76 and @xmath77 from our combined seyfert sample and the extrapolation of the exponential pure luminosity evolution of croom et al .
we obtain @xmath78 and @xmath79 compared to @xmath80 and @xmath81 from their work ( the agreement is less good for the polynomial evolution , where a value of @xmath82 is predicted for our sample ) . the lower space density we derive can not be due to incompleteness in our sample of seyfert galaxies since our spectroscopic selection will include all such objects ( accounting for the uncertainty in the starburst seyfert classification ) , whereas the optical colour selection used to select qsos will exclude a population of red qsos which may exist in significant numbers ( webster et al .
1995 ; benn et al . 1998 ; richards et al . 2003 ) .
we also try to simultaneously fit our seyfert lf and the six qso lfs of croom et al .
( 2004 ) , after using the modified receding torus model to account for the missing type 2 objects .
we use their two power - law model with an exponential form for the luminosity evolution , although we fit the value of @xmath76 for our seyfert lf as a separate parameter since an extremely poor fit is obtained otherwise . following croom et al .
, we use only the @xmath83 bins , and we show the best fit in fig . [
fig : croom ] .
while this fit is unacceptably poor in a formal sense ( @xmath84 ) , it does reproduce the gross characteristics of the evolving lf and , in particular , the power law indices differ from those derived from the type 1 ( qso ) lf alone . unsurprisingly , the value of @xmath12 is least affected . from our earlier calculation
that @xmath85w , we conclude that qsos outnumber type 2 objects where @xmath86 and therefore the bright end of the luminosity function is not strongly affected by the absence of type 2 objects and the power law slope is only moderately steepened by their inclusion we find @xmath87 as opposed to @xmath69 .
however , the faint end slope is made much steeper ( @xmath88 compared to @xmath70 ) , and is closer to the faint - end slope of the hard x - ray luminosity function ( @xmath89 ; ueda et al .
this steepening has a large effect on the contribution from faint agn to the overall agn luminosity .
integrating the two power - law model , the total luminosity from agn with @xmath90 is @xmath91 .
the total luminosity from agn fainter than @xmath92 does not converge for @xmath93 and so is dependent on the low - luminosity cutoff , but in the decade below @xmath92 , it is @xmath94 for @xmath95 and nearly twice as much for @xmath96 .
wolf et al .
( 2003 ) find a downturn in the shape of the lf at faint absolute magnitudes ( @xmath97 , not probed by 2qz ) which allows the luminosity density integral to converge .
however , the increased number of type 2 agn predicted at these luminosities significantly affects the strength of this downturn and therefore the contribution by faint agn to the overall luminosity density .
such objects would be fainter than the low - luminosity type 1 agn studied by wolf et al .
( 2003 ) and therefore could not have yet been spectroscopically identified in deep x - ray surveys such as barger et al .
given that a putative further population of compton - thick type 2 agn would not even appear in such x - ray surveys , there is a distinct likelihood that the importance of low luminosity agn has been significantly underestimated . finally , while attempting to tie the present - day black hole mass density to the evolution of qsos , yu & tremaine ( 2002 ) note that their results are dependent on the prevalence of obscured low - luminosity agn . in particular , the accretion efficiency can be made roughly independent of agn luminosity if the obscured agn fraction is larger for low mass black holes , and hence less luminous agn , as predicted by the receding torus model . while the predictions of the receding torus model agree well with the observed number of x - ray - selected type 2 agn , models for the cxb ( e.g. , comastri et al . 1995 ; wilman , fabian & nulsen 2000 ) require the presence of an additional compton - thick population which is not detectable in such surveys .
as discussed previously , this means the true type 1 fraction is lower than the observations suggests , and hence there is an apparent discrepancy between the receding torus model and models for the cxb . however , the number of compton - thick agn proposed is a relatively small fraction of the total agn population .
ueda et al . ( 2003 ) fit the cxb by proposing that there are as many agn with @xmath98@xmath99@xmath100 as with @xmath101@xmath102@xmath100 , but there is a strong luminosity dependence in the shape of the @xmath103 distribution . therefore , while there are many compton - thick agn at low redshift / luminosity ( risaliti et al .
1999 ) , there are many fewer at the redshift / luminosity which produces the bulk of the cxb . in addition
, type 1 agn with significant x - ray obscuration are known to exist , the original member of this class being mr 2251@xmath104178 ( halpern 1984 ; see also page et al .
2004 ) , so the type 1 fraction need not be the same as the fraction of x - ray unabsorbed agn . although the column densities of these objects tend not to exceed @xmath105 , willott et al .
( 2003 , 2004 ) have found objects with greater absorption ( @xmath106 ) whose optical / infrared spectra are consistent with lightly - reddened qsos .
it is therefore possible that the most heavily absorbed agn also suffer significant non - torus reddening and may have escaped detection in both x - ray and optical qso surveys .
finally , a large discrepancy between the receding torus model and the observations of hard x - ray - selected agn occurs only at very high luminosities ( @xmath7}>10^{36}$]w , or @xmath107w ) and this population contributes almost negligibly to the overall cxb ( e.g. , figs 16 and 17 of ueda et al .
we conclude therefore that the receding torus model does not conflict with our current understanding of the cosmic x - ray background .
we have constructed separate luminosity functions for seyfert 1 and seyfert 2 galaxies , using the second data release of the sloan digital sky survey .
we have shown that these luminosity functions have different shapes , resulting in an increase in the type 1 agn fraction with luminosity which is in quantitative agreement with that determined from hard x - ray - selected samples .
we show that the form of this luminosity dependence is inconsistent with the standard receding torus model , and propose a modification where the height of the torus varies with luminosity as @xmath108 , although we note that a more robust spectral classification of distant luminous agn is desired . assuming that this modification to the receding torus model provides a reasonable correction factor to account for the ` missing ' type 2 agn in qso surveys , we have shown that the contribution to the cosmic accretion luminosity from faint agn may have been significantly underestimated .
i would like to thank matt jarvis for useful discussions , chris done for agreeing to classify a subsample of spectra , and the referee , andy lawrence , for his helpful report .
i also wish to thank pparc for financial support in the form of an advanced fellowship .
funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions .
the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington .
|
using a complete , magnitude - limited sample of active galaxies from the sloan digital sky survey ( sdss ) we show that the fraction of broad - line ( type 1 ) active galactic nuclei increases with luminosity of the isotropically - emitted [ o iii ] narrow emission line .
our results are quantitatively in agreement with , and far less uncertain than , similar trends found from studies of x - ray and radio - selected active galaxies . while the correlation between broad - line fraction and luminosity is qualitatively consistent with the receding torus model , its slope is shallower and we therefore propose a modification to this model where the height of the torus increases slowly with agn luminosity .
we demonstrate that the faint - end slope of the agn luminosity function steepens significantly when a correction for ` missing ' type 2 objects is made and that this can substantially affect the overall agn luminosity density extrapolated from samples of more luminous objects .
galaxies : active galaxies : seyfert
|
the investigation of quantum transport through molecular systems has become an important research field in the last few decades.@xcite progress in measuring and fabrication techniques@xcite has led to the continuous miniaturization of electronic devices , which have reached the point where quantum effects are important .
for example , semiconductor devices have been reduced in size to the nanoscale@xcite and even to the atomic scale.@xcite in addition to the report of metal atomic wires,@xcite stable and rigid carbon atomic chains have been reported recently.@xcite a key idea behind the advances in the understanding of charge transport through molecular systems is based on the view proposed by landauer , _ conductance is transmission_.@xcite potential applications of molecular devices range from novel computer architectures@xcite to chemical sensors@xcite and medical diagnostics @xcite . among the various molecular devices , 1d conductors such as molecular wires,@xcite
have been considered to be one of the most fundamental components for nanotechnology . due to the reduction of size and dimensions of materials ,
1d systems show sensitive response to external field or intrisic characteristics , which can be exploited for the development of molecular devices such as biological sensors@xcite with high sensitivity .
polymers are one of the most promising materials for acting as 1d conductors . their applicability@xcite ranges from displays@xcite to thin film transistors,@xcite photovoltaics,@xcite and solar cells.@xcite the compatibility of polymer materials with light - weight , mechanically flexible plastic substrates and new fabrication methods make them possible candidates for future electronic devices .
intrinsic properties of molecules as well as external perturbations can strongly affect the transport properties of the resulting low dimensional devices .
for example , the conductance of 1d polymers drastically depends on the concentration and position of impurities ( dopant)@xcite or defects.@xcite other effects which are important include dimerization and the formation of solitonic defects .
hence , all of these properties should be considered in the design of 1d devices .
therefore , in order to facilitate the design of controllable molecular devices , it is necessary to use a theoretical approach that simplifies the monitoring of how these intrinsic features affect their transport properties .
such approach should allow one to gain understanding and deep insights into the physical origins of the behavior of these materials .
the knowledge acquired through these models can be used for the interpretation and elucidation of experimental observations , as guidelines for the planning of experiments , as well as in the design of molecular devices . in this paper we model
molecular junctions where 1d molecular wire is connected between two electrodes using a tight - binding approach
. then we systematically investigate how the degree of dimerization , and the position of solitonic and binding defects affect the transport properties of such 1d systems .
we calculate the electronic density of states ( dos ) and the conductance using the landauer model in terms of the equilibrium green functions .
this article is divided into the following sections .
section ii briefly presents the theoretical framework and section iii presents the effect of solitonic and binding defects on quantum transport through molecular systems , respectively .
finally section iv summarizes this article , stressing on the importance of the defect position on the linear conductor .
the effect of an odd vs. an even number of sites in the dos and in the conductance is summarized in supplementary information a , and the effect of dimerization is summarized in supplementary information b. in addition , the relationship between dimerization and length dependence of conductance is summarized in supplementary information c.
we model a molecular junction by connecting a 1d molecular wire between two electrodes .
we describe this molecular system using a standard tight - binding hamiltonian@xcite @xmath0 that considers only @xmath1 orbitals . here , the dimerization is represented by a sequence of alternating weak and strong bonds while the solitonic defect is described by a pair consecutive weak bonds . for clarity , we model the electrodes as 1d wires with hamiltonians @xmath2 and @xmath3 , which result in a fair representation of infinite reservoirs . fig .
[ fig1 ] shows the molecular junction considered in this work .
the tunneling between the central molecule and the left and right electrodes is represented by two matrix elements @xmath4 and @xmath5 .
note that the electrode parameters and tunneling amplitude reduced to into two independent parameters @xmath6 and @xmath7 ; the escape rates to the left and to the right electrode , respectively.@xcite the details this of theoretical framework are described in appendix a. in principle , we can expand this method to general molecular systems which include any type of orbitals . in this work
, we simplified the model hamiltonian by only considering @xmath1-orbitals since the contribution from other orbitals to the conductance near the fermi energy is negligible in the sp2 carbon systems which we addressed .
however , in the case of the molecular wires including metal atoms,@xcite where s- , p- , and/or d - orbitals play a role in charge transport , those orbitals have to be taken into account . we model a molecular junction by connecting a 1d molecular wire between two electrodes . for clarity
we model the electrodes as 1d wires .
we describe this system using a tight - binding hamiltonian , considering only @xmath1 orbitals .
the tight - binding parameters used depend explicitly on the interatomic distances in real space .
thus , we can introduce defects or disorder by directly controlling on - site energies ( diagonal elements ) and hopping integrals ( off - diagonal elements ) of the electronic tight - binding hamiltonian matrices .
figure 1 shows the molecular junction considered in this work . our starting point is to divide the entire 1d system into three regions ; left electrode , right electrode , and central molecular region .
we write the electronic hamiltonian as @xmath34 where @xmath35 , @xmath0 , and @xmath36 are respectively the hamiltonian matrices for the left / right electrodes , for the the central molecular region , and the matrix representing the coupling between the central molecule and the left / right electrodes . in this case , we considered only nearest neighbor interactions , so that the terms @xmath36 are scalar values .
the matrix elements of @xmath0 are given by @xmath37_{m , n}=\alpha_{n}\delta_{m , n}+v_{m , n}(1-\delta_{m , n})$ ] , where @xmath38 and @xmath39 are on - site energies and hopping integral between @xmath40-th and @xmath41-th atomic orbitals , respectively .
in this study , the electrodes are approximated as semi - infinite 1d chains with a single - orbital per site , thus the hamiltonians of the semi - infinite electrodes @xmath35 have similar forms with @xmath0 .
for simplicity we only consider the interaction between nearest neighbors .
the matrix elements of left / right electrodes are given by @xmath42_{m , n}=\alpha_{{\rm l / r}}\delta_{m , n}+\beta_{l / r}(1-\delta_{m , n})$ ] .
hereafter we normalize all hopping integrals , @xmath43 and @xmath39 , in terms of @xmath44 and set all on - site energies as 0 , @xmath45 . in the orthogonal basis the retarded green function
is given by @xmath46^{-1},\ ] ] where @xmath47 , @xmath48 , and @xmath49 are the identity matrix , the positive infinitesimal , and the self energy in the left / right electrode , respectively .
the self - energy term @xmath50 is defined as @xmath51 , where @xmath52 is the surface green function of left / right 1d electrode .
the left / right electrode consists of equally spaced 1d sites and coupling between nearest neighbors is set as @xmath14 .
the surface green function of the electrode consisting of equally spaced 1d sites considering only nearest neighbor interaction can be analytically described by the newns - anderson model@xcite as @xmath53 whose derivation@xcite is summarized in appendix b. the conductance of a molecular junction in a low bias is estimated from landauer s formula @xmath54 , where @xmath55 is the conductance quantum and @xmath56 is the electronic transmission probability .
@xmath56 is obtained from the fisher - lee relation@xcite : @xmath57 $ ] , where @xmath58 is the broadening function defined as @xmath59 $ ] .
the dos is calculated from following equation : @xmath60 /2\pi.\ ] ] in this study we only consider coherent transport and ignore internal scattering effects such as inelastic transport or incoherent transport@xcite associated with electron - phonon coupling.@xcite
dimerization and soliton formation are closely related processes which have been extensively studied .
the dimerization has been examined the first in the context of metal - insulator transition where dimerization is known as peierls distortion,@xcite while the soliton formation has been examined in the field of conducting 1d polymers.@xcite these effects strongly modify the conductance and other electronic properties of 1d polymers .
thus , such effects in single molecular wires connected between contacts also should be considered . in order to estimate the influence of solitons on transport
, we modeled a 1d molecular system consisting of an odd number of sites coupled with electrodes , and examined the dependence of the transmission spectra , total dos ( tdos ) and local dos ( ldos ) on the position of a soliton . for simplicity we assumed that there is a single soliton on a molecular wire
however , this approach can be also applied to larger numbers of solitons on molecular wires .
figure [ case18 ] shows modeled poly - acetylene - based ( pa - based ) molecular wires consisting of 19 sites connected between 1d electrodes and allowing for possible 8 soliton positions along the molecular wires .
transfer integrals associated with double bonds and single bonds are set to @xmath8 and @xmath9 , respectively .
the two hopping parameters on either side of the soliton site are set to @xmath10 .
all on - site energies are set to @xmath11 .
the coupling constants to the electrodes are set as @xmath12 .
the 1d electrodes are treated by the newns - anderson model@xcite with @xmath13 and @xmath14 .
fermi energy is set to @xmath15 .
[ soliton - tmdos ] shows calculated transmission spectra and tdos of the 1d molecular systems with different soliton positions .
interestingly , depending on the position of the soliton , the transmission at the fermi energy ( @xmath16 ) changes drastically ( see animation in supplemental information ) .
when the soliton lies near an electrode , the resonances in the transmission spectra and tdos at @xmath16 broaden with lower peaks . in this case , in spite of having an odd number of sites , the features of the tdos and transmission resemble those for the pa - based molecular junctions which have an even number of sites ( see fig .
s4(a ) or ( c ) in supplemental information ) . on the other hand ,
when a soliton lies in the middle of the molecular wire , the resonances in the transmission spectra and tdos sharpen and get narrower .
the value of the transmission at @xmath16 approaches its theoretical maximum of 1.0 . in order to elucidate the relationship between the positions of a soliton and sharpness of resonant peaks ,
the distribution of ldos along the molecular framework was calculated .
figure [ soliton - walk ] shows the ldos surface plot as a function of energy and site index .
the ldos surface plot has peaks near the fermi energy ( @xmath16 ) where the soliton is located .
the ldos peaks at @xmath16 broaden and reduce in height when the soliton is located at the terminal of the molecular wire ( fig .
[ case18](a ) and ( h ) ) , whereas the ldos peaks at @xmath16 sharpen and narrow when the soliton is located in the middle of molecule ( fig .
[ case18](d ) and ( e ) ) . this feature can be explained as follows .
when a soliton lies at the terminal of a molecular wire , the localized state on the molecular wire strongly interacts with surface states of the lead broadening the ldos peaks around @xmath16 .
this means an electron occupying the soliton state easily escapes back to the electrode where the electron was injected from , so that the chance of the electron traveling through to the other electrode is reduced resulting in low transmission at @xmath16 .
meanwhile , when a soliton lies at the center of a molecular wire , the localized state of the molecular wires interacts symmetrically and weakly with surface states of the two leads producing sharp ldos peaks .
this conclusion can also be understood in terms of the generalized symmetry condition.@xcite the transmission peak close to the resonance is approximately described as @xmath17 , where @xmath18 is the energy of the molecular orbital associated to the soliton .
@xmath19 follows a simple relation , @xmath20 , where @xmath21 is the orbital amplitude of the molecule at the left / right interface .
thus , symmetric coupling of the molecule with two electrodes , @xmath22 , which corresponds to the soliton localization at the center of the molecule in case 4 or 5 in fig .
[ case18 ] , gives optimal transmission probabilities , while asymmetric coupling , as can be seen in case 1 or 8 in fig .
[ case18 ] , gives lower transmission probabilities .
we have examined how the position of a soliton modifies the conductance of molecular wires and seen that the conductance is greatly reduced when the soliton lies near electrodes . here , we investigate the influence of the defect and demonstrate the counter - intuitive result that the introduction of a defect satisfying special conditions creates a transmission channel leading to higher conductance than in defect free systems . in general
, realistic molecular junctions will contain defects .
the defects may induce disorder in the molecular framework . in the presence of defects
both the on - site energies and the hopping integrals are modified .
therefore , the transport properties of the molecular junctions will also be changed .
thus we need to consider the effect of defects on quantum transport .
we will now examine the dependence of the transmission and dos on the position of the defect .
here we only consider fluctuations in the hopping integrals and introduce a defect as a reduced transfer integral in a molecular system with an even number of sites . for simplicity
we consider a single defect in the molecular system .
figure [ defect - case ] shows the schematic description of 1d molecular wire consisting of 6 sites , with a single defect , coupled to two electrodes .
transfer integrals for nearest neighbors and coupling strength at the left and right interfaces are set to @xmath23 and @xmath24 , respectively .
the defect is introduced as a reduction of the transfer integral @xmath25 between the two sites on either side of the defect to a value of @xmath26 .
we considered 5 cases of different positions of a single defect on a molecular wire and investigated position - dependence of the transmission and dos of the molecular wire .
figure [ defect - tmdos ] shows the transmission spectra and tdos of the molecular wire shown in fig .
[ defect - case ] .
depending on the position of the defect , the transport properties change dramatically particularly around @xmath16 .
[ iv ] shows i - v curves of the molecular wire calculated in the linear response regime .
surprisingly , the transmission at @xmath16 in cases 1 , 3 , and 5 is higher than for defect free systems ( fig . [ defect - tmdos ] ) . in order to further analyze this behaviour , we calculated the ldos as a function of energy and site index .
figure [ defect - ldos ] shows the surface plot of ldos as a function of electronic energy and site index .
the position of the defect is illustrated by changing the colour of the ldos curve from red to blue on either side .
when a defect is introduced such that the molecular wire is divided into two fragments which have an odd number of sites , for example @xmath27 , the ldos plot of each fragment , as shown in fig .
[ defect - ldos](c ) , resembles that of a linear chain having 3 sites ( @xmath28 ) without defects , as shown in fig .
the ldos in fig .
[ defect - ldos](c ) shows non - negligible split peaks around @xmath16 .
likewise when a defect is introduced such that the molecular wire is divided into fragments containing @xmath29 or @xmath30 sites , the ldos plot of longer fragment resembles that of a linear chain having 5 sites ( @xmath31 ) without defects as can be seen in fig .
the ldos in fig .
[ defect - ldos](a ) and ( e ) show non - negligible split peaks around @xmath16 .
this is why , despite including the single defect , these cases lead to higher transmission probabilities and tdos around @xmath16 than the defect free systems shown in fig .
[ defect - tmdos ] and fig .
[ iv ] .
on the other hand , when the defect is introduced such that the molecular wire is divided into two fragments both having an even number of sites , for example @xmath32 , the ldos plot of each fragment resembles that of a linear chain having an even number of sites without defects .
the ldos in fig .
[ defect - ldos](b ) and ( d ) are small around @xmath16 .
thus , these cases lead to low transmission probabilities and tdos around @xmath16 as shown in fig . [ defect - tmdos ] .
in summary , in order to elucidate the factors controlling charge transport through 1d molecules connected between electrodes we have modeled molecular junctions using the tight - binding method .
then we calculated the the influence of the parameters including molecular lengths , degree of dimerization , odd - even effects , soliton formation , and defects on transport properties of the molecular junctions at equilibrium using the landauer approach combined with the green s function formalism .
the numerical calculations have shown that the transport properties at the fermi energy ( @xmath33 ) dramatically change depending on the degree of dimerization and on whether the number of the sites is odd or even .
it has been shown that dimerization of the molecular wires associated with peierls distortion reduces the dos at the fermi energy , leading to exponential decay of length dependence of conductance of the molecular wires .
we also proved that the damping factor is closely related to the degree of dimerization of the molecular wires .
interestingly , the longer chain system without dimerization showed no decay in the conductance at the fermi energy . in the case of molecular wires with an odd number of sites , the extra non - bonding state appears in the middle of the band gap contributing to high conductance at low bias .
additionally , in the molecular wires with an odd number of sites , the conductance at the fermi energy was independent of the length of the molecular wires .
our calculations also demonstrate that the transport properties are highly sensitive to the position of solitons and defects .
when a soliton lies near one electrode it strongly interacts with surface states of that electrode giving rise to a low , broad peaks in the zero - bias conductance , while when a soliton lies in the middle of the molecular wire , the interaction of the localized state with the electrodes is weaker leading to high conductance at the fermi energy with sharp resonant peak .
concerning on the calculation of the influence of defects , we obtained a counter - intuitive result that in some cases defects increase the zero - bias conductance .
our findings are that a defect which divides the molecular wire into two fragments both having an odd number of sites creates a new conduction channel and enhances the zero - bias conductance .
since thermal fluctuation naturally creates this sort of defects that , at least for a while , strongly favors charge transfer , we might infer that this is one of the underlying mechanisms of transport at room temperature .
we believe that the presented systematic study could be predictive for a wider material ( _ e.g. _ semiconductor / organic interfaces ) .
our results could guide the synthesis of novel molecular wires with enhanced conductors .
this work was partially funded by the volkswagen foundation and by the wcu ( world class university ) program through the korea science and engineering foundation funded by the ministry of education , science and technology ( project no .
r31 - 2008 - 000 - 10100 - 0 ) , and the european social funds in saxony and the cluster of excellence `` ecemp
european centre for emerging materials and processes dresden '' within the excellence initiative of the free state of saxony .
we acknowledge the center for information services and high performance computing ( zih ) at the dresden university of technology for computational resources .
the work of horacio m. pastawski was possible under the eramus mundus master s program at biotechnologische zentrum ( biotec ) .
we thank claudia gomes da rocha , cormac toher , and luis fo torres for useful discussions .
here we derive the analytic solution of surface green function of a semi - infinite 1d electrode consisting of equally spaced sites .
consider a semi - infinite 1d electrode with lattice constant @xmath61 as shown in fig .
[ appb1 ] .
each site has only one orbital whose position is given by @xmath62 and its on - site energy is @xmath63 .
we only consider the interaction between nearest neighbors .
the hopping integral and overlap between nearest neighbors are @xmath44 and @xmath64 , respectively .
energy - dependent coupling between nearest neighbors is defined as @xmath65 .
the wave function of semi - infinite electrode can be constructed from forward- and backward - propagating plane wave @xmath66 : @xmath67 the bra and ket are defined as @xmath68 thus , the wave function at site @xmath41 is given by : @xmath69 this function satisfies the boundary condition of semi - infinite 1d electrode that the wave function on site @xmath70 is zero since @xmath71 .
the operator of green s function is defined as @xmath72 the free propagator of the electrode between sites @xmath41 and @xmath73 is given by @xmath74 especially in the case of @xmath75 we get @xmath76 we only need the matrix element at the terminal site since this is the only site where the molecular system couples , thus putting @xmath77 gives @xmath78 hereafter we drop @xmath79 from @xmath80 . since the number of sites is very large thus the summation can be converted into integral ( @xmath81 ) leading to @xmath82 using e - k relation of 1d electrode consisting of equally spaced sites ( @xmath83 ) , we get @xmath84 when we define @xmath85 , the variable of integral can be changed as @xmath86 when we define @xmath87 , then we get @xmath88 , @xmath89 , and @xmath90 . substituting these expressions into @xmath91 , we obtain @xmath92 where @xmath93 . the term @xmath94 can be changed to @xmath95 .
these @xmath96 and @xmath97 must satisfy following relations : @xmath98 and @xmath99 .
note that @xmath100 satisfies these relations .
then @xmath101 is written as @xmath102 we can solve this complex integral using residue theorem : @xmath103_{z = z_k}.\ ] ] figure [ decta2 ] shows the singular points of function @xmath104 on complex variable plane . the residual term for res@xmath105_{z = z_2}$ ] vanishes since the imaginary positive infinitesimal term @xmath106 move the singular point at @xmath97 outside of the circle . therefore , @xmath107 is where @xmath109_{z = z_1 } = [ ( z - z_1)f(z)]_{z = z_1}=z_1-z_2,\ ] ] and @xmath110_{z = z_0 } = \left [ \frac{{\rm{d } } } { { \rm{d } } z } \frac{(z^2 - 1)^2}{(z - z_1)(z - z_2 ) } \right]_{z = z_0 } = z_1+z_2.\ ] ] finally , we get the surface green function of 1d electrode consisting of equally spaced sites as @xmath111 where , @xmath112 is energy - dependent coupling strength between the electrode and molecule in non - orthogonal description , where @xmath113 is the overlap between the terminal site in the electrode and the molecule .
99 cuniberti g , fagas g and richter k , eds .
2005 _ introducing molecular electronics . _
( springer - verlag berlin heidelberg ) agrat n , levy - yeyati a and van ruitenbeek j m 2003 _ phys
* 377 * 81 hisamoto d , lee w - c , kedzierski j , takeuchi h , asano k , kuo c , anderson e , king t - j , bokor j and hu c 2000 _ ieee , trans .
. dev . _ * 47 * 2320 chang p - c , fan z , chien c - j , stichtenoth d , ronning c and lu j g 2006 _ appl .
* 89 * 133113 huang y , duan x , cui y and lieber c m 2002 _ nano lett . _ * 2 * 101 singh n , agarwal a , bera l k , liow t y , yang r , rustagi s c , tung c h , kumar r , lo g q , balasubramanian n and kwong d - l 2006 _ ieee , elec
* 27 * 383 ohnishi h , kondoh y and takayanagi k 1998 _ nature _ * 395 * , 780 yanson a i , bollinger g r , vand den brom h e , agrat n and van ruitenbeek j m 1998 _ nature _ * 395 * 783 smit r h m , untiedt c , rubio - bollinger g , segers r c and van ruitenbeek j m 2003 _ phys .
_ * 91 * 076805 bettini j , sato f , coura p z , dantas s o , galvo d s and ugarte d 2006 _ nature nanotechnol . _ * 1 * 182 nilius n , wallis t m and ho w 2003 _ phys .
_ * 90 * 186102 jin c , lan h , peng l , suenaga k and iijima s 2009 _ phys . rev .
lett . _ * 102 * 205501 chuvlin a , meyer j c , algara - siller g and kaiser u 2009 _ new j. phys . _ * 11 * 083019 landauer r 1981 _ phys . lett . _
* 85a * 91 bttiker m 1986 _ phys .
_ * 57 * 1761 ellenbogen j c and love j c 2000 _ proceedings of the ieee _ * 88 * 386 joachim c , gimzewski j k and aviram a 2000 _ nature _ * 408 * 541 cui y , wei q , park h and lieber c m 2001 _ science _ * 293 * 1289 patolsky f , zheng g , hayden o , lakadamyali m , zhuang x and lieber c m 2004 _ proc .
usa _ * 101 * 14017 de cola l ed . _ molecular wires : from design to properties . _ _ top .
_ * 257 * ( springer - verlag berlin heidelberg , 2005 ) .
calev y , cohen h , cuniberti g , nitzan a and porath d 2004 _ israel j. chem .
_ * 44 * 133 lang n d 1997 _ phys .
_ * 79 * 1357 forrest s r 2004 _ nature _ * 428 * 911 burroughes j h , bradley d d c , brown a r , marks r n , mckay k , friend r h , burns p l and holmes a 1990 _ nature _ * 347 * 539 friend r h , gymer r w , holmes a b , burroughes j h , marks r n , taliani c , bradley d d c , dos santos d a , brdas j l , lgdlund m and salaneck w r 1999 _ nature _ * 397 * 121 kulkarni a p , tonzola c j , babel a and jenekhe s a 2004 _ chem .
* 16 * 4556 katz h e , bao z and gilat s l 2001 _ acc . chem .
res . _ * 34 * 359 grtzel
m 2001 _ nature _ * 414 * 338 gnes s , neugebauer h and sariciftci n s 2007 _ chem .
rev . _ * 107 * 1324 blom p w m , mihailetchi v d , koster l j a and markov d e 2007 _ adv .
_ * 19 * 1551 chiang c k , fincher jr .
c r , park y w , heeger a j , shirakawa h , louis e - j , gau s c and macdiarmid a g 1977 _ phys .
lett . _ * 39 * 1098 chiang c k , dury m a , gau s c , heeger a j , louis e - j , macdiarmid a g , park y w and shirakawa h 1978 _ j. am .
soc . _ * 100 * 1013 gregg b a 2009 _ j. phys . chem .
c _ * 113 * 5899 su w p , schrieffer j r and heeger a j 1979 _ phys . rev .
lett . _ * 42 * 1698 mujica v , kemp m and ratner m a 1994 _ j. chem .
* 101 * 6849 mujica v , kemp m and ratner m a 1994 _ j. chem .
phys . _ * 101 * 6856 pastawski h m , weisz j f and albanesi e a 1995 _ phys . rev .
b _ * 52 * 10665 pastawski h
m and medina e 2001 _ rev .
fis . _ * 47s1 * 1 tyler d r 2007 _ frontiers in transition metal - containing polymers _ ( wiley - interscience , weinheim , germany ) bera j k and dunbar k r 2002 _ angew
ed . _ * 41 * 4453 peierls r e 1955 _ quantum theory of solids _
( clarendon press ; oxford , u.k . )
conwell e m 1988 _ semiconductors and semimetals _ ( academic press : dordrecht , netherlands ) anderson p w 1961 _ phys .
_ * 124 * 41 newns d m 1969 _ phys .
_ * 178 * 1123 ricco b and azbel m y 1984 _ phys . rev .
b _ * 29 * 1970 fo torres l e f , pastawski h m and makler s s 2001 _ phys .
* 193304 damato j l , pastawski h m and weisz j f 1989 _ phys .
b _ * 39 * 3554 levstein p r , pastawski h m and damato j l 1990 j. phys . : condens .
matter * 2 * 1781 emberly e g and kirczenow g 1998 _ phys .
b _ * 58 * 10911 fisher d s and lee p a 1981 _ phys . rev .
b _ * 23 * 6851 nozaki d , girard y and yoshizawa k 2008 _ j. phys .
c _ * 112 * 17408 ryndyk d a and cuniberti g 2007 _ phys .
* 76 * 155430 frederiksen t , brandbyge m and jauho a - p 2004 _ j. comp . elec . _ * 3 * 423
|
understanding of charge transport mechanisms in nanoscale structures is essential for the development of molecular electronic devices .
charge transport through 1d molecular systems connected between two contacts is influenced by several parameters such as the electronic structure of the molecule and the presence of disorder and defects . in this work ,
we have modeled 1d molecular wires connected between electrodes and systematically investigated the influence of both soliton formation and the presence of defects on properties such as the conductance and the density of states .
our numerical calculations have shown that the transport properties are highly sensitive to the position of both solitons and defects .
interestingly , the introduction of a single defect in the molecular wire which divides it into two fragments both consisting of an odd number of sites creates a new conduction channel in the center of the band gap resulting in higher zero - bias conductance than for defect free systems .
this phenomenon suggests alternative routes toward engineering molecular wires with enhanced conductance .
|
the narrow - line seyfert 1 ( nls1 ) galaxies , a subclass of active galactic nuclei ( agn ) have been the centre of interest because of their extreme variability in the x - ray band @xcite .
the defining properties of this class of agn are : balmer lines with full width at half - maximum fwhm(h@xmath11 ) @xmath12 km s@xmath13 @xcite , strong permitted optical / uv fe ii emission lines @xcite and weaker [ oiii ] emission @xmath14\lambda5007}{h_{\beta}}\leq3 $ ] @xcite .
the x - ray spectra of seyfert galaxies show a power - law like primary continuum which is thought to arise due to thermal comptonization of the optical / uv seed photons in a corona of hot electrons surrounding the central supermassive black hole ( e.g.@xcite ) .
the optical / uv photons are thought to arise from an accretion disk but direct observational evidence is lacking .
similarly , the interplay between the accretion disk and the hot corona is not well understood .
many type 1 agn also show strong ` soft x - ray excess ' emission over the power - law continuum below @xmath15 kev in their x - ray spectra .
the existence of this component ( @xmath16 kev ) was discovered around 30 years ago ( e.g. @xcite ) and its origin is still controversial .
initially , it was considered to be the high energy tail of the accretion disk emission ( @xcite ) , but the temperature of the soft excess is in the range @xmath17 kev which is much higher than the maximum disk temperature expected in agn .
it was then speculated that the soft excess may result from the compton up - scattering of the disk photons in an optically thick , warm plasma ( e.g. @xcite ) .
currently , there are two competing models for the soft x - ray excess emission in terms of optically thick thermal comptonization @xcite and blurred reflection from partially ionized accretion disk @xcite .
however , these models sometimes lead to spectral degeneracy because of the presence of multiple spectral components in the energy spectra of nls1 galaxies @xcite .
one efficient approach to overcome the spectral model degeneracy is to study the root mean square ( rms ) spectrum which links the energy spectrum with variability and has been successfully applied in a number of agn ( mcg6 - 30 - 15 : @xcite , 1h 0707495 : @xcite , rx j1633.3 + 4719 : @xcite ) .
observational evidence for the emission in different bands such as uv , soft and hard x - rays during large variability events may help us to probe the connection between the disk , hot corona and the soft excess emitting regions . in this manuscript , we investigate the origin of the soft x - ray excess emission , rapid x - ray variability and uv / x - ray connection in pg 1404 + 226 with the use of both model dependent and model independent techniques .
pg 1404 + 226 is a narrow - line seyfert 1 galaxy at a redshift @xmath18 = 0.098 with fwhm(h@xmath11 ) @xmath19 km s@xmath13 @xcite .
previously , the source was observed with _ rosat _ ( @xcite ) , _ asca _ ( @xcite ) , _ chandra _ ( @xcite ) and _ xmm - newton _
( @xcite ) . from the _ asca _ observation , the 2 - 10 kev spectrum
was found to be quite flat ( @xmath20 ) with flux @xmath21 erg @xmath22 s@xmath13 @xcite .
the detection of an absorption edge at @xmath23 kev was claimed in previous studies and interpreted as high velocity ( 0.2 - 0.3 @xmath24 ) outflow of ionized oxygen @xcite .
the source is well - known for its strong soft excess and large - amplitude x - ray variability on short time scales ( @xcite ) .
here we explore the energy spectrum , rms spectrum , flux flux plot and uv / x - ray connection through a new @xmath25 ks _ xmm - newton _ observation of pg 1404 + 226 .
we observed pg 1404 + 226 with the _ xmm - newton _
telescope @xcite on 25th january 2016 ( obs .
i d 0763480101 ) with a total duration of @xmath25 ks . here
we analyze data from the european photon imaging camera ( epic - pn ; @xcite ) and optical monitor ( om ; @xcite ) on - board _ xmm - newton_. we processed raw data with the scientific analysis system ( sas v.15.0.0 ) and the most recent ( as of 2016 august 2 ) calibration files .
the epic - pn detector was operated in the large window ( lw ) mode ( @xmath26 per cent ` livetime ' ) using the thin filter .
we processed epic - pn data using the task epproc and produced calibrated photon event file .
we included unflagged single and double pixel events ( @xmath27 ) and filtered the processed pn event .
we generated a gti ( good time interval ) file above 10 kev for the full field with rate@xmath28 counts s@xmath13 using the task tabgtigen and excluded proton background flares .
this resulted in a filtered duration of @xmath29 ks for the cleaned epic - pn data .
we extracted the source event from a circular region of radius 35 arcsec centered on the source while background events were extracted from nearby source free circular region with a radius of 50 arcsec .
the rmf ( redistribution matrix file ) and arf ( ancillary region file ) were produced with the sas tasks rmfgen and arfgen , respectively .
we extracted the source and background light curves for different energy bands and bin times from the cleaned pn event file using xselect v.2.4 c. the background subtracted light curves were produced using the ftool task lcmath .
we grouped the spectral data using the grppha tool in order to have a minimum of 50 counts per energy bin .
the net count rate estimated for epic - pn is ( @xmath30 ) counts s@xmath13 resulting in a total of @xmath31 pn counts .
the optical monitor ( om ) was operated in the imaging - fast ( if ) mode using only uvw1 ( @xmath32 ) filter for a total duration of 94 ks .
there are a total of 20 uvw1 exposures and we found that only the last 14 exposures were acquired simultaneously with the filtered epic - pn data .
we processed the fast mode om data with the sas task omfchain and generated the uvw1 light curve .
interval of the time - delay between the uv and x - ray light curves .
the time - lag was estimated using the maximum likelihood method . ] [ ccf_lag ] [ lag ] [ fig4 ] [ fig5 ] [ fig5a ] [ fig5b ]
the top panel of figure [ fig1 ] shows the background subtracted epic - pn light curve of pg 1404 + 226 with time bins of 500 s. the x - ray time series clearly shows a short - term , large - amplitude variability event in which pg 1404 + 226 varied by a factor of @xmath0 in @xmath1 ks during the 2016 observation . the fractional root mean square ( rms ) variability amplitude estimated in the 0.3@xmath3310 kev band is @xmath4=89.0@xmath34 ( the uncertainty on @xmath35 was calculated in accordance with @xcite ) .
based on the variability pattern , we divided the entire 73 ks light curve into five intervals .
int 1 consists of the first 38 ks of the time series and had the lowest flux with mean count rate of @xmath36 counts s@xmath13 and moderate fractional rms variability of @xmath37=12.5@xmath38 . in int 2 , the x - ray flux increases exponentially by a factor of @xmath39 with fractional rms variability of @xmath40=47.5@xmath41 . the duration and
mean flux level of int 2 are @xmath42 ks and 0.51@xmath43 counts s@xmath13 , respectively .
during int 3 , the source was in the highest flux state , with mean count rate of 1.38@xmath44 counts s@xmath13 and fractional rms amplitude of @xmath45=11.8@xmath46 .
the source was in the brightest state only for @xmath47 ks and then count rate started decreasing . in int 4 ,
the source flux drops by a factor of @xmath48 in @xmath47 ks with @xmath49=47.2@xmath38 . during the end of the observation ,
the source was moderately variable with @xmath50=13.3@xmath51 .
in the bottom panel of fig .
[ fig1 ] , we show the uvw1 light curve of pg 1404 + 226 simultaneous with the x - ray light curve . the amplitude of the observed uv variability is only @xmath52 of the mean count rate on timescales of @xmath53 ks .
the fractional rms variability amplitude in the uvw1 band is , @xmath2=3.9@xmath3 which is much less compared to the x - ray variability .
the x - ray and uv variability patterns appear significantly different suggesting lack of any correlation between the x - ray and uv emissions at zero time - lag .
the duration and fractional rms amplitude for each interval are listed in table [ table0 ] .
the upper and middle panels of figure [ fig2 ] show the background - subtracted soft ( 0.3@xmath331 kev ) and hard ( 1@xmath3310 kev ) x - ray light curves , respectively , with time bins of 2 ks .
the soft band is observed to be brighter than the hard band , however the variability pattern and amplitude ( @xmath54 and @xmath55 ) in these two bands are found to be comparable during the observation .
the peak - to - trough ratio of the variability amplitudes in the soft and hard bands are of the order of @xmath56 and @xmath57 , respectively . in the bottom panel of fig .
[ fig2 ] , we have shown the hardness ratio as a function of time . a constant model fitted to the hardness ratio curve provided statistically poor fit ( @xmath58/d.o.f=47/31 ) , implying the presence of moderate spectral variability and the source became harder in the beginning of the large amplitude variability .
[ table0 ] to investigate the connection between the x - ray and uv emitting regions , we calculated the cross - correlation function ( ccf ) between the uvw1 and x - ray ( 0.3@xmath3310 kev ) light curves using the z - transformed discrete correlation function ( zdcf ) @xcite . in the zdcf method ,
all pairs from two time series are ordered according to their time difference , @xmath59 .
the ordered list is then divided into time lag bins , each comprising of a minimum of @xmath60 pairs .
the cross - correlation for each lag bin is calculated as @xmath61 where @xmath60 is the number of pairs in a time lag bin , @xmath62 , @xmath63 are the bin averages , and @xmath64 , @xmath65 are the bin standard deviations .
the zdcf method uses the fishers z - transformation of @xmath66 to estimate the confidence intervals of a measured bin correlation coefficients .
the zdcf method also determines the peak of the ccf and the corresponding lag uncertainty based on the monte carlo simulations @xcite .
we have used minimum of 22 data points in a time bin and 10000 realizations for monte carlo runs to estimate the time - lag .
figure [ ccf ] shows the zdcf for the observed uvw1 and x - ray light curves with @xmath67 bins .
the time - lag was estimated relative to the uvw1 band and the positive lag implies that the uv is leading the x - ray variations .
we observed both positive and negative correlations between the uv and x - ray bands with ccf@xmath68 and @xmath69 at time - lags of @xmath6 ks and @xmath70 ks , respectively .
we estimated the time - lag in the range @xmath71 to 60 ks using the maximum likelihood method ( plike code , @xcite ) .
the lag of x - rays by uvw1 is @xmath72 ks , where the errors represent the 1-@xmath73 confidence intervals . to determine the significance of the observed time - lag and ccf peak , we generated @xmath74 pair of simulated light curves by bootstrapping method using astropy v.1.3 @xcite .
we estimated the zdcf and time - lag for each pair of simulated light curves .
the resulting ccf and lag distributions are shown in figure [ ccf_lag ] . clearly , the detection of the uv lead corresponding to the positive correlation is statistically robust . to check the accuracy of the estimated time - lag
, we have generated the distribution of the time - lag from @xmath75 realizations of the data using javelin cross - correlation program @xcite in the range of @xmath76 ks .
[ lag ] shows the probability distribution of the x - ray time - lag with respect to the uvw1 band .
the estimated time - lag is @xmath77 ks .
we performed the spectral analysis of pg 1404 + 226 using xspec v.12.8.2 @xcite .
we employed the @xmath58 statistic and quote the errors at the @xmath78 confidence limit for a single parameter corresponding to @xmath79 , unless otherwise specified .
we begin our spectral analysis by fitting the 1@xmath338 kev epic - pn spectrum using a continuum model ( zpowerlw ) multiplied by the galactic absorption ( tbabs ) .
we fixed the galactic column density at @xmath80 @xmath22 @xcite .
this model provided @xmath58=71 for 50 degrees of freedom ( d.o.f ) with @xmath81 and can be considered as a good baseline model to describe the hard x - ray emission from the source .
then we extrapolated our 1@xmath338 kev absorbed power - law model ( tbabs@xmath82zpowerlw ) down to 0.3 kev .
this extrapolation reveals the presence of a strong soft excess emission below 1 kev with @xmath58/d.o.f = 11837/160 .
the ratio of the observed epic - pn data and the absorbed power - law model is shown in figure [ fig4 ] ( left ) .
the fitting of the 0.3@xmath338 kev data with the absorbed power - law model ( tbabs@xmath82zpowerlw ) resulted in a poor fit with @xmath58/d.o.f = 1126.5/158 .
the residual plot demonstrates a sharp dip around @xmath23 kev and excess emission below 1 kev .
initially , we modelled the soft excess emission using a simple blackbody model ( zbbody ) .
the addition of the zbbody model improved the fit to @xmath58/d.o.f = 232.1/156 ( @[email protected] for 2 additional free parameters ) . in xspec ,
the model reads as tbabs@xmath82(zbbody@xmath84zpowerlw ) .
the deviations of the observed epic - pn data from the absorbed blackbody and power - law model is shown in fig .
[ fig4 ] ( right ) .
the estimated blackbody temperature @xmath85 ev is consistent with the temperature of the soft excess emission observed in seyfert 1 and qsos @xcite . in order to model the absorption feature @xmath23 kev , we used one ionized absorber model ( zxipcf ) which has 4 free parameters : column density ( @xmath86 ) , covering fraction ( c@xmath87 ) , redshift ( @xmath18 ) and ionization parameter ( @xmath88 , where @xmath89
, @xmath90 is the source luminosity , @xmath60 is the hydrogen density and @xmath66 is the distance between the source and cloud ) .
we assumed that the absorbing cloud to cover the x - ray source fully and fixed the covering fraction at 1 .
the inclusion of the warm absorber significantly improved the fit from @xmath58/d.o.f = 232.1/156 to 176.5/154 ( @[email protected] for 2 additional free parameters ) . to test the presence of out - flow
, we varied the redshift of the absorbing cloud which further improved the fit to @xmath58/d.o.f = 168.1/153 ( @[email protected] for 1 additional free parameter ) . in xspec , the 0.3@xmath338 kev best - fitting model reads as tbabs@xmath82zxipcf(zbbody@xmath84zpowerlw ) .
the epic - pn spectral data , the best - fitting model and the deviations of the observed data from the best - fitting model are shown in figure [ fig5 ] ( left ) .
the best - fitting values for the column density , ionization parameter and outflow velocity ( determined from the redshift of the cloud relative to the rest frame ) of the warm absorber are @xmath91 , @xmath88(erg cm s@xmath92 and @xmath93 , respectively . in order to search for spectral variability on shorter timescales ,
we performed time - resolved spectroscopy . first we applied our best - fitting mean spectral model to the 5 epic - pn spectra extracted from the five intervals ( see section 3.1 ) .
we tied all the parameters except the normalization of the power - law and blackbody components which we set to vary independently .
this resulted in statistically acceptable fit with @xmath58/d.o.f = 337.7/269 = 1.25 and without strong residual .
if we allow the blackbody temperature and photon - index of the power - law to vary , we did not find any significant improvement in the fit with @xmath58/d.o.f = 327.7/261 = 1.25 ( @xmath83=@xmath3310 for 8 free parameters ) .
the 5 epic - pn spectral data - sets , the best - fitting model and residuals are shown in fig .
[ fig5 ] ( right ) .
the best - fitting spectral model parameters for both time - averaged and time - resolved spectra are listed in table [ table1 ] .
[ table1 ] [ table2 ] [ table3 ] in order to examine the origin of the soft x - ray excess emission , we have tested two different physical models
thermal comptonization in an optically thick , warm medium and blurred reflection from a partially ionized accretion disk .
first , we have used the intrinsic disk comtonization model ( optxagnf ) ( @xcite ) which assumes that the gravitational energy released in the disk at each radius is radiated as a blackbody emission from the outer disk radius , @xmath94 to the coronal radius , @xmath95 . inside the coronal radius ,
the gravitational energy is dissipated to produce the soft excess component in an optically thick , warm ( @xmath96 kev ) corona and the hard x - ray power - law tail in an optically thin , hot ( @xmath97 kev ) corona above the disk .
thus , this model represents an energetically self - consistent model .
the four parameters which determine the normalization of the model are the following : black hole mass ( @xmath98 ) , dimensionless spin parameter ( @xmath99 ) , eddington ratio ( @xmath100 ) and proper distance ( @xmath101 ) .
we fitted the 0.3@xmath338 kev epic - pn time - averaged spectrum with the optxagnf model modified by the galactic absorption tbabs .
we fixed the black hole mass , outer disk radius and proper distance at @xmath102 @xcite , 400@xmath103 and 397 mpc , respectively .
we assumed a maximally rotating black hole as concluded by @xcite and fixed the spin parameter at @xmath104 .
this model resulted in statistically unacceptable fit with @xmath58/d.o.f = 237.2/154 and a sharp dip around 1 kev in the residual spectrum .
as before , we used the warm absorber model ( zxipcf ) which significantly improved the fit from @xmath58/d.o.f = 237.2/154 to 177.5/152 ( @[email protected] for 2 additional free parameters ) .
we then left the redshift of the absorbing medium to vary freely in order to justify the presence of out - flow , which further improved the fit to @xmath58/d.o.f = 168.3/151 = 1.11 ( @[email protected] for 1 additional free parameter ) .
the epic - pn mean spectrum , the best - fitting absorbed disk comptonization model ( tbabs@xmath82zxipcf@xmath82optxagnf ) and the deviations of the observed data from the model are shown in figure [ fig5a ] ( left ) .
the best - fitting values for the eddington rate , coronal radius , electron temperature , optical depth and spectral index are @xmath105 , @xmath106 , @xmath107 kev , @xmath108 and @xmath109 , respectively .
then we jointly fitted the five time - resolved spectral data - sets with the absorbed disk comptonization model and kept all the parameters tied to their mean spectral best - fitting values except the eddington rate .
this provided statistically acceptable fit with @xmath58/d.o.f = 364.3/279 = 1.3 and we did not notice any strong feature in the residual spectra .
the 5 epic - pn spectral data - sets , the best - fitting disk comptonization model and residuals are shown in fig .
[ fig5a ] ( right ) .
the best - fitting spectral model parameters for both mean and time - resolved spectra are listed in table [ table2 ] .
the origin of the soft excess emission could be the blurred reflection from the ionized accretion disk @xcite .
hence we modelled the soft x - ray excess using the reflection model reflionx @xcite convolved with the model kdblur which blurs the spectrum due to general relativistic effects close to the rapidly rotating smbh @xcite .
then we fitted the 0.3@xmath338 kev epic - pn mean spectrum with the incident power - law ( zpowerlw ) and blurred reflection ( kdblur@xmath110reflionx ) models after correcting for the galactic absorption ( tbabs ) .
the parameters of the reflionx model are iron abundance ( fe / solar ) , ionization parameter ( @xmath111 , @xmath112 is the total illuminating flux , @xmath60 is hydrogen density ) , normalization ( @xmath113 ) of the reflected spectrum and photon - index ( @xmath114 ) of the incident power - law .
the convolution model kdblur has four free parameters : emissivity index ( @xmath115 , where emissivity of the reflected emission is defined by @xmath116 ) , inner disk radius ( @xmath117 ) , outer disk radius ( @xmath94 ) and disk inclination angle ( @xmath118 ) .
the outer disk radius was fixed at @xmath119 . in xspec
, the 0.3@xmath338 kev model reads as tbabs@xmath82(kdblur@xmath110reflionx@xmath84zpowerlw ) which provided reasonably good fit with @xmath58/d.o.f = 191.9/152 .
however , the residual spectrum shows an absorption feature @xmath23 kev which was modelled with the ionized absorber model ( zxipcf ) .
the multiplication of the warm absorber model improved the fit to @xmath58/d.o.f = 165.8/150 = 1.1 ( @[email protected] for 2 additional free parameters ) . in xspec ,
the 0.3@xmath338 kev best - fitting model reads as tbabs@xmath82zxipcf@xmath82(kdblur@xmath110reflionx@xmath84zpowerlw ) .
the epic - pn mean spectrum , the best - fitting absorbed blurred reflection model and the deviations of the observed data from the best - fitting model are shown in figure [ fig5b ] ( left ) .
the best - fitting values for the emissivity index , inner disk radius , disk ionization parameter , inclination angle and spectral index are @xmath120 , @xmath121 , @xmath122 erg cm s@xmath13 , @xmath123 and @xmath124 , respectively .
we also fitted the five time - resolved spectra jointly with the blurred reflection model and tied every parameter to its mean spectral best - fitting value except the normalization ( @xmath113 ) of the reflection component .
this provided an unacceptable fit with @xmath58/d.o.f = 380.2/279 = 1.36 .
we then set the normalization ( @xmath125 ) of the illuminating continuum to vary between the 5 spectra and obtain a noticeable improvement in the fitting with @xmath58/d.o.f = 334.7/274 = 1.22 ( @[email protected] for 5 additional free parameters ) .
if we leave the spectral index ( @xmath114 ) of the primary continuum to vary , we did not get any significant improvement in the fitting .
the results obtained from both time - averaged and time - resolved spectroscopy are listed in table [ table3 ] .
the 5 epic - pn spectral data - sets , the best - fitting blurred reflection model and residuals are shown in fig .
[ fig5b ] ( right ) .
the flux@xmath33flux analysis is a model - independent approach to distinguish the main components responsible for the observed spectral variability and was pioneered by @xcite and @xcite . on the basis of our x - ray spectral modelling , we identified the 0.3@xmath331 kev and 1@xmath3310 kev energy bands as representatives of the soft excess and primary emission , respectively .
then we constructed the 0.3@xmath331 vs 1@xmath3310 kev flux@xmath33flux plot which is shown in figure [ fig6 ] ( top panel ) .
the mean count rate in the soft and hard bands are @xmath126 counts s@xmath13 and @xmath127 counts s@xmath13 , respectively .
we begin our analysis by fitting the flux@xmath33flux plot with a linear relation of the from , @xmath128 , where @xmath129 and @xmath130 represent the 1@xmath3310 kev and 0.3@xmath331 kev band count rates , respectively .
the straight line model provided statistically unacceptable fit with @xmath58/d.o.f = 56.7/31 and implied that the immanent relationship between the soft excess and primary emission is not linear .
therefore we fit the flux@xmath33flux plot with a power - law plus constant ( plc ) model of the form , @xmath131 ( where @xmath132 and @xmath133 count rates ) following the approach of @xcite .
the plc model explains the data statistically well with @xmath58/d.o.f @xmath134 .
the best - fitting plc model is shown as a solid line in fig .
[ fig6 ] ( top ) .
the best - fitting power - law normalization , slope and constant values are @xmath135 counts s@xmath13 , @xmath136 and @xmath137 counts s@xmath13 , respectively .
the plc best - fitting slope is significantly greater than unity which indicates the presence of intrinsic variability in the source .
the detection of positive ` c'-value in the flux@xmath33flux plot implies the existence of a distinct spectral component which is less variable than the primary x - ray continuum and contributes @xmath138 of the 1@xmath3310 kev count rate at the mean flux level . in order to investigate this issue further , we computed the unabsorbed ( without the galactic and intrinsic absorption ) primary continuum and soft excess flux in the full 0.3@xmath338 kev band for all 5 intervals using xspec convolution model cflux and plotted the intrinsic primary power - law flux as a function of the soft excess flux which is shown in fig . [ fig6 ] ( bottom ) .
the best - fitting normalization , slope and constant parameters , obtained by fitting the @xmath139 vs @xmath140 plot with a plc model , are @xmath141 erg @xmath22s@xmath13 , @xmath142 and @xmath143 erg @xmath22s@xmath13 , respectively .
interestingly , we found steeping in the @xmath139 vs @xmath140 plot with an apparent positive constant which is in agreement with the 0.3@xmath331 vs 1@xmath3310 kev flux@xmath33flux plot .
our flux@xmath33flux analysis suggests that the primary continuum and soft excess emission are well correlated with each other , although they vary in a non - linear fashion .
mos 0.3@xmath338 kev fractional rms spectrum of pg 1404 + 226 .
the solid red line represents the best - fitting fractional rms which consists of a ` two - component phenomenological ' model in which both soft excess ( zbbody ) and primary emission ( zpowerlw ) are variable in normalization only and are correlated with each other .
bottom : the deviation of the observed fractional rms from the best - fitting model and is given with reference to @xmath144=(data - model)/@xmath73 . the solid straight line is the best - fitting residual corresponding to @xmath145 .
] in order to estimate the percentage of variability in the primary continuum and soft excess emission , and also to quantify the variability relation between them , we derived and modelled the fractional rms variability spectrum of pg 1404 + 226 .
first we extracted background subtracted light curves in different energy bands from the simultaneous and equal length ( @xmath146 ) combined epic - pn+mos data with a timing resolution of @xmath147 ks .
then we computed the frequency - averaged ( @xmath148[@xmath149@xmath150 hz ) fractional rms , @xmath151 in each light curve using the method described in @xcite .
the source shows an increase in fractional variability with energy which can be explained in the framework of two components : soft excess and primary hard x - ray emission with decreasing relative importance of soft excess and increasing dominance of intrinsic primary emission with energy .
we constructed fractional rms spectral model using our best - fitting phenomenological and physical mean spectral models in isis v.1.6.2 - 32 @xcite .
first we explored the phenomenological fractional rms spectral model in which both soft excess ( zbbody ) and primary emission ( zpowerlw ) are variable in normalization only and are coupled with each other . using the equation 4 of @xcite
, we obtained the expression for theoretical fractional rms spectral model : @xmath152}}{f_{\rm pl}(a_{\rm pl},e)+f_{\rm bb}(a_{\rm bb},e ) } \label{eu1}\ ] ] where @xmath153 and @xmath154 represent fractional changes in the normalization of the primary power - law , @xmath155 and blackbody , @xmath156 components respectively .
@xmath157 measures the correlation or coupling between @xmath155 and @xmath156 .
we then fitted the 0.3@xmath338 kev fractional rms spectrum of pg 1404 + 226 using this ` two - component phenomenological ' model ( equation [ eu1 ] ) and the best - fitting mean spectral model parameters as the input parameters for the above model .
this model describes the data reasonably well with @xmath58/d.o.f = 3.48/4 = 0.87 .
the best - fitting rms model parameters are : @xmath153=83.9@xmath158 , @xmath154=78.5@xmath159 and @xmath160 .
the fractional rms variability spectrum , the best - fitting ` two - component phenomenological ' model and residuals are shown in figure [ fig7a ] . in pg 1404 + 226
, the soft excess was modelled by two different physical models : intrinsic disk comptonization and blurred reflection from the ionized accretion disk . in order to break the degeneracy between these two possible physical scenarios , we made fractional rms spectral model considering our best - fitting disk comptonization and blurred reflection models to the time - averaged spectrum . in the disk comptonization scenario , the observed variability in pg 1404
+ 226 was driven by variation in the accretion rate as inferred from the joint fitting of 5 epic - pn spectra .
therefore the expression for theoretical fractional rms ( see @xcite ) can be written as @xmath161 where @xmath162 represents the best - fitting optxagnf model and @xmath163 is the scaled accretion rate which is the only variable free parameter in the model .
the fitting of the 0.3@xmath338 kev fractional rms spectrum using this model ( equation [ eu2 ] ) resulted in statistically poor fit with @xmath58/d.o.f = 15.3/6 = 2.6 .
the fractional variability spectrum , the ` one - component disk comptonization ' model and residuals are shown in figure [ fig7b ] ( left ) .
then we investigated the blurred reflection scenario where the origin of soft excess was explained in terms of disk irradiation @xcite . in this scenario , the rapid variability in pg 1404 + 226
can be described due to changes in the normalization of the reflection component and illuminating power - law continuum as evident from the time - resolved spectroscopy .
thus , we constructed the ` two - component blurred reflection ' model where both reflection ( kdblur@xmath110reflionx ) and intrinsic continuum ( zpowerlw ) are variable in normalization only and are perfectly coupled with each other .
mathematically , the expression for fractional rms can be written as @xmath164 where @xmath165 here @xmath166 and @xmath167 represent the best - fitting incident power - law ( zpowerlw ) and blurred reflection ( kdblur@xmath110reflionx ) components , respectively .
the two free parameters of this model are @xmath125 and @xmath113 .
we then fitted the observed fractional rms spectrum using that model ( equation [ eu3 ] ) and obtained statistically acceptable fit with @xmath58/d.o.f = 4.35/5 = 0.87 .
the fractional rms variability spectrum , the best - fitting ` two - component blurred reflection ' model and residuals are shown in figure [ fig7b ] ( right ) .
the fractional variations in the normalization of the illuminating continuum and reflected emission are @xmath168 and @xmath169 respectively .
we presented results from our _ xmm - newton _ observation of the nls1 galaxy pg 1404 + 226 and studied the large - amplitude x - ray variability , the uv / x - ray connection , the origin of the soft x - ray excess and its connection with the intrinsic power - law emission through a detailed analysis of the time - resolved x - ray spectra , frequency - averaged ( @xmath148[@xmath149@xmath150 hz ) x - ray fractional rms spectrum and uv / x - ray cross - correlation . below we summarize our results : 1 .
pg 1404 + 226 showed large - amplitude ( a factor of @xmath0 in about 11 ks ) x-ray variability with fractional rms @xmath4=89.0@xmath5 .
the rapid x - ray variability had been observed in a few nls1 galaxies ( e.g. ngc 4051 : @xcite , 1h 0707495 : @xcite , mrk 335 : @xcite ) .
the uv emission form pg 1404 + 226 is much less variable ( @xmath2=3.9@xmath3 ) compared to the x - ray variability and is observed to lead the x - ray emission by @xmath6 ks .
the source exhibited a strong soft excess below @xmath23 kev , which was modelled by intrinsic disk comptonization as well as blurred reflection component .
the x - ray ( 0.3@xmath338 kev ) spectra revealed the presence of a highly ionized ( @xmath170 erg cm s@xmath13 ) absorbing cloud along the line - of - sight with column density of @xmath171 @xmath22 .
3 . the time - resolved spectroscopy shows significant variability in the soft excess and primary power - law flux , although there was no noticeable variations in the soft excess temperature ( @xmath172 kev ) and photon index ( @xmath173 ) of the primary continuum .
4 . in the disk comptonization scenario
, the rapid x - ray variability can be attributed to a change in the mass accretion rate as indicated by time - resolved spectroscopy .
however , the modelling of the x - ray fractional rms spectrum using ` one - component disk comptonization ' model can not reproduce the observed x - ray variability pattern and indicates reflection origin for the soft excess emission ( see fig . [ fig7b ] ) .
5 . in the reflection scenario
, we found that the observed large - amplitude x - ray variability was predominantly due to two components : illuminating continuum and reflected emission which are variable in normalization only ( see fig .
[ fig7b ] ) . 6 . the inner disk radius estimated from the blurred reflection model is @xmath174 which implies a rapidly rotating black hole at the center of pg 1404 + 226 and is consistent with the measurement performed by @xcite .
the non - detection of the 6.4 kev iron emission line is perhaps due to its smearing into the broad shape in the spectrum .
we found that the soft ( 0.3@xmath331 kev ) and hard ( 1@xmath3310 kev ) band count rates are correlated with each other and vary in a non - linear manner as suggested by the steepening of the flux - flux plot .
the fitting of the hard - vs - soft counts plot with a power - law plus constant ( plc ) model reveals a significant positive offset at high energies which can be interpreted as corroboration for the presence of a less variable reflection component ( probably smeared iron emission line ) in the hard band .
the observed uv variability in pg 1404 + 226 is weak with @xmath2=3.9@xmath3 only , whereas the x - ray variability is much stronger ( @xmath4=89.0@xmath5 ) .
we found that the uv variations lead the x - ray variations by @xmath175 in pg 1404 + 226 .
similar behaviour had also been observed in another seyfert 1 galaxy mcg-6 - 30 - 15 @xcite .
the time - lag between the uv and x - ray variations in mcg-6 - 30 - 15 was found to be @xmath176 ks and interpreted as a signature of the accretion flow propagation .
the uv lead observed in pg 1404 + 226 is much shorter than that seen in mcg-6 - 30 - 15 , although pg 1404 + 226 is hosting more massive ( a factor of 10 ) black hole than mcg-6 - 30 - 15 .
the observed lead of @xmath175 for variations in the uv emission compared to the variations in the x - ray emission is likely to be the result of the thermal comptonization of the uv seed photons in a compact hot corona . in this process
, the uv lead can occur due to
( i ) the light - crossing time between the uv emitting region of the disk and the compact hot corona , and ( ii ) the time required for the multiple inverse compton scattering of uv photons into x - ray photons . the light - crossing time required by the uv photons to reach the compact hot corona
can be written as @xcite @xmath177 where @xmath178 is the scaled mass accretion rate , @xmath98 is the central black hole mass in units of @xmath179 and @xmath180 is the effective wavelength where the disk emission peaks . in the case of pg 1404 + 226 , @xmath181 , @xmath182 ( calculated from the unabsorbed flux in the energy band @xmath183 kev using the convolution model cflux in xspec ) and @xmath184 for uvw1 filter .
therefore the light crossing time between the x - ray source and the disk is @xmath185 ks , which corresponds to the peak disk emission radius of @xmath186 .
recent results on x - ray reprocessing lags @xcite have shown that the agn disks are larger than that predicted for the standard disks .
however , we did not observe the reprocessing lag .
the uv emission from pg 1404 + 226 is most likely to be dominated by the intrinsic emission from the disk . in the framework of compton up - scattering , a uv photon with energy @xmath187 after @xmath188 scatterings produces an x - ray photon of energy @xcite @xmath189 where @xmath190 , @xmath191 and @xmath192 and @xmath193 are the electron temperature and rest mass , respectively . if @xmath194 is the time delay between successive scattering and the x - ray photons are produced within the optically thin , hot corona after n scatterings , then the comptonization timescale is @xmath195 for an optically thin ( @xmath196 ) , hot ( @xmath197 kev ) corona of size @xmath198 , the estimated comptonization timescale is @xmath199 ks .
thus the observed uv lead of @xmath175 can indeed correspond to the sum of the light - crossing time ( @xmath185 ks ) between the disk and power - law emitting region and the comptonization timescale ( @xmath200 ks ) of the uv seed photons within the hot corona . in the intrinsic disk comptonization ( optxagnf )
model , the soft excess emission results from the compton up - scattering of the uv seed photons by optically thick , warm ( @xmath201 kev ) electron plasma in the inner disk ( below @xmath95 ) itself .
however , if the optically - thick corona is the inner disk , it is unlikely that it will intercept significant amount of seed photons from intermediate or outer disk .
thus , the seed photons for the optically - thick thermal comptonization must arise from the inner regions of the standard cold disk . hence , the soft excess emission from the comptonized inner disk and the intrinsic uv emission from the intermediate regions of standard cold disk are likely to be connected only via the propagation of accretion rate fluctuations , and the intrinsic uv emission from the standard disk should lead the soft excess by the propagation fluctuation timescale . assuming that the fluctuations in the accretion rate propagate with the speed of sound waves in the radial direction , the timescale of propagation fluctuation @xmath202 can be written as @xcite @xmath203 where the dynamical time scale @xmath204 , @xmath205 is the disk height at the radius @xmath206 for the observed uv emission . if we consider a thin disk for which height - to - radius ratio , @xmath207 @xcite , the propagation fluctuation timescale at this emission radius ( @xmath208 ) is of the order of @xmath209 s which is much longer than the observed uv lead
hence , the observed uv lead is unlikely to be due to the propagation of accretion rate fluctuations .
unlike the comptonized inner disk in the optxagnf model , the hot corona is likely spherical and can intercept significant fraction of uv seed photons .
thus , in the framework of optxagnf model , the uv lead is entirely due to the comptonization of seed photons in the hot corona . in this case
, it is difficult to explain the correlation between the soft excess and the power - law components unless the variability is intrinsic resulting from simultaneous gravitational energy release both to the inner , comptonized disk and the hot corona .
however , it is difficult to explain the uv x - ray connection as the propagation fluctuation timescale is longer compared to the observed uv lead .
therefore the only possible explanation for the observed uv lead is that the uv seed photons from the disk travel all the way up to the hot electron plasma , get compton up - scattered within the compact hot corona and produces the x - ray power - law component .
pg 1404 + 226 shows strong x - ray variability with increase in overall fractional rms with energy .
we attempted to explain the observed rapid variability of pg 1404 + 226 in the framework of two possible physical scenarios : intrinsic disk comptonization and blurred reflection from the ionized accretion disk . in the disk comptonization scenario ,
if the rapid x - ray variability is due to the mass accretion rate changes which is favoured by the time - resolved spectroscopy , then it can not explain the rms variability pattern below 0.4 kev and in the hard band ( see fig .
[ fig7b ] , left ) .
therefore , it is unlikely that the rapid x - ray variability is caused by variations in the mass accretion rate alone . on the other hand ,
the soft excess and primary continuum vary non - linearly ( as @xmath10 ) which indicates that the soft excess is coupled and reciprocating with the primary continuum variations .
this is in agreement with the smeared reflection scenario which is further supported by the high emissivity index ( @xmath210 ) .
moreover , the fractional variability spectrum of pg 1404 + 226 is best described by two components illuminating continuum and reflected emission which vary in normalization only and perfectly coupled with each other .
the variations in the reflected component and primary continuum are @xmath211 and @xmath212 , respectively . in the blurred reflection model ,
the soft excess is a consequence of the disc irradiation by a hot , compact corona close to the black hole .
the correlated , non - linear variability between the illuminating continuum and soft excess emission is consistent with the light bending model @xcite .
our results indicate that the thermal comptonization of the uv seed photons from the accretion disk produces the x - ray power - law component which in turn illuminates the innermost accretion disk giving rise to the soft x - ray excess emission in pg 1404 + 226 .
lm gratefully acknowledges support from the university grants commission ( ugc ) . this research has made use of processed data of _ xmm - newton _ observatory through the high energy astrophysics science archive research center online service , provided by the nasa goddard space flight center .
this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the nasa .
figures in this manuscript were made with the python package matplotlib and gui scientific plotting package veusz .
this research made use of astropy , a community - developed core python package for astronomy .
|
we present results from a detailed analysis of our 2016 _ xmm - newton _ observation of the narrow - line seyfert 1 galaxy pg 1404 + 226 which showed a large - amplitude , rapid x - ray variability by a factor of @xmath0 in @xmath1 ks .
we use this variability event to investigate the origin of the soft x - ray excess emission and the connection between the disk , hot corona and the soft excess emitting region through uv / x - ray cross - correlation , time - resolved spectroscopy and root mean square ( rms ) spectral modelling . the weakly variable uv emission ( @xmath2=3.9@xmath3 )
appears to lead the strongly variable x - ray emission ( @xmath4=89.0@xmath5 ) by @xmath6 ks .
such a uv lead is consistent with the crossing time ( @xmath7 ks ) of the seed photons from the disk to a compact ( @xmath8 ) hot corona and the time required for their thermal comptonization ( @xmath9 ks ) giving rise to the x - ray power - law emission .
the strong soft x - ray excess below 1 kev seen in the mean x - ray spectrum as well as in the time - resolved spectra is well described by both the intrinsic disk comptonization and the blurred reflection models .
the soft excess emission is found to vary together with the power - law component as @xmath10 .
the x - ray fractional rms spectrum shows an increase in variability with energy which can be described only in the framework of blurred reflection model in which both the intrinsic continuum and the reflected emission are highly variable in normalization only and are perfectly coupled with each other .
our results suggest that accretion disk provides the seed photons for thermal comptonization giving rise to the x - ray power - law component which in turn illuminates the innermost accretion disk and gives rise to the soft x - ray excess emission .
[ firstpage ]
|
this is the third paper of a series @xcite describing a new , gauge - invariant , non - perturbative formulation of qcd , which provides analytic estimates of any amplitude which sums over all relevant feynman graphs , including all possible gluon exchanges between appropriate quarks ( q ) and antiquarks ( @xmath2 ) .
all cubic and quartic gluon interactions are included .
results are expressed in terms of fradkin s most useful functional representations @xcite of the green s function @xmath3 | y \rangle$ ] , @xmath4 = [ m + \gamma \cdot ( \partial - ig a \cdot \lambda)]^{-1}\ ] ] and the closed - loop , or vacuum functional @xmath1 $ ] , @xmath5 = { { \mathrm{tr}\ , } } { \ln{[1 - i g ( \gamma \cdot a \cdot \lambda ) \ , \mathbf{s}_{c } ] } } , \quad \mathbf{s}_{c } = \mathbf{g}_{c}[0],\ ] ] where @xmath6 represents a given vector potential . since the fradkin representations for these quantities are of potential theory origin , they have relatively simple approximations in different physical situations @xcite , especially at high energies where they effectively generate eikonal models . it should be understood that we are here dealing with the simplest form of `` textbook qcd '' : one type of quark coupled to multicolored gluons . flavors and electroweak interactions
are to be added later on .
this new formulation provides an exact statement of what may be expected from conventional , or `` ideal '' qcd , where the quanta of the quark field operators are expected to have the same quantum - mechanical measurement properties as , _
e.g. _ , electrons ( perfect position measurement at the expense of momenta , and vice - versa ) . for an important category of qcd processes
which involve gluon exchange between quarks , it turns out that this theory is effectively empty , because attention has not been paid to the fact that asymptotic quarks are only found in bound states , and that their transverse coordinates can , in principle , never be exactly determined . in all previous approximate calculations such as those corresponding to gauge - dependent perturbative approximations , or to a subset of feynman graphs , or to machine calculations wherein gauge - invariance is enforced by breaking up space - time into plaquettes this difficulty has not been visible ; but for the exact , non - perturbative theory , it becomes clear that quark transverse imprecision must be included as a basic feature of the theory .
the reason for this statement , as well as the relatively simple form of results for each and every qcd amplitude , is due to the novel and unexpected property we call `` effective locality '' ( el ) , which has been shown in ref .
@xcite to be rigorously true for the exact theory . in exact qcd ,
the way in which el forces the long - held paradigm of abelian qft where internal interactions are carried by a boson propagator from one space - time point to another to change to a new form of locality , will become clear below .
in the first paper on this subject @xcite , we replaced the exact fradkin representations by their high - energy , eikonal limits ; in this paper a more rigorous approach is maintained .
the aim here is to show how the conventional , or `` ideal '' qcd formulation becomes untenable when treated in this non - perturbative formulation ; and then to adopt a more `` realistic '' , though presently phenomenological , version of qcd in which `` ideal '' troubles disappear , and where the foundation is put in place for automatic quark binding into hadrons @xcite , as well as hadron - hadron interactions of the yukawa - type @xcite . for definiteness , we again consider the case of q - q or q-@xmath2 scattering , for which the corresponding 4-point function ( unsymmetrized , for simplicity of presentation , and in configuration space , before mass - shell amputation ) is proportional to @xmath7 \ , e^{\frac{i}{4 } \int{\chi^{2 } } } } & \cdot & e^{\mathfrak{d}_{a } } \cdot e^{\frac{i}{2 } \int{\chi \cdot \mathbf{f } } + \frac{i}{2 } \int{a \cdot \left ( \mathbf{d}_{c } \right)^{-1 } \cdot a } } \\
\nonumber \quad & \cdot & \left .
\mathbf{g}_{c}^{\mathrm{i}}(x_{1 } , y_{1 } | a ) \cdot \mathbf{g}_{c}^{\mathrm{i\!i}}(x_{2 } , y_{2 } | a ) \cdot { e^{\mathbf{l}_{\mathrm{c}}[a ] } } \right|_{a \rightarrow 0 } , \end{aligned}\ ] ] where @xmath8 @xmath9 $ ] is the `` linkage operator '' which links all pairs of @xmath10 to each other and provides the complete panoply of relevant feynman graphs , @xmath11 is the gluon propagator in an arbitrary ( relativistic ) gauge , @xmath12}$ ] is halpern s functional integral @xcite which allows one to include all cubic and quartic gluon interactions , and @xmath13 is a normalization constant such that @xmath14 \ ,
e^{\frac{i}{4 } \int{\chi^{2 } } } } = 1.\ ] ] the form of ( [ eq:3 ] ) , and the mechanism which generates a gauge - invariant result , have been derived , in detail , in ref .
@xcite and @xcite , and need not be repeated here .
the fradkin representations we use are exact variants of those introduced a half - century ago @xcite , which have been re - derived and discussed in detail elsewhere @xcite . in particular
, we find it more convenient to employ a functional integral ( fi ) representation involving a space - time coordinate @xmath15 rather than the original , functional differential representation of fradkin , wherein his @xmath16 denoted the proper - time - dependent 4-velocity of a real or virtual fermion ; the relation between these two quantities is simply : @xmath17 where both @xmath15 and @xmath16 are understood to be continuous functions of their proper - time , although no such restriction is placed upon higher derivatives of @xmath16 .
the fradkin representation of @xmath0 is then @xmath18 } \ , e^ { \frac{i}{2 } \int { u \cdot ( 2h)^{-1 } \cdot u } } \\ \nonumber & & \quad \times { \left [ m - \gamma_{\mu } \frac{\delta}{\delta u'_{\mu}(s ) } \right ] } \ , \delta^{(4)}(x - y + u(s ) ) \\ \nonumber & & \quad \times \left ( e^ { -ig \int_{0}^{s}{\mathrm{d}s ' \ , u'_{\mu}(s ' ) \ , a_{\mu}^{a}(w ) \ , \tau^{a } } + g \int_{0}^{s}{\mathrm{d}s ' \sigma_{\mu \nu } \ , \mathbf{f}_{\mu \nu}^{a}(w ) \ , \tau^{a } } } \right)_{+},\end{aligned}\ ] ] where @xmath19 \ , e^ { \frac{i}{2 } \int { u \cdot ( 2h)^{-1 } \cdot u } } } , \end{aligned}\ ] ] and @xmath20 is the ( bare ) quark mass .
explicit @xmath21-dependence can be extracted from the ordered exponential ( oe ) of ( [ eq:4 ] ) , and as in ref .
@xcite , this is illustrated by suppressing the spin - dependent part of the oe .
this simplification is done simply for ease of presentation , since conclusions based upon el are left unchanged @xcite .
we therefore consider the simple oe @xmath22 \ , \int{\mathrm{d}[\alpha ] \
, e^{i\int_{0}^{s}{\alpha \cdot \omega } } \ , \left ( e^ { -i \int_{0}^{s } { \alpha \cdot \tau } } \right)_{+ } } } \\ \nonumber & & \times e^ { -ig \int_{0}^{s}{\mathrm{d}s ' \ , u'_{\mu}(s ' ) \ , a_{\mu}^{a}(y - u(s ' ) ) \ , \omega^{a}(s ' ) } } , \end{aligned}\ ] ] where @xmath23 is that normalization constant needed for the delta - functional defined by the fi @xmath24}$ ] .
because of its definition in terms of @xmath16 , and because of the @xmath25 appearing in ( [ eq:4 ] ) , one has the `` boundary conditions '' , @xmath26 and @xmath27 . with or without that neglected spin - dependence of ( [ eq:4 ] ) ,
it is clear that the @xmath21-dependence of @xmath0 is not more complicated than gaussian .
this means suppressing the @xmath1 $ ] functional momentarily that the needed linkage operations of ( [ eq:3 ] ) can be performed exactly , expressing the result in terms of the halpern and fradkin fi representations .
as shall become clear immediately , the fradkin fi representation for @xmath1 $ ] is also not more complicated than gaussian : if a simple expansion in powers of @xmath28 is adopted , the @xmath21-dependence of every term in that expansion will not be more complicated than gaussian either . in brief
, the statement that the linkages can be calculated exactly means that the sum over all feynman graphs is reduced to the summation of a set of fi s @xcite .
after such linkages have been performed , el appears , which has the extraordinary and simplifying effect of converting the halpern fi into a set of ordinary integrals .
if the most convenient cluster expansion is used for the linkage operation upon @xmath29)$ ] , the previous expansion in powers of @xmath1 $ ] is brought into a much simpler form , and in the end , one need calculate a handful of overlapping proper - time integrals to achieve a reasonable estimate of the desired physics . in this paper such detail is not necessary , for to illustrate the distinction between `` ideal '' and `` realistic '' qcd , one can simply suppress all @xmath1 $ ] dependence , using , in effect , a quenched approximation .
in the subsequent papers @xcite , explicit examples will be given for the construction of binding and scattering potentials , with and without quenching .
for completeness , we here state the fradkin representation for @xmath1 $ ] , @xmath30 \\ \nonumber & = & - \frac{1}{2 } \int_{0}^{\infty}{\frac{\mathrm{d}s}{s } \ , e^{-is m^{2 } } } \ , e^{- \frac{1}{2 } { { \mathrm{tr}\ , } } { \ln{(2h ) } } } \\ \nonumber & & \quad \times \int{\mathrm{d}^{4}x } \ , \mathcal{n}^{\prime } \ , \int{\mathrm{d}[u ] } \ , \delta^{(4)}(u(s ) ) \ , e^ { \frac{i}{2 } \int { u \cdot ( 2h)^{-1 } \cdot u } } \\ \nonumber & & \quad \times \left [ { { \mathrm{tr}\ , } } { \left ( e^ { -ig \int_{0}^{s}{\mathrm{d}s ' \ , \left [ u'_{\mu}(s ' ) \ , a_{\mu}^{a}(w ) + i \sigma_{\mu \nu } \ , \mathbf{f}_{\mu \nu}^{a}(w ) \right ] \ , \tau^{a } } } \right)_{+ } } - 1 \right],\end{aligned}\ ] ] which in contrast to ( [ eq:4 ] ) contains a trace over all dirac and color variables , as well as a four - dimensional @xmath31 over this closed loop .
it should also be mentioned that @xmath1 $ ] is rigorously gauge - invariant under the full gauge group of qcd ; an original and simple proof may be found in ref . @xcite . with the two simplifying assumptions noted above , we write the @xmath21-dependence of the exponential of the pair of green s functions of ( [ eq:3 ] ) , with one @xmath32 $ ] representing q and the other @xmath2 , @xmath33 where the two lines of ( [ eq:6 ] ) refer to distinct contributions from each @xmath32 $ ] .
in addition , there is the @xmath21-dependence from the halpern fi , @xmath34 the linkage operation to be performed is then @xmath35 where @xmath36 .
the operation of ( [ eq:7 ] ) is equivalent to a normalized gaussian fi , and yields @xmath37^{-1 } \cdot ( \tilde{\mathcal{q } } + 2 \partial \chi ) } } , \end{aligned}\ ] ] using an obvious notation . and here is where a remarkable mechanism becomes apparent , since @xmath38^{-1 } & = & \left [ 1 - \left(\mathbf{d}_{c}^{-1 } + g f \cdot \chi
\right ) \cdot \mathbf{d}_{c } \right]^{-1 } \\ \nonumber & = & \left [ - g ( f \cdot \chi ) \cdot \mathbf{d}_{c } \right]^{-1 } \\ \nonumber & = & \mathbf{d}_{c}^{-1 } \cdot \left
[ - g ( f \cdot \chi ) \right]^{-1},\end{aligned}\ ] ] so that @xmath39^{-1 } = - \frac{1}{g } ( f \cdot \chi)^{-1 } $ ] , and the gauge - dependent propagator has disappeared from the interaction , remaining only in the factor @xmath40^{-\frac{1}{2 } } \cdot \left [ { { \mathrm{det } } } { \mathbf{d}_{c } } \right]^{-\frac{1}{2}},\end{aligned}\ ] ] so that the @xmath41^{-\frac{1}{2}}$ ] can be safely absorbed into an overall normalization .
we refer the reader to ref .
@xcite for details of exact and approximate evaluations of @xmath42 .
the important point we here wish to stress comes from the property of locality @xmath43 we shall call the locality apparent in ( [ eq:9 ] ) `` effective locality '' ( el ) , for it represents a contrast to the way in which ordinary propagators convey information , from one space - time point to another , from @xmath44 to @xmath45 . here , ( [ eq:9 ] ) is the object which replaces the conventional ( perturbative ) propagator @xmath46 . in ref .
@xcite , this locality is shown to hold for a large family of qcd processes ; it simplifies all calculations tremendously , reducing halpern s fi to a finite set of ordinary integrals , which can be studied numerically .
we illustrate the effects of el in this paper by making two simplifications , in the interest of clarity : 1 . neglect the @xmath47 term compared to @xmath48 , which approximation , in view of ( [ eq:6 ] ) and ( [ eq : chif ] )
, one would make in a strong - coupling limit .
this is just a convenient step for clarity and simplicity of presentation .
most importantly , conclusions are completely independent of this simplification .
the product of the two @xmath48 terms of ( [ eq:8 ] ) contains the self - interactions of each particle , as well as the cross - terms , corresponding to the interactions between the particles .
again , for reasons of simplicity and clarity , we retain only the cross - terms , writing the relevant exponential factor of ( [ eq:8 ] ) as @xmath49 + where @xmath50 , @xmath51 , and @xmath52 are variables associated with @xmath53 , whereas @xmath54 , @xmath55 , and @xmath56 are with @xmath57 . the last line of ( [ eq:10 ] )
may be written as @xmath58 and one sees that , as a consequence of el , this interaction is associated only with space - time point @xmath44 .
this means that all the other @xmath59 are irrelevant to the interaction , and as a most important consequence , they are removed , along with their normalization factors , leaving dependence only upon the ordinary integral @xmath60 .
further details are given in appendix [ appa ] .
the point central to the argument of this article is now being reached , as one seeks to evaluate the support of the second delta - function in ( [ eq:11 ] ) , which it is convenient to display as a product of delta - functions , in time , longitudinal and transverse coordinates , @xmath61 in the cm of q and @xmath2 , one can choose the origin of each time coordinate as the time of their closest approach ; the time difference @xmath62 is then always zero . if the q and @xmath2 are scattering , then @xmath63 , since their longitudinal projections are in opposite directions ; alternatively , if the q and @xmath2 are bound together , then @xmath64 , and their difference vanishes .
either choice makes no difference at all to the following analysis , and the simplest , second possibility can be adopted , that is , @xmath65 .
consider the time - coordinate delta - function , @xmath66 , which can have a zero argument whenever @xmath67 and @xmath68 coincide .
assume this happens at a set of points @xmath69 , so that @xmath70 in a similar way , the longitudinal delta - function may be evaluated as @xmath71 and the product of ( [ eq:13 ] ) and ( [ eq:14 ] ) as , @xmath72 under the restrictions @xmath73 and @xmath74 .
now , @xmath75 and @xmath76 , and @xmath77 and @xmath78 are continuous but otherwise completely arbitrary functions : the probability that the intersections of @xmath68 with @xmath67 and of @xmath79 with @xmath80 occur at exactly the same points is therefore arbitrarily small .
the only place where all four of these continuous functions have the same value is at @xmath81 , where , by definition of these functions , @xmath82 , and hence ( [ eq:15 ] ) reduces to @xmath83 where , it may be noted , corresponding @xmath84 and @xmath85 appearing in ( [ eq:10 ] ) combine with those of ( [ eq:16 ] ) to generate the signs , @xmath86 and @xmath87 , which may be specified in comparison with those of an eikonal model . for being crucial to the main argument of this article
, it matters to support the heuristic derivation given above with a rigorous mathematical proof . since the wiener functional space @xcite is the most adopted realization of a functional space , the proof will be given in this framework . here
, this space is made out of proper - time parameterized paths : for example , starting from @xmath88 at @xmath89 , one will consider paths passing through @xmath79 at proper - time @xmath900,s]$ ] ; and likewise , paths starting from @xmath91 at @xmath92 and passing through @xmath80 at proper - time @xmath930 , \bar{s}]$ ] . following the notation convention of ref .
@xcite , @xmath94 is the space @xmath950 , s],{\mathbb{r}})$ ] of continuous functions @xmath96 from the interval @xmath970 , { s}]$ ] into @xmath98 , satisfying @xmath99 ; and similar for `` barred '' quantities .
the @xmath100-functions @xmath96 and @xmath101 are here understood to represent either of the 2 possibilities @xmath75 and @xmath77 .
endowed with the wiener measure @xmath20 , @xmath94 is the measurable space , and with the measure @xmath102 , so is the product space @xmath103 endowed with the topology product , because @xmath104 and @xmath105 are independent .
then , we have the following , _ theorem : _ for all pair @xmath1060,{s } ] \ , \times \ , ] 0,\bar{s}]$ ] , one has @xmath107 whereas @xmath108^{2}=1\ ] ] the proof is as follows .
let @xmath109 be the set @xmath110 .
one has @xmath111 , where @xmath112 because of the obvious inclusion , @xmath113 one can write @xmath114 next , one has @xcite @xmath115 now , since @xmath116 under the integral and get zero , which establishes ( [ eq:16a ] ) , whereas ( [ eq:16b ] ) comes from the normalization procedure of the measure space @xmath117 into a probability space ; the theorem is then demonstrated .
the product of the first two delta distributions of ( [ eq:12 ] ) is therefore proportional to @xmath118 , and further dimensional and symmetry arguments determine the overall multiplicative constant such as in ( [ eq:16 ] ) .
one can observe that this determination of ( [ eq:12 ] ) fits the eikonal approximated evaluation of ref .
@xcite .
it is the remaining , transverse delta - function of ( [ eq:12 ] ) which is now most relevant , the term @xmath119 , where @xmath120 denotes the impact parameter , or transverse distance between the two particles .
this @xmath121 is sitting in the exponential of ( [ eq:8 ] ) , and the question immediately arises as to what meaning it can be assigned .
depending on its argument , a delta function is either zero or infinite : in the first case this means that there is no interaction , while the second case means that at @xmath122 , one has an infinite phase factor , suggestive of hard disc scattering @xcite .
the relevant question is therefore why such a delta - function @xmath121 appears at all .
the answer is that the assumption has earlier been made , and in the conventional , abelian way , that the @xmath123 and @xmath124 may be treated as ordinary particles , whereas it is by now well - known that asymptotic @xmath123 and @xmath124 exist only in bound states , and that their transverse coordinates can not be specified with arbitrary precision .
it is therefore unreasonable and unphysical to retain the conventional abelian practice in which such measurement is assumed possible .
this is the interpretation that will be posited here , taking the @xmath121 outcome as a serious warning that some odd working hypothesis has popped out in exactly this way .
one may wonder why this happens in qcd . because qcd possesses el , which conventional abelian theories do not .
the latter display sums over interconnected propagators , which provide a certain vagueness of position , whereas in the exact non - perturbative qcd , as described above and in ref .
@xcite , one finds the sharp determination of delta - functions corresponding to the el property , and transverse imprecision must therefore be introduced separately . in ref .
@xcite , it was suggested that this difficulty be treated in an _ ad hoc _ phenomenological way , by replacing @xmath121 by the smoothly varying , effective gaussian @xmath125 where @xmath126 is a mass parameter on the order of the q-@xmath2 bound state ( which we shall call a `` model pion '' ) , although we were able to draw the conclusion of that paper without specifying the precise form of @xmath127 . in this article , we face this question directly , by first developing a formalism in which transverse quark coordinates can not be specified , and then showing how this formalism removes all such absurdities , such as that of the exponential factor of @xmath121 above .
but it must be emphasized that our prescription for such a `` realistic '' qcd is phenomenological , for there remains to be shown how such an approach could be derived from a more fundamental , operator - field version of qcd , in which transverse imprecision would occur automatically , perhaps in relation to a possible non - commutative geometrical phase of non - perturbative qcd @xcite . in a following article @xcite , it will be shown how a specific choice of @xmath127 , slightly but importantly distinguishable from gaussian , can serve to define physically reasonable quark binding , and in a subsequent paper @xcite , these techniques will be extended to the construction of nucleon - nucleon scattering and binding potentials .
perhaps the simplest way of introducing transverse imprecision is to average that part of the qcd lagrangian dealing with the quark - gluon interaction , so that the transverse position of the color - charge current operator @xmath128 should be averaged over a small range by means of an initially unspecified distribution .
one can also demand the same imprecision for the vector current @xmath129 and scalar density @xmath130 , but these extra requirements seem to complicate the presentation , to no real advantage , and will not be considered here . instead of the conventional quark - gluon contribution to the lagrangian density , @xmath131 \ , \psi,\end{aligned}\
] ] in which all field operators occur at the same space - time point , and for which gauge invariance under the standard qcd gauge transformations is obvious , we now adopt a local in time and longitudinal position
but non - local in its transverse coordinates replacement , @xmath132 \ , \psi(x'),\end{aligned}\ ] ] where the transverse imprecision function ( tif ) @xmath133 is a real , symmetric function of its arguments , of significant value only for distances on the order of the inverse of the pion mass , @xmath134 , @xmath135 , and @xmath6 is left untouched . in this formulation
, rigorous local gauge - invariance is suppressed for the underlying quark fields , whose quanta have unmeasurable transverse positions , but the hadrons all constructed from these quanta will nevertheless be proper singlets under su(3 ) .
one notes that in the contribution of ( [ eq:18 ] ) to its part of the action operator , @xmath136 , the @xmath137 and @xmath138 coordinates can be interchanged , which yields an equivalent form in which every @xmath6 of the original ( [ eq:17 ] ) is replaced by @xmath139 .
this interchange allows a very simple extraction of all such transverse imprecision , since both delta - functions of ( [ eq:10 ] ) will now be replaced by @xmath140 this can be seen in the simplest way by noting that the presence of a tif @xmath141 shall mean that the difference @xmath142 is effectively bounded by @xmath143 , where @xmath126 appears in the definition of @xmath144 below , and where , as shown in ref .
@xcite , @xmath126 is on the order of the pion mass , @xmath145 .
hence a negligible error is made by replacing the transverse coordinate @xmath146 by @xmath147 .
a more precise justification appears in appendix [ appb ] .
it should be noted that this approximation concerning the utility of el can be justified in the same way as described in appendix [ appb ] for both the construction of the quark binding potential and the nucleon - nucleon scattering and binding potential ( the exchange of a gluon bundle across a closed quark loop apparently involves a new set of @xmath148 integrals , and will be discussed elsewhere ) . the first delta - function of ( [ eq:19 ] ) defines the argument @xmath44 of @xmath149 , and we again observe that the final output of the halpern fi will be an ordinary integral @xmath148 , independent of the choice of @xmath44 .
the second delta - function of ( [ eq:19 ] ) now involves the @xmath150-dependence , generating in place of the @xmath151 which follows from ( [ eq:16 ] ) , the combination @xmath152 inserting 2-dimensional fourier transforms of each @xmath153 the combination ( [ mu ] ) becomes @xmath154 from its definition , @xmath150 is real , and hence ( [ eq:20 ] ) becomes @xmath155 which provides the definition of @xmath127 . note that while no restriction has been placed on the form of @xmath150 other than that it is real , from ( [ eq:21 ] ) , @xmath156 turns out to be independent of the direction of @xmath120 , that is , @xmath157 .
one might expect that an intuitive choice such as @xmath158 would suffice , but that is not true . by trial and error
it is found that a slight change to @xmath159 with @xmath160 real , positive , and small , @xmath161 , is most appropriate and leads to a binding potential of form @xmath162 , for small @xmath160 @xcite . for the time being , the interpretation of the parameter @xmath160 is quite enigmatic .
there are steps in the derived expression for @xmath163 where @xmath160 may be neglected compared to unity ; but @xmath160 can not be set equal to zero .
and @xmath160 is not a coupling constant , for even when small , its effect on the q-@xmath2 binding is large .
it would seem that , in some sense , @xmath160 has something to do with the transverse correlations between bound quarks and/or antiquarks . in this respect
it is worth noticing that ( [ levy ] ) is a part of a levy - flight probability distribution @xcite . at this point ,
a part of the enigma is precisely how such correlations would emerge from the assumed transverse imprecision of the quark field operators .
one other point deserves mention concerning the scale change used when @xmath164 in the exponent of the halpern representation is broken up into small cells of volume @xmath165 : @xmath166 upon rescaling , @xmath167 , and re - expressing all interactions in terms of @xmath168 , there appears in ( [ eq:10 ] ) the factor @xmath169 . in ref .
@xcite , where transverse imprecision was treated in an _ ad hoc _ way , the size of @xmath170 was taken to be @xmath171 , where @xmath172 corresponded to a very large energy associated with the eikonal limit . here
, we ask the more physical question of just how small that @xmath170 may be chosen in the light of an actual measurement , and we let quantum physics provide the answer : that contribution to the @xmath170 corresponding to a time separation should be chosen as @xmath173 , that corresponding to a ( cm ) longitudinal coordinate should be @xmath174 , while that corresponding to each of the transverse coordinates should be @xmath175 ; and hence @xmath176 .
an alternate way of expressing this is that , starting from arbitrarily small separations in each coordinate , we average each @xmath177 variable over a physically meaningful distance , and call that average the @xmath177 contained in the volume @xmath178 . finally , we note that @xmath144 replaces the @xmath179 of the conventional , ` ideal ' formulation of qcd , with obvious normalization : @xmath180 .
if we adopt the same normalization for @xmath181}$ ] , then it is easy to see that , with @xmath182 , @xmath183 becomes @xmath184
in the above example of quark and/or antiquark scattering , where the infinite number of exchanged gluons appears to originate and end at a single space - time point on a quark / antiquark line , it may be helpful to introduce the concept of an exchanged ` gluon bundle ' , as in figure [ fig:1 ] . because of the four - dimensional delta function @xmath185 , arising from the product of the pair of delta functions of ( [ eq:19 ] ) , and of the subsequent analysis which produces ( [ eq:20 ] ) , the transverse separation @xmath186 satisfies the distribution of ( [ eq:21 ] ) .
but the argument @xmath187 in @xmath188 is given by @xmath189 in virtue of ( [ eq:16 ] ) , and hence the halpern functional integral reduces to a set of ordinary integrals , as explained in section [ sec : fradkinrepresentations ] , and represented by the bundle diagram of figure [ fig:1 ] .
a gluon bundle exchanged between quarks @xmath190 and @xmath191,height=132 ] here , the time and longitudinal coordinates of the end - points of the bundle are the same , although their transverse coordinates measured vertically in the figure are separated .
bundle diagrams are not feynman diagrams , but are perhaps a more efficient way of representing the sum over all of the feynman graphs corresponding to such multiple gluon exchange .
a slightly more complicated expression describes gluon bundles exchanged between any two of three quarks , as in figure [ fig:2 ] , where , because of el , the @xmath187-coordinates of each of the @xmath192 entering into the appropriate halpern functional integral are the same , even though the transverse coordinates of the three quarks can be quite different .
gluon bundles exchanged among three quarks.,height=132 ] in contrast , were a closed quark loop corresponding to a simple relaxation of the quenched approximation to appear between a pair of quarks , joined to each external quark lines by the exchange of a gluon bundle , as in figure [ fig:3 ] , there will now be two distinct sets of ordinary halpern integrals to be evaluated . as will be seen elsewhere
, the effective diagram of figure [ fig:3 ] will provide us with the essential features of the nucleon - nucleon potential for separation lengths beyond 2 fm @xcite .
gluon bundles joining two quarks via a closed quark loop.,height=132 ]
for a whole family of qcd processes , a remarkable property of effective locality was first observed on an approximated ( quenched and eikonal ) version of qcd @xcite .
however , that property still holds without these approximations , so that effective locality appears as a genuine property of non - perturbative qcd @xcite . in the present article , a thorough analysis of the constraints inherent to effective locality has displayed an odd @xmath121-term in the argument of an exponential factor describing @xmath193 scattering at impact parameter @xmath120 .
this fact is here taken as an indication that an untenable hypothesis is lurking in the derivation leading to this @xmath121-function , an assumption most certainly inherited from our long practice of perturbative qcd .
accordingly , a replacement of the meaningless @xmath121 in an exponential factor of an exact statement obtained in an ideal qcd amplitude , by a phenomenologically defined @xmath127 is proposed .
it can be performed for each and every qcd process and will be used for the derivation of quark binding in a following article @xcite . as emphasized in the text ,
the proposed transition from ideal to realistic qcd is phenomenological , but it clearly reflects the necessary change in viewpoint , taking into account the fact that quark quanta are always bound .
it is only when a complete sum over all relevant feynman graphs has been accomplished in a gauge - invariant way , that the new and exact property of effective locality appears . that property ,
if taken seriously , then forces this change to a would - be more realistic form of non - perturbative qcd where transverse imprecision would be built - in . clearly , work remains to be done in order to appreciate as deeply as possible the necessity of such a change as specified by ( [ levy ] ) .
this point will be dealt with , in detail , in ref .
@xcite . at face value , though , it is certainly fascinating that starting from a notion of propagation , such as implemented through the familiar 2-point function @xmath194 , with its obvious perturbative content , one could be lead , in the bound non - perturbative context , to another mode of behavior , more appropriately described in terms of levy flight distributions !
one may then ask what happens when the variables @xmath196 change , such that space - time point @xmath44 changes to a different point , say @xmath197 ?
the answer is that nothing concerning the output of the integral @xmath198}$ ] is changed , for all the other space - time points @xmath199 are then removed , with their normalizations , and the remaining integration of @xmath200 has exactly the same form as that of @xmath60 . in brief , for this scattering process , there is but a single , relevant @xmath201 whose origin is irrelevant . one may also ask if a significant jacobian of the transformation between @xmath149 and @xmath202 could enter the integrand . holding the external position
coordinates constant , and treating the @xmath203 as continuous functions , could there be a significant jacobian under a change of value of @xmath96 ?
the answer is negative , and for two reasons : 1 .
inspection of the fradkin representation for @xmath32 $ ] shows that these @xmath50 are bounded , @xmath204 , where @xmath20 denotes the quark mass , while @xmath205 is a macroscopic quantity , so that the difference of @xmath206 and @xmath96 is effectively just @xmath206 .
thanks to ( [ eq:16 ] ) , @xmath207 is fixed at 0 and @xmath208 ; any other possible value of @xmath207 is suppressed by the el property .
for these two reasons , we infer that any such jacobian is essentially unity , and that the final integral is independent of the value of the argument @xmath187 of @xmath203 .
before transverse imprecision was introduced , el had the effect of attaching to the representative symbol @xmath210^{-1}$ ] of each gluon bundle
exchanged between quark and/or antiquark of respective cm coordinates @xmath206 and @xmath91 a pair of delta functions , @xmath211 , as used in the text , or the pair @xmath212 as used in an eikonal approximation @xcite . for either case one
finds fixed values of @xmath213 and @xmath214 , and @xmath215 .
then , as claimed in the text , the halpern fi can be reduced to an ordinary set of @xmath148 integrals . in the process , though , one makes a systematic error , of the eikonal - type , by neglecting variations of the impact parameter or , correspondingly , of momentum transfer in the core parts of the matrix element . in the context of the more exact expression of the first pair of delta functions above , that _ ad hoc _ approximation avoided the much more complicated analysis of the transverse fradkin s difference @xmath216 . with transverse imprecision
now being included , the situation changes for the better , in the sense that no such approximation need be made . but this change now requires a slightly more complicated justification of the argument which replaces halpern s fi by a set of ordinary integrals . for
the question arises if this useful simplification is also true when the @xmath217 inside the @xmath210^{-1}$ ] factor is itself given by @xmath218 , and is being integrated over the @xmath219 in that exponential factor , as in the discussion of the text leading to ( [ eq:20 ] ) .
it was there noted that the replacement of that @xmath217 by @xmath220 or @xmath221 is a reasonable approximation .
consider the halpern fi @xmath222 \ , [ { { \mathrm{det } } } { ( f \cdot \chi)}]^{-\frac{1}{2 } } \ , e^{\frac{i}{4}\int{\mathrm{d}^{4}w \ , \chi^{2 } } + ig\int{\mathrm{d}^{2}\vec{y}^{\ , \prime}_{\perp } \ , \mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp } ) [ f \cdot \chi(\vec{y}^{\ , \prime}_{\perp})]^{-1 } } } } \end{aligned}\ ] ] where @xmath223 , the normalization is defined so that the fi of ( [ eq : b1 ] ) equals 1 when @xmath224 .
the dependence of color , time and longitudinal coordinate has been omitted for simplification of presentation . as in the definition of this or any such fi
, @xmath225 is understood to mean @xmath226 , where the subscript @xmath227 denotes the value of @xmath228 at the space - time point @xmath229 , and @xmath230 corresponds to a small volume surrounding that point , which is to become arbitrarily small as @xmath231 becomes arbitrarily large @xcite . as mentioned in the text by the end of section [ sec : realisticqcd ] , residual @xmath170-dependence will be re - expressed in terms of physically significant quantities as a last step ; but for the following argument , all the transverse coordinate differences are to be taken as arbitrarily small .
now , re - scale the @xmath232 variables such that @xmath233 , and re - write ( [ eq : b1 ] ) as @xmath234 \ , [ { { \mathrm{det } } } { ( f \cdot \bar{\chi})}]^{-\frac{1}{2 } } \ , e^{\frac{i}{4 } \ , \sum_{\ell}{\bar{\chi}_{\ell}^{2 } } + ig \delta^{2 } \ , \int{\mathrm{d}^{2}\vec{y}^{\ , \prime}_{\perp } \ , \mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp } ) [ f \cdot \bar{\chi}(\vec{y}^{\ , \prime}_{\perp})]^{-1 } } } } \end{aligned}\ ] ] let us also break up the @xmath235 integral into an infinite series of terms : one is free to choose the individual @xmath218 coordinates as exactly those which define the transverse positions of the @xmath236 . in this way , ( [ eq : b2 ] ) may be re - written as @xmath237 \ , [ { { \mathrm{det } } } { ( f \cdot \bar{\chi})}]^{-\frac{1}{2 } } \ , e^{\frac{i}{4 } \ , \sum_{\ell}{\bar{\chi}_{\ell}^{2 } } + ig \delta^{2 } \ , \delta^{2}_{y_{\perp } } \ , \sum_{\ell}{\mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp\ell } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp \ell } ) [ f \cdot \bar{\chi}(\vec{y}^{\ , \prime}_{\perp\ell})]^{-1 } } } } \end{aligned}\ ] ] where @xmath238 is understood as a true infinitesimal quantity , and where , for simplicity , we suppress explicit dependence on @xmath239 and @xmath240 . but now ( [ eq : b3 ] ) may be written as the product of @xmath231 integrals , @xmath241^{-\frac{1}{2 } } \ , e^{\frac{i}{4 } \ , { \bar{\chi}^{2}(\vec{y}^{\ , \prime}_{\perp \ell } ) } + ig \delta^{2 } \ , \delta^{2}_{y_{\perp } } \ , { \mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp\ell } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp \ell } ) [ f \cdot \bar{\chi}(\vec{y}^{\ , \prime}_{\perp\ell})]^{-1 } } } } \\ \nonumber & \equiv & \prod_{\ell}^n{\mathbb{f}(ig \delta^{2 } \ , \delta^{2}_{y_{\perp } } \ , \mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp\ell } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp \ell } ) ) } \\ \nonumber & \equiv & \prod_{\ell}^n{\mathbb{f}(z_{\ell})},\end{aligned}\ ] ] where ( [ eq : b4 ] ) denotes the normalized product of all such @xmath242-valued integrals , and @xmath243 denotes the ordinary integral @xmath244 over the variable associated with @xmath245 .
that integral is well defined in the sense that , for @xmath246 , as is the case here , it can be expressed as an absolutely - convergent series , or as a converging integral over a set of eigenvalues in a random matrix calculation . where the normalization condition of ( [ eq : b4 ] ) stipulates that @xmath249 .
since @xmath250 is proportional to the infinitesimal @xmath238 , one may expand in powers of @xmath250 , @xmath251 } = 1 + i \int{\mathrm{d}\varrho \ , \varrho \ , \tilde{\mathbb{f}}(\varrho ) } \ , z_{\ell } + \cdots,\ ] ] so that ( [ eq : b4 ] ) becomes @xmath252 } \\ \nonumber & = & 1 + i \int{\mathrm{d}\varrho \ , \varrho \ , \tilde{\mathbb{f}}(\varrho ) } \ , \sum_{\ell}{z_{\ell } } \\ \nonumber & = & 1 + i \int{\mathrm{d}\varrho \ , \varrho \ , \tilde{\mathbb{f}}(\varrho ) } \cdot ig \delta^{2 } \
, \int{\mathrm{d}^{2}\vec{y}^{\ , \prime}_{\perp } \ , \mathfrak{a}(\vec{y}_{1\perp } - \vec{y}^{\ , \prime}_{\perp } ) \ , \mathfrak{a}(\vec{y}_{2\perp } - \vec{y}^{\ , \prime}_{\perp})}.\end{aligned}\ ] ] which is just the first - order expansion of the result obtained in the text when @xmath218 was shifted to @xmath255 or @xmath256 . and
since @xmath257 is expected to be small , @xmath258 , it is , in effect , equivalent to @xmath259 which is just the integral of ( [ eq : b2 ] ) when the intuitively equivalent change @xmath260 has been made in the argument of @xmath261 , and after the residual @xmath262 dependence has been continued to the measurably - significant value of @xmath263 .
|
a basic distinction , long overlooked , between the conventional , `` idealistic '' formulation of qcd , and a more `` realistic '' formulation is brought into focus by a rigorous , non - perturbative , gauge - invariant evaluation of the schwinger solution for the qcd generating functional in terms of exact fradkin representations for the green s functional @xmath0 and the vacuum functional @xmath1 $ ] . the quanta of all ( abelian ) quantized fields may be expected to obey standard quantum - mechanical measurement properties , perfect position dependence at the cost of unknown momenta , and vice - versa , but this is impossible for quarks since they always appear asymptotically in bound states , and their transverse position or momenta can never , in principle , be exactly measured .
violation of this principle produces an absurdity in the exact evaluation of each and every qcd amplitude .
we here suggest a phenomenological change in the basic qcd lagrangian , such that a limitation of transverse precision is automatically contained in a proposed ` realistic ' theory , with the function essential to quark binding into hadrons appearing in the new lagrangian .
all absurdities in estimates of all `` realistic '' qcd amplitudes are then removed , and one obtains the possibility of hadron formation by appropriate quark binding potentials , and nucleon scattering and binding by effective , yukawa - type potentials ; the first of these potentials is constructed , in detail , in a following article .
|
if supersymmetry exists at an energy scale of 1 tev , the study of third generation sleptons and squarks at the lhc is of a special interest .
their masses can be very different from that of sparticles of the first and second generation , because of the effects of large yukawa and soft couplings as can be seen from the renormalization group equations .
furthermore they can show large mixing in pairs @xmath4 and @xmath5 . a detailed discussion of possible susy effects at the lhc is given in @xcite .
in this paper we consider the mass reconstruction of the lightest stop quark ( @xmath2 ) in the cascade decay @xmath6 with the top quarks decaying into hadrons . the gluino decay chain ( [ eq - chain ] ) is represented in fig . [ fig - chain ] , in which all final state particles are explicitly shown .
here , the considered leptons are electrons and muons ( @xmath7 e , ) .
the lightest neutralino @xmath8 is invisible to the particle detector , whereas b quarks and light quarks ( labeled as ` q ' in fig . [ fig - chain ] ) are observed as jets .
approaches to stop quark mass reconstruction in different decay chains and for different points in the mssm parameter space are discussed in literature ( for example in @xcite ) .
recent limits from searches for stop quarks were published in @xcite .
the reconstruction of a susy event is complicated because of the escaping neutralinos and of the many complex and competing decay modes . at present , there are two different approaches to susy mass reconstruction .
the _ endpoint method _ , which has been widely studied @xcite , looks for kinematic endpoints of invariant mass distributions .
the second method is the _ mass relation approach _
@xcite , based on the `` mass relation equation '' which relates the masses of the susy particles and the measured momenta of the detected particles .
it was shown in @xcite that the mass relation approach can be successfully used for integrated luminosities as low as a few @xmath1 . in this work ,
the mass relation approach of @xcite is used for measuring the mass of the lightest stop quark @xmath9 , assuming about 20 @xmath1 of integrated luminosity in lhc proton - proton collisions at @xmath10 tev , under the assumption that the masses of the gluino , slepton and of the two lightest neutralinos have been reconstructed in advance with 1020% uncertainty . at such low integrated luminosity ,
the stop mass reconstruction is quite challenging because of a high level of susy background ( sparticles created in decay chains different from that of fig .
[ fig - chain ] ) and standard model @xmath11 background . in this paper , a particular example is chosen ( the su3 model ; section [ sec - simul ] ) to illustrate the method .
first , an `` event filter '' is applied to suppress the background ( section [ sec - bkg - supp ] ) , making use of a likelihood function built with the known uncertainties on jet and lepton measurements , constrained by the mass relation equation .
next , events which do not satisfy the kinematic limits derived from the chain ( [ eq - chain ] ) are discarded ( section [ sec - limits ] ) .
finally , a combinatorial mass reconstruction method ( section [ sec - stop - mass ] ) is applied to find the best estimate of the stop mass from the maximization of a combined likelihood function , which depends on all five sparticle masses ( gluino , stop , slepton and the two lightest neutralinos ) and is constructed for each possible permutation of the final state particles .
the particular msugra model chosen for this work is the bulk point su3 , one of the official benchmark points of the atlas collaboration @xcite , which is compatible with the recent precision wmap data @xcite .
the values of the relevant msugra parameters are given in table [ tab - paramth ] .
for the susy particles in the cascade process ( [ eq - chain ] ) , the theoretical masses and the production cross section have been found by isajet 7.74 @xcite as reported in table [ tab - massth ] .
.msugra parameters for the su3 point @xcite . [ cols="^,^,^,^,^,^ " , ] as shown in @xcite , the sm processes are suppressed significantly by the above requirements .
the sm dominant background surviving the hard cuts is @xmath11 production , where both w bosons decay leptonically producing a @xmath12 state .
since the @xmath11 production cross section is about 833 pb , a 17 m @xmath11 sample , corresponding to 20 fb@xmath13 of integrated luminosity , was generated with the herwig event generator .
after applying the above selection cuts , only 21 @xmath11 background events survive .
note that the requirement of high hadronic activity is important for the @xmath11 background suppression .
if the cut on the total number of jets @xmath14 is loosened to 7 jets , the number of @xmath11 events surviving the selection cuts increases to 115 . for every event , light jets ( i.e. not tagged as b - jets )
were combined in pairs whose invariant mass has been reconstructed . only the independent pairs
whose @xmath15 is in the range 60100 gev ( w boson region ) have been retained and events with less than two jet pairs have been rejected . the invariant mass distribution obtained with all combinations is shown in fig .
[ fig - mw ] .
next , all combinations of a jj pair and a b - jet have been considered , retaining only those with @xmath16145205 gev , representing the acceptable w - b pairs in the top quark region ( fig .
[ fig - mtop ] ) . after this step , from the initial 215 susy events ( signal + background ) ,
a total of 834 bjj combinations were identified as candidates for a top quark decay @xmath17 bw @xmath18 bjj , coming from 23 signal and 70 background events .
thus at this stage only one signal event was lost . on the other hand , applying the same procedure to the 21 @xmath11 background events left only 3 events with 21 bjj combinations as candidates for a top quark decay .
naively , one would think that this selection should let all @xmath11 background events survive , but it is clearly not the case .
if a w - boson decays leptonically an invariant mass of light jets is not related with the w - boson mass .
also a parton showering leads to redistribution of energy and this means that the momenta of the b - quark and the w - boson are not sufficient to give the top invariant mass .
the final result is that most @xmath11 events are dropped at this stage .
to suppress the background before the last ( combinatorial ) step , an event filter is used which assumes that the masses of @xmath19 , @xmath20 , @xmath21 are known ( table [ tab - massth ] ) .
the event filter maximizes , for each event , a likelihood function constrained by the mass relation equation , or equivalently minimizes the function : @xmath22 where the index @xmath23 runs over the two leptons and the two t - quarks , @xmath24 and @xmath25 are the reconstructed momentum and its uncertainty , and @xmath26 is the true momentum , which is varied to find the minimum ( only uncertainties in jet and lepton energy measurements are taken into account ) .
the parameter @xmath27 is a lagrange multiplier and @xmath28 is the mass relation .
the constraint @xmath28 is the key of the `` mass relation approach '' : it relates the masses of the susy particles to the measured momenta of the final state particles in the chain ( [ eq - chain ] ) @xcite .
the mass relation constraint is obtained as a solution of a system of four - momentum constraints for each vertex containing susy particles in the decay chain ( [ eq - chain ] ) .
for example , for the gluino decay vertex one has @xmath29 where the right hand part is in terms of the four - momenta of the lightest supersymmetric particle ( lsp ) , the detectable leptons , and the reconstructed top quarks .
similar relations can be obtained for each susy vertex in the process ( [ eq - chain ] ) , which contains four vertices , hence one gets four kinematic equations that can be solved to find the four components of @xmath30 in terms of the susy masses and the momenta of the detectable and reconstructed particles . by substituting these components into the on - shell mass condition for the lsp , @xmath31 , the mass relation constraint includes all susy masses ( @xmath32 ) and the momenta ( @xmath33 ) of the detectable leptons and of the reconstructed t - quarks in the process ( [ eq - chain ] ) .
the explicit form of @xmath34 is given by @xcite .
note that in the decay chain ( [ eq - chain ] ) the locations of each of the two t - quarks and of each of the two leptons are unknown . here
, it is assumed that the t - quark with higher energy originates from the gluino decay .
in addition , we assumed that the leptons with higher @xmath35 originate from the @xmath36 decay .
the momentum resolution for a t - quark is computed as @xmath37 , where all jets are assumed to have the resolution mentioned above ( equation ( [ eq - jet - res ] ) in section [ sec - simul ] ) . in the numerical minimization procedure by mc sampling , the gluino mass is left free to vary within @xmath38 from the value reported in table [ tab - massth ] ( a gaussian sampling ) , and the stop mass is left free to vary uniformly in the range 480@xmath39120 gev because the stop mass is unknown at this step .
the latter range has lower limit approximately related with the kinematic condition for a stop decay @xmath40 and is limited from above by the sbottom quark mass , which is about 600 gev for the su3 point . for signal events ,
the event likelihood distribution has a maximum in the region of the ( @xmath41 , @xmath9 ) mass plane correlated with the true masses of @xmath41 and @xmath9 .
hence , signal events should give a peak in the region of the true masses .
on the other hand , for background events there is no strong correlation between the likelihood maximum and the true masses of @xmath41 and @xmath9 .
therefore , if we arbitrarily chose a point in the ( @xmath41 , @xmath9 ) mass plane in the range close to true masses , there is a very high probability that the value of @xmath42 found from equation ( [ chi1 ] ) is smaller for a signal event than for a background one . for each event ,
@xmath43 points in the mass plane were generated in the range @xmath44 gev for the gluino and @xmath45 gev for the stop quark , and the @xmath42 was calculated .
results of event filter are presented in fig .
[ fig - evfilter ] .
this figure shows the number of events versus the number of trials per event in which @xmath46 .
two groups of events can be seen : a group with the number of accepted trials close to zero and a group with more than a few hundred of accepted trials .
the first group presents background events for which unlike for signal events a minimum of @xmath42 need not be in the considered mass range .
thus if @xmath46 in at least 300 trials per event , this event was considered as a signal candidate and was retained for the subsequent analysis . .
the total number of trials is @xmath43 . ] before the application of the event filter , we had 23 signal events , 70 susy background events and 3 standard model @xmath11 background events .
after the application of the event filter , the susy background events were reduced approximately by a factor of 2 while a single signal event was lost : 22 signal events , 37 susy background events and 3 @xmath11 background events survived .
it was shown @xcite that the endpoint method could be very useful in susy particle mass reconstruction for finding relations between the masses of the susy particles involved in a decay chain and to determine their masses .
such a method allows mass reconstruction without relying on a specific susy model .
in particular , the endpoint method can be applied to a decay chain of the type @xmath47 where particles @xmath48 , @xmath49 , @xmath50 are invisible but particles @xmath51 and @xmath52 are considered as visible ( they can be either directly detected or indirectly reconstructed from the properties of final state particles ) . in the literature , kinematic limits on the invariant mass distribution of @xmath53 pairs in decay ( [ eq2 ] ) over a variable @xmath54 often are given for the case in which at least one of the visible particles is massless . however , in some cases both particles @xmath51 and @xmath52 can have a non - negligible mass .
for example , this is the case when a gluino decays into a stop quark and top quark .
these kinematic limits for the case of all massive particles in process ( [ eq2 ] ) are given @xcite by @xmath55 with @xmath56 \end{array}\end{aligned}\ ] ] where the upper edge corresponds to the case when @xmath51 and @xmath52 particles are moving in opposite directions in the rest frame of particle @xmath48 and the lower edge corresponds to the case when @xmath51 and @xmath52 particles are moving in the same direction .
a nonzero lower limit is a consequence of nonzero masses of the particles .
the kinematic limits from equation ( [ limits ] ) for the process ( [ eq - chain ] ) are @xmath57 = 375.1 gev and @xmath58 = 496.8 gev for particles created on - shell .
because susy particles can also be created off mass shell , we have set a wider range . by choosing a cut at 525 gev in @xmath11 invariant mass , the number of events remaining after the event filter selection is reduced to 21 signal events , 34 susy background events and 3 @xmath11 background events
if all sparticle masses but the stop mass in the decay chain ( [ eq - chain ] ) were known , the mass relation equation would allow finding the stop mass directly .
however , we assume that these masses are known at integrated luminosity of 20 @xmath1 with an uncertainty of 10 - 20@xmath59 . in order to take this into account , at the final step of the stop mass reconstruction we allow for all sparticle masses to vary in ranges defined by their uncertainties . because in this case for each event there are five unknown masses , at least five events are required to reconstruct sparticle masses . the combinatorial procedure of @xcite is used for the final stop mass reconstruction , applied only to the events that pass the event filter .
it is important to make the most effective cuts in advance , because such procedure is computationally very intensive due to combinatorics : all possible groupings of five events in the sample are considered . for each set of five events ,
a combined likelihood function is built and maximized . for sparticle masses
the combined likelihood function for the combination is defined as the product of the maximum likelihood functions for individual events . finding a maximum of the combined likelihood function for the combination is the same as searching for a minimum of the function @xmath60 in eq .
( [ chi20 ] ) , @xmath61 is a result of searching for a minimum of the @xmath62 function for an individual event . for each of the five events in the combination , the @xmath63
is fitted with 9 parameters ( four particle momenta and five susy masses ) .
the @xmath42 function for an individual event is defined by @xmath64 where the variables with superscript `` event '' are those with respect to which one has to minimize .
background with integrated luminosity of 20 @xmath65 .
the curve is the result of a gaussian fit . ]
the first term in equation ( [ chi2 ] ) takes into account the deviations of the reconstructed momenta of the t - quarks , or the measured momenta of the leptons , from the true ones .
the second term takes into account the intrinsic width of the susy particle masses , in the gaussian approximation ( instead of a breit - wigner distribution ) .
the standard deviations corresponding to the mass widths are taken to be 15 gev for the gluino , 2 gev for the stop and 1 gev for all light masses .
the first two numbers are comparable with the theoretical widths for the heavy susy particles .
the last number takes into account the fact that light susy particles are either quite narrow or stable .
we note that the results of the mass reconstruction are not strongly sensitive to the actual values of sparticle widths . in equation ( [ chi2 ] ) the mass relation and the @xmath66 edge , which relates three light sparticle masses for chain ( [ eq - chain ] ) , are included by means of the lagrange multipliers @xmath67 and @xmath68 . the @xmath66 edge ( 103.1 gev )
can be obtained , for example , from eq .
( [ limits ] ) by plugging in the correspondent light sparticle masses and zero lepton masses . the minimization is done numerically by means of mc samplings of the parameter space .
the sampling is uniform for the stop mass , in the range : 480 @xmath69 120 gev . for the masses of @xmath41 , @xmath70 , @xmath71 , @xmath8 a gaussian sampling
is done , with mean values as given in the table [ tab - massth ] and standard deviations of 72 , 20 , 20 , 20 gev , which approximately corresponds to uncertainties of @xmath72 in sparticle masses found in @xcite .
the minuit code @xcite is used to search for the minimum of the @xmath73 function ( [ chi20 ] ) . the reconstructed susy particle mass distribution , together with a gaussian fit , is shown in fig .
[ fig - stop ] .
as can be seen in this figure , the reconstructed stop mass distribution is described approximately by a gaussian with the mean value of 483 gev and standard deviation 36 gev .
the asymmetry of the stop mass distribution toward higher masses can be explained by the nearness of the kinematic edge for the stop decay @xmath74 to the stop mass so that high masses are generated more often than lower values .
this is not accounted for in the monte carlo sampling , for which we have uniformly generated the values of the stop mass .
in order to understand the role of background and of the simulated detector effects , the stop mass has been reconstructed using generator - level momenta without susy background and standard model @xmath11 background , and the result is shown in fig .
[ fig - idealstop ] . in this case , as expected , the reconstructed stop mass is very close to the theoretical mass and the width of the distribution is smaller because of the absence of detector effects .
it follows from comparison of fig .
[ fig - stop ] and fig .
[ fig - idealstop ] that the presence of background tends to shift the stop mass peak position to the region of higher masses . .
the curve is the result of a gaussian fit . ]
we applied the cascade mass reconstruction approach developed in @xcite for reconstructing the mass of the 3@xmath75 generation supersymmetric quark , assuming 14 tev proton - proton collisions at the lhc with integrated luminosity of about @xmath76 . at such relatively low integrated luminosity ,
the stop mass reconstruction is complicated because of a high level of susy and standard model @xmath11 backgrounds and of the low branching ratio for the gluino decay chain involving a stop quark .
our approach to the stop mass reconstruction is based on the consecutive use of an event filter and of a combinatorial mass reconstruction method . in this work
, we considered the stop mass reconstruction at the su3 msugra point and we obtained an estimate of the stop mass with a precision of about @xmath3 .
we expect that our approach should work for different mssm parameters as well , provided that a decay chain containing at least four successive two - body decays and involving five susy particles has a sufficiently large branching ratio to be identified in a heavy background environment .
the authors thank m. ibe , a. mincer and p. nemethy for interesting discussions and useful suggestions .
this work has been supported by the national science foundation under grant phy-0854724 .
h. baer , k. hagiwara and x. tata , phys .
* d35 * ( 1987 ) 1598 ; + h. baer , d.d .
karatas and x. tata , phys .
* d42 * ( 1990 ) 2259 ; + h. baer , c. kao and x.tata , phys .
* d48 * ( 1993 ) 5175 ; + h. baer , c. chen , f. paige and x. tata , phys .
* d50 * ( 1994 ) 4508 .
|
the cascade mass reconstruction approach was applied to simulated production of the lightest stop quark at the lhc in the cascade decay @xmath0 with top quarks decaying into hadrons .
the stop quark mass was reconstructed assuming that the masses of gluino , slepton and of the two lightest neutralinos were reconstructed in advance .
a data sample set for the su3 model point containing 400k susy events was generated which corresponded to an integrated luminosity of about 20 @xmath1 at 14 tev .
these events were passed through the acerdet detector simulator , which parametrized the response of a generic lhc detector .
the mass of the @xmath2 was reconstructed with a precision of about @xmath3 .
|
from computer games and medicine to weather predictions and new technologies , nearly every aspect of our lives is influenced by mathematical modeling .
a primary barrier to forward progress in teaching mathematical modeling across our partner disciplines is misconceptions and biases around what constitutes modeling .
we ( the co - authors ) are part of an interdisciplinary working group at the national institute for mathematical and biological synthesis ( nimbios ) that has brought together mathematicians , biologists , and education researchers to address teaching quantitative biology , especially modeling .
each of us has experienced reaching out to collaborate with a member of another discipline , only to have the conversation shut down before it has really started because of the naive assumptions we make about the other s discipline .
we posit that if mathematicians and biologists alike can improve their understanding of the similarities and differences in their approaches to and language around modeling , then each discipline will play a more effective role in advancing the other @xcite , and we will be able to teach this valuable skill more effectively . in this paper
we describe a framework for models and modeling that can bridge the communication gap between disciplinary boundaries , enabling mathematicians , statisticians , and biologists to come together to improve student learning . in 2012 ,
the president s council of advisors on science and technology ( pcast ) released the _ engage to excel _ report @xcite .
the report recommended that we should engage in a national experiment that encourages faculty from math - intensive fields other than math to be involved in teaching mathematics as a way to help close the mathematics achievement gap @xcite .
the subtext is that traditionally trained mathematics educators are failing at helping our students succeed at mathematics as applied to science and technology . soon after , the national research council ( nrc ) released a study , titled _ the mathematical sciences in 2025 _ , which suggested that in order for the mathematical sciences to remain strong in the united states , the education of students should be conducted in a cross - disciplinary manner that reflects these ever - changing realities @xcite .
this requires a rethinking of the curricula in the mathematical sciences , especially for mathematics and statistics departments , in order to provide the additional quantitative skills needed for students entering the workforce in the fastest growing career fields , such as those in stem .
mathematics professional societies responded to this call with the _ common vision _ project , which identified ways of improving undergraduate curricula and education in the mathematical sciences by bringing together leaders from five mathematical and statistical associations @xcite .
the project summarized the collective recommendations of seven other curricular guides on undergraduate education from the associations , as well as added its own suggestions for improving undergraduate education in the mathematical sciences , especially in the first two years .
the _ common vision _
project concluded that departments should increase efforts to update curricula , support evidence - based pedagogical methods , and establish connections with other disciplines . from the survey of curricular guides ,
the _ common vision _
project identified six themes for improving undergraduate curricula .
they are : ( 1 ) to find more pathways into and through the curriculum for both stem and non - stem majors ; ( 2 ) to increase the presence of statistics in student training ; ( 3 ) to increase the use of modeling and computation in order to enhance conceptual understanding and introduce the scientific method into math classes ; ( 4 ) to connect to skills needed in other disciplines ; ( 5 ) to improve communication skills through technical writing and presentations ; and ( 6 ) to aid in the transition from secondary to post - secondary education as well as from two - year to four - year institutions for transfer students .
modeling can play an important part of several of these themes , not just where it is mentioned explicitly . _
common vision _ notes that an early introduction to modeling , along with statistics and computation , can be a pathway `` into and through mathematical sciences curricula '' @xcite .
the society for industrial and applied mathematics ( siam ) suggests that professional societies play a greater role in the incorporation of modeling throughout the undergraduate curriculum @xcite , such as through the now - created siam special interest activity group on applied mathematics education , that recommends the development of a first - year modeling course that `` precedes and motivates the study of calculus and other fundamental mathematics for stem majors '' @xcite .
the american statistical association s curriculum guidelines for undergraduate programs in statistical science suggests that incorporating statistical modeling with simulations into mathematics courses can improve computational skills @xcite .
the mathematics association of america s subcommittee on curriculum renewal across the first two years ( crafty ) released a report in 2004 @xcite , titled _ the curriculum foundations project : voices of the partner disciplines _ , which emphasized mathematical modeling in math courses .
the report noted that `` every disciplinary group in every workshop '' identified mathematical modeling as an essential part of training students in the first two years of their undergraduate experience .
furthermore , having students engage in mathematical modeling can `` provide a mechanism for communication , expression , and reasoning that is cross - cultural and cross - disciplinary '' @xcite , while having students develop a `` set of transferable skills that has the potential to be far more impactful on their futures '' @xcite .
these findings are not exclusive to mathematics departments , but in fact run parallel to findings within mathematics partner disciplines .
for example , in biology , three reports , _
bio2010 _ by the national academy of science ( nas ) , _ vision and change _ by the american association for the advancement in science ( aaas ) and the _ scientific foundations for future physicians ( sffp ) _ report of the american association of medical colleges - howard hughes medical institute committee , all mention the important role of mathematics , and specifically modeling , in the future of biology as a discipline @xcite . _ bio2010 _ outlines , in an incredible amount of detail , the core concepts that future research biologists need from mathematics and computing , which include multiple mentions of modeling ( both mathematical and statistical ) throughout , as well as a section devoted to important modeling concepts . _
vision and change _ specifically names the _ ability to use modeling and simulation _ as a _ core competency _ ( emphasis added ) .
the _ sffp _ report also identifies modeling as a _ core competency _ in the following way : students should be able to `` apply quantitative knowledge and reasoning including integration of data , modeling , computation , and analysis and informatics tools to diagnostic and therapeutic clinical decision making . ''
based on the anecdotal experience of the co - authors , it seems that there are disciplinary differences by what is meant by models and modeling between mathematicians , statisticians , and biologists .
for example , does modeling require the use of data , or even numbers ?
does it require symbolic equations or formulas ? does a schematic qualify as a model ?
this is compounded on the mathematical side by the historical , philosophical , and physical separation of departments of mathematics and statistics on large campuses , and the teaching of statistics in many different departments ( statistics , mathematics , biology , psychology , economics , business , education , kinesiology , etc . ) on many campuses , both large and small . on the biological side
, there are many different subdisciplines , each with their own rich modeling traditions , _ e.g. _ , physiological modeling , ecological modeling , and more recently , systems biology modeling , among others .
there is also the specific field of mathematical biology , whose practitioners may move fluidly between the identities of mathematician and biologist . if , in all of this diversity , we are not clear about our definitions of models and modeling , then our students will not be clear .
we think , however , that it is possible to articulate an overarching framework for models and modeling that will unify what seem like disparate traditions with the advantage of improving student learning . ) .
each arrow is an activity in the modeling process ( defined in table [ table : arrows ] ) . ] for our discussions below , we define _ model _ as a simplified , abstract or concrete representation of relationships and/or processes in the real world . by simplified ,
we mean that it corresponds to a caricature of the real world rather than the real world itself ( figure [ fig : elephant ] ) .
we categorize representations of models into five types : * experiential * , * numerical * , * symbolic * , * verbal * , and * visual * ( figure [ fig : framework ] , boxes ) .
this `` rule of five '' categorization has been used previously in different contexts and will be discussed in detail in the next section @xcite . with model representations as _ objects _ clearly defined , we now define _ modeling _ as a _ process_. we define modeling as 1 ) the process of moving from observations of the real world to a model , 2 ) moving from one model representation to another model representation , or 3 ) comparing different models ( figure [ fig : framework ] , arrows ) . in the sections below
, we will explain what we mean by the `` rule of five '' framework , give illustrative examples of both model representations and the modeling process using the logistic growth model , the hardy - weinberg equilibrium ( hwe ) model , and a physical model of the structure of dna , and show how this framework works whether approached from the point of view of mathematics , statistics , or biology . while our focus in this paper is to provide a framework that will facilitate communication between mathematicians , statisticians , and biologists , this framework is adaptable to any stem discipline .
the crafty report , described above , noted that by engaging in mathematical modeling , students have an opportunity to describe their work , `` * * analytically * * , * graphically * , * numerically * , and * verbally * '' @xcite .
these model descriptions are historically referred to as `` the rule of four '' @xcite .
this concept came about in the 1980s and 1990s when it was recognized that over half of students enrolled in calculus courses in the united states did not finish the course @xcite .
this led to a number of calculus reform projects , many of which were funded by the national science foundation .
several of these projects used a `` multiple representation approach , '' @xcite as was suggested in the national council of teachers of mathematics _ curriculum and evaluation standards for school mathematics _
these ideas then came to the forefront with the work of hiebert and colleagues @xcite .
hiebert argued that in order to learn mathematics , a student must understand mathematics .
such understanding occurs when there is a continually evolving and strengthening network of connections to internal mental representations of mathematical ideas and procedures .
hiebert proposed a framework to aid in the understanding of mathematics by making connections between * numerical * , * graphical * , and * symbolic * representations .
note that the * symbolic * representation was called * analytical * by crafty above and has also been called * algebraic * by other projects .
the _ calculus connections project _ of oregon state university implemented a reform calculus emphasizing the representations of * graphical * , * numerical * , and * symbolic * , and the importance of switching between representations @xcite .
olaf project _ also described moving among the representations of * graphical * , * numerical * , and * algebraic * as being crucial to learning the concepts of calculus @xcite .
the _ calculus consortium _ at harvard university coined the term `` rule of three '' in their reform textbook @xcite .
the `` rule '' stated that equal weight should be given to describing topics * algebraically * , * numerically * , and * geometrically * @xcite .
note that the original word choice for the representations `` parallels '' course descriptions at the secondary level @xcite . in the second edition of the text @xcite , this became a `` rule of four '' with a fourth equal - emphasis on * verbal * descriptions of math problems by teachers and students . in subsequent editions of the text @xcite , the descriptions of the four representations became * graphical * , * numerical * , * symbolic * , and * verbal*. since the inception of calculus reform in the 1990s , it is now commonplace to see a wider variety of problems and the use of multiple representations in `` traditional '' math textbooks @xcite .
more recently , a fifth rule was proposed by simundza in a laboratory course for precalculus @xcite .
this fifth rule , * experiential * , is , in the words of simundza , a `` direct sensory experience of quantitative phenomena '' @xcite .
the importance of the * experiential * representation is alluded to in the american mathematical association of two - year colleges ( amatyc ) standards where they propose that mathematics be taught like the sciences as a laboratory discipline @xcite .
we are also using * experiential * for model representations , because introducing students to multiple forms of representation have been shown to improve student learning .
the first principle guideline from the universal design for learning ( udl ) is a recommendation to provide students multiple forms of representation since there are diverse ways learners comprehend information @xcite .
udl is a framework that addresses high variability in learners responses to instruction by suggesting flexibility in the curriculum to meet the varied needs of the students . using multiple representations
can allow students to make connections within and between concepts @xcite .
when students solved problems using more than one representation , student performance was better than for those learners who used a single strategy @xcite . the use of multiple forms of representation and the `` rule of five '' mirrors the parable of the blind men and the elephant as seen in figure [ fig : elephant ] .
john godfrey saxe s poem version of the parable describes six blind men touching different parts of an elephant in order to `` understand '' it @xcite .
each blind man compares each different elephant part to an everyday object that is similar - a wall ( side ) , snake ( trunk ) , spear ( tusk ) , tree ( knee ) , fan ( ear ) or rope ( tail ) . just as in the parable
, it is only through the use of different forms of representation that we may hope to gain the truest understanding of a problem .
interestingly , this analogy to the parable also works when considering the different ways in which mathematicians and biologists approach problems - each is experiencing one aspect of the problem , and through communication using a common language around modeling , we can create together a clearer picture of the world .
as discussed above and shown in figure [ fig : framework ] and table [ table : boxes ] , we extend the `` rule of five '' from its original use as an aid to calculus and precalculus instruction to a useful general description of model representations .
.the five types of model representations , corresponding to figure [ fig : framework ] with their definitions and classroom examples . [ cols= " < ,
< , < " , ] table [ table : arrows ] describes the modeling activities contained in the arrows between boxes in figure [ fig : framework ] .
we felt that it was important to present a fully connected graph to be inclusive of all potential uses of this framework , although some activities ( arrows ) would be more commonly performed than others . for each activity
, we have suggested ideas for how the framework could be used in the classroom in the context of teaching models of logistic growth , hardy - weinberg equilibrium , or the structure of dna . since the reader may be least familiar with * experiential * activities
, we have chosen to describe the connections to - and - from * experiential * representations both below and in table [ table : arrows ] .
for an arrow description and classroom example of all other connections , see table [ table : arrows ] .
* experiential * representations can either be simulations , physical models , or the conducting of a laboratory or field experiment . moving from * experiential * to * verbal * ( arrow 1 ) involves generating a formal scientific hypothesis , which implies mechanism , making predictions , or collecting qualitative observations from an experiment . moving from * verbal * to * experiential * ( arrow 2 ) is the converse , designing and executing an experiment to provide evidence for a hypothesis . moving from * experiential * to * numerical * ( arrow 3 ) involves collecting quantitative data from either an experiment or simulation . moving from * numerical * to
* experiential * ( arrow 4 ) refers to * numerical * data fed into a simulation or used to create a physical model .
it could also mean validation of a model against new experiments . moving from * symbolic * to * experiential * involves coding equations to create simulations ( arrow 17 ) .
arrow 18 ( * experiential * to * symbolic * ) is the converse .
for example , experienced modelers might immediately recognize logistic growth and write the equation for it . moving from * experiential * to * visual * involves drawing a sketch , graph , or schematic based on experiential observations ( arrow 19 ) .
arrow 20 ( * visual * to * experiential * ) involves making a physical model from a drawing , such as making a physical three dimensional model of dna from a two - dimensional drawing ( constructing an ikea book shelf from purely visual instructions and doing origami from folding schematics are two everyday examples of this ) .
we want to emphasize that work performed entirely within one particular box alone is not `` modeling '' per se , but the province of the disciplinary context within which the work is performed .
for example , a set of qualitative observations alone is a model representation for biologists ( see table [ table : boxes ] ) .
however , the process of designing and setting up the experiment to collect those qualitative data based on a scientific hypothesis ( arrows 1 and 2 in figure [ fig : framework ] and table [ table : arrows ] , i.e. going from * verbal * to * experiential * and back ) is a modeling activity . in another example , solving a system of differential equations to obtain formulas ( * symbolic * box , figure [ fig : framework ] ) is a mathematics activity and may be an important part of the larger modeling process , but alone is not a modeling activity .
however , using a differential equation to run a simulation ( arrow 10 in figure [ fig : framework ] going from * symbolic * to * numerical * ) or using the solution to plot a graph ( arrow 15 in figure [ fig : framework ] going from * symbolic * to * visual * ) are modeling activities . in a third example , `` application '' problems in which students are given a formula and asked to use it to compute an answer without explaining the meaning is not modeling @xcite .
however , more extended problems in which models are used to create simulated data ( arrow 10 in figure [ fig : framework ] going from * symbolic * to * numerical * ) , and in which these results are checked against additional data ( arrow 4 in figure [ fig : framework ] going from * numerical * to * experiential * ) are a much better illustration of modeling
. we are certainly not the first group to describe a conceptual framework for modeling .
recently siam has documented their take on the modeling process @xcite .
these types of diagrams usually talk about the iterative process of modeling , _ i.e. _ going back and forth between the different stages .
they can be more or less rigid in specifying a particular order of the steps in the modeling process @xcite .
we want to distinguish our paper from this previous literature in two ways : 1 ) by describing a framework for modeling that is inclusive of approaches taken by mathematicians , statisticians , and biologists , and 2 ) emphasizing that the * experiential * representation is crucial for student learning . in what follows ,
we give three examples of pathways through the modeling process / framework , starting from the point of view of mathematicians , statisticians , and biologists .
what these three pathways have in common are that they are `` messy '' and reflect what happens in actual expert practice @xcite . as we noted previously , a modeling investigation becomes more and more valuable as the number and variety of connections made between different model representations increases . in teaching modeling to students
, it is very helpful to begin the modeling process with something * experiential*. ideally this would use real observations of biological phenomena , as would be done in biology laboratory courses .
if this is not possible , a manipulatable representation of a model would serve as a reasonable starting point ( for example , in biology courses without a lab component , or in mathematics / statistics courses ) .
actual experiments and simulated experiments both belong in the * experiential * box ( figure [ fig : framework ] ) .
depending on the initial experience , * experiential * observations might generate either qualitative or quantitative data , which lead to * verbal * and * numerical * representations of a model , respectively . in our first example
, coming from the mathematical point of view , we talk about the derivation of a logistic growth model that follows the path * experiential * @xmath0 * numerical * @xmath0 * visual * @xmath0 * verbal * @xmath0 * symbolic*. we begin with an * experiential * representation by performing an experiment to measure the growth of a bacterial culture over time . if this can not be done in class , one could find a time lapse video of the phenomenon .
this approach to developing a * numerical * model would be to use raw population data to calculate absolute and relative growth rates .
one could then obtain a * visual * representation of the model by plotting the data in different ways ( for example , population size versus time , or relative growth rate versus population size ) .
then these steps lead to a * verbal * model consisting of the assumption that the growth rate is proportional to population .
an alternate path would go from * experiential * @xmath0 * verbal * @xmath0 * symbolic * when the * symbolic * model is obtained from the qualitative observation that more parents produce more offspring . in either pathway , the * symbolic * representation of the model does not come directly from qualitative data ( * verbal * box ) or quantitative data ( * numerical * box ) , but comes rather from a * verbal * representation obtained from the initial observations . in this example , the * verbal * representation leads to a * symbolic * representation in the form of a differential ( or difference ) equation .
the pathway can then be extended from * symbolic * @xmath0 * numerical * @xmath0 * visual * by running a simulation to produce data generated by the mathematical * symbolic * model , which can be graphed . or the pathway can be extended from a * symbolic * to * experiential * representation by creating an animation with computer graphics . the pathway can be extended from * symbolic * directly to * visual * by solving the equation analytically and using the resulting formula to prepare a graph . in either case , the * visual * representation may contribute a statement of model behavior to the * verbal * representation .
note that in this description of these pathways , the boxes and arrows are traversed several times and the order of the traversal is not fixed .
the previous example illustrates a modeling pathway that leads to a * symbolic * representation of a mechanistic mathematical model ( logistic growth ) .
the next example illustrates a modeling pathway leading to a * symbolic * representation of a statistical model , given in the context of hwe , which is typically taught in introductory biology courses .
pulling from previous experiments and observations ( * experiential * @xmath0 * verbal * ) , students are introduced to the five major forces in evolution : ( 1 ) selection , ( 2 ) genetic drift , ( 3 ) mutation , ( 4 ) gene flow , and ( 5 ) non - random mating @xcite .
to arrive at the probability model ( * symbolic * ) of hwe , we first assume a null * verbal * model that no evolutionary forces are present .
in particular , suppose we have an isolated ( no gene flow ) , infinite population ( no genetic drift ) where random mating , no selection , and no mutation occurs .
a simplified genetic model also assumes one locus and two alleles .
given this * verbal * model , the * symbolic * probability model representation is derived ( hwe ) stating that allele and genotypic frequencies do not change in subsequent generations .
many times this is where instruction stops , but it is informative to mention that this probability model can then be used to statistically test for the presence of hwe , which is a necessary but not sufficient condition for lack of evolutionary forces .
this is accomplished by measuring genotypic frequencies in a population of interest ( * experiential * @xmath0 * numerical * ) and then performing a chi - squared test to the null probability model of hwe ( * numerical * @xmath0 * symbolic * ) .
results of the test are then used as evidence for the presence or absence of evolutionary forces ( * symbolic * @xmath0 * verbal * ) .
it should be noted that most hypothesis testing in statistics follows this pathway : observed data collected from experiment ( * experiential * @xmath0 * numerical * ) is tested against a null probability model ( * numerical * @xmath0 * symbolic * ) , and the results are used as a quantification of evidence for a biological hypothesis ( * symbolic * @xmath0 * verbal * ) . finally , let s describe a pathway followed by biologists .
one of the most famous physical models in molecular biology is the physical model of the structure of dna created by james watson and francis crick , that inspired the more abstract schematic that appears in their _ nature _ paper @xcite . to build this model
, they integrated both quantitative and qualitative data from multiple sources ( * numerical * and * verbal * @xmath0 * experiential * ) .
this physical model enabled a hypothesis about mechanism ( * experiential * @xmath0 * verbal * ) : `` it has not escaped our notice that the specific pairing we have postulated immediately suggests a possible copying mechanism for the genetic material '' @xcite .
one of the co - authors of the current paper regularly uses building a chemically - correct physical model of dna as an activity in her courses . students gain a much deeper understanding of the chemical structure of dna from the direct manipulation of the physical model than from * verbal * descriptions or * visual * aids alone .
note that in this example , not all of the boxes or arrows were traversed , yet it is still an example of the process of modeling that enhances student learning .
ultimately , the results of a modeling investigation need to be checked against original or new qualitative and/or quantitative data ( or even other mathematical models ) .
this is what is called _
model validation_. a careful critique can almost always identify features of the biological system that are missing from the model . in biology , a large amount of stochasticity is often superimposed on deterministic phenomena , so we can not expect a model to exactly reproduce experimental data .
our first example of a modeling pathway , using * experiential * activities to examine logistic growth , would almost surely lead to a data set that does not precisely match the deterministic logistic growth example .
oftentimes , however , an important part of the modeling process includes recognizing when a model may need to be further refined to address the question at hand . in the course of a modeling investigation with students , the instructor s role often is to remind students to pause , validate the model , and if needed , begin again with the modeling process to refine it . in the instructional setting , we can use figure [ fig : elephant ] to remind students that modeling is theoretical science rather than mathematics per se .
results obtained from the study of a particular mathematical model pertain only to that model . whether they are useful in understanding the biological setting depends on comparing and contrasting model formulation with corresponding biological process and model - generated data with observed or experimental data .
this is a key ingredient in encouraging students to engage in the modeling process . even if students are moving through the boxes and are hence `` modeling
, '' it is imperative to remind them that they are working in a conceptualized model of reality @xcite .
they are using caricatures of reality , not dealing with reality itself , and the assumptions and results should be critically analyzed in that light . in our example model of logistic growth , there are a number of models that one can use for limited growth
. if the per capita growth rate is constant , the population is growing exponentially @xmath1 .
however there are a number of ways in which a population can experience a decreasing per capita growth rate that results in limited growth .
each of these limited growth scenarios says something different about the process of growth or the growth relationship and each results in slightly different limited growth curves .
when the inflow of a population is exponential and the outflow is mediated by intraspecific competition for resources , @xmath2 , this is equivalent to saying that overall , the per capita growth rate is linear @xmath3 .
however , growth may be limited by other processes which may result in a non - linear decrease of per capita growth rate . in this case
, one might fit many types of mathematical models to the data set and select the model that minimizes error and avoids over - fitting , for example by using a measure such as the akaike information criterion ( often referred to as aic ; for a review of model fitting , see @xcite ) .
the model that has the lowest aic may tell you something about which processes may be driving the population . in some cases , the fact that a model is not matching the outcomes observed in reality can also be important .
such is the case with the primary use of the hwe model discussed above .
this model predicts the distribution of offspring genotypes in a population given a list of assumptions . if those assumptions are true , and if in the parent generation the probability or frequency of allele a occurs is @xmath4 and the probability that allele a occurs is @xmath5 , then the next generation will have the following distribution of genotypes : @xmath6 , @xmath7 , @xmath8 .
if the observed distribution in the offspring does not match , this is useful , because then we know that one of the assumptions of the hwe model have been violated .
hwe is considered a classic use - case of a null model in biology @xcite .
we have synthesized a framework around modeling with the view that a unified framework allows us to be more purposeful practitioners around the teaching of modeling . in this paper
, we particularly emphasize the importance of the * experiential * component in the engagement of students and quality of learning .
our use of `` experiential '' in the context of modeling is slightly different than , but is still in alignment with , what is meant by `` experiential learning '' in other contexts @xcite .
there is an abundance of evidence that experiential learning can improve student learning outcomes @xcite .
there is also early evidence that adopting this type of framework , in particular multiple model representations and movements between them , can lead to success in subsequent quantitative courses @xcite .
some students may enter into the learning experience more comfortable with a subset of these representations due to their own ways of knowing or professional identity , but the goal is that they should know all and be able to move between them .
`` understanding science , '' a website that lays out a framework for the process of science , suggests the following three main actions for bringing the process of science into the classroom : ( 1 ) make it explicit , ( 2 ) help the students reflect upon it , and ( 3 ) give it context , again and again @xcite . because modeling is theoretical science , we suggest the same actions to bring our modeling framework into the classroom to strengthen both students understanding of , and abilities in , modeling .
we list these actions again here with specific modeling framework examples : * be explicit . *
tell your students what you are doing , and why , in the context of modeling .
teach students the modeling framework , including definitions of models and modeling , the different model representations ( boxes ) , and the many modeling activities ( arrows ) .
be clear that biologists are already engaged in modeling but may not realize it or use the same language or approach as mathematicians or statisticians .
there are many first steps with which one can begin modeling , and it does not have to begin the same way that was outlined in our example modeling pathways . teaching any arrow can be a first step in teaching modeling as long as one is explicit in connecting it to the modeling process . *
help them reflect ( and connect ) .
* have students reflect upon the modeling process by assigning metacognitive exercises , such as a one - minute paper . work with instructors from different disciplines to help students make connections between classes to solidify their understanding of modeling .
using a common framework can help us engage in conversations with colleagues from other disciplines , and thus bring the connections to our students .
* give it context , again and again .
* ground the modeling investigation in the biological problem , using the * experiential * representation .
use the framework to acknowledge and clarify the various approaches that each discipline takes to solving the same scientific problem .
show that different practitioners have different paths through the boxes that are equally valid .
give examples of different paths in the same context .
while traversing the modeling pathways may be intuitive for expert practitioners , some studies have found that students have difficulty translating between model representations @xcite , although the facility of translation did vary depending upon the specific relations selected @xcite . yet , in keeping with the recommendation from understanding science to be explicit @xcite , one study showed that simply telling students the purpose of the multiple representations can have a positive impact on learning @xcite . in particular , schwonke _
et al_. conducted two studies , collecting gaze data from students viewing multiple representations of the same problem . from the initial study ,
many students did not understand why they should have different representations and why they should transfer between them . to improve the transitions between representations , one group of students in a follow - up study ( reported in the same paper @xcite ) received additional instructions explaining the `` bridge between [ the ] problem texts and equations , '' while a control group did not have these instructions .
et al_. concluded that an explanation of the different representations improved the learning outcomes for both low- and high - prior knowledge students , but in different ways .
the low - prior knowledge group seemed to `` transfer knowledge more easily between representations , '' while the high - prior knowledge students benefited because they paid more attention to the different representations .
our conclusion from this prior work is that while it is important for the instructor to facilitate use of multiple model representations and modeling activities , it is most valuable when the reasons behind it are made clear to the students , _
i.e. _ , make it explicit , reflect and connect , and provide context @xcite . for information and ideas around using our framework for teaching ,
see the following collection at qubeshub.org ( https://qubeshub.org/groups/nimbios_wg_teachingquantbio/collections/a-interdisciplinary-framework-for-teaching-modeling--primus-paper ) @xcite .
there have been numerous calls for mathematicians to work more closely with members of partner disciplines , such as biology , to improve student learning and retention in stem ( _ e.g. , the mathematical sciences in 2025 _ and _ a common vision _ , @xcite ) .
analogous to this , there have been numerous calls for biologists to incorporate more mathematical modeling into their curriculum to better prepare students for future careers in research and the health professions ( _ e.g. , vision and change _ , _ bio2010 _ , _ scientific foundations for future physicians _ , @xcite ) . despite this widespread agreement on what needs to happen , implementing the recommendations
has been slow .
one barrier that must be overcome to initiate true transdisciplinary conversations and collaborations on improving the modeling curriculum is to agree upon a common definition of what is meant by models and modeling between biologists , mathematicians , and statisticians .
therefore , we have proposed adapting the `` rule of five '' , which has previously informed calculus reform efforts , to describe a framework for modeling that can bring all the disciplines together .
this framework defines five types of model representations ( * experiential * , * verbal * , * numerical * , * visual * , and * symbolic * ) and modeling activities that provide flexible routes through a modeling investigation by students .
we give examples about what the implementation of this framework may look like in the classroom , along with the associated benefits to student learning @xcite . in particular , an advantage for students working with * experiential * representations is that they provide an entryway for understanding models that do not require the abstraction of * symbolic * or * visual * representations , the integration of detail needed to understand a * numerical * representation , or the conceptual knowledge and reading comprehension needed to understand a * verbal * representation .
finally , we share resources that we think will be helpful for others to use .
we also hope this framework will unify practitioners coming from different parent fields ( mathematics , statistics , and biology ) and allow them to find the similarities and differences in their approaches to modeling , leading to more productive interdisciplinary conversations .
making the connections explicit to students in different disciplinary courses will reinforce the concepts for students and empower them to apply knowledge from one domain to another , making them informed citizens for the 21^st^ century . in their professional futures , students will not encounter textbook questions with multiple choice answers . instead , they will hear a wildlife biologist discuss the rate of population growth , they will see a graph in a paper that they are reading , they will monitor a population , or they will use software to run management scenario planning . a student that can move between these representations to help solve problems is one with a superior preparation for the profession .
we have a trained disciplinary identity , but we are all modelers .
we can work together to help our students be modelers , too .
this work was conducted as part of `` unpacking the black box : teaching quantitative biology '' working group at the national institute for mathematical and biological synthesis ( nimbios ) , sponsored by the national science foundation through nsf award # dbi-1300426 , with additional support from the university of tennessee , knoxville .
co - authors glenn ledder and richard c. schugart were supported as nimbios sabbatical fellows .
we would like to thank the other members of the working group for useful discussions : melissa l. aikens , joseph t. dauer , samuel s. donovan , ben g. fitzpatrick ( also a sabbatical fellow at nimbios ) , gregory d. goins , kristin p. jenkins , john r. jungck , and robert l. mayes .
we would also like to thank deborah hughes - hallett for her insights .
organization of this co - authorship and working group between meetings would not have been possible without qubes hub supported by nsf awards # dbi 1346584 , due 1446269 , due 1446258 , and due 1446284 @xcite and our qubes communications student , timothy beaulieu . 0 ainsworth s. , d. wood , and p. bibby 1996 . _ coordinating multiple representations in computer based learning environments_. proceedings of the european conference on artificial intelligence in education .
336 - 342 .
ainsworth s. 1999 .
_ the function of multiple representations_. computers in education .
336 - 342 .
american association for the advancement in science ( aas ) .
_ transforming undergraduate education in biology : mobilizing the community for change_. http://visionandchange.org
. accessed 23 may 2016 .
american mathematical association of two - year colleges ( amatyc ) .
2014 . _ opening doors through mathematics .
position on professional development_. http://www.amatyc.org/?page=positionprofdev . accessed 26 may 2016 .
american statistical association ( asa ) .
2014 . _ discovery with data : leveraging statistics and computer science to transform science and society_. alexandria , va .
accessed 31 may 2016 .
bender e.a .
an introduction to mathematical modeling_. wiley - interscience .
mineola , new york .
1 - 46 .
bransford j.d . ,
brown a.l . , cocking r.r .
, donovan m.s . , and pellegrino j.w .
( eds . ) .
_ how people learn brain , mind , experience , and school_. commission on behavioral and social sciences and education national research council . national academy press .
washington , d.c . 1 - 363 .
brauer f. , castillo - chavez c. , and bles d. 2001 . _ mathematical models in population biology and epidemiology_. vol .
em springer science buissness media .
springer - verlag , new york .
carson e.c . and cobelli c. 2014 .
_ modeling methodology for physiology and medicine_. 2nd edition .
waltham , ma . 1 - 562
. center for advanced spatial technology ( cast ) .
_ universal design for learning guidlines_. version 2.0 .
wakefeild , ma .
cox r. and brna p. 1995 .
_ supporting the use of external representations in problem solving : the need for flexiable learning environments .
_ _ journal of artificial intelligence in education_.6 : 239 - 302 .
dick t. and patton c. 1994 .
_ calculus of a single variable .
_ boston , ma .
pws publishing company .
douglas r.g .
_ toward a lean and lively calculus : report of the conference / workshop to develop curriculum and teaching methods for calculus at the college level .
journal of college mathematics_. 18(5 ) : 439 - 442 . donovan s. , eaton c.d . , gower s.t . , _ et al_. 2015 . _
qubes : a community focused on supporting teaching and learning in quantitative biology .
letter in biomathematics_.2(1 ) : 46 - 55 . ganter s. and barker w. ( eds . ) 2004 . _
the curriculum foundations project : voices of the partner disciplines_. washington , dc : mathematics association of america .
garfunkel s. and montgomery m. ( eds . ) 2016 .
_ guideline for assessment and instruction in mathematical modeling education_. bedford , ma .
consortium for mathematics and its applications .
gleason a.m. and hughes - hallet d. 1992 .
_ the calculus consortium based at harvard university_. in _ focus on calculus : a newsletter for the calculus consortium at harvard university_. issue 1 .
harvard university .
hiebert j. 1986 .
_ conceptual and procedural knowledge : the case of mathematics . _ mahwah , nj .
erlbaum associates . heibert j. and carpenter t.p .
_ learning and teaching with understanding_. in grouws d.a .
( ed . ) _ handbook oof research on mathematics teaching and learning . _ new york , ny . macmillian .
holling c.s .
the components of predation as revealed by a study of small - mammal predation of the european sawfly .
the canadain entomologist_. 91 : 293 - 320 .
holling c.s .
1959b . _ some characteristics of simple types of predation and parasitism .
the canadian entomologist_. 91 : 293 - 320 .
howard hughes medical institute committee ( hhmi ) .
2009 . _ scientific foundations for future physicians_. report of the american association of medical colleges.university of delaware . 1 - 46 .
hughes - hallett d. 2006 .
_ what have we learned from calculus reform ?
the road to conceptual understanding_. in hastings n. ( ed . ) _ rethinking the courses below calculus_. mathematics association of america .
hughes - hallett d. 2016 . personal interview .
may 31 , 2016 . huges - hallett d. , gleason a.m. , mccallum w.g . ,
_ et al_. 1995 .
_ calculus : single and multivariate_. new york , ny . wiley and sons .
hughes - hallett d. , gleason a.m. , mccallum w.g . , _ et al_. 1998 . _
calculus : single and multivariate_. 2nd edition .
new york , ny .
wiley and sons .
hughes - hallett d. , gleason a.m. , mccallum w.g . , _ et al_. 2005 . _
calculus : single and multivariate_. 4th edition .
new york , ny . wiley and sons .
hughes - hallett d. , gleason a.m. , mccallum w.g . , _
et al_. 2013 .
_ calculus : single and multivariate_. 6th edition .
new york , ny . wiley and sons .
jungck j.r .
, gaff h. , and weisstein a.e . 2010 . _ mathematical manipulative models : in defense of `` beanbag biology''_. cbe - life sciences education . 9
: 201 - 211 .
kolb a. and kolb d.a .
2005 . _ learning styles and learning spaces : enhancing experiential learning in higher education .
academy of management learning and education_. 4(2 ) : 193 - 212 .
_ experiential learning : experience as the source of learning and development_. 2nd edition .
prentice hall .
englewood cliffs , nj . 1 - 377 .
kolb d.a . ,
boyatiz r.e .
, mainemelis c. ( eds . ) sternberg s. and zhang l. 2001 . _ experiential learning theory : previous research and new directions_. lawrence erlbaum associates .
new york , ny . 197 - 226 .
ledder g. 2005 .
_ differential equations : a modeling approach_. mcgraw - hill higher education .
university of nebraska
. ledder g. 2007 .
_ bugbox - predator_. https://www.math.unl.edu/~gledder1/bugbox/. accessed 26 may 2016 .
ledder g. 2016 .
_ emperical modeling : choosing models and fitting them to data .
_ _ the college mathematics journal . _ 47 : 1 - 11 .
mathematical association of america ( maa ) .
2004 . _ curriculum renewal across the first two years ( crafty)_. http://www.maa.org/programs/faculty-and-departments/curriculum-department-guidelines-recommendations/crafty .
accessed 23 may 16
. national academy of science ( nas ) .
_ bio 2010 : transforming undergraduate education for the future research biologists_. washington , dc . the national academic press
. national council of teachers of mathematics ( nctm ) .
_ curriculum and evaluation standards for school mathematics . _
reston , va .
the national council of teachers of mathematics . national research council ( nrc ) .
2013._the mathematical sciences in 2025_. washington , dc . the national academies press
. 1 - 222 .
nandi b. and chakravarti a. 2013 . _ an introduction to mathematical modeling_. jeevanu times . 13(1):16 - 18 .
ostebee , a. and zorn , p. 1994 . _
calculus from graphical , numerical , and symbolic points of view . _ orlando , fl .
saunders college publishing .
presidents council of advisers on science and technology ( pcast ) .
_ engage to excel : producing one million additional college graduates with degrees in science , technology , engineering , and mathematics_. 1 - 130 .
quantitative undergraduate biology education and synthesis ( qubes ) . 2015 .
_ nimbios working group : unpacking the black box . _ https://qubeshub.org/groups/nimbios_wg_teachingquantbio/overview . accessed 26 may 2016 .
quantitative undergraduate biology education and synthesis ( qubes ) . 2016 . _
qubes : the power of biology@xmath9math@xmath9community_. https://qubeshub.org . accessed 26 may 2016 .
raven p. and johnson g. 2016 .
_ biology_. mcgraw - hill education .
redish f.e . and kuo e. 2015 .
_ language of physics , langauge of math : disciplinary culture and dynamic epistemology .
science and education_. 24(5 ) : 561 - 590 .
1873 . _ the poems_. highgate edition .
james r. osgood and company .
boston , ma . 1 - 485 .
saxe k. and braddy l. 2015 . _ a common vision for undergraduate mathematical sciences programs in 2025_. washington , dc .
mathematics association of america .
schwonke r. , berthold k. and renkl a. 2009 .
_ how multiple external representations are used and how they can be made more useful
. applied cognitive psychology _ 23 : 1227 - 1243 .
simundza g. 2006 .
_ the fifth rule : direct experience of mathematics , a fresh start for collegiate mathematics_. mathematical association of america .
320 - 328 .
society for industrial and applied mathematics ( siam ) .
modeling across the curriculum i._. philidelphia , pa . society for industrial and applied mathematics ( siam ) .
_ modeling across the curriculum ii._. philidelphia , pa . sorgo a. and jungck j. 2010 .
_ connecting biology and mathematics : first prepare the teachers .
cbe life sciences education_. 9(3):196 - 200 .
steele j. and demand media .
_ steady state vs. equilibrium in biology_. seattle pi .
accessed 19 may 2016 .
svoboda j. and passmore c. 2013 .
_ the strategies of modeling in biology education .
science and education .
_ 22(1 ) : 119 - 142 .
tabachneck h.j.m .
, koedinger k.r . and
nathan m.j .
_ towards a theoretical account of strategy use and sense making in mathematical problem solving_. proceedings of the 16th annual conference of the cognitive science society .
hillsdale , nj . erlbaum .
836 - 841 .
understanding science . 2016a .
_ university of california museum of paleontology_. http://www.understandingscience.org . accessed 19 may 2016 .
understanding science . 2016b . _ the real process of science_. http://undsci.berkeley.edu/article/howscienceworks_02 .
accessed 19 may 2016 .
utter f.w .
_ relationship among ap calculus teachers pedagogical content beliefs , classroom practice and their student s acheivement_. dissertation , department of mathematics .
oregon state university .
waldron i. 2016 . _ population growth - exponential and logistic models vs. complex reality_. university of pennsylvannia .
philadelphia , pa . http://serendip.brynmawr.edu/exchange/bioactivities/pop . accessed 19 may 2016 .
watson j.d . and crick f.h . 1953 .
_ molecular structure of nucleic acids_. _ nature_. 171(4356 ) : 737 - 738 . wilensky u. 1999 .
_ center for connected learning and computer - based modeling_. northwestern university .
evanston , il .
/ netlogo/. accessed 19 may 2016 .
windham d.m . , 2008 .
_ faculty perceptions of a calculus reform experiment at a research university : a historical qualitative analysis_. dissertation , department of middle and secondary education , florida state university .
proquest llc .
accessed 19 may 2016 .
eaton earned her b.a . in mathematics with a minor in zoology and her m.a . in interdisciplinary mathematics from the university of maine .
she then earned her ph.d . in mathematical ecology and
evolutionary theory from the university of tennessee .
eaton s primary research interests are interdisciplinary mathematics and education .
she is a principal investigator on the qubes project for quantitative undergraduate biology education and synthesis .
callender earned a b.a . in mathematics with minors in computer science and music from wesleyan college , and an m.s . and ph.d . in mathematics with an emphasis in mathematical biology from vanderbilt university .
she then worked as a postdoctoral fellow at the institute for mathematics and its applications at the university of minnesota .
her primary research interests include mathematical modeling of cellular and molecular processes and infectious diseases , sensitivity analysis , and mathematical biology education .
dahlquist earned a b.a . in biology from pomona college and a ph.d . in molecular , cellular , and developmental biology from the university of california , santa cruz .
she then worked as a postdoctoral fellow at the gladstone institute of cardiovascular disease at the university of california , san francisco .
she studies the systems biology of gene regulatory networks using genomics , mathematical , and computational biology approaches , bringing this research into the classroom .
g. ledder earned a b.s . in ceramic engineering from iowa state university and a ph.d . in applied mathematics from rensselaer polytechnic institute .
his research interests span a variety of theoretical science areas including groundwater flow , plant physiology , and dynamic energy budget theory .
he was the editor of a primus special issue and maa notes volume on mathematics pedagogy for biology students .
he is also the author of a mathematics textbook for biology students and is a member of the qubes advisory board .
lamar earned a b.s . in mathematics from the university of texas at san antonio and a ph.d . in mathematics from the university of texas
at austin .
his research interests include mathematical biology , specifically focusing on computational neuroscience and ecological modeling , as well as quantitative biology education .
he is also a principal investigator on the qubes project .
schugart earned a b.a . in mathematics from the state university of new york college at geneseo , and an m.s . and ph.d . in applied mathematics from north carolina state university
he was a post - doctoral fellow at the mathematical biosciences institute at the ohio state university .
his research area is mathematical biology , with particular interests on modeling in wound healing and engaging undergraduate students in research .
he is also an editor for smb digest , a weekly moderated electronic mail containing items of current interest to the math biology community for the society for mathematical biology .
|
professional societies and collaboratives have called for the incorporation of more quantitative skills , such as mathematical modeling , securely into the foundation and throughout undergraduate biology programs .
however , the integration of modeling both within and between mathematics and biology curricula remains limited .
one barrier to integration comes from differing perceptions about the nature of modeling , thereby limiting communication between disciplines . in this paper
, we extend the `` rule of five , '' initially used in calculus reform efforts , to build a modeling framework that explicitly separates and defines models as representations and modeling as a process .
we discuss benefits to student learning and cross - disciplinary communication .
|
the subdivision of electronic states into delocalized , itinerant , and localized , strongly correlated states is one of the most fundamental ones in the description of solids .
crudely , delocalized states exist when the intersite electron hopping is larger than the typical electron
electron interaction energy , and localized states exist in the opposite limit . in quantum chemistry
these two limits correspond e.g. to the description of electrons in terms of molecular orbitals ( mulliken description , known in chemistry as molecular orbital linear combination of atomic orbitals ( mo lcao ) method ) , and strongly correlated electrons are described by the heitler london approximation . for concentrated solids
these two regimes give either standard band systems ( metals or semiconductors ) , or mott insulators .
the electron electron repulsion , e.g. the hubbard s @xmath0 , is essentially an atomic property and does not depend on interatomic distance , whereas the electron hopping strongly depends on this distance .
this leads to the concept , largely advocated e.g. by goodenough @xcite , that there should exist a critical interatomic distance @xmath1 ( different for different transition metal ( tm ) ions ) dividing the regions of localized and itinerant behaviour of electrons .
when this distance changes , e.g. under pressure , there should occur an insulator metal ( mott ) transition .
usually mott transition in tm compounds such as tm oxides is supposed to occur homogeneously in the whole sample .
but , as became clear recently , this is not the only possibility .
there may appear novel , inhomogeneous states close to a mott transition , so that we have , in a sense , a `` partial '' mott transition : in some parts of the system , in some particular clusters , the interatomic distances may already be smaller than @xmath1 , so that these clusters may be better described using molecular orbitals , whereas the distance _ between _ such clusters will still remain large , so that there will be no net metallic conductivity .
the formation of such `` metallic '' clusters has much in common with the valence bond solids .
this phenomenon , i.e. such `` fractional '' mott transition , is most often observed in low - dimensional and in frustrated systems .
indeed , in regular three - dimensional lattices such as perovskites we usually have homogeneous ordered states , e.g. 3d antiferromagnetic ordering .
molecular - type clusters such as valence bond states are much more plausible in low - dimensional and frustrated systems .
below i will describe this situation using several specific examples of systems with triangular and pyrochlore ( spinel ) lattices , and also on some quasi - one - dimensional systems .
i will also shortly mention some other possible novel phenomena , which may appear in systems close to the localized
itinerant crossover .
metal insulator transitions with the formation of unusual structures are observed in many spinels .
the recent examples are mgti_2o_4 ] @xcite , cuir_2s_4 ] @xcite or alv_2o_4 ] @xcite .
structural studies have shown that in all these cases rather unusual structural modifications take place , which often can be described as the formation of molecular clusters . in mgti_2o_4 ] there appears a `` chiral '' structure @xcite , in cuir_2s_4 ] ir octamers are formed @xcite , but both these phenomena can in fact be explained by the formation of ti or ir _ singlet dimers _ , which in a frustrated lattice of @xmath2-sites of a spinel finally give rise to these chiral or octamer structures @xcite . and in alv_2o_4 ] the molecular clusters formed consist of 7 vanadiums real large `` heptamer molecules '' .
electrons in a lattice of @xmath2-sites of a spinel .
the @xmath3 and @xmath4 orbitals are shown . ] in all these cases the formation of these clusters is largely caused by specific features of the @xmath2-site spinel lattice , which may be represented as a collection of 1d chains running in @xmath3 ( or @xmath5 ) , @xmath6 and @xmath4 directions , see fig .
[ fig:1 ] .
the @xmath7-orbitals of ti or v ions on this lattice have a strong direct @xmath8@xmath8 overlap , such that e.g.the electrons from the @xmath3-orbitals can hop to the same orbitals in neighbouring ions in the @xmath3-direction , see fig .
[ fig:1 ] , or similarly @xmath4-electrons can hop in the @xmath4-direction , etc . as a result
the electronic structure in these , basically cubic crystals , has essentially one - dimensional character .
and in the insulating phase one forms a peierls state with singlet dimers in these 1d chains @xcite , which for a spinel lattice give the chiral structure in mgti_2o_4 ] , the heptamers in alv_2o_4 ] and octamers ( consisting actually of dimers ) in cuir_2s_4 ] .
note that the interatomic distances in these clusters are rather short : they are shorter than the critical distance @xmath1 for the localized itinerant crossover @xcite , so that one can consider the electronic state in these clusters as forming molecular orbitals , i.e. in these clusters we are already on the `` metallic '' side of the mott transition .
but the distance _ between _ these clusters is larger than @xmath1 and as a result the total compound is insulating .
the `` metallic '' clusters in the examples discussed above are singlet .
but this is not the only possibility .
magnetic ( e.g. triplet ) clusters can also be formed in some cases .
this seems to happen in znv_2o_4 ] @xcite .
v spinels such as mgv_2o_4 ] , znv_2o_4 ] , cdv_2o_4 ] show cubic tetragonal structural transitions @xcite with the formation of a very specific magnetic structure at lower temperatures , with @xmath9 pattern in @xmath6- and @xmath4-chains ( and with @xmath10 ordering in the @xmath3-direction ) @xcite , see fig .
[ fig:2 ] .
usually the properties of these systems are explained by some type of orbital ordering : the @xmath3-orbitals are always occupied due to tetragonal distortion with @xmath11 , and the remaining second @xmath7 electron of the v^3+](@xmath12 ) ions is supposed to form an additional orbital ordering of some type @xcite .
but our recent ab initio calculations @xcite have shown that there should exist in znv_2o_4 ] a v dimerization in @xmath6- and @xmath4-chains , see fig .
[ fig:2 ] , with , surprisingly , ferromagnetic bonds becoming shorter ( thick bonds in fig .
[ fig:2 ] ) !
one can explain this tendency by noticing that , in contrast to most examples of valence bond solids with singlet bonds , here we are dealing not with ions with one electron with @xmath13 , but with @xmath12 ions with @xmath14 .
one can argue that the extra delocalization of one of two electrons in a short bond enhances the ferromagnetic ( actually double - exchange ) interaction between these @xmath12 , @xmath14 ions .
the ab initio calculations of another material of this family , cdv_2o_4 ] , done by a different method , have confirmed the formation of these dimers , and demonstrated , in addition , that their formation can lead to the appearance of ferroelectricity , which was indeed observed in cdv_2o_4 ] @xcite . , following @xcite ; the same pattern also exists in cdv_2o_4 ] @xcite . ]
interestingly enough , short v
v distances obtained theoretically for znv_2o_4 ] are only @xmath15 again shorter than the critical v
v distance @xmath16 @xcite .
thus again these v dimers may be considered as `` metallic '' , though the material itself remains insulating , albeit with a rather small energy gap @xmath17 @xcite .
some more examples of `` metallic '' clusters in an overall insulating matrix are found in layered materials , notably with triangular lattices , such as livo_2 ] @xcite , livs_2 ] @xcite , tii_2 ] @xcite .
this lattice is usually considered as frustrated , meaning that it is not bipartite .
but in livo_2 ] , livs_2 ] , tii_2 ] we are dealing with ions ti^2 + ] , v^3 + ] with two @xmath7 @xmath8-electrons , which , besides spin , have also triple orbital degeneracy . and
, if one can not subdivide a triangular lattice into two sublattices , one can naturally subdivide it into three !
that is indeed what happens in these materials @xcite , see fig .
[ fig:3 ] , in which the occupied orbitals are shown . after forming this orbital superstructure
, we have strong antiferromagnetic exchange in the shaded triangles of fig . [ fig:3 ] , and in effect
these bonds become shorter than the others , and there are singlet states formed _ on such trimers _ ( three @xmath14 ions with strong antiferromagnetic coupling have a singlet ground state with @xmath18 ) .
our recent detailed ab initio calculations @xcite confirmed this picture and gave a strong antiferromagnetic exchange in a trimer , the exchange interaction between trimers being much weaker and ferromagnetic .
lattice optimization gave v v distances in a trimer of @xmath19 , in good agreement with the experimental value .
note that in this case the short distance in a cluster here a trimer is even shorter than the v
v distance in v metal , @xmath20 , i.e. these clusters can indeed be considered as `` metallic '' . with three orbital sublattices , leading to the formation of spin singlet states on tightly bound v triangles (
shaded ) , following @xcite . ]
if however we increase the total lattice parameter , which we can do e.g. by going from livo_2 ] to navo_2 ] , the situation changes drastically . as shown in @xcite , the orbital ordering in navo_2 ]
is quite different from that in livo_2 ] : only two types of orbitals are occupied , see fig .
[ fig:4 ] . as a result ,
after such orbital ordering the system becomes topologically a square lattice , and as such it can develop the ordinary nel magnetic ordering , which it actually does .
note also that , in contrast to livo_2 ] , in navo_2 ] the bonds with strong orbital overlap become _ not shorter , but longer _ !
one can explain this as a competition of two effects : stronger covalency tends to make such bonds shorter , but the bond charge repulsion ( coulomb repulsion of electron clouds on orbitals directed towards each other ) would tend to make such bonds longer . apparently when the system has a relatively small lattice parameter , or small average tm
tm distance , i.e. when the system is close to the localized
itinerant crossover , the hybridization is rather strong , and the first effect , the tendency to increase it still further , prevails , so that such bonds become even shorter ; this is what happens in livo_2 ] and livs_2 ] . if , however , the system has localized @xmath8-electrons and is far from the mott transition , the @xmath8@xmath8 overlap is exponentially small , and the second effect , the bond
charge repulsion , may become stronger , as a result of which it is favourable to make such bonds longer , to reduce this repulsion further .
this is apparently what happens in navo_2 ] .
@xcite . ] until now we saw the formation of different types of tm clusters , mostly singlet ones : dimers ( mgti_2o_4 ] , cuir_2s_4 ] ) , trimers ( livo_2 ] , livs_2 ] , tii_2 ] ) , or larger clusters such as heptamers in alv_2o_4 ] .
also singlet tetramers can be formed , especially in layered materials .
one such example is cav_4o_9 ] @xcite , with the interesting `` chiral '' lattice , in which the singlets are formed by four v ions around an `` empty '' plaquette . in all these cases either low - dimensionality or the frustrated nature of the lattice apparently plays an important role . as mentioned above , valence bond solids are indeed especially easily formed in frustrated or quasi - one - dimensional systems this is in fact the essence of the peierls phenomena .
a very clear example of this phenomenon , in particular its features close to the localized
itinerant crossover , is observed in tiocl .
this quasi - one - dimensional material is insulating with localized @xmath8-electrons at ambient pressure , and it has a spin - peierls transition at @xmath21 ( with an incommensurate phase at @xmath22 ) . under pressure
its energy gap strongly decreases , and it approaches an insulator
metal transition at @xmath23 @xcite .
however direct dc resistivity measurements have shown @xcite that tiocl actually remains insulating even for @xmath24 , at least up to @xmath25 , albeit with a smaller gap .
surprisingly , ab initio calculations with lattice optimisation have shown @xcite that in the high - pressure phase , in which indeed the energy gap strongly decreases , the ti
ti dimerization itself does not decrease but _ increases _ instead , see fig .
[ fig:5 ] ! a qualitative explanation can be obtained , using again the idea of the proximity of this material to the localized itinerant crossover . at @xmath26
the @xmath8-electrons are localized , and we have a spin - peierls transition , with @xmath27 and with the degree of dimerization @xmath28 proportional to @xmath29 where @xmath30 is the exchange interaction , @xmath31 , and @xmath32 is the spin phonon coupling constant .
note that the energy gap in this regime is still of the mott
hubbard type , i.e. it is large , @xmath33 . under pressure
tiocl approaches the insulator
metal transition , as observed in @xcite .
but after crossing to the itinerant regime , the system still `` knows '' that it is one - dimensional , and as such it has a peierls instability , with @xmath27 ( @xmath34degree of dimerization@xmath35energy gap ) @xmath36 with @xmath37 being the electron
electron coupling constant in this regime .
that is , the very nature of insulating state in this regime is _ due to peierls dimerization _ , and is not of mott insulator type . but
the dimerization is in this regime proportional to the bandwidth , or the hopping matrix element @xmath38 , and not to @xmath31 .
thus apparently we have here , simultaneously with the localized itinerant crossover , also a spin - peierls peierls transition .
as we have discussed above , a possible situation , especially in low - dimensional and frustrated systems , is that the mott transition occurs not in the whole system simultaneously , but `` piecewise '' , first in some finite clusters such as dimers , trimers , tetramers etc .
but other nontrivial states may also occur in the vicinity of mott transitions , different both from those typical for strong mott insulators and for normal metals .
these situations can occur also for simple lattices such as those of perovskites . in this chapter for completeness
i will also shortly discuss some of these phenomena
. in some systems with strongly correlated electrons charge ordering occurs with decreasing temperature .
usually it appears in systems with noninteger electron concentration ( fractional number of electrons per tm ion ) .
the well - known examples are the verwey transition in magnetite fe_3o_4 ] or charge ordering in half - doped manganites such as la_0.5ca_0.5mno_3 ] , see e.g. @xcite .
but there are also systems in which charge ordering , or rather spontaneous charge disproportionation , appears in case of , formally , integer electron occupation , or integer valence .
such is for example the situation in cafeo_3 ] , in which there occurs charge segregation @xmath39 a phase transition with the formation of a fe^3+]fe^5 + ] superstructure occurs in cafeo_3 ] at @xmath40 and is simultaneously a metal insulator transition @xcite .
note that a similar compound srfeo_3 ] with somewhat broader bands remains metallic down to lowest temperatures .
apparently the phenomenon of charge disproportionation in this case occurs in the system close to the localized
itinerant crossover .
similar spontaneous charge disproportionation occurs also in la_2/3sr_1/3feo_3 ] @xcite ( 3fe^4 + 2fe^3 + + fe^5 + ] ) , but also e.g. in nickelates @xmath41nio_3 ] , especially those with small rare earths @xmath42 , y , etc .
@xcite . in nickelates
charge disproportionation may be described by the `` reaction '' @xmath43 ( ni^3 + ] and ni^4 + ] here are in the low - spin state ) . in these systems charge disproportionation
is also a metal insulator transition . yet
other systems showing the same phenomenon are materials formally containing au^2+](@xmath44 ) , which at normal conditions always disproportionate into au^1+](@xmath45)@xmath46au^3+](@xmath47 ) .
this occurs e.g. in csaucl_3 ] which at ambient pressure contains two inequivalent au sites ( its formula is usually even written as cs_2au_2cl_6 ] ) . under pressure
this system also becomes metallic with equivalent au ions @xcite .
one can notice three specific features , which apparently are important for the very phenomenon of spontaneous charge disproportionation .
first , in tm compounds such charge disproportionation occurs in systems in which the tm ion in the `` original '' state would have been a strong jahn teller ion , with double orbital degeneracy ( fe^4+](@xmath48 ) ; low - spin ni^3+](@xmath49 ) ; au^2+](@xmath50 ) ) .
thus it seems that this spontaneous charge disproportionation appears in fe^4 + ] and ni^3 + ] compounds _ instead of _ the jahn teller or orbital ordering , when the systems are close to localized itinerant crossover @xcite .
but in `` au^2 + ] '' compounds such charge disproportionation is stabilized by the _ enhanced _ jt distortion instead @xcite .
second , the tm ionic states involved typically correspond to states with unusually high oxidation states , or untypically high valence ( fe^4 + ] and fe^5 + ] ; ni^3 + ] and ni^4 + ] ; au^3 + ] ) . as follows from the standard classification @xcite , these states correspond to the situation with small or negative charge - transfer gaps and with a large contribution of ligand ( e.g. oxygen ) @xmath51-holes . in effect
e.g. the reaction ( [ eq:3 ] ) should be rather written as @xmath52 where @xmath53 denotes the ligand hole ( e.g. fe^3+]@xmath53 instead of fe^4 + ] denotes the state fe^3+](@xmath54)o^-](@xmath55 ) ) .
apparently the large contribution of ligand holes facilitates the process of charge disproportionation ( the increase of coulomb repulsion in the reaction ( [ eq:5 ] ) is definitely much smaller than if it would occur on the @xmath8-shells ( [ eq:3 ] ) ) .
these factors lead to the third conclusion : that the spontaneous charge disproportionation is typically observed in systems close to an insulator metal transition .
experimentally this is indeed the case : in the examples mentioned above such disproportionation coincides with metal insulator transition . without going into details ,
let us shortly mention yet another specific phenomenon , which is typically observed close to metal insulator transitions .
this is the phenomenon of phase separation , which often occurs especially in doped mott insulators .
it is found experimentally in many cases ; in manganites @xcite ; in cobaltites @xcite ; and apparently also in cuprates @xcite . the tendency towards phase separation
is also seen theoretically in many models : in the hubbard @xcite and in the @xmath38@xmath30 models @xcite ; in the double - exchange models @xcite , etc .
the resulting state may be visualised as a random system with metallic regions in an insulating matrix .
the properties of such state can be described by percolation theory a classical description of random systems , although quantum effects in this case can also be very important .
typical for many frustrated systems , which contain triangles or tetrahedra as building blocks , is an extra degeneracy , which can be characterized by spin chirality .
thus the ground state of an equilateral triangle of @xmath13 spins with an antiferromagnetic interaction is a quartet , with the total spin @xmath56 and with extra double degeneracy , which can be characterized by the scalar spin chirality @xmath57\;. \label{eq:6}\ ] ] the physical meaning of the scalar chirality was clarified in @xcite ( see also @xcite ) , where it was shown that the states of a triangle with nonzero chirality ( [ eq:6 ] ) carry a spontaneous circular electric current @xmath58 .
correspondingly , such triangle would have an orbital magnetic moment @xmath59 .
in the nondegenerate hubbard model with strong interaction @xmath60 the expression for this current , obtained in perturbation theory in @xmath61 , has the form @xmath62)\;.\ ] ] in this limit the current and the corresponding orbital moment are small .
but one can show on general grounds , by using symmetry arguments , that such currents would be proportional to the chirality ( [ eq:6 ] ) in the general case , for @xmath63 , and also close to mott transition .
in such situations the corresponding current and orbital moments can be quite large .
indeed , estimates done in @xcite , using the exact solution for a triangle , show that for reasonable values of parameters the orbital magnetic moments can be as large as @xmath64 comparable with the spin moment .
the authors of @xcite suggested to use this degree of freedom for quantum computations .
this is yet another nontrivial property of frustrated systems , which can lead to large and potentially useful effects close to mott transitions .
in conclusion , i want to stress once again that the systems close to mott transition , or to a localized itinerant crossover , can show a variety of specific properties , different from those of states deep in the mott insulator regime or those of conventional metals .
especially interesting is the situation of a `` partial mott transition '' , in which first certain finite clusters go over to the regime with uncorrelated ( `` itinerant '' ) electrons , and only later there occurs a real mott transition in the whole sample .
as i tried to show on many examples , the most favourable conditions for this phenomenon are met in low - dimensional and in frustrated systems ; they are less probable in simple lattices such as perovskites . but also other specific phenomena , less sensitive to the type of the lattice , such as spontaneous charge disproportionation , can occur close to insulator metal transitions .
thus this crossover region between localized and itinerant electronic states presents a rich playground for interesting physics , with eventually some novel states appearing in this situation .
i am grateful to many people who contributed to works on which this paper is based , especially m. abd - elmeguid , c. batista , l. bulaevskii , hua wu , i. mazin , m. mostovoy , v. pardo , r. rivadulla , s. streltsov and a. ushakov .
this work was supported by the german program sfb 608 and by the european project soprano .
99 j. b. goodenough , _ magnetism and the chemical bond _ , wiley interscience , new york , 1963
|
recent studies demonstrated that there may appear different novel states in correlated systems close to localized itinerant crossover .
especially favourable conditions for that are met in low - dimensional and in frustrated systems . in this paper i discuss on concrete examples some of such novel states .
in particular , for some spinels and triangular systems there appears a `` partial mott transition '' , in which first some finite clusters ( dimers , trimes , tetramers , heptamers ) go over to the itinerant regime , and the real bulk mott transition occurs only later .
also some other specific possibilities in this crossover regime are shortly discussed , such as spin - peierls peierls transition in tiocl , spontaneous charge disproportionation in some cases , etc .
|
conductive polymers @xcite have received much attention due to its extensive industrial applications ( for a review , see ref .
@xcite ) , such as in nonlinear optical waveguides , vapor - phase detectors , twisted nematic liquid crystal displays , and so on . in fact
, the conductive organic polymers @xcite can be formulated with polymers that incorporate conductive additives such as carbon black @xcite or fine dispersed metallic particles @xcite .
this is because the presence of certain additives in the polymer can enhance the effective electrical conductivity . in particular , recently organic polymer composites filled with conductive metallic particles received much attention in scientific research @xcite .
for such polymer composites , their electrical characteristic are close to that of metals , whereas the mechanical property and processing method are typical for plastics @xcite .
the effective electrical conductivity of metal - polymer composite should depend on the conductivity of particles , the particle shape , the volume fraction of particles , and the spatial orientation or alignment of particles in the composite
. ferrogels @xcite are a new class of magneto - controlled elastic materials , which are chemically cross - linked polymer networks swollen with a ferrofluid .
a ferrofluid @xcite is a colloidal dispersion of monodomain ferromagnetic particles . in the ferrogel
, the finely distributed magnetic particles are located in the swelling liquid and attached to the flexible network chains by adhesive forces . in the absence of an applied field the magnetic moments are randomly oriented , and thus the gel has no net magnetization .
if an external field is applied , the magnetic moments tend to align with the field , thus yielding a bulk magnetic moment . in case of ordinary field strengths ,
the tendency of the dipole moments to align with the applied field is partially overcome by thermal agitation . as the strength of field increases
, all the particles eventually align their moments along the direction of the field , and as a result , the magnetization saturates .
if the field is turned off , the magnetic dipole moments quickly randomize and thus the bulk magnetization is again reduced to zero . in a zero magnetic field
a ferrogel presents a mechanical behavior very close to that of a swollen network filled with non - magnetic colloidal particles . in uniform magnetic fields a ferrogel experiences no net force , and hence no macroscopic shape changes and motion
are observed .
ferrogels have much application ranging from soft actuators , micromanipulators , and artificial muscles @xcite , to cancer therapy @xcite or as an apparatus for immunoblotting @xcite .
anisotropy is a common phenomenon in most materials , and it can be an intrinsic material property or induced by the application of fields . also , in real applications , the shape of particles may deviate from a perfect spherical shape during fabrication . in ferrogels , due to the anisotropic particle shape and/or conductivity of the ferromagnetic particles , the specific spatial orientation of the particles
can be realized by using an external magnetic field .
thus , the electric properties of ferrogels can be controlled by reorientation of the particles in an applied magnetic field . in this paper , we shall focus on an anisotropic ferrogel in which both the particle shape and conductivity possess a tensorial form .
our aim is to study the effects of the shape and/or material ( conductivity ) anisotropy on the effective electrical conductivity of the ferrogel , by taking into account the spatial orientation of the particles , which can be affected by the external magnetic field . to this end , we find both the shape anisotropy and the external magnetic field can cause the effective electrical conductivity to increase .
furthermore , the material ( conductivity ) anisotropy can increase the effective electrical conductivity significantly . this paper is organized as follows . in sec .
ii , we put forth a general model for a ferrogel containing ellipsoidal ferromagnetic particles with anisotropic conductivity . in sec .
iii , we numerically calculate two cases , namely , spherical particles with anisotropic conductivity , and spheroidal particles with isotropic conductivity , in an attempt to focus on the influence of the shape and material anisotropy , respectively .
this paper ends with a discussion and conclusion in sec .
let us start by considering an ellipsoidal particle with depolarization factors @xmath0 @xmath1 @xmath2 along @xmath3 @xmath4 @xmath5 axes , respectively .
namely , the depolarization factor has a tensor form = ( ccc l_x & 0 & 0 + 0 & l_y & 0 + 0 & 0 & l_z ) .
the three components satisfy a sum rule , @xmath6 @xcite . in the case of crystalline anisotropic
, the conductivity of the particle has a tensor form as @xmath7 the particles are embedded in a host with a scalar conductivity @xmath8 in this paper , the principle axes of @xmath9 are assumed to be parallel to the geometric axes of the ellipsoidal particle .
the orientation of the conductivity tensor @xmath9 differs from inclusion to inclusion , such that @xmath9 is transformed to _ 1 = _ 1^t in a common coordinate system , where @xmath10 denotes the rotation matrix of a particle . for the volume average of the field of the whole system @xmath11 , one has @xcite p_1 + ( 1-p ) _
l= , [ ae0 ] where @xmath12 stands for the volume fraction of particles , and @xmath11 is equal to the applied field @xmath13 under appropriate boundary conditions @xcite . here
@xmath14 denotes the volume average of @xmath15 @xmath16 represents the volume average of the lorentz field ( namely , local field in the vicinity of the particles ) which includes the contribution from the dipole moments of all the other particles , and @xmath17 represents the volume average of the local field inside the particles .
thus , we can rewrite eq .
( [ ae0 ] ) as @xcite p_1 + ( 1-p ) _ l= _ 0,[ae ] solving the electrostatic equation @xmath18 ( @xmath19 denotes the electrical potential inside the particles ) , we obtain @xmath17 such that @xcite _ 1= _ l[ae1 ] with = ^t . here
= ( ccc & 0 & 0 + 0 & & 0 + 0 & 0 & ) .
on the other hand , we average the electric current density @xmath20 over the volume of the whole system , and obtain = p_1_1 + ( 1-p)_2 _ l.[aj ] further , the effective electrical conductivity @xmath21 of the ferrogel can be defined as _
e = = . in view of eqs .
( [ ae ] ) , ( [ ae1 ] ) and ( [ aj ] ) , we obtain _ e = , [ eff ] which can be expressed in a tensorial form such that _ e = ( ccc _ e11 & 0 & 0 + 0 & _ e22 & 0 + 0 & 0 & _
e33 ) . in eq .
( [ eff ] ) , @xmath22 represents a unit matrix .
( [ eff ] ) is a nontrivial equation , which can reproduce many other known results .
for instance , setting @xmath23 and @xmath24 , and seeing the particles _ randomly _ oriented in a host , @xmath25 obtained from eq .
( [ eff ] ) is identical to eq .
( 14 ) of ref .
@xcite . in this case , if the metallic particles are subjected to an external magnetic field ( i.e. , no longer randomly oriented ) , @xmath26 and @xmath27 predicted by eq .
( [ eff ] ) are identical to eqs .
( 3 ) and ( 4 ) of ref . @xcite , respectively . finally , for spherical particles ( namely , @xmath28 ) with a scalar conductivity ( i.e. , @xmath24 ) , eq .
( [ eff ] ) can produce the expression for the well - known maxwell garnett theory @xcite , namely _
e = _ 2 + .
we are now in a position to do some numerical calculations .
we shall discuss two cases : ( case i ) spherical particle with tensorial conductivity ; ( case ii ) ellipsoidal particle with scalar conductivity .
in doing so , we can focus on the effect of shape or material ( conductivity ) anisotropies , respectively .
to start , we have to derive the expression for @xmath29 . for @xmath23 and @xmath30 , @xmath29 can be obtained as @xcite = ( ccc [ _ 11+_33-(_33-_11 ) ] & 0 & 0 + 0 & [ _ 22+_33-(_33-_22 ) ] & 0 + 0 & 0 & _
11+(_33-_11 ) ) , where @xmath31 with @xmath32 being the polar orientation angel of a particle , @xmath33 . since the orientation of the ferromagnetic particles in ferrogels can be affected by the external magnetic field , there is @xmath34 @xcite . here
@xmath35 denotes the langevin function and @xmath36 the langevin parameter , where @xmath37 represents the magnetic moment of a particle , and @xmath38 the magnetic field , @xmath39 the boltzmann constant , and @xmath40 the temperature . for the sake of convenience , @xmath41 will be used to denote the magnetic field in the following numerical calculations . in a word , @xmath42 contains the information of the spatial orientation of the particles in the host . in particular , the zero field ( @xmath43 ) yields @xmath44 naturally due to the random orientation of the particles . as @xmath45 @xmath42 deviates from @xmath46 accordingly . in other words , the specific spatial orientation of the particles in the ferrogel
can be achieved by adjusting the external magnetic field . in this case
, there are @xmath47 and @xmath30 . in fig . 1 , we display the effective electrical conductivity versus the langevin parameter for different conductivity ratio @xmath48 .
in fact , @xmath49 represents the degree of anisotropy of conductivity .
we find increasing the langevin parameter @xmath41 causes @xmath25 to decrease , see fig .
1(a)-(b ) .
in contrast , in fig .
1(c)-(d ) , as @xmath41 increases , @xmath27 increases first , and decreases after a peak .
in particular , at @xmath50 there is always a peak for @xmath51 $ ] . in detail , for @xmath52 , @xmath53 , and @xmath54 , the corresponding peak values are @xmath55 , @xmath56 , and @xmath57 , respectively . in other words ,
an optimal magnetic field exists which leads to a maximum conductivity @xmath27 .
it is worth noting that both @xmath58 and @xmath59 ( @xmath60 is a number ) can predict the same @xmath25 or @xmath27 concerned , see fig .
1(b ) and ( d ) .
this behavior keeps unchanged up to @xmath61 or so .
moreover , larger @xmath60 and hence a stronger conductivity anisotropy can lead to a higher effective electrical conductivity @xmath25 or @xmath27 . for this case , @xmath23 , and @xmath24 .
2 shows the effective electrical conductivity against the langevin parameter @xmath41 for different depolarization factor @xmath2 . for prolate spheroidal particles ( @xmath62 )
, smaller @xmath2 can lead to higher @xmath25 or @xmath27 , see fig . 2 .
in contrast , for oblate spheroidal particles ( @xmath63 ) , larger @xmath2 can yield higher @xmath25 or lower @xmath27 .
also , for the prolate spheroidal cases , increasing the magnetic field causes @xmath25 ( or @xmath27 ) to decrease ( or increase ) .
however , for the oblate spheroidal cases , as the magnetic field increases , @xmath25 increases , but @xmath27 decreases . in a word , the stronger shape anisotropy ( namely ,
how @xmath2 deviates from @xmath46 ) can yield a higher effective electrical conductivity . from cases
i and ii , it is concluded that material ( conductivity ) anisotropies are more important to yield higher effective electrical conductivities of ferrogels than shape anisotropies , while magnetic fields can offer a correction .
in addition , we also discussed the effect of the volume fraction of particles ( no figures shown here ) .
it is shown that increasing the volume fraction of particles can cause the effective electrical conductivity to increase accordingly .
this result is the same as that obtained by xue @xcite .
here some comments are in order . in this paper , we have presented a general model for a ferrogel in which both the particle shape and conductivity have a tensorial form , in an attempt to study the effect of the shape and/or material ( conductivity ) anisotropy on the effective electrical conductivity of the ferrogel , by taking into account the spatial orientation of the particles in the presence of an external magnetic field .
our theory reproduced several known results , and provides a link between the particle property and orientation distribution and the effective electrical conductivity of ferrogels . for water ,
the minerals like sodium chloride are often dissolved in it , thus making it have a conductivity .
more importantly , as one tries to make pure water by gradually removing electrolytes , its conductivity gradually decreases indeed .
however , if all electrolytes are removed , its conductivity is still nonzero .
the reason is that an infinitesimal part of the molecules of water only about one in 500 million is ionized as hydrogen ions ( h@xmath64 ) and hydroxide ions ( oh@xmath65 ) .
theoretically , at this point , the conductivity becomes @xmath66s / m at @xmath67@xmath68c .
in contrast , the conductivity of polymers could be @xmath69s / m . for the ferrogel discussed in this paper
, we see the ferromagnetic particles to be embedded in the host medium which is composed of polymers as well as water . since the structure and magnetization property of a polydisperse ferrogel can differ from that of a monodisperse system @xcite .
it is instructive to extend the present work to polydisperse case , so that one could investigate the polydisperse effect on the effective electrical conductivity of polydisperse ferrogels . in this work , we derived an anisotropic maxwell - garnett formula [ eq .
( [ eff ] ) ] , which is nonsymmetrical and may thus be suitable for low concentrations .
for a higher concentration of particles , we can derive an alternative anisotropic bruggeman formula ( see appendix ) , which is symmetrical .
similar results should be obtained by means of the anisotropic bruggeman formula .
however , the bruggeman formula can predict a percolation threshold .
it is also interesting to see what happens if one extend the present model to investigate the electro - optical effects @xcite , nonlinearity enhancement @xcite , figure of merit @xcite , and nonlinear ac responses @xcite of ferrogels due to the presence of the ( weak ) nonlinearity inside the metallic particles . in doing so , the particle reorientation arising from external magnetic fields
is expected to play a role as well .
in addition , due to the analogy of the mathematical form between conductivities and dielectric constants of composite materials , it is straightforward to apply the present theory to the effective dielectric constant of ferrogels . to sum up , based on our general model , it was shown that material ( conductivity ) anisotropies are more important to produce a high effective electrical conductivity of ferrogels than shape anisotropies , while magnetic fields can offer a correction due to the change in orientation distributions of particles .
this work was supported by the deutsche forschungsgemeinschaft ( german research foundation ) under grant no .
ho 1108/8 - 3 . i would like to thank dr . c. holm and professor h. pleiner for their critical comments , and professor k. w. yu for fruitful discussions .
to derive the anisotropic bruggeman formula , let us start from the effective dipole factor of the two components in the system .
for the particles , the dipole factor @xmath70 is _ 1 = ( ) .
on the other hand , the host have a dipole factor @xmath71 such that _ 2 = .
now we can derive the bruggeman formula by considering the fact that the effective dipole factor @xmath72 of the whole system should be zero , and then obtain p_1 + ( 1-p)_2 = 0,[bru ] where @xmath73 . in case of @xmath28 and @xmath24 ( scalar conductivity ) , eq .
( [ bru ] ) reduces to the well - know ( isotropic ) bruggeman formula p + ( 1-p ) = 0 .
|
conductive organic polymers can be formulated with polymers that incorporate fine dispersed metallic particles . in this work ,
we present a general model for ferrogels which are chemically cross - linked polymer networks swollen with a ferrofluid .
our aim is to study the effect of the shape and/or material ( conductivity ) anisotropy on the effective electrical conductivity of the ferrogel in the presence of an external magnetic field .
our theory can reproduce the known results , and provides a link between the particle property and orientation distribution and the effective electrical conductivity . to this end
, we find that material ( conductivity ) anisotropies are more important to yield a high effective electrical conductivity than shape anisotropies , while magnetic fields can offer a correction .
|
ballistic conductors are characterized by an active region where carriers , once injected by contacts , move without suffering any scattering from contact to contact , i.e. the carrier mean free path is much longer than the sample characteristic dimensions . in metals ,
ballistic transport is usually studied in point contacts.@xcite such contacts allow the analysis of elastic and inelastic scattering processes by means of the so called point contact spectroscopy.@xcite since the fermi wavelength in metals is very small ( @xmath1 ) , the nature of carrier transport is semiclassical and quantum effects related to the wave nature of the electrons can be disregarded . in semiconductors , ballistic transport has been investigated in bulk materials,@xcite point contacts@xcite and two dimensional electron gases;@xcite some studies have allowed the development of ballistic emission spectroscopy@xcite for the analysis of semiconductor heterointerfaces .
moreover , a variety of ballistic electron devices with promising performances have been realized@xcite or proposed.@xcite since the fermi wave length in semiconductors can be as large as 40 @xmath2 , the nature of the carrier transport can be either semiclassical or quantum depending on the characteristic sample dimensions , @xmath3 , with respect to the fermi wave length , @xmath4 . for @xmath5 transport
is semiclassical , while for @xmath6 transport is quantum . from a fundamental point of view
, the study of non - equilibrium electronic noise ( shot noise)@xcite of ballistic conductors in the semiclassical transport regime offers a unique scenario where the simultaneous effect of two fundamental physical interactions , namely the long range coulomb interaction and the pauli exclusion principle , can be investigated . on the one hand ,
pauli correlations have an influence on the carrier injecting statistics at the contacts . on the other hand
, coulomb correlations can modify the carrier passing statistics inside the active regions .
the effect of both mechanisms results in values of the current noise below the full poissonian value , i.e. in shot - noise suppression .
thus , the non - equilibrium low frequency current spectral density is given by @xmath7 , with @xmath8 being the current flowing trough the sample and @xmath9 a dimensionless parameter ( fano factor ) which takes values below unity because of the negative correlations induced by both mechanisms .
the relative relevance of coulomb and pauli suppression in determining the fano factor depends on parameters like the sample length , the temperature , the carrier density or the applied voltage , thus allowing the monitoring of both correlation mechanisms .
the first theoretical analyses on the non - equilibrium noise properties of ballistic conductors were performed by van der ziel and bosman in @xmath10 ballistic semiconductor diodes@xcite and by kulik and omelyanchuk in ballistic metallic point contacts.@xcite in the former case , by considering non - degenerate injection conditions , only the effect of the coulomb correlations was evidenced . in particular , by using ideas borrowed from north s theory of vacuum
tubes,@xcite it was shown that under space charge conditions current noise results suppressed below the full poissonian value ( see also refs . [ ] ) . in the later case ,
the high carrier density of metals prevented the presence of significant space charge effects , and hence the possibility to test the effect of the long range coulomb correlations . in this case , by means of a selfconsistent semiclassical kinetic theory based on the boltzmann - langevin equation,@xcite which included the effects of the pauli principle , it was shown that in the collision free regime ballistic metallic point contacts displayed nyquist noise with low frequency current spectral density given by @xmath11 , where @xmath12 is the boltzmann constant , @xmath13 the temperature and @xmath14 the sharvin contact resistance , with @xmath15 being the electron effective mass , @xmath16 the fermi velocity , @xmath17 the unit charge , and @xmath18 the carrier concentration inside the sample .
this result is applied in the physical limit where @xmath19 , @xmath20 , and arbitrary relation between @xmath19 and @xmath21 , where @xmath22 is the applied bias and @xmath23 the fermi energy .
these initial studies did not allowed the study of the simultaneous effect of both pauli and coulomb correlation mechanisms .
it was only very recently that ballistic structures including the simultaneous effect of pauli and coulomb correlations have been analyzed.@xcite by making use of a semi - analytical approach,@xcite monte carlo ( mc ) simulations,@xcite and an asymptotic analytical approach,@xcite non - equilibrium noise properties of ballistic semiconductor structures have been investigated .
it was shown that these structures , by allowing both space charge effects and degenerate injection statistics , were unique in order to study the simultaneous effect of both pauli and coulomb correlations.@xcite however , the existing studies , being only valid under some limit conditions or performed for some specific sets of system parameters , do not offer a complete overview of the relative relevance of pauli and coulomb correlations in the whole range of system parameters and applied bias .
this lack of a general overview can prevent the designing of suitable experimental strategies to test the theoretical predictions .
the aim of the present work is precisely to address this issue . to this purpose
we work out a general theory for the low frequency shot - noise properties of ballistic @xmath0 semiconductor diodes that consistently describes also current - voltage characteristics .
we consider ligthly doped bulk semiconductor materials since at low temperatures the bulk electrons are trapped by their parent donors , neutralizing them and thus minimizing both the @xmath24 interaction and ionized impurity scattering , thus allowing long mean free paths up to relatively high electron energies.@xcite the theory presented accounts for both the pauli exclusion principle and the long range coulomb interaction , and is applied to the whole range of system parameters ( sample length , temperature , contact doping ) and external bias .
in particular , the theory allows us to study in a unifying framework the transition from : ( i ) nondegenerate to degenerate injection conditions , ( ii ) short to ( asymptotically ) long sample lengths and , ( iii ) low to ( asymptotically ) high applied voltages . by overcoming the limitations of the previous existing theories ,
the present study allows us to compress into a general scheme the full scenario displayed by the pauli and coulomb correlations in these ballistic structures .
this scheme , validated and tested with a wide set of mc simulations , is believed to be of relevant assistance in designing future investigations on an experimental and/or simulation basis .
the paper is organized as follows . in sec .
[ system ] we describe the system under study . in sec . [ model ] we present the physical model used to analyze the non - equilibrium noise properties . in sec .
[ solution ] the analytical expressions obtained for the transport and noise properties are reported , and their validity is checked by means of mc simulations . in sec .
[ general ] we propose a general scheme able to provide a systematic physical picture of the shot - noise properties of two - terminal ballistic conductors , and check its reliability by means of the developed theory . finally , in sec .
[ conclusions ] we summarize the main results of the paper .
the appendix is devoted to technical derivations .
the system under study is the @xmath0 ballistic semiconductor diode showed in fig .
[ figsystem ] .
it consists of an undoped semiconductor of length @xmath3 , where injected carriers move in the total absence of scattering events , sandwiched between two highly doped regions which act as ideal injecting contacts as specified later .
the contacts and the ballistic region are taken of the same material ( homodiode ) or of different material ( heterodiode ) . in the former case the band offset @xmath25 @xmath26 ( a symmetric structure
is assumed for simplicity ) vanishes , while in the later case it takes a finite value .
the ballistic region is taken to be perfectly coupled with the contacts , thus no reflections take place at the interfaces .
the contacts are assumed to be in quasiequilibrium at the given temperature , and at electrochemical potentials @xmath27 and @xmath28 , with @xmath29 .
the voltage drop in the contacts is assumed to be negligibly small ( ohmic contacts ) , and hence all the band bending occurs in the active region of the sample . under these assumptions ,
the effective contact chemical potential @xmath30 , defined as @xmath31 is independent of the applied bias . note that for the case of an homodiode @xmath30 coincides with the contact fermi energy , @xmath32 , while for the case of an heterodiode one has @xmath33 .
we note the existence of a maximum with amplitude @xmath34 in the potential energy along the active region due to the presence of space charge .
the structure is assumed to be sufficiently thick in the transversal directions so as to allow for a 1d electrostatic treatment .
thus the system is 1d in real space and 3d in momentum space.@xcite the carriers are injected into the active region in accordance with the equilibrium fermi - dirac distribution function and satisfying the pauli exclusion principle , thus following the corresponding binomial injection statistics.@xcite for simplicity , a single spherical parabolic band is assumed .
the parameters describing the model system introduced above are therefore the length of the conductor , @xmath3 , the cross sectional area @xmath35 , the effective contact chemical potential , @xmath30 , the temperature , @xmath13 , the static dielectric constant of the medium , @xmath36 , the electron effective mass , @xmath15 , and the charge carriers @xmath17 .
the independent - variable parameter is the applied voltage , @xmath22 . as mentioned above , the correlations between different current pulses in the present system is twofold .
on the one hand , pauli correlations have an influence on the contact injecting statistics .
carriers above the fermi energy , at the tail of the fermi distribution at the contacts , are injected obeying poissonian statistics due to the negligible influence of the pauli principle at such high energies .
carriers near and below the fermi energy are injected following a binomial ( sub - poissonian ) statistics due to the increasing influence of the pauli principle at the highly occupied states . on the other hand , coulomb correlations , through the potential energy maximum @xmath34 associated with the presence of space charge , can modify the carrier passing statistics inside the active region . indeed , the fluctuations of @xmath34 originated by the passage of carriers over the maximum modulate the transmission of further carriers ( spacing them out more regularly in time ) and smooth out the current fluctuations imposed by the random injection at the contacts .
the relative importance of the two types of correlating mechanisms depends on the values of @xmath3 , @xmath30 and @xmath22 .
within a semiclassical approach the description of the transport and noise properties of the system under study can be carried out by means of the vlasov equation , self - consistently coupled to the poisson equation , and supplemented by appropriate _ fluctuating _
boundary conditions.@xcite in the one dimensional approximation followed here , the vlasov - poisson system of equations reads ( in dimensionless units , as specified later ) : @xmath37 f(x , v_{x},t)=0\text { , } \label{vlasov}\ ] ] @xmath38 where @xmath39 is the time variable , @xmath40 the spatial coordinate , @xmath41 the velocity in the @xmath40 direction , @xmath42 the potential energy , @xmath43 the distribution function integrated over transversal momentum directions and @xmath44 the carrier density . in the poisson equation
the contribution of the intrinsic free carrier density is neglected in comparison to the injected carriers . moreover ,
the boundary conditions for the potential energy read @xmath45 where we have assumed that the applied bias is fixed by a low impedance external circuit so that neither the electric potential nor the contact potential energies fluctuates , i.e. , @xmath46 .
accordingly , the boundary conditions for the distribution function at the contacts read : @xmath47 these boundary conditions consist of two contributions .
the first is of deterministic nature , and gives the average value of the distribution function .
the second is of stochastic nature , and describes the random injection of carriers .
the average distribution function is given by : @xmath48 with @xmath49 and @xmath50 being the longitudinal and transversal kinetic energy , respectively , and where @xmath51 is the fermi - dirac distribution function .
note that due to the integration over the transverse momentum directions the effective one dimensional contact distribution function is @xmath52 instead of @xmath53 .
moreover , the fluctuating contributions @xmath54 and @xmath55 have zero mean and low frequency spectral density given by@xcite @xmath56 equations ( [ vlasov])-([fccorr ] ) constitute the complete set of equations to study the noise properties of the ballistic structure described in sec . [ system ] in the previous set of equations , and in what follows , we use dimensionless variables to simplify the notation .
the dimensionalizing factors used for energy , length , carrier density , velocity , distribution function , electric potential , electric field , electric current , resistance and current spectral density are , respectively , @xmath57 where @xmath58 is the planck constant . with these dimensionalizing factors the dimensionless parameters describing
the system are @xmath59 @xmath3 and @xmath60 , and the dimensionless independent variable is @xmath61 .
the physical meaning of the different parameters is the following : @xmath62 is the effective density of states in the conduction band , @xmath63 is the debye screening length associated to @xmath64 , @xmath65 is the thermal energy , @xmath66 is the thermal velocity , @xmath67 is the thermal voltage , @xmath68 the thermal electric field , @xmath69 the current associated to @xmath62 and @xmath66 , @xmath70 the resistance associated to @xmath71 and @xmath72 the shot noise level associated to @xmath69 . to complete our analysis , we will perform mc simulations of the system under study , so that the validity of the analytical theory can be checked by comparison . to this end
we use an ensemble mc simulator , 3d in momentum space and self - consistently coupled with a 1d poisson solver to account for coulomb interaction .
the carrier dynamics is simulated in the ballistic active region of the structure and the electron injection from the thermal reservoirs is modeled according to fermi statistics . due to pauli principle
, the instantaneous occupancy of an incoming electron state with energy @xmath73 and impinging at the interface between the ideal thermal reservoir and the active region fluctuates in time obeying a binomial distribution@xcite with a probability of success given by @xmath74 .
this statistics is implemented in the mc simulation of the contact injection by introducing a discretization of momentum space and using the rejection technique to select the times of injection at every momentum state.@xcite as limiting cases , when @xmath75 , @xmath76 and the injection statistics of the corresponding state is uniform in time . by contrast , when @xmath77 , @xmath78 and the injection statistics is poissonian . for the calculations we use the following parameters : @xmath79 k , @xmath80 , @xmath81 , with @xmath82 the free electron mass and @xmath83 the vacuum permittivity .
the low frequency solution of the model presented in sec . [ model ] can be obtained in fully analytical form . for the sake of conciseness , below we only present the final expression of the relevant quantities of interest and refer to the appendix [ appendix ] for the details of the derivation . following the results presented in appendix [ appendix ] , the current voltage ( @xmath84 ) characteristics of the ballistic conductor presented in sec .
[ system ] can be calculated as @xmath85 \text { , } \label{ivav}\ ] ] where @xmath86 is given in eq .
( [ fbcav ] ) , @xmath87 is the average value of the maximum potential energy at applied bias @xmath22 , and @xmath88 and @xmath89 are the electrochemical potentials at the contacts located at @xmath90 and @xmath91 , respectively .
the value of @xmath92 can be calculated through the following equation ( see appendix [ appendix ] ) @xmath93 where @xmath94 \left [ f_{c}(u-\xi _ { l})+f_{c}(u-\xi _ { 0})\right ] \right .
\nonumber \\ & & \left .
+ \int_{u}^{\overline{u}_{m}}du\sqrt{u - u}2f_{c}(u-\xi _ { 0})\right\ } ^{\frac{1}{2}}2^{\frac{3}{4 } } \label{e - av}\end{aligned}\ ] ] and @xmath95 \left [ f_{c}(u-\xi _ { l})+f_{c}(u-\xi _ { 0})\right ] \right
. \nonumber \\ & & \left .
+ \int_{u}^{\overline{u}_{m}}du\sqrt{u - u}2f_{c}(u-\xi _ { l})\right\ } ^{\frac{1}{2}}2^{\frac{3}{4}}\text { . } \label{e+av}\end{aligned}\ ] ] figure [ figumv ] reports the maximum potential energy as a function of the voltage , as calculated from eq .
( [ umav ] ) with @xmath96 at different values of the sample length .
mc results are also shown .
an excellent agreement is found between analytical and mc calculations . at vanishing voltages @xmath87 saturates at higher values the longer
is the normalized length , because of the increasing effect of space charge inside the ballistic region . at increasing voltages
@xmath87 starts decreasing till vanishing at a threshold voltage @xmath97 .
figure [ figiv ] shows the @xmath84 characteristics as calculated from eq .
( [ ivav ] ) for @xmath96 and several sample lengths .
the figure also reports the results of mc simulations .
we have found perfect agreement between the results of the theory and those of mc simulations as expected , since both results are exact solutions of the same physical model .
the main features of the @xmath84 characteristics consist in the presence of a linear ( ohmic ) behavior at low applied voltages and a current saturation regime at high voltages .
the resistance of the linear regime is given by ( in dimensionless units ) @xmath98 in general , this value depends on both the sample length @xmath3 and the reduced chemical potential @xmath30 . when @xmath99 the resistance takes a value independent of the sample length @xmath100 in agreement with the results of ref .
typically this happens for @xmath101 and @xmath102 , which represent the conditions for negligible space charge inside the ballistic region . in this limit , and when @xmath103 , @xmath104 tends to the sharvin contact resistance @xmath105 .
the value of the saturation current is given by ( in dimensionless units ) : @xmath106 current saturation takes place when all carriers injected from one contact reach the opposite one , while none of the carriers injected from the other contact is able to cross the ballistic region .
the value of the saturation current is independent of the sample length since it is only determined by the emission properties of the contacts .
the voltage for the onset of current saturation , @xmath107 , depends on both @xmath3 and @xmath30 , as can be seen in fig .
[ figiv ] . in the case that @xmath108 , where @xmath97 is the bias value at which the maximum of the potential energy disappears ( see fig .
[ figumv ] ) , one has @xmath109 . otherwise , when @xmath110 , one has @xmath111 , and saturation coincides with the disappearance of the maximum of the potential energy . between the linear and the current saturation regimes the @xmath84 characteristic displays a nonlinear region , whose properties
are determined by the precise values of @xmath3 and @xmath112 .
the asymptotic behavior of this non - linear region for large values of @xmath3 in the limit when @xmath113 consists of a slightly sublinear region followed by a superlinear characteristic of child - langmuir type @xcite @xmath114 @xmath115 , as has been found in ref .
the sublinear region is associated with a non zero temperature value which smooths out the fermi distribution .
the superlinear region is associated with space - charge effects driven by the presence of @xmath34 .
the steady state profiles can also be calculated in closed analytical form .
the potential energy profile can be obtained in inverse form from the following relations ( see appendix [ appendix ] ) : @xmath116 where @xmath117 is the location of the potential energy maximum inside the ballistic region . the value of @xmath117 can be calculated from either eq .
( [ uxav- ] ) or eq .
( [ uxav+ ] ) by setting the value of @xmath118 equal to @xmath87 .
once the potential energy profile is obtained , the electric field profile can be calculated as ( see appendix [ appendix ] ) @xmath119 where @xmath120 and @xmath121 are given in eqs .
( [ e - av ] ) and ( [ e+av ] ) , respectively .
finally , the carrier density profile is obtained from ( see appendix [ appendix ] ) @xmath122 + \\
\int_{\overline{u}^{-}(x)}^{\overline{u}_{m}}\frac{du}{\sqrt{2\left ( u-% \overline{u}^{-}(x)\right ) } } 2f_{c}(u-\xi _ { 0})\text{;\quad } 0<x<\overline{x}% _ { m } \\
\int_{\overline{u}_{m}}^{+\infty } \frac{du}{\sqrt{2\left ( u-\overline{u}% ^{+}(x)\right ) } } \left [ f_{c}(u-\xi _ { 0})+f_{c}(u-\xi _ { l})\right ] + \\ \int_{\overline{u}^{+}(x)}^{\overline{u}_{m}}\frac{du}{\sqrt{2\left ( u-% \overline{u}^{+}(x)\right ) } } 2f_{c}(u-\xi _ { l})\text{;\quad } \overline{x}% _ { m}<x < l \end{array } \right . \text{. } \label{nxav2}\end{aligned}\ ] ] figure [ figprofiles ] reports the steady state profiles for @xmath96 and @xmath123 , at several values of the applied bias . for the sake of comparison
, the same figure also reports the profiles obtained from mc simulations .
again analytical calculations are found to perfectly agree with the simulations .
the figure clearly illustrates the space - charge nature of the transport and the presence of the potential energy maximum .
the value of the potential energy maximum decreases systematically at increasing bias ( as already indicated in fig .
[ figumv ] ) , while its location shifts towards the left contact . at the same time , the free charge redistributes inside the structure by shifting the minimum of the carrier density from the center of the sample towards the right contact .
the low frequency noise properties of the system under study are characterized by the low frequency spectral density of current fluctuations defined as ( in dimensionless units ) , @xmath124 when interested in the study of the microscopic correlations the proper figure of merit is the fano factor @xmath9 , which is obtained from @xmath125 as ( in the dimensionless units ) @xmath126 for conditions far from thermal equilibrium ( @xmath127 ) a value of @xmath128 corresponds to the absence of correlations between different current pulses , while @xmath129 ( @xmath130 ) corresponds to the presence of negative ( positive ) correlations .
as shown in appendix [ appendix ] , @xmath131 can be calculated in closed analytical form for the model presented in sec .
[ model ] .
the final result can be written in the following compact form ( see appendix [ appendix ] ) @xmath132 ^{2}f_{fd}(u-\xi _
{ l})+ \label{siav } \\ & & \int_{\overline{u}_{0}}^{+\infty } du\left [ \gamma _ { 0}(u)\right ] ^{2}f_{fd}(u-\xi _ { 0})\text { , } \nonumber\end{aligned}\ ] ] where ( see appendix [ appendix ] ) @xmath133 here , we have defined @xmath134 } { \delta } \text { , } \label{omega}\ ] ] with @xmath135 \text { . }
\label{deltaav}\end{aligned}\ ] ] moreover , @xmath136 from the previous equations we can evaluate @xmath131 and , in turn , the fano factor @xmath9 .
note that these analytical expressions are valid in the whole range of system parameters @xmath3 and @xmath30 , and in the whole range of applied bias @xmath22 .
for the purpose of a reliability test , the results obtained from the analytical formulae are compared with those of mc simulations in fig .
[ figsi_v ] . here
, the low frequency current spectral density is reported as a function of applied bias for @xmath96 and several sample lengths . as can be seen in the figure ,
the agreement between the analytical theory and mc simulations is excellent , thus proving the reliability of both theory and simulations . for the sake of completeness fig .
[ figsi_v ] also reports the results obtained by means of the asymptotic theory developed in ref .
[ ] . according to this theory , in the limit of @xmath137 the low frequency current spectral density can be approximated by ( in dimensionless variables ) @xmath138 with @xmath139 ^{2}}{f_{0}(\alpha ) f_{1}(\alpha ) } \right ) \text { , } \label{beta}\ ] ] where @xmath140 , with @xmath141 being the gamma function . as seen in fig .
[ figsi_v ] the asymptotic theory agrees well with the present theory for @xmath142 .
however , it can not describe either the transition between low and high bias regions , or the case of small sample lengths in which the condition @xmath113 can not be satisfied .
the main features of the low frequency spectral density depicted in fig .
[ figsi_v ] are summarized as follows . at low voltages ( @xmath143 ) , where ohmic conditions are satisfied
, @xmath131 is bias independent and takes the value ( in dimensionless variables ) @xmath144 in agreement with nyquist theorem.@xcite at high voltages ( @xmath145 ) , when the sample exhibits current saturation conditions , @xmath146 is again bias independent and given by ( in dimensionless variables ) @xmath147 in the region of intermediate values of voltages , @xmath131 can present a monotonic or a non - monotonic behavior ( with the presence of a minimum ) determined by the interplay between coulomb and pauli correlations , as will be detailed in the next section .
the present theory enables us to investigate separately the relevance of the two mechanisms responsible for the correlations in the system under study , namely , the pauli exclusion principle and the long range coulomb interaction , in the whole range of system parameters .
accordingly , we propose a general scheme which summarizes the whole scenario of the shot - noise properties exhibited by @xmath0 ballistic semiconductor structures .
to construct such a scheme , we make use of the fano factor , @xmath148 , which is factorized into the two independent contributions , @xmath149 and @xmath150 , related to the pauli and coulomb correlations , respectively . indeed , according to eq .
( [ fccorr ] ) , the fluctuations of the contact distribution function at different energy levels are uncorrelated , so that the only source of correlations among carriers injected with different energy is the coulomb interaction in the active region . as a consequence , both contributions to the fano factor are independent , which implies @xmath151 .
thus , the pauli contribution @xmath152 corresponds to the fano factor that would be obtained in the absence of the self - consistent long range coulomb interaction . it can be easily evaluated from the noise calculation performed in the appendix [ appendix ] by neglecting the self - consistent contribution .
this is equivalent to set up in eq .
( [ siav ] ) , @xmath153 for the current spectral density associated with pauli correlations only , it is thus obtained @xmath154 so that the pauli contribution to the fano factor is found to be @xmath155 } \text { , % } \label{fpauli}\ ] ] where use is made of eq .
( [ ivav ] ) . the coulomb contribution
is then evaluated as @xmath156 . according to these definitions , for @xmath127 values of @xmath157 correspond to the presence of pauli correlations , while values of @xmath158 correspond to the presence of coulomb correlations . a detailed analysis of eq . ( [ fpauli ] )
indicates that the pauli contribution is essentially dependent on the difference @xmath159 .
hence , when @xmath160 it is @xmath161 , thus indicating the absence of pauli correlations .
in particular , for nondegenerate injection conditions @xmath162 , one always has that @xmath160 , hence indicating the absence of pauli correlations , as should be .
by contrast , when @xmath163 it is @xmath164 , thus indicating the presence of pauli correlations . since @xmath87 is a decreasing function of the bias , the condition @xmath165 implies automatically that the inequality @xmath166 is satisfied for all bias values , and hence the presence of pauli correlations for @xmath127 .
on the contrary , when @xmath167 pauli correlations are absent for low ( or intermediate ) bias values and present for bias values sufficiently high to validate the condition @xmath163 . concerning coulomb correlations , their presence or absence is roughly determined by a value of the ratio @xmath168 higher or lower than unity , respectively , where @xmath169 is the debye screening length corresponding to an homogeneous system with charge density equal to the equilibrium contact density .
taking into account the effects of degenerancy , @xmath169 is calculated as ( in dimensionless variables ) @xmath170 where we used that from eq .
( [ nxav2 ] ) one has @xmath171 on the basis of these considerations , for the scenario of the shot - noise properties in ballistic conductors we propose the general scheme displayed in fig .
[ figdiagr ] . in this scheme
we identify five different regions in the plane ( @xmath3 , @xmath30 ) , corresponding to five different possibility of interplay between coulomb and pauli correlations .
the different regions in fig . [ figdiagr ] are determined by the three lines defined by the equalities @xmath172 , @xmath173 and @xmath174 .
the shot noise behavior in each of the five regions corresponds , respectively , to : ( i ) the absence of both pauli and coulomb correlations ( @xmath175 and @xmath176 ) ; ( ii ) the presence of pauli correlations and the absence of coulomb correlations ( @xmath177 , @xmath165 and @xmath176 ) ; ( iii ) the absence of pauli correlations and the presence of coulomb correlations ( @xmath162 and @xmath178 ) ; ( iv ) the absence ( presence ) of pauli correlations for low ( high ) bias and the presence of coulomb correlations ( @xmath177 , @xmath179 and @xmath178 ) ; ( v ) the presence of both pauli and coulomb correlations ( @xmath177 , @xmath165 and @xmath178 ) .
the reliability of the above scheme has been tested by performing a series of theoretical calculations for the relevant regions identified above .
we have found that the proposed scheme is essentially valid , except in the zones close to the lines separating the different regions , where intermediate behaviors have been observed .
representative examples concerning the fano factor , @xmath9 , and the contributions into which it is decomposed , @xmath152 and @xmath150 , are shown in figs .
[ figgl5mu_5 ] to [ figgl2mu8 ] for each of the five regions individuated in the @xmath3-@xmath30 plane .
figure [ figgl5mu_5 ] displays the fano factor for @xmath180 and @xmath181 ( @xmath182 ) corresponding to region ( i ) ( continuous line ) and for @xmath183 and @xmath96 ( @xmath184 ) corresponding to region ( ii ) ( dot - dashed line ) . in both cases
the presence of space charge in the active region is nearly negligible , insufficient to originate a potential energy maximum large enough so as to lead to coulomb suppression .
therefore , both regions concern with the absence of coulomb correlations ( @xmath185 ) . in region
( i ) pauli correlations are absent ( @xmath186 for @xmath127 ) , since @xmath181 implies that carriers are injected at the contacts with energies @xmath187 , so that the distribution function at the contacts is well approximated by the nondegenerate maxwell - boltzmann distribution , and the injection statistics is poissonian . on the contrary , in region ( ii ) pauli correlations are responsible for the suppression of shot noise ( @xmath164 , for @xmath127 ) since degenerate injection conditions prevail .
it is worth noting that , in all cases where the long range coulomb correlations are absent , as in the present case , the fano factor at low voltages decreases inversely with the applied bias according to the law @xmath188 which corresponds to the thermal noise behavior , while at high voltages ( @xmath145 ) it becomes constant with a value given by @xmath189 which corresponds to current saturation conditions . to obtain eq .
( [ fanosat ] ) use is made of eqs .
( [ isat ] ) and ( [ sisat ] ) .
the value of @xmath190 interpolates monotonically between @xmath191 for @xmath192 ( nondegenerate injection statistics ) and @xmath193 for @xmath194 ( strongly degenerate injection statistics ) . since the coulomb correlations vanish identically at thermal equilibrium and under current saturation conditions , the two limiting behaviors represented by eqs .
( [ fanother ] ) and ( [ fanosat ] ) are common to all cases , as we will see below .
figure [ figgl40mu_3 ] reports the fano factor for @xmath195 and @xmath196 ( @xmath197 ) corresponding to region ( iii ) of the general scheme .
it shows the presence of coulomb correlations ( @xmath198 for @xmath199 ) and the absence of pauli correlations ( @xmath186 for @xmath127 ) .
pauli correlations are absent because @xmath200 ( see above ) , while coulomb correlations set in due to relevant space charge effects . due to the small value of the ratio @xmath168
the asymptotic theory@xcite provides only some rough agreement with the exact result in the limited range of voltages associated with a suppressed behavior . for larger values of the ratio @xmath168 the agreement improves .
the behavior displayed by the system in region ( iii ) was previously examined in detail in ref . [ ] .
figure [ figgl20mu8 ] reports the fano factor for @xmath201 and @xmath96 ( @xmath202 ) corresponding to the region ( iv ) of the general scheme .
here we assist to a low and intermediate voltage range where coulomb correlations are dominating and to a high voltage range where pauli correlations prevail .
indeed , the fano factor displays the presence of coulomb correlations ( @xmath198 for @xmath199 ) , the absence of pauli correlations in an intermediate bias range ( @xmath203 for @xmath204 ) , and the presence of pauli correlations for higher bias ( @xmath205 for @xmath206 ) . physically , this behavior is due to the fact that in this region , even if contacts are under degenerate injection conditions ( @xmath177 ) , one has @xmath179 , so that for the lowest applied voltages ( @xmath207 ) it is @xmath208 and the current flows is only due to poissonian carriers at the tail of the fermi distribution function , leading to @xmath203 .
as @xmath22 increases , it is @xmath209 and sub - poissonian carriers near and below the fermi level increasingly contribute to the current and low - frequency noise , so that pauli correlations become manifest in the noise , thus leading to @xmath164 . note that in region ( iv ) one has generally @xmath210 , thus ensuring that the asymptotic theory @xcite provides a good approximation to the exact theory presented here in the range of bias satisfying @xmath211 , as illustrated in fig .
[ figgl20mu8 ] . finally , fig .
[ figgl2mu8 ] reports the fano factor for @xmath212 and @xmath96 ( @xmath213 ) corresponding to region ( v ) of the general scheme .
the fano factor displays the presence of both pauli correlations ( @xmath164 for @xmath127 ) and coulomb correlations ( @xmath198 for @xmath199 ) . in this region one
always has @xmath163 thus ensuring that pauli correlations are present at all applied bias .
therefore , this is the most interesting region to analyze the interplay between coulomb and pauli correlations .
we note that here the exact theory presented in this paper becomes strictly necessary , since in this region the ratio @xmath168 never takes values much higher than unity ( one typically obtains @xmath214 ) , thus reducing significantly the usefulness of the asymptotic theory,@xcite as illustrated in fig . [ figgl2mu8 ] . to provide more insight into the interplay between coulomb and pauli correlations in region ( v ) , fig .
[ figgl5 ] reports the fano factor as a function of bias for @xmath215 and several values of @xmath30 belonging to region ( v ) .
it is observed that by increasing the value of @xmath30 the contribution of the pauli correlations decreases faster than that of the coulomb correlations .
this is understood by noting that for @xmath194 , @xmath152 varies as @xmath193 , while @xmath150 , being determined by the ratio @xmath216 , varies as @xmath217 .
physically , this behavior reflects the increasing amount of sub - poissonian carriers contributing to the current as @xmath30 becomes higher .
these carriers lead to a more and more pronounced pauli suppression while they do not contribute so much to coulomb suppression , since due to their sub - poissonian character the fluctuations of @xmath34 they induce are much less significant than those originated by the poissonian carriers above the fermi level .
for this reason , @xmath150 tends to saturate for the highest values of @xmath218 while @xmath219 continues decreasing .
remarkably enough , in the region ( v ) for @xmath220 the total fano factor @xmath9 is well described in all cases by a @xmath221 dependence . as seen in the figure , this dependence on voltage is due to the joint action of both pauli and coulomb correlations .
it can also be observed that the onset of coulomb suppression takes place for higher values of @xmath22 as @xmath30 increases , since for this suppresion to become significant it is necessary that the contribution to the current due to carriers injected at the right contact becomes negligible .
this takes place when @xmath222 , condition that requires a higher value of @xmath22 as @xmath30 increases , for a given value of @xmath3 . to illustrate a realistic example where the theoretical results presented here could be applied
, we consider as a significant example the case of a ballistic gaas @xmath0 homodiode at @xmath223 and contact density @xmath224 @xmath225 @xmath226 .
for this structure @xmath227 mv , @xmath228 ( @xmath229 mev ) , @xmath230 nm and @xmath231 ( @xmath232 nm ) , where we have taken @xmath233 , and @xmath234 . for a sample length of @xmath235 nm , @xmath236 , and
according to fig .
[ figdiagr ] we are in region ( iv ) of the general scheme . for a sample length of @xmath237 nm , @xmath238 and we are in region ( v ) . for a sample length of @xmath239 nm , @xmath240 and we are in region ( ii ) . to explore regions ( i ) and ( iii ) one can refer to the case of heterodiodes.@xcite
we conclude that in principle the different behaviors of the fano factor predicted by the present theory can be investigated experimentally within realistic conditions .
we have presented a semiclassical theory of non - equilibrium noise ( shot - noise ) properties of @xmath0 ballistic semiconductor structures aiming at evaluating the relevance of the long range coulomb interaction and of the pauli exclusion principle .
the theory covers the whole range of system parameters , physically identified in the contact chemical potential , the sample length , and the applied voltage . within a unitary scheme free from any approximation we succeed in investigating the transition between : nondegenerate and degenerate injection conditions , short and long sample lengths , and low and high applied bias . through the determination of the fano factor ,
we have analyzed the relevance of coulomb and pauli correlations . at applied voltages above about @xmath241
both correlations lead to the suppression of shot noise .
we have identified five different regions in the plane defined by the sample length and the chemical potential corresponding , respectively , to the following conditions .
( i ) the absence of pauli and coulomb correlations ; ( ii ) the presence of pauli correlations and the absence of coulomb correlations ; ( iii ) the absence of pauli correlations and the presence of coulomb correlations ; ( iv ) the absence ( presence ) of pauli correlations for low ( high ) bias and the presence of coulomb correlations ; and ( v ) the presence of both pauli and coulomb correlations . case ( i ) corresponds to small sample lengths and non - degenerate conditions when different current pulses are clearly uncorrelated . case ( ii ) occurs for small sample lengths and degenerate injection conditions so that pauli correlations are the only being active . case ( iii ) implies a large sample length and non - degenerate injection conditions so that coulomb correlations are the only being active . case ( iv ) is associated with large samples and degenerate injection conditions . here
coulomb correlations are always present because of the small value of the debye screening length . in addition , at low bias only poissonian carriers at the tail of the contact energy distribution contribute to the noise so that pauli correlations are absent , while at higher bias pauli principle becomes active due to the increasing amount of carriers obeying binomial injection statistics that contribute to noise .
finally , case ( v ) refers to moderately long samples and degenerate injection conditions so that both coulomb and pauli correlations are present simultaneously .
the results of the theory are in perfect agreement with analogous mc simulations .
therefore , besides offering a complete physical picture of the subject , this study provides new insight into the noise properties of ballistic conductors , and of mesoscopic systems in general .
we believe that the theory here developed constitutes a powerful tool to design experimental investigations of the nonequilibrium noise properties of solid - state ballistic conductors .
partial support from the dursi of the generalitat de catalunya ( spain ) , the consejera de educacin y cultura de la junta de castilla y len thorugh the project sa057/02 , the ministerio de ciencia y tecnologa ( spain ) through the projects tic2001 - 1754 , bfm2000 - 0624 and the ramon y cajal program , and from the italy - spain joint action of the miur italy ( ref .
it-109 ) and mcyt spain ( ref .
hi2000 - 0138 ) is gratefully acknowledged .
in this appendix we derive the main formulae used in the paper by following a procedure similar to that used in ref . [ ] , here extended to the case of degenerate injection conditions . to this end ,
the first step consists in solving for the distribution function that satifies the vlasov - poisson - langevin system of equations , eqs .
( [ vlasov])-([fccorr ] ) . to find the solution we first note that , since transport is ballistic , the total longitudinal energy of a carrier @xmath242 , defined as @xmath243 remains constant during the flight of the carrier through the structure .
then , we note that , since transport is ballistic , carriers can reach a given point inside the structure either directly from the contacts or indirectly after reflection at the self - consistent potential barrier . moreover
, we can identify from which contact the carrier comes from . from these considerations it is easy to convince oneself that , in the low frequency limit of interest in the present paper , the distribution function solving the set of eqs . ( [ vlasov])-([fccorr ] ) is given by @xmath244 where @xmath245 with @xmath246 and @xmath247 being obtained through the boundary conditions in eqs .
( [ fbc0 ] ) and ( [ fbcl ] ) . in the previous expression @xmath248 and @xmath249 refer to the value and location of the potential energy maximum , respectively , while @xmath65 and @xmath250 correspond to the values of the potential energy at the contacts ( note that , according to the boundary conditions assumed here , these values do not fluctuate ) . since eq .
( [ fsol ] ) depends on the value of the maximum potential energy , to completely determine the distribution function we need to derive the equation satisfied by the energy maximum . to this purpose ,
we first obtain an expression for the carrier density by performing the integral in eq .
( [ density ] ) with the help of eq .
( [ fsol ] ) . after some algebra
one can show that the carrier density is given by @xmath251\right ) \,0<x < x_{m } \\
n^{+}\left ( u(x , t),u_{m};[f_{0},f_{l}]\right ) \,\,x_{m}<x < l \end{array } \right . \text { , } \label{nx}\end{aligned}\ ] ] where @xmath252)= \nonumber \\ & & \int_{u_{m}}^{+\infty } \frac{du}{\sqrt{2(u - u)}}\left [ f_{0}(u , t)+f_{l}(u , t)\right ] \nonumber \\ & & + \int_{u}^{u_{m}}\frac{du}{\sqrt{2(u - u)}}2f_{0}(u , t ) \label{nu-}\end{aligned}\ ] ] and @xmath253)= \nonumber \\ & & \int_{u_{m}}^{+\infty } \frac{du}{\sqrt{2(u - u)}}\left [ f_{0}(u , t)+f_{l}(u , t)\right ] \nonumber \\ & & + \int_{u}^{u_{m}}\frac{du}{\sqrt{2(u - u)}}2f_{l}(u , t)\text { . } \label{nu+}\end{aligned}\ ] ] in the previous equations the square brackets on the l.h.s .
mean a functional dependence .
now , by multiplying the poisson equation , eq.([poisson ] ) , by @xmath254 and integrating with respect to @xmath255 , it can be shown that the electric field @xmath256 is given through @xmath257\right ) \,0<x < x_{m } \\ e^{+}\left ( u(x , t),u_{m};[f_{0},f_{l}]\right ) \,x_{m}<x
< l \end{array } \right . \text { , } \label{ex}\end{aligned}\ ] ] with @xmath258)= \nonumber \\ & & \left\ { \int_{u_{m}}^{\infty } du\left [ \sqrt{u - u}-\sqrt{u - u_{m}}\right ] \left [ f_{0}(u , t)+f_{l}(u , t)\right ] \right . \nonumber \\ & & \left . + \int_{u}^{u_{m}}du\sqrt{u - u}2f_{0}(u , t)\right\ } ^{\frac{1}{2}}2^{% \frac{3}{4 } } \label{eu-}\end{aligned}\ ] ] and @xmath259)= \nonumber \\ & & -\left\ { \int_{u_{m}}^{\infty } du\left [ \sqrt{u - u}-\sqrt{u - u_{m}}\right ] \left [ f_{0}(u , t)+f_{l}(u , t)\right ] \right .
\nonumber \\ & & \left .
+ \int_{u}^{u_{m}}du\sqrt{u - u}2f_{l}(u , t)\right\ } ^{\frac{1}{2}}2^{% \frac{3}{4}}\text { . }
\label{eu+}\end{aligned}\ ] ] finally , from the definition @xmath260 and its integration with respect to the space coordinate , we arrive at the following inverse equations for the potential energy profile @xmath261)}=x\qquad \text { , } \label{ux-}\ ] ] valid for @xmath262 , and @xmath263)}=x - l\text { , } \label{ux+}\ ] ] valid for @xmath264 . from the previous expressions the location of the maximum energy potential can be obtained by evaluating either eq .
( [ ux- ] ) or eq .
( [ ux+ ] ) at @xmath265 , @xmath266 ) } & = & x_{m}\text { , } \label{xm1 } \\
\int_{u_{l}}^{u_{m}}\frac{du}{e^{+}(u , u_{m};[f_{0},f_{l } ] ) } & = & x_{m}-l\text { . } \label{xm2}\end{aligned}\ ] ] by eliminating @xmath267 from the two resulting equations we derive a closed equation for the value of the potential energy maximum in the form @xmath268)}-
\nonumber \\ & & \int_{u_{l}}^{u_{m}}\frac{du}{e^{+}(u , u_{m};[f_{0},f_{l}])}\text { . } \label{um}\end{aligned}\ ] ] equation ( [ um ] ) constitutes a closed equation to determine the value of the potential energy maximum , @xmath34 .
notice that it depends solely on the boundary conditions for the distribution function and the sample length .
once the value of the potential energy maximum is known , one can determine its location through either eq .
( [ xm1 ] ) or eq .
( [ xm2 ] ) .
then , from eq .
( [ fsol ] ) one obtains the explicit expression of the distribution function . in a similar way the explicit spatial dependence of the potential energy can be determined by substituting the value of @xmath34 in eqs .
( [ ux- ] ) and ( [ ux+ ] ) , and that of the electric field and carrier density by substituting @xmath34 in eq .
( [ ex ] ) and eq .
( [ nx ] ) , respectively . in this way
, a complete analytical solution of the model presented in sec.[model ] is obtained .
in particular , the electrical current , defined as @xmath269 can be shown to be given by @xmath270 \text { . }
\label{ivt}\ ] ] now , we are in the position to derive the corresponding expressions for the transport and noise properties . the average steady - state transport properties can be computed directly from the equations presented above by simply substituting into them @xmath271 with @xmath86 given through eq .
( [ fbcav ] ) . in this way
we arrive at the equations used in sec .
[ transport ] the fluctuating properties of any of the quantities of interest can be evaluated directly from the equations derived above by just performing the corresponding perturbation around the steady state , and by taking into account that the only source of fluctuations in the system is located at the contacts , and represented by the fluctuating term in eqs .
( [ fbcav ] ) . in this paper
we are interested on the low frequency current fluctuations . to compute them we perturb eq .
( [ ivt ] ) around the steady state thus obtaining @xmath272 - \nonumber \\ & & \left [ \overline{f}_{l}(\overline{u}_{m})-\overline{f}_{0}(\overline{u}% _ { m})\right ] \delta u_{m}(t)\text { . }
\label{dit}\end{aligned}\ ] ] where @xmath273 represents the fluctuations of the potential energy maximum . in the above expression
we distinguish two contributions to the current fluctuation , one coming directly from the contacts and the other coming indirectly through the self - consistent potential fluctuations . to express the dependence of the second contribution on the noise sources , we perturb eq .
( [ um ] ) . in performing such a perturbation it is convenient to shift all the energy integration variables by an amount equal to @xmath34 .
after some algebra one then arrives at the following expression for the fluctuations of the maximum potential energy : @xmath274 with @xmath275 where we have used that @xmath276 and @xmath277 . here
, we have defined @xmath278 with @xmath279 du}{e\left ( u,\overline{u}_{m};[\overline{f}_{0},\overline{f}_{l}]\right ) ^{3 } } \nonumber \\ & & -\text { } \int_{u_{l}}^{\overline{u}_{m}}\frac{\left [ \sqrt{2\left ( u - u\right ) } -\sqrt{2\left ( u-\overline{u}_{m}\right ) } \right ] du}{e\left ( u,% \overline{u}_{m};[\overline{f}_{0},\overline{f}_{l}]\right ) ^{3}}\text { , } \label{g > t}\end{aligned}\ ] ] @xmath280\right ) ^{3}}\text { , } \label{gtildl } \\
\widetilde{\gamma } _ { 0}^{<}(u ) & = & \int_{u_{0}}^{u}\frac{2\sqrt{2\left ( u - u\right ) } du}{e^{-}\left ( u,\overline{u}_{m};[\overline{f}_{0},\overline{f}% _ { l}]\right ) ^{3}}\text { , } \label{gtild0}\end{aligned}\ ] ] by substituting eq .
( [ dum ] ) in eq .
( [ dit ] ) for the current fluctuations we finally obtain @xmath281 with @xmath282 where @xmath134 } { \delta } \text { . } \label{omegat}\ ] ] now we are in the position to compute the low frequency current spectral density , defined as @xmath283 by substituting eq .
( [ ditfin ] ) into eq .
( [ sideft ] ) we arrive at @xmath284 where we have used that from eq .
( [ fccorr ] ) one has @xmath285 from eqs .
( [ g > t])-([sit ] ) it is straightforward to arrive at the equations used in sec .
[ noise ] .
t. gonzlez , o. m. bulashenko , j. mateos , d. pardo , and l.reggiani , phys .
b * 56 * , 6424 ( 1997 ) ; o. m. bulashenko , j. mateos , d. pardo , t. gonzlez , l. reggiani , and j. m. rub , phys . rev .
b * 57 * , 1366 ( 1998 ) .
the effects of injection statistics that are an arbitrary function of energy and of fractional - type have also been addressed.see for instance : o. m. bulashenko , j. m. rub and v. a. kochelap , phys .
b * 62 * , 8184 ( 2000 ) ; g. gomila and l. reggiani , phys . rev .
b * 63 * , 165404 ( 2001 ) .
|
we work out a semiclassical theory of shot noise in ballistic @xmath0 semiconductor structures aiming at studying two fundamental physical correlations coming from pauli exclusion principle and long range coulomb interaction .
the theory provides a unifying scheme which , in addition to the current - voltage characteristics , describes the suppression of shot noise due to pauli and coulomb correlations in the whole range of system parameters and applied bias .
the whole scenario is summarized by a phase diagram in the plane of two dimensionless variables related to the sample length and contact chemical potential . here
different regions of physical interest can be identified where only coulomb or only pauli correlations are active , or where both are present with different relevance .
the predictions of the theory are proven to be fully corroborated by monte carlo simulations .
|
the physical process known as double - diffusive convection was first described in the 1950 s by @xcite who observed the effect in an experiment . shortly afterwards
, it was also found in astrophysics when the first detailed stellar models were computed and @xcite found irregularities in their calculations concerning whether or not a zone with a gradient in molecular weight was stable according to the ledoux criterion or the schwarzschild criterion . but
even more than fifty years after its discovery , the field is still actively researched which is due to two reasons : on the one hand it lacked immediate practical incentives that have accelerated the development of other branches of fluid mechanics . on the other hand ,
numerical simulations were not possible for a long time because of the considerable computational expenses they demand .
for a summary of the historical development of the area see the paper by @xcite , for a recent physical review about semiconvection see @xcite .
+ double - diffusive convection occurs in situations where the effect of a thermal gradient on stability and the effect of a molecular weight gradient on stability compete with each other : if the temperature gradient stabilises the system and the molecular weight gradient destabilises it , thermohaline convection can occur .
its distinguishing property is the appearance of flow structures known as salt - fingers ( thus also the name salt - fingering convection ) . in the opposite case ( temperature gradient unstable and molecular weight gradient stable ) layering
convection / semiconvection can occur .
note that we used the term `` can occur '' . whether thermohaline / layering convection really does occur depends on the ratio of the molecular weight buoyancy frequency to the thermal buoyancy frequency , the so called stability ratio @xmath1 . in the incompressible case
it is equivalent to the ratio of the rayleigh numbers associated with the thermal instability and the instability caused by the molecular weight . in this paper
, our focus will be on layering convection .
situations where this process occurs on earth include the convection in the arctic ocean where cool and fresh melt water from above leads to a destabilising negative temperature gradient and a stabilising negative molecular gradient in salt @xcite .
other examples are east - african rift lakes which are heated from below by volcanic activity .
this leads to a temperature gradient ( unstable ) and causes dissolved gases like methane and carbon dioxide to be introduced into the system , thus causing a stabilising molecular weight gradient @xcite .
but not only systems on earth are prone to double - diffusive convection : it can also occur in astrophysical systems like in icy satellites , giant planets and massive stars .
very recently , @xcite have investigated the effects of a stabilising compositional gradient and the resulting double - diffusive convection in titan .
the role of semiconvection for the interior of giant planets has been discussed by @xcite and recently by @xcite who suggested that it might be responsible for the radius anomalies of some hot jupiters .
this thought is further developed by @xcite ; they investigated the effect of semiconvection on the interior structure of planets and showed that it could explain the luminosity anomaly of saturn @xcite .
they also point out : `` determining the solute transport properties in the regime of layered convection more precisely , however , will be central to evolutionary calculations .
3d hydrodynamical simulations in a realistic parameter range are thus strongly needed . ''
@xcite .
however , numerical simulations of double - diffusive systems in a realistic astrophysical parameter range pose a serious challenge ( and are , in fact , still impossible in the stellar regime with today s computers ) because of the huge spread of length and time scales of which the smallest length scale the size of the diffusive boundary layer needs to be resolved .
for example , the ratio of the diffusivities of temperature and solute ( the lewis number ) for the plasma in the interior of a semiconvective region of a star is @xmath2 @xcite .
this would require an impossible spatial resolution if one were to attempt a dns of such a zone ( also * ? ? ?
while simulations are nowhere near the realistic parameter range for stellar astrophysical conditions yet , the parameter regime of giant planets has become feasible with today s computers since their prandtl and lewis numbers are much more moderate : the prandtl number ranges from @xmath3 to @xmath4 , the lewis number is about @xmath5 @xcite . for idealised microphysics ,
there are a number of simulations in two dimensions ( e.g. * ? ? ?
* ) and in three dimensions ( e.g. * ? ? ?
* ) in this parameter regime .
recently , a simulation in a realistic parameter range for the atlantic ocean that correctly reproduces measurements has been conducted by @xcite .
however , all of the mentioned studies have neglected the effect of rotation on the development of double - diffusive convection .
while this may be justifiable in the case of thin layers as they are occuring in the arctic ocean ( layer thickness 1 to 5 m , see @xcite ) or in lake kivu ( average thickness of the mixed layers 0.48 m , see @xcite ) , it might not be negligible for large layers that could be forming in global convection zones on rapidly rotating giant planets and stars .
it might even prove to be essential if trying to determine if layered convection is indeed occurring in giant planets and stars and what its precise influences on the transport properties are .
our work is a first step in the direction of investigating the effects of rotation on semiconvective layers .
we note that while @xcite did study thermosolutial convection in rotating spherical shells , they investigated a different parameter regime than the one where layers are expected to form so their work gives us no lead as to how semiconvective _ layers _ are influenced by rotation .
we want to give a remark on nomenclature here : in oceans the molecular weight gradient is caused by dissolved salt .
that is why salinity gradient is another common term for the molecular weight gradient , particularly in oceanography .
we will use the term salinity in this papers as well because it is handier than molecular weight of second species .
we assume salinity to be the concentration of the solute , no matter what exactly the solute is .
the publication is structured as follows : in chapter [ sec : model_description ] we present the physical model and the underlying equations .
we introduce the governing dimensionless numbers and the boundary conditions . in chapter [
sec : numerical_implementation ] we discuss the numerical setup . in chapter
[ sec : results ] we present the results of our simulations .
first , we show the results for a run with one set of parameters without rotation to have a reference framework to which we can compare the following runs ( chapter [ sec : nonrotating ] ) .
next , we present the results of the simulations with rotation ( chapter [ sec : rotation ] ) and highlight some differences before investigating the influence of a change of the prandtl number @xmath6 and the density ratio @xmath0 chapter [ sec : modifying_pr_and_rrho ] .
this is followed by a discussion in chapter [ sec : discussion ] and conclusions in chapter [ sec : conclusion ] .
two concentric spherical shells are maintained at different , constant temperatures : a hot inner sphere ( radius @xmath7 , temperature @xmath8 ) and a cool outer sphere ( radius @xmath9 , temperature @xmath10 ) .
the axis of rotation is taken to be the @xmath11-axis , the rate of rotation is constant and parallel to the z - direction : @xmath12 .
gravity operates inwards in radial direction : @xmath13 .
the main simulation was run over a simulation time of almost one full thermal diffusion time scale .
we studied the effects of an increase of the rate of rotation on the temporal evolution of a double - diffusive state . for constant viscosities and diffusivities , the navier
stokes equations in the boussinesq form including rotation and conservation of solute , read @xmath14 @xmath15 @xmath16 @xmath17 with @xmath18 $ ] .
+ @xmath19 is the velocity of the flow , @xmath20 the pressure , @xmath21 the density , @xmath22 a reference density , @xmath23 the kinematic viscosity , @xmath24 the angular velocity , @xmath25 the position vector , @xmath26 the temperature , @xmath27 the reference temperature where @xmath28 , @xmath29 the salinity , @xmath30 the reference salinity where @xmath28 , @xmath31 the thermal diffusivity and @xmath32 the molecular diffusivity .
@xmath33 is the thermal expansion coefficient @xmath34 , @xmath35 is the saline expansion coefficient @xmath36 . with a later application in astrophysics in mind
, we concentrated on systems where the centrifugal force is assumed much smaller than the gravitational force .
hence , the term describing centrifugal forces @xmath37 will be neglected .
the equations are nondimensionalized with the scales @xmath38 @xmath39 is the difference between outer and inner radius , @xmath40 and @xmath41 are the differences of temperature and salinity between outer and inner radius .
the time scale used is the thermal diffusion time scale . inserting ( [ eq : nondim ] ) into ( [ eq : ns1 ] ) ( [ eq : ns4 ] ) and dropping the asterisks
leads to the dimensionless form of the equations : @xmath42 @xmath43 = - { \boldsymbol{\nabla}}p_{\mathrm{eff } } + \nabla^{2 } \boldsymbol{u } + ra_t t \boldsymbol{e_r } - ra_s s \boldsymbol{e_r } - \sqrt{{\ensuremath{ta } } } \boldsymbol{e_z } \times \boldsymbol{u},\ ] ] @xmath44 @xmath45 where we introduced the dimensionless numbers @xmath46 @xmath6 is the prandtl number , @xmath47 is the lewis number , @xmath48 are thermal and saline rayleigh numbers , respectively , and @xmath49 is the taylor number . @xmath50 and @xmath51 are related to each other by the stability parameter @xmath52 .
note that @xmath53 now stands for the time in thermal diffusion time scales , so that @xmath54 means that one thermal diffusion time scale has passed .
we also introduced the effective pressure @xmath55 which is the gradient of the scaled original pressure plus a constant term .
this can be written in this form because the constant term will vanish in the course of solving the equations .
the idea behind our setup is to observe a growing double - diffusive layer and to investigate how it is influenced by rotation . to achieve that , we chose the following boundary and initial conditions .
we applied no - slip boundary conditions which read @xmath56 these are reasonable boundary conditions because we assume our spherical shell to be one layer of a so called double - diffusive stack .
the appropriateness of these boundary conditions is explained in chapter 3.2 of @xcite .
the same assumption ( taking the shell to be one layer of many ) demands a step - like initial distribution of temperature and salt .
however , since the purpose of this work is to investigate the effects of rotation on a developed double - diffusive layer it was imperative for a layer to form within our simulated spherical shell .
we ran some simulations with different initial conditions : step - like initial distributions of both temperature and salinity and a step - like distribution of one quantity and a linear distribution of the other one .
our goal was to test the influence of these different initial conditions on the thermal nusselt number .
the thermal and saline nusselt numbers are measures for the convective heat and salt flux at the boundaries of the system , respectively .
for incompressible flows ( which we are looking at ) they are defined as the ratio of the total heat or salt flux and the heat or salt flux that would be transported by conduction alone : @xmath57 with @xmath58 being the total heat flux , @xmath59 the flux that would be transported if the temperature profile was linear between the bottom and the top , @xmath60 the measured saline flux and @xmath61 the saline flux that would be transported if the concentration profile was linear between the bottom and the top . the result is shown in figure [ fig : ic_vgl ] . ] ] each initial condition leads to the same asymptotic range of values for the average thermal nusselt number which means that the physical state after relaxation is the same .
only the time it takes to reach this state differs . for `` t step ,
s step '' , plumes immediately reached the upper boundary of the shell without any layering in between .
but as we wanted to investigate the effects of rotation on layering , this initial condition was no viable option for us . to reduce computational costs , we did not chose `` t linear , s linear '' .
this left us with the initial condition of a step in the temperature field and a linear distribution of salinity which offered a good compromise between observing layering and keeping simulation time within affordable limits .
we investigated three different parameter regimes .
the main simulations were run with @xmath62 and @xmath63 . to be able to reach the layered convective state
, @xmath0 has to be sufficiently small .
there exist two upper bounds for the maximum value of @xmath0 , for which layer formation occurs . the one given by linear stability analysis is @xmath64 which gives @xmath65 in our case . for larger values , the flow would be damped by viscous friction .
the other one is given by the model of @xcite ( figure 3 therein ) and is @xmath66 with our parameters and provides an upper limit for the subcritical instability that triggers layer formation .
additionally , to avoid the simple case of convective mixing due to an unstable stratification in the sense of @xcite , @xmath67 should be larger than @xmath4 . in simulations without rotation an increase of @xmath0
has a stabilising effect on the flow @xcite and different regimes as a function of @xmath67 are established ( figure 3 and chapter 3.2 in @xcite and the schematic illustration in figure 1 of @xcite ) .
we have investigated if the same applies when rotation is present and have chosen @xmath68 for our main simulations and @xmath69 for comparison runs .
in another set of comparison runs we reduced the prandtl number to @xmath70 to get a hint of the effects of viscosity on the rotational constraints . + a commonly used dimensionless number used to measure the respective importance of buoyancy and rotation on a system is the rossby number @xmath71 where @xmath72 is the characteristic flow speed and @xmath73 is the colatitude .
this can be written as @xmath74 in order for a motion to be significantly influenced by rotation the rossby number must be of order one or less .
we have chosen a range of rossby numbers near unity : @xmath75 and the case without rotation . since the used code uses the taylor number as an input parameter , we rewrite ( [ eq : ro_taylor ] ) as @xmath76
a point to note is that for the same rossby number , we have different taylor numbers when the prandtl number varies .
this is important because we ran simulations with prandtl numbers @xmath4 and @xmath70 .
accordingly , the taylor numbers of the simulations with @xmath77 had to be doubled so that the rossby number was the same .
the corresponding taylor numbers for prandtl numbers 0.5 and 1 at different colatitudes are shown in table [ tab : taylor_rossby ] .
.rossby numbers and corresponding taylor numbers for prandtl numbers @xmath4 and @xmath70 [ cols="^,^,^,^,^,^,^ " , ] obviously , increasing @xmath67 delays the time development of @xmath78 while lowering @xmath6 accelerates it .
but since these are far too few data points to make a sound assumption about an underlying law we restrict ourselves to just listing them .
we have studied the influence of rotation on semiconvection in a three dimensional spherical shell .
first , we have run simulations without rotation with @xmath79 and a stability ratio @xmath68 to set up a reference calculation .
then , we compared simulations with different rates of rotation to the non - rotating case and compared the values for thermal and saline nusselt numbers and kinetic energies .
we concluded that slow rotation has hardly any influence on layer formation while fast rotation suppresses convective transport completely . at intermediate rotation rates
the temporal development of layers may take an entire thermal diffusion time scale at some critical @xmath80 .
this is why short simulation times can be highly misleading as there may be a long relaxation phase . at low rotation rates the relaxation of @xmath78 to its quasi - equilibrium value
occurs on a much smaller timescale that for @xmath81 .
furthermore , for higher @xmath49 the equilibrium state is no longer quasi - adiabatic but becomes increasingly more superadiabatic with ever more extended , diffusive transition regions at the layer boundaries .
we have also compared results from simulations with a modified prandtl number or a modified stability ratio @xmath0 and conclude that the critical value of the taylor number @xmath82 at which convection is suppressed depends weakly on @xmath6 and strongly on @xmath67 . for lower @xmath6
the equilibrium state requires less ( turbulent ) kinetic energy , since less viscous friction occurs which converts kinetic energy into heat . for higher @xmath67 the maximum @xmath49
for which convection develops , drops , so the critical @xmath67 in the sense of @xcite and @xcite should depend on @xmath49 .
we have also compared our results with the three regimes of rotational constraints that @xcite suggested and conclude that they are a valid means of classifying the effect of rotation on semiconvection but have to be extended by a dependence on @xmath0 .
similarly , if using the rossby number as a means for the influence of rotation it has to be extended by a dependence on @xmath0 as well . we have studied the relationship of @xmath78 and @xmath81 which seems to be a good indication for the state of the flow in terms of layered convection , boundary layer creation , relaxation to thoroughly mixed convection and the final state of overturning convection . for the case without rotation , we compared our data with model predictions by @xcite , @xcite and @xcite . for the equilibrium state of the system , spruit s model fits our data perfectly , while wood et al.s model does so only after readjusting their fitting parameters .
rosenblum et al.s model does not deliver a satisfactory fit to our data . for the rotating case ,
the models have to be readjusted , however , since fast rotation is found to significantly affect the ratio of nusselt numbers .
our work is a step on the ladder of correctly understanding the influence of double - diffusive convection on the heat and solute transport in rapidly rotating systems like some stars and planets .
we already know that semiconvection significantly constrains heat transport .
if we now respect that high rates of rotation can further decrease the effective heat and solute fluxes , the assumed overall heat flux in a rapidly rotating system undergoing double - diffusive convection has to be lowered even further when modeling planets or stars .
future works can include a more precise study of the influence of rotation on the lifetime of a layer and an investigation of the opposite regime of double - diffusive convection : that of salt - fingering .
+ pb & fk are grateful to financial support from the austrian science fund ( fwf ) through project p25229-n27 . rh is supported by stfc grant st / k000853/1 .
mirouh , g. m. , p. garaud , s. stellmach , a. l. traxler , and t. s. wood ( 2012 ) . a new model for mixing by double - diffusive convection ( semi - convection ) .
i. the conditions for layer formation . _
750_(1 ) , 61 . , f. , f. kupka , and h. j. muthsam ( 2013 ) . .
in m. goupil , k. belkacem , c. neiner , f. lignires , and j. j. green ( eds . ) , _ lecture notes in physics _ , volume 865 of _ lecture notes in physics _ , pp . 219 .
|
so far , numerical studies of double - diffusive layering in turbulent convective flows have neglected the effects of rotation .
we undertake a first step into that direction by investigating how coriolis forces affect a double - diffusive layer inside a rotating spherical shell . for this purpose
we have run simulations in a parameter regime where these layers are expected to form and successively increased the rate of rotation with the result that fast rotation is found to have a similar stabilising effect on the overall convective flux as an increase of the stability ratio @xmath0 has in a non - rotating setup .
we have also studied to what extent the regimes of rotational constraints suggested by @xcite for rotation in the case of rayleigh - bnard convection are influenced by double - diffusive convection : their classification could also be applicable to the case of double - diffusive convection in a spherical shell if it is extended to be also a function of the stability ratio @xmath0 . furthermore , we examined the ratio of saline and thermal nusselt numbers and compared our results with models of @xcite , @xcite and @xcite .
we find our data to be fitted best by spruit s model .
our result that fast rotation further decreases the convective transport , which is already lowered by double - diffusive convection , could play a major role for e.g. the modeling of the interior of some rapidly rotating giant planets , as gaseous giant planets have recently been proposed to be influenced by double - diffusive convection .
|
in recent years x - ray missions such as and have demonstrated that seyfert-2 galaxies have a very complex x - ray spectrum ( see mushotzky , pounds & done 1993 for a recent review ) . in particular
their spectrum contains at least the following features : a ) a power - law continuum with slope @xmath2 absorbed by a very large column ( @xmath3 ) ( eg turner & pounds 1988 ) ; this obscuring screen is probably associated with a molecular torus .
b ) an fek line at 6.4 kev ie originating from neutral fe c ) a spectral curvature ( hump ) above 10 kev ; this is generally believed to originate from reflection of the power - law component on cold material ( eg lightman & white 1988 , george & fabian 1991 ) .
in contrast to the recent progress in understanding the spectral features of seyfert-2 galaxies , our knowledge of their x - ray variability properties remains scanty . usually , this has been based on comparison of the x - ray flux between observations from different missions spanning time intervals of several years .
recently , turner et al . (
1997 ) have detected some evidence for short term variability in a handful of seyfert-2 galaxies using .
the systematic monitoring of seyfert-2 galaxies has become possible with the _ rxte _ mission .
_ rxte _ observations of a few objects revealed strong variability in time - scales of hours as well as evidence for spectral variability ( georgantopoulos et al . 1999 , georgantopoulos & papadakis 2000 )
indeed , time variability studies can provide strong constraints in understanding the geometry of the circumnuclear matter in these objects . for example , the time lags between the power - law continuum and the normalization of the fe line could constrain the origin of the line . in the case where the fe line originates in an accretion disk
( as is the standard scenario for seyfert-1 galaxies ) , the fe line should closely track the variations of the intrinsic power - law . in this paper
we present an analysis of four observations of the seyfert-2 galaxies ngc 7172 and eso 103-g35 from the _ bepposax _ archive .
the two observations for each galaxy are separated by a year s interval .
our goal is to study any long - term variations in the spectral features between the two epochs . the broad energy band of @xmath4 as well as its good spectral resolution ,
provide the opportunity to place stringent constraints on the geometry of the nuclei of eso103-g35 and ngc7172 and any surrounding gaseous media .
ngc 7172 is an obscured , almost edge on galaxy .
it belongs to the compact group hcg90 and has a redshift of z=0.0086 .
it is classified as a type 2 seyfert galaxy .
strong infrared emission from the nucleus is observed , which is variable on time scales of approximately 3 months ( sharples et al .
the first x - ray observation which was performed by @xmath5 showed a @xmath6=1.84 power - law absorbed by @xmath7 ( turner & pounds 1989 ) .
later observations with @xmath8 confirmed the above results and also indicated that a reflection component may be important ( smith & done 1996 ) .
furthermore the analysis of observations ( turner et al . 1997 ) showed the presence of a 6.4 kev emission line with an equivalent width ( ew ) of @xmath9 ev .
in contrast with the previous results they found a rather harder photon index , @xmath10 , while the amount of the absorption column remained constant .
guainazzi et al .
( 1998 ) , using 2 @xmath11 observations separated by one year , found an even flatter photon index of @xmath12 absorbed by the same column density found in previous observations .
they further detect short - term variability ( within a day ) of about 30 percent . until 1995 , before the _ bepposax _ observations , the flux of ngc 7172 has remained almost constant between @xmath13 3 and @xmath14 . since may 1996 the source has reached a very low flux state of @xmath15 .
the flux variability history of ngc7172 is summarized in fig .
[ ngc7172_history ] .
the heao a2 hard x - ray source 1h 1832 - 653 was identified as the s0/sa type galaxy eso 103-g35 , with a redshift of z=0.013 , by phillips et al .
they classified the agn as a type 1.9 seyfert galaxy because of the weakness of @xmath16 emission line and the presence of a broad @xmath17 line .
@xmath5 observations , showed a steep photon index of @xmath18 absorbed by a high column density of n@xmath19 , which was found to vary within factor of 2 in a period of 90 days possibly due to motion of material in the broad line region across the line of sight ( warwick , pounds & turner 1988 ) .
later observations obtained by @xmath8 strongly suggested the presence of a narrow emission line at 6.4 kev and also of a reflection component with a photon index @xmath20 ( smith & done 1996 ) .
@xmath11 observations ( turner et al . 1997 ) confirmed the previous @xmath21 results and showed strong evidence for the presence of a triplet fe line ( 6.4 , 6.68 , 6.98 kev ) .
they also found a decrease by a factor of two in the continuum flux as compared to @xmath8 .
finally , forster , leighly & kay ( 1999 ) , analyzing three @xmath11 observations spanning a period of @xmath132 years ( 1994 - 1996 ) , found a decrease in the ew of the fe line while the continuum flux increased by a factor of 2 .
an edge at 7.4 kev was also found in the 1996 observation .
the long term variability history of eso 103-g35 is summarized in fig .
[ eso103_history ] .
the scientific instrumentation on board the italian - dutch x - ray satellite _ bepposax _ includes a medium energy concentrator spectrometer , mecs , which consists of three units , ( boella et al 1997 ) a low energy concentrator spectrometer , lecs , ( parmar et al 1997 ) a high pressure gas scintillation proportional counter , hpgspc , ( manzo et al 1997 ) and a phoswich detector system , pds , ( frontera et al 1997 ) , all of which point in the same direction .
the mecs instrument consists of a mirror unit plus a gas scintillation proportional counter and has imaging capabilities .
it covers the energy range between 1 - 10 kev with a spatial resolution of about 1.4 arcmin at 6 kev and a spectral resolution of 8 per cent at 6 kev .
the lecs instrument is similar to the mecs and operates in the 0.1 - 10 kev band with the same spectral and spatial resolution of the lecs .
the pds is a direct view detector with rocking collimators and extends the @xmath4 bandpass to high energies ( 13 - 300 kev ) .
its energy resolution is 15 per cent at 60 kev . in our analysis
we use mecs and pds data only .
our objects have not been detected by the hpgspc detector .
we did not use the lecs data as there are nearby contaminating soft x - ray sources , most probably stars in case of eso 103-g35 , while in case of ngc 7172 the contamination is due to the presence of the hickson compact group h90 .
these sources are not present in the hard mecs images .
another reason for excluding the lecs data was the poor photon statistics ( see tables 1 and 2 ) .
we analyze here two observations of ngc7172 and eso 103-g35 both obtained with @xmath4 within an interval of about one year .
more specifically , ngc7172 was first observed in 1996 ( from october 15th to 16th ) and then in 1997 ( from november 6th to 7th ) .
eso 103-g35 was first observed in 1996 ( from october 3rd to 4th ) and then in 1997 ( from october 14th to 15th ) .
tables 1 and 2 list the exposures times and net count rates ( background subtracted ) for all four observations . the three different ( or two after may 7 1997 , when mecs1 failed ) mecs units are merged together in order to increase the signal to noise ratio .
this is feasible because the three mecs units show very similar performance and the difference in the position of the optical axis in the three units is smaller than the scale on which the vignetting of the telescopes varies significantly ( @xmath225 arcmin ) . ._bepposax
_ observations of ngc 7172 [ cols="^,^,^,^,^,^,^",options="header " , ]
the spectral analysis of both observations implies that ngc 7172 is absorbed by a column density of n@xmath23 , in good agreement with previous @xmath5 ( turner & pounds 1989 ) , @xmath24 ( smith & done 1996 ) and @xmath11 ( turner et al 1997 ) observations .
the data require the presence of an fe line at 6.4 kev . in the 1996 _ bepposax _
observation the line is marginally resolved , at less than the 2@xmath25 confidence level , while in the 1997 observation ( where the photon statistics are poorer ) the line width is unconstrained .
we note that the line has been resolved in previous observations ( guainazzi et al .
there is some evidence ( fig .
[ ngc7172_contours ] ) for a variation in the normalization of the line .
this is in agreement with guainnazi at al ( 1998 ) who presented similar evidence for a variation of the line based on two @xmath26 observations ( 1995 and 1996 ) .
the line variation in conjunction with the line width has important implications for the geometry of the circumnuclear matter in ngc7172 .
indeed , the fe line must originate close to the nucleus possibly in the outer regions of an accretion disk .
this would simultaneously explain the fact that the continuum flux variations are tracked by changes in the line flux , and the large width of the line .
indeed , the line broadening may be due to the large velocities in the disk similar to that witnessed in mcg-6 - 30 - 15 by tanaka et al .
the observed broadening in the case of ngc 7172 implies a velocity of @xmath1370000 @xmath27 .
we tried to investigate further the above hypothesis including a line from a relativistic accretion disk ( for an accretion disk viewed at an intermediate inclination angle with solar abundances ) .
the fit was only marginally improved ( @xmath28 ) .
we note that georgantopoulos & papadakis ( 2000 ) find no significant variation of the normalization of the line in ngc 7172 using several _ rxte _ observations spanning a period of around a week .
this would set a lower limit of about a light week to the size of the line emitting region .
the observed ew can not be interpreted in terms of transmission processes in the torus alone . indeed leahy & creighton ( 1993 ) using monte carlo simulations found that a column density of n@xmath29 could produce an fe line ew of only @xmath30 ev .
this independently suggests that a reflection component may be present .
indeed , in this case the ew of the line should be increased by @xmath13150 ev due to reflection on the accretion disk ( reynolds , fabian & inoue 1995 , george & fabian 1991 ) . alternatively the observed ew may be explained on the basis of a high fe abundance .
indeed , excess absorption from fe is present in the first observation of ngc 7172 at @xmath31 kev , confirming earlier reports by warwick et al .
( 1993 ) ; however , note that superior spectral resolution observations with the @xmath11 sis ( turner et al 1997 ) do not corroborate this result . the edge energy is consistent with a neutral or low ionization material surrounding the source ( kallman & mccray 1982 ) .
the optical depth of the edge corresponds to an absorption column of @xmath32 .
this large column density may suggest an fe abundance well above the solar value .
if this is the case , the observed ew could be explained by transmission processes alone .
finally note that cases where a strong fe line is detected although there is no strong evidence for a reflection component are not uncommon ( eg gilli et al 2000 , reynolds et al 1995 ) .
the power - law photon index has a rather flat value of @xmath33 while it remains constant in the two _ bepposax _ observations .
the value of the photon index is in agreement with previous observations ( guainazzi et al 1998 , ryde et al 1997 ) who found a photon index of @xmath34 .
however , it is clearly flatter than that found in @xmath5 ( turner et al 1989 ) and @xmath24 ( smith & done 1996 ) ( @xmath35 ) . therefore the above may be suggesting long term spectral variability ( ryde et al . 1997 ) .
for example , assuming that the x - ray emission comes from comptonization of uv photons from an accretion disk on a hot electron corona ( rybicki & lightman 1979 ) , a steepening of the spectral index may be attributed to a change of the temperature or the optical depth of the corona .
of course it is difficult to make a detailed comparison due to differences in the energy bandpass of the various instruments as well as in the spectral analysis : for example the addition of a reflection component would result in the steepening of the spectral index .
however , we note that there is no correlation between the flux level and the photon index observed in the case of the ngc7172 observations in the past 20 years .
finally we note that despite the flat photon index found above , our data do not significantly support the existence of any reflection continuum . a recent analysis of @xmath36 monitoring observations of ngc 7172 ( georgantopoulos & papadakis 2000 ) again shows no evidence for reflection component .
the spectral analysis of the two observations of eso103-g35 implies that a simple power - law model absorbed by a column density of n@xmath37 plus a gaussian line at 6.4 kev provides a good fit to our data .
_ exosat _ observations of eso103-g35 ( warwick et al . 1988 ) reveal a 50 per cent decrease in the column density over a 90 day period .
this variation was interpreted in terms of an x - ray absorbing screen composed of broad - line clouds moving around the x - ray source .
alternatively , the n@xmath38 variation could be attributed to photoionization of the neutral column as the flux increases ; in this case an ionized fe edge should be detected .
however , in our analysis we detect no variation of the column within the errors . an fe line at 6.4 kev is strongly needed to fit our data .
unfortunately , the width of the line remains unconstrained .
however , _ asca _
sis observations with better spectral resolution have probably resolved the line into three components at 6.4 , 6.68 and 6.96 kev ( turner et al 1997 ) .
monte carlo simulations ( leahy & creighton 1993 ) show that a column density of n@xmath37 could produce an fe line ew of @xmath13170 ev .
however , in the case of eso 103-g35 where the line originates possibly far away from the central region , the ew varies accordingly to the variations of the continuum .
although a comparison is not straightforward , it is possible that a large fraction of the line emission could originate from transmission processes . some residuals around 7.5 kev ( see fig .
[ eso103 ] ) may indicate the presence of an absorption edge due to ionized fe .
this was first suggested by @xmath24 results ( warwick et al .
1993 ) and was confirmed later with @xmath11 observations ( forster et al 1999 ) .
the power - law photon index has a steep value of @xmath39 which is consistent with the typical value 1.9 for seyfert galaxies ( nandra & pounds 1994 ) .
despite our large energy bandpass we do not find strong evidence for a reflection continuum in contrast to previous @xmath24 ( smith & done 1996 ) and @xmath36 ( georgantopoulos & papadakis 2000 ) observations who found evidence for such a component but in a smaller band ( 3 - 20 kev ) .
higher signal to noise observations are necessary in order to test the presence of such a component .
previous studies of spectral variability in seyfert-2 galaxies have shown some interesting results .
in particular , iwasawa et al ( 1994 ) and griffiths et al .
( 1998 ) analyzed non - simultaneous _ ginga _ , _ asca _ , _ rosat _ and _ bbxrt _ observations to investigate the spectrum of the seyfert-2 galaxy mrk 3 .
their analysis demonstrated that the fe line does show some variability on time - scales of a few years .
systematic monitoring observation became feasible with the _ rxte _ mission .
in particular , georgantopoulos et al .
( 1999 ) , georgantopoulos & papadakis ( 2000 ) , smith , georgantopoulos & warwick ( 2000 ) present monitoring observations of several seyfert-2 galaxies ( mrk3 , eso 103-g35,ic 5063 , ngc 4507 , ngc 7172 & mrk 348 ) spanning time periods from about seven days to seven months . in most cases , they detect spectral variability in the sense that the column density decreases with increasing flux . however , they find no evidence for fe line variability . here instead , we find tentative evidence for line variability in ngc 7172 in a period of one year confirming previous _ asca _ results ( guainazzi et al 1998 ) . to our knowledge
this is one of the very few examples of fe line variability in seyfert-2 galaxies .
our finding requires at least one of the line emitting regions has dimension not much more than about a light year .
it would seem more likely that this scale size corresponds to an accretion disk or inner cloud structure rather than the inner extent of the putative molecular torus .
indeed , the molecular torus is probably a much greater structure .
at least in the case of ngc 1068 observations of molecular hydrogen ( tacconi et al 1994 ) showed that the obscuring torus is located between 30 and 300 pc .
we have analyzed four _ bepposax _ observations of ngc7172 and eso103-g35 ( two observations for each galaxy separated by about a year ) .
the goal was to search for spectral variability in a large energy band and therefore to attempt to constrain the geometry of the circumnuclear matter in these two galaxies .
the spectra of both galaxies are fitted by an absorbed power - law plus a neutral fe line in agreement with previous and observations .
a neutral fe edge is probably detected in the 1996 observation of ngc7172 .
our data do not require a reflection component at a statistically significant level .
we detected flux variability in both objects by about a factor of about two .
we find no significant evidence for spectral variability in the sense that the power - law spectral index and the column density have remained constant in both galaxies implying a homogeneous obscuring screen and constant physical conditions in the accretion disk corona despite the continuum flux variations ( but see warwick et al .
1988 , georgantopoulos & papadakis 2000 ) .
however , in the case of ngc7172 we detect a variation of the fe line flux by a factor of two following the power - law flux , confirming previous results .
this provides strong support to a scenario where the fe line at least in the case of ngc7172 originates in a region close to the nucleus .
these findings support the idea that one region responsible for the line emission should lie within one light year away from the nucleus .
this region may be associated with the outer part of an accretion disk .
the above results clearly emphasize the strength of monitoring observations in probing the geometry of the central engine in agn .
high resolution spectroscopic observations with gratings with both chandra and xmm are expected to revolutionize our knowledge of the physical conditions of the circumnuclear matter in agn .
these combined with future monitoring observations with large effective area instruments such as epic on - board xmm are expected to provide for the first time a detailed mapping of the central regions in agn .
we thank the anonymous referee for many useful comments and suggestions .
this project was funded by a greek - italian scientific collaboration under the title `` observations of active galaxies with the italian astrophysics mission _ bepposax _ '' .
aa is grateful to the bologna observatory group for their warm hospitality .
we are grateful to m. cappi , c. vignali and r. lazio for many useful discussions . boella g. et al , 1997 , a&as , 122,327 fiore , f. , guainazzi , m. grandi , p. , 1999
, handbook for bepposax nfi spectral analysis , + ftp://www.sdc.asi.it/pub/sax/docs/saxabc_v1.2.ps.gz forster k. , leighly k.m . , kay l.e .
, 1999 , apj , 523 , 521 frontera f. , gosta e. , dal fiume d. , feroci m. , nicastro l. , orlandini m. , palazzi e. , zavattini g. , 1997 , a&as , 122 , 357 georgantopoulos , i. , papadakis , i. , warwick , r.s . , smith , d.a . ,
stewart , g.c . , griffiths , r.g . , 1999 ,
mnras , 307 , 815 georgantopoulos , i. , papadakis , i. , mnras , 2000 in press george i.m .
, fabian a.c . , 1991 , mnras , 249 , 352 gilli r. , maiolino r. , risaliti g. , dadina m. , weaver k.a . , colbert e.j.m . , 2000 ,
a&a , 355 , 485 griffiths r.g .
, warwick r.s . ,
georgantopoulos i. , done c. , smith d.a .
, 1998 , mnras , 298 , 1159 guainazzi m. , matt g. , antonelli l.a . , fiore f. , piro l. , ueno s. , 1998 , mnras , 298 , 824 iwasawa k. , yaqoob t. , hisamitsu a. , yasushi o. , 1994 , pasj , 46 , l167 kallman t.r . , mccray r. , 1982 , apjs , 50 , 263 leahy d.a . ,
creighton j. , 1993 , mnras , 263 , 314 lightman a.p . ,
white t.r . , 1988 , apj , 335 , 57 magdziarz p. , zdiarski a.a . , 1995 , mnras , 273 , 837 manzo g. , giarrusso s. , santangelo a. , ciralli f. , fazio g. , piraino s. , segreto a. , 1997 a&as , 122 , 341 mushotzky r.f .
, done c. , pounds k.a . , 1993 ara&a , 31 , 717 nandra k. , pounds k.a . , 1994 , mnras , 268 , 405 parmar a.n . ,
martin d.d.e .
, bavdaz m. , favata f. , kuulkers e. , vacanti g. , lammers u. , peacock a. , taylor b.g . , 1997 a&as , 122 , 309 philips m.m . , feldamn f.r .
, marchall f.e . ,
wansteker , w. , 1979 , a&a , 76 , l14 reynolds c.s . ,
fabian a.c , inoue r. , 1995 , mnras , 276 , 1311 rybicki , g.b . ,
lightman a.p . , 1979 , ratiative processes in astrophyscs , new york , john wiley and sons , inc .
ryde f. , poutanen j. , svensson r. , larsson s. , ueno s. , 1997 , a&a , 328 , 69 sharples r.m .
, longmore a.j . ,
hawarden t.g . ,
carter d. , 1984 , mnras , 208 , 15 smith d.a .
, done c. 1996 , mnras , 280 , 355 tacconi l.j .
, genzel r. , blietz m. , harris a.i .
, madden s. , 1994 , apj , 426 , l77 tanaka et al . , 1995 , nat , 375 , 659 turner t.j . ,
george i.m . , nandra k. , mushotzky r.f . , 1997 , apjs , 113 , 23 turner t.j .
, pounds k.a .
, 1989 , mnras , 240 , 833 warwick r.s .
, pounds k.a . ,
turner t.j . , 1988 ,
mnras , 231 , 1145 warwick r.s . , sembay s. , yaqoob t. , makishima k. , ohashi t. , tashiro m. , kohmura y. , 1993 , mnras , 265 , 412
|
we investigate the x - ray spectra of the type-2 seyfert galaxies ngc 7172 and eso 103-g35 , using _ bepposax _ observations , separated by approximately one year .
we find that the x - ray spectra of both ngc 7172 and eso 103-g35 can be well fitted using a power - law model with an fe k@xmath0 emission line at 6.4 kev .
we did not find any statistically significant evidence for the existence of a reflection component in the x - ray spectra of these two galaxies .
the continuum flux has decreased by a factor of approximately two during this period , in both objects .
however , the spectral index as well as the absorption column have remained constant .
we find weak evidence for the decrease of the normalization of the fe k@xmath0 emission line in a similar manner to the continuum in ngc7172 .
we also report evidence for a broad fe k@xmath0 confirming previous observations .
in contrast , in the case of eso 103-g35 the line flux does not change while its width remains unresolved .
2@xmath1 = = = 1=1=0pt = 2=2=0pt = 2=2=0pt galaxies : active - quasars : general - x - rays : general
|
mode division multiplexing ( mdm ) is the method of optical communication where spatial modes are used as information channels carrying independent data streams . in general
, mdm can potentially increase the information capacity of optical communication in an amount proportional to the number of modes used .
mdm has been used in optical fiber communication ; for comprehensive reviews see @xcite .
potentially , mdm can also be used in free space optical communication ( fso ) @xcite .
polarization division multiplexing and wavelength division multiplexing have been used to increase the information capacity of fso @xcite . by definition
, spatial modes are solutions to maxwell s wave equation and can be represented by many bases @xcite . in principle , any basis can be used for mdm .
for example , what is sometimes referred to as the basis of orbital angular momentum ( oam ) modes has been used to increase the information capacity of fso and free space communication with millimeter waves @xcite .
there is also a basis sometimes referred to as the basis of vector modes or vector beams .
vector modes are spatial modes that have spatially inhomogeneous states of polarization , such as , radial and azimuthal polarization .
they can produce smaller spot sizes and stronger longitudinal polarization components upon focusing . as a result
, they are used for many applications , including optical trapping and nanoscale imaging @xcite . in this work
, vector modes are used to increase the information capacity of fso via mdm . a mode ( de)multiplexer for vector modes based on a liquid crystal technology referred to as a @xmath0-plate
is introduced . as a proof of principle , using the mode ( de)multiplexer four vector modes , each carrying a 20 gbit / s quadrature phase shift keying signal on a single wavelength channel ( @xmath11550 nm ) , comprising an aggregate 80 gbit / s ,
were transmitted @xmath21 m over the lab table , with @xmath3 - 16.4 db ( @xmath4 ) mode crosstalk .
bit error rates for all vector modes were measured at the forward error correction threshold with power penalties @xmath3 3.41db .
a vector mode is defined here as a solution to the wave equation in free space whose electric field is given by the equation @xcite : @xmath5 where @xmath6 are cylindrical coordinates and @xmath7 is a solution to the radial part of the wave equation , e.g. , bessel - guassian @xcite , or laguerre - gaussian functions .
@xmath8 is a jones vector describing the spatially inhomogeneous states of polarization of a vector mode given by the equation @xcite : @xmath9 where @xmath10 , and @xmath11 .
@xmath12 and @xmath13 are referred to as radial and azimuthal polarization and are shown in fig .
1(a1 ) and fig .
1(a2 ) , respectively . @xmath14 and @xmath15 are shown in fig .
1(b1 ) and fig .
1(b2 ) , respectively .
the basis of vector modes comprises light s space and polarization degrees of freedom ; the number of mdm channels that can be used for mdm when using vector modes is the same as using , for example , oam modes together with polarization division multiplexing .
to use vector modes for mdm , a mode ( de)multiplexer for vector modes is required . in general , a mode ( de)multiplexer ( separates ) combines @xmath16 spatial modes at the ( receiver ) transmitter of an optical communication system .
for example , a mode ( de)multiplexer for oam modes that has been demonstrated is based on passive beam ( separation ) combining @xcite . in passive beam ( separation ) combining , data streams from @xmath16 single mode optical fibers ( smfs ) are transformed into @xmath16 oam modes , or vice vera , via matching their wavefronts with @xmath16 spatial light modulators ( slms ) .
they are then ( separated ) combined via @xmath17 beam splitters .
slms can also be used to generate vector modes @xcite .
however , there are many methods to generate vector modes ; for example see @xcite . here
, a liquid crystal technology referred to as a @xmath0-plate is used @xcite .
analogous to how a slm matches an oam mode s wavefront , a @xmath0-plate matches a vector mode s spatially inhomogeneous state of polarization @xcite .
a @xmath0-plate comprises a thin layer of pattern liquid crystal molecules in - between two thin glass plates .
the molecules orientations are described by @xmath18 , where @xmath19 is the azimuthal coordinate and @xmath0 is a half - integer . @xmath20 and
@xmath21 plates are schematically shown in fig .
1(a ) and fig .
1(b ) , respectively .
effectively , a @xmath0-plate is a half - wave plate with an azimuthally varying fast axis and can be mathematically represented by a jones matrix @xcite : @xmath22 it can be shown using jones calculus , upon propagation through a @xmath0-plate , the state of polarization of the fundamental mode of a smf , given by the jones vector @xmath23 ( @xmath24 , i.e. , a linear combination of horizontal and vertical polarization ) is transformed into a linear combination of vector modes , or vice versa , given by the equation : @xmath25
using a @xmath20 or @xmath21 plate , horizontal / vertical polarization can be transformed into @xmath26 or @xmath27 vector modes , or vice versa , as shown in fig .
1(a ) and fig .
1(b ) , respectively . two @xmath20 and @xmath21 plates were fabricated .
the fabrication details can be found in @xcite . using the @xmath0-plates , a mode ( de)multiplexer for vector modes based on passive beam ( separation ) combining
was constructed as shown in fig .
as a mode multiplexer- the fundamental modes of two smfs at @xmath281550 nm were collimated to beam waists of @xmath23 mm using appropriate lenses ( l ) and then made to propagate through the @xmath0-plates .
then , the light beams from each smf were aligned co - linearly via a non - polarizing beam splitter ( bs ) .
the @xmath0-plates were connected to an electronic signal generator generating a 1khz square wave with a tunable voltage .
the @xmath0-plates could be
turned on " and turned off " by controlling the voltage .
each @xmath0-plate was turned on " ( @xmath21.2 volts ) .
note , there is a -1.16db loss when the @xmath0-plate is turned on .
" @xmath26 and @xmath27 vector modes as generated using the mode ( de)multiplexer are shown in fig .
1(a1 ) / fig .
1(a2 ) and fig .
1(b1 ) / fig .
1(b2 ) , respectively , along with their intensities as analyzed by a linear polarizer whose transmission axis is oriented at 0 , 45 , 90 , and 135 degrees with respect to the lab .
using the mode ( de)multiplexer , four vector modes , each carrying a 20 gbit / s quadrature phase shift keyed ( qpsk ) signal on a single wavelength channel ( @xmath281550 nm ) , comprising an aggregate 80 gbit / s , were transmitted as one light beam @xmath21 m over the lab table .
note , over this distance there is negligible atmospheric turbulence and beam divergence .
the experimental setup is shown in fig .
the transmitter was as follows .
first , the output of an external cavity laser ( ecl ) tuned to @xmath29 nm was modulated by an i - q modulator to generate a 20 gbit / s qpsk signal .
the i - q modulator was driven by a two - channel pattern generator generating two 10 gbaud signals , being decorrelated pseudo - random - binary - bit - sequences of length @xmath30 , and comprising the in - phase ( i ) and quadrature ( q ) components of the qpsk signal .
the signal was then amplified by a low - noise erbium doped fiber amplifier ( edfa ) and polarization division multiplexed ( pdm ) by splitting and decorrelating it in two smfs , making the states of polarization in each smf mutually orthogonal , ( horizontal / vertical polarization ) , and then recombining them into one smf . finally , the resultant pdm - qpsk signal was mode multiplexed by again splitting and decorrelating it in two smfs which were then connected to the smfs of the mode ( de)multiplexer , resulting in four vector modes each carrying a 20 gbit / s qpsk signal . at the mode ( de)multiplexer , due to twists and bends in the smfs , the channels of each pdm - qpsk signal are not necessarily horizontal and vertical polarization
however they can be aligned using a polarization paddle controller ( pol - con ) .
nonetheless , each channel is a linear combination of horizontal and vertical polarization as represented by two arbitrary antipodal points on the poincar sphere .
therefore , as given by eq .
5 , after propagation through a @xmath0-plate each channel of a pdm - qpsk signal is transformed into a linear combination of vector modes as represented by two arbitrary antipodal points on a higher - order poincar sphere @xcite .
the channels originating from the @xmath20 and @xmath21 plates are labelled ch1/ch2 and ch3/ch4 as shown in fig .
1(a4 ) and fig .
1(b4 ) , respectively .
the intensities of ch1/ch2 and ch3/ch4 , as recorded using the ingaas camera , are shown in fig .
3(a ) and fig .
3(b ) , respectively . after transmission over the lab table ,
the receiver was as follows .
first , the channels were mode demultiplexed into the two smfs of another mode ( de)multiplexer .
note , ch1/ch2 and ch3/ch4 were received and discriminated via coupling to both smfs . in accordance with eq .
3 , upon propagation through a @xmath20 plate , @xmath26 vector modes are transformed into horizontal / vertical polarization while @xmath27 vector modes are transformed into higher - order vector modes .
the intensities of ch1/ch2 and ch3/ch4 at the smf corresponding to the @xmath20 @xmath0-plate , as recorded using an ingaas camera , are shown in fig .
3(c ) and fig .
3(d ) , respectively .
as can be seen , ch1/ch2 were transformed back into the fundamental mode of a smf , i.e , a pdm - qpsk signal , and were coupled to the smf .
however , ch3/ch4 were transformed into higher - order vector modes and were not coupled to the smf .
the opposite is true for the @xmath21 @xmath0-plate .
the intensities of ch1/ch2 and ch3/ch4 at the smf corresponding to the @xmath21 @xmath0-plate , as recorded using an ingaas camera , are shown in fig .
3(e ) and fig .
3(f ) , respectively . next
, each resulting pdm - qpsk signal was amplified by an edfa and aligned via a pol - con to a polarization diversity coherent receiver where they were polarization demultiplexed and coherently detected .
the polarization diversity coherent receiver comprised a polarizing splitter , followed by two optical hybrids , connected to another ecl ( `` local oscillator '' ) . via intradyne detection
, the analog qpsk signals were converted to digital signals by four balanced detectors . using four digital oscilloscopes ,
the resultant digital signals were captured at 40 gsample / s ( @xmath3120 ghz ) for offline digital signal processing ( dsp ) .
the captured constellations of the qpsk signals for all channels are shown in fig 3(a ) .
mode crosstalk ( mc ) is defined as the transfer of power ( signals ) between channels .
mc for each channel was measured at the mode demultiplexer by transmitting one channel at a time and measuring the power received by each of all channels . a normalized mc matrix " , comprising the mc measurements , are shown in fig .
there is @xmath3 - 16.7db ( @xmath32 ) mc for any channel . as can be seen , via the off - diagonal blocks , the greatest contribution to mc for any channel came from channels originating from different @xmath0-plates .
as there is negligible atmospheric turbulence and beam divergence over the transmitted distance , this is attributed to mismatch of the numerical apertures and misalignment of the lenses with the smfs of the mode demultiplexer .
mc may also originate from mode impurity , " i.e. , power is generated in vector modes other than those intended .
a preliminary analysis indicates the vector modes as generated via the @xmath0-plates of the mode multiplexer have @xmath33db @xmath34 mode impurity .
a detailed analysis of the measurement of mode impurity will be reported elsewhere . for each channel and aback to back " ( b2b ) channel , bit error rates ( ber )
were measured as a function of optical signal to noise ratio ( osnr ) when all channels were simultaneously transmitted .
ber vs. osnr curves are shown in fig .
the b2b channel is the original qpsk signal , transmitted and received , when the @xmath0-plates are turned off . "
the performance of each channel can be assessed via its power penalty ( pp ) , i.e. , the difference in osnr between each channel and the b2b channel at the forward error correction ( fec ) threshold ( ber = @xmath35 ) .
the pps for all channels and their cumulative mcs are shown in table 1 .
all channels reach a ber at the fec threshold with a pp @xmath3 3.41db .
differences in pps are attributed to polarization dependent loss throughout the various smfs of the system and misalignment of the pol - cons .
in conclusion , vector modes were used to increase the information capacity of fso via mdm . a mode ( de)multiplexer for vector modes based on a liquid crystal technology referred to as a @xmath0-plate was introduced . as a proof of principle , using the mode ( de)multiplexer four vector modes , each carrying a 20 gbit / s quadrature phase shift keying signal on a single wavelength channel ( @xmath11550 nm ) , comprising an aggregate 80 gbit / s , were transmitted @xmath21 m over the lab table , with @xmath3 - 16.4 db ( @xmath4 ) mode crosstalk .
bers for all vector modes were measured at the forward error correction threshold with power penalties @xmath3 3.41db .
it has been conjectured upon propagation through atmospheric turbulence a vector mode will experience less scintillation as compared to an oam mode @xcite .
the propagation of vector modes through atmospheric turbulence is the subject of future work . when using oam modes for mdm , the deleterious effects of atmospheric turbulence can be compensated via wavefront measurements @xcite . when using vector modes for mdm , comparable compensation of atmospheric turbulence may be possible via the measurement of spatially inhomogeneous polarization @xcite .
it has been shown , the transmission of an oam mode through scattering media depends on its polarization and oam @xcite . as vector modes are non - separable superpositions of circular polarized oam modes @xcite , a comparable study of vector modes may be of interest , especially with regard to atmospheric light scattering .
vector modes are referred to as the true modes " of an optical fiber .
use of the @xmath0-plate mode ( de)multiplexer for optical fiber communication , as well as its compatibility with wavelength division multiplexing , especially with regard to ring core " optical fibers , is the subject of current work @xcite .
it is noted , there are other light beams that have spatially inhomogeneous states of polarization , such as , full poincare beams @xcite .
however , as compared to vector modes , full poincare beams experience non - trivial dynamics upon propagation @xcite .
.*power penalties * [ cols="^,^,^",options="header " , ]
gm and rra acknowledge support from afosr grant .
no . 47221 - 00 - 01 , aro grant .
52759-ph - h , nsf grfp grant .
40017 - 00 - 04 , and corning , inc .
ek acknowledges support from the canada excellence research chairs ( cerc ) program .
mpjl acknowledges support from epsrc and the royal academy of engineering .
usc acknowledges support from the darpa inpho program .
gm thanka s. slussarenko for fabricating the @xmath0-plates .
authors from usc thank dr .
prem kumar and dr .
tommy willis for helpful discussions .
e. ciaramella , y. arimoto , g. contestabile , m. presi , a. derrico , v. guarino , and m. matsumoto ,
1.28-tb / s ( 32@xmath3640 gb / s ) free - space optical wdm transmission system , " ieee photon .
. lett . * 21*(16 ) , 1121 - 1123 ( 2009 ) .
j. wang , j. y. yang , i.m .
fazal , n. ahmed , y. yan , h. huang , y. ren , y. yue , s. dolinar , m. tur , and a. e. willner , terabit free - space data transmission employing orbital angular momentum multiplexing , " nat .
* 6 * , 488 - 496 ( 2012 ) .
h. huang , g. xie , y. yan , n. ahmed , y. ren , y. yue , d. rogawski , m. j. willner , b. i. erkmen , k. m. birnbaum , s. j. dolinar , m. p. j. lavery , m. j. padgett , m. tur , and a. e. willner ,
100 tbit / s free - space data link enabled by three - dimensional multiplexing of orbital angular momentum , polarization , and wavelength , " opt .
lett . , * 39*(2 ) , 197 - 200 ( 2014 ) .
y. yan , g. xie , m. p. j. lavery , h. huang , n. ahmed , c. bao , y. ren , y. cao , l. li , z. zhao , a. f. molisch , m. tur , m. j. padgett , and a. e. willner , high - capacity millimetre - wave communications with orbital angular momentum multiplexing , " nature commun . 5 , 1 - 9 ( 2014 ) .
g. milione , a. dudley , t. a. nguyen , k. chakraborty , e. karimi , a. forbes , and r. r. alfano , experimental measurement of the self - healing of radially and azimuthally polarized vector bessel beams , " to be published in j. opt .
g. milione , h. i. sztul , d. a. nolan , j. kim , m. etienne , j. mccarthy , j. wang , and r. r. alfano , `` cylindrical vector beam generation from a multi elliptical core optical fiber , '' proc .
spie 7950 , 79500k ( 2011 ) .
g. milione , h. sztul , d. nolan , j. kim , m. etienne , j. mccarthy , j. wang , and r. alfano , `` cylindrical vector beam generation from a multi elliptical core optical fiber , '' in cleo:2011 - laser applications to photonic applications , osa technical digest ( cd ) ( optical society of america , 2011 ) , paper ctub2 .
f. cardano , e. karimi , s. slussarenko , l. marrucci , c. de lisio , and e. santamato , polarization pattern of vector vortex beams generated by @xmath0-plates with different topological charges , " appl . opt . * 51 * , c1-c6 ( 2012 ) .
y. ren , g. xie , h. huang , n. ahmed , y. yan , l. li , c. bao , m. lavery , m. tur , m. neifeld , r. boyd , j. shapiro , and a. willner , `` adaptive - optics - based simultaneous pre- and post - turbulence compensation of multiple orbital - angular - momentum beams in a bidirectional free - space optical link , '' optica 1 , 376 - 382 ( 2014 ) . c. n. alexeyev , a. n. alexeyev , b. p. lapin , g. milione , and m. a. yavorsky , spin - orbit - interaction - induced generation of optical vortices in multihelicoidal fibers , " phys .
a * 88*(6 ) , 063814 ( 2013 ) .
y. rumala , g. milione , t. nguyen , s. pratavieira , z. hossain , d. nolan , s. slussarenko , e. karimi , l. marrucci , and r. alfano , `` tunable supercontinuum light vector vortex beam generator using a @xmath0-plate , '' opt . lett . *
38 * , 5083 - 5086 ( 2013 ) .
n. bozinovic , y. yue , y. ren , m. tur , p. kristensen , h. huang , a. e. willner , s. ramachandran , terabit - scale orbital angular momentum mode division multiplexing in fibers , " science 340(6140 ) , 1545 - 1548 ( 2013 )
|
vector modes are spatial modes that have spatially inhomogeneous states of polarization , such as , radial and azimuthal polarization .
they can produce smaller spot sizes and stronger longitudinal polarization components upon focusing . as a result
, they are used for many applications , including optical trapping and nanoscale imaging . in this work ,
vector modes are used to increase the information capacity of free space optical communication via the method of optical communication referred to as mode division multiplexing . a mode ( de)multiplexer for vector modes based on a liquid crystal technology referred to as a @xmath0-plate
is introduced . as a proof of principle , using the mode (
de)multiplexer four vector modes each carrying a 20 gbit / s quadrature phase shift keying signal on a single wavelength channel ( @xmath11550 nm ) , comprising an aggregate 80 gbit / s , were transmitted @xmath21 m over the lab table with @xmath3 - 16.4 db ( @xmath4 ) mode crosstalk .
bit error rates for all vector modes were measured at the forward error correction threshold with power penalties @xmath3 3.41db .
|
an antidot is a potential hill in a two - dimensional electron system .
when a strong perpendicular magnetic field @xmath0 is applied , the system shows the quantum hall effects , and there appear bound modes of quantum hall edge states around the antidot .
as the modes enclose magnetic flux , they are governed by the aharonov - bohm ( ab ) effect .
@xcite when the bound modes couple with extended edge channels via tunneling , the antidot shows ab resonance peaks in electron conductance , as a function of @xmath0 or the gate voltage @xmath1 applied to the antidot .
an antidot is a useful tool for detecting and studying localized quantum hall edge states .
there are experimental evidences that electron interactions play an essential role in an antidot in the integer quantum hall regime .
@xcite in an antidot with one or two bound modes , the interactions cause interesting phenomena such as @xmath2 ab effects , @xcite charging effects , @xcite and kondo effects . @xcite ab effects in an antidot with modes more than 2 , @xcite spectator behavior in an antidot molecule , @xcite and charge screening effects @xcite have also been reported .
some of the experimental results have been theoretically understood .
a phenomenological model for an antidot captures aharonov - bohm physics of antidot bound modes as well as the capacitive interactions of excess charges around the antidot ; other approaches , such as computations with local - density - functional approximation , @xcite will be useful for studying charge screening .
this model successfully describes the @xmath2 ab effects and kondo effects in an antidot @xcite as well as the spectator behavior in an antidot molecule , @xcite and it agrees with a hartree - fock numerical calculation for an antidot .
@xcite this may motivate one to extend the model to a general antidot - based interferometer in the integer quantum hall regime
. the extension will be useful for exploring generic aspects of an antidot system with multiple modes , and also for studying antidots in the fractional quantum hall regime .
it may be applicable , with modification , to other quantum hall interferometers .
@xcite in this paper , we develop a capacitive interaction model applicable to a general situation of antidots , and apply it to an antidot with @xmath3 , where @xmath4 is the local filling factor around the antidot and equals the number of antidot bound modes .
the application demonstrates the generic features by charging effects , charge relaxation between bound modes , the interaction between bound modes and the extended edge channels nearby the antidot , and the competition between charging energy and single - particle level spacing .
it provides systematic understanding of the ab effects in electron conductance @xmath5 through the antidot as a function of @xmath0 or @xmath6 . for @xmath7
, the ab effects can deviate from the @xmath8 ab oscillation in which there appear @xmath4 resonance peaks with equal height and equal peak - to - peak spacing within one period of the peaks .
moreover , the case of @xmath9 shows resonance signals significantly different from that of @xmath10 .
the difference comes from the fact that the interaction between the bound modes and the extended edge channels is much stronger in the case of @xmath9 .
the predictions agree with recent experimental data .
@xcite this paper is organized as follows . in sec .
ii , we derive the capacitive interaction model . in sec .
iii , we summarize nontrivial features of experimental data @xcite in an antidot with @xmath3 . in sec .
iv , we apply the capacitive interaction model to the antidot , and analyze the ab resonance pattern in the regime that the charging energy is much larger than the single - particle level spacing . in sec .
v , we consider finite single - particle level spacing .
section vi provides the summary .
in this section , we derive the capacitive interaction model applicable to a general situation of an antidot system under a strong magnetic field @xmath11 , where @xmath12 is the period of ab resonances of the system . for simplicity , the system is considered to be in the tunneling regime that the antidot bound modes are weakly coupled to extended edge channels , and to be in the regime of zero bias and zero temperature ; note that we ignore kondo effects .
@xcite with this simplification , we first describe bound modes in the noninteracting limit , and then the interaction between them . .
excess charges @xmath13 accumulated in mode @xmath14 interact with those of the other modes ( not shown ) , the extended edge channels nearby the antidot , and the backgate .
@xmath15 , @xmath1 , and @xmath16 are the voltages of mode @xmath14 , the backgate , and the extended edge channels , respectively .
@xmath17 ( @xmath18 ) is the capacitance between mode @xmath14 and the extended edge channels ( the backgate ) .
( b ) electric circuit model equivalent to an antidot system with multiple modes .
@xmath19 is the capacitance between modes @xmath14 and @xmath20.,title="fig:",scaledwidth=48.0% ] + we consider an antidot system , which could be a single antidot or multiple antidots , such as an antidot molecule ; see a schematic view of an antidot in fig . [ ad-1-model](a ) . in the integer quantum hall regime ,
the bound modes of edge states are formed around the potential of each antidot , which is usually assumed to slowly vary on the scale of magnetic length ( @xmath21 ) .
each mode corresponds to a landau level whose energy varies to pass the fermi level near its antidot . for a single antidot ,
the number of bound modes is identical to the local filling factor @xmath4 around the antidot ; @xmath4 is smaller than the bulk filling factor @xmath22 .
each mode weakly couples to extended edge channels with the same spin nearby the mode via electron tunneling , and also to the other modes with the same spin .
the tunneling strengths depend on the geometry of the system .
as we focus on the weak tunneling regime , we ignore the tunneling when we discuss the energy of bound modes ; the tunneling will be taken into account in the calculation of @xmath5 .
each mode @xmath14 has orbitals @xmath23 with discrete single - particle energy @xmath24 .
@xmath23 is the angular momentum quantum number of the orbital or the number of the magnetic flux quanta enclosed by the orbital .
@xmath24 is determined by antidot potential , landau level energy , and zeeman energy .
as @xmath0 increases from a value , say @xmath25 , by @xmath26 , each orbital @xmath27 spatially shrinks toward the center of its antidot , to keep the same number @xmath23 of magnetic - flux quanta .
this results in the increase of @xmath24 , @xmath28 where @xmath29 is the single - particle level spacing of mode @xmath14 and @xmath30 is the period of ab resonances by mode @xmath14 .
the linear dependence of @xmath31 on @xmath26 is valid for @xmath32 . in the noninteracting limit
, the period is determined by the area @xmath33 enclosed by the single - particle orbital , @xmath34 , where @xmath35 is the flux quantum .
equation describes ab resonances in the noninteracting limit . as @xmath26 increases , the energy of each orbital passes , one by one , through the fermi level . when the energy matches with the fermi level , electrons tunnel in and out of the orbital , showing a resonance peak in @xmath5 .
in the case that there are multiple modes in the system , each mode results in its own ab resonance peaks independently of the others in the noninteracting limit . the peak height and width
are determined by the tunneling strength of the mode to extended edge channels .
we turn on electron interactions . in the integer quantum hall regime
, they may be well described @xcite by capacitive interactions between excess charges accumulated in bound modes , and with those in the extended edge channels nearby the antidot system .
excess charges accumulated in each mode depend on @xmath26 .
as @xmath26 increases , each orbital of a mode moves toward the center of its antidot to keep the same magnetic flux quanta enclosed by it , which results in the shift of the electron density of the mode .
the accumulated excess charge @xmath36 in mode @xmath14 due to @xmath26 is written as @xmath37 where @xmath38 is the unit of electron charge , and @xmath39 is the total number of electrons occupying mode @xmath14 at @xmath25 .
@xmath40 gives the rate of the accumulation .
@xmath41 is a parameter of our model , and it is the period of ab resonances . in the presence of the interaction
, @xmath41 is different from @xmath42 , therefore , @xmath43 is interpreted as the _ effective _ area enclosed by mode @xmath14 .
we treat the effect of the capacitive interactions between excess charges , by generalizing an electric circuit model used for a quantum dot .
@xcite figure [ ad-1-model](b ) shows the circuit equivalent to the antidot system , where the excess charges of bound modes , the backgate , and extended edge channels interact with each other capacitively ; the capacitive interaction between the external voltage sources is ignored .
the capacitance between mode @xmath14 and the backgate ( the extended edge channels ) is denoted by @xmath44 ( @xmath45 ) . the charge @xmath13 in eq .
is related with the voltages ( measured from the ground ) @xmath46 , @xmath1 , and @xmath16 of mode @xmath20 , the backgate , and extended edge channels as @xmath47 this relation is rewritten as @xmath48 here we have introduced the excess charge @xmath49 of mode @xmath14 , total capacitance @xmath50 , mutual capacitance @xmath51 , gate capacitance @xmath52 , and gate charge @xmath53 .
the expression of @xmath49 has the compensation of gate charge @xmath54 , in addition to the charge accumulation due to @xmath26 in @xmath13 ; see eq . .
from eq .
, one obtains the charging energy of the antidot system , @xmath55 where @xmath56 . by combining the single - particle energy and the charging energy in eqs . and , one has the expression of the ground - state energy of the antidot system , @xmath57 where @xmath58 is the occupation of orbital @xmath27 .
the part of @xmath59 depending on @xmath26 is @xmath60\frac{\delta b}{\delta b_\alpha}$ ] . in the regime of @xmath61
, one can treat the parameters ( such as @xmath62 and @xmath29 ) of eq . as constant over several @xmath41 , as in the constant interaction model . in the capacitive interaction model ,
one studies the change of the ground state of the system , characterized by the total numbers @xmath63 of electron occupation in bound modes @xmath14 , as a function of @xmath26 and @xmath1 .
@xmath49 is accumulated continuously as @xmath26 increases [ see eq . ] , and relaxed when resonant tunneling of electrons is allowed between a mode and an extended edge channel or internally between different modes .
the relaxation results in the transition of the ground state , i.e. , the change of @xmath64 .
first , the resonant tunneling of an electron between mode @xmath14 and an extended edge channel occurs when @xmath65 showing a resonance peak in conductance @xmath5 ; here @xmath66 is the fermi energy . on the other hand ,
the internal relaxation of an electron between @xmath14 and @xmath20 occurs when @xmath67 it occurs via direct tunneling or cotunneling mediated by virtual states . @xcite it does not cause any resonance peak in @xmath5 , but involves internal charge redistribution of the ground state ; some cotunneling processes can slightly modify @xmath5 , but we may neglect the modification .
in general , the relaxation of multiple electrons can also occur , depending on systems .
@xcite the condition for the relaxation of multiple electrons is obtained by combining eqs . and .
it is useful to draw a charge stability diagram @xcite to study the transition of the ground state ; see below . and ( b ) back - gate voltage @xmath1 dependence of electron conductance @xmath5 through the antidot . for @xmath68 ,
the data show the @xmath2 ab oscillation .
@xcite for @xmath69 , they deviate from the @xmath70 oscillation .
@xcite , title="fig:",scaledwidth=48.0% ] +
in this section , we summarize the experimental data of @xmath5 for an antidot with @xmath3 . @xcite the data disagree with the features of noninteracting electrons discussed in sec .
, the data of @xmath69 show different features from that of @xmath68 .
we first mention the dependence of @xmath5 on @xmath0 . for @xmath68
, it typically shows the @xmath2 ab oscillation @xcite that the two peaks within one ab period are almost identical and have equal peak - to - peak spacing ; see the top panel of fig .
[ ad-2-data](a ) .
it looks like the periodic structure of one peak per half magnetic flux quantum , instead of two independent peaks ( with different height and spacing ) per one flux quantum , the expectation from noninteracting electrons . for @xmath69 ,
the features of the data deviate from the @xmath8 oscillation , an extension of the @xmath2 ab oscillation .
@xcite for @xmath71 , the three peaks within one ab period are not identical .
two of them have almost the same height , but are higher than the third [ see the middle panel of fig .
[ ad-2-data](a ) ] . for @xmath72 , two peaks among the four within one ab period
are almost identical , but different from the other two ( the bottom panel ) . moreover , the dependence of @xmath5 on @xmath6 is also nontrivial .
@xcite for @xmath73 , it resembles the dependence on @xmath0 .
in contrast , for @xmath71 , @xmath74 shows only two peaks ( _ not _ three ) within one period , i.e. , two alternating peak separations within one period ; see the middle panel of fig .
[ ad-2-data](b ) .
this can not be understood by simple modification of the @xmath8 oscillation .
in this section , we analyze the features of an antidot with @xmath3 , by using the capacitive interaction model .
we study the evolution of the ground state as a function of @xmath0 or @xmath6 .
the transition of the ground state is governed by relaxations of excess charges , and gives rise to ab resonances .
we draw charge stability diagrams for the analysis , which has been widely used for the studies of a multiple quantum dot .
@xcite below , we consider the strong interaction limit of @xmath75 .
the effects due to finite @xmath76 are discussed in the next section .
we note that some features of @xmath77 discussed in this section have been already mentioned in the literature .
@xcite we here describe them in more detail ( for @xmath68 we provide a new analysis ) , and compare the features of the cases of different @xmath4 .
figure [ ad-3-evolutionnutwo](a ) shows the geometry of an antidot with @xmath68 .
it has two modes , say @xmath78 ( inner mode ) and @xmath79 ( outer ) , originating from the two spin states of the lowest landau level .
the spatial separation between @xmath78 and @xmath79 is governed by zeeman splitting ( which is enhanced by exchange interactions ) , while the separation between the modes and extended edge channels is determined by landau energy gap .
the geometry implies @xmath80 and @xmath81 .
the ground - state evolution of the antidot and the resulting ab resonances is studied by analyzing a stability diagram . for @xmath68 ,
the evolution of @xmath82 follows two different types ( i and ii ) of sequences of ab resonances , depending on how many times the internal relaxations occur per @xmath83 ; see fig .
[ ad-3-evolutionnutwo](b ) . in type
i of @xmath79-@xmath78 , the evolution never encounters the internal relaxation , and ab resonances occur sequentially by @xmath84 . in type ii of @xmath79-@xmath79 ,
the evolution passes the internal relaxation once per @xmath83 , and the ab resonances by @xmath78 disappear and are replaced by those by @xmath79 ; this effect is called the spectator behavior .
@xcite the internal relaxation is characterized by @xmath85 the ratio of energy gains between @xmath86 and @xmath87 in the internal relaxation between them ; see eq . .
[ ad-3-evolutionnutwo](b ) , @xmath88 equals to the ratio of slopes between the dash - dotted relaxation line and the evolution arrow . based on the antidot geometry mentioned above
, we find @xmath89 , therefore the relaxation occurs from @xmath86 to @xmath87 .
the internal relaxation is also governed by the inter - mode interaction strength ( @xmath90 ) . as @xmath91 and/or @xmath92 increase , the dash - dotted line in fig .
[ ad-3-evolutionnutwo](b ) becomes longer so that the evolution has more chance to pass the internal relaxation . as a result , more sequences ( with different `` initial '' values of @xmath93 at @xmath25 ) follow type ii rather than i. when the interaction strength vanishes or @xmath94 , the internal relaxation is suppressed and only type i appears .
noninteracting electrons show only type i. this feature appears in the probability @xmath95 of finding type @xmath96 in the ensemble of sequences with different values of @xmath93 at @xmath25 ; see fig .
[ ad-3-evolutionnutwo](c ) . .
( a ) antidot modes @xmath78 and @xmath79 , edge states ( solid line ) , and electron tunneling ( dashed ) .
( b ) charge stability diagram in @xmath97 plane .
each of its periodic hexagonal cells represents the ground state of @xmath98 . at
blue dashed ( red solid ) cell boundaries , ab resonances occur via tunneling through @xmath79 ( @xmath78 ) , and at green dash - dotted boundaries internal charge relaxations occur between @xmath79 and @xmath78 ; see eqs . and .
as the magnetic field @xmath26 increases , @xmath97 evolves along a line ( solid arrow ) of slope @xmath99 . depending on `` initial '' values of @xmath49 at given field @xmath25
, the evolution shows one of two sequences of ab resonances , @xmath79-@xmath78 " ( type i ) and @xmath79-@xmath79 " ( ii ) .
( c ) probability @xmath95 of finding the sequences of type @xmath100 .
( d ) sequence of ab peaks in @xmath101 for each type in ( b ) .
the modes giving the peaks are shown .
triangles represent internal charge relaxation . for ( b)(d )
, we choose @xmath102 , @xmath103 , @xmath104 ; the other parameters are given in the appendix a.,title="fig:",scaledwidth=50.0% ] + we plot the conductance @xmath105 through the antidot of the sequential tunneling regime in fig .
[ ad-3-evolutionnutwo](d ) .
it is obtained by the standard master - equation approach @xcite ( see the appendix a ) , which is enough to demonstrate the positions and relative heights of ab peaks . here
we assumed low temperature , @xmath106 , and the backward - reflection regime , in which mode @xmath14 couples to edge channel @xmath107 with strength @xmath108 as @xmath109 ; a similar regime of sequential tunneling and backward reflection will be considered for @xmath69 .
we describe the features of @xmath5 . in type i ,
each of @xmath78 and @xmath79 shows one peak per period @xmath83 .
the resulting two peaks in @xmath83 have different height , because of different @xmath108 s .
the peak height of @xmath79 is larger , since @xmath79 is the outer mode .
the separation between the two peaks depends on the initial values of @xmath93 at @xmath25 .
these features are obvious in the noninteracting limit .
as the inter - mode interaction increases , the dependence of peak separation on the initial values is weakened , and all the peak - to - peak separations become uniform . in type ii , the two peaks within @xmath83 come from @xmath79 , therefore , they show the same shape . unlike type
i , the separation between them is determined by interactions .
regardless of the initial values of @xmath110 at @xmath25 , the separations @xmath111 and @xmath112 are determined by @xmath113 . in the strong inter - mode interaction limit of @xmath114 ,
one has @xmath115 , and the separation becomes @xmath116 .
in this limit , @xmath5 shows the well - known @xmath2 ab effects , and the total energy in eq . becomes @xmath117 , where @xmath118 .
experimental data of the @xmath2 ab effects in refs .
@xcite can be explained by type ii ; see the top panels of figs .
[ ad-2-data](a ) and [ ad-2-data](b ) . on the other hand , a recent experiment by kato _
et al . _
@xcite reported that , as the magnetic field becomes tilted from the perpendicular direction to the two - dimensional system , type ii disappears , and instead type i appears . by tilting the magnetic field ,
one changes the zeeman energy , therefore controlling the initial values of @xmath119 at @xmath25 , leading to the transition between i and ii .
we consider an antidot with @xmath72 .
the geometry is shown in fig .
[ ad-4-evolutionnufour](a ) .
it has four modes , @xmath78 , @xmath79 , @xmath120 , and @xmath121 ( from the innermost one ) . @xmath78 and
@xmath79 ( @xmath120 and @xmath121 ) originate from the two spin states of the lowest ( second ) landau level .
the spatial separation between @xmath78 and @xmath79 and that between @xmath120 and @xmath121 are determined by the zeeman gap , while the other separation ( e.g. , between @xmath78 and @xmath120 ) is governed by the landau gap . for @xmath7 ,
the exchange enhancement of the zeeman gap is much smaller than the landau gap .
@xcite this implies @xmath122 .
moreover , @xmath78 and @xmath79 ( @xmath120 and @xmath121 ) can be treated almost equally so that they have the same values of @xmath41 , @xmath29 , @xmath123 , @xmath124(@xmath125 ) , and @xmath126 . for simplicity , we further approximate @xmath127 .
the symmetry between @xmath78 and @xmath79 ( @xmath120 and @xmath121 ) simplifies the stability diagram .
we introduce new definitions of @xmath128 and @xmath129 , and notice that the dependence of @xmath130 and @xmath131 on @xmath26 is negligible for @xmath80 and @xmath132 , which is valid for a usual antidot with @xmath133 . in this case , it is enough to draw the diagram in the subspace @xmath134 instead of the full space @xmath135 . @xmath130 and @xmath131 are constant within a given cell of the diagram [ see fig .
[ ad-4-evolutionnufour](b ) ] and vary only at cell boundaries by an integer multiple of charge @xmath136 . in terms of @xmath137 and @xmath138 , eq .
is rewritten as @xmath139 , where @xmath140 , @xmath141 , @xmath142 , @xmath143 , and @xmath144 .
@xmath145 and @xmath146 are capacitively decoupled from the others as shown in @xmath147 . .
each panel corresponds to that of fig .
[ ad-3-evolutionnutwo ] .
( a ) antidot modes @xmath78 , @xmath79 , @xmath120 , @xmath121 .
( b ) charge stability diagram , a periodic structure of four cells ( a ) , ( b ) , ( c ) , ( d ) in @xmath148 plane .
each cell represents the ground state of @xmath149 . as @xmath26 increases
, @xmath148 evolves along a line ( solid arrow ) of slope @xmath150 , while @xmath130 and @xmath131 are constant within a cell ; one or both of @xmath130 and @xmath131 differ by charge @xmath136 between neighboring cells .
the evolution of @xmath148 shows one possible sequence of ab resonances , e.g. , @xmath121-@xmath79-@xmath120-@xmath78 " ( type i ) , @xmath121-@xmath120-@xmath120-@xmath78 " ( ii ) , @xmath121-@xmath121-@xmath120-@xmath120 " ( iii ) .
( c ) sequence of ab peaks in @xmath101 for each type in ( b ) . for ( b and c ) , we choose @xmath104 , @xmath151 , @xmath152 , @xmath153 ; the other parameters are in the appendix a.,title="fig:",scaledwidth=47.0% ] + by analyzing the stability diagram , we find that there are three types , i , ii , iii , of resonance sequence for @xmath72 . as in the case of @xmath68 ,
the types are characterized by how many times the internal relaxations occur per @xmath154 .
the evolution passes the internal relaxation never , once , and twice in type i , ii , and iii , respectively [ fig .
[ ad-4-evolutionnufour](b ) ] .
the internal relaxations occur from @xmath155 to @xmath156 . in a similar way to @xmath157
, it is understood by the ratio @xmath88 of energy gains between @xmath158 and @xmath159 at the relaxation events , @xmath160 the geometry of edge states indicates @xmath89 , which explains the relaxation direction from @xmath155 to @xmath156 .
we note that the internal relaxation [ see eq . ]
is forbidden between modes @xmath78 and @xmath79 and between @xmath120 and @xmath121 in the symmetric case of @xmath102 and @xmath161 . in fig .
[ ad-4-evolutionnufour](c ) , we plot @xmath105 in the sequential tunneling regime . in type
i , each of @xmath78 , @xmath79 , @xmath120 , and @xmath121 shows one peak per period @xmath154 .
the resulting four peaks in @xmath154 have different height in general , because of different coupling strengths to extended edge channels .
the separation between the peaks by different modes depends on the initial values of @xmath93 .
as the inter - mode interaction increases , the peak separation becomes independent of the initial values of @xmath93 , leading to the uniform peak - to - peak separations . in type ii
, two consecutive peaks of the four peaks in one ab period show the same shape .
the two peaks come from the same mode @xmath120 or @xmath121 , and the separation @xmath162 between the two peaks depends on the interactions as @xmath163 , regardless of the initial values of @xmath110 at @xmath25 . in the strong inter - mode interaction limit of @xmath164 ,
one has @xmath165 .
this behavior results from the internal relaxation .
the position of the other two peaks depends on the initial values of @xmath93 . in type iii ,
there occur two consecutive identical peaks by @xmath120 and the other two identical peaks by @xmath121 within one ab period .
there can also occur another ab resonance sequence such as @xmath166 but with a small chance .
the peak - to - peak separation between the identical peaks is determined by @xmath162 as in type ii . for strong inter - mode interaction ,
it becomes @xmath167 .
the separation between the peaks by @xmath120 and @xmath121 depends on the initial values of @xmath93 for weak inter - mode interaction , and becomes independent of them , approaching @xmath168 as the interaction increases .
type iii is different from the @xmath169 ab effects that all the four peaks have the same shape . experimental data for @xmath72 in ref .
@xcite [ see the bottom panels of figs .
[ ad-2-data](a ) and [ ad-2-data](b ) ] show two different pairs of identical peaks within one ab period .
they can be explained by type iii such that one pair occurs by @xmath120 and the other by @xmath121 .
this indicates that electron interactions play a role in the antidot .
note that we do not exclude the possibility of type i of @xmath121-@xmath120-@xmath79-@xmath78 which could also show two different pairs of identical peaks accidently . .
( a ) antidot modes @xmath78 , @xmath79 , @xmath120 ; see fig .
[ ad-3-evolutionnutwo](a ) for details .
( b ) possible sequences of ab peaks in @xmath101 .
parameters are in the appendix a.,title="fig:",scaledwidth=50.0% ] + the @xmath71 case [ see fig .
[ ad-5-evolutionnuthree](a ) ] was previously studied in ref .
@xcite . here
we compare its features with the cases of @xmath73 .
the antidot has three modes , @xmath78 , @xmath79 , and @xmath120 ( from the innermost one ) .
the geometry implies @xmath127 and @xmath170 . as in @xmath72
, we treat @xmath78 and @xmath79 equally so that they have the same values of @xmath41 , @xmath29 , @xmath123 , @xmath171 , and @xmath126 .
in contrast to @xmath10 , the separation between modes and extended edge channels is not governed by the landau gap but by the zeeman gap in @xmath71 . due to this , ab resonances of @xmath71 have different features from @xmath172 . the symmetry between modes @xmath78 and @xmath79 simplifies the stability diagram as in @xmath72 . by introducing the definition @xmath128
, we analyze the ground - state evolution in the subspace of @xmath173 .
the analysis shows that there are three types i , ii , iii of resonance sequences , characterized by the number of the internal relaxation events within one ab period [ fig .
[ ad-5-evolutionnuthree](b ) ] .
the internal relaxations occur from @xmath155 to @xmath120 . @xcite we summarize the features of ab peaks . in type
i , each mode gives one peak per ab period , showing three different peaks . in type ii , two of the three peaks are identical , coming from @xmath120 , and their peak height is larger than the other . in type iii ,
all three peaks are identical , coming from @xmath120 .
the separation @xmath174 between the identical peaks is determined by the interactions as @xmath175 . in the strong inter - mode interaction limit of @xmath164 , @xmath176 . in this limit ,
type iii shows the @xmath177 ab effects .
the dependence of @xmath5 on @xmath26 experimentally measured in ref . @xcite
[ the middle panels of fig .
[ ad-2-data](a ) ] can be understood by type ii ; we do not exclude the possibility of type i that two modes among the three accidently give the peaks of almost equal shape .
we remark that the above description with @xmath104 fails to explain @xmath74 found in ref .
@xcite where only two peaks appear in one ab period of @xmath6 [ see the middle panels of fig .
[ ad-2-data](b ) ] .
this behavior shows an obvious contrast to the case of @xmath10 , in which there appear @xmath4 peaks per one ab period of @xmath6 . to understand this difference
, we need to take the effect of finite single - particle level spacing , which is the subject of the next section .
in the previous section , the analysis of ab effects was restricted to the regime of @xmath178 . in a general situation ,
single - particle level spacing @xmath29 is not ignorable .
one expects that interaction effects will become reduced as @xmath29 increases . in this section ,
we study how finite @xmath29 modifies the interaction effects found in sec .
we study the effect of finite @xmath29 by assuming linear dispersion @xmath179 .
then , the first term of the ground - state energy in eq . is written as @xmath180
. it is quadratic in @xmath39 .
it is absorbed into the interaction term of eq . so that the total energy has the same form as the interaction term , @xmath181 , but with modification @xmath182 here , @xmath183 satisfies @xmath184 .
we discuss the results of the modification .
first , in eq . , @xmath29 effectively enhances the self - interaction of mode @xmath14 , without modifying intermode interactions . as a result , as @xmath29 increases , the internal relaxations between different modes ( therefore type ii and iii )
become suppressed .
the suppression is shown for @xmath71 in fig .
[ ad-6-backgate](a ) .
second , @xmath29 also affects @xmath49 in eq . .
it should be emphasized that it does not alter the dependence of @xmath93 on @xmath26 .
hence an antidot with finite @xmath29 shows the same dependence of ab peaks on @xmath26 as the case of @xmath104 .
however , @xmath29 _ does _ affect the dependence of @xmath93 on @xmath6 . in the stability diagram
[ see fig .
[ ad-6-backgate](b ) ] , the evolution follows along a line of slope @xmath185 .
the expression of @xmath186 is @xmath187 where @xmath188 for @xmath157 and @xmath189 for @xmath69 . here
, we have put the same value of @xmath190 into @xmath29 s . .
( a ) probability @xmath191 of finding type i , ii ( iia + iib ) , iii ( see fig .
[ ad-5-evolutionnuthree ] ) for @xmath104 ( upper panel ) and @xmath192 ( lower ) .
( b ) evolution of @xmath173 as a function of the _ back - gate voltage _ for @xmath104 ( dashed arrow ) and @xmath192 ( solid ) .
the charge stability diagram has a periodic structure of two cells ( a ) , ( b ) .
each cell represents the ground state of @xmath194 . for ( a ) and
( b ) , we choose @xmath195 , @xmath196 , and @xmath197 for cell ( a ) and @xmath198 for ( b).,title="fig:",scaledwidth=47.0% ] + for a @xmath71 antidot with finite @xmath29 comparable with @xmath199 , @xmath200 can be much smaller than the value of @xmath104 .
it is because of the geometrical feature [ see fig .
[ ad-5-evolutionnuthree](a ) ] that the spatial separation of @xmath120 from extended edge channels is determined by the zeeman gap , while that of @xmath78 is determined by the landau gap .
this feature implies @xmath201 .
the evolution line with small @xmath202 can pass only the solid cell boundaries in the stability diagram , showing paired peaks by @xmath78 and @xmath79 over several periods ; see fig .
[ ad-6-backgate](b ) . or
, depending on the initial value of @xmath203 at @xmath25 , it can pass only the dashed cell boundaries , showing paired peaks by @xmath120 .
hence , the sequence of ab peaks shows only two peaks within one period .
this behavior may explain the puzzling features of the experimental data @xcite in the middle panel of fig .
[ ad-2-data](b ) .
in contrast , in the cases of @xmath73 , there is no drastic change of @xmath186 when @xmath29 changes from 0 to @xmath204 . unlike @xmath71 , these cases have the common geometrical feature that the separation between the outermost mode and edge channels is governed by the landau splitting , which implies @xmath81 for @xmath68 and @xmath205 for @xmath72 . with these parameters of capacitances , one notices @xmath206 .
therefore @xmath74 shows @xmath4 peaks within one ab period , similar to the dependence on @xmath26 .
this finding emphasizes the role of the electron interaction between an antidot and extended edge channels as well as the competition between electron interactions and single - particle physics in an antidot system .
we have developed a capacitive interaction model for an antidot system in the integer quantum hall regime , and applied it to an antidot with @xmath3 .
the predictions of the model agree with the various features of @xmath101 and @xmath74 observed in experiments .
this may support the validity of our model , and show the importance of electron interactions in the antidot .
we summarize the features of an antidot predicted by our model .
the common features are ( i ) charging effects , ( ii ) internal relaxations from an inner mode to an outer one ( @xmath89 , independent of @xmath4 ) , and ( iii ) @xmath4 resonance peaks in @xmath105 within one period ; in other geometries , internal relaxations can occur in the opposite way .
@xcite for larger @xmath4 , more different types of sequences of the peaks can appear .
as the interaction strength increases , the peaks within one period become correlated , i.e. , some of them can have the same shape ( coming from the same mode ) and the peak - to - peak separation becomes equally spaced . for @xmath69 ,
the peaks show the features deviated from the @xmath207 oscillation . on the other hand
, the feature of @xmath208 depends on @xmath4 . for @xmath10 ,
@xmath208 shows @xmath4 peaks within one period , while for @xmath71 , it can show only two peaks .
this @xmath4 dependence comes from the geometrical difference of the spatial separation between the antidot and extended edge channels as well as the competition between the electron interactions and the single - particle level spacing . in other types of quantum hall interferometers , such as fabry - perot resonators
, electron interactions will also play an important role , since the interferometers utilize localized quantum hall edge states . some puzzling features ( e.g. , a checkerboard pattern ) of electron conductance through a fabry - perot resonator have been experimentally observed .
@xcite they may be originated from electron interactions .
some parts of the features can be explained by using a capacitive interaction model which was proposed recently , @xcite but the features have not been fully understood yet . to understand them
, it will be interesting to extend our model to the fabry - perot resonator .
we thank c.j.b .
ford , v.j .
goldman , and m. kataoka for discussion , and financial support by national research foundation ( grant no .
2009 - 0078437 ) .
we provide the derivation of electron conductance @xmath5 through the antidot with @xmath157 in the sequential tunneling regime of @xmath209 .
it uses the master - equation approach .
@xcite the derivation is trivially extended to @xmath69 , so we do not provide the extension . in the antidot system
[ see fig .
[ ad-3-evolutionnutwo](a ) ] , electron current @xmath210 by small bias voltage @xmath211 between the source and drain is decomposed into three contributions , @xmath212 .
@xmath213 ( @xmath214 ) comes from the resonant forward ( backward ) scattering of an electron from edge channel @xmath215 or @xmath216 to @xmath217 ( @xmath218 ) via antidot bound modes , while @xmath219 comes from electrons propagating along the extended edge channels without any scattering by the antidot ; @xmath220 is a spin index .
the conductance @xmath5 can be defined by excluding the trivial contribution of @xmath219 , @xmath221 .
@xmath213 ( @xmath214 ) is the net flux of charge @xmath136 across the tunneling barrier between the antidot and edge channels @xmath217 ( @xmath218 ) , @xmath222 where @xmath223 .
@xmath224 is the probability of finding the antidot in the ground state of @xmath98 , and @xmath225 varies over the ground - state configurations .
@xmath226 is the rate of the transition from @xmath225 to @xmath227 via electron tunneling between the antidot and the channel @xmath217 or @xmath218 injected from the drain .
the possible values of @xmath227 are @xmath228 and @xmath229 , and @xmath230 counts the net electrons tunneling from the antidot to the channels @xmath217 and @xmath218 at the transition .
@xmath226 is expressed as @xmath231 , @xmath232 $ ] , etc . , where @xmath233 , @xmath234 , @xmath235 , @xmath236 is the change of antidot energy at the transition , e.g. , @xmath237 , and @xmath238 .
@xcite the probability @xmath224 satisfies the detailed balance equations in the stationary situation of @xmath239 , i.e. , @xmath240 = 0 $ ] . here , we define the total transition rate @xmath241 , where @xmath242 is the transition rate of the antidot ground state via electron tunneling between the antidot and the channel @xmath215 or @xmath216 injected from the source . in the linear - response regime , the detailed balance equations can be solved by imposing the approximate form @xmath243 $ ] , where @xmath244 / ( k_\mathrm{b}t)}$ ] is the equilibrium distribution , @xmath245 is determined by @xmath246 , and @xmath247 satisfies the equation , e.g. , @xmath248 . substituting @xmath224 into eq . , and totally linearizing eq . in terms of @xmath211
, we find the formula , @xmath249 @xmath250 we note that eq .
is derived in the case where the internal relaxation between antidot modes does not occur around the central position of ab peaks ( within peak broadening ) , thus it does not affect the shape of ab peaks . finally , we mention the parameters used in the figures .
we choose tunneling strengths @xmath108 , assuming that antidot modes are symmetrically coupled to upper and lower edge channels of the system , and that the antidot is in the backward - reflection regime of @xmath251 . in fig .
[ ad-3-evolutionnutwo](d ) , we choose @xmath252 , @xmath253 , and @xmath254 . in fig .
[ ad-4-evolutionnufour](b ) , @xmath255 , @xmath256 , @xmath257 , @xmath258 for ( a ) , ( b ) , ( c ) , ( d ) , respectively . in fig .
[ ad-4-evolutionnufour](c ) , @xmath259 , @xmath260 , @xmath261 , @xmath262 , @xmath263 , and @xmath264 . in fig .
[ ad-5-evolutionnuthree](b ) , @xmath104 , @xmath265 , @xmath266 , @xmath267 , @xmath268 , @xmath269 , @xmath270 , @xmath271 , and @xmath272 , @xmath273 .
y. aharonov and d. bohm , phys . rev .
* 115 * , 485 ( 1959 ) ; * 123 * , 1511 ( 1961 ) .
sim , m. kataoka , and c. j. b. ford , phys . rep . * 456 * , 127 ( 2008 ) . c. j. b. ford , p. j. simpson , i. zailer , d. r. mace , m. yosefin , m. pepper , d. a. ritchie , j. e. f. frost , m. p. grimshaw , and g. a. c. jones , phys .
b * 49 * , 17456 ( 1994 ) . m. kataoka , c. j. b. ford , g. faini , d. mailly , m. y. simmons , and d. a. ritchie , phys . rev .
b * 62 * , r4817 ( 2000 ) . m. kato , a. endo , s. katsumoto , and y. iye , j. phys . soc
78 * , 124704 ( 2009 ) . i. j. maasilta and v. j. goldman , phys . rev .
b * 57 * r4273 ( 1998 ) .
m. kataoka , c. j. b. ford , g. faini , d. mailly , m. y. simmons , d. r. mace , c .- t .
liang , and d. a. ritchie , phys .
83 * , 160 ( 1999 ) .
m. kataoka , c. j. b. ford , m. y. simmons , and d. a. ritchie , phys .
. lett . * 89 * , 226803 ( 2002 )
. v. j. goldman , j. liu , and a. zaslavsky , phys .
b * 77 * , 115328 ( 2008 ) . c. gould , a. s. sachrajda , m. w. c. dharma - wardana , y. feng , and p. t. coleridge , phys . rev .
lett . * 77 * , 5272 ( 1996 ) .
i. karakurt , v. j. goldman , j. liu , and a. zaslavsky , phys .
lett . * 87 * 146801 ( 2001 ) .
m. kato , a. endo , s. katsumoto , and y. iye , phys .
lett . * 102 * , 086802 ( 2009 ) .
s. ihnatsenka and i. v. zozoulenko , phys .
b * 74 * 201303(r ) ( 2006 ) . s. ihnatsenka , i. v. zozoulenko , and g. kirczenow , phys . rev .
b * 80 * , 115303 ( 2009 )
sim , m. kataoka , h. yi , n. y. hwang , m .- s .
choi , and s .- r .
eric yang , phys .
* 91 * , 266801 ( 2003 ) . w .-
r . lee and h .- s .
sim , phys .
* 104 * , 196802 ( 2010 ) .
n. y. hwang , s .-
eric yang , h .- s .
sim , and h. yi , phys .
b * 70 * , 085322 ( 2004 ) .
n. ofek , a. bid , m. heiblum , a. stern , v. umansky , and d. mahalu , proc .
usa * 107 * , 5276 ( 2010 ) .
y. zhang , d. t. mcclure , e. m. levenson - falk , c. m. marcus , l. n. pfeiffer , and k. w. west , phys .
b * 79 * , 241304(r ) ( 2009 ) .
f. e. camino , w. zhou , and v. j. goldman , phys .
b * 72 * , 155313 ( 2005 ) ; * 76 * , 155305 ( 2007 ) ; p.
v. lin , f. e. camino , and v. j. goldman , _ ibid . _ * 80 * , 125310 ( 2009 ) .
w. g. van der wiel , s. de franceschi , j. m. elzerman , t. fujisawa , s. tarucha , and l. p. kouwenhoven , rev .
phys . * 75 * , 1 ( 2002 ) .
note that the rate of the cotunneling processes in eq .
can be estimated by fermi s golden rule . " for example , the rate of elastic cotunneling processes is obtained as @xmath274 , where @xmath275 is the antidot charging energy , @xmath276 is the broadening of ab resonances , and @xmath277 is the thermal energy . for an antidot with @xmath278 @xmath136v , @xmath279 @xmath136v , and @xmath280 mk , one obtains @xmath281 @xmath136v .
the time scale of the relaxation via the cotunneling processes ( including spin - flip processes ) is @xmath282 sec .
c. w. j. beenakker , phys .
b * 44 * , 1646 ( 1991 ) .
w. xu , p. vasilopoulos , m. p. das , and f. m. peeters , j. phys . : condens .
matter * 7 * , 4419 - 4432 ( 1995 ) .
b. rosenow and b. i. halperin , phys .
* 98 * , 106801 ( 2007 ) .
|
we derive a general capacitive interaction model for an antidot - based interferometer in the integer quantum hall regime , and study aharonov - bohm resonances in a single antidot with multiple bound modes , as a function of the external magnetic field or the gate voltage applied to the antidot .
the pattern of aharonov - bohm resonances is significantly different from the case of noninteracting electrons .
the origin of the difference includes charging effects of excess charges , charge relaxation between the bound modes , the capacitive interaction between the bound modes and the extended edge channels nearby the antidot , and the competition between the single - particle level spacing and the charging energy of the antidot .
we analyze the patterns for the case that the number of the bound modes is 2 , 3 , or 4 .
the results agree with recent experimental data .
|
the ratio between transverse and longitudinal components of kinematical observables is a measure of the conversion of the initial entrance channel motion into intrinsic degrees of freedom in heavy - ion collisions ( hics ) .
such an observable gives an insight on the rate of a system s equilibration , of the dissipation of the available energy , as well as of hic stopping power @xcite .
thanks to such an observable , the fopi collaboration has evidenced partial nuclear transparency in hic in the beam - energy range @xmath70.1@xmath31@xmath3 gev @xcite .
more recently , by examining the ratio of transverse to longitudinal energy @xmath0 and linear momentum @xmath8 for the most violent hics , the indra collaboration has revealed a substantial reduction of the nuclear stopping power at @xmath710@xmath3100@xmath3 mev@xcite .
this stopping observable reaches a minimum around the fermi energy @xmath5 and stagnates or very weakly increases with the further increase of @xmath9 at least up to 100@xmath3 mev , the upper limit of the energy range available in this study .
the above observation is valid for all ( mass symmetric ) systems studied , with system masses @xmath10=72394 a.m.u .
it is worth emphasizing that the fusion cross section normalized by the total reaction cross section exhibits an analogously rapid fall - off up to about @xmath5 @xcite , a behavior especially evident for mass - symmetric systems ( cf.fig .
6 of ref .
@xcite ) . in a recent publication @xcite the above observable @xmath0
was analyzed for the @xmath11=1 subset of the same indra data .
the @xmath11=1 @xmath0 displays a slightly stronger increase with @xmath12 for the heavier systems@xcite relative to the @xmath0 values obtained in the previous study @xcite which included light charged particles and fragments , but also was somewhat more stringent on the selection of the most central events .
the authors of ref .
@xcite report a minimum of @xmath0 around @xmath5 , which is particularly enhanced when @xmath0 is normalized to the fermi - gas - model prediction of the incoming @xmath0 value at a given @xmath9 .
@xcite in ref .
@xcite it was assumed that protons are predominantly dynamically emitted during the early reaction phase , in accordance with refs .
such a hypothesis offers a possibility of extracting information on the in - medium correction for the free nucleon - nucleon ( @xmath4 ) cross section @xmath13 .
following such an argument , starting from the experimental @xmath0 values in ref .
@xcite , with some basic assumptions about the effects owing to the pauli - exclusion principle , the nucleon mean free path was extracted and an effective value of the in - medium @xmath4 cross section @xmath14 was deduced . in the process , a correction factor @xmath6 was obtained by which @xmath13 has to be multiplied at each @xmath9 to get a proper @xmath14 value .
the authors found that ( i ) a _ significant reduction _ of @xmath13 is present in hics below 100@xmath3 mev and ( ii ) this change of @xmath13 is strongly dependent upon @xmath9 . at the lowest energies
the measured @xmath0 is compatible with the full stopping value ( @xmath151 ) and the effective @xmath14 amounts to about 0.4@xmath13 .
one should keep in mind that the authors claimed a large uncertainty on the factor @xmath6 below @xmath1630@xmath3 mev , a subject for which they have announced a devoted publication @xcite . at incident energies around @xmath5 where @xmath0 attains its minimum ,
@xmath14 is reduced to less than one fifth the free @xmath13 value ( @xmath6=0.17 ) and then the effective @xmath14 steadily and regularly increases up to half of the free value ( @xmath170.5 ) at @xmath9=100@xmath3 mev @xcite ( see also fig .
[ cf ] in the present work ) .
the stopping observables @xmath0 and/or @xmath8 have also been investigated in isospin - dependent quantum molecular dynamics ( iqmd ) @xcite and antisymmetrized molecular dynamics ( amd ) @xcite model studies of hics .
all these works were carried out before publication of ref .
neither of simulation approaches predicts the remarkable in - medium reduction of @xmath13 found in ref .
@xcite . in the amd study ,
specific attention has been paid to performing the analysis by meticulously following the experimental procedure for data handling @xcite . the simulation with in - medium @xmath18 due to li and machleidt @xcite ( the free @xmath18 ) undershoots ( overshoots ) the data of @xcite .
an agreement with the data could only be reached at @xmath1980@xmath3 mev by doubling the theoretically established @xmath14 of @xcite .
a systematic investigation of the impact of @xmath18 on @xmath0 , however , has not been performed yet .
the intention of the present study is twofold : \1 . by varying the nuclear equation of state ( eos ) and the parametrization of @xmath18 , to investigate how the semiclassical landau - vlasov ( lv ) transport model of hic@xcite complies with the experimentally deduced dependence of the stopping observable @xmath0 on @xmath9 and \2 . by varying a simple multiplicative factor @xmath6 of the free @xmath4 cross section , to compare
thus obtained values for @xmath6 with those reported in ref .
within the semiquantal extension of the boltzmann transport theory , the highly nonlinear lv equation governs the spatio - temporal evolution of the one - body density distribution function @xmath20 : @xmath21 which gives the probability of finding at the instant @xmath22 a particle in the phase - space point @xmath23 .
@xmath24 stands for the poisson bracket , whereas @xmath25 is the one - body hamiltonian describing the coulomb potential and the nuclear mean field .
we present the results obtained with a soft nonlocal mean field labeled d1-g1 ( @xmath26=228 mev , @xmath27=0.67 ) due to gogny @xcite and those obtained with the standard simplification of the soft zero - range skyrme interaction due to zamick ( @xmath26=200 mev , @xmath27=1.0)@xcite .
the d1-g1 force is reputed to reproduce fundamental properties of nuclear matter as well as those of finite nuclei @xcite while the zamick parametrization of the eos is , owing to its simplicity , of rather widespread use in a number of microscopic approaches .
details on both the nonlocal and the local parametrizations of the used eos may be found in tables i and iii of ref.@xcite , respectively .
we have demonstrated that the lv model is able to correctly describe experimental observations in the intermediate energy regime @xcite .
the use of only a density dependent eos is legitimated by the finding @xcite that the isospin dependence of the mean field has a weak , if any , influence on isotropy ratios .
experimental @xmath0 s for a number of hics between various xenon and tin isotopes corroborate this result ; cf .
table i of ref .
@xcite .
the function @xmath28 is expanded onto a moving basis of coherent states taken as normalized gaussians @xmath29 ( @xmath30 ) with frozen width @xmath31 ( @xmath32 ) in @xmath33 ( @xmath34 ) space : @xmath35 @xmath3 is the system mass number and @xmath36 is the number of coherent states ( @xmath37 equals 60 in the present study ) .
the widths @xmath31 and @xmath32 are chosen such as to best reproduce the nuclear ground state characteristics of the two colliding nuclei .
the local density reads @xmath38 gaussians move in the self - consistent mean field and suffer hard scattering between them , controlled by the uehling - uhlenbeck collision integral accounting for the fermionic character of interacting particles@xcite : @xmath39 which takes into account energy and momentum conservation as well as the pauli exclusion principle . here
, @xmath40 denotes the nucleonic mass , @xmath41 is the occupation number with @xmath42 the spin - isospin degeneracy@xcite , @xmath43 and @xmath44 ( @xmath45 and @xmath46 ) are initial ( final ) momenta of the scattering particle pair , @xmath47 is the single - particle energy , while @xmath14 is the _ in - medium _ nucleonic cross section .
@xmath14 is scaled so that a gaussian - averaged mean - free path is the same as for a nucleon .
the cross section dependence on isospin has been reported as crucial for the study of stopping @xcite .
thhis kind of @xmath18 parametrization was proposed by chen _
et al . _
@xcite , which hereafter we label @xmath48 .
this phenomenological @xmath13 is based on the empirical isospin and energy dependence of the free @xmath4 scattering and has been used in both @xcite and @xcite .
@xmath18 due to li and machleidt @xcite , which accounts for the in - medium effects and is also isospin dependent , is tested too .
our stopping observable , the energy - based isotropy ratio , is defined as the ratio between transverse @xmath49 and longitudinal @xmath50 energy components of reaction ejectiles @xcite @xmath51 where summation runs over particles of those reaction events that satisfy certain selection criteria . for the lv simulation results , the summation index @xmath52 of eq .
( [ r_e ] ) runs over the free gaussians , i.e. , those which are not bound in large ( residue - like ) fragment(s ) . among the experimentally studied systems , the @xmath1xe+@xmath2sn reaction has been measured at by far the most abundant number of @xmath9 values@xcite .
consequently , in the present work only the simulation of this system is performed . to acquire stable @xmath53 values ,
the simulation is carried out up to 600 fm/@xmath54 at the lowest @xmath9=12@xmath332@xmath3 mev and up to 240 fm/@xmath54 at the highest @xmath9=80@xmath3100@xmath3 mev . beyond that time
the calculation was continued until 8000 fm/@xmath54 , considering only the coulomb repulsion due to reaction residues .
special care was taken in order to perform our analysis of simulation data as close as possible to experimental conditions . in the experimental analysis
@xcite , event selection is based on the charged particle s multiplicity .
the authors selected the most central events that are estimated , in cross - section units , to be equivalent to 50 to 150 mb@xcite .
we adopt the median value of 100 mb for our analysis .
this amount corresponds to about 2% of the total reaction cross section @xmath55 and , in a geometrical sharp - cut approximation , to @xmath562.0 fm .
consequently , in this subsection our simulation is limited to @xmath562.0 fm .
figure [ snn ] displays @xmath53 obtained with the momentum - dependent d1-g1 eos ( upper panel ) and with the zero - range zamick eos ( lower panel ) for several parametrizations of the in - medium corrected @xmath18 . for comparison ,
the experimental @xmath57 s are shown by filled circles with the corresponding errors @xcite . as a reference
, the @xmath53 results obtained with the in - medium _ uncorrected _ empirical free scattering @xmath58@xcite are displayed by the heavy dotted curves .
this empirical @xmath18 is used as an input for the in - medium modified @xmath14 suggested by cugnon _
et al . _
@xcite . in their brueckner @xmath59-matrix in -
medium renormalization of the @xmath4 interaction , they obtained a set of parameters explicitly describing the dependence of @xmath60 on the local density@xcite .
these simulation results are displayed by the red curves and reddish zone in fig .
[ snn ] : the zone shows the range of the @xmath53 values limited by the impact parameters @xmath61=1 fm ( dashed bordering curve ) and @xmath61=2 fm ( full bordering curve ) .
( @xmath61=0 fm has no weight and @xmath53 at most of energies is roughly the same for @xmath61=0 and 1 fm . )
the heavy curve in each zone represents the @xmath61-weighted @xmath53 value in the range @xmath61=02 fm and corresponds to 2% of @xmath55 . for both eos ,
the @xmath53 values with @xmath60 are very similar to those obtained with @xmath48 ( dotted curves ) .
clearly , in the full energy range investigated here the in - medium effects of @xmath60 have rather weak impact on the @xmath0 observable .
consequently , as for @xmath48 , the compatibility of @xmath53 and @xmath57 for both eos may be observed at the lowest @xmath9 only when the experimental errors are accounted for .
in addition , for the zamick eos , fig .
[ snn]b ) , the simulation strongly overshoots the data at the highest @xmath9 s .
a full _ ab initio _ microscopic study of @xmath14 based upon the dirac - brueckner approach to nuclear matter was performed by li and machleidt@xcite . besides dependence on energy , isospin , and density of @xmath60 for this @xmath62
, we have added an explicit dependence on angle .
in contrast to the scattering of neutrons , which is taken as isotropic , those between neutron and proton @xmath63 and between protons @xmath64 are anisotropic in accordance with the fit of ref .
@xcite , which is given in detail in the appendix .
similarly to above , the corresponding values of @xmath53 are displayed by the blue curves and bluish zone in fig.[snn ] .
again , the compatibility of @xmath53 with @xmath57 is unsatisfactory . nevertheless , for the d1-g1 eos and @xmath65 , fig .
[ snn](a ) , the slope of the isotropy ratio excitation function is correct but the simulation somewhat undershoots the experimental points : @xmath53 may be taken as compatible with the lower edges of experimental errors on @xmath57 .
for the zamick eos , fig .
[ snn](b ) , the compatibility with @xmath57 exists at low @xmath9 and around @xmath1660@xmath3 mev , but the general features of the data are poorly reproduced .
manifestly , none of the above parametrizations of @xmath14 and eos can account for the observed behavior of the @xmath0 stopping observable in the full energy range .
the parameters in the above @xmath14 are of a fixed value . by an expansion around the saturation value @xmath66 , klakow _ et al .
_ have suggested a simple parametrization for the dependence of @xmath14 on the evolving nuclear density @xcite , @xmath67 where @xmath68 is evaluated locally according to eq . ( [ ro ] ) , and @xmath69 is a free parameter assumed to reduce the cross section , thus it is strictly negative .
as before , for @xmath13 the value taken is the empirical @xmath48 .
the authors have recommended for @xmath69 the domain [ 0.3 , 0.1 ] @xcite . in our simulation @xmath69
is varied between 0.1 and 0.6 .
these @xmath53 are presented in fig .
[ snn ] by the thinner dashed curves with variable dash size .
they display a more or less regular dependence on both @xmath9 and @xmath69 .
for the nonlocal eos and 0.6@xmath70@[email protected] the @xmath57 values at @xmath65 are well reproduced in both slope and absolute value ; cf.fig .
[ snn](a ) . at
energies higher than @xmath5 , however , for each @xmath9 another and regularly increasing value of the parameter @xmath69 is required such that , at the highest @xmath9 here considered , it should become positive , implying an in - medium enhancement rather than a reduction of @xmath13 at @xmath9@xmath7180@xmath3 mev .
let us mention that @xmath53 with @xmath60 corresponds to that of @xmath72 with @xmath69=0.1 in the full range of @xmath9 considered and for both eos .
simulation results with @xmath62 and d1-g1 eos are compatible to @xmath72 with @xmath69=0.6 and @xmath9@xmath7150@xmath3 mev . for the zamick eos of fig .
[ snn](b ) one does not find a range of @xmath9 of stable value of the parameter @xmath69 that gives @xmath53 s compatible with either @xmath57 or those @xmath53 due to @xmath62 . in conclusion
, neither choice of @xmath14 allows for a unique description of experimental observation .
one faces the fact that every model study , ours and previous@xcite , fails to reproduce with a single set of parameters the indra experimental results in the full energy range studied @xcite .
in particular , all models but @xcite predict steadily decreasing values of @xmath53 when @xmath9 increases , while experimental @xmath57 results display a break in the slope around the fermi energy @xmath5 .
being clearly unable to reproduce the experimental data with different parametrizations of the residual @xmath4 cross section , with or without momentum dependence of the force , let us concentrate on our second task that is , by following ref .
@xcite , to infer the multiplicative factor @xmath6 between the in - medium @xmath4 cross section @xmath14 and the free @xmath13 one : @xmath73 as previously done and as in @xcite , we take @xmath58@xcite . of course , this simple cross - section normalization factor @xmath6 can not completely describe the rather complex modification of the free @xmath4 interaction occurring in the nuclear medium .
in particular , such a @xmath14 is frozen during a reaction course and depends only indirectly on @xmath9 . nevertheless , the prescription of eq .
( [ sig_m ] ) allows one to get an insight into the global in - medium effects on nuclear medium stopping properties and enables a comparison of the factor @xmath74 obtained in our simulation with @xmath75 of ref .
@xcite .
figure [ sf ] displays @xmath53 as a function of @xmath9 and the @xmath4 cross - section factor @xmath6 for the two effective interactions . in the d1-g1 eos case ,
[ sf](a ) , the parameter @xmath6 takes values 0.2 , 0.5 , 0.8 , 1.0 , 1.2 , and 1.5 . for the zamick eos ,
[ sf](b ) , it is varied between 0.1 and 0.8 in steps of 0.1 . ) , which is equivalent to taking the parameter @xmath6=0 . ] as in fig .
[ snn ] , @xmath53 are for central hic with @[email protected] fm , where @xmath61=1 fm ( 2 fm ) results are represented by the thin dashed ( full ) curves that boarder the ( colored ) zone of each of the @xmath6 values .
as before , the heavy curve in each zone shows the @xmath61-weighted @xmath53 that corresponds to 2% of @xmath55 .
@xmath53 displays a regular dependence on @xmath9 and @xmath6 . in accordance with expectation and
corroborating the results of fig .
[ snn ] , higher @xmath14 ( larger @xmath6 ) implies higher stopping power of hics .
unlike experimental @xmath57 and like our results of fig .
[ snn ] , as well as of a number of previous theoretical works @xcite , the lv - simulation @xmath53 steadily decreases with @xmath9 for all @xmath6 without a minimum around @xmath5 . at the lowest @xmath9
s the mean field completely dominates the course of the collision , and for each @xmath6 value and both eos @xmath53 is compatible with @xmath57 . for @xmath7745@xmath3 mev
, @xmath57 is well reproduced by the @xmath6=0.5 curve [ fig.[sf](a ) ] and by the @xmath6=0.1 one [ fig.[sf](b ) ] , respectively .
again , a single value of @xmath6 can not reproduce experimental results .
however , similarly to the case of the parameter @xmath69 of eq .
( [ kwb ] ) , by allowing @xmath6 to change with @xmath9 one may find a set of @xmath6 values to achieve an agreement between @xmath53 and @xmath57 .
the behavior of both the parameter @xmath69 and the factor @xmath6 with @xmath9 corroborates the experimental finding @xcite that the effective in - medium cross section @xmath14 drastically changes with @xmath9 and that around @xmath5 there is a break in this dependence . as a function of incident energy for the central xe+sn reaction and several values of the @xmath13 scaling factor @xmath6 of eq .
( [ sig_m ] ) .
upper ( lower ) panel shows results obtained with the d1-g1 ( zamick ) eos .
the colored zones and curves have the same meaning as in fig . [ snn ] , but here for the scaled @xmath18 of chen _ et al .
_ @xcite . for more details
see the caption of fig .
[ snn ] and the text .
[ sf ] ] we take the @xmath61-weighted @xmath53 as the starting point to infer information about the correction factor @xmath6 by which one would have to multiply @xmath48 to comply with @xmath57 .
the procedure is evidenced in the inset of fig .
[ cf ] in which the d1-g1 eos at 50@xmath3 mev is shown as an example . the horizontal red line and reddish background zone display the @xmath57 value and its uncertainty , respectively , at 50@xmath3 mev .
blue circles joined by a broken line are the lv simulation @xmath53 as a function of @xmath6 at the same energy .
the crossing of this broken line with the red line and the edges of the reddish zone give the most appropriate value for the factor @xmath6 of eq .
( [ sig_m ] ) and its uncertainty , respectively . in the main panel of fig .
[ cf ] we show , by the open circles and squares joined by dashed curves , the thus obtained @xmath6 values plotted against @xmath9 for the d1-g1 and zamick eos , respectively . within experimental errors , the @xmath53 values for @xmath7720@xmath3 mev are roughly compatible with any @xmath78 value and are not reported . the lv model with the highly recommended nuclear interaction d1-g1 for the range of energies of the present study and with the empirical @xmath4 cross section @xmath48 predicts , for all energies studied , about twice higher @xmath6 values compared to those suggested by fig .
10 of ref .
@xcite ; these are presented in fig .
[ cf ] as black filled circles , with the gray area showing their uncertainties .
in contrast to this , when experimental and simulation uncertainties are accounted for , the zero - range ( _ local _ ) skyrme interaction in the zamick implementation @xmath74 is compatible with @xmath75 above @xmath735@xmath3 mev .
let us underline that @xmath74 for both eos display a minimum around @xmath5 .
the minimum is relatively more pronounced than the one suggested by @xmath75 and it is somewhat shifted in energy .
the zamick eos gives a @xmath74 that reduces the free @xmath18 at all @xmath9 while the d1-g1 eos gives @xmath791 at @xmath8080@xmath3 mev . of eq.([sig_m ] ) as a function of @xmath9 .
open circles and squares interpolated by the dashed curve display @xmath74 values obtained for @xmath53 evaluated in the @xmath61=02 fm range using the d1-g1 eos and zamick eos , respectively .
open triangles , upright and reversed , interpolated by the dash - dotted curve denote the gaussian weighted @xmath53 evaluated in the interval @xmath61=05 fm for the d1-g1 eos and zamick eos , respectively . the symbols ( but not the curves )
are slightly shifted in @xmath9 to avoid overlapping error bars .
full dots , full curve , and gray zone represent @xmath75 values and their uncertainty deduced from the experimental @xmath57 @xcite .
all curves are merely intended to guide the eye .
the inset explains the procedure used to extract the values of @xmath74 . for more details
[ cf ] ] the most evident difference between a simulation and an experimental data analysis is in the reliability of the assessments of reaction impact parameter @xmath61 .
experimental selection of the most central collisions is made by assuming that there is a biunivocal correspondence between the reaction violence , i.e. , the multiplicity of particles in a reaction event , and the reaction centrality . in a simulation
the centrality is an input variable , thus it is under full control . in comparing simulation results and the earlier indra study of @xmath0 and @xmath8
@xcite it has been underlined that selecting events via multiplicity strongly mixes events of different impact parameters over a rather broad span in @xmath61@xcite .
thus , let us examine the @xmath61 vs multiplicity relationship and its influence on the isotropy ratio .
for that purpose we use the semidynamical general - purpose event - generating code hipse ( heavy - ion phase - space exploration ) intended to describe hics at intermediate energies@xcite . at each energy
100000 events are generated in the range @xmath61=07 fm .
let us note that , according to the expression of @xmath55 in ref .
@xcite , the above range in @xmath61 is equivalent to [email protected]@xmath55 , depending on @xmath9 . at @xmath9=50@xmath3
mev the simulation was performed in the full impact parameter range of the @xmath1xe+@xmath2sn reaction , i.e. , @xmath61=013 fm , in order to verify that in the non - covered range ( @xmath61=713 fm ) the high multiplicity events , in which we are interested , are not present . by passing the generated events through a sophisticated indra - device geometry and detection - acceptance filter@xcite we found that it has no appreciable effect on the @xmath11=1 @xmath0 values .
mostly , the change in @xmath0 due to this filter is below 0.5% .
xe+@xmath2sn reaction with the hipse code @xcite . shown
are the @xmath61 distributions of the full data set ( black line ) and of the high multiplicity subset ( filled and hollow red - line histograms ) , which is by number of events equivalent to the one of @xmath61=02 fm ( hatched part of the black histograms ) for four @xmath9 .
the curves are the best fits by a gaussian function . for more details
[ eq_m],width=294 ] selecting the range @xmath61=02 fm , in a geometrical sharp - cut approximation , corresponds to 1.82% to 1.92% of @xmath55 in the studied @xmath9 range , i.e. , between 104 and 116 mb .
these values fall in the middle of the cross section values of the selected subset of the most violent indra data events analyzed in ref .
ideally , the reactions with @xmath812 fm should correspond to 8163 out of the total 100000 generated events . in reality , there were on the average 8109 such events with a fluctuation up to 3% from energy to energy .
we denote this precise number of events @xmath82 to search for , in the full ensemble of 100000 events , the subset of events with the highest multiplicity that is by number of events closest to @xmath82 .
by @xmath83 we label both the lowest multiplicity of the thus selected subset as well as the subset of events itself at each @xmath9 .
let us check the behavior of the most violent @xmath83 hipse events . as a kind of `` background '' , in fig .
[ eq_m ] we show by the thin black line the @xmath61-distribution histogram of the full 100000 event data set for each second studied @xmath9 .
the @xmath61 distribution of the @xmath83 events is shown by the red - line yellow - filled histogram .
these high - multiplicity events are generated in a large domain of @xmath61 values which extends up to 5 fm . to make the @xmath83 @xmath61 distribution better visible ,
it is enlarged to the full frame size by the hollow red - line histogram
. a gaussian fit to it clearly demonstrates that the normal - law of data statistics correctly reproduces the distribution of @xmath83 subset over @xmath61 s .
these events are in minor part ( 3% to 29% ) belonging to the @xmath562 fm subset of the full data set ( hatched part of black histogram ) . from the gaussian fit one
infers that the maximum of these high - multiplicity events is about @xmath843 fm and that it slightly decreases with the increasing @xmath9 .
finally , let us apply the hipse @xmath83 @xmath61 distribution to the lv simulation results . taking the gaussian fit values of fig .
[ eq_m ] as the weights for the integer values of @xmath61 , the @xmath61-averaged @xmath53 are obtained for each studied value of the factor @xmath6 of eq .
( [ sig_m ] ) . by this method , for the @xmath6=1 case these @xmath53 are , in millibarn units , also equivalent to 0.02@xmath55 .
@xmath85 extracted from thus averaged @xmath53 is in fig .
[ cf ] shown by dot - dashed curves and open triangles , upright and reversed , for the d1-g1 eos and zamick eos , respectively . for the nonlocal d1-g1
eos the two @xmath61-averaging intervals give strictly the same @xmath53 for @xmath7750@xmath3 mev . at @xmath9=80@xmath3 and 100@xmath3 mev
the respective @xmath53 values differ by about 20% but are mutually compatible when errors are accounted for .
for the zero - range zamick eos in the full @xmath9 interval , two @xmath61-averaging intervals give compatible predictions for the @xmath86 values although for @xmath87 the more stringent centrality results are in somewhat better agreement with the @xmath75 values .
the semiclassical landau - vlasov ( lv ) transport model was used to study the energy - based isotropy ratio @xmath0 of eq .
( [ r_e ] ) for the @xmath1xe+@xmath2sn reaction in the wide incident energy range 12@xmath88100@xmath3 mev .
the focus of the present work is twofold : \(2 ) the comparison of the simulation multiplicative factor @xmath6 representing the global in - medium change of the free @xmath4 cross section @xmath48 of ref.@xcite with the one deduced from the experimental @xmath57 @xcite . in approaching the above point 1 we investigated ( i ) the role of the dependence of the nuclear mean field on momentum , i.e. , of the nonlocality of the interaction , and ( ii ) the impact of the residual @xmath4 interaction through varied parametrizations of @xmath18 . the success in reproducing the experimental isotropy ratios @xmath57 of ref .
@xcite is mixed : below the fermi energy @xmath5 , the lv model with the strongly in - medium reduced @xmath4 cross section @xmath14 of refs .
@xcite and with the momentum - dependent d1-g1 eos leads to a correct description of @xmath57 ; cf .
[ snn](a ) .
a similar result may be obtained with both the d1-g1 eos and the zero - range zamick eos when the free @xmath4 cross section is strongly scaled down by a constant multiplicative factor @xmath6 of eq .
( [ sig_m ] ) : for the d1-g1 eos @xmath890.5 , fig .
[ sf](a ) , and for the zamick eos @xmath890.1 , fig .
[ sf](b ) . above @xmath5
there is no unique set of model parameters which would lead to a favorable description of the experimental @xmath57 .
earlier studies of the observable @xmath0@xcite have also failed to reproduce the experimental results of ref .
we emphasize , however , that a smooth variation of the parameter that controls the in - medium value of @xmath18 would lead to a complete description of the experimental @xmath57 for both eos . should one draw a conclusion that none of the existing studies on the in - medium modifications of @xmath18 around the fermi energy appropriately accounts for the physical reality ? the local or the nonlocal character of the interaction does not elucidate this question . regarding the above point 2 , we may summarize the outcome of our study as follows : the lv simulation with the local zamick eos predicts the @xmath4 cross - section correction factor @xmath74 which clearly supports the experimentally deduced @xmath75 @xcite .
the model predicts however , the agreement or disagreement between the absolute values of @xmath74 and @xmath75 should be considered with some caution due to two possible causes .
on one hand , the value of factor @xmath6 may be altered by reaction centrality .
accordingly , an investigation of @xmath0 with a quasidynamical event generator hipse @xcite was carried out .
it reveals that the event selection based on multiplicity and the geometrical sharp - cut approximation is not a correct centrality selector .
indeed , corroborating earlier findings @xcite , we show that this selection approach strongly mixes events of different impact parameters over a rather broad span of @xmath61 values ; cf .
when a properly weighted contribution of @xmath61 s involved in the high - multiplicity events is accounted for , the isotropy ratios calculated for the thus relaxed centrality requirement and those strictly central do not differ much .
the thus extracted @xmath74 does not change much as well . on the other hand ,
the derived @xmath75 values are based on a number of strong assumptions that allowed the link between the stopping ratio @xmath0 and the in - medium @xmath4 cross section @xcite .
hence , besides further experimental and theoretical considerations of the stopping observable @xmath0 intended to disentangle the remaining ambiguities a study of other related observables may shed some fresh light on the subject .
in addition , the experimentally observed strong and rapid change of the effective in - medium residual @xmath4 cross section beyond the fermi energy urges for an _ ab initio _ theoretical analysis of this problem , the solution of which might lie in the way the exclusion principle is accounted for @xcite and/or by incorporating the recent observation of short - range correlations in nuclei @xcite .
their consequences for transport descriptions of heavy - ion reactions are of high interest and need to be investigated .
z.b . gratefully acknowledges the financial support and the warm hospitality of the facult des sciences of university of nantes and the laboratory subatech , umr 6457 . this work has been supported in part by croatian science foundation under project no .
7194 and in part by the scientific center of excellence for advanced materials and sensors . which modulates total elastic cross section as a function of polar angle @xmath90 and nucleon incident energy @xmath91 for the scattering of neutron and proton ( upper panel ) and between protons ( lower panel ) , where the applicate axis is in logaritmic scale .
[ ad],width=294 ] parameters @xmath93 , @xmath94 , and @xmath95 are fixed by fitting the experimental @xmath63 data at nine beam energies @xmath96 between 26 and 319 mev , index @xmath52 running over energies . @xmath91 and @xmath96 are expressed in mev units . between these @xmath96 values the parameters
are assumed to change linearly with @xmath91 .
the values of these parameters are given in table [ tabnp ] and @xmath97 is shown in fig .
[ ad](a ) . where coefficients @xmath100 are expressed by the following functional dependence : @xmath101 due to indistinguishability of particles , coefficients @xmath102 and @xmath103 are divided by 2 . at each energy
@xmath91 the limiting angle reads @xmath104 .
the overall angular distribution normalization is given by the value of @xmath105 , and eq.([fpp ] ) is used to define , on monte carlo grounds , the angle into which a couple of charged gaussians is scattered in a @xmath106 collision .
the parameters @xmath107 are fixed by fitting experimental differential cross sections @xmath64 at six energies @xmath108 ranging from 5 to 144 mev denoted by the index @xmath52 .
as above , @xmath91 and @xmath108 are expressed in mev units . between these energies , parameters are assumed to change linearly with @xmath91 .
the values of these parameters are given in table [ tabpp ] and @xmath109 is shown in fig .
[ ad](b ) .
10 h. strbele _ et al .
_ , phys .
c*27 * , 1349 ( 1983 ) .
w. bauer , phys .
lett . * 61 * , 2534 ( 1988 ) .
b. hong _ et al .
_ ( fopi collaboration ) , phys . rev .
c*66 * , 034901 ( 2002 ) ; w. reisdorf _ et al . _
( fopi collaboration ) , phys .
lett . * 92 * , 232301 ( 2004 ) . g. lehaut _ et al . _
( indra collaboration ) , phys .
* 104 * , 232701 ( 2010 ) .
eudes _ et al .
_ , europhys
. lett . * 104 * , 22001 ( 2013 ) .
eudes _ et al .
_ , phys .
c*90 * , 034609 ( 2014 ) .
o. lopez _ et al .
_ ( indra collaboration ) , phys . rev .
c*90 * , 064602 ( 2014 ) . in the work by j.
su and f .- s .
zhang [ phys .
c*87 * , 017602 ( 2013 ) ] the same observable is dubbed the memory loss ratio .
eudes _ et al .
_ , phys .
c*56 * , 2003 ( 1997 ) ; f. haddad _ et al .
_ , phys .
c*60 * , 031603 ( 1999 ) ; jian - ye liu _
et al . _ ,
* 86 * , 975 ( 2001 ) .
g. q. zhang _
_ , phys .
c*84 * , 034612 ( 2011 ) .
v. kaur , s. kumar , and r. k. puri , nucl .
a*861 * , 37 ( 2011 ) .
k. s. vinayak and s. kumar , j. phys .
g*39 * , 095105 ( 2012 ) .
m. h. zhao _ et al .
_ , phys .
c*89 * , 037001 ( 2014 ) .
li and r. machleidt , phys .
c * 49 * , 566 ( 1994 ) ; * 48 * , 1702 ( 1993 ) .
c. grgoire _ et al .
a*465 * , 317 ( 1987 ) .
f. sbille _ et al .
a*501 * , 137 ( 1989 ) .
j. decharg and d. gogny , phys .
c*21 * , 1568 ( 1980 ) .
l. zamick , phys .
b*45 * , 313 ( 1973 ) .
f. haddad _ et al .
_ , phys .
c*52 * , 2013 ( 1995 ) .
p. schuck _ et al .
. part .
* 22 * , 181 ( 1989 ) ; v. de la mota _ et al .
_ , phys .
c*46 * , 677 ( 1992 ) ; f. haddad _ et al . ,
_ * 53 * , 1437 ( 1996 ) ; z. basrak _ et al .
a*624 * , 472 ( 1997 ) ; i. novosel _
_ , phys .
b*625 * , 26 ( 2005 ) .
l. w. nordheim , proc .
a119 * , 689 ( 1928 ) ; e. a. uehling and g. e. uhlenbeck , phys . rev .
* 43 * , 552 ( 1933 ) .
k. chen _ et al .
_ , phys .
166 * , 949 ( 1968 ) .
j. cugnon , a. lejeune , and p. grang , phys .
c*35 * , 861 ( 1987 ) .
f. sbille _ et al .
a*791 * , 313 ( 2007 ) ; f. sbille , private communication .
d. klakow , g. welke and w. bauer , phys .
c*48 * , 1982 ( 1993 ) ; d. j. magestro , w. bauer , and g. d. westfall , _ ibid .
_ * 62 * , 041603(r ) ( 2000 ) .
e. bonnet _ et al .
_ , phys .
c*89 * , 034608 ( 2014 ) .
d. lacroix , a. van lauwe , and d. durand , phys .
c*69 * , 054604 ( 2004 ) .
tripathi , f.a .
cucinotta , and j.w .
wilson , nucl .
b * 117 * , 347 ( 1996 ) .
napolitani ( [email protected] ) , code _ panforte _ for simulating geometrical and detailed detection features of several multidectors .
r. subedi _ et al .
_ , science * 320 * , 1476 ( 2008 ) . o. hen _ et al .
_ , science * 346 * , 614 ( 2014 ) . t.c .
montgomery _ et al .
_ , phys .
c*16 * , 499 ( 1977 ) .
s. benck _ et al .
_ , nucl .
a*615 * , 220 ( 1997 ) .
scanlon _ et al .
_ , nucl .
phys . * 41 * , 401 ( 1963 ) .
king _ et al .
_ , phys .
c*21 * , 1185 ( 1980 ) .
chih and w.m .
powell , phys .
106 * , 539 ( 1959 ) .
measday , phys .
rev . * 142 * , 584 ( 1966 ) .
et al . _ ,
a*377 * , 529 ( 1982 ) .
meagher , phys .
rev . * 78 * , 667 ( 1950 ) .
n. jarmie _ et al .
_ , phys .
lett . * 25 * , 34 ( 1970 ) .
n. jarmie and j.h .
jett , phys .
c*13 * , 2554 ( 1976 ) .
l.h . johnston and d.a .
swenson , phys .
rev . * 111 * , 212 ( 1958 ) .
young and l.h .
johnston , phys .
rev . * 119 * 313 ( 1960 ) .
o.n . jarvis and c. whitehead , phys .
b * 36 * , 409 ( 1971 ) .
|
with the semiclassical landau - vlasov transport model we studied the stopping observable @xmath0 , the energy - based isotropy ratio , for the @xmath1xe+@xmath2sn reaction at beam energies spanning 12@xmath3 to 100@xmath3 mev .
we investigated the impacts of the nonlocality of the nuclear mean field , of the in - medium modified nucleon - nucleon ( @xmath4 ) cross section and of the reaction centrality .
a fixed set of model parameters yields @xmath0 values that favorably compare with the experimental ones , but only for energies below the fermi energy @xmath5 . above @xmath5 agreement is readily possible , but by a smooth evolution with energy of the parameter that controls the in - medium modification of @xmath4 cross section . by comparing the simulation correction factor @xmath6 applied to the free @xmath4 cross section with the one deduced from experimental data [ phys .
rev .
c*90 * , 064602 ( 2014 ) ] , we infer that the zero - range mean field almost entirely reproduces it .
also , in accordance with what has been deduced from experimental data , around @xmath5 a strong reduction of the free @xmath4 cross section is found . in order to test the impact of sampling central collisions by multiplicity an event generator ( hipse )
was used .
we obtain that high multiplicity events are spread over a broad impact parameter range , but it turns out that this has a small effect on the observable @xmath0 and , thus , on @xmath6 as well .
|
the draco dwarf spheroidal companion to the milky way , at a distance of about 82 kpc @xcite , was discovered fifty years ago from examination of palomar observatory sky survey schmidt plates @xcite .
the horizontal branch for this dwarf galaxy is located at @xmath2 and it covers more than 1/3 of a square degree on the sky .
draco is one of the nearest of the galaxy s known dwarf spheroidal companions and is among the faintest known galaxies , with a luminosity of @xmath3 @xcite .
dwarf spheroidal galaxies typically contain many variable stars , predominately the rr lyrae variables typical of older , metal - poor stellar populations @xcite .
draco is no exception .
the first extensive study of draco by @xcite found 138 variable stars in the central part of draco .
they determined that all but five of these variables were rr lyraes .
the five remaining variables were deemed to be anomalous " cepheids .
these cepheids have a shorter period and so do not obey the period - luminosity relation for type ii cepheids @xcite .
these anomalous " cepheids have been found in other dwarf spheroidals since their discovery in draco @xcite .
five carbon stars have also been found in draco by @xcite and @xcite .
more recent compilations of variables in draco are underway @xcite .
the first detailed photometric survey of the draco dwarf spheroidal galaxy was created by @xcite . since then draco has become one of the best studied dwarf galaxies with several more recent photometric surveys .
we have created our own catalog of stars in draco using data from the sloan digital sky survey ( sdss ) .
we present wide area , @xmath4 ccd observations of draco down to about @xmath5 , and construct a color magnitude diagram .
we combine our catalog of sdss stars in draco with previously published photometric data in draco , from : @xcite , @xcite , @xcite , @xcite , and @xcite to create a heterogeneous catalog of draco stars , with multiple epochs of observation which allow us to identify variable candidates .
the ccd imaging observations of draco are taken from sdss runs 1336 and 1339 , which were obtained 2000 april 4 with the sdss mosaic imaging camera @xcite .
the majority of the stars in the dwarf spheroidal fall on two frames of data in the same camera column ( one continuous stream of drift - scan data ) . since the precise calibration for the sdss filter system is still in progress , magnitudes in this paper are quoted in the @xmath6 system , which approximates the final sdss system @xcite .
these systems differ absolutely ( with negligible color terms ) by only a few percent in @xmath7 , and no more than 10% in @xmath8 .
see @xcite and @xcite for further information on the filter system and the overall survey , respectively .
the data were reduced with photo ( lupton et al . in preparation ) version 5.1.7 .
@xcite describes the astrometry and astrometric accuracy of the software .
@xcite describes our monitoring program for photometricity .
these data are included in the public early data release ( edr ) @xcite . in the course of commissioning a survey ,
the images are reduced several times as the software evolves .
the data presented here have been reduced three times . the first reduction is called rerun 0 , " and second is rerun 1 . "
most of the data included in this paper are from rerun 1 , " with some exceptions which are noted in the text below .
rerun 2 " was the version released in the edr , and differs from rerun 1 " only in a small overall calibration offset .
since we recalibrate all of the photometry ourselves , the difference is not important .
so , an object which is listed in the catalog by the sdss i d 1336 - 1 - 5 - 60 - 950 ( run - rerun - cameracolumn - field - objectid ) is the same object as 1336 - 2 - 5 - 60 - 950 in the edr .
the seeing for the draco scans was typically 1.8 fwhm .
intercomparison of objects detected twice in overlapping scans is a good indication of relative photometric error , and for objects with @xmath9 , rms error for stellar sources is typically @xmath10% . for objects between @xmath11 ,
typical errors are 8% , growing to 20% at @xmath12 , near the detection limit . for reference ,
blue stars with @xmath13 have a sdss @xmath1 magnitude approximately equal to their johnson @xmath14 magnitude , while stars with @xmath15 have @xmath16 . in general , @xmath17 @xcite . in order to obtain the best measurements at faint magnitudes , we used photometry derived from fits of the pointspread function ( psf ) to stellar profiles .
the sdss code determines the psf as a function of position in each ccd frame ( 1361 x 2048 pixels , or @xmath18 x @xmath19 ) .
while we were studying the stellar populations within draco , we noticed a systematic shift in the color of the giant branch as a function of position within the draco field .
further analysis showed that this result could be explained by systematic inaccuracies in the ( psf ) photometry as modeled in the edr reductions of the data ( these systematics are corrected in later data releases ) .
since the code used to create the edr did not accurately track rapidly varying point - spread functions , the psf magnitudes in some regions of the sky were systematically shifted brighter or fainter by up to a tenth of a magnitude .
for this study , we were able to correct the edr psf magnitudes using the aperture magnitudes for bright objects .
although aperture magnitudes are quite noisy for faint stars , they are accurate for bright stars and not affected by the error in the calculated psf .
we used the brighter stars to calculate a correction to the psf magnitudes , as a function of row and column in each frame .
we fit a second degree polynomial to the deviations between the aperture and psf magnitudes ( mag(@xmath20)@xmath21mag(@xmath22 ) ) vs. ccd row and column for stars brighter than magnitude 19.5 .
we then used this polynomial to correct the photometry of _ all _ stars in the frame to the mean of the aperture photometry for that frame .
the assumption here is that the aperture magnitudes are correct , and their calibration does not vary as the psf varies across rows or columns .
this is a good assumption if we capture practically all of the light in each object , even in the sections of data with poorer seeing .
the results of this correction procedure are shown in figure 1 .
the left panel shows the color - magnitude diagram for draco , using psf magnitudes , and the right panel shows the same diagram after correcting the psf magnitudes . though we did not use any information about the giant branch of draco , notice that the giant branch of the dwarf galaxy is significantly narrower after correction
this is evidence that the correction technique significantly improved the accuracy of the photometry .
in order to generate a useful catalog of stars in the draco dwarf galaxy , we first selected from the sdss database those objects with @xmath23 and @xmath24 , which were marked as unsaturated , and which were far enough from the edge of the frame that they were completely contained on a single ccd detector .
this latter criterion effectively eliminates only quite extended galaxies near the edges , since there is enough overlap between frames that point sources are completely contained on at least one frame .
we also removed duplicate measurements of the same astronomical object , either by multiple observation or multiple software measurements of the same observation , by choosing only observations marked in the sdss as ` primary . ' for each direction in the sky , only one observation and instance of processing is considered ` primary , ' so this flag effectively removes duplicates . to determine a centroid of our elliptical region , stars in our original data set were binned into boxes of area 0.05 degrees on a side
then , we fit an elliptical gaussian ( right ascension and declination of the center , major and minor axes , a position angle , and sky ) , to the binned data .
the adopted center of the distribution is : @xmath25 ( j2000 ) .
the exact center of the fit shifted by ten arcseconds depending on how we weighted the fit , which yields an error bar on our chosen center .
other sources such as @xcite , @xcite , and @xcite report the center of draco as : @xmath26 17:20:13 @xmath27 57:55:11 ( j2000 , converted from b1950 ) , @xmath26 17:20.3 @xmath27 57:55.1 ( j2000 , converted from b1950 ) , and @xmath26 17:20:12.39 @xmath27 57:54:55.3 ( j2000 ) respectively .
these positions for draco s center are in good agreement with the value we found in right ascension , and off by about ten arcseconds in declination .
the major axis of the ellipse fit is nearly constant in declination ( pa @xmath28 ) with gaussian sigma 0.122 degrees ( corrected for @xmath29 ) .
the minor axis is 0.082 degrees in the declination direction implying an ellipticity of 1 - 0.082/0.122=0.33 in agreement with hodge ( 1964 ) who found an ellipticity of @xmath30 , and also in excellent agreement with @xcite who find an ellipticity of 0.331 for draco . to increase the fraction of cataloged stars which are associated with the dwarf galaxy , we included only those stars in an elliptical region centered on the draco dwarf galaxy ( see figure 2 ) , which extends to about one third of the tidal radius measured by @xcite . of the 7417 objects within the area of this 2-sigma ellipse for our draco catalog , 5492 were brighter than our adopted magnitude limit of 23.005 ( this odd effective limit is due to rounding before the limit was applied ) .
the actual number of sdss objects in the final catalog , 5634 , is slightly different , as 15 objects with suspicious sdss photometry ( mostly on bright star bleed trails ) were removed , and 157 objects were added from previous data reductions of the same astronomical images .
positions on the sky of cataloged sdss stars are displayed in figure 2 .
faint sdss detections ( @xmath31 ) are shown as small black dots , while brighter stars are larger black dots .
galaxies ( objects which the sdss software determined to be extended ) are not included in the plot .
the catalog presented in this paper overlaps several previously published photometric surveys of the draco dwarf galaxy .
the footprints of the five surveys which are cross - referenced to sources in our catalog are shown in figure 2 .
our catalog also includes the fainter ten of the twelve draco stars with accurate photometry from @xcite .
their sdss ids , from brightest to faintest , are : 1339 - 1 - 5 - 62 - 26 , 1336 - 1 - 5 - 61 - 148 , 1336 - 1 - 5 - 61 - 119 , 1339 - 1 - 5 - 61 - 384 , 1336 - 1 - 5 - 61 - 478 , 1339 - 1 - 5 - 61 - 366 , 1336 - 1 - 5 - 61 - 370 , 1336 - 1 - 5 - 61 - 294 , 1336 - 1 - 5 - 61 - 685 , and 1336 - 1 - 5 - 61 - 1227 .
since there are so few , we do not include a separate column in the catalog table . in the remainder of this section
, we describe the construction of a catalog of draco stars with cross - references to the other catalogs .
ultimately , the overlapping catalogs are used to generate a list of possible variable stars .
we list in table 1 a summary of stars detected in each previous survey , along with numbers of matches to sdss detections . the main data product of this paper , with format specified in table 2 , is contained in the ( electronic ) table 3 .
table 3 contains identifications , magnitudes , colors and cross - references to other surveys for stars in draco .
the sdss j2000 right ascension ( in degrees ) , declination ( in degrees ) , sdss i d ( run - rerun - cameracolumn - field - objectid ) , @xmath1 magnitude , and @xmath32 colors of each object are listed as columns 1 through 8 of table 3 , respectively .
these magnitudes have been derreddened using the e(b - v ) tabulated in column 9 , as derived from @xcite .
following @xcite , @xmath33 , @xmath34 , @xmath35 , @xmath36 and @xmath37 .
measurements of the same objects in other catalogs are listed in columns 10 through 33 , as described below .
note that the sdss photometry is dereddened , while that of all the other catalogs is not .
a flag indicating classification of the object as stellar ( s " ) , a resolved galaxy ( g " ) or of an indeterminant extent ( f " for faint ) is indicated in column 34 of table 3 .
all stars with @xmath31 are classified as faint . "
the 5766 lines in this table include the 5634 individual objects detected in the sdss database , 19 objects from the baade - swope dataset which were not matched to sdss objects , and 113 duplicate entries in cases where there exist piatek objects in both the e1 and co fields .
the additional baade - swope entries were apparent in the sdss images , so we were able to provide approximate astrometric positions even though no automated photometry was generated .
we included multiple lines for piatek duplicates to avoid adding yet four more columns to an already large table . the survey of area of @xcite ( hereafter bs61 ) , approximately covering a circular region of radius @xmath38@xmath39 centered at r.a .
@xmath40 , dec . @xmath41@xmath42@xmath43 ( b1950 ) , is completely contained within our catalog .
since the individual right ascensions and declinations were not included for each star in the bs61 catalog , the stars were matched by comparing the finding charts with sdss images .
of the 624 stars in table c of bs61 , 563 were cross - identified with an object in the sdss catalog , available in the sdss early data release ( edr ) .
an additional 24 objects were measured in a previous software reduction ( rerun 0 ) of the same data , and were included in our catalog .
the majority of the data in table 3 are from a software reduction called rerun 1 .
" the second number in the third field of table 3 is the sdss i d rerun number , thus , a 0 there would indicate that rerun 0 was used .
the photometry for these latter objects could differ systemmatically by a small amount from the majority of the objects in the catalog , and were not corrected to match large aperture photometry of bright stars .
several other bs61 stars were apparent in the sdss images , but were not reliably measured in any automated sdss data reduction . in most cases ,
the star was close enough to a very bright star that the automated detection software had difficulty deblending it .
the approximate positions for nineteen such objects are included in table 3 , though no sdss photometry is provided .
one star could not be identified because it was not labeled in the bs61 finding chart .
the remaining bs61 stars either could not be reliably identified on the sdss images or lay close enough to an image defect that the data could not be used reliably .
the process of cross - referencing catalogs through visual identification of stars in finding charts and images is inherently prone to errors . to reduce and evaluate the error rate ,
several checks were performed .
we plotted @xmath44 vs. @xmath45 for a set of magnitude ranges ( figures 3a and 3b ) . here
, @xmath1 is the sdss magnitude which is most similar to v. except in unusual cases , stellar photometry should produce an approximately linear relationship between @xmath44 and @xmath45 .
the scatter should increase with the photometric errors at fainter magnitudes .
the derived color relations ( sdss dereddened values compared to bs61 undereddened values ) are : @xmath46 @xmath47 this transformation is only slightly different from that found in @xcite ( hereafter f96 ) , based on preliminary sdss filter transmission design .
outliers on the @xmath44 vs. @xmath45 plots were checked for mis - identification matches between sdss and bs61 , and several were found and corrected . using the derived color transformations , the magnitudes of the detections of the same objects were directly compared in magnitude and color .
stars with significant differences in computed magnitudes or colors between different catalog observations are either variable , have unusual spectra which do not lead to linear color transformations , or are errors in either the sdss or bs61 measurements .
the outliers in magnitude are marked with diamonds in figure 4a , and those outliers in color by diamonds in figure 5a .
note that the photometry of bs61 is systematically shifted by several tenths of a magnitude at the faint end .
we compared the photometry between catalogs only for objects which are classified in the sdss as stellar ( column 34 of table 3 ) . since the fraction of the light measured for galaxies differed between datasets , the inclusion of extended sources in the outlier detection process led to a large number of objects which appeared to vary in brightness from one observation to the next .
this data cut did not have a large effect on the results from the bs61 data , since bs61 pre - selected only stellar objects , but is of greater import in comparing objects in other catalogs discussed below .
of the 586 bs61 objects matched to sdss objects , only 25 were classified as galaxies by the sdss , and 6 were fainter than @xmath48 magnitude in @xmath1 ( and thus are classified only as faint " in table 3 ) .
stellar objects with a significant calculated photometric difference between sdss and bs61 are flagged in column 13 of table 3 .
stars without significant photometric differences ( as seen in figures 4a and 5a ) are flagged 0 , stars with significant differences , but which have magnitudes and colors consistent with rr lyrae variables ( selection criteria are discussed below ) are flagged as 1 and other variable candidate stars are flagged as 2 in this column .
a flag value of 3 denotes an object which was not measured in the sdss , bs61 , or both .
the bs61 i d , v magnitude , and @xmath45 values are found in columns 10 , 11 , and 12 of table 3 , respectively . the @xcite ( hereafter s79 ) photometric survey consisted of new photometry for a subset ( 512 stars ) of the bs61 catalog .
the star identifications of the stars in table v of s79 are the same star identifers as in table c of bs61 .
since the s79 stars are identical to the bs61 stars , the s79 photometric data were simply matched to the appropriate bs61 stars .
the s79 i d , v magnitude , and @xmath45 values are found in columns 14 , 15 , and 16 respectively in table 3 .
the photometric difference flag in column 17 of table 3 indicates whether or not there is a significant difference between the s79 ( recalibrated bs61 ) photometry and sdss data , as graphically shown in figures 3c and 3d .
the same @xmath49 transform was performed as was used for bs61 , and the outliers are shown in figures 4b and 5b .
the values of the flags have the same meanings as they do for bs61 ( 0:no difference , 1:rr lyrae candidate , 2:unknown variable candidate , 3:no sdss detection ) .
the survey area of @xcite ( hereafter cs86 ) is broken up into two overlapping fields : field 1 and field 2 . the total area ( 27 square arcminutes ) of these two fields is completely contained within our catalog .
each of the two fields covers @xmath50@xmath51 in declination and @xmath52@xmath53 in right ascension and an outline of the field boundaries is drawn in figure 2 . each field
is approximately centered on a bs61 star ; field 1 is centered on bs69 star 92 and field 2 on star 289 .
the two fields overlap by 3.4 square arcminutes ( 26% ) .
an x and y value ( in pixels ) is given for each star in the two fields .
to convert these x and y values to right ascension and declination , we created a transformation for each field .
these transformations are accurate within about 1 arcsecond ( rms ) .
the transformation for field 1 is : @xmath54 @xmath55 and for field 2 the transformation is : @xmath56 @xmath57 about half of the stars in each of the two fields were faint ( v @xmath58 @xmath59 ) , and were not matched to the sdss data .
370 of the 547 field 1 stars were matched to sdss detections , and 292 of the 423 field 2 stars were matched .
most of the stars that were not identified had @xmath60 ) .
since the unmatched stars were not clumped in position on the sky , we believe that the coordinate transformation is not responsible for the fact that many of stars are not found in common between the catalogs .
almost all of the unmatched cs86 stars were either fainter than 23rd magnitude in v or close enough to brighter stars that the sdss deblender did not identify them as isolated sources .
we used the same method to reduce and evaluate the error rate in cs86 data as we did in bs61 data .
the transformation from b , v photometry to @xmath61 photometry was different , however .
the transformation from f96 was a good match to the data , so it was adopted : @xmath62 @xmath63 the computed @xmath1 values would differ by a few percent if we used the transformations derived in @xcite from comparisons of sdss standard stars to landolt standards .
a comparison of our photometry with the corresponding photometry from ten stars in @xcite support the claim that the @xcite transformations are a better match to modern ccd photometry than the theoretical transformations of f96 .
since we are using the transformations only to look for outliers , it does nt matter if the overall transformation is off by a few percent ( either due to the transformation or the fact that our magnitudes are dereddened and cs86 data is not ) .
figures 3e , 3f , 4c and 5c show the photometric comparisons between sdss and cs86 data , and the outliers chosen as candidate variables .
twenty candidate variable objects were identified ( of which 16 are consistent with draco rr lyrae variables in color and magnitude ) .
these objects are either variable , have unusual spectra , or highlight photometric errors in at least one catalog .
the cs86 field 1 i d , v magnitude , @xmath45 color , and variability flag are found in columns 18 , 19 , 20 , and 21 , respectively , and the cs86 field 2 i d , v magnitude , @xmath45 color , and variability flag are found in columns 22 , 23 , 24 , and 25 , respectively . out of the nine tables in and around draco in @xcite ( hereafter p01 ) ,
only three ( tables a , b , and d ) overlap data within our catalog . field c0 from p01 contains the majority of stars found within our catalog . of the 11381 c0 stars within our catalog region , 4796 stars matched to stars within our catalog .
132 of the unidentified stars were found in the rerun 0 version of the reduction software .
as was true in comparing our catalog to those of bs61 and cs86 , most of the unmatched stars brighter than @xmath64 were not detected in the sdss because they were not deblended from bright stars .
out of the 1294 stars in e1 ( p01 table b ) and 191 stars in w1 ( p01 table d ) within our catalog region , 444 and 60 , respectively , stars were identified in our catalog .
an additional five of the unidentified stars in table b and two of the unidentified stars from table d were found in the rerun 0 version of the reduction software .
a total of 5326 matches to p01 objects were made out of a total of 5634 unique objects in the sdss catalog .
again , about 90% of the stars brighter than @xmath64 are matched with objects in the sdss database . for bluer colors , @xmath65 , the f96 color transformations work well : @xmath66 @xmath67 for redder stars , we used the transformation : @xmath68 @xmath69 figures 3 g , 3h , 4d , and 5d show a comparison of the photometry between p01 and sdss catalogs ; stars with inconsistent photometry are marked with a diamond - shaped symbol .
134 of the 2374 matched objects are flagged as possible variable stars .
due to saturation of p01 photometry on the bright end , we did not compare photometry for stellar objects brighter than @xmath70 .
the p01 i d , v magnitude , @xmath71 color , and match flag are found in columns 2629 of table 3 , respectively .
note that in cases where there is a match of the same sdss star to a p01 c0 and either a p01 e1 or p01 w1 field , that there are two lines of entries in table 3 , one for each match , with the sdss information duplicated .
the survey of area of the hst observations of draco by @xcite ( hereafter g98 ) is completely contained within our catalog and is roughly centered and almost completely contained within cs86 field 1 ( see figure 2 ) . it was necessary to create a transformation from g98 x and y to r.a . and dec . for the three g98 fields ( wf2 , wf3 , and wf4 ) . for the wf2 field
the transformation is : @xmath72 @xmath73 for wf3 the transformation is : @xmath74 @xmath75 and for wf4 the transformation is : @xmath76 @xmath77 these fields are from one pointing of wfpc2 .
the g98 data are taken with the hubble space telescope and therefore includes very faint stars .
there are 50 stars with v magnitude @xmath78 23 in the wf2 data , 46 in the wf3 data , and 56 in the wf4 data .
a total of 88 stars were matched by position to stars in our catalog , including 6 from the rerun 0 " version of the catalog .
since f96 provided no transformations from v , i to @xmath61 filter systems , we generated our own transformation equations from the data : @xmath79 @xmath80 the g98 i d ( id - wf chip number ) , v magnitude , @xmath81 color , and photometric matching flag are found in columns 30 , 31 , 32 and 33 of table 3 , respectively .
no stars were selected as variables by comparison to the g98 catalog ( see figures 3i , 3j , 4e and 5e ) .
the catalog of table 3 is intended to provide a reference to researchers interested in the stellar populations and structure of draco .
the extensive cross - references to the literature provide an opportunity to select variable candidates based on multi - epoch photometry over a long baseline . though the filter systems are inhomogeneous , roughly linear behavior for most colors yields a good list of candidate variables across the draco field .
table 4 lists all 142 candidate variables selected from table 3 .
the columns of table 4 are : right ascension , declination , sdss i d , sdss @xmath1 , @xmath82 , @xmath83 , suspected variable type ( rr lyrae , qso , cepheid or unknown ) , matching catalog ( bs61 , s79 , cs79 , p01 , or g98 ) , and the @xmath84 and @xmath85 between the sdss observation and the transformed magnitude and color of the cross reference observation onto the sdss @xmath86 system . a positive value of delta @xmath1 or delta @xmath82 means the star was fainter or redder respectively in the sdss than in the comparison catalog .
the last column gives the identity of the object from other catalogs .
since the observations with the full set of filters for sdss observations for each object are obtained within minutes of each other , most stars can be expected to be at a single point in their light curve . by plotting the colors of known rr lyrae stars from bs61
, we determined that in draco these variable stars lie in a narrow region of the h - r diagram ( see figure 6 ) .
therefore , we tentatively identify all 112 stars in the range @xmath87 , @xmath88 , @xmath89 as rr lyrae candidates .
these objects have a one in the variability flag fields of table 3 , and are labeled as rr lyrae " in table 4 .
objects which are suspected to vary but which do not lie in the region inhabited by rr lyrae stars have a one in the variability flag fields of table 3 and are labeled as unknown " in table 4 .
there are only 30 objects suspected of varying which have colors outside the region known to contain rr lyrae variables .
a few of these might be rr lyrae variables as well , as their colors place them only slightly outside our rr lyrae color box .
the location of all variable candidates is summarized in the color magnitude diagram presented in figure 6 .
variable candidates from each cross - referenced catalog are marked with a separate color while all sdss stars are indicated as black dots .
the variables from the bs61 survey are well confirmed @xcite with full light curves .
we were able to find the positions of 135 of the 137 bs61 variable objects within eight arcminutes of the center of draco .
two of the 137 did not appear to be in the central part in the finding charts , so they were not included .
note that the numbering system for the bs61 variable stars is separate from the photometric table with which we matched our data ; we did not match the photometry for the variable star list .
we recover 82 ( 60% ) of the 135 bs61 variables , which are identified in the last column of table 4 .
we also recover 28 numbered bs61 variables outside the inner eight arcminutes , and three bs61 variables which were not given a number .
since most of the variables were detected from only two epochs of photometry , we would not expect to recover all of the variable objects in draco from this technique .
two of the variable candidates ( 1339 - 1 - 5 - 60 - 207 and 1339 - 1 - 6 - 61 - 504 ) are known quasars from @xcite .
these objects are indicated by a large black circle in figure 6 , and labeled qso " in the last column of table 4 .
five known carbon stars are included in table 3 .
sdss stars 1336 - 1 - 5 - 61 - 267 , 1336 - 1 - 5 - 61 - 436 , and 1336 - 1 - 5 - 61 - 484 correspond to @xcite stars @xmath90 , c1 , and c3 , respectively .
sdss stars 1336 - 1 - 5 - 60 - 249 , 1336 - 1 - 5 - 61 - 436 , and 1336 - 1 - 5 - 60 - 294 correspond to @xcite stars j171942.4 + 575838 , j171957.7 + 575005 , and j172038.8 + 575934 , respectively .
object 1336 - 1 - 5 - 61 - 267 is apparently variable and we detect it as such in multiple catalog comparisons .
it appears as a multi - colored point at @xmath91 in figure 6 , and is identified as type carbon " in the last column of table 4 .
a matched catalog of 5634 objects in the draco dwarf field , extending over four times the area of the @xcite catalog and reaching to approximately one third the new tidal radius of @xcite , has been assembled .
we selected five color data from the sdss with @xmath92 , within the elliptical region @xmath93 ^ 2 + [ ( \delta-57.913)/0.164]^2 < 1 $ ] , where @xmath94 and @xmath95 are in j2000 degrees .
this sdss data was merged with several existing data sets to produce our final catalog .
the data sets include : @xcite , @xcite , @xcite , @xcite , and @xcite . using only stellar sources with @xmath96
, we identify 142 candidate variable stars , of which 113 were identified as variables in bs61 , one is a known carbon star , and two are known qsos .
since the sdss observations were taken nearly simultaneously in all filters , the catalog colors and distance to the draco dwarf can be used to identify potential rr lyrae stars .
nearly @xmath97 of the candidate variable objects have colors consistent with those of rr lyrae variables .
only 6 ( 23% ) of the 26 objects which do not have previous identifications are classified as rr lyrae candidates on the basis of their colors .
we report all astrometric and photometric transformations used to compare our data with previous catalogs .
we would like to thank carl j. grillmair for sending his unpublished data , and hugh harris for providing us with the positions of bs61 variables outside of the central eight arcminutes of draco .
the referee , slawomir piatek , made many helpful comments which improved the paper .
funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions .
the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington .
aaronson , m. , liebert , j. , & stocke , j. 1982 , , 254 , 507 baade , w. , and swope , h. h. 1961 , 66 , 300 cardelli , j. a. , clayton , g. c. , & mathis , j. s. , 1989 , , 345 , 245 carney , b. w. , and seitzer , p. 1986
, , 92 , 23 cotton , w. d. , condon , j. j. , arbizzani , e. 1999 , , 125 , 409 fukugita , m. , ichikawa , t . , gunn , j. e. , doi , m. , shimasaku , k. , schneider , d. p. 1996
, , 111 , 1758 grillmair , c. j. , mould , j. r. , holtzman , j. a. , worthey , g. , ballester , g. e. , burrows , c. j. , clarke , j. t. , crisp , d. , evans , r. w. , gallagher , j. s. iii , griffiths , r. e. , hester , j. , hoessel , j. g. , scowen , p. a. , stapelfeldt , k. r. , trauger , j. t. , watson , a. m. , and westphal , j .
a. 1998 , , 115 , 144 gunn , j. e. , et al .
1998 , , 116 , 3040 hendon , a. , and munari , u. 2000 , a&as , 143 , 343 hodge , p. w. 1964 , , 69 , 853 hogg , d.w . , finkbeiner , d.p .
, schlegel , d.j . , and gunn , j.e .
2001 , , 122 , 2129 irwin , m. , & hatzidimitriou , d. 1995 , mnras , 277 , 1354 kinemuchi , k. , harris , h. c. , smith , h. a. , silbermann , n. , lacluyze , a. , clark , c. l. 2001 , , 198 , 46.05 lasker , b. , et al .
, 68 , 1 margon , b. , et al .
2002 , , 124 , 1651 mateo , m. 1998 , , 36 , 435 nemec , j. m. 1985 , , 90 , 204 odenkirchen , m. , grebel , e. k. , harbeck , d. , dehnen , w. , rix , h. , newberg , h. j. , yanny , b. , holtzman , j. , brinkmann , j. , chen , b. , csabai , i. , hayes , j. j. e. , hennessy , g. , hindsley , r. b. , ivezic , z. , kinney , e. k. , kleinman , s. j. , long , d. , lupton , r. h. , neilsen , e. h. , nitta , a. , snedden , s. a. , york , d. g. 2001 , , 122 , 2583 piatek , s. , pryor , c. , armandroff , t. e. , and olszewski , e. w. 2001 , , 121 , 841 pier , j.r .
, et al . 2003 , , in press schneider , d. p. , et al .
2002 , , 123 , 567 schlegel , d.j . ,
finkbeiner , d.p . , &
davis , m. 1998 , apj , 500 , 525 smith , j. a. , et al .
2002 , , 123 , 2121 stetson , p. b. 1997 , baltic astronomy 6 , 3 stetson , p. b. 1979 , , 84 , 1149 stoughton , c. , et al .
2002 , , 123 , 485 wilson , a. g. 1955 , , 67 , 27 york , d.g . , et al .
2000 , , 120 , 1579 zinn , r. 1978 , , 225 , 790 no . of stars & 624 & 512 & 547 & 423 & 11,381 & 1,294 & 191 & 152 + total matched & 586 & 488 & 370 & 292 & 4,928 & 449 & 62 & 88 + re - run 1 stars & 562 & 467 & 363 & 288 & 4,796 & 444 & 60 & 82 + re - run 0 stars & 24 & 21 & 7 & 4 & 132 & 5 & 2 & 6 + possible variables & 2 & 6 & 11 & 9 & 126 & 7 & 0 & 0 + 1 & 9.5f & right ascension ( j2000 degrees ) + 2 & 8.5f & declination ( j2000 degrees ) + 3 & 17s & sdss i d ( run - rerun - camcol - field - id ) + 4 & 7.2f & @xmath1 - dereddened , aperture corrected luptitude from psf - fitting + 5 & 6.2f & @xmath83 + 6 & 6.2f & @xmath82 + 7 & 6.2f & @xmath98 + 8 & 6.2f & @xmath99 + 9 & 6.4f & e(b - v ) + 10 & 6s & baade - swope i d + 11 & 6.2f & baade - swope v + 12 & 6.2f & baade - swope b - v + 13 & 1d & baade - swope flags + 14 & 6s & stetson i d + 15 & 6.2f & stetson v + 16 & 6.2f & stetson b - v + 17 & 1d & stetson flags + 18 & 6s & carney - seitzer field 1 i d + 19 & 6.3f & carney - seitzer field 1 v + 20 & 6.3f & carney - seitzer field 1 b - v + 21 & 1d & carney - seitzer field 1 flags + 22 & 6s & carney - seitzer field 2 i d + 23 & 6.3f & carney - seitzer field 2 v + 24 & 6.3f & carney - seitzer field 2 b - v + 25 & 1d & carney - seitzer field 2 flags + 26 & 9s & piatek i d + 27 & 7.3f & piatek v + 28 & 5.2f & piatek v - r + 29 & 1d & piatek flags + 30 & 6s & grillmair id - field + 31 & 6.3f & grillmair v + 32 & 6.3f & grillmair v - i + 33 & 1d & grillmair flags + 34 & 1s & type ( s - star ; g - galaxy ; f - faint , @xmath31 ) rrcrrrrrrrrrr 259.59433 & 57.90571 & 1339 - 1 - 5 - 60 - 1510 & 22.98 & 1.42 & 1.51 & 0.86 & 0.43 & 0.0262 & & & & 3 + 259.59869 & 57.90565 & 1339 - 1 - 5 - 60 - 1511 & 22.94 & -0.16 & -0.06 & -0.66 & 1.46 & 0.0262 & & & & 3 + 259.59895 & 57.92744 & 1339 - 1 - 5 - 60 - 291 & 19.08 & 1.38 & 2.66 & 1.06 & 0.61 & 0.0257 & & & & 3 + 259.59919 & 57.94193 & 1339 - 1 - 5 - 60 - 1252 & 22.51 & 0.89 & 0.51 & 0.61 & 0.29 & 0.0254 & & & & 3 + 259.59940 & 57.91351 & 1339 - 1 - 5 - 60 - 383 & 21.31 & 1.32 & 1.57 & 1.19 & 0.62 & 0.0261 & & & & 3 259.64816 & 57.87486 & 1339 - 1 - 5 - 61 - 655 & 20.27 &
-0.01 & 1.31 & unknown & p01 & 0.559 & 0.116 & + 259.66009 & 57.87728 & 1339 - 1 - 5 - 61 - 652 & 19.91 & 0.15 & 1.09 & rr lyrae & p01 & -0.680 & -0.210 & + 259.70073 & 57.92844 & 1339 - 1 - 5 - 60 - 102 & 20.69 & 0.43 & 2.33 & unknown & p01 & -0.025 & -1.000 & + 259.73549 & 57.98246 & 1339 - 1 - 5 - 60 - 239 & 20.73 & -0.88 & 0.91 & unknown & p01 & 0.849 & 0.326 & + 259.74238 & 57.88235 & 1339 - 1 - 5 - 61 - 689 & 20.14 & 0.18 & 0.80 & rr lyrae & p01 & 1.153 & 1.242 & bs-76 + 259.74469 & 57.88249 & 1339 - 1 - 5 - 61 - 688 & 19.91 & 0.08 & 0.95 & rr lyrae & p01 & -1.237 & -1.108 & bs-109 + 259.75711 & 58.00798 & 1339 - 1 - 5 - 60 - 207 & 18.70 & 0.08 & 0.20 & unknown & p01 & -0.524 & 0.062 & qso + 259.77357 & 57.92739 & 1339 - 1 - 5 - 61 - 569 & 20.37 & 0.26 & 0.87 & rr lyrae & p01 & 0.684 & 0.404 & bs-159 + 259.77654 & 57.83001 & 1339 - 1 - 5 - 61 - 369 & 19.21 & 0.14 & 0.95 & unknown & p01 & 1.486 & 1.220 & bs-134 + 259.78363 & 57.97636 & 1339 - 1 - 5 - 60 - 261 & 18.24 & -0.04 & 1.15 & unknown & p01 & -0.414 & -0.184 & bs-157 + 259.79617 & 58.01210 & 1339 - 0 - 5 - 60 - 49 & 20.84 & 1.80 & 1.56 & unknown & p01 & 0.393 & 0.370 & + 259.79988 & 57.91029 & 1339 - 1 - 5 - 61 - 626 & 20.41 & 0.19 & 1.17 & rr lyrae & p01 & -0.497 & -0.764 & bs-22 + 259.82323 & 57.89237 & 1339 - 1 - 5 - 61 - 704 & 19.61 & -0.03 & 1.15 & rr lyrae & p01 & -0.784 & -0.246 & bs-73 + 259.82668 & 57.87930 & 1339 - 1 - 5 - 61 - 750 & 20.32 & 0.14 & 1.10 & rr lyrae & p01 & 0.702 & 0.716 & bs-110 + 259.84590 & 57.83288 & 1339 - 1 - 5 - 61 - 106 & 19.69 & 0.02 & 1.19 & rr lyrae & p01 & -1.281 & -0.952 & bs-137 + 259.84705 & 57.93216 & 1339 - 1 - 5 - 61 - 590 & 20.54 & 0.28 & 0.93 & rr lyrae & p01 & 0.933 & 0.874 & bs-15 + 259.86091 & 57.89279 & 1339 - 1 - 5 - 61 - 732 & 20.45 & 0.31 & 1.05 & rr lyrae & p01 & 0.414 & 0.472 & bs-107 + 259.87057 & 57.82140 & 1339 - 1 - 5 - 61 - 908 & 20.40 & 0.31 & 0.90 & rr lyrae & p01 & 0.410 & 0.364 & bs-47 + 259.87712 & 57.94266 & 1339 - 1 - 5 - 61 - 183 & 19.38 & -0.01 & 1.14 & rr lyrae & p01 & 0.517 & 0.980 & bs-14 + 259.89192 & 57.92656 & 1339 - 1 - 5 - 61 - 235 & 20.43 & 0.17 & 1.18 & rr lyrae & p01 & 0.672 & 0.602 & bs-18 + 259.89579 & 57.84657 & 1339 - 1 - 5 - 61 - 884 & 19.98 & 0.12 & 1.20 & rr lyrae & p01 & 1.027 & 1.254 & bs-45 + 259.90025 & 57.90432 & 1339 - 1 - 5 - 61 - 280 & 18.12 & 0.19 & 1.26 & unknown & p01 & 0.328 & 0.388 & bs-194 + 259.90181 & 57.82497 & 1339 - 1 - 5 - 61 - 911 & 20.38 & 0.09 & 1.02 & unknown & p01 & 0.405 & 0.252 & bs-46 + 259.90992 & 57.79015 & 1339 - 1 - 5 - 62 - 406 & 20.28 & 0.16 & 1.11 & rr lyrae & p01 & 0.413 & 0.520 & bs-50 + 259.91930 & 57.97742 & 1336 - 1 - 5 - 60 - 737 & 20.29 & 0.43 & 1.29 & unknown & p01 & 1.000 & 1.960 & + & & & & & & & s79 & -0.035 & -0.043 & + & & & & & & & bs61 & 0.008 & -0.169 & + 259.92261 & 57.89095 & 1339 - 1 - 5 - 61 - 763 & 20.38 & 0.22 & 1.37 & rr lyrae & p01 & 0.397 & 0.490 & bs-129 + & & & & & & & cs86 & 0.455 & 0.811 & + 259.92724 & 58.05734 & 1336 - 1 - 5 - 60 - 578 & 19.53 & -0.17 & 1.31 & rr lyrae & p01 & -1.572 & -1.600 & bs-2 + 259.92756 & 57.89161 & 1339 - 1 - 5 - 61 - 762 & 19.89 & 0.09 & 0.94 & rr lyrae & p01 & -0.541 & -0.234 & bs-188 + & & & & & & & cs86 & -0.585 & -0.178 & + 259.93739 & 57.90498 & 1336 - 1 - 5 - 60 - 525 & 20.28 & 0.28 & 1.04 & rr lyrae & p01 & 0.399 & 0.298 & bs-127 + & & & & & & & cs86 & 0.363 & 0.508 & + & & & & & & & s79 & 0.278 & -0.393 & + & & & & & & & bs61 & 0.318 & -0.487 & + 259.93800 & 57.97613 & 1336 - 1 - 5 - 60 - 769 & 20.24 & 0.27 & 1.22 & rr lyrae & p01 & 1.858 & -1.160 & + & & & & & & & s79 & -0.006 & -0.035 & + & & & & & & & bs61 & 0.038 & -0.067 & + 259.94096 & 57.94053 & 1336 - 1 - 5 - 60 - 355 & 20.10 & 0.19 & 1.40 & rr lyrae & p01 & 0.381 & 0.730 & bs-104 + & & & & & & & cs86 & 0.243 & 0.614 & + 259.94262 & 57.79564 & 1339 - 1 - 5 - 62 - 407 & 20.13 & 0.23 & 1.06 & rr lyrae & p01 & 0.413 & 0.482 & bs-86 + 259.95101 & 57.91426 & 1336 - 1 - 5 - 60 - 1053 & 19.79 & 0.00 & 1.26 & rr lyrae & p01 & -0.308 & 0.090 & bs-20 + & & & & & & & cs86 & -0.505 & -1.158 & + 259.95303 & 57.94910 & 1336 - 1 - 5 - 60 - 867 & 20.42 & 0.23 & 1.24 & rr lyrae & p01 & 1.075 & 0.914 & bs-147 + & & & & & & & cs86 & 0.373 & -0.044 & + 259.95442 & 57.99014 & 1336 - 1 - 5 - 60 - 1504 & 21.79 & 0.61 & 0.61 & unknown & p01 & 0.615 & 0.934 & + 259.95470 & 58.03670 & 1336 - 1 - 5 - 60 - 622 & 20.59 & 0.37 & 1.16 & rr lyrae & p01 & 0.514 & 0.586 & bs-93 + 259.95796 & 57.81791 & 1339 - 1 - 5 - 61 - 958 & 19.77 & -0.02 & 1.17 & rr lyrae & p01 & -0.926 & -0.812 & bs-48 + 259.96361 & 57.83811 & 1336 - 1 - 5 - 61 - 101 & 21.18 & 0.98 & 0.37 & unknown & p01 & 1.081 & 1.178 & + 259.96370 & 57.81209 & 1339 - 1 - 5 - 61 - 967 & 20.09 & 0.11 & 1.02 & rr lyrae & p01 & 0.671 & 0.632 & bs-85 + 259.96382 & 57.88907 & 1336 - 1 - 5 - 61 - 277 & 20.46 & 0.22 & 1.54 & rr lyrae & p01 & 0.482 & 0.508 & bs-196 + & & & & & & & cs86 & 0.726 & 0.905 & + 259.96792 & 57.98671 & 1336 - 1 - 5 - 60 - 245 & 21.79 & 0.66 & 0.92 & unknown & p01 & 0.905 & 1.452 & bs-144 + 259.96842 & 57.87396 & 1336 - 1 - 5 - 61 - 993 & 20.42 & 0.22 & 1.18 & rr lyrae & p01 & -0.284 & -0.446 & bs-33 + 259.97258 & 57.81262 & 1339 - 1 - 5 - 61 - 969 & 20.37 & 0.23 & 0.82 & rr lyrae & p01 & -0.752 & -1.200 & bs-84 + 259.97731 & 57.88219 & 1336 - 1 - 5 - 61 - 946 & 20.42 & 0.23 & 0.96 & rr lyrae & p01 & 0.886 & 0.752 & bs-78 + 259.98065 & 57.86038 & 1336 - 1 - 5 - 61 - 1078 & 20.39 & 0.20 & 1.23 & rr lyrae & p01 & -0.802 & -1.230 & bs-39 + 259.98082 & 57.81681 & 1336 - 1 - 5 - 61 - 479 & 19.55 & 0.04 & 1.16 & rr lyrae & p01 & -0.311 & 0.130 & bs-114 + 259.98440 & 57.87989 & 1336 - 1 - 5 - 61 - 328 & 20.08 & 0.17 & 1.21 & rr lyrae & p01 & 0.609 & 0.926 & bs-32 + 259.98606 & 57.93344 & 1336 - 1 - 5 - 60 - 973 & 20.51 & 0.30 & 1.16 & rr lyrae & p01 & 0.129 & 0.120 & bs-16 + & & & & & & & cs86 & 0.062 & -0.077 & + & & & & & & & cs86 & 0.573 & 0.763 & + 259.99224 & 57.91310 & 1336 - 1 - 5 - 61 - 6 & 20.36 & 0.41 & 1.45 & unknown & p01 & 0.014 & -0.148 & + & & & & & & & cs86 & 0.331 & 0.045 & + & & & & & & & cs86 & 0.176 & 0.003 & + 259.99545 & 57.90403 & 1336 - 1 - 5 - 61 - 826 & 19.99 & 0.18 & 1.18 & rr lyrae & g98 & 0.125 & 0.150 & bs-24 + & & & & & & & p01 & 0.216 & 0.486 & + & & & & & & & cs86 & -0.507 & -0.095 & + & & & & & & & cs86 & -0.265 & 0.286 & + 259.99668 & 57.94619 & 1336 - 1 - 5 - 60 - 922 & 20.47 & 0.28 & 1.11 & rr lyrae & p01 & 1.008 & 0.874 & bs-65 + & & & & & & & cs86 & 0.439 & 1.012 & + & & & & & & & cs86 & -0.004 & -0.118 & + 259.99791 & 58.05947 & 1336 - 1 - 5 - 60 - 597 & 20.38 & 0.24 & 0.97 & rr lyrae & p01 & 1.210 & 1.032 & bs-55 + 259.99839 & 57.91189 & 1336 - 1 - 5 - 61 - 7 & 20.11 & 0.33 & 1.15 & rr lyrae & g98 & -0.124 & -0.051 & + & & & & & & & p01 & -0.338 & -0.426 & + & & & & & & & cs86 & -0.081 & 0.046 & + & & & & & & & cs86 & -0.058 & 0.072 & + & & & & & & & s79 & -0.046 & -0.070 & + & & & & & & & bs61 & 0.009 & -0.154 & + 260.00124 & 57.89062 & 1336 - 1 - 5 - 61 - 915 & 20.45 & 0.23 & 1.14 & rr lyrae & p01 & 1.623 & 1.472 & bs-74 + & & & & & & & cs86 & -0.061 & -0.116 & + & & & & & & & cs86 & 0.299 & 0.524 & + 260.00285 & 57.89625 & 1336 - 1 - 5 - 61 - 267 & 18.02 & 1.13 & 3.07 & unknown & p01 & 0.319 & 0.212 & carbon + & & & & & & & cs86 & -0.049 & -0.256 & + & & & & & & & cs86 & -0.025 & -0.248 & + & & & & & & & s79 & -0.128 & -0.362 & + & & & & & & & bs61 & -0.073 & -0.352 & + 260.00939 & 58.02230 & 1336 - 1 - 5 - 60 - 476 & 19.58 & -0.01 & 1.17 & rr lyrae & p01 & -1.622 & -1.440 & bs-59 + 260.01286 & 57.83103 & 1336 - 1 - 5 - 61 - 1214 & 20.31 & 0.17 & 1.33 & rr lyrae & p01 & -0.336 & -0.442 & bs-113 + 260.02392 & 57.93595 & 1336 - 1 - 5 - 60 - 373 & 20.94 & 0.96 & 0.35 & unknown & p01 & 0.588 & 0.636 & + & & & & & & & cs86 & 0.615 & 0.695 & + & & & & & & & s79 & 0.660 & 0.665 & + & & & & & & & bs61 & 0.729 & 0.476 & + 260.02500 & 57.89691 & 1336 - 1 - 5 - 61 - 884 & 19.61 & 0.08 & 1.24 & rr lyrae & p01 & 0.054 & 0.512 & bs-27 + & & & & & & & cs86 & -0.652 & -0.176 & + 260.02551 & 58.03505 & 1336 - 1 - 5 - 60 - 641 & 19.96 & 0.11 & 1.11 & rr lyrae & p01 & 0.345 & 0.560 & bs-58 + 260.02727 & 57.82912 & 1336 - 1 - 5 - 61 - 1242 & 20.61 & 0.30 & 1.15 & rr lyrae & p01 & 0.376 & 0.390 & bs-186 + 260.02962 & 57.99697 & 1336 - 1 - 5 - 60 - 734 & 20.29 & 0.23 & 1.02 & rr lyrae & p01 & 0.449 & 0.482 & bs-98 + 260.03149 & 57.97373 & 1336 - 1 - 5 - 60 - 290 & 18.47 & 0.70 & 1.58 & unknown & p01 & 0.374 & 0.844 & + & & & & & & & s79 & -0.051 & -0.047 & + & & & & & & & bs61 & 0.003 & 0.111 & + 260.03395 & 58.04203 & 1336 - 1 - 5 - 60 - 637 & 20.30 & 0.37 & 1.19 & rr lyrae & p01 & 0.631 & 0.712 & bs-57 + 260.03553 & 57.79121 & 1336 - 1 - 5 - 61 - 1389 & 20.29 & 0.27 & 1.09 & rr lyrae & p01 & 0.517 & 0.612 & bs-49 + 260.03697 & 57.93967 & 1336 - 1 - 5 - 60 - 375 & 20.44 & 0.22 & 1.25 & rr lyrae & p01 & 1.380 & 1.192 & bs-126 + & & & & & & & cs86 & -0.036 & 0.104 & + 260.03703 & 57.92476 & 1336 - 1 - 5 - 60 - 1097 & 20.48 & 0.31 & 1.05 & rr lyrae & g98 & -0.097 & -0.053 & bs-160 + & & & & & & & p01 & -0.099 & 0.004 & + & & & & & & & cs86 & 0.467 & 0.262 & + 260.03847 & 57.91065 & 1336 - 1 - 5 - 61 - 250 & 20.05 & 0.20 & 1.11 & rr lyrae & g98 & -0.121 & 0.025 & bs-68 + & & & & & & & p01 & -0.839 & -1.024 & + & & & & & & & cs86 & 0.003 & 0.034 & + 260.04789 & 57.96733 & 1336 - 1 - 5 - 60 - 860 & 20.42 & 0.18 & 1.02 & rr lyrae & p01 & 0.970 & 0.972 & bs-123 + 260.05178 & 57.90319 & 1336 - 1 - 5 - 61 - 868 & 19.59 & -0.06 & 1.26 & rr lyrae & p01 & -0.773 & -0.258 & bs-72 + & & & & & & & cs86 & -0.129 & 0.258 & + 260.05709 & 57.95686 & 1336 - 1 - 5 - 60 - 915 & 20.36 & 0.23 & 1.36 & rr lyrae & p01 & -0.766 & -0.814 & bs-184 + & & & & & & & cs86 & 0.187 & 0.011 & + 260.06165 & 57.90899 & 1336 - 1 - 5 - 61 - 627 & 20.24 & 0.67 & 1.15 & unknown & g98 & 0.218 & 0.130 & + & & & & & & & p01 & 0.225 & 0.220 & + & & & & & & & cs86 & 0.210 & 0.219 & + & & & & & & & s79 & 0.166 & 0.123 & + & & & & & & & bs61 & 0.320 & 0.018 & + 260.06334 & 57.98805 & 1336 - 1 - 5 - 60 - 807 & 20.57 & 0.24 & 1.07 & rr lyrae & p01 & -0.373 & -0.570 & bs-8 + 260.06422 & 57.89098 & 1336 - 1 - 5 - 61 - 334 & 20.69 & 0.24 & 1.61 & rr lyrae & p01 & 0.124 & -0.300 & bs-171 + & & & & & & & cs86 & 0.503 & -0.035 & + 260.06524 & 57.90470 & 1336 - 1 - 5 - 61 - 875 & 20.54 & 0.32 & 0.93 & rr lyrae & g98 & -0.010 & 0.075 & bs-71 + & & & & & & & p01 & 0.757 & 0.644 & + & & & & & & & cs86 & 0.396 & 0.736 & + 260.07075 & 57.87772 & 1336 - 1 - 5 - 61 - 1064 & 20.51 & 0.24 & 1.15 & rr lyrae & p01 & -0.227 & -0.246 & bs-34 + & & & & & & & cs86 & 0.645 & 0.030 & + 260.07115 & 57.77800 & 1336 - 1 - 5 - 61 - 1423 & 19.95 & 0.15 & 1.35 & rr lyrae & p01 & -0.351 & 0.006 & bs-140 + 260.07345 & 57.86679 & 1336 - 1 - 5 - 61 - 91 & 20.15 & 0.20 & 1.23 & rr lyrae & p01 & -0.801 & -0.952 & bs-36 + 260.07361 & 57.95369 & 1336 - 1 - 5 - 60 - 950 & 21.87 & 0.51 & 0.93 & unknown & p01 & 0.114 & 0.114 & + & & & & & & & cs86 & -0.020 & 0.173 & + & & & & & & & s79 & 0.117 & -0.016 & + & & & & & & & bs61 & 0.470 & 0.047 & + 260.07414 & 57.95210 & 1336 - 1 - 5 - 60 - 948 & 19.22 & 0.21 & 1.04 & unknown & p01 & -0.376 & -0.384 & bs-141 + & & & & & & & cs86 & -0.308 & -0.078 & + 260.07524 & 57.87247 & 1336 - 1 - 5 - 61 - 92 & 21.25 & 0.10 & 1.37 & unknown & p01 & 0.324 & 0.640 & + 260.08537 & 58.01548 & 1336 - 1 - 5 - 60 - 714 & 20.64 & 0.27 & 1.02 & rr lyrae & p01 & -0.207 & -0.432 & bs-5 + 260.08782 & 57.87185 & 1336 - 1 - 5 - 61 - 1106 & 20.21 & 0.24 & 1.17 & rr lyrae & p01 & 0.386 & 0.258 & bs-81 + & & & & & & & s79 & -0.356 & -0.422 & + & & & & & & & bs61 & -0.289 & -0.653 & + 260.09410 & 57.78254 & 1336 - 1 - 5 - 61 - 1420 & 21.28 & 0.65 & 0.51 & unknown & p01 & -0.322 & -0.574 & + 260.10083 & 57.86138 & 1336 - 1 - 5 - 61 - 1150 & 19.88 & 0.11 & 1.14 & rr lyrae & p01 & -0.579 & -0.124 & bs-132 + 260.10812 & 57.94047 & 1336 - 0 - 5 - 60 - 421 & 20.07 & 0.43 & 0.96 & unknown & p01 & 0.059 & 0.142 & + & & & & & & & s79 & -0.332 & -0.169 & + & & & & & & & bs61 & -0.188 & -0.348 & + 260.11124 & 57.90755 & 1336 - 1 - 5 - 61 - 290 & 20.51 & 0.20 & 1.20 & rr lyrae & p01 & -0.550 & -0.934 & bs-106 + 260.11766 & 57.95019 & 1336 - 1 - 5 - 60 - 417 & 19.85 & 0.06 & 1.04 & rr lyrae & p01 & 0.287 & 0.672 & bs-103 + 260.12558 & 57.84441 & 1336 - 1 - 5 - 61 - 139 & 19.73 & 0.01 & 1.08 & rr lyrae & p01 & 0.101 & 0.334 & bs-135 + 260.12974 & 57.96027 & 1336 - 1 - 5 - 60 - 958 & 20.43 & 0.29 & 1.06 & rr lyrae & p01 & 0.435 & 0.308 & bs-158 + 260.13549 & 57.91925 & 1336 - 1 - 5 - 61 - 32 & 20.63 & 0.36 & 0.98 & rr lyrae & p01 & 1.466 & 1.368 & bs-21 + & & & & & & & s79 & -0.058 & 0.538 & + & & & & & & & bs61 & -0.025 & 0.506 & + 260.13768 & 57.95849 & 1336 - 1 - 5 - 60 - 986 & 19.51 & -0.12 & 1.29 & rr lyrae & p01 & -0.884 & -0.156 & bs-124 + 260.14471 & 57.89721 & 1336 - 1 - 5 - 61 - 1016 & 21.94 & 1.54 & 1.38 & unknown & p01 & 0.121 & 0.110 & + & & & & & & & bs61 & 0.498 & 0.342 & + 260.14936 & 57.82179 & 1336 - 1 - 5 - 61 - 1346 & 20.44 & 0.45 & 0.83 & unknown & p01 & 0.735 & 0.450 & bs-138 + 260.15351 & 57.87019 & 1336 - 1 - 5 - 61 - 660 & 20.23 & 0.19 & 1.06 & rr lyrae & p01 & 0.529 & 0.478 & bs-40 + 260.15828 & 57.92527 & 1336 - 1 - 5 - 61 - 848 & 20.32 & 0.32 & 1.46 & rr lyrae & p01 & 1.040 & 0.932 & bs-162 + 260.15994 & 57.87658 & 1336 - 1 - 5 - 61 - 408 & 20.44 & 0.29 & 1.27 & rr lyrae & p01 & -0.364 & -0.430 & bs-35 + 260.16906 & 57.91449 & 1336 - 1 - 5 - 61 - 936 & 20.50 & 0.26 & 0.93 & rr lyrae & p01 & 0.587 & 0.530 & bs-161 + 260.17357 & 57.83108 & 1336 - 1 - 5 - 61 - 1332 & 20.01 & 0.09 & 1.14 & rr lyrae & p01 & 0.819 & 1.026 & bs-82 + 260.17433 & 57.96383 & 1336 - 1 - 5 - 60 - 996 & 20.48 & 0.25 & 0.83 & rr lyrae & p01 & -0.439 & -0.524 & bs-12 + 260.17624 & 57.98114 & 1336 - 1 - 5 - 60 - 895 & 20.38 & 0.21 & 0.93 & rr lyrae & p01 & 0.745 & 0.768 & bs-169 + 260.17851 & 57.85796 & 1336 - 1 - 5 - 61 - 1235 & 19.64 & 0.02 & 1.11 & rr lyrae & p01 & 0.224 & 0.758 & bs-41 + 260.18800 & 57.85765 & 1336 - 1 - 5 - 61 - 1239 & 19.58 & 0.02 & 1.04 & rr lyrae & p01 & -0.434 & -0.088 & bs-164 + 260.19653 & 57.96645 & 1336 - 1 - 5 - 60 - 999 & 19.31 & -0.17 & 1.16 & rr lyrae & p01 & -0.100 & 0.550 & bs-13 + 260.19697 & 57.92301 & 1336 - 1 - 5 - 61 - 902 & 19.92 & 0.04 & 1.29 & rr lyrae & p01 & -0.533 & -0.140 & bs-185 + 260.20741 & 57.90137 & 1336 - 1 - 5 - 61 - 1058 & 20.42 & 0.21 & 1.06 & rr lyrae & p01 & -0.547 & -0.780 & bs-25 + 260.21321 & 57.86324 & 1336 - 1 - 5 - 61 - 130 & 20.53 & 0.38 & 0.85 & rr lyrae & p01 & 1.061 & 0.848 & bs-133 + 260.21460 & 57.97171 & 1336 - 1 - 5 - 60 - 509 & 19.26 & 0.63 & 1.46 & unknown & p01 & 0.821 & 1.548 & + 260.21700 & 58.02371 & 1336 - 1 - 5 - 60 - 767 & 19.57 & -0.01 & 1.22 & rr lyrae & p01 & -1.409 & -0.946 & bs-94 + 260.21794 & 57.92034 & 1336 - 1 - 5 - 61 - 964 & 20.22 & 0.24 & 0.31 & rr lyrae & p01 & -0.377 & 0.258 & + & & & & & & & bs61 & 0.253 & 0.344 & + 260.21987 & 57.84490 & 1336 - 1 - 5 - 61 - 159 & 19.61 & -0.01 & 1.42 & rr lyrae & p01 & 0.441 & 0.944 & bs-177 + 260.22526 & 57.95146 & 1336 - 1 - 5 - 61 - 777 & 20.71 & 0.83 & 0.79 & unknown & p01 & 0.342 & 0.434 & + 260.23602 & 57.89786 & 1336 - 1 - 5 - 61 - 1100 & 20.53 & 0.28 & 1.33 & rr lyrae & p01 & 1.297 & 1.144 & bs-75 + 260.23839 & 57.83363 & 1336 - 1 - 5 - 61 - 1361 & 19.48 & -0.05 & 1.10 & rr lyrae & p01 & -1.615 & -1.480 & bs-178 + 260.24357 & 57.89214 & 1336 - 1 - 5 - 61 - 649 & 20.42 & 0.29 & 0.97 & rr lyrae & p01 & 1.051 & 1.082 & bs-175 + 260.24627 & 58.00153 & 1336 - 1 - 5 - 60 - 858 & 19.78 & -0.07 & 1.28 & rr lyrae & p01 & -0.708 & -0.574 & bs-97 + 260.24700 & 58.02391 & 1336 - 1 - 5 - 60 - 270 & 19.47 & -0.03 & 1.16 & rr lyrae & p01 & -0.079 & 0.420 & bs-118 + 260.25222 & 57.86466 & 1336 - 1 - 5 - 61 - 1267 & 21.37 & -0.13 & 0.61 & unknown & p01 & 0.579 & 0.392 & + 260.26119 & 57.88066 & 1336 - 1 - 5 - 61 - 673 & 19.76 & 0.07 & 1.10 & rr lyrae & p01 & 0.734 & 0.988 & bs-80 + 260.26461 & 57.99732 & 1336 - 1 - 5 - 60 - 888 & 20.36 & 0.23 & 1.04 & rr lyrae & p01 & 0.796 & 0.896 & bs-96 + 260.27663 & 57.86468 & 1336 - 1 - 5 - 61 - 1285 & 20.06 & 0.11 & 1.20 & rr lyrae & p01 & 1.021 & 1.154 & bs-112 + 260.27901 & 57.99500 & 1336 - 1 - 5 - 60 - 913 & 20.46 & 0.28 & 1.07 & rr lyrae & p01 & -0.325 & -0.620 & bs-95 + 260.27955 & 57.90240 & 1336 - 1 - 5 - 61 - 396 & 20.39 & 0.24 & 1.06 & rr lyrae & p01 & -0.327 & -0.696 & bs-70 + 260.28048 & 57.96674 & 1336 - 1 - 5 - 60 - 1073 & 20.34 & 0.16 & 1.23 & rr lyrae & p01 & 1.171 & 1.042 & bs-183 + 260.28602 & 57.79628 & 1336 - 1 - 5 - 62 - 310 & 20.14 & 0.18 & 1.24 & rr lyrae & p01 & 0.295 & 0.540 & bs-87 + 260.30299 & 58.02519 & 1339 - 1 - 6 - 61 - 405 & 19.94 & 0.03 & 1.34 & rr lyrae & p01 & 0.416 & 0.750 & bs-181 + 260.30423 & 57.89739 & 1336 - 1 - 5 - 61 - 1147 & 19.65 & 0.02 & 1.04 & rr lyrae & p01 & -0.437 & 0.146 & bs-189 + 260.30846 & 57.90981 & 1336 - 1 - 5 - 61 - 1102 & 20.33 & 0.25 & 0.94 & rr lyrae & p01 & -0.221 & -0.470 & bs-128 + 260.31889 & 57.89216 & 1336 - 1 - 5 - 61 - 1178 & 19.98 & 0.14 & 1.20 & rr lyrae & p01 & -0.959 & -0.904 & bs-26 + 260.32756 & 57.79548 & 1336 - 1 - 5 - 62 - 319 & 19.84 & 0.10 & 1.21 & rr lyrae & p01 & -0.497 & -0.026 & bs-199 + 260.32928 & 57.89011 & 1336 - 1 - 5 - 61 - 1193 & 20.35 & 0.20 & 0.91 & rr lyrae & p01 & 0.897 & 0.902 & bs-130 + 260.33418 & 57.82171 & 1336 - 1 - 5 - 61 - 1426 & 20.35 & 0.13 & 1.07 & rr lyrae & p01 & 0.731 & 0.670 & bs-166 + 260.34115 & 57.82410 & 1336 - 1 - 5 - 61 - 1422 & 20.16 & 0.05 & 1.17 & rr lyrae & p01 & 0.456 & 0.194 & bs-139 + 260.35674 & 57.84176 & 1336 - 1 - 5 - 61 - 190 & 20.44 & 0.35 & 1.09 & rr lyrae & p01 & 0.764 & 0.818 & bs-197 + 260.37540 & 57.97799 & 1339 - 1 - 6 - 61 - 469 & 20.24 & 0.22 & 1.06 & rr lyrae & p01 & 0.450 & 0.400 & bs-61 + 260.39565 & 57.85005 & 1339 - 1 - 6 - 62 - 330 & 20.79 & 0.78 & 0.98 & unknown & p01 & 0.508 & 0.150 & + 260.40182 & 57.93293 & 1339 - 1 - 6 - 61 - 551 & 20.19 & 0.14 & 1.18 & rr lyrae & p01 & 0.482 & 0.338 & + 260.41938 & 57.91002 & 1339 - 1 - 6 - 61 - 307 & 19.90 & 0.11 & 1.18 & rr lyrae & p01 & -0.582 & -0.106 & + 260.42271 & 57.87652 & 1339 - 1 - 6 - 62 - 301 & 21.52 & 0.36 & 0.26 & unknown & p01 & 0.555 & 0.000 & + 260.42448 & 57.90823 & 1339 - 1 - 6 - 61 - 584 & 20.40 & 0.22 & 1.01 & rr lyrae & p01 & 0.537 & 0.346 & + 260.44703 & 57.88509 & 1339 - 1 - 6 - 62 - 293 & 20.24 & 0.21 & 1.14 & rr lyrae & p01 & 0.302 & 0.084 & + 260.45115 & 57.96817 & 1339 - 1 - 6 - 61 - 504 & 20.11 & 0.07 & 0.16 & rr lyrae & p01 & -0.538 & -0.326 & qso +
|
the sloan digital sky survey ( sdss ) has scanned the entire region containing the draco dwarf spheroidal galaxy to @xmath0 magnitude in @xmath1 .
we present a catalog of stars found in a 453 square arcminute , elliptical region centered on the draco dwarf spheroidal galaxy .
objects in the catalog are matched with five previously published catalogs .
the catalog contains sdss photometry for 5634 individual objects , and also the photometry from matches to any of the other catalogs .
a comparison of the photometry between catalogs allows us to identify 142 candidate variable objects .
one hundred and twelve of the suspected variables have colors consistent with rr lyrae variables .
|
we consider here a database on mixed marriages and newborns to mixed couples , annually recorded by istat ( italian institute of statistics ) for all italian municipalities in the time period from 2001 to 2011 . in that time span
documented foreign born population has grown in italy from just over 1 million to almost 4 millions people , corresponding to an increase from @xmath1 to over @xmath2 of the total resident population .
these figures do not include illegal immigrants whose numbers are difficult to determine . although there are immigrants from almost all nations , the communities with higher presence are from romania ( @xmath3,@xmath4 in @xmath5 , i.e @xmath6 of the total population of immigrants ) , followed by albania ( @xmath7,@xmath4 in @xmath5 , i.e @xmath8 ) , morocco ( @xmath9,@xmath4 in @xmath5 , i.e @xmath10 ) , and china ( @xmath11,@xmath4 in @xmath5 , i.e @xmath12 ) .
the geographical distribution of immigrants is largely not homogeneous : @xmath13 lives in the northern and central part of the country ( the most economically developed areas ) , while @xmath14 lives in the south .
the dataset contains over @xmath15,@xmath16,@xmath4 data , yearly describing - for each of the @xmath17,@xmath16 municipalities - the total population , the number of immigrants , the number of marriages and newborns originating from different types of couples ( either mixed or not ) . for each municipality
we have considered the immigrant density @xmath18 where @xmath19 is the number of immigrants and @xmath20 is the number of the natives . a useful parameter , measuring the number of possible cross - links among natives and immigrants is @xmath21 .
we focussed , in particular , on two quantifiers : the fraction of marriages with spouses of mixed origin ( native and immigrant ) @xmath22 and the fraction of newborns with parents of mixed origin @xmath23 .
the entire marriages dataset contains @xmath24,@xmath25 records but only @xmath26,@xmath27 points were considered in our study .
this is due to the fact that in tiny villages , occasionally , no marriages occurred .
those events account for about @xmath28 of the whole dataset .
moreover , in about @xmath29 of all records no mixed marriages have occurred .
analogously the inputs from municipalities where no newborns occurred account for about @xmath30 of the whole dataset , whereas for about @xmath31 of all records no mixed newborns have occurred .
we separated the dataset into two parts : one for the small cities ( below @xmath32,@xmath4 inhabitants ) and the other for large ones ( above @xmath32,@xmath4 inhabitants ) .
figure [ figraw ] shows the resulting collection of the raw data of each part for the mixed marriages ( upper panels ) and for the newborns of mixed couples ( lower panels ) in the planes @xmath33 and @xmath34 , respectively .
we observe that records belonging to villages and small cities , whose population is under @xmath32,@xmath4 inhabitants , count up to @xmath35.figure [ figdens ] displays the data densities and shows statistical robustness ( up to one percentile ) for all @xmath36 up to @xmath37 . the partition of both marriages and newborns datasets , which has led us to the main results of this work , has been investigated under various perspectives .
we remark that this partition was proposed since an analysis performed over a unified datasets ( small and large cities together ) happened to be unsatisfactory .
it displayed in fact a high dependence on the binning parameter settings revealing the typical presence of _ data mixture _ of interacting and non - interacting type .
this situation led us to partition each dataset , separating small municipalities from large ones .
the implemented threshold ( i.e. @xmath32,@xmath4 inhabitants ) was attempted according to previous work @xcite , where the considered municipalities consisted only of cities over @xmath32,@xmath4 inhabitants .
ex - post an optimisation test on the threshold has been performed showing that the proposed one was a stable choice , i.e. the coefficients of determination ( @xmath38 ) of the data fitting in the two regimes were maximal . moreover , for the chosen partition we verify both robustness and homogeneity among the two datasets ( large and small cities ) in order to exclude pathologies by imbalance .
our tests show that the immigrant proportions over the total population in large and small cities are comparable for each year with a maximum difference between the two of about one percent .
we can therefore infer that immigrants seem to be distributed in the same manner , independently from the size of the municipality they belong to . eventually , the ratio between the total population living in small municipalities and the one living large municipalities is nearly constant ( @xmath39 ) over the years .
we tested , moreover , the behavior of the temporal series ( year by year ) of the proportions of immigrants , of natives and of the total population living in the small cities .
this study shows that the national fraction of immigrants and natives in small municipalities are about the same ( @xmath40 ) .
furthermore , a similar value ( slightly more than @xmath40 ) is found for the total population that lives in cities with less than @xmath32,@xmath4 inhabitants . since our study
is devoted to investigate the average behavior at the country scale for the two quantifiers @xmath22 and @xmath23 in large cities and small ones , we estimate the average percentage of mixed marriages or mixed newborns , for a given immigrant density . to this purpose
we performed a _
mediant _ average and binning with constant information ( see @xcite for details ) .
figure [ figradline ] displays the output averages versus @xmath41 .
the emerging behaviors are well fitted ( see the @xmath38 coefficient ) by two different laws : square root for the quantifiers on small municipalities and linear on large ones
. the statistical physics approach developed and broadly described in the technical paper @xcite suggests an interpretation for these results .
we know in fact that the linear behaviour emerges , in strong ties conditions , from collective effects when the network is sparse and unpercolated .
the links are rare and the connected groups are made of only few units . on the other side ,
the square root law emerges when the social framework of strong personal ties , is built on a fully connected network where the cases of isolated individuals or small groups inside the communities are rare or completely absent . in order to better test the predictability power of our theory
we have performed our analysis not only for whole time span of the observation data , but also on increasing time sub - spans .
figure [ figr2 ] shows that the distinction among linear and square root growth is already evident from elaboration coming from only @xmath42 data , and stay stable for increasing intervals @xmath43 etc , up to the whole set of data ( @xmath44 ) .
our results are in good agreement with the classical sociological theories of alienation and anomie about the social behavior on large cities , where social connections are seldom and ineffective , in comparison to villages where they are strong ( see @xcite ) .
our results moreover have the potential to impact policy makers .
in fact , the identification of the growth law of each quantifier in @xmath36 allows to predict the level of the quantifiers for growing values of @xmath36 .
tilly , c. , _ trust and rules _ , ( cambridge university press new york , 2005 ) .
blau , p.m. , beeker , c. , fitzpatrick k.m . , intersecting social affiliations and intermarriage _ social forces _ * 62 * , 585 - 606 ( 1984 ) .
barra , a. , contucci , p. , sandell , r. , vernia , c. , an analysis of a large dataset on immigrant integration in spain .
the statistical mechanics perspective on social action .
rep . _ * 4 * , 4174 , ( 2014 ) .
we thank adriano barra and rickard sandell for interesting discussions .
we also acknowledge claudio giberti for valuable suggestions and for a careful reading of the manuscript .
we thank francesco de pretis for his help at a very early stage of the manuscript .
this work was supported by the firb grant rbfr10n90w and by the prin grant 2010hxaw77 .
local view of a sparse unpercolated graph .
individuals are either isolated or , seldomly , connected with few others .
our findings indicate this structure for the average local trust network in large italian cities , width=377,height=377 ] local view of a percolated graph .
individuals are connected with several others .
the graph , a part for few individuals , is completely connected .
our finding indicate this structure for the local average trust network in small italian cities.,width=377,height=377 ] raw data versus immigrant density @xmath36 .
blue points represent the fraction of mixed marriages occurred in municipalities with less than @xmath32,@xmath4 inhabitants where a percentage @xmath36 of migrants is present ; similarly green points account for newborns from mixed couples .
magenta points represent the fraction of mixed marriages occurred in municipalities with more than @xmath32,@xmath4 inhabitants , while orange ones mirror the newborns from mixed couples .
one may note that data in the left panels seem to lie along horizontal lines displaced according to @xmath45 , with @xmath46 due to effects seen in small - sized municipalities.,width=321,height=226 ] data density for the mixed marriages and mixed newborns quantifiers as a function of the percentage of immigrants @xmath36 . left upper panel : density of the mixed marriages ( blue circles ) dataset for municipalities with less than @xmath32,@xmath4 inhabitants .
right upper panel : density of the mixed marriages ( magenta circles ) dataset for municipalities with more than @xmath32,@xmath4 inhabitants .
lower panels display the same information for the newborns in small ( green circles ) and large ( orange circles ) municipalities . in the insets , the marriages and newborns data density plots are fitted , for @xmath47 , with the power - law behavior ( in log - log scale ) where @xmath48 with @xmath49 ( small municipalities ) and @xmath50 ( large municipalities ) . in all panels
the data densities decrease close to @xmath51 .
this comes from the fact that the migration phenomena was already ongoing when the data collection was started ( @xmath42 ) and the density of migrants in italy was larger than zero.,width=321,height=226 ] dots are average quantities for the the mixed marriages and mixed newborns quantifiers versus @xmath41 .
left upper panel : quantifier @xmath22 ( blue dots ) , fraction of mixed marriages occurred in municipalities with less than @xmath32,@xmath4 inhabitants , with the best square root fit ( red curve ) @xmath52 ( a = 0.233 @xmath530.009 , b = [email protected] , with a goodness of fit @xmath54 = 0.97 computed for @xmath41<0.13 ) .
right upper panel : quantifier @xmath22 ( magenta dots ) , fraction of mixed marriages occurred in municipalities with more than @xmath32,@xmath4 inhabitants , with the best linear fit ( black curve ) @xmath55 ( a = 1.00 @xmath530.05 , b = [email protected] , with a goodness of fit @xmath38 = 0.98 computed for @xmath41<0.08 ) . left lower panel : quantifier @xmath23 ( green dots ) , fraction of newborns with mixed parents , born in municipalities with less than 10.000 inhabitants , with the best square root fit ( red curve ) @xmath52 ( a = 0.174 @xmath530.008 , b = [email protected] , with a goodness of fit @xmath54 = 0.96 computed for @xmath41<0.10 ) .
right lower panel : quantifier @xmath23 ( orange dots ) , fraction of newborns with mixed parents , born in municipalities with more than @xmath32,@xmath4 inhabitants , with the best linear root fit ( black curve ) @xmath55 ( a = 0.78 @xmath530.04 , b = [email protected] , with a goodness of fit @xmath54 = 0.99 computed for @xmath41<0.07),width=321,height=226 ] determination coefficients @xmath38 of different fits ( linear and square root functions ) for the averaged quantifiers shown in figure 3 against an increasing cumulative time span : the first cumulative time span includes data from 2001 , the second cumulative time span data from the interval 2001 - 2002 , up to the last that includes data from the entire period @xmath44 .
@xmath38 are represented through red curves when related to a square root fit and through a black curve when related to a linear fit .
left upper panel : @xmath38 ( blue dots ) for the mixed marriages occurred in municipalities with less than @xmath32,@xmath4 inhabitants .
right upper panel : @xmath38 ( magenta dots ) related to mixed marriages occurred in municipalities with more than @xmath32,@xmath4 inhabitants .
analogously in the lower panels for the newborns in small ( green circles ) and large ( orange circles ) municipalities.,width=321,height=226 ]
|
we study the structure of a social network of strong ties ( trust network ) investigating its property of connectedness versus fragmentation . to this purpose
we analyse an extensive set of census data , about marrying or having children with immigrants , collected by italian national statistical institute for all italian municipalities from 2001 to 2011 . not using neither obtaining personal local information but only average ones
, our method fully complies with privacy and confidentiality .
our findings show that large cities display the behaviour of highly fragmented trust networks where individuals face possible phenomena of alienation .
smaller cities and villages instead behave like fully connected social systems with a rich tie structure , where isolation is rare or completely absent .
while confirming classical sociological theories on alienation in large urban areas our approach provides a quantitative method to test them and a predictive tool for policy makers . human choices ,
as clearly elucidated in the work of max weber @xcite , can be distinguished according to the fact that they are _ social actions _ or not .
sometimes , mostly when the matter of the choice is simple and the conclusions can be reached with elementary reasoning , we draw a decision by ourselves , without relying on other opinions and only counting on our pre - existing cultural background . some other times ( the case of social action ) , when the choice is of complex nature or when the information we have is not enough to draw a rational conclusion we proceed , _ whenever possible _ , by relating to others either by imitation or seeking for interaction and consensus .
those two types of individual choice mechanisms lead to two different emerging collective social behaviour each displaying peculiar features .
the case of independent choices usually come with smooth , locally linear , behaviour while the other may present different characteristics , including tipping points , according to the nature and structure of the interactions among the individuals and the topological features of the underlying social network . in this paper
we investigate , by analysing an extensive dataset about italy , some features of the _ trust _ social network @xcite involved with two specific choices .
these are the choice of marrying a person from your own country or from a different one , and the choice of having a child under the same two alternatives .
those _ coupling _ choices are among the most important people face in their existence .
this is due not only to their strong intrinsic emotional content but also because norms and regulations , of both moral and legal nature , make them basically life long .
such choices , with the complexity they come with , fully belong to the category of social actions .
this is both a sociological finding @xcite and also a recent conclusion of the work @xcite .
possible differences observed in their collective social appearance depend therefore on the underlying social network . in particular if the social network is extremely sparse ( figure [ figunp ] ) , with links that are so rare that do not allow the whole group to connect and percolate , we expect a social behaviour similar to a group of independent individuals .
conversely , if the network is globally well connected ( figure [ figperc ] ) , with rare cases of isolated individuals or small groups , we expect a genuine collective effect to emerge .
distinguishing among these two types of social network has important sociological implications .
the network involved with the two mentioned choices is indeed a crucial one since it is made by strong personal ties @xcite possibly affecting also other relevant personal choices .
being part of a well connected social group , with a shared and mutually approved behaviour , provides individuals with a sense of moral rightness , safety and availability of support .
if instead the network is sparse and the connected groups are made of few units the same perception of belonging to a group is lacking and individuals feel alienated @xcite . the key idea that we advance in this paper is that it is possible , from the the observation of the global social behaviour , to infer what type of network the choice is holding on . in order to derive the network structure
we investigate how the frequency of each different coupling choice changes when immigrant s percentage increases .
in particular we analyse italian data about the frequency of mixed marriages and newborn children from mixed couples .
our findings show that , up to an immigrant density of 7% , there are two different growth laws according to the size of the municipality . for large cities
( with more than 10k inhabitants ) if the frequency has grown to a given amount @xmath0 when the immigrant density is 1% it will grow to 2@xmath0 at 2% , 3@xmath0 at 3% and so on . in small cities
the behaviour is different : in order to reach the frequency 2@xmath0 the immigrant density has to grow to 4% , to reach 3@xmath0 it has to grow to 9% and so on .
while the first law is linear , i.e. has a constant growth rate the second , that follows the square root function , has an anomalous high growth rate when the phenomenon starts , which quickly decreases at increasing immigrant densities .
we interpret these results within the statistical physics framework which is largely used in many research fields to study the behavior of quantities that can be represented as large sums of mutually correlated random variables ( see @xcite and references there in ) . on the basis of a statistical physics model developed in @xcite
we argue that the two different growth laws depend on the different social network structures .
in particular the large italian cities display the behavior of sparse unpercolated networks while the small ones behaves like fully connected networks .
in other words individuals living in large italian cities turn out to be in alienated conditions while smaller cities provide a full social environment .
this conclusion provides a quantitative confirmation of the general sociological theories about alienation and anomie in large cities @xcite .
our study does nt use nor produce any information of proximity like who trusts who and is fully compliant with privacy regulations .
starting from collective information we deduce average topological properties of the network .
moreover the fact that our approach is purely of observational nature and does nt present any type of solicitation to individuals nor to the system as a whole , makes our approach ethically viable .
|
consider the beltrami equation @xmath7 where @xmath8 is a lebesgue measurable function on the complex plane @xmath9 satisfying @xmath10 . according to a remarkable old theorem of morrey @xcite
there exists an essentially unique function @xmath11 in the sobolev space @xmath12 ( functions with first order derivatives locally in @xmath13 ) which satisfies almost everywhere and is a _ homeomorphism of the plane_. these functions are called quasiconformal .
it turns out that @xmath11 may change drastically the hausdorff dimension of sets .
indeed , sets of arbitrarily small positive hausdorff dimension may be mapped into sets of hausdorff dimension as close to @xmath14 as desired ( and the other way around by the inverse mapping ) .
there has been during the last decades much hard and penetrating work in understanding how @xmath11 distorts sets ( see , for instance @xcite and the references given there or @xcite for a recent result ) .
on the other hand , orientation preserving bilipschitz homeomorphisms of the plane are easily seen to satisfy a beltrami equation for a certain beltrami coefficent @xmath15 since bilipschitz mappings preserve all metric properties of sets , in particular hausdorff dimension , they appear to be a distinguished subclass of particularly simple quasiconformal mappings . in @xcite
one gives geometric conditions which are necessary and sufficient for @xmath11 being bilipschitz , but which do not involve the beltrami coefficient @xmath8 .
in fact , it is widely accepted that the problem of characterizing in an efficient way those @xmath8 which determine bilipschitz mappings is hopeless .
a classical result that goes back to schauder ( @xcite ) asserts that @xmath11 is of class @xmath16 provided @xmath8 is a compactly supported function in @xmath17 .
it is then not difficult to see that @xmath11 is indeed bilipschitz .
the main result of this paper identifies a class of non - smooth functions @xmath8 which determine bilipschitz quasiconformal mappings @xmath11 .
[ t ] let @xmath18 , @xmath19 , be a finite family of disjoint bounded domains of the plane with boundary of class @xmath2 , @xmath20 , and let @xmath21 where @xmath22 is of class @xmath23 assume in addition that @xmath10 .
then the associated quasiconformal mapping @xmath11 is bilipschitz .
notice that the boundaries of the @xmath24 may touch , even on a set of positive length and , of course , @xmath8 may have jumps on the boundary of some @xmath25 in particular , if we only have one domain and @xmath8 is constant we obtain the following corollary .
[ c1 ] if @xmath1 is a bounded domain of the plane with boundary of class @xmath2 , @xmath20 , and @xmath26 , where @xmath27 is a complex number such that @xmath28 , then the associated quasiconformal mapping @xmath11 is bilipschitz .
if @xmath1 is a disc then corollary 1 reduces to the fact that @xmath11 can be computed explicitly and that one can check by direct inspection that is bilipschitz . if @xmath1 is a square @xmath29 , then one can show that the mapping @xmath11 associated to @xmath30 is not lipschitz for some @xmath27 of modulus less than @xmath6 , so that the corollary and thus the theorem are sharp as far as the smoothness of the boundaries of the @xmath24 is concerned .
recall that a @xmath8-quasi - regular function on a domain @xmath31 is a complex function @xmath3 in @xmath32 satisfying ( 1 ) , with @xmath11 replaced by @xmath3 , almost everywhere in @xmath31 . by stoilow
s factorization theorem , @xmath33 for some holomorphic function @xmath34 on @xmath35 from the theorem we then conclude that @xmath3 is locally lipschitz on @xmath31 .
this improves on mori s theorem , which asserts that , for general @xmath8 , @xmath3 is locally in @xmath36 for @xmath37 .
thus , from the perspective of pde , the theorem may also be viewed as a regularity result for the beltrami equation .
the beltrami equation is intimately related to second order elliptic equations in divergence form ot the type @xmath38 where @xmath39 is a @xmath40 symmetric elliptic matrix with bounded measurable coefficients and determinant @xmath6 ( see ( * ? ? ?
* chapter 13 ) ) .
indeed , the real and imaginary parts of a solution to the beltrami equation satisfy , where the entries of the matrix @xmath41 are given explicitly in terms of the beltrami coeficient .
conversely , given a solution @xmath42 of , one may find a solution of an appropiate beltrami equation whose real part is @xmath43 thus for regularity issues one can work indistinctly with the beltrami equation or with equation .
the proof of the theorem gives , in particular , the following regularity result for solutions of equation .
[ c2 ] let @xmath24 , @xmath19 , be a finite family of disjoint bounded domains of the plane with boundary of class @xmath2 , @xmath20 , and assume that each @xmath24 is contained in bounded domain @xmath31 with boundary of class @xmath2 .
let @xmath44 a @xmath40 symmetric elliptic matrix with determinant @xmath6 and entries supported in @xmath45 and belonging to @xmath46 @xmath47 let @xmath42 be a solution of equation in @xmath48 let @xmath49 stand for the set of points in @xmath31 at distance greater than @xmath50 from the boundary of @xmath48 then @xmath51 for @xmath52 and @xmath47 in particular , @xmath53 and @xmath42 is a locally lipschitz function in @xmath48 the main point of the corollary above is that each solution of is locally lipschitz in @xmath54 while the classical de giorgi - nash theorem gives only that @xmath42 satisfies locally a lipschitz condition of order @xmath55 for some @xmath56 satisfying @xmath57 see section 8 for an extension to more general domains , which may have cusps .
there is some overlapping here with previous results by li and vogelius ( @xcite ) and li and nirenberg ( @xcite ) .
see at the end of the introduction for more about that .
another application of our theorem concerns removability problems .
there has recently been a renewed interest in gaining a better understanding of the nature of removable sets for bounded quasi - regular functions ( see @xcite , @xcite and @xcite ) . since bilipschitz mappings preserve removable sets for bounded holomorphic functions ( @xcite ) , the theorem immediately says that the removable sets for bounded @xmath8-quasi - regular functions , with @xmath8 as in the theorem , are exactly the removable sets for bounded holomorphic functions .
if @xmath1 is a domain , the @xmath58 norm of a function @xmath3 on @xmath1 is @xmath59 where @xmath60 the main difficulty in proving the theorem lies in the fact that no smallness assumption is made on @xmath61 . in the same vein ,
corollary 1 is much more difficult to prove if @xmath62 is close to @xmath6 .
if one assumes that @xmath63 is small enough ( depending on @xmath24 ) for each @xmath64 , then the theorem becomes easier .
similarly , the corollary 1 becomes easier under the assumption that @xmath65 .
see a sketch of the argument at the end of section 2 .
the scheme for the proof of the theorem is inspired by a clever idea of iwaniec @xcite in the context of @xmath66 spaces , which has been further exploited in @xcite .
this idea brings into play the index theory of fredholm operators on banach spaces and , thus , compact operators .
our underlying banach space is @xmath67 a domain with boundary of class @xmath2 , and on this space we estimate the beurling transform and its powers .
we also show that the commutator between the beurling transform and certain functions is compact on appropriate larger lipschitz spaces .
the beurling transform is the principal value convolution operator @xmath68 the fourier multiplier of @xmath69 is @xmath70 , or , in other words , @xmath71 thus @xmath69 is an isometry on @xmath72 .
our main lemma shows that for each even smooth homogeneous caldern - zygmund operator @xmath73 the mapping @xmath74 sends continuously @xmath75 into itself , where @xmath1 is a bounded domain with boundary of class @xmath76 throughout the paper we understand that , for @xmath77 @xmath78 the above boundedness result fails if @xmath73 is not even . as a simple example , one may take as @xmath73 the hilbert transform and as @xmath1 the interval @xmath79 .
the even character of @xmath73 is used in the proof of the main lemma in the form @xmath80 which should be understood as a local version of the global cancellation property @xmath81 common to all smooth homogeneous caldern - zygmund operators .
this was proved by iwaniec for the beurling transform in @xcite . in section 2
we present a detailed sketch of the proof and we introduce the lemmas required . in section 3 the main lemma is proved in @xmath82 .
section 4 deals with commutators .
we compare in section 5 the operator @xmath83 with @xmath84 , @xmath1 a bounded domain with boundary of class @xmath2 , and we show that the difference is compact on @xmath85 , for @xmath86
. in section 6 one completes the proof of the theorem for the case of one domain .
section 7 contains the reduction to the one domain case . in section 8
we present an extension of the theorem to what seems to be its more natural setting , namely domains with cusps whose boundary is of class @xmath16 off the set of cusps .
applications to the regularity theory of solutions of equation is this setting are also mentioned .
after a first version of the paper was completed , daniel faraco brought to our attention the work of li and vogelius @xcite ( and @xcite ) , which deals with lipschitz regularity for the equation in @xmath82 for matrices with entries satisfying a lipschitz condition of order @xmath87 on finitely many disjoint domains with boundary of class @xmath16 , but with possible jumps across the boundaries .
the closures of the domains were disjoint and a main point was to obtain gradient estimates independent of the mutual distances between the closed domains .
this is not an issue for our methods , which even allow touching domains .
moreover , as stated before in corollary 2 , for each solution of we obtain a regularity of class @xmath88 , for each @xmath89 in each domain .
this is almost the expected best possible result , namely @xmath16 in each domain . in @xcite
there is a more substantial loss , due to the techniques employed . on the other hand ,
the setting in @xcite and @xcite is more general , in the sense that one works in @xmath82 , there is no restriction on the determinant of the matrix and also non - homogeneous terms are considered .
we also learnt from antonio crdoba that the regularity theory of the euler equation in 2d , in particular the regularity theory of vortex patches , makes broad use of even caldern - zygmund operators , the beurling transform in particular .
we then became aware of the article @xcite , in which one also proves the main lemma .
however , the proof there is different , and certainly not as much in the caldern - zygmund tradition as ours .
first of all , there is a standard factorization method in quasiconformal mapping theory that reduces the theorem to the case of only one domain @xmath90 the argument is presented in detail in section 7 .
then , from now on we will assume that @xmath8 vanishes off some domain @xmath1 with boundary of class @xmath2 and that @xmath91 as is well known , @xmath11 is given explicitly by the formula @xcite @xmath92 where @xmath93 is the cauchy transform of @xmath34 .
recall the important relation between the cauchy and the beurling transforms : @xmath94 , @xmath95 .
the function @xmath96 is determined by the equation @xmath97 as soon as we can invert the operator @xmath98 on @xmath85 , for some @xmath99 satisfying @xmath86 , then @xmath100 and thus @xmath34 is in @xmath85 and , in particular , is bounded on @xmath1 . by the beltrami equation ,
@xmath34 vanishes on @xmath101 and therefore @xmath34 is in @xmath102 . on the other hand , @xmath103 . by the main lemma
, @xmath104 is in @xmath102 ( see below ) , and so @xmath11 is a lipschitz function on the plane . showing that @xmath11 is bilipschitz still requires an argument .
indeed , we have shown up to now that @xmath11 is of class @xmath105 and thus its jacobian is non - zero at each point of @xmath1 ( ( * ? ? ?
* theorem 7.1 , p. 233 ) ) . on the other hand
, @xmath11 is conformal on @xmath101 and thus the jacobian is also non - zero there .
however we can not infer immediately that the jacobian is bounded below away from zero either on @xmath1 or on @xmath101 .
this is proved in section [ sec6 ] and hence @xmath106 is bilipschitz .
it remains to prove that @xmath107 is invertible on @xmath85 for each @xmath99 with @xmath86 . for @xmath3 in @xmath75
set @xmath108 where , as we said in the introduction , @xmath109 stands for @xmath110 following @xcite we define @xmath111 so that we have @xmath112 where @xmath113 can be easily seen to be a finite sum of operators that contain as a factor the commutator @xmath114 .
lemma 3 in section 4 asserts that @xmath115 is compact on @xmath85 for each @xmath99 less than @xmath116 , so that @xmath117 is also compact on @xmath85 .
one would like to have now that the operator norm of @xmath118 on @xmath85 is small if @xmath119 is large .
would this be so , then @xmath120 would be a fredholm operator on @xmath85 .
but it looks like a difficult task to obtain estimates for the operator norm of @xmath83 better than the obvious exponential upper bound @xmath121 .
we overcome this difficulty by finding an expression of the form @xmath122 where @xmath123 is compact on @xmath85 .
this is done in theorem 1 in section 5 . incidentally , in turns out that @xmath124 when @xmath1 is a disc , so that in this case @xmath125 is exactly @xmath126 for each @xmath119 .
then can be rewritten as @xmath127 where @xmath128 is compact on @xmath85 .
the kernel of @xmath129 may be computed explicitly , for instance via a fourier transform argument @xcite , and one obtains @xmath130 thus the caldern - zygmund constant of @xmath131 , namely , @xmath132 is less than @xmath133 , where @xmath134 is a positive constant . hence , by the main lemma @xmath135 which tells us that the operator norm of @xmath136 as an operator on @xmath85 is small for large @xmath119 .
therefore @xmath120 is a fredholm operator on @xmath85 .
clearly @xmath137 , @xmath138 is a continuous path from the identity to @xmath120 . by the index theory of fredholm operators on banach spaces ( e.g. @xcite ) ,
the index is a continuous function of the operator .
hence @xmath120 has index @xmath139 . on
the other hand @xmath120 is injective , because if @xmath140 , then @xmath141 , which is possible only if @xmath142 .
thus @xmath120 is invertible on @xmath85 . as we mentioned before , the proof of the theorem simplifies if @xmath143 is assumed to be less than a small number @xmath144 .
in this case one can invert @xmath145 by a neumann series and get @xmath146 by the main lemma @xmath147 is bounded on @xmath75 .
denote by @xmath148 its operator norm and assume that @xmath149 .
then @xmath150 but is also part of the main lemma that @xmath151 hence , if we also assume that @xmath152 , we have @xmath153 thus @xmath154 and so @xmath11 is a lipschitz function . that @xmath11 is bilipschitz follows from @xmath155
in this section we move to @xmath82 .
we say that a bounded domain @xmath156 has a boundary of class @xmath2 if @xmath157 is a @xmath158 hyper - surface whose unit normal vector satisfies a lipschitz condition of order @xmath116 as a function on the surface . to state an alternative condition , for @xmath159 we use the notation @xmath160 , where @xmath161 . then @xmath1 has a boundary of class @xmath2 if for each point
@xmath162 one may find a ball @xmath163 and a function @xmath164 , of class @xmath2 , such that , after a rotation if necessary , @xmath165 is the part of @xmath163 lying below the graph of @xmath166 .
thus we get @xmath167 a smooth ( of class @xmath158 ) homogeneous caldern - zygmund operator is a principal value convolution operator of type @xmath168 where @xmath169 @xmath170 being a homogeneous function of degree @xmath139 , continuously differentiable on @xmath171 and with zero integral on the unit sphere .
the maximal singular integral associated to @xmath73 is @xmath172 where @xmath173 the caldern - zygmund constant of the kernel of @xmath73 is defined as @xmath174 the operator @xmath73 is said to be even if the kernel is even , namely , if @xmath175 for all @xmath176 we are now ready to state our main lemma .
the definition of the norm in @xmath75 is as in . as we explained in the previous section ,
we need the precise form of the constant in the inequality below .
let @xmath1 be a bounded domain with boundary of class @xmath2 , @xmath20 , and let @xmath73 be an even smooth homogeneous caldern - zygmund operator .
then @xmath73 maps @xmath75 into @xmath75 , and @xmath73 also maps @xmath75 into @xmath177 .
in fact , one has the inequalities @xmath178 and @xmath179 where @xmath134 is a constant depending only on @xmath119 , @xmath116 and @xmath1 . we choose a positive @xmath180 small enough so that a series of properties that will be needed along the proof are satisfied .
the first one is that for each @xmath181 , which we can assume to be @xmath182 , we have . after a rotation
we may assume that the tangent hyperplane to @xmath183 at @xmath139 is @xmath184 .
we take @xmath185 so small that latexmath:[\[\label{eq7 } for some positive constant @xmath134 depending only on @xmath1 .
we claim that @xmath187 the proof of is a technical variation of the proof of lemma 5 in @xcite .
we have @xmath188 & = i_\delta + ii \ , .
\end{split}\ ] ] clearly , @xmath189 to deal with the term @xmath190 we write @xmath191 & \quad+ f(x)\int_{\delta < |y - x| < r_0 } \chi_\omega(y)\ , k(x - y)\,dy \\*[5pt ] & = iii_\delta+ f(x)\,iv_\delta \ , , \end{split}\ ] ] and we remark that @xmath192 can easily be estimated as follows @xmath193 taking care of @xmath194 is not so easy .
assume first that @xmath195 is in @xmath157 .
without loss of generality we may also assume that the tangent hyperplane to @xmath157 at @xmath139 is @xmath196 ( see figure 1 ) .
+ figure 1 let @xmath197 be the half space @xmath198 . take spherical coordinates @xmath199 with @xmath200 and @xmath201 .
then @xmath202 where @xmath203 and @xmath204 is the surface measure on the unit sphere @xmath205 .
since @xmath206 is even , @xmath207 thus @xmath208 and so @xmath209 by , we obtain @xmath210 which yields , by , @xmath211 take now @xmath212 .
denote by @xmath213 the distance from @xmath214 to @xmath157 and let @xmath215 be a point in @xmath157 where such distance is attained .
set @xmath216 and @xmath217 we compare @xmath194 to the expression we get replacing @xmath214 by @xmath215 and @xmath50 by @xmath218 in the definition of @xmath194 . for @xmath219 we have , by the standard cancellation property of the kernel , @xmath220 and then @xmath221 & = \left|\int_{a } k(x - y)\,dy - \int_{a_0 } k(x_0-y)\,dy \right| \\*[7pt ] & \leq \int_{a\cap a_0 } \!\chi_\omega(y)\ , k(x - y)\,dy \right| + \left| \int_{a_0 \setminus a } \chi_\omega(y)\ , k(x_0-y)\,dy \right| \\*[7pt ] & = j_1+j_2+j_3\ , .
\end{split}\ ] ] if @xmath222 , then @xmath223 hence @xmath224 to estimate @xmath225 observe that @xmath226 assume for the moment that @xmath227 . now , it is obvious that if @xmath228 , then @xmath229 , and so @xmath230 a similar argument does the job for @xmath231 .
if @xmath232 , then the estimate of @xmath233 is straightforward . indeed , by the cancellation of the kernel we may assume that @xmath234 and so @xmath235 this completes the proof of .
our next task is to estimate the semi - norm @xmath236 on @xmath1 . for this
we need a lemma , which should be viewed as a manifestation of the extra cancellation enjoyed by even kernels .
notice that no smoothness assumptions on the kernel are required .
the lemma is known for the beurling transform ( @xcite ) .
[ cancel ] let @xmath237 , where @xmath238 is an even homogeneous function of degree @xmath139 , integrable on the unit sphere and with vanishing integral there .
let @xmath73 be the associated caldern - zygmund operator defined by .
then @xmath239 assume , without loss of generality that @xmath69 is the unit ball .
fix a point @xmath214 in @xmath69 .
then @xmath240 expressing the latest integral above in polar coordinates @xmath241 centered at @xmath214 , we get @xmath242 the lower value @xmath243 is determined as shown in figure 2 .
+ figure 2 since @xmath238 has zero integral on the unit sphere , @xmath244 set @xmath245 be the half of the unit sphere above the hyperplane @xmath246 . since @xmath238 is even ,
@xmath247 now , the points @xmath139,@xmath214 , @xmath248 and @xmath249 lie in a plane that intersects the unit sphere in a circumference .
it is clear from figure 2 that the product @xmath250 is the power of @xmath214 with respect to that circumference and thus it does not depend on @xmath251 ( in fact it is exactly @xmath252 . since the integral of @xmath238 on each semi - sphere is zero
the proof is complete .
we are now ready to estimate the semi - norm @xmath236 .
we deal first with the case @xmath253 .
the general case follows from this by the @xmath254-theorem for lipschitz spaces on spaces of homogeneous type @xcite ( see also @xcite and @xcite for the non - doubling case ) .
the conditions on the kernel required in @xcite ( and in @xcite , @xcite ) are implied by the fact that @xmath255 , which we proved before . however , in our particular setting the reduction to @xmath253 is elementary and will be discussed afterwards for the sake of completeness . we want to prove that @xmath256 fix @xmath214 and @xmath257 in @xmath1 . changing notation if necessary
, we can assume that @xmath258 .
we may also assume , without loss of generality , that @xmath259 .
otherwise we have @xmath260 where @xmath261 and @xmath134 depends only on @xmath1 .
notice that @xmath262 , because @xmath263 and the kernel of @xmath73 is continuously differentiable off the origin . indeed , for some constant depending only on @xmath264 and @xmath1 , we have @xmath265 we may also assume , without loss of generality , that @xmath266 , because , otherwise , @xmath267 having settled these preliminaries we proceed to the core of the proof of .
we may assume that the point of @xmath157 nearest to @xmath214 is the origin .
let @xmath69 be the ball with center @xmath268 and radius @xmath269 , so that @xmath270 is tangent to @xmath157 at @xmath139 .
let @xmath128 stand for the set @xmath271 .
the central idea in the proof of the main lemma is to use the extra cancellation of even caldern - zygmund operators via lemma 1 to write latexmath:[\[|t(\chi_\omega)(x)-t(\chi_\omega)(y ) | = the obvious advantage is that @xmath128 is a region which is tangential " to @xmath157 at @xmath139 , and hence small . by we
may take @xmath269 so small that for some constant @xmath134 depending only on @xmath1 @xmath273 where @xmath274 stands for the inward unit normal vector to @xmath157 at @xmath139 and @xmath275 denotes the scalar product in @xmath82 .
thus , if @xmath269 is small enough , @xmath276 we distinguish two cases according to whether the position of @xmath214 and @xmath257 relative to @xmath157 is non - tangential or tangential . to make this precise we introduce the cone @xmath277 with vertex @xmath139 and amplitude @xmath278 , namely , @xmath279 clearly ,
if @xmath269 is chosen small enough , then the part of the cone near @xmath139 is contained in @xmath1 and in @xmath69 .
more precisely , @xmath280 see figure 3 .
+ figure 3 _ case 1 : _
@xmath214 and @xmath257 are in non - tangential position , that is , @xmath214 and @xmath257 belong to @xmath277 .
we have @xmath281 & \quad + \left |\int_{s \cap b(x,{r_0 } ) } k(x - z)\,dz - \int_{s \cap b(y,{r_0 } ) } k(y - z)\,dz \right| \\*[7pt ] & = i+ii\ , .
\end{split}\ ] ] split @xmath282 into three terms as follows @xmath283 & \quad + \left
|\int_{s \cap b^c(x,{r_0})\cap b^c(y,{r_0 } ) } \left(k(x - z)-k(y - z)\right)\,dz \right| \\*[7pt ] & = i_1+i_2 + i_3 \ , . \end{split}\ ] ] the terms @xmath284 and @xmath285 are estimated in the same way .
for instance , for @xmath284 , we get @xmath286 & \leq \frac{c}{{r_0}^n}\,|x - y|\,{r_0}^{n-1 } = \frac{c}{{r_0}}\,|x - y|\ , , \end{split}\ ] ] where in the latest inequality we used that @xmath287 the term @xmath288 is controlled by a gradient estimate , namely , @xmath289 the more difficult term @xmath290 is not greater than @xmath291 & \qquad\quad + \left| \int_{s \cap b(y,2\,|x - y ) } k(y - z)\,dz \right|\\*[9pt ] & \qquad\quad + \left|\int_{s \cap b(x,{r_0})\cap b^c(x,2\,|x - y| ) } k(x - z)\,dz - \int_{s \cap b(y,{r_0})\cap b^c(y,2\,|x - y| ) } k(y - z)\,dz \right| \\*[9pt ] & \qquad = ii_1 + ii_2+iii\ , .
\end{split}\ ] ] estimating the three terms above requires a simple lemma .
[ lemma 2 ] if @xmath269 is small enough , then one has @xmath292 for @xmath293 . according to the definition of the cone @xmath277 and by @xmath294 & \leq ( 1+\sqrt{2})\,|z - w|+ \sqrt{2}\,|(z,\overrightarrow{n})|\\*[5pt ] & \le ( 1+\sqrt{2})\,|z - w|+ \sqrt{2}\,c\ , |z|^{1+{\varepsilon}}\\*[5pt ] & \le ( 1+\sqrt{2})\,|z - w|+ \sqrt{2}\,c\,{r_0}^{\varepsilon}\ , |z|\ , .
\end{split}\ ] ] if @xmath269 satisfies @xmath295 , then @xmath296 , which proves the lemma . to estimate the term @xmath297
we apply lemma 2 to @xmath298 to obtain @xmath299 changing to polar coordinates we get @xmath300 by @xmath301 and hence @xmath302 one estimates @xmath303 likewise , so we turn our attention to @xmath304 .
the method is similar to what we have done before with other terms : in the intersection of the domains of integration of the two integrals in @xmath304 we apply a gradient estimate and in the complement , which we split in four terms , we resort to the smallness of the resulting domain of integration .
performing the plan just sketched we get @xmath305 & \quad + \left|\int_{s \cap b(x,{r_0 } ) \cap b^c(x,2\,|x - y|)\cap b^c(y,{r_0 } ) } k(x - z)\,dz \right|\\*[5pt ] & \quad + \left|\int_{s \cap b(x,{r_0 } ) \cap b^c(x,2\,|x - y| ) \cap b(y,2\,|x - y| ) } k(x - z)\,dz \right|\\*[5pt ] & \quad + \left|\int_{s \cap b(y,{r_0 } ) \cap b^c(y,2\,|x - y|)\cap b^c(x,{r_0 } ) } k(x - z)\,dz \right|\\*[5pt ] & \quad + \left|\int_{s \cap b(y,{r_0 } ) \cap b^c(y,2\,|x - y| ) \cap b(x,2\,|x - y| ) } k(x - z)\,dz \right|\\*[5pt ] & = iii_1 + iii_2 + iii_3 + iii_4 + iii_5 \ , .
\end{split}\ ] ] by a gradient estimate @xmath306 is not greater than @xmath307 where has been used in the first inequality .
the terms @xmath308 and @xmath309 are estimated in the same way .
for instance , for @xmath308 we have @xmath310 & \le \frac{c}{{r_0}^n}\ , |b(x,{r_0 } ) \cap b^c(y,{r_0})|\\*[7pt ] & \le \frac{c}{{r_0}}\ , |x - y|\ , .
\end{split}\ ] ] the terms @xmath311 nd @xmath312 are also estimated in the same way .
for instance , for @xmath311 we have @xmath313 since @xmath314 , for @xmath315 , we get by lemma 2 @xmath316 and so , making use of , @xmath317 _ case 2 : _ @xmath214 and @xmath257 are in tangential position , that is , @xmath318 we intend to perform a reduction to the non - tangential case . with this in mind take the point @xmath319 in @xmath157 nearest to @xmath257 and let @xmath320 be the inner unit normal vector to @xmath157 at the point @xmath319 .
consider the ray @xmath321 .
see the figure 4 . + figure 4 the condition on @xmath322 for @xmath323 is @xmath324 we clearly have @xmath325 and @xmath326 a sufficient condition for is then @xmath327 if @xmath269 is small enough , then @xmath328 , and thus we obtain a simpler sufficient condition for namely , @xmath329 set @xmath330 , so that @xmath331 .
the reduction will be completed if we show that @xmath332 on one hand , and @xmath257 and @xmath333 on the other , are in non - tangential position .
clearly @xmath334 . since @xmath335 , @xmath336 , and
hence @xmath337 which yields @xmath338 therefore @xmath339 & \geq |y'| \geq \left(1-\frac{1}{\sqrt{2}}\right)\,|y|\\*[5pt ] & = c\ , |y - y_0|\ , , \end{split}\ ] ] which is .
this completes the proof that @xmath340 .
we now proceed to prove that for an arbitrary @xmath341 one has that @xmath342 .
recall that we already know that @xmath343(see ) . to estimate the semi - norm @xmath236 we start with the obvious decomposition of @xmath344 , namely , @xmath345 so that only the first term , which we denote by @xmath346 , is still a problem .
let @xmath41 stand for @xmath347 and set @xmath348 .
then @xmath349\,dy \\*[7pt ] & \quad + \
! \int_{b } \left [ ( f(y)-f(x_1 ) ) \ ,
k(x_1-y ) \!-\ ! ( f(y)\!-\!f(x_2 ) ) \ , k(x_2-y)\right]\,dy\\*[7pt ] & = i+ii\ , .
\end{split}\ ] ] set @xmath350
. clearly @xmath351 .
the term @xmath282 is easy to estimate as indicated below .
@xmath352 & \le c\,\|f\|_{\varepsilon}\ , \int_0^{3\,|x_1-x_2| } r^{-1+{\varepsilon}}\,dr \\*[7pt ] & = c\,\|f\|_{\varepsilon}\,|x_1-x_2|^{\varepsilon}\ , .
\end{split}\ ] ] for the term @xmath290 we have @xmath353 & \quad + ( f(x_2)-f(x_1))\ , \int_{b } k(x_1-y)\,dy \\*[7pt ] & = iii+iv\ , .
\end{split}\ ] ] on one hand , by a gradient estimate we get @xmath354 & = \|f\|_{\varepsilon}\,|x_1-x_2|\ , \int_{2\,|x_1-x_2|}^\infty r^{-2+{\varepsilon}}\,dr \\*[7pt ] & = c\,\|f\|_{\varepsilon}\,|x_1-x_2|^{\varepsilon}\ , , \end{split}\ ] ] and on the other hand , we clearly have by @xmath355 finally , one can check without pain that the arguments above may be adapted to yield the boundedness of @xmath73 as a map from @xmath75 into @xmath177 .
in this section we consider the commutator between the smooth homogeneous even caldern - zygmund operator @xmath73 ( see ) and the multiplication operator by a function @xmath356 @xmath357(f)(x)= \int_\omega \left ( a(x)-a(y)\right)\,k(x - y)\,f(y)\,dy\,,\quad x \in \omega\,,\ ] ] where @xmath358 is the kernel of @xmath73 and @xmath359 . as in the previous section
, @xmath1 is a bounded domain in @xmath82 with smooth boundary of class @xmath2 , @xmath360 [ lemma 3 ] for @xmath361 and @xmath362 we have the estimate @xmath363 ( f)\|_\alpha \le c\ , \sigma_\alpha(a)\ , \|f\|_\beta\,,\quad f \in \operatorname{lip}(\beta,\omega)\,,\ ] ] where @xmath134 is a constant depending only on @xmath119 , @xmath1 , @xmath116 , @xmath56 and @xmath364 .
recall that for @xmath361 , @xmath365 where @xmath366 is the supremum norm of @xmath367 on @xmath1 and @xmath368 a consequence of the preceding lemma is that if @xmath369 and @xmath370 then the commutator @xmath371 $ ] is compact as an operator from @xmath372 into itself .
this follows from the fact that each ball of @xmath373 is relatively compact in @xmath372 ( ( * ? ? ?
* corollary 3.3 , p. 154 ) ) .
the lemma is applied to the beurling transform and the function @xmath374 .
then @xmath375 and @xmath376 , where @xmath99 is any number with @xmath86 .
we first estimate @xmath377 ( f)\|_\infty$ ] . for each @xmath378 , @xmath379 & \le c\,\sigma_\alpha(a)\ , \|f\|_\infty \
, \int_0^{d } r^{-1+\alpha}\,dr \\*[7pt ] & = c\,\sigma_\alpha(a)\ , \|f\|_\infty \ , d^\alpha \ , , \end{split}\ ] ] where @xmath380 is the diameter of @xmath1 .
we turn now to the more difficult task of estimating @xmath381(f))$ ] .
fix @xmath382 and @xmath383 in @xmath1 .
then @xmath384 & = i + ii\ , , \end{split}\ ] ] and clearly , by the main lemma , @xmath385 to estimate @xmath290 we introduce the sets @xmath386 and @xmath387 notice that @xmath388 and @xmath389 ) denote the absolute value of the integral in @xmath290 with domain of integration restricted to @xmath390 ( respectively to @xmath391 ) . by a gradient estimate @xmath392 & \le c\,|x_1-x_2|\,\sigma_\alpha(a)\ , \|f\|_\infty \int_{a } c\,|x_1-x_2|\,\sigma_\alpha(a)\ , \|f\|_\infty \int_{|x_1-x_2|}^\infty r^{-2+\alpha}\,dr\\*[7pt ] & \le c\,\sigma_\alpha(a)\ , \|f\|_\infty \,|x_1-x_2|^\alpha\ , .
\end{split}\ ] ] for the term @xmath393 we have @xmath394 & \quad + \left|\int_b ( a(x_2)-a(y))\,(k(x_2-y)\,f(y)\,dy \right| \\*[7pt ] & = iii+iv\ , , \end{split}\ ] ] and @xmath395 can be estimated directly as follows @xmath396 & \le \sigma_\alpha(a)\ , \|f\|_\infty \
, \int_0^{3\,|x_1-x_2| } r^{-1+\alpha}\,dr \\*[7pt ] & = c\ , \sigma_\alpha(a)\ , \|f\|_\infty \,|x_1-x_2|^\alpha\ , .
\end{split}\ ] ] the term @xmath304 needs an additional manoeuvre , which consists is bringing back @xmath397 : @xmath398 & \quad + |a(x_2)-a(x_1)|\left|\int_b k(x_1-y)f(y)\,dy)\right| \\*[7pt ] & = iv ' + v\ , , \end{split}\ ] ] and @xmath399 can be treated as @xmath395 . now @xmath400 and
thus , by , @xmath401 therefore @xmath402
recall that if @xmath69 is the beurling transform then @xmath403 the main goal of this section is to prove the following result .
[ th1 ] let @xmath404 be a bounded domain with boundary of class @xmath2 , @xmath20 .
then , for each positive integer @xmath119 , we have @xmath405 where @xmath123 is a compact operator from @xmath85 into itself , @xmath86 . for @xmath406 , we obtain , proceeding by induction , @xmath407 & = b\left(b^{n-1}(f)\,\chi_\omega+k_{n-1}(f)\right)\,\chi_\omega \\*[5pt ] & = b\left(b^{n-1}(f)- b^{n-1}(f)\,\chi_{\omega^c}+k_{n-1}(f)\right)\,\chi_\omega\\*[5pt ] & = b^n(f)\,\chi_\omega - b(b^{n-1}(f)\,\chi_{\omega^c})\,\chi_\omega + b(k_{n-1}(f))\,\chi_\omega\ , .
\end{split}\ ] ] it is then enough to prove that , for @xmath408 , the operator @xmath409 is compact from @xmath85 into itself .
let @xmath410 stand for area measure in the plane and take a function @xmath411 .
then , for @xmath412 , @xmath413 & = - \frac{1}{\pi } \int_{\omega^c } \frac{1}{(z - w)^2 } \frac{(-1)^n n}{\pi}\,\int_\omega \frac{(\overline{w-\zeta})^{n-1}}{(w-\zeta)^{n+1}}\,f(\zeta)\,da(\zeta ) \,da(w ) \\*[7pt ] & = c_n\ , \int_\omega k(z,\zeta)\,f(\zeta)\,da(\zeta)\ , , \end{split}\ ] ] where @xmath414 and @xmath415 notice that if @xmath1 is a disc , say the unit disc , then @xmath416 to see this readily , apply green - stokes theorem to the complement of the unit disc to obtain @xmath417 expand @xmath418 by newton s formula and then use @xmath419 thus @xmath420 is a finite sum of integrals over the unit cercle of rational functions with all poles in the open unit disc .
hence each of these integrals is zero .
we claim that if @xmath1 is not a disc , then the operator @xmath421 which may be non - zero , is a smoothing operator . by this
we mean that @xmath422 where @xmath134 depends only on @xmath56 , @xmath116 and @xmath1 .
of course completes the proof of theorem 1 , because then @xmath423 maps the unit ball of @xmath85 into a ball of @xmath373 , for @xmath424 , which is relatively compact in @xmath85 provided @xmath425 .
our next goal is to show that is a consequence of the properties of the kernel @xmath420 described in the following lemma .
[ lemma4 ] the kernel @xmath420 satisfies the following @xmath426 & \text{(ii)}\hskip1cm|k(z_1,\zeta)-k(z_2,\zeta)| \le c \ ,
\frac{|z_1-z_2|^\varepsilon}{|\zeta - z_1|^2}\,,\quad z_1,\,z_2 \in \omega\,,\quad |\zeta - z_1| \geq 2\,|z_1-z_2|\,.\end{aligned}\ ] ] before discussing the proof of lemma 6 we show how it yields .
we first prove that @xmath427 is bounded on @xmath1 . denoting by @xmath380 the diameter of @xmath1
, we obtain , by lemma 6 ( i ) , @xmath428 & \le c\ , \|f\|_\infty\ , \int_0^d r^{-1+{\varepsilon}}\,dr \\*[7pt ] & = c\,d^{\varepsilon}\ , \|f\|_\infty \ , .
\end{split}\ ] ] next we claim that @xmath429 clearly follows from . to prove take @xmath430 .
define @xmath431 and @xmath432 .
therefore @xmath433 applying lemma 6 ( i ) , the terms @xmath282 and @xmath290 can be estimated by @xmath434 applying lemma 6 ( ii ) , the term @xmath304 can be estimated by @xmath435 & \le c\,\|f\|_\infty\ , |z_1-z_2|^\varepsilon \ , \int_{2\,|z_1-z_2|}^d \frac{dr}{r } \\*[7pt ] & = c\ , \|f\|_\infty\ , |z_1-z_2|^\varepsilon \,\log{\frac{d}{2\,|z_1-z_2|}}\ , , \end{split}\ ] ] which completes the proof of theorem 1 . for each @xmath436 consider the cauchy integral of the function @xmath437 on @xmath438 , that is
, @xmath439 for @xmath440 let @xmath441 be the non - tangential limit of @xmath442 from @xmath1 , that is , the limit of @xmath443 as @xmath444 tends to @xmath445 non - tangentially
. similarly , denote by @xmath446 the non - tangential limit of @xmath442 from @xmath447 .
these limits exist a.e .
on @xmath438 with respect to arc - length and one has the plemelj formula ( e.g. @xcite ) @xmath448 indeed , it can be shown that @xmath442 is of class @xmath16 in @xmath1 and in @xmath449 , so that the above limits exist everywhere on @xmath438 and without the non - tangential approach restriction .
we do not need , however , such fact . applying the green - stokes theorem to the form @xmath450 and the domain @xmath451
, we get @xmath452 which by the residue theorem is @xmath453 a straightforward computation of the residues yields @xmath454 \left.+\sum^n_{\ell=0}(-1)^{n-\ell}\frac{(n-\ell+1)}{\ell ! } \frac{d^\ell}{d\zeta^\ell}h_{\zeta}(\zeta)\frac{1}{(\zeta - z)^{n+2-\ell}}. \right\}.\end{gathered}\ ] ] in the expression above for the kernel @xmath420 one may divine the presence of non - obvious cancellation properties ( consider the case @xmath455 ) .
the strategy to unravel them is to bring into the scene the function @xmath456 and express @xmath420 in terms of @xmath34 and its derivatives .
taylor s expansions of @xmath34 and its derivatives will then help in understanding cancellations .
the derivatives of @xmath34 are given by @xmath457 on the other hand , by the binomial formula , @xmath458 & = \sum^n_{\ell=0}\binom{n}{\ell } ( \overline{z-\zeta})^{\ell } \int_{\partial\omega } \frac{(\overline{t - z})^{n-\ell}}{t - z}\,dt\\*[5pt ] & = \sum^n_{\ell=0}\frac{(-1)^\ell}{\ell ! } \frac{\partial^{\ell}h}{\partial \overline{z}^\ell}(z ) ( \overline{z-\zeta})^{\ell}. \end{split}\ ] ] differentiating the preceding identity with respect to @xmath459 @xmath460 therefore @xmath461 & \quad-(n+1)\sum^n_{\ell=0}\frac{(-1)^{\ell}}{\ell ! } \frac{\partial^{\ell}h}{\partial \overline{z}^{\ell}}(z ) ( \overline{z-\zeta})^{\ell}\\*[5pt ] & \quad+ \sum^n_{\ell=0}\frac{(n+1-\ell)}{\ell ! } \frac{\partial^{\ell}h(\zeta)}{\partial \zeta^\ell}(z-\zeta)^{\ell}. \end{split}\ ] ] in each of the terms of the last sum it will be convenient to write a taylor expansion of @xmath462 up to order @xmath463 around the point @xmath459 . doing so we obtain @xmath464 & \quad + ( n+1)\sum^n_{\ell=0}\frac{1}{\ell ! } \frac{\partial^{\ell}h}{\partial \overline{z}^{\ell } } ( z ) ( \overline{\zeta - z})^{\ell}\\*[5pt ] & \quad-\sum^n_{\ell=0 } \frac{n+1-\ell}{\ell!}\\*[5pt ] & \qquad\times \sum^{n-\ell}_{j=0}\sum^{j}_{k=0 } \frac{(-1)^{\ell}}{k!\,(j - k ) ! } \frac{\partial^{\ell+k}}{\partial z^{\ell+k } } \frac{\partial^{j - k}}{\partial \overline{z}^{j - k}}h(z)(\zeta - z)^{k+\ell}(\overline{\zeta - z})^{j - k}\\*[5pt ] & \quad+ r(z,\zeta)\equiv s(z,\zeta)+r(z,\zeta ) .
\end{split}\ ] ] a cumbersome but easy computation shows now that @xmath465 the most direct way to ascertain this is to check that the coefficient of @xmath466 in the monomial @xmath467 vanishes for all non - negative exponents @xmath468 and @xmath469 for this we distinguish four cases .
_ case 1 : _ assume that @xmath470 only in the third sum may appear terms of this type and they must cancel out by themselves .
this can be shown using the identities @xmath471 _ case 2 : _
take @xmath472 and @xmath473 two terms appear in the third sum and one in the third , and they cancel .
_ case 3 : _ take @xmath472 and @xmath474 there is only one term of this type , which corresponds to letting @xmath475 in the first sum . to show that this term vanishes we resort to for @xmath476 and @xmath477 and then we apply cauchy s theorem .
_ case 4 : _ take @xmath478 and @xmath479 one term in the second sum cancels with a term in the third sum .
we turn now to the analysis of the kernel @xmath480 since @xmath466 vanishes identically we get @xmath481 where @xmath482 is the remainder of the taylor expansion of @xmath462 up to order @xmath463 around the point @xmath459 . a key fact in the present proof
is that the remainder @xmath482 is @xmath483 , because the @xmath119-th order derivatives of @xmath484 are in @xmath485 . to show this we resort to to get @xmath486 if @xmath487 , then the above expression is @xmath488 if @xmath489 , then we obtain , by green - stokes and for some constant @xmath490 , @xmath491 which is in @xmath492 owing to the main lemma . here
@xmath493 is the @xmath494-th iteration of the beurling transform .
part ( i ) of the lemma is a straightforward consequence of and the size estimate on the remainder @xmath482 we have just proved .
we are left with part ( ii ) .
take points @xmath495 and @xmath496 in @xmath1 with @xmath497 from we obtain @xmath498 add and subtract @xmath499 in the numerator of the second fraction above to get @xmath500 where @xmath501 and @xmath502 controlling @xmath282 is easy via an obvious gradient estimate , which yields @xmath503 & = c\,\frac{|z_1-z_2|}{|z_1-\zeta|^{3-\varepsilon } } \\*[7pt ] & \le c\ , \frac{|z_1-z_2|^\varepsilon}{|z_1-\zeta|^2}\ , .
\end{split}\ ] ] to estimate the term @xmath290 we need a sublemma .
[ sublema ] we have the identity @xmath504 since @xmath505 is in @xmath4 for each non - negative number @xmath506 , the sublemma immediately provides the right control on the term @xmath507 namely , @xmath508 and this completes the proof of lemma 6 .
the most convenient way of proving the sublemma is to place ourselves in a real variables context . given a smooth function @xmath3 on @xmath509 let @xmath510 be its taylor polynomial of degree @xmath506 around the point @xmath511 then , clearly , @xmath512 and so the sublemma is an easy consequence of the fact that each @xmath119-th order derivative of @xmath34 is a constant times @xmath513 , for an appropriate exponent @xmath506 , and the following elementary calculus lemma .
[ calculus ] if @xmath3 is a @xmath506 times continuously differentiable function on @xmath509 , then @xmath514 let @xmath515 stand for the polynomial @xmath516 so that @xmath517 a straightforward computation yields @xmath518 which completes the proof of lemma 7 .
in the precedings sections we have proved that @xmath106 is a lipschitz function on @xmath519 . moreover
@xmath520 and so , by the main lemma , @xmath521 since @xmath106 is holomorphic on @xmath522 , @xmath523 extends continuosly to @xmath451 , but nothing excludes that this extension might vanish somewhere on @xmath438 .
the functions @xmath524 and @xmath525 also extend continuously from @xmath1 to @xmath526 , but again it could well happen that both vanish at some point of @xmath438 .
we will show now that this is not possible .
indeed , we claim that for some positive number @xmath527 we have @xmath528 this implies that the jacobian of @xmath106 is bounded from below by @xmath529 at @xmath459 almost all points of @xmath519 .
thus the invers mapping @xmath530 has gradient in @xmath531 and hence @xmath106 is bilipschitz . for @xmath532 denote by @xmath533 the limit of @xmath534 as @xmath535 tends to @xmath536 .
we claim that follows if we can show that @xmath537 indeed , this clearly implies @xmath538
. now denote by @xmath539 and @xmath540 , @xmath541 , the limits of @xmath542 and @xmath543 as @xmath544 tends to @xmath536 .
take a parametrization @xmath545 of @xmath438 of class @xmath158 , such that @xmath546 for all @xmath322 .
computing @xmath547 in two different ways , @xmath548 and so @xmath549 thus , by , @xmath550 , @xmath532 , which yields @xmath551 .
we turn now to the proof of .
assume that @xmath552 .
performing a rotation before applying @xmath106 we may assume that @xmath553 is a non - negative real number ( @xmath554 is the limit of @xmath555 as @xmath544 tends to @xmath139 ) .
performing a rotation after applying @xmath106 we may also assume that the tangent plane to @xmath438 at the origin is the real axis .
denote by @xmath556 and @xmath557 the upper and lower half planes , respectively .
consider the continuous piecewise linear mapping @xmath558 then , by ( * ? ? ?
* ( 5.6 ) , p. 83 ) , the beltrami coeficient @xmath559 of the mapping @xmath560 is @xmath561 since @xmath562 vanishes on @xmath563 , @xmath564 & \quad + \int_{h^-\cap l^{-1}(\omega^c)\cap b(0,r_{0 } ) } \dotsi\\*[7pt ] & \quad -\int_{h^-\cap l^{-1}(\omega)\cap b(0,r_{0 } ) } \dotsi\\*[7pt ] & = i+ii+iii , \end{split}\ ] ] where @xmath565 is the small number introduced in the proof of the main lemma ( see figure 1 ) . by , @xmath566 since @xmath567 on @xmath556 , making the change of variables @xmath568 gives @xmath569 where in the next to the last inequality we used .
for @xmath290 we begin by remarking that @xmath570 and making the change of variables @xmath571 , we get @xmath572 which can be shown to be finite as before ( in particular , using again ) . to take care of @xmath304
we make the same change of variables and we obtain @xmath573 & \le \frac{1}{1-\lambda } \int_{h^-\cap l^{-1}(\omega)\cap b(0,r_{0 } ) } \frac{|\mu ( l(w))-\mu(0)|}{|w|^2}\,da(w)\\*[7pt ] & \le \frac{1+\lambda}{(1-\lambda)^2 } \int_{h^-\cap \omega\cap b(0,r_{0 } ) } \frac{|\mu ( z ) -\mu(0)|}{|z|^2}\,da(z)\\*[7pt ] & \le c\int_{b(0,r_{0})}\frac{da(z)}{|z|^{2-\varepsilon}}<\infty . \end{split}\ ] ] therefore @xmath574 and so , by @xcite , @xmath575 is conformal at the origin , in the sense that the limit @xmath576 exists and @xmath577 . the part of the imaginary positive axis close to the origin is included in @xmath522 ( see figure 1 ) , and thus @xmath578 if @xmath579 is small .
hence @xmath580 & = \lim_{0<y\to 0}\frac{h(l^{-1}(iy))-h(0)}{iy}\\*[7pt ] & = \frac{h'(0)}{(1-\lambda)}. \end{split}\ ] ] this completes the proof of .
suppose , as in the statement of the theorem , that @xmath581 ... ,@xmath582 are bounded disjoint domains with boundary of class @xmath2 , for some @xmath116 with @xmath20 , and that @xmath583 where @xmath22 is of class @xmath46 and @xmath584 let @xmath585 be the quasiconformal mapping associated with @xmath586 assume that @xmath587 and set @xmath588 and @xmath589 by ( * ? ? ?
* ( 10 ) p. 9 ) , @xmath590 where @xmath591 and , for each @xmath592 the point @xmath459 is defined by @xmath593 in particular , @xmath27 is supported on @xmath594 recall from the previous section that @xmath595 is bilipschitz , of class @xmath596 on @xmath597 and @xmath598 and conformal on @xmath599 in particular , @xmath595 is holomorphic on @xmath600 and @xmath601 thus the bounded domains @xmath602 have boundaries of class @xmath603 on the other hand , the function @xmath27 satisfies a lipschitz condition of order @xmath604 for @xmath605 in each domain @xmath606 proceeding by induction we conclude now that @xmath585 is bilipschitz and of class @xmath607 in each domain @xmath608 one can prove now corollary 2 .
let @xmath31 be a bounded planar domain and let @xmath3 be a function in @xmath32 satisfying the beltrami equation @xmath609 where @xmath8 is as in the statement of the theorem . by stoilow s factorization theorem , @xmath610 where @xmath11 is the quasiconformal mapping associated with @xmath8 and @xmath34 is a holomorphic function on @xmath611 let @xmath49 stand for the set of points in @xmath31 whose distance to the boundary of @xmath31 is larger than @xmath612 thus @xmath34 has bounded derivatives of all orders on @xmath613 and consequently @xmath3 is as smooth as @xmath11 on @xmath614 hence @xmath3 is lipschitz on @xmath49 and of class @xmath607 in each open set @xmath615 if @xmath31 contains the closure of each @xmath616 then @xmath3 is of class @xmath607 on each @xmath617 recalling the relation between the beltrami equation and second order elliptic equations in divergence form , as explained in the introduction , corollary 2 follows immediately from the above argument .
as we remarked in the introduction , the conclusion of the theorem fails for domains with corners ; for instance , for a square .
however , the class of domains with boundary of class @xmath2 is not optimal for the theorem .
there is a heuristic argument that points out at a more general class of domains , which , at least in a first approximation , may be viewed as optimal .
first of all we recall that a central point in the proof of the theorem was the fact , which is part of the main lemma , that each power of the beurling transform sends the characteristic function of the domain into a bounded function .
let us concentrate on the beurling transform @xmath69 and find a simple condition on a bounded domain @xmath1 with rectifiable boundary so that @xmath618 is bounded .
our first remark is that @xmath618 can be written as the cauchy transform of a boundary measure . for this
we use , on one hand , the basic property of @xmath619 @xmath620 which holds for all compactly supported smooth functions @xmath166 and extends to a variety of situations by regularization .
on the other hand , we use the elementary identity @xmath621 which holds at least for bounded domains with rectifiable boundary . combining the above two identities
we get @xmath622 which yields @xmath623 now , @xmath624 @xmath625 being the unit tangent vector to @xmath438 at @xmath626 assume now that the arc - length measure on the boundary of @xmath1 satisfies the ahlfors condition @xmath627 for each @xmath628 and @xmath629 where @xmath630 stands for the open disc with center @xmath459 and radius @xmath631 then a simple estimate shows that the cauchy integral of a function @xmath3 on @xmath632 that is , @xmath633 is bounded provided @xmath3 satisfies a lipschitz condition of some positive order on @xmath634 thehefore , to get boundedness of @xmath618 one has to require that the square of the tangent unit vector satisfies a lipschitz condition of some positive order on @xmath634 this is weaker than requiring the lipschitz condition on the tangent unit vector itself , because it allows jumps of @xmath635 degrees on the argument of the tangent unit vector .
in other words , cusps are allowed .
we want now to define formally _ cuspidal domains of class _ @xmath76 given a planar domain @xmath1 we say that @xmath438 is @xmath2-smooth at a boundary point @xmath636 if there is a positive @xmath185 such that @xmath637 is , after possibly a rotation , the part of @xmath638 lying below the graph of a function of class @xmath76 we say that @xmath1 has an _ interior cusp of class _ @xmath2 at @xmath639 provided there is a positive @xmath185 and functions @xmath640 @xmath641 of class @xmath2 on the interval @xmath642 such that @xmath643 , and , after possibly a rotation , a point @xmath644 is in @xmath637 if and only if @xmath645 and @xmath646 we say that @xmath1 has an _ exterior cusp of class _ @xmath2 at @xmath647 provided @xmath648 has an interior cusp of class @xmath2 at @xmath649 a planar domain @xmath1 is a _ cuspidal domain of class _ @xmath2 if @xmath438 is @xmath2-smooth at all boundary points , except possibly at finitely many boundary points where @xmath1 has a cusp of class @xmath2 ( either interior or exterior ) . the simplest examples of non - smooth cuspidal domains are the drop like domain @xmath31 and the peach like domain @xmath423 shown in figure 5 below .
the reader may easily imagine more complicated cuspidal domains with lots of cusps of both types ( see figure 7 ) . + figure 5 with appropriate formulations the theorem and corollaries 1 and 2 hold true for cuspidal domains of class @xmath76 the right statements involve the notion of geodesic distance in the domain @xmath1 , which we discuss now . given two points @xmath459 and @xmath445 in @xmath1 their geodesic distance is defined by @xmath650 where the infimum is taken over all rectifiable curves @xmath651 in @xmath1 joining @xmath459 and @xmath652 here @xmath653 stands for the length of @xmath654 notice that if @xmath1 has only interior cusps , then the geodesic and the euclidean distances are comparable and that this is not the case in the proximity of an exterior cusp .
the lipschitz norm @xmath655 of order @xmath87 and the corresponding lipschitz spaces @xmath656 with respect to the distance @xmath380 are defined is the usual way , with the euclidean distance replaced by @xmath380 in and .
the theorem for cuspidal domains reads as follows .
let @xmath18 , @xmath19 , be a finite family of disjoint bounded cuspidal domains of class @xmath2 , @xmath20 , and let @xmath583 where @xmath22 is of class @xmath657 assume in addition that @xmath10 .
then the associated quasiconformal mapping @xmath11 is bilipschitz .
corollary 1 remains true without any change .
[ c1 ] if @xmath1 is a bounded cuspidal domain of class @xmath2 , @xmath20 , and @xmath26 , where @xmath27 is a complex number such that @xmath28 , then the associated quasiconformal mapping @xmath11 is bilipschitz .
[ c2 ] let @xmath24 , @xmath19 , be a finite family of disjoint bounded cuspidal domains of class @xmath2 , @xmath20 , and assume that all domains @xmath24 are contained in a bounded domain @xmath31 with boundary of class @xmath2 .
let @xmath44 a @xmath40 symmetric elliptic matrix with determinant @xmath6 and entries supported in @xmath658 and belonging to @xmath659 @xmath660 let @xmath42 be a solution of equation in @xmath48 let @xmath49 stand for the set of points in @xmath31 at distance greater than @xmath50 from the boundary of @xmath48 then @xmath661 for @xmath662 and @xmath47 in particular , @xmath663 and @xmath42 is a locally lipschitz function in @xmath48 we proceed now to sketch the proof of theorem. the modifications needed are minor and fortunately one can reduce without much pain the cuspidal case to the smooth case .
we start by discussing the proof in the one domain case ( @xmath664 ) .
the difficulty is that the main lemma does not hold for cuspidal domains with exterior cusps and the usual euclidean lipschitz spaces .
we present an example to make the difficulty clear .
let @xmath1 be the open disc centered at the origin of radius @xmath14 minus the union of the closed discs of radius @xmath6 centered at @xmath6 and @xmath665 thus @xmath1 has an exterior cusp at the origin .
we claim that @xmath618 does not satisfy a lipschitz condition on @xmath1 of any order @xmath87 such that @xmath666 the point is that we can explicitly calculate @xmath667 if @xmath668 stands for the open disc centered at @xmath536 of radius @xmath669 then a simple argument ( see , for example , @xcite ) shows that @xmath670 thus @xmath671 set @xmath672 and @xmath673 , where @xmath214 and @xmath257 are positive real numbers such that @xmath674 and @xmath675 are in @xmath1 .
then @xmath676 on the other hand , a simple computation yields @xmath677 as @xmath214 and @xmath257 tend to @xmath678 choosing @xmath679 we conclude that @xmath618 does not satisfy a lipschitz condition on @xmath1 of any order @xmath87 with @xmath680 the main lemma for cuspidal domains reads as follows .
let @xmath1 be a bounded cuspidal domain of class @xmath2 , @xmath20 , and let @xmath73 be an even smooth homogeneous caldern - zygmund operator .
then @xmath73 maps @xmath656 into itself , and @xmath73 also maps @xmath656 into @xmath681 . in fact , one has the inequalities @xmath682 and @xmath683 where @xmath134 is a constant depending only on @xmath116 and @xmath1 .
we first prove that @xmath684 for each @xmath685 take a point @xmath686 clearly we may assume that @xmath687 where @xmath636 is a cuspidal point and @xmath185 is as in the definition of cusp .
otherwise @xmath459 is far from all cusps and thus we may apply the arguments of the smooth case . since there are only finitely many cusps we may also assume that there is a positive number @xmath185 which works in the definition of cusp simultaneously for all cusps .
clearly @xmath688 and the second term is bounded by @xmath689 where @xmath380 stands for the diameter of @xmath690 we are therefore left with the first term .
assume for the moment that @xmath1 has an interior cusp at @xmath649 we connect the cercle of center @xmath636 and radius @xmath691 with the concentric cercle of radius @xmath185 to produce domains @xmath692 and @xmath693 with boundary of class @xmath694 as shown in figure 6 .
+ figure 6 notice that three other residual " domains @xmath695 @xmath696 have been formed .
the domains @xmath697 are not smooth but they are far from @xmath698 since we are assuming that @xmath1 has an interior cusp at @xmath699 the restriction of @xmath3 to @xmath700 satisfies an euclidean lipschitz condition of order @xmath701 by the well known extension theorem for lipschitz functions ( ( * ? ? ?
* chapter vi ) ) we may extend the restriction of @xmath3 to @xmath700 to a function @xmath702 such that @xmath703 hence @xmath704 and so @xmath705 the last three terms are controlled by @xmath706 where @xmath134 is a constant depending on @xmath707 because the domains of the functions to which @xmath708 is applied are far from @xmath709 by this gives the correct bound @xmath710 to estimate the first three terms we remark that @xmath711 @xmath693 and @xmath638 are domains with boundary of class @xmath712 thus the proof of the main lemma for smooth domains of section 3 yields , by , @xmath713 assume now that @xmath1 has an exterior cusp at @xmath649 then @xmath714 and so @xmath715 the proof of the estimate proceeds as before .
let us turn to prove the lipschitz condition on @xmath344 .
since @xmath344 is bounded , to estimate @xmath716 we may restrict our attention to the case in which @xmath459 and @xmath445 are very close to each other .
we may also assume that @xmath459 and @xmath445 are close to a cusp , say , @xmath717 from this point on the proof is very similar to what we did before , with @xmath708 replaced by @xmath718 the first step is to restrict our attention to @xmath719 which may be achieved by the identity @xmath720 assume first that @xmath1 has an interior cusp at @xmath636 and consider again the extension @xmath367 of @xmath721 satisfying .
we clearly have @xmath722 by the smooth version of the main lemma @xmath723 and @xmath724 satisfy an euclidean lipschitz condition of order @xmath87 on the complement of @xmath725 and @xmath726 respectively .
clearly , again by the smooth main lemma , @xmath727 satisfies an euclidean lipschitz condition of order @xmath87 on @xmath728 finally @xmath729 satisfy an euclidean lipschitz condition of order @xmath87 on @xmath730 because the domains @xmath697 are far from @xmath731 if both @xmath459 and @xmath445 belong to @xmath1 we then get an estimate of the form @xmath732 which completes the proof , because in the case at hand the euclidean distance between @xmath459 and @xmath445 is comparable to their geodesic distance . if @xmath459 and @xmath445 are in @xmath733 then we may assume that @xmath734 and @xmath735 otherwise @xmath459 and @xmath445 belong both to either @xmath692 or @xmath693 and thus we obtain the above euclidean lipschitz estimate of order @xmath87 .
choose then a point @xmath736 such that @xmath737 ( see figure 6 ) . since @xmath723 and @xmath724 satisfy an euclidean lipschitz condition of order @xmath87 on @xmath692 and @xmath693 respectively , and since @xmath344 is continuous on the complement of @xmath738 because @xmath3 is supported on @xmath738 we get @xmath739 if @xmath1 has an exterior cusp at @xmath699 we argue similarly , using the identity @xmath740 the details are left to the reader .
we continue now the proof of theorem in the one domain case .
if the domain has only interior cusps the argument we described to prove the theorem goes through without any change .
the reason is that , since the geodesic distance in @xmath1 is comparable to the euclidean distance sections 4 and 5 hold true , because of the cuspidal version of the main lemma . for section 6 one has to remark that @xmath741 is now only in @xmath742 but that this still implies that @xmath741 extends continuously from @xmath648 to @xmath634 the conformality of @xmath11 on @xmath451 is proved as in section 6 , after remarking that cusps do not create any problem because they are perfectly suited for an appeal to @xcite .
assume now that our domain has exterior cusps .
in figure 7 it is shown how to subdivide @xmath1 in finitely many subdomains @xmath743 , which are cuspidal domains of class @xmath16 _ with only interior cusps_. since @xmath744 @xmath745 is in @xmath746 we can then apply the factorization method of section 7 to get that @xmath747 and @xmath748 @xmath749 + figure 7 the reduction to the one domain case needs only one comment . using the notation of section 7
, we use the fact that @xmath595 is conformal on @xmath750 to ascertain that the image of each @xmath24 under @xmath595 is again a cuspidal domain of class @xmath751 we repeat for emphasis that the conformality of @xmath595 is proved at a cusp appealing , as in section 6 , to @xcite .
very likely the restriction on the determinant of the matrix @xmath41 in corollary 2 and corollary 2 is superfluous .
this would follow if lipschitz regularity results should hold for the general elliptic system @xmath752 where @xmath753 a. e. on @xmath754 it seems also rather clear that the right conclusion in corollary 2 ( and analogously in corollary 2 ) should be that the solution @xmath755 is of class @xmath756 @xmath757 in all dimensions ( and without any restriction on the determinant of @xmath41 ) .
evidence for this conjecture is provided by the fact that it is true in the plane whenever the lipschitz norm of the coefficients of the matrix @xmath41 are small enough .
see the argument for the beltrami equation at the end of section 2 .
we acknowledge some useful correspondence with l. escauriaza and d.faraco on that issue .
apparently it is not known what is the best exponent @xmath319 such that @xmath758 for @xmath759 where @xmath760 is a complex number such that @xmath761 and @xmath29 is a square .
this looks extremely surprising to the authors , who would very much appreciate knowing the precise regularity properties of @xmath585 in the scale of the local sobolev spaces @xmath762 .
this work was partially supported by the grants 2005sgr00774 ( generalitat de catalunya ) and mtm2007 - 60062 ( ministerio de educacin y ciencia ) .
the authors are indebted to stephen semmes for suggesting the proof of and to eero saksman for an encouraging conversation .
a. e. gatto , _ on the boundedness on inhomogeneous lipschitz spaces of fractional integrals , singular integrals and hypersingular integrals associated to non - doubling measures on metric spaces , _ preprint 2007 .
l joan mateu + departament de matemtiques + universitat autnoma de barcelona + 08193 bellaterra , barcelona , catalonia + _ e - mail : _ [email protected] + + joan orobitg + departament de matemtiques + universitat autnoma de barcelona + 08193 bellaterra , barcelona , catalonia + _ e - mail : _ [email protected] + + joan verdera + departament de matemtiques + universitat autnoma de barcelona + 08193 bellaterra , barcelona , catalonia + _ e - mail : _ [email protected]
|
in this paper we discuss a special class of beltrami coefficients whose associated quasiconformal mapping is bilipschitz .
a particular example are those of the form @xmath0 , where @xmath1 is a bounded domain with boundary of class @xmath2 and @xmath3 a function in @xmath4 satisfying @xmath5 .
an important point is that there is no restriction whatsoever on the @xmath4 norm of @xmath3 besides the requirement on beltrami coefficients that the supremum norm be less than @xmath6 .
the crucial fact in the proof is the extra cancellation enjoyed by even homogeneous caldern - zygmund kernels , namely that they have zero integral on half the unit ball .
this property is expressed in a particularly suggestive way and is shown to have far reaching consequences .
an application to a lipschitz regularity result for solutions of second order elliptic equations in divergence form in the plane is presented .
|
the world is filled with combinatorial problems .
these include important combinatorial optimization tasks such as planning , scheduling and rostering , combinatorics problems such as extremal graph theory , and countless puzzles and games .
solving combinatorial problems is hard , and all the methods we know to tackle them involve some kind of search .
various _ declarative paradigms _ have been developed to solve such problems . in such approaches , objects and attributes that are searched for
are represented by symbols , and constraints to be satisfied by those objects are represented as expressions over these symbols in a declarative language .
solvers then search for values for these symbols that satisfy the constraints .
this idea is found in the fields of mycommoncitationcp @xcite , mycommoncitationasp @xcite , sat , mixed integer programming ( mip ) , etc . in the terminology of logic ,
the declarative method amounts to expressing the desired properties of a problem class by sentences in a _
logical theory_. the data of a particular problem instance corresponds naturally to a _ partial interpretation ( or structure)_. the solving process is to apply _ model generation _ , or more specifically _ model expansion _
@xcite , the task of finding a structure that expands the input partial structure and satisfies the theory .
the resulting structure is a solution to the problem .
model generation / expansion , studied for example in the field of mycommoncitationkr @xcite , is thus analogous to the task of solving constraint satisfaction problems , studied in , and that of generating answer sets of logic programs , studied in .
the similarities between these areas go deeper and extend to the level of the used techniques .
state - of - the - art approaches often follow a two - phase solving methodology . in a first phase , the input theory , in the rich language at hand ,
is reduced into a fragment of the language that is supported by some search algorithm . in the second phase ,
the search algorithm is applied to the reduced theory to effectively search for models .
for example , model generation for the language minizinc is performed by reducing to the ground language flatzinc , for which search algorithms are available .
similarly , the language mycommoncitationfodot @xcite is reduced to its propositional fragment , and asp is reduced to propositional asp .
as the reduced theory is often in a ground fragment of the language , we refer to the resulting reduced theory as the _ grounding _ and to the first phase as the _ grounding _ phase ( where quantifiers are instantiated with elements in the domain ) . in other fields , grounding is also referred to as flattening , unrolling , splitting or propositionalization .
the solving methodology itself is generally referred to as _ ground - and - solve_. grounding becomes a bottleneck as users turn to applications with large domains and complex constraints .
indeed , it is easy to see that the grounding size of an fo formula is exponential in the nesting depth of quantifiers and in the arity of predicates and polynomial in the size of the universe of discourse . there is an increasing number of applications where the size of the grounded theory is so large that it does not fit in memory .
for example , discuss several applications where the ground - and - solve approach turns out to be inadequate .
in this paper , we present a novel approach to remedy this bottleneck , called _ lazy model expansion _
, where the grounding is generated lazily ( on - the - fly ) during search , instead of up - front .
the approach works by associating _ justifications _ to the non - ground parts of the theory .
a valid justification for a non - ground formula is a recipe to expand a partial structure into a more precise ( partial ) structure that satisfies the formula .
of course , it is crucial that the recipe is a lot more compact than the grounding of the formula . given a partial structure and a valid justification for each of the non - ground formulas , a ( total ) structure is obtained by extending the partial structure with the literals in the justifications of the non - ground formulas .
justifications are selected in such a way that this total structure is a model for the whole initial theory .
consequently , model generation can be limited to the grounded part of the theory ; if a model is found for that part , it can be extended to a model of the whole theory .
however , a new assignment during model generation can conflict with one of the justifications . in that case
, an alternative justification needs to be sought .
if none is found , the associated formula can be split in two parts , one part that is grounded and one part for which a valid justification is still available .
[ ex : sokoban ] consider the ` sokoban ` problem , a planning problem where a robot has to push blocks around on a 2-d grid to arrange them in a given goal configuration .
a constraint on the move action is that the target position @xmath0 of the moved block @xmath1 is currently ( at time @xmath2 ) empty , which can be expressed as @xmath3 as it is not known in advance how many time steps are needed , one ideally wants to assume a very large or even infinite number of steps . using ground - and - solve ,
this blows up the size of the grounding .
incremental grounding , iteratively extending the time domain until it is large enough to allow for a plan , has been developed to avoid the blow - up in the context of planning .
our approach is more general and does not depend on the presence of one domain that can be incrementally increased .
returning to the example , instead of grounding sentence ( [ ex : soko:1 ] ) , we associate with it a justification , a recipe to satisfy it .
`` make @xmath4 false for all @xmath5 , @xmath6 and @xmath7 '' is such a recipe .
when the search finds a model for the grounded part of the problem that is not in conflict with the recipe , the model can be extended with the literals in the recipe to obtain a model of the whole theory . however , if the search would decide to move block @xmath8 to position @xmath9 at time @xmath10 , a conflict is created with the recipe . to resolve it , the instance of sentence ( [ ex : soko:1 ] ) that is in conflict with the partial model of the search is split off and sentence ( [ ex : soko:1 ] ) is replaced by the equivalent sentences : @xmath11 sentence ( [ ex : soko:2 ] ) is grounded and passed to the search component which will use it to check that @xmath12 holds .
sentence ( [ ex : soko:3 ] ) is non - ground and can be satisfied by the recipe `` @xmath4 is false except for @xmath13 '' .
when the search makes more moves , more instances will be grounded , until the search finds a partial plan for the problem at hand .
then the literals in the recipe of the remaining non - ground formula making @xmath4 false for all instances of sentence ( [ ex : soko:1 ] ) that have not been grounded will complete the plan .
the main contributions of this paper are : * a theoretical framework for _ lazy model expansion_. by aiming at minimally instantiating quantified variables , it paves the way for a solution to the long - standing problem of handling quantifiers in search problems , encountered , e.g. , in the fields of @xcite and sat modulo theories @xcite .
the framework also generalizes existing approaches that are related to the grounding bottleneck such as incremental domain extension @xcite and lazy clause generation mycommoncitationlazyclausegeneration @xcite . * a complete algorithm for lazy model expansion for the logic , the extension of with inductive definitions .
this includes efficient algorithms to derive consistent sets of justifications and to maintain them throughout changes in a partial structure ( e.g. , during search ) . *
an implementation extending the knowledge - base system and experiments that illustrate the power and generality of lazy grounding .
lazy grounding is a new step in our ability to solve complex combinatorial problems . by avoiding the up - front grounding step of previous approaches
, lazy grounding can ground enough of the problem to solve it . while our method is developed for the logic ,
as will become clear , justifications are associated with rules , and these rules are very similar to the rules used by asp systems . hence ,
as discussed towards the end of the paper , our framework and algorithms can be applied also in the context of .
the paper is organized as follows . in section [ sec : preliminaries ] , the necessary background and notations are introduced .
formal definitions of lazy grounding with are presented in section [ sec : theory ] , followed by a presentation of the relevant algorithms and heuristics in sections [ sec : algorithms ] and [ sec : optimizations ] .
experimental evaluation is provided in section [ sec : experiments ] , followed by a discussion on related and future work and a conclusion . a preliminary version of the paper appeared as the work of and of .
in this section , we provide the necessary background on the logic , on the inference tasks model generation and model expansion for and on the ground - and - solve approach to model expansion .
[ [ section ] ] first , we define syntax and semantics of the logic _
_ mycommoncitationfodot @xcite , the extension of with inductive definitions .
we assume familiarity with . without loss of generality , we limit to the function - free fragment .
function symbols can always be eliminated using graph predicates @xcite .
a ( function - free ) vocabulary @xmath14consists of a set of predicate symbols .
propositional symbols are 0-ary predicate symbols ; these include the symbols @xmath15and @xmath16 , denoting _ true _ and _ false _ respectively .
predicate symbols are usually denoted by @xmath17 , @xmath18 , @xmath19 ; atoms by @xmath20 , literals ( atoms or their negation ) by @xmath21 ; variables by @xmath22 , @xmath23 ; and domain elements by @xmath24 . with @xmath25
we denote an ordered set of objects @xmath26 ; with @xmath27 a predicate @xmath17 of arity @xmath28 . the methods for model generating developed below require that a ( possibly infinite ) domain @xmath29 is given and fixed .
given a ( function - free ) vocabulary @xmath30 , a _ domain atom _ is an atom of the form @xmath31 with @xmath32 and @xmath33 , an @xmath28-tuple of domain elements .
likewise , we consider _
domain literals_. a structure @xmath34interpreting @xmath14consists of the domain @xmath29 and an @xmath28-ary relation @xmath35 for all predicate symbols @xmath36 .
alternatively , an @xmath28-ary relation can be viewed as a function @xmath37 .
the propositional symbols @xmath15and @xmath16are respectively interpreted as @xmath38and @xmath39 . model generation algorithms maintain _ partial _ structures and may ( temporarily ) find themselves in an _ inconsistent _ state , for example when a conflict arises . to represent such states , three - valued and four - valued structures @xmath40 are introduced ; they consist of the domain @xmath29 and , for each @xmath28-ary predicate @xmath17 in @xmath14 , of a three- or four - valued relation @xmath41 .
this is a function @xmath42 .
a structure is _ two - valued _ if the range of its relations is @xmath43 , _ partial _ or _ three - valued _ if the range is @xmath44 ) and four - valued in general .
thus , two - valued structures are also three - valued and four - valued .
when unqualified , the term _ structure _ stands for the most general , four - valued case . given a fixed @xmath29 and @xmath30 , an alternative way to represent @xmath40 is as a set @xmath45 of domain literals .
indeed , there is a one - to - one correspondence between such sets @xmath45 and @xmath30-structures @xmath40 with domain @xmath29 such that for a domain atom @xmath20 , @xmath46 ( inconsistent ) if both @xmath20 and @xmath47 are in @xmath45 , @xmath48 if only @xmath20 is in @xmath45 , @xmath49 if only @xmath47 is in @xmath45 and @xmath50 ( unknown ) otherwise .
hence , we may treat four - valued structures as sets of domain literals and vice versa .
a structure is _ inconsistent _ if at least one domain atom is inconsistent .
a structure @xmath40 of a vocabulary @xmath30 can be naturally viewed as a structure of a larger vocabulary @xmath51 , namely by setting @xmath50 for any domain atom of a predicate in @xmath52 . for a set @xmath53 of predicate symbols
, we use @xmath54 to denote the restriction of @xmath40 to the symbols of @xmath53 . for a set @xmath45 of domain atoms
, we use @xmath55 to denote the restriction of @xmath40 to @xmath45 : @xmath56 if @xmath57 and @xmath58 otherwise .
we call @xmath40 a two - valued structure of @xmath45 if @xmath40 is two - valued on domain atoms of @xmath45 and unknown otherwise
. the inverse @xmath59 of a truth value @xmath60 is defined as follows : @xmath61 , @xmath62 , @xmath63 and @xmath64 .
the _ truth _ order @xmath65 on truth values is defined by @xmath66 and @xmath67 .
the _ precision _
order @xmath68 is defined by @xmath69 and @xmath70 .
both orders are pointwise extended to arbitrary @xmath14-structures .
we say that @xmath34 is an _ expansion _ of @xmath34if @xmath71 , that is if for each domain atom @xmath20 , @xmath72 .
viewing structures as sets of domain literals , this corresponds to @xmath73 .
we assume familiarity with the syntax of ( function - free ) . to facilitate the reasoning with partially grounded formulas , we deviate from standard and quantify over explicitly specified subsets of the domain @xmath29
this is denoted as @xmath74 and @xmath75 , with @xmath76 .
we sometimes abbreviate @xmath77 as @xmath78 , and similarly for @xmath79 . given a formula @xmath80 ,
@xmath81 $ ] indicates that @xmath82 are the free variables of @xmath80 .
substitution of a variable @xmath22 in formula @xmath80by a term @xmath7 is denoted by @xmath83 $ ] .
a _ ground formula _ ( in domain @xmath29 ) is a formula without variables ( hence without quantifiers ) .
similar properties and notations are used for _ rules _ ( introduced below ) .
we denote by @xmath84 the set of all predicate symbols that occur in theory @xmath85 . for a structure @xmath40
, @xmath86 is the set of symbols interpreted by @xmath40 . unless specified otherwise , theories and structures range over the vocabulary @xmath14 .
the language extends fo with ( inductive ) _
definitions_. a theory in is a ( finite ) set of sentences and definitions . a definition @xmath87is a ( finite ) set of rules of the form @xmath88 , with @xmath17 a predicate symbol and @xmath80an fo formula . the atom @xmath89 is referred to as the _ head _ of the rule and @xmath80as the _
body_. given a rule @xmath90 , we let @xmath91 and @xmath92 denote respectively the head and the body of @xmath90 .
given a definition @xmath87 , a domain atom @xmath31 is _ defined _ by @xmath93 if there exists a rule @xmath94 in @xmath93 such that @xmath95 .
otherwise @xmath31 is _ open _ in @xmath87 .
a domain literal @xmath96 is defined by @xmath87if @xmath31 is defined by @xmath87 .
the sets of defined and open domain atoms of @xmath87are denoted as @xmath97 and @xmath98 , respectively . without loss of generality
, we assume that in any definition a domain atom is defined by at most one rule .
technically , this means that rules @xmath99 , @xmath100 are pairwise disjunct , that is @xmath101 .
rules can always be made disjunct by transforming them in @xmath102 , @xmath103 , @xmath104 .
the semantics of is a two - valued model semantics .
nevertheless , we introduce concepts of three- and four - valued semantics which are useful in defining the semantics of definitions and in formalizing lazy grounding .
we use the standard four - valued truth assignment function , defined by structural induction for pairs of domain formulas @xmath105 and structures @xmath40 that interpret @xmath105 : * @xmath106 , * @xmath107 , * @xmath108 , * @xmath109 , * @xmath110^{{\ensuremath{\mathcal{i}}\xspace}}\mid d\in d\})$ ] , * @xmath111^{{\ensuremath{\mathcal{i}}\xspace}}\mid d\in d\})$ ] .
the assignment function is monotonic in the precision order : if @xmath112 , then @xmath113 .
hence , if a formula is true in a partial structure , it is true in all two - valued expansions of it .
also , if @xmath40 is two - valued ( respectively three - valued , four - valued ) then @xmath114 is two - valued ( respectively three - valued , four - valued ) .
a structure @xmath34is a _ model _ of / _ satisfies _ a sentence @xmath80(notation @xmath115 ) if @xmath34is two - valued and @xmath116 .
the satisfaction relation can be defined for definitions as well .
the semantics of definitions is based on the parametrized well - founded semantics , an extension of the well - founded semantics of logic programs informally described first in the work of , and formally defined for s definitions by .
this semantics formalizes the informal semantics of rule sets as ( inductive ) definitions @xcite .
a structure @xmath34is a _ model _ of / _ satisfies _ a definition @xmath87(notation @xmath117 ) if @xmath34is two - valued and is the well - founded model denoted as @xmath118 of @xmath87 in the structure @xmath119 @xcite . in case @xmath118
is not two - valued , @xmath93 has no model expanding @xmath120 .
a structure @xmath34satisfies a theory @xmath121if @xmath34is two - valued and @xmath34is a model of all sentences and definitions in @xmath121 . in the next subsection , we present a formalization of the well - founded semantics using the notion of _ justification_.
according to s methodology , ( formal ) definitions are used to express informal definitions . in the work of
, it was shown that definitions offer a uniform representation of the most important types of informal definitions and that expressing informal definitions leads to rule sets that are _
total_. formally , a definition @xmath87is called _ total _ if the well - founded model of @xmath87 in each two - valued structure @xmath34of @xmath122 is two - valued @xcite .
in general , totality is undecidable ; however broad , syntactically defined classes of definitions have been proven to be total .
inspection of current applications shows that in practice , non - total definitions occur rarely and almost always contain a modeling error .
also , in most cases totality can be established through a simple syntactic check .
totality can be usefully exploited during computation .
the lazy grounding techniques introduced below exploit totality and should be applied only to total definitions .
this restriction matches with s design and methodology and , in practice , this does not impose a strong limitation . in case
the input theory does contain definitions that are not known to be total , all is not lost : those definitions can be grounded completely up - front , in which case lazy grounding can be applied safely to the remaining sentences and total definitions in the input .
[ [ equivalence . ] ] equivalence .
+ + + + + + + + + + + + two theories @xmath85 and @xmath123 , which can be over different vocabularies , are _
@xmath14-equivalent _ if each model of @xmath85 restricted to @xmath14 can be expanded to a model of @xmath123 and vice versa .
two theories @xmath85 and @xmath123 are _ strongly @xmath14-equivalent _ if the above expansions are also unique . by extension , ( strong ) @xmath14-equivalence _ in a structure @xmath34 _ is defined similarly : if each model of @xmath85 expanding @xmath40 can be expanded to a model of @xmath123 expanding @xmath34and vice versa ; to obtain strong equivalence , these expansions have to be unique . from a theory @xmath85
, we often derive a _ strongly _ @xmath84-equivalent theory @xmath123 in a given structure @xmath40 .
such transformations preserve satisfiability and number of models _ and _ each model of @xmath123 can be directly mapped to a model of @xmath85 by projection on @xmath84 .
[ [ canonical - theories . ] ] canonical theories .
+ + + + + + + + + + + + + + + + + + + to simplify the presentation , the lazy grounding techniques are presented here for theories of the form @xmath124 , with @xmath125 a propositional symbol , and @xmath93 a single definition with function - free rules .
this is without loss of generality .
first , as mentioned above , standard techniques @xcite allow one to make a theory function - free .
second , multiple definitions can always be combined into one as described by and .
this is achieved by renaming defined predicates in some of the definitions , merging all rules into one set and adding equivalence constraints between predicates and their renamings .
third , the theory @xmath126 resulting from the previous step can be translated to the strongly @xmath127-equivalent theory @xmath128 with @xmath129a new propositional symbol .
this transformation results in a ground set of sentences and a definition consisting of a set of ( ground and non - ground ) rules , so lazy grounding has only to cope with non - ground rules .
furthermore , we assume that rule bodies are in negation normal form ( negation only occurs in front of atoms ) and that , for each defined domain atom @xmath31 , there is a unique rule @xmath130 such that @xmath131 .
the methods proposed below can be extended to full with functions , and such extended methods have been implemented in our system .
however , this introduces a number of rather irrelevant technicalities which we want to avoid here .
we assume the presence of a domain @xmath29 and a canonical theory @xmath132 as explained above .
recall , structures with domain @xmath29 correspond one - to - one to sets of domain literals .
a _ direct justification _ for a defined domain literal @xmath31 ( respectively @xmath133 ) is a consistent non - empty set @xmath45 of domain literals such that , for the rule @xmath134 of @xmath93 such that @xmath95 , it holds that @xmath135^s = { { { \ensuremath{{\bf t}}\xspace}}}$ ] ( respectively @xmath135^s = { { { \ensuremath{{\bf f}}\xspace}}}$ ] ) .
any consistent superset @xmath136 of a direct justification @xmath45 of @xmath31 is a direct justification as well .
indeed , a body @xmath135 $ ] true in @xmath45 is true in the more precise @xmath136 . also , a direct justification @xmath45 is not empty by definition ; if @xmath105 is true in every structure , then a minimal direct justification is @xmath137 .
[ ex : justif1 ] consider a domain @xmath138 and the definition @xmath87@xmath139 a direct justification for @xmath140 is @xmath141 and for @xmath142 is @xmath143 .
both domain literals have many other direct justifications , but those are the unique minimal ones under the subset relation .
minimal direct justifications for @xmath144 are both @xmath145 and @xmath146 while the only minimal direct justification for @xmath147 is @xmath148 .
atoms @xmath149 are open and have no direct justification . a ( directed ) graph @xmath150 is a pair @xmath151 of a set @xmath152 of nodes and a set @xmath153 of directed edges , i.e. , ordered pairs @xmath154 of nodes .
for any node @xmath155 , we denote by @xmath156 the set of children of @xmath60 , i.e. , @xmath157 . a _ justification _ over a definition @xmath87is a graph @xmath158 over the set of domain literals of @xmath87such that for each domain literal @xmath21 , @xmath159 is either empty or a direct justification of @xmath21 .
thus , a justification is a graph that encodes for every defined domain literal none or one direct justification . in the sequel
we say that @xmath158 is _ defined in @xmath21 _ if @xmath160 .
a justificationis denoted as a set of pairs @xmath161 , with @xmath45 a direct justification of @xmath21 .
let @xmath162be a justificationover @xmath87 .
the justification__for a literal _ _ @xmath21 is the subgraph @xmath163 of nodes and edges of @xmath158 reachable from @xmath21 .
the justification__for a set of literals _ _ @xmath164 is the subgraph @xmath165 of nodes and edges of @xmath158 reachable from any @xmath166 .
a justification@xmath162over @xmath87is _ total for _ @xmath21 if @xmath158 is defined in each literal that is reachable from @xmath21 and defined in @xmath87 ; it is _ total for a set of literals _ @xmath164 if it is total for each literal in @xmath164 . a justification@xmath162is _ consistent with _ a structure @xmath34if @xmath34is consistent and none of the literals for which @xmath162is defined is false in @xmath34 .
if @xmath162is total for @xmath21 , then the leaves of @xmath163 are open domain literals . a path in a justification@xmath162is a sequence @xmath167 such that , if @xmath168 , then there is an edge from @xmath169 to @xmath170 in @xmath162 .
a path is _ positive _ if it consists of only positive literals ; it is _ negative _ if it consists of only negative literals ; it is _ mixed _ otherwise .
cycle _ in a justification@xmath162is a set of domain literals on a path in @xmath162that starts and ends in the same domain literal .
a cycle is positive ( respectively , negative ) if all domain literals are positive literals ( respectively , negative literals ) ; otherwise the cycle is mixed .
an infinite path may be cyclic or not .
if @xmath29 is finite , every infinite path is cyclic .
intuitively , a justification @xmath158 containing a domain literal @xmath21 provides an argument for the truth of @xmath21 .
the strength of this argument depends on the truth of the leaves and on the infinite paths and cycles in @xmath163 .
if all leaves are true and every infinite path is negative , @xmath163 provides the argument that @xmath21 is true .
if a leaf is false or unknown , or @xmath163 contains a positive or mixed loop , the argument for @xmath21 is weak .
notice that other justifications for @xmath21 may still argue @xmath21 s truth .
[ def : justifies ] we say that a defined literal @xmath21 is _ well - founded _ in the justification @xmath158 if every infinite path in @xmath163 is negative .
otherwise @xmath21 is _ unfounded _ in @xmath162 .
a justification@xmath158 over @xmath87__justifies _ _ a set of literals @xmath164 defined in @xmath87(the set @xmath164 of literals _ has a justification _
@xmath158 ) if ( * i * ) @xmath165 is total for @xmath164 ; ( * ii * ) each literal of @xmath164 is well - founded in @xmath158 ; ( * iii * ) the set of literals in @xmath165 is consistent .
[ ex : justif2 ] ( pd1 ) @xmath171 ; ( qd1 ) @xmath172 ; ( pd1 ) edge[->,bend left=-50 ] ( qd1 ) ; ( qd1 ) edge[->,bend left=-50 ] ( pd1 ) ; \(i ) ( i ) ; ( pd2 ) @xmath171 ; ( rd2 ) @xmath173 ; ( qd2 ) @xmath172 ; ( pd2 ) edge[->,bend left=0 ] ( rd2 ) ; ( qd2 ) edge[->,bend left=0 ] ( pd2 ) ; \(ii ) ( ii ) ; ( pd3 ) @xmath171 ; ( qd3 ) @xmath172 ; ( qd3 ) edge[->,bend left=-50 ] ( pd3 ) ; \(iii ) ( iii ) ; ( npd4 ) @xmath174 ; ( nqd4 ) @xmath175 ; ( nrd4 ) @xmath176 ; ( npd4 ) edge[->,bend left=0 ] ( nqd4 ) ; ( npd4 ) edge[->,bend left=0 ] ( nrd4 ) ; ( nqd4 ) edge[->,bend left=-50 ] ( npd4 ) ; \(iv ) ( iv ) ; in figure [ fig : justex ] , we show a few possible justifications ( ordered ( i)-(iv ) from left to right ) over definition @xmath87 in example [ ex : justif1 ] that contain the defined domain atoms @xmath171 and @xmath172 ( @xmath177 ) . justification ( ii ) justifies @xmath171 and @xmath172 and ( iv ) justifies @xmath174 and @xmath175 ; ( iii ) , however , is not total for @xmath171 nor @xmath172 and ( i ) has a positive cycle and is unfounded for both @xmath171 and @xmath172 . the relationship between justifications and the well - founded semantics has been investigated in different publications @xcite . below we recall the results on which this paper relies .
the first result states that if @xmath162justifies all literals in @xmath164 , then any model @xmath40 of @xmath93 in which the leaves of @xmath165 are true , satisfies all literals in @xmath164 and in @xmath165 .
[ prop : just ] if @xmath162is a justificationover @xmath87that justifies a set of domain literals @xmath164 then all literals in @xmath165 are true in every model of @xmath93 in which the ( open ) leaves of @xmath165 are true . for an interpretation @xmath178 that is two - valued for @xmath122 , the well - founded model @xmath179 can be computed in time polynomial in the size of the domain , as shown by . in general
, @xmath179 is a three - valued structure .
if @xmath179 is two - valued , then it is the unique model of @xmath93 that expands @xmath178 ; otherwise , @xmath87has no model that expands @xmath178 .
the above proposition follows from the fact that if a justification @xmath162justifies @xmath164 and all leaves of @xmath162are true in @xmath178 , then all literals of @xmath164 are true in @xmath179 .
justification ( ii ) justifies @xmath180 and has a unique open leaf @xmath173 .
for any structure @xmath178 interpreting the open predicates of @xmath93 , if @xmath173 is true in @xmath178 , then @xmath172 is true in @xmath179 . in particular , in any model of @xmath87 in which @xmath173 is true , @xmath172 is true . [
prop : justbis ] if @xmath40 is a model of @xmath87 , then a justification@xmath162over @xmath87exists that consists of literals true in @xmath34 , is defined for all defined domain literals true in @xmath40 and justifies each of them .
[ cor : voila ] in case @xmath87is total , if a justification@xmath162over @xmath87justifies a set of domain literals @xmath164 , then every two - valued @xmath181-structure consistent with @xmath165 can be extended in a unique way to a model of @xmath87that satisfies all literals of @xmath164 .
hence , for a canonical theory @xmath124 ( recall , @xmath87is total ) , the theory is satisfiable iff a justification @xmath162exists that justifies @xmath129 .
_ model generation _ is the inference task that takes as input a theory @xmath121and returns as output a model of @xmath121 .
was defined by as the inference task that takes as input a theory @xmath121over vocabulary @xmath14and a two - valued structure @xmath34over a subvocabulary of @xmath30 , and returns an expansion @xmath182of @xmath34that satisfies @xmath121 . here , it will be the more general inference problem as defined by that takes as input a ( potentially partial ) structure @xmath34over @xmath14 , and returns an expansion @xmath182of @xmath34that satisfies @xmath121 . as already mentioned , the state - of - the - art approach to model expansion in is ( similar to ) grounding @xmath121 in the context of @xmath34and afterwards applying search to the resulting ground theory .
the latter can , e.g. , be accomplished by the sat(id ) search algorithm .
below , we present the grounding algorithm that is the basis of the lazy algorithm .
we assume familiarity with the basic algorithm of sat solvers @xcite .
for an overview of intelligent grounding techniques in , we refer the reader to the work of and of .
below we present the basic principle .
a grounder takes as input a theory @xmath121over vocabulary @xmath14 , a partial structure @xmath34with domain @xmath29 , interpreting at least @xmath15and @xmath16 , and returns a ground theory @xmath183 that is strongly @xmath30-equivalent with @xmath121 in @xmath34 .
theory @xmath183 is then called a _ grounding _ of @xmath121given @xmath34 .
recall that we assume that @xmath121is a canonical theory of the form @xmath124 .
one way to compute the grounding is using a top - down process on the theory , iteratively applying grounding steps to direct subformulas of the rule or formula at hand .
the grounding algorithm may replace subformulas by new predicate symbols as follows .
let @xmath81 $ ] be a formula in @xmath121and let @xmath184be the domains of @xmath185 .
a _ tseitin transformation _ replaces @xmath80by the atom @xmath186 , with @xmath187 a new @xmath188-ary predicate symbol called a _
tseitin _ symbol , and extends @xmath87with the rule @xmath189 .
the new theory is strongly @xmath14-equivalent to the original one .
the procedure , outlined in figure [ fig : onestepground ] , performs one step in the grounding process . called with a formula or rule @xmath105 in canonical form , the algorithm replaces all direct subformulas with tseitin symbols and returns a pair consisting of a ground part @xmath150 ( rules or formulas ) and a possibly non - ground part @xmath19 ( rules ) . if @xmath105 is a formula , then @xmath150 consists of ground formulas . replacing @xmath105 by the returned ground formulas and extending @xmath87with the returned rules produces a theory that is strongly @xmath127-equivalent to the original .
if @xmath105 is a rule from @xmath87 , @xmath150 consists of ground rules , and replacing @xmath80by both sets of returned rules results again in a theory that is strongly @xmath127-equivalent to the original .
grounding a theory then boils down to applying on the sentence @xmath129(which copies @xmath129to the ground part ) and on each rule of the theory and repeatedly applying on the returned rules @xmath19 ( all returned sentences and rules in @xmath150 are ground ) .
we use to refer to the algorithm for this overall process .
various improvements exist , such as returning @xmath15/@xmath16for atoms interpreted in @xmath34and returning @xmath16from conjunctions whenever a false conjunct is encountered ( analogously for disjunctions and quantifications ) .
also , algorithm introduces a large number of tseitin symbols .
state - of - the - art grounding algorithms use a number of optimizations to reduce the number of such symbols . as
these optimizations are not directly applicable to the techniques presented in this paper , we start from the naive algorithm . in section
[ sec : optimizations ] , we present an optimized version of that introduces fewer tseitin symbols and hence results in smaller groundings .
we use the term _ lazy grounding _ to refer to the process of partially grounding a theory and the term _ lazy model expansion _
( lazy mx ) for the process that interleaves lazy grounding with model expansion over the grounded part . in section
[ sec : lazy - foid ] , we formalize a framework for lazy model expansion of theories ; in section [ sec : default ] , we formalize the instance of this framework that is the basis of our current implementation ; in section [ sec : lazyexample ] , we illustrate its operation . given a canonical theory @xmath190 and an input structure @xmath191 , models expanding @xmath191are searched for by interleaving lazy grounding with search on the already grounded part .
we first focus on the lazy grounding .
apart from the initial step that moves @xmath129to the grounded part , the input of each step consists of a set of rules still to be grounded , an already grounded theory and a three - valued structure that is an expansion of the initial input structure .
each subsequent grounding step can replace non - ground rules by ground rules and might introduce new rules .
hence , the state of the grounding includes a set @xmath192of ground rules and a set @xmath193(the _ delayed definition _ ) of ( possibly ) non - ground rules .
the definitions have the property that @xmath194 ( in what follows abbreviated as @xmath195 ) is @xmath196-equivalent with the original definition @xmath87and hence , @xmath195is total .
the grounding procedure will guarantee that , at all times , @xmath192and @xmath193are total . given a partial structure @xmath191and the rule sets @xmath192and @xmath193 , the key idea behind
lazy model expansion is ( * i * ) to use a search algorithm to search for a model @xmath34of @xmath192that is an expansion of @xmath191 in which @xmath129is true ; ( * ii * ) to maintain a justification@xmath162such that the literals true in @xmath34and defined in @xmath193are justified over @xmath195and that @xmath162is consistent with @xmath34 ; ( * iii * ) to interleave steps ( * i * ) and ( * ii * ) and to move parts of @xmath193to @xmath192when some literal defined in @xmath193that needs to be justified can not be justified .
thus , to control lazy model expansion , it suffices to maintain a state @xmath197 consisting of the grounded rules @xmath192 , the rules @xmath193yet to be grounded , a justification@xmath162 , and a three - valued structure @xmath34 .
initially , @xmath40 is @xmath198 , @xmath199 is @xmath93 , @xmath200 , and @xmath162is the empty graph
. lazy model expansion searches over the space of _ acceptable _ states .
[ def : acc_state ] a tuple @xmath197 of a theory with an atomic sentence @xmath129 , a total definition @xmath87 , and an input structure @xmath191is an _ acceptable _ state if ( * i * ) @xmath195 , @xmath192and @xmath193are total definitions and @xmath195is strongly @xmath196-equivalent with @xmath87 , ( * ii * ) no domain atom is defined in both @xmath192and @xmath193 , ( * iii * ) @xmath162is a justificationover @xmath195 , ( * iv * ) @xmath34is an expansion of @xmath191 , ( * v * ) the set @xmath164 of literals true in @xmath34and defined in @xmath193is justified by @xmath162 , and ( * vi * ) @xmath165 , the justification of the literals in @xmath164 , is consistent with @xmath34 .
[ ex : cons ] consider the theory @xmath124 , with @xmath87the definition + + + + let @xmath34be the structure @xmath201 ( hence , @xmath202 and @xmath203 are unknown ) , and @xmath192and @xmath193the definitions consisting of the first rule and the remaining rules , respectively .
furthermore , let @xmath162be @xmath204 .
the tuple @xmath205 is then an acceptable state .
indeed , @xmath206 is the only literal in @xmath34that is defined in @xmath193and it is justified by @xmath162 .
as already said , the lazy model expansion algorithm starts from the initial state @xmath207 , which is acceptable if defined literals are unknown in @xmath198 . in each state
, it either refines @xmath40 by propagation or choice , or it backjumps .
if the resulting state is unacceptable , a repair operation restores acceptability ; these steps are described in section [ sec : default ] .
the algorithm tries to compute an acceptable state in which @xmath129is justified in @xmath208 . by corollary [ cor : voila ] , this would entail that a model of @xmath121exists ; it can be computed efficiently through well - founded model computation . in intermediate states
, the justification may be non - total for @xmath129 , contain unfounded literals , or be inconsistent .
note that , in ( * iii * ) , the justificationmust be over @xmath195 . indeed , assume some literal @xmath21 is justified over @xmath193 .
its justification graph can have a leaf that is defined in @xmath192and that depends positively or negatively on @xmath21 . then
every attempt to extend this justification graph to a total justification graph that justifies @xmath21 over @xmath195might fail , e.g. , because of a forbidden cycle .
consider , e.g. , the definitions @xmath209 and @xmath210 . in that case
, it would not be correct to take @xmath17 as justification for @xmath18 being true , even though it is a valid justification within @xmath193 .
indeed , no model exists that justifies @xmath18 in the full definition @xmath195 .
[ prop : model ] let @xmath197 be an acceptable state .
@xmath195has a well - founded model that expands the literals that are true in @xmath40 and defined in the ( delayed ) definition @xmath193 .
let @xmath164 be the set of literals true in @xmath40 and defined in @xmath193 .
as the state is acceptable , @xmath162justifies the literals of @xmath164 .
hence , by corollary [ cor : voila ] , there exists a well - founded model that expands @xmath164 . the well - founded evaluation , after assigning @xmath15to the open literals of @xmath162(i.e .
, to @xmath211 ) , derives that @xmath206 is true .
moreover , because @xmath34is a model of @xmath192 , @xmath129is also true in such a well - founded model .
note that @xmath19 can be interpreted randomly , as no @xmath19-atoms occur in @xmath34or @xmath162 .
the following theorem states when the obtained expansion is also a model of @xmath121 .
[ theo : acceptablemodel ] let @xmath197 be an acceptable state of a theory @xmath212 with input structure @xmath191such that @xmath129is true in @xmath34and @xmath213 is a model of @xmath192 .
then there exists a model @xmath182of @xmath121that expands @xmath213 .
@xmath213 is a model of @xmath214 .
it follows from proposition [ prop : justbis ] that there exists a justification @xmath215 over @xmath214 that justifies every true defined literal of @xmath214 and that consists of only domain literals true in @xmath213 .
we now have two justifications : @xmath158 and @xmath215 .
we combine them in one @xmath216 as follows : for each defined literal @xmath21 of @xmath208 , if @xmath158 is defined in @xmath21 , we set @xmath217 ; otherwise , we set @xmath218 . as @xmath219 takes edges from either @xmath162 or @xmath220 for each defined literal , it is a justification for @xmath195 .
we verify that @xmath216 justifies @xmath129 .
first , it is total in @xmath129 .
indeed , any path from @xmath129either consists of literals defined in @xmath192 , and then it is a branch of the total @xmath215 over @xmath192 , or it passes to a literal @xmath221 defined in @xmath193 , which is justified by @xmath158 according to condition ( * v * ) and hence @xmath222 is total . as such , from @xmath129we can not reach a defined literal of @xmath195 in which @xmath216 is undefined .
second , @xmath216 does not contain unfounded literals starting from @xmath129 .
this is because any path from @xmath129is either a path in @xmath215 ( so well - founded as it justifies @xmath192 ) or it has a tail in @xmath162(well - founded by property ( * v * ) ) .
finally , the set of literals reachable from @xmath129 in @xmath216 is consistent .
also this we can see if we look at paths in @xmath216 from @xmath129 : at first we follow @xmath215 which consists of true literals in @xmath40 , then we may get into a path of @xmath162which contains literals that are consistent with @xmath40 . in any case , it is impossible to reach both a literal and its negation .
it follows from proposition [ prop : just ] that there exists a model of @xmath195that expands @xmath213 and in which @xmath129is true .
since @xmath195is strongly equivalent with @xmath87 , the proposition follows .
recall that effectively computing such a model @xmath182can be achieved by well - founded evaluation of @xmath195 , with polynomial data complexity , starting from any two - valued @xmath223-structure expanding @xmath213 @xcite . in the above theorem
, it is required that @xmath34is a model of @xmath192 .
actually , we do not need to compute a two - valued model of @xmath192 .
it suffices to search for a partial structure and a justificationthat justifies @xmath129 .
so , we can relax this requirement at the expense of also maintaining justifications for literals true in @xmath34and defined in @xmath192 .
[ col : acceptable ] let @xmath197 be an acceptable state of a theory @xmath224 with input structure @xmath191such that @xmath129is true in @xmath34and @xmath162justifies @xmath129over @xmath195
. then there exists a model @xmath182of @xmath121that expands @xmath55 with @xmath45 the set of defined literals in @xmath225 .
failure to find a model of @xmath192expanding @xmath191 in which @xmath129is true implies the lack of models of @xmath121expanding @xmath191 .
indeed , if @xmath192has no model expanding @xmath191 , then it has an unsatisfiable core , i.e. , a set of rules from @xmath192such that no model exists that expands @xmath191 . hence , it is also an unsatisfiable core for @xmath226 . to find an unsatisfiable core , one can , for example , use techniques described by
roughly speaking , our lazy model expansion framework consists of two components . on the one hand , a standard model expansion algorithm that operates on @xmath227 and , on the other hand , a justification manager that maintains a justificationover @xmath195and lazily grounds @xmath193 .
lazy model expansion performs search over the space of acceptable states and aims at reaching a state where theorem [ theo : acceptablemodel ] ( or corollary [ col : acceptable ] ) is applicable . to avoid slowing down the search during model expansion , the work done by the justification manager and the lazy grounding must be limited . to achieve this ,
we have designed a system in which the justification manager has no access to the grounded definition @xmath192and need not restore its state when the search algorithm backtracks over the current structure @xmath34 .
the justification manager only has access to @xmath34and maintains justifications that are restricted to @xmath193 .
in particular , a literal defined in @xmath192is not allowed in a direct justification . our justification manager maintains the following properties : * literals in direct justifications are either open in @xmath195or defined in @xmath193 . *
all direct justifications in @xmath162are kept consistent with each other and with the current structure @xmath34 .
* the justification graph defined by @xmath162has no unfounded literals and is total . to distinguish acceptable states that meet these additional requirements from acceptable states as defined in definition [ def : acc_state ]
, we call them _ default acceptable states _ ; we define them as : [ def : default_acc_state ] a state @xmath197 is a default acceptable state if it is an acceptable state and , in addition , ( * i * ) literals in direct justifications are either open in @xmath195or defined in @xmath193 , and ( * ii * ) @xmath162justifies the set of all literals for which @xmath158 is defined .
it follows that default acceptable states satisfy two extra conditions : they do not justify literals defined in @xmath193 in terms of literals defined in @xmath192 , and the set of all literals in @xmath162is consistent . for an acceptable state , it suffices that the literals in @xmath162that are true in @xmath34and defined in @xmath193 , are consistent . since default
acceptable states are acceptable states , theorem [ theo : acceptablemodel ] and corollary [ col : acceptable ] also hold for default acceptable states . during standard model expansion ,
the main state - changing operations are to make @xmath34more precise ( by making literals true , either through choice or propagation ) and to make @xmath34less precise ( by backjumping ) .
when @xmath228 is a default acceptable state and model expansion modifies @xmath34into @xmath34 , the new state @xmath229 is not necessarily a default acceptable state .
the following propositions identify situations where acceptability is preserved .
[ prop : forward ] let @xmath205 be a default acceptable state , @xmath164 a set of literals unknown in @xmath34and @xmath230 the consistent structure @xmath231 . if ( * i * ) literals of @xmath164 either are not defined in @xmath193or have a direct justification in @xmath162and ( * ii * ) no direct justification in @xmath162contains the negation of a literal in @xmath164 , then @xmath232 is a default acceptable state . as the literals true in @xmath34and
defined in @xmath193have a direct justification , it follows from ( * i * ) that all literals true in @xmath34 and defined in @xmath193have a direct justification . as justifications in @xmath162are consistent with @xmath34 , then , by ( * ii * ) , they are also consistent with @xmath34. hence , @xmath162justifies all literals true in @xmath34 and defined in @xmath193 . [ prop : backward ] let @xmath205 be a default acceptable state
. then @xmath232 with @xmath233 is a default acceptable state .
the justification @xmath162justifies all literals defined in @xmath87and true in @xmath34 . as
@xmath34 is a subset of @xmath34 , @xmath162 justifies all literals defined in @xmath87and true in @xmath34. in a default acceptable state , literals defined in @xmath192are not allowed in direct justifications of literals defined in @xmath193 .
this restriction is quite limiting ( see next section ) but is to avoid hidden loops over @xmath195 .
such loops can only be detected by maintaining a justification over both @xmath192and @xmath193 , which our current implementation does not do .
several methods exist to extend the class of default acceptable states .
literals @xmath21 defined in @xmath192can be allowed in direct justifications of @xmath193 , provided it can be established that @xmath21 s justification can not loop over @xmath195 .
one case is when the body of the rule of @xmath21 has no defined literals .
a step further is to analyze the dependency graph : a literal defined in @xmath192can be allowed in the direct justification of a literal defined in @xmath193provided both literals do not belong to the same strongly connected component of the dependency graph . in that case , they can not be part of the same cycle . in the rest of the section
, we illustrate the behavior of lazy model expansion on an artificial example , constructed in such a way that all main features are illustrated . in the next section ,
the processes involved are described in more detail .
we focus on the operation of the justification manager and its interaction with the solving process .
the manager is activated in an unacceptable state , either when the solver falsifies a literal that occurs in a direct justification of @xmath162or when a true literal @xmath21 defined in @xmath193is not justified by @xmath162 .
one option for repair is to search for a justification for @xmath21 to extend @xmath162 . in general
this problem is as hard as the model expansion problem itself , as shown by corollary [ cor : voila ] .
our manager only searches _ locally _ for a direct justification that justifies @xmath21 to extend @xmath162 , and if it does not find one , it grounds @xmath21 s definition and moves it to @xmath192 .
our example uses a theory @xmath121which states that a symmetric graph ( @xmath234 ) exists where at least one node other than the root node ( predicate @xmath235 ) is reachable ( predicate @xmath236 ) from the root node .
the input structure @xmath34interprets the domain as @xmath237 and the equality predicate as the identity relation on @xmath29 ( below omitted in @xmath34 ) .
predicates @xmath238 and @xmath239are not interpreted ; @xmath19 and @xmath240 are defined .
in particular , @xmath239is defined as the singleton @xmath241 , specifying the root as @xmath242 .
@xmath243 the lazy algorithm proceeds as follows : 1 . the initial default acceptable state is @xmath197 in which @xmath192 , @xmath34and @xmath162are empty , and @xmath244 .
2 . propagation over @xmath245 sets @xmath40 to @xmath246 .
this expands the structure @xmath34 , but now the conditions of proposition [ prop : forward ] are no longer satisfied .
the resulting state is not acceptable since @xmath129is true and defined in @xmath193while it has no direct justification in @xmath162 .
one option to repair acceptability is to extend @xmath162with a direct justification for @xmath129 .
the atom @xmath129has a unique direct justification @xmath247 but extending @xmath162with it does not restore ( default ) acceptability since @xmath248 have no direct justification in @xmath162and @xmath129remains unjustified .
therefore , the alternative is taken and rule ( 1 ) is moved to @xmath192 .
now , a default acceptable state is obtained .
unit propagation sets @xmath40 to @xmath249 .
now @xmath250 and @xmath251 have to be justified .
consider first @xmath251 and rule ( 3 ) .
as @xmath252 is open , our manager can build the direct justification @xmath253 , that sets all negative @xmath252 literals true , and extends @xmath162with it ( setting all positive @xmath252 literals true would be equally good ) .
this justifies @xmath251 and avoids the grounding of the rule defining @xmath251 .
literal @xmath250 can not be justified ( with the local approach ) since each of its direct justifications contains unjustified defined literals .
however , as rule ( 2 ) is existentially quantified , one can avoid grounding the whole rule by performing a tseitin transformation to isolate one instance and then only ground that instance .
for the purpose of illustration , we make the ( bad ) choice of instantiating @xmath22 with @xmath242 : @xmath254 rule ( 2a ) is moved to @xmath214 and a default acceptable state is reached .
we are in an acceptable state in which no further propagation is possible , so a choice has to be made .
as @xmath250 is true , the body of rule ( 2a ) has to become true .
preferably not selecting a tseitin ( this would trigger more grounding ) , the first disjunct is selected by model expansion and propagation extends the structure with @xmath255 and @xmath256 .
the literal @xmath255 is defined in @xmath193by rule ( 4 ) but can not be justified since its unique direct justification @xmath257 is false .
the manager partially grounds the definition of @xmath258 and splits it up in a ground rule ( 4a ) and a non - ground rule ( 4b ) defining @xmath258 for the other domain elements : @xmath259 rule ( 4a ) is moved to @xmath192 .
note that @xmath260 is justified by @xmath261 in @xmath195 , hence it is safe to use @xmath260 in direct justifications in @xmath193 . whenever grounding has been done , the justification manager is interrupted by propagation , which can infer the truth of additional literals , or detect an inconsistency ( which will result in backjumping ) . in both cases ,
the manager has to resume the revision of the justification afterwards , until an acceptable state is reached . here , even though the resulting state is still unacceptable ( due to the unjustified @xmath256 ) , the creation of the new rule ( 4a ) in @xmath192interrupts the manager .
propagation using the new rule derives @xmath260 and a conflict . after backtracking to @xmath262 ,
the subsequent propagation sets the structure @xmath34to @xmath263 . still not in a default acceptable state (
@xmath85 is not justified ) , rule ( 2b ) is further transformed to split off another instance . @xmath264
rule ( 2ba ) is moved to @xmath214 , while rule ( 2bb ) remains in @xmath193 .
this state is default acceptable
again , the search avoids the new tseitin @xmath202 , choosing the first disjunct in rule ( 2ba ) which propagates @xmath265 and @xmath266 .
the literal @xmath267 is defined in @xmath193 , but is justified by the direct justification @xmath268 .
the literal @xmath266 can not be justified by a direct justification ( as all @xmath252 literals are false in the current justification graph ) and rule ( 5 ) is transformed to split off the instance for @xmath269 . actually , this instance in turn has a disjunctive body with a complex subformula , so to avoid grounding the subformula , we break it up in two parts and introduce another tseitin . @xmath270 rule ( 5aa ) is moved to @xmath214 , the others remain in @xmath199 . 7 .
the current structure @xmath34is @xmath271 , hence propagation on rule ( 5aa ) in @xmath192extends it with @xmath203 .
there is no direct justification justifying @xmath203 and , hence , rule ( 5ab ) is partially grounded by splitting off the @xmath242 case : @xmath272 rule
( 5aba ) is moved to @xmath192while rule ( 5abb ) remains in @xmath193 .
the search selects the first disjunct of @xmath203 s rule body and propagates @xmath273 and @xmath256 .
the literal @xmath256 is defined in @xmath193 , but @xmath274 is a direct justification for it . extending @xmath162with this direct justification yields an acceptable but not default acceptable state , since @xmath260 is defined in @xmath192 .
however , @xmath260 is justified in @xmath195 , making it safe to extend @xmath162with this direct justification as discussed earlier .
now the justification manager faces a new problem : the true literal @xmath273 is in conflict with the direct justification @xmath275 of @xmath251 ( rule ( 3 ) ) . to handle this conflict , it splits off the affected instance ( @xmath276 ) from this rule : @xmath277 rule ( 3a ) is moved to @xmath192while rule ( 3b ) remains in @xmath193 .
the direct justification of @xmath278 is set to @xmath279 , the unaffected part of the direct justification of @xmath251 .
this restores acceptability .
propagation on rule ( 3a ) extends @xmath34with @xmath280 and @xmath278 . the literal @xmath280 , which is true , is in conflict with the direct justification for @xmath278 ( rule ( 3b ) ) . to resolve it , the justification manager
partially grounds rule ( 3b ) and splits off the instance @xmath281 as follows .
@xmath282 rule ( 3ba ) is moved to @xmath192while rule ( 3bb ) remains in @xmath193 ; @xmath283 inherits the direct justification of @xmath278 with @xmath284 removed .
propagation on rule ( 3ba ) extends @xmath34with @xmath283 .
the resulting state is acceptable , with @xmath283 defined in @xmath193but justified . by now , @xmath192consists of the rules ( 1 ) , ( 2a ) , ( 4a ) , ( 2ba ) , ( 5aa ) , ( 5aba ) , ( 3a ) , and ( 3ba ) , and the residual definition @xmath193consists of the rules ( 4b ) , ( 2bb ) , ( 5b ) , ( 5abb ) , and ( 3bb ) .
the current structure @xmath34is @xmath285 @xmath286 @xmath287 , a model of @xmath288 . of these literals , @xmath267 , @xmath256 and @xmath283
are defined in @xmath193 .
literal @xmath267 , defined by rule ( 4b ) has @xmath289 as direct justification .
literal @xmath256 , defined by rule ( 5b ) , has @xmath290 as direct justification .
literal @xmath283 , defined by rule ( 3bb ) has as direct justification the set of all negative @xmath252 literals over @xmath29 except @xmath280 and @xmath273 . to obtain a full model of the theory
, @xmath34is extended with the literals of the above direct justifications . in this case
, this assigns all open literals and the model can be completed by the well - founded model computation over @xmath195 .
actually , this can be done without grounding the definition @xcite .
in section [ sec : default ] , we have instantiated our general framework , developed in section [ sec : lazy - foid ] , for a justification manager that only has access to @xmath193 . in the example of section [ sec : lazyexample ]
, the justification was constructed on demand , i.e. , each time some literal needed a ( different ) direct justification , the body of its defining rule was analyzed and a justification was extracted .
if that failed , part of the rule was grounded .
this was called the _
local approach_. one can also imagine a _ global approach _
, where more rules of @xmath193are considered at once in an attempt to select direct justifications that minimize the grounding of the rules as a whole .
obviously , a global approach will be more time consuming , so should not be applied every time an adjustment of the justification is required . in this section , we describe both approaches . before describing the algorithms ,
we introduce some notations and assume some normalizations have been done .
the function reduces a formula to its negation normal form . with @xmath45 a set and @xmath291 a single element , @xmath292 and @xmath293
are used as shorthands for @xmath294 and @xmath295 . with @xmath162a justification , we denote by @xmath296 $ ] the graph identical to @xmath162except that @xmath21 is now justified by @xmath24 .
we assume quantifiers range over a single variable and variable names are not reused in a formula .
furthermore , we assume basic reductions have been applied to formulas , e.g. , @xmath297 reduces to @xmath80 , @xmath298 reduces to @xmath38 , algorithm [ fig : lazymx ] shows the top level of the model expansion algorithm , taking as input theory @xmath124 and input structure @xmath191 .
definitions @xmath193and @xmath192are initialized with @xmath87and the empty definition , respectively , and @xmath34is initialized as @xmath191 .
the set of ground sentences @xmath299is initialized with the fact @xmath129and the initial justification @xmath162is empty .
an auxiliary ( fifo ) queue @xmath300is initialized as empty .
the latter keeps track of literals for which the direct justification needs to be checked .
the main loop performs model expansion over @xmath301 , interleaved with work by the justification manager towards establishing a default acceptable state .
the model expansion part consists of propagation ( the call to ) , the test on whether the current state is inconsistent ( with learning and backjumping ) , the test on whether a model of @xmath301 has been found ( returning the model and the justification ) and of the choice step that selects a literal unknown in @xmath301 and assigns it a value .
propagation returns literals that are entailed by a ground theory over a ( partial ) structure , for example by applying unit propagation and unfounded / wellfoundedness propagation .
the test for a model is only performed in a default acceptable state(i.e .
, when the queue @xmath300is empty ) .
if the test succeeds , this ensures that the well - founded model computation can expand the current structure @xmath34 extended with the direct justifications of all literals into a model of the whole theory .
also the choice step only takes place in a default acceptable state ; this ensures that the search space is limited to the state space of default acceptable states .
the justification manager is activated when a propagation or choice step assigns a literal @xmath21 . by calling
, it is checked whether the current justification remains valid .
if @xmath21 is defined in @xmath193and has no justification , it needs a justification and is added to the queue @xmath300for further processing by the justification manager . if @xmath302 occurs in the justification of another literal @xmath221 , the justification becomes inconsistent with @xmath34and @xmath221 needs another justification so it is also added to @xmath300 .
the further processing is done by selecting elements from the queue and calling the function .
the latter function first attempts to find a ( different ) consistent direct justification for @xmath21 ; if that fails , it splits off the rule instance defining @xmath21 from @xmath193and partially grounds it , hence @xmath192is extended .
the new clauses may trigger propagation ; therefore the processing of queued literals is interleaved with propagation and , possibly , with backtracking .
note that backtracking might restore the consistency with @xmath34of the direct justification @xmath159 of a literal @xmath21 on @xmath300 .
the function , algorithm [ fig : lazyground ] , checks whether the literal @xmath21 needs a direct justification ; if not , it simply returns . otherwise , it checks whether @xmath21 has a valid justification , i.e. , one that satisfies the invariants detailed below .
if so , it also returns ; otherwise , it passes the rule body that has to be used to construct a justification ( the negation of the defining rule when the literal is negative ) to , a function that attempts to find a valid direct justification . besides the literal and the rule body ,
also an initial justification , derived from the rule body , is passed to .
if the latter function is successful , the justification is updated and is done ; if not , the direct justification of the literal @xmath21 is set to @xmath303and is called to ground part of the rule defining @xmath21 . before going into more details ,
we first analyze which properties we want to maintain in the current justification @xmath162 .
the direct justifications of literals in the @xmath300queue are not considered part of @xmath162since they might be invalid .
the global invariants of @xmath162are : * no literals are unfounded in @xmath162(recall , negative cycles are allowed ) , * the set of literals in @xmath162is consistent . for each direct justification @xmath304 of @xmath162for
some @xmath21 not on the queue , invariants of the lazy grounding process are : * @xmath45 contains no literals defined in @xmath192(unless such a literal is safely justified in @xmath195 , as discussed before ) , * literals in @xmath45 that are defined in @xmath193either have a direct justification in @xmath158 or belong to the queue @xmath300 .
these invariants imply that a default acceptable stateis reached when the @xmath300queue is empty .
indeed , it follows from the invariants that the current justification is total in that situation and hence all literals that have a direct justification are justified ( definition [ def : justifies ] ) . due to
the policy followed to queue literals , the current justification is also consistent with @xmath34while all literals true in @xmath34and defined in @xmath193have a justification , hence @xmath197 is a default acceptable state .
the purpose of , algorithm [ fig : buildconstr ] , is to extend @xmath162with a suitable direct justification for literal @xmath21 defined in @xmath193 . here
@xmath21 is a literal for which @xmath159 is currently undefined or is inconsistent with @xmath34 .
it is a recursive function which takes three parameters : ( * i * ) the literal @xmath21 , ( * ii * ) the formula @xmath80to be made true by the direct justification ( initially the whole body of the rule defining the literal ; note that the initialization takes the negation of the rule when the literal is negative ) , ( * iii * ) a description of the direct justification derived so far , initialized through @xmath305 . for this algorithm ,
we assume @xmath80is such that different quantifiers range over different variables . before going into details ,
we discuss how to represent direct justifications .
basically , we could represent a direct justification as a set of ground literals .
however , this set can be quite large and using a ground representation might hence defy the purpose of lazy grounding .
instead , we represent a direct justification as a pair @xmath306 with @xmath164 a set of possibly non - ground literals and @xmath307 a set of bindings @xmath308 with @xmath309 a variable and @xmath310 a domain .
a set of bindings @xmath311 represents the set of variable substitutions @xmath312\}$ ] .
the set of ground literals represented by @xmath306 is then @xmath313 .
the direct justification of a literal @xmath314 , defined by a rule @xmath315 , is initialized by @xmath305 as @xmath316 . in effect
, @xmath307 allows to identify the relevant rule instantiation by providing the appropriate variable instantiation from the domains , while the set of literals is empty .
the algorithm searches for a set of literals making @xmath80true .
it works by recursively calling itself on subformulas of @xmath80and composing the results afterwards into a larger justification for @xmath80 .
when no such set of literals is found , for example because none exists that is consistent with all other direct justifications , @xmath317 is returned . the base case is when the formula is a literal . to make that literal true , all instances of the literal under the set of bindings @xmath307 must be true , hence
, the set of literals @xmath164 is extended with the literal itself .
the resulting direct justification has to satisfy all invariants , which is checked by the call to : it returns @xmath318for a call @xmath319 if @xmath320 satisfies the invariants to be ( part of ) a direct justification for @xmath21 and @xmath321 $ ] satisfies the invariants on the justification .
a universally quantified formula @xmath322 has to be true for each instance of the quantified variable .
hence , in the recursive call , the set of bindings @xmath307 is extended with @xmath323 . for an existentially quantified formula
, it suffices that one instance is true .
hence , a minimal approach is to try each instance separately until one succeeds ; if all fail , @xmath303is returned .
note however that we do not want to iterate over each domain element if @xmath29 is large , which would be similar to constructing the grounding itself . instead ,
if @xmath29 is large , we extend the binding with @xmath324 .
conjunction is similar to universal quantification , except that explicit iteration over each conjunct is needed .
as soon as one conjunct fails , the whole conjunction fails .
disjunction is similar to existential quantification with a small domain .
note that is non - deterministic due to choices for a domain element to justify an existentially quantified formula , or for a disjunct to justify a disjunction .
[ ex : build ] consider the following rule over a large domain @xmath29 .
@xmath325 assume that @xmath162is empty and we have no loops to keep track of . applying to @xmath326 returns @xmath327 if the first disjunct @xmath328 in the body is chosen .
this corresponds to the direct justification @xmath329 . alternatively , if the second disjunct is chosen , it returns @xmath330 , which represents the direct justification @xmath331 .
the last bit of the lazy model expansion algorithm handles the case where no justification can be found and the definition of a literal @xmath21 is to be grounded .
a straightforward way would be to call on the rule defining @xmath21 , and store the result in @xmath192and @xmath193 .
however , in many cases such an operation results in too much grounding . [ ex : splitting ]
consider a rule @xmath332 of the form @xmath333 in a situation where no justification can be found for atom @xmath171 .
applying to @xmath332 would instantiate @xmath22 with all elements in @xmath29 , resulting in @xmath334 rules , while in fact it suffices to split @xmath332 in two rules , one for the instance @xmath335 and one for the remainder .
another example applies to a rule @xmath336 of the form @xmath337 and a direct justification @xmath338 . when @xmath172 becomes false
, the justification manager may need to ground this rule . applying to it would instantiate the universally quantified @xmath22 with all elements in @xmath29 .
instead , it is better to split off the instance for @xmath335 and to introduce a tseitin @xmath85 for the remainder , producing @xmath339 for @xmath192and @xmath340 for @xmath193 . the direct justification for @xmath85 can be obtained incrementally by removing @xmath172 from that of @xmath326 , as discussed in section [ ssec : splitting ] .
the algorithm ( algorithm [ fig : splitandground ] ) has to ground part of the rule defining a given literal @xmath21 , say @xmath31 .
the first step is to split off the rule instance for which the rule defining @xmath21 has to be grounded ( the call to ) .
let @xmath315 be the rule in @xmath193that defines @xmath31 .
we then replace that rule by @xmath341 in @xmath193and additionally return the rule @xmath342 $ ] .
afterwards , we apply to the latter rule and add the computed rules to either @xmath192or @xmath193 .
the result of is that definition @xmath195is `` more '' ground than the previous one .
the limit is a ground definition @xmath195 in which @xmath193is empty and @xmath192is strongly @xmath343-equivalent with @xmath87 . even if no justification was found , we can do better than just splitting off @xmath21 and applying , as shown in example [ ex : splitting ] .
first , splitting can be made significantly more intelligent , which is discussed in section [ ssec : splitting ] .
second , we can improve to only ground part of expressions if possible , which we describe below .
[ [ improving- . ] ] improving .
+ + + + + + + + + + + applying to a rule @xmath344 iterates over all subformulas / instantiations of @xmath80 .
for example if @xmath80is the sentence @xmath345 , the result consists of @xmath334 new rules and as many new tseitin symbols . instead , depending on the value of @xmath21 , it is sufficient to introduce only one ( or some ) of these subformulas , as shown in algorithm [ algo : improveground ] , which extends the switch statement of with two higher - priority cases . if @xmath21 is true , it is sufficient to ground one disjunct / existential instantiation and to delay the rest by tseitin introduction .
if @xmath21 is false , we take a similar approach for conjunction / universal quantification .
correctness and termination of the presented algorithms is discussed in the following theorem . if returns an interpretation @xmath34 , then expanding @xmath34with the literals in the direct justifications of @xmath162 , applying the well - founded evaluation over @xmath195and restricting it to @xmath127 results in a model of @xmath121 .
if the algorithm returns @xmath303 , no interpretation exists that is more precise than @xmath191and satisfies @xmath121 .
algorithm terminates if @xmath34is over a finite domain @xmath29 .
otherwise , termination is possible but not guaranteed .. ] if returns an interpretation @xmath34 , @xmath34is a model of @xmath192and @xmath300is empty . given the properties of , after applying for a literal @xmath21 , either @xmath21 has a valid justification or is defined in @xmath192 .
hence if @xmath300is empty , we are in a default acceptable state and , by theorem [ theo : acceptablemodel ] , @xmath34can be expanded into a model of the whole theory .
if returns @xmath303 , it has been proven that @xmath192has no models in @xmath191 . in that case
, there can also be no models of @xmath195and hence @xmath121also has no models expanding @xmath191 . without calls to
, the search algorithm terminates for any finite @xmath192 ; the function itself produces an ever - increasing ground theory @xmath192with the full grounding as limit .
hence , always terminates if @xmath29 is finite .
if @xmath29 is infinite , the limit of @xmath192is an infinite grounding , so termination can not be guaranteed .
the algorithms presented above are sound and complete .
however , they can be further improved by taking the formulas from which justifications are derived into account .
[ [ symbolic - justifications - and - incremental - querying . ] ] symbolic justifications and incremental querying .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + when multiple justifications exist for ( subformulas of ) a formula @xmath80 , grounding can be further delayed .
consider the formula @xmath346 , where both @xmath347 and @xmath348 are justifications . from that
, we could derive the justification : for each @xmath349 , make either @xmath171 _ or _
@xmath172 true .
hence , more grounding is necessary _ only _ when both @xmath171 and @xmath172 become false for the same @xmath24 .
we can do this automatically by changing as follows : * the algorithm is allowed to select multiple disjunctions / existential quantifications even if a valid justification was already found for one ( lines [ buildconstr - enougha ] and [ buildconstr - enoughb ] ) .
* no longer returns a justification , but a symbolic _ justification formula _ that entails the original formula .
the formula is built during and reflects the subformulas / instantiations that have been selected . from @xmath350 ,
the justification itself can be derived directly as the set of non - false literals in the ( full ) grounding of @xmath350 .
for example , for a formula @xmath351 , instead of the justification @xmath352 , might now return the justification formula @xmath353 . * the validity check ( ) is extended to return false if the justification formula is false . by allowing more complex formulas ( instead of a conjunction of universally quantified literals ) , the validity check whether a formula has become false after incremental changes to @xmath34 is now more expensive .
this is in fact an _ incremental query _ problem . in the experiments , we limit the depth of the allowed formulas and use a straightforward ( expensive ) algorithm that evaluates the whole formula whenever an assignment can falsify it . [ [ body - splitting . ] ] body splitting .
+ + + + + + + + + + + + + + + as described in algorithm [ fig : lazyground ] of section [ sec : splitandground ] , simply splits off the rule instance that defines @xmath21 , then grounds this rule instance step by step , in accordance with the structure of the formula .
however , when the grounding is triggered by a conflict with the current justification , is blind for the origin of the conflict . by using the conflicting literals
, one could focus on grounding the part of the formula that contributes to the conflict .
one can do so by restructuring the rule in a part to be grounded , that contains the conflict , and a part not to be grounded , for which the old justification can be adjusted to still apply .
the latter part can then be split off by introducing new tseitins and the transformation is called body splitting .
this approach can be inserted in algorithm [ fig : lazyground ] after the call to .
for this , the original justification ( call it @xmath354 ) is passed as an extra argument to .
consider the rule @xmath355 ; let @xmath347 be the justification for @xmath356 true in @xmath34 .
when @xmath171 becomes false , it is easy to see that we can split off the `` violating instantiation '' by rewriting the original rule into @xmath357 and adding the rule @xmath358 .
crucially , a justification for the second part can be derived from the original justification , namely @xmath359 .
the second part can hence be added to @xmath193and its justification to @xmath162while the first part is added to @xmath214 .
this revision of a rule @xmath90 and its direct justification @xmath354 can be done in an efficient way .
assume that @xmath60 is a true domain literal in the partial structure and that the direct justification of rule @xmath90 contains its negation @xmath360 .
the implementation is such that the binding(s ) for which the justification instantiates to @xmath360 can be extracted from the representation of the direct justification of the rule . for simplicity ,
assume @xmath361 is the single such instance . a recursive algorithm visits the formula @xmath80 in the body of @xmath90 depth - first .
whenever a quantification @xmath362 is encountered with @xmath22 equal to @xmath363 , it is replaced by @xmath364 $ ] .
then a tseitin transformation is applied to the left - hand conjunct and the algorithm recurses on the right - hand conjunct with what remains of the binding .
the new rule defining the new tseitin has @xmath365 as a direct justification .
similarly , an existential quantification is replaced by a disjunction .
the result is a set of new rules for which no new justification has to be sought and a smaller rule @xmath366 which is passed to .
correctness follows from the fact the @xmath365 is a valid justification , none of the new rules contains @xmath60 , and from the correctness of the tseitin transformation .
[ ex : bodysplit ] in example [ ex : build ] , justifications were sought for @xmath326 in the rule @xmath367 assume we selected the justification @xmath368 .
when @xmath369 is true in @xmath34 , and @xmath370 becomes false , @xmath158 is no longer consistent with @xmath34and can not be repaired .
@xmath371 , however , is still consistent with @xmath34 , but is not a justification of the whole body . on the other hand , @xmath371 is a justification for the subformula @xmath372 for each instantiation of @xmath22 different from @xmath242 .
consequently , we can split the quantification @xmath373 into @xmath374 and @xmath375 and apply a tseitin transformation to the former .
afterwards , we recursively visit the latter formula and apply a similar reasoning to the existential quantification . the operations on the formula are illustrated in figure [ fig : split ] .
the result consists of the following rules , where the rule for @xmath326 is now even ground .
+ + child node [ node ] @xmath376 child[normal ] node [ node ] ( or ) @xmath377 child[halfline ] node [ r_node ] @xmath328 child node [ node ] @xmath378 child node [ node ] @xmath379 child node [ node ] @xmath380 child[halfline ] node [ r_node ] @xmath381 childnode [ empty_node ] .
child node [ node ] ( or2 ) @xmath377 child[halfline ] node [ r_node ] ( ln ) @xmath382 child node [ empty_node ] at ( @xmath383 ) @xmath377 ( * split 2 * ) child[normal ] node [ node ] @xmath384 child[normal ] node [ node ] ( or3 ) @xmath379 child node [ node ] @xmath385 child[halfline ] node [ r_node ] @xmath386 childnode [ empty_node ] .
child[normal ] node [ node ] at ( @xmath387 ) @xmath379 child node [ node ] @xmath388 child node [ node ] @xmath389 ; to further optimize the traversal of the formula @xmath80 , can be extended to store the path taken through the parse tree of @xmath80and the direct justifications of the subformulas .
assume the rule @xmath390 has to be justified , @xmath162is empty and @xmath34does not interpret @xmath252 .
the algorithm recursively visits the body of the rule until @xmath391 is returned as it is a valid literal to use .
going up one level , we store that for @xmath392 , we selected @xmath393 . assuming no more disjuncts are selected , @xmath391 is returned again .
going back up through both quantifications , we store that , for both quantifications , we selected @xmath29 as the set of relevant domain elements , and returns the justification formula @xmath394 .
if is given access to @xmath354 , similar optimizations are possible for repairing a direct justification .
consider again example [ ex : bodysplit ] , but assume @xmath369 is unknown in @xmath34 . in this case , the left branch of figure [ fig : split ] can also be transformed in a rule with a still valid , direct justification .
for the right branch , a repair is to select the disjunct @xmath382 as direct justification for the rule @xmath395 , where @xmath206 is as in example [ ex : bodysplit ] .
finding justifications using the greedy local approach can easily lead to more grounding than necessary .
consider for example the sentences @xmath396 and @xmath397 . applying the local approach to the second sentence first ( with an empty @xmath162 ) , might result in the construction which makes atoms over @xmath17 false . then applying the local approach to the first sentence finds no valid justification for it
; so it has to be fully grounded .
the global approach takes a set of rules as input and tries to select direct justifications for them such that the _ expected _ grounding size of the whole set is minimal .
we cast the task , called the _ optimal justification _ problem , as a problem on a graph as follows .
the graph consists of two types of nodes , _ rule nodes _ @xmath19 and _ justification nodes _ @xmath162 .
a justification node is a symbolic set of literals representing a possible justification ( for the literals defined by a rule in @xmath193 , given the current partial structure @xmath34 ) .
a rule node is a pair @xmath398 of a rule @xmath90 in @xmath193and a truth value @xmath7 ( @xmath38 , @xmath39 , @xmath399 ) ; a pair @xmath400 for a rule @xmath90 with head @xmath21 expresses that there exists a direct justification for @xmath21 ; the pair @xmath401 that there exists a direct justification for @xmath402 and the pair @xmath403 that r has no justification
. there are three types of edges : * _ valid edges _ between a rule node @xmath400 ( @xmath401 ) and a justification node @xmath404 where @xmath404 justifies ( the negation of ) the head of @xmath90 . *
_ conflict edges _ between ( * i * ) rule nodes of the same rule with different truth value , ( * ii * ) justification nodes that contain opposite literals , ( * iii * ) a rule node @xmath400 ( @xmath405 ) where @xmath90 defines @xmath21 , and a justification node that contains @xmath402 ( @xmath21 ) , and ( * iv * ) a rule node @xmath406 , where @xmath90 defines @xmath21 , and a justification node that contains @xmath21 or @xmath402 ( a conflict because @xmath21 ( or @xmath402 ) needs a justification ) . *
_ depends - on edges _ between a justification node @xmath404 and a rule node @xmath400 ( @xmath405 ) where @xmath404 contains negative ( positive ) literals defined by @xmath90 .
the aim is then to select subsets @xmath407 and @xmath408 such that * each selected rule node is connected with a valid edge to at least one selected justification node . *
no conflict edges exist between pairs of selected nodes .
* neither positive nor mixed cycles exist in the subgraph consisting of the valid and depends - on edges between the selected nodes . from a selection @xmath409 satisfying these constraints , an initial
justification @xmath162can be extracted as follows .
a literal @xmath21 ( @xmath402 ) is given a direct justification if it is defined by a rule @xmath90 such that @xmath400 ( @xmath401 ) is a selected rule .
its direct justification is the union of the justifications of the justification nodes in @xmath410 connected with @xmath400 ( @xmath401 ) through a valid edge .
moreover , all literals defined by rules for which no rule node was selected can be added to the initial @xmath300queue , to be handled by the local approach , as a complete solution must have a justification for them .
when @xmath403 is selected , it means that the grounding of instances of the rule is delayed until the literals it defines become assigned .
this type of problem is somewhat related to the np - hard _ hitting set _ ( or _ set cover _ ) problem @xcite : given a set of `` top '' and `` bottom '' nodes and edges between them , the task is to find a minimal set of bottom nodes such that each top node has an edge to at least one selected bottom node . given a default acceptable state @xmath197 , the input for the optimal justification problem is generated as follows . for any rule in @xmath193 ,
a node is constructed for each of the three truth values ( only one when the truth value is known in @xmath34 ) and also their conflict edges are added .
valid edges and justification nodes are obtained using a ( straightforward ) adaptation of that returns a set of possible justifications that make the head of a rule true ( false ) .
e.g. , for a rule @xmath315 , is called with the literal @xmath89 and the binding is initialized at @xmath411 .
conflict and depends - on edges are derived by checking dependencies between justifications and between rules and justifications . to keep this efficient ,
it is done on the symbolic level .
[ ex : global ] consider the theory of our running example , after @xmath129 , @xmath250 and @xmath251 have been propagated to be true .
definition @xmath193is then @xmath412 the associated optimal construction set input is shown in figure [ fig : ocs ] .
note that in rule nodes , we use the defined head literals to identify the rule .
literal @xmath250 and @xmath251 are true in @xmath34 , hence there is only one rule node for rules ( 2 ) and ( 3 ) .
neither @xmath413 nor @xmath414 can be justified for all @xmath415 , hence there is only a @xmath416 tuple .
there are four solutions that are subset - maximal with respect to rule nodes , namely the following rule node selections : @xmath417 for each of these , multiple justification selections are possible ( also shown in figure [ fig : ocs ] ) . for @xmath250 , we have to select justification @xmath418 , but for @xmath251 we can choose from @xmath419 or @xmath420 ( but not both ) .
( rf ) @xmath421 ; ( jrfa ) @xmath422 ; ( jrfb ) @xmath423 ; ( ru ) @xmath424 ; ( rt ) @xmath425 ; ( jrt ) @xmath426 ; ( c1 t ) @xmath427 ; ( jc1 t ) @xmath428 ; ( c2 t ) @xmath429 ; ( jc2at ) @xmath430 ; ( jc2bt ) @xmath431 ; ( rootu ) @xmath416 ; ( rt ) edge[-,bend left=+60 ] ( rf ) ; ( rt ) edge[- ] ( ru ) ; ( rf ) edge[- ] ( ru ) ; ( jrfb ) edge[- ] ( rt ) ; ( jc1 t ) edge[-,bend left=+20 ] ( rf ) ; ( jc1 t ) edge[- ] ( c1 t ) ; ( jc2at ) edge[- ] ( c2 t ) ; ( jc2bt ) edge[- ] ( c2 t ) ; ( jc2at ) edge[- ] ( jc2bt ) ; ( jc2at.east ) edge[-,bend left=-50 ] ( jrfa.east ) ; ( jrt ) edge[- ] ( rt ) ; ( jrfa ) edge[- ] ( rf ) ; ( jrfb ) edge[- ] ( rf ) ; ( rt ) edge[- ] ( jc1t.west ) ; ( rf ) edge[- ] ( jrfb.west ) ; the objective is then not to maximize the number of selected rule nodes , but to minimize the expected grounding size . to obtain an estimate of the expected grounding size
, the following conditions were taken into account : * it should depend on the size of the grounding of the rule . *
assigning multiple justifications to a rule should result in a lower estimate as the rule will only be grounded when all are false .
* variables occurring in multiple justifications result in less matching instantiations . * in most practical applications ,
the number of false atoms far exceeds the number of true ones in any model .
hence , positive literals in a justification should have a higher cost than negative ones .
we approximate the expected grounding size with the function @xmath432 which takes as input a rule @xmath90 ( with head @xmath356 ) , the selected type of justification ( rule not selected ( * n * ) , no justification ( @xmath399 ) , a justification for @xmath356 ( @xmath38 ) or a justification for @xmath433 ( @xmath39 ) ) and the set of justifications @xmath162 .
the function returns the expected grounding size of the rule ( @xmath434 , defined below ) weighted depending on the type of justification .
the weights are derived from two estimates : @xmath435 is the probability an atom will become assigned and @xmath436 is the probability of an assigned atom to be true .
hence , as defined formally below , for non - delayed rules ( * n * ) , the full size is used ; a rule without justification ( * u * ) is weighted by @xmath435 ; a true justification ( * t * ) is weighted by @xmath436 , a false one ( * f * ) by @xmath437 ; the latter two weights are multiplied by a product over the justifications @xmath404 in @xmath162 .
each factor of this product is a sum of two terms , namely @xmath438 times the number of negative literals in @xmath404 and @xmath437 times the number of positive literals in @xmath404 .
the effect is that the expected size decreases as the number of justifications increases and that the expected size increases as a justification has more literals .
@xmath439 for the probabilities , we assumed @xmath435 to be small ( currently 0.1 ) to reflect the hope that lots of literals will not get a value , and @xmath436 is less than half , to reflect that atoms are more often assigned false than true .
the function @xmath440 is defined below .
the function returns the number of atoms in the grounding of a rule or formula , except for existential quantification and disjunction . for these ,
we take into account that they can be grounded only partially using tseitin transformation , by taking the logarithm of the total grounding size .
@xmath441}size({{\ensuremath{\varphi}\xspace}}_i ) \\ size(\exists { { \ensuremath{x\in d:}\xspace } } { { \ensuremath{\varphi}\xspace}})&=log(d)\times size({{\ensuremath{\varphi}\xspace } } ) \\ size(\phi_1 \lor \ldots \lor { { \ensuremath{\varphi}\xspace}}_n)&=log(n)\times\frac{\sum_{i \in [ 1,n]}size({{\ensuremath{\varphi}\xspace}}_i)}{n}\end{aligned}\ ] ] solutions to the optimal justification problem should minimize the term @xmath442 with @xmath443 the type ( @xmath38 , @xmath39 , or @xmath399 ) and @xmath444 the justification of the literal defined by @xmath90 .
the size of rule @xmath250 is @xmath445 , that of @xmath251 is @xmath446 , that of @xmath239 is @xmath447 , and that of @xmath19 is @xmath448 .
consider assigning justification ( iv ) to @xmath250 : this results in an expected cost for that rule of @xmath449 ( as the construction relies on making @xmath19 true ) .
additionally , it would force the grounding of the rule defining @xmath19 , increasing the cost with the size of the rule for @xmath19 .
the optimal solution for the problem in figure [ fig : ocs ] is then rule node selection ( [ rulenodea ] ) with justification ( vi ) for @xmath251 .
its cost is the sum of @xmath450 ( for justification ( vi ) ) and @xmath451 ( the expected size of the rule for @xmath250 ) .
now , only the rule for @xmath250 has to be passed to the local approach . to solve the optimal justification problem , s optimization inference itself
is applied to a ( meta - level ) declarative specification of the task . for larger theories @xmath121 ,
the problem turns out to be quite hard , so two approximations were considered to reduce the search space .
first , the number of selected justifications for any rule was limited to 2 .
second , as the values of @xmath434 can grow quite large , the approximation @xmath452 is used ( a standard approach ) . rounding to integer values was applied as s support for floating point number is still preliminary .
the resulting specification could be solved to optimality within seconds for all tested theories . during lazy model expansion
, the global approach is applied in the initial phase when all tseitin literals representing sentences in the original theory have been propagated true .
this section discusses how to tune the lazy grounding and the search heuristics of the underlying sat solver to obtain an effective implementation of lazy model expansion .
we also describe a few other inferences tasks beyond model expansion that are useful in the context of lazy grounding .
a few less important issues are discussed in appendix [ app : furtherconsid ] .
heuristics play an important role in the lazy grounding algorithms , as they serve to find the right balance between how much to ground and how long to search .
we first discuss how our heuristics were chosen .
afterwards , we discuss an alternative approach to minimize grounding .
the algorithms leave room for a number of heuristic choices that can have an important effect on the performance .
we now briefly discuss these choices . as a guideline for our decisions ,
the following principles are used : * avoid leaving the search process without enough information to make an informed decision ; for example , avoid losing too much ( unit ) propagation or introducing too many tseitin symbols .
* prevent creating a grounding that is too large ; this may for example happen as the result of a very long propagate - ground sequence .
recall , the goal is not to create a minimal grounding , but to solve model expansion problems while avoiding a too large grounding .
below , we introduce a number of parameters that affect these heuristics .
the exact values used in the experimental evaluation for the parameters introduced below are specified in appendix [ app : furtherconsid ] . in , when handling a disjunction or existential quantification , there is a choice on how many disjuncts to expand .
if we expand one instantiation at a time for a rule @xmath453 , as done in algorithm [ algo : improveground ] ( lines [ algo : selectonea ] and [ algo : selectoneb ] ) , iterative application results in a ground theory @xmath454 a sat - solver such as minisat , which is used in the system , initially assigns @xmath455 to the @xmath144 atoms ; such a choice triggers an iteration of propagation and grounding .
the resulting thrashing behavior can be reduced somewhat , and the grounding is more compact when the grounding introduces @xmath28 disjuncts at a time : @xmath456 to further remedy this , two search - related heuristics are changed .
first , the initial truth value is randomized , but favoring false ( as in most models , many more atoms are false than true ) .
second , search algorithms typically _ restart _ after an ( ever - increasing ) threshold on the number of conflicts , sometimes caching the truth value assigned to atoms ( _ polarity caching _ ) .
this allows the solver to take learned information into account in the search heuristic while staying approximately in the same part of the search space . in case of lazy grounding
, we might want to jump to another part of the search space when we come across long propagate - ground sequences . to this end
, we introduce the concept of _ randomized restarts _ , which take place after an ( ever - increasing ) threshold on the number of times @xmath192is extended and randomly flipping some of the cached truth values .
in addition , always returns @xmath303if it is estimated that the formula has a small grounding . indeed ,
grounding such formulas can help the search .
whether a formula is considered small is determined in terms of its ( estimated ) grounding size .
the same strategy is used in : whenever the formula to which would be applied is small , is applied instead , to completely ground the formula .
grounding is applied during the search process as soon as unit propagation has taken place .
the result is a focus on the current location in the search space , but with the danger of grounding too much if there is no solution in that part of the space .
alternatively , we could apply the opposite strategy , namely to ground as late as possible : only apply additional grounding when the search algorithm terminates without ever having found a model in an acceptable default state .
such a strategy is well - known from the fields of incremental proving and planning , where the domain ( number of time steps ) is only increased after search over the previous , smaller bound has finished .
this guarantees a minimal grounding .
a prototype of this strategy has been implemented in with good results on planning problems .
the bulk of the paper focuses on model expansion ( mx ) for theories
@xmath121 , for which solutions are structures that are two - valued on @xmath127 .
often , one is only interested in a small subset of the symbols in @xmath127 .
this is for example the case for model generation for , the language which extends with existential quantification over relations .
an problem @xmath457 with an initial structure @xmath34 , relation symbols @xmath458 , , @xmath459 , and @xmath121an theory , can be solved by model generation for the theory @xmath460 with initial structure @xmath34and by dropping the interpretation of the symbols @xmath458 , , @xmath459 in the models .
another example is query evaluation for : given a theory @xmath121 , an initial structure @xmath34and a formula @xmath80 with free variables @xmath185(all in ) , the purpose of evaluating the query @xmath461 is to find assignments of domain elements @xmath462to @xmath185such that a model of @xmath121exists that expands @xmath34and in which @xmath463 $ ] is true . to solve it by model expansion in ,
a new predicate symbol @xmath85 is introduced and answers to the query are tuples of domain elements @xmath464 such that @xmath465 is true in a model of the theory @xmath460 extended with the sentence @xmath466 and the definition @xmath467 . in both cases , approaches using ( standard ) model expansion
compute a total interpretation and afterwards drop all unnecessary information , which is quite inefficient .
lazy model expansion can save a lot of work by only partially grounding the theory .
however , once a model is found for the grounded part , the justifications and the remaining definitions are used to expand the structure to a model of the full theory .
although this expansion is obtained in polynomial time , it is still inefficient when afterwards a large part of the model is dropped . to remedy this
, we define a variant of the model expansion task , denoted _ restricted _ mx .
restricted mx takes as input a theory @xmath121 , a structure @xmath34and an additional list of symbols @xmath468 , called _
output symbols_. solutions are then structures @xmath469 which are
two - valued on all symbols in @xmath468 and for which an expansion exists that extends @xmath34and is a model of @xmath121 . adapting lazy grounding to solve restricted mx can be done through an analysis of which justifications need not be added ( completely ) to the structure , splitting @xmath195into multiple definitions and only evaluating those defining output symbols or symbols on which those depend ( using a stratification argument ) .
the above - mentioned inference tasks can be cast trivially to restricted mx problems and lazy restricted mx then greatly improves the efficiency with respect to ground - and - solve , as shown in the experimental section .
the extension of with _ procedurally interpreted _ symbols @xcite provides another class of interesting problems .
such predicate symbols have a fixed interpretation , but to know whether a tuple belongs to the predicate , a procedural function has to be executed .
such an approach provides a clean way to combine declarative and procedural specifications .
consider for example a symbol @xmath470 that is interpreted by a procedure which executes an efficient prime - verification algorithm and returns true iff the given argument is prime .
we are generally not interested in the complete interpretation of @xmath471 , so it can be cast as a restricted mx problem with @xmath471 not in @xmath468 .
solving such a problem using lazy grounding then has the benefit of only executing the associated function _ during _
search for relevant atoms @xmath472 .
also for this task , we show an experimental evaluation in the next section .
the system has a state - of - the - art model expansion engine , as can be observed from previous answer - set programming competitions .
the lazy model expansion algorithms presented in this paper were implemented in the system , by extending the existing algorithms @xcite .
the current implementation is incomplete in the sense that the cycle check for justifications has not been implemented yet .
this only affects inductive definitions as non - inductive ones can be replaced by the fo formulas of their completion . as a workaround for the lack of a cycle check , , the function that constructs a direct justification , returns false for rules defining inductive predicates . as a consequence , an instance of such a rule is immediately grounded , although lazily , when a domain atom defined by the rule is assigned a value .
another consequence is that inductively defined predicates can not be used in justifications of other rules .
this affects three benchmarks of the asp competition ( described below in section [ ssec : aspcompexper ] ) , namely ` reachability ` , ` sokoban`and ` labyrinth ` . for these
, the grounding might be delayed even more in a complete implementation .
the section is organized as follows . in section [ ssec :
overhead ] , we evaluate the overhead of completely grounding a theory using the presented approach . in section [ ssec :
aspcompexper ] , we evaluate the effect of lazy grounding on a number of benchmarks of the asp competition . in section [ ssec :
additionalexper ] , a number of additional properties of the presented algorithms are demonstrated .
we tested three different setups : with the standard ground - and - solve approach ( referred to as ` g&s ` ) , with lazy model expansion ( ` lazy ` ) and the award - winning asp system gringo - clasp ( ` asp ` ) .
we used version 3.2.1-lazy , gringo 3.0.5 and clasp 2.1.2-st .
the parameters of the lazy grounding algorithms are discussed in section [ ssec : heur ] , the values used in the experiments are documented in appendix [ app : furtherconsid ] .
the experiments for sections [ ssec : overhead ] and [ ssec : additionalexper ] were run on a 64-bit ubuntu 13.10 system with a quad - core 2.53 ghz processor and 8 gb of ram .
experiments for section [ ssec : aspcompexper ] were run on a 64-bit ubuntu 12.10 system with a 24-core 2.40-ghz processor and 128 gb of ram .
a timeout of 1000 seconds and a memory limit of 3 gb was used ; out - of - time is indicated by ` t ` , out - of - memory by ` m ` .
lazy grounding may reduce grounding size and time but also causes overhead .
for instance , we expect the ( naive ) incremental querying of justifications to be costly as discussed previously .
the aim of this section is to quantify the overhead caused by lazy grounding . in the experiments below we compare the grounding time of the standard idp system with that of a _ naive _ instance of the lazy grounding algorithm that is forced to generate the complete grounding before starting the search
this instance was obtained from the standard algorithm using some small changes : the shortcut to ground small formulas at once is turned off , disjuncts and instances of existentially quantified formulas are grounded one by one , a defined literal is enqueued for lazy grounding as soon as it appears in @xmath192 . for comparison
, we also measure the cost of the standard lazy grounding algorithm that computes partial groundings .
we devised six benchmarks to test various aspects of the novel algorithm .
each benchmark is a simple theory with at most two sentences that is simple to solve .
the benchmarks are designed to measure the cost of different aspects of lazy grounding : delaying and resuming grounding , the querying needed to resume grounding , the splitting of formulas , etc .
specifically , the tested aspects are the following : 1 . overhead of delaying and resuming grounding in case of an existential quantifier with a large domain .
the sentence is @xmath473 .
standard grounding creates a single clause with @xmath28 disjuncts ; naive lazy grounding grounds the formula piece by piece and introduces @xmath474 tseitin symbols
overhead in case of an inductive definition , here @xmath475 .
both standard grounding and naive lazy grounding construct a ground rule for each @xmath476 atom .
3 . overhead in case of a universal quantification .
the sentence is @xmath477 .
while standard grounding creates @xmath28 atomic formulas , naive lazy grounding splits off one instance at a time and introduces @xmath474 tseitin symbols .
4 . lifted unit propagation ( lup ) @xcite is an important preprocessing step to reduce the grounding size . concretely , applying lup to the rules @xmath478 derives that the second formula follows from the first and hence does not need to be grounded at all .
this theory is used to check whether lup remains equally important in a system with lazy grounding .
overhead in case of nested universal quantification .
the sentence is of the form @xmath479 .
standard grounding creates a formula for each instance @xmath464 of @xmath82 with a tseitin for the grounding of @xmath480 .
naive lazy grounding creates an extra tseitin for each instance @xmath464 of @xmath82 and an extra set of tseitins for the piece by piece grounding of the subformula @xmath480 .
overhead of the incremental querying in case a symbolic justification has to be validated .
the sentence is @xmath481 , with an identical justification formula .
the formula is validated by checking the falsity of the query @xmath482 .
this query is re - evaluated each time an @xmath19-atom or @xmath45-atom is falsified .
experiments were done for predicates @xmath17 and @xmath18 with arity 3 and @xmath19 and @xmath45 with arity 2 , and domains of size 10 , 20 , 30 , 40 and 50 .
none of the predicates symbols were interpreted in the structure . in all experiments ,
the overhead for the time required to solve the initial optimization problem ( for the global approach ) was always around 0.02 seconds , so in itself negligible .
the results for the first three experiments are not shown as the differences between standard grounding and naive lazy grounding are negligible .
while expected for experiment 2 , for experiments 1 and 3 , it shows that our actual implementation eliminates the overhead for tseitins when quantifiers are not nested . in each of these three experiments , standard lazy grounding
is able to justify the formulas without grounding them and hence fast and almost insensitive to the domain size .
as shown in figure [ overhead ] , there is no difference between standard grounding and naive lazy grounding for experiment 4 . in both cases ,
the use of lup has a big impact on the size of the grounding and hence on the time .
while experiment 1 and 3 showed that a top level quantifier does not create overhead for lazy grounding , experiment 5 shows that this does not hold anymore for nested quantifiers and that naive lazy grounding has substantial overhead when compared with standard grounding .
note that this overhead is worst case . when tseitins can be justified , their definitions are not grounded , which explains why normal lazy grounding is faster than standard grounding and insensitive to the domain size .
experiment 6 shows that a more complex justification formula causes significant overhead for naive lazy grounding . also here
, the overhead is worst case and not visible in normal lazy grounding .
still , it is an important part of future research to reduce the overhead of the incremental querying of complex justification formulas .
second , we selected benchmarks from previous asp competitions to evaluate the lazy grounding algorithm in a more realistic setting .
many benchmarks solutions of that competition are carefully fine tuned for speed and minimal grounding .
lazy grounding is usually unable to substantially reduce the grounding of such theories and , due to its overhead , is then slower than standard ground and solve .
for this reason , we have sometimes selected modelings of the benchmarks that are more natural but less optimized in time and grounding size .
we justify this on the ground that the aim of our work is to improve inference for declarative _ modeling _
@xcite , where the emphasis is not on developing intricate encodings , but on modeling a problem close to its natural language specification .
we selected the following problems ( see the competition websites for complete descriptions ) .
they consist of problems that are known to be hard , in order to evaluate the effect of lazy model expansion on search , and problems that typically result in a large grounding . * ` reachability ` : given a directed graph , determine whether a path exists between two given nodes . * ` labyrinth ` : a planning problem where an agent traverses a graph by moving between connected nodes to reach a given goal node .
in addition , the graph can be manipulated to change its connectedness . * ` packing ` : given a rectangle and a number of squares , fit all squares into the grid without overlaps .
* ` disjunctive scheduling ` : schedule a number of actions with a given earliest start and latest end time with additional constraints on precedence and disjointness . * ` sokoban ` : a planning problem where a robot has to push a number of blocks to goal positions , constrained by a 2-d maze . * ` graph colouring ` : given a graph , assign colour to nodes ( from a given set of colours ) , such that no connected nodes have the same colour . * ` stable marriage ` : given a set of men and women and a set of preferences , find a `` stable '' assignment : no swap results in a better match . for each of these
, we used all instances from the 2011 and 2013 competitions , except for the 2013 ` reachability`instances , because of the huge data files which none of the systems is designed to handle . for ` stable marriage ` , ` graph colouring`and ` reachability ` , we based our encodings on the available asp - core-2 encodings . for `
packing`and ` disjunctive scheduling ` , we constructed a natural encoding and made a faithful translation to asp . for the more complex benchmarks of ` labyrinth`and ` sokoban ` , we used the original and gringo - clasp s specifications submitted to the 2011 competition .
for the lazy model expansion , we replaced cardinality expressions by their encoding as for the former no justifications are derived yet ; this also increases the size of the full grounding .
.the number of solved instances for the asp benchmarks and the average time taken on the solved instances .
different solvers solve quite different sets of instances . [ cols="<,^,^,^,^,^,^,^ " , ] during the modeling phase of an application , different encodings are typically tested out , in an attempt to improve performance or to locate bugs .
while modeling our experimental benchmarks , we noticed that simplifying a theory by dropping constraints often resulted in a dramatic reduction in the time lazy model expansion took to find a model .
standard model expansion , on the other hand , was much less affected by such simplifications . in our opinion , this observation , while hardly definitive evidence , is another indication that the presented algorithms are able to derive justifications for parts of a theory that can be satisfied cheaply .
in that way , the approach is able to distinguish better between problems which are inherently difficult and problems which would just have a large grounding .
lazy model expansion offers a solution for the blow - up of the grounding that often occurs in the ground - and - solve model expansion methodology for theories . and
techniques also process theories that can have a large grounding ; the constraint store of and mixed integer programming and the clauses of sat can be considered the equivalent of a grounded theory ( they are often derived from quantified descriptions such as `` @xmath483 * for all * @xmath484 and @xmath404 for which '' ) and can also become very large . and have reported a blow - up problem in these paradigms and a multitude of techniques has been developed to address it .
we distinguish four approaches .
first , concerning grounding up - front , research has been done towards _ reducing the size of the grounding _ itself through ( * i * ) _ static analysis _ of the input to derive bounds on variable instantiations @xcite , ( * ii * ) techniques to _ compile _ specific types of sentences into more compact ground sentences , ( * iii * ) detect parts that can be evaluated polynomially and ( * iv * ) detect parts that are not relevant to the task at hand ( e.g. , in the context of query problems ) as shown in the work of .
naturally , each of these approaches can be used in conjunction with lazy grounding to further reduce the size of the grounding . in ,
e.g. , lazy grounding is already combined with ( * i * ) and ( * iii * ) .
second , the size of the grounding can be reduced by _ enriching _ the language .
for example , asp solvers typically support ground aggregates ( interpreted second - order functions such as cardinality or sum that take sets as arguments ) , and and solvers support ( uninterpreted ) functions .
more recently , the constraint - asp paradigm was developed @xcite , that integrates asp and cp by extending the asp language with _ constraint
these are interpreted as constraints in a csp problem and can thus be handled using techniques .
various casp solvers are already available , such as clingcon ( @xcite ) , ezcsp mycommoncitationezcsp @xcite , mingo @xcite and inca mycommoncitationinca @xcite .
this technique is also integrated into .
inca and in fact implement lazy clause generation @xcite , an optimized form of lazy grounding for specific types of constraints .
the language hex - asp also extends asp , this time with _ external _ atoms that represent ( higher - order ) external function calls .
third , _ incremental approaches _ are well - known from model generation , theorem proving and planning . for these tasks ,
the domain is typically not fixed in advance , but part of the structure being sought , such as the number of time steps in a planning problem ( recall the sokoban example from the introduction ) .
such an approach typically works by grounding the problem for an initial guess of ( the number of elements in ) the domain .
afterwards , search is applied ; if no model was found , the domain is extended and more grounding is done .
this is iterated until a model is found or a bound on the maximum domain size is hit ( if one is known ) .
this technique is applied , e.g. , in the prover paradox @xcite and the asp solver iclingo .
fourth , and closest to lazy grounding itself , is a large body of research devoted to delaying the grounding of specific types of expressions until necessary ( for example until they result in propagation ) .
propagation techniques on the first - order level that delay grounding until propagation ensues have been researched within and within @xcite .
such techniques can be used in conjunction with lazy grounding as they derive more intelligent justifications for specific types of constraints than presented here .
for example , dao - tran et al . also presented an efficient algorithm for bottom - up propagation in a definition . within smt ,
various theory propagators work by lazily transforming their theory into sat , such as for the theory of bit vectors by .
investigated quantifier handling by combining heuristic instantiation methods with research into decidable fragments of fo theories , as these can be efficiently checked for models . within
, work has been done on goal - directed reasoning . both and demonstrate approaches , in the style of sld resolution , that apply top - down instantiation to answer queries over infinite domains .
extend an abduction framework to lazily generate part of the relevant sentences . in search algorithms , justifications ( or _
watches _ ) are used to derive when a constraint will not result in propagation or is already satisfied , and hence need not be checked in the propagation phase .
show how maintaining ( short ) justifications can significantly reduce the cost of the propagation phase .
in fact , a well - known technique already exists that combines search with lazy instantiation of quantifiers , namely _ skolemization _ , where existentially quantified variables are replaced by newly introduced function symbols .
universal quantifications are handled by instantiating them for those introduced function symbols .
reasoning on consistency can , e.g. , be achieved by congruence closure algorithms , capable of deriving consistency without effectively assigning an interpretation to the function symbols .
these techniques are used in tableau theorem proving @xcite and smt solvers @xcite .
formula @xcite interleaves creating a ground program and giving it to an smt solver , iterating when symbolic guesses proved to be wrong .
skolemization - based techniques typically work well in case only a small number of constants needs to be introduced , but have difficulty in case the relevant domain is large .
one can also see that lazy grounding ( with support for function symbols ) could incorporate skolemization by adapting the rules for grounding existential and universal quantification .
we expect skolemization to be complementary to lazy grounding , but an in - depth investigation is part of future work . in the field of probabilistic inference , several related techniques have been developed that also rely on lazy instantiation .
first , the problog system uses a form of static dependency analysis to ground a ( probabilistic ) program in the context of a given query , by constructing all possible ways to derive the query in a top - down fashion .
second , so - called _ lazy inference _ , applied e.g. in _ lazysat _
@xcite , exploits the fact that , for the considered inference , a ( fixed ) _ default _ assumption exists under which an expression certainly does not contribute to the probabilities .
hence , expressions for which the assumption certainly holds do not have to be considered during search .
third , _ cutting plane inference _
@xcite applies lazy inference in an interleaved setting , only constructing the part of the program for which the assumptions are not satisfied .
several aspects of the presented work need further investigation . one aspect is extending support to lazily ground more complex expressions , including aggregate expressions and ( nested ) function terms .
consider for example the sentence @xmath485 , which expresses that the sum of terms @xmath486 for which the atom @xmath171 is true , with @xmath177 , @xmath17 a predicate and @xmath487 a function , should be larger than 3 .
one can observe that it is not necessary to ground the whole sentence up - front .
for example , if @xmath487 maps to the natural numbers ( hence positive ) , the set @xmath488 is a minimal justification .
even if no easy justification can be found , we can suffice by grounding only part of the sentence and delay the remainder .
for example , we can create the ground sentence @xmath489 , with @xmath85 a tseitin symbol defined as @xmath490 .
indeed , in any model of the sentence in which @xmath85 is false , the original inequality is satisfied .
a second aspect is whether there are advantages to grounding earlier , for example to guarantee no propagation is lost , or grounding later , possibly reducing the size of the grounding even more .
for example , consider the sentences @xmath491 and @xmath492 , with @xmath493 and @xmath350 both large formulas for which no justification was found . instead of grounding at least one of the sentences , we might add @xmath17 to the list of atoms the search algorithm has to assign and only ground either of the sentences when @xmath17 has been assigned a value ( it might even be that unsatisfiability is detected before grounding either one ) . given that lazy grounding is useful , what about lazy _ forgetting _ the grounded theory ? as the ground theory is extended when
making the structure more precise , the ground theory could be reduced again during backtracking . by storing the justification violations that caused grounding , we can derive
which grounding can be forgotten again if the violation is no longer problematic ( e.g. , after backtracking ) . for this , an algorithm needs to be developed which tracks grounding / splitting dependencies between rules given their justifications .
this closely resembles techniques used in tableau theorem proving and smt , where the theory at hand can be compacted when moving to a different part of the search space .
the approach described for lazy grounding can also be applied to answer set generation in the field of . in ,
a logic program under stable semantics can be seen as one rule set , a single definition .
however , such asp programs do not satisfy a major condition to apply lazy grounding .
indeed such programs are typically non - total , due to the presence of constraints and rules of the form @xmath494 or other _ choice rules _ which result in multiple stable models . however , as described by , most practical programs can be partitioned in a set of choice rules , a set of _ total _ definitions and a set of constraints ( the so - called generate - define - test partition ) . any program that can be gdt - partitioned , can be translated straightforwardly into an equivalent theory that only contains total definitions .
this suggests a way to apply lazy grounding to such programs .
solvers used in the domains of sat , smt and asp are often confronted with problems that are too large to ground .
lazy model expansion , the technique described in this paper , interleaves grounding and search in order to avoid the grounding bottleneck .
the technique builds upon the concept of a justification , a deterministic recipe to extend an interpretation such that it satisfies certain constraints .
a theoretical framework has been developed for lazy model expansion for the language and algorithms have been presented to derive and maintain such justifications and to interleave grounding with state - of - the - art cdcl search algorithms .
the framework aims at bounded model expansion , in which all domains are finite , but is also an initial step towards handling infinite domains efficiently .
experimental evaluation has been provided , using an implementation in the system , in which lazy model expansion was compared with a state - of - the - art ground - and - solve approach .
the experiments showed considerable improvement over ground - and - solve in existing benchmarks as well as in new applications .
the main disadvantage is the less - informed search algorithm , caused by the delay in propagation and the introduction of additional symbols .
a possible solution is to develop new heuristics or portfolio approaches that combine the strengths of both methods .
finally , we have indicated a way how the proposed methods can be applied beyond , to solvers in general .
during this research , broes de cat was funded by the agency for innovation by science and technology in flanders ( iwt ) .
this research was also supported by fwo - vlaanderen and by the project goa 13/010 , research fund kuleuven .
nicta is funded by the australian government through the department of communications and the australian research council through the ict centre of excellence program .
in this appendix , we mention parameter values as well as some optimizations that further reduce the grounding overhead and/or improve the search . for each optimization , we indicate what is currently implemented ( and part of the experimental results ) and what is part of future work .
in [ ssec : heur ] , a number of parameters were introduced to control the behavior of lazy model expansion .
here , we provide details on the values used in the experimental evaluation .
these values were set manually , based on experience and a limited number of observations ( e.g. , the extension threshold works similar to the conflict threshold of the sat solver ) .
it is part of future work to study the impact of different values .
* for an existential quantification , 10 instantiations are grounded at a time ; for a disjunction , 3 disjuncts are grounded at a time .
this turned out to give the best balance between introducing too many tseitin atoms and grounding too much . *
the initial truth value is @xmath495 with probability @xmath496 and @xmath455 otherwise . *
the initial threshold for randomized restarts is 100 extensions of the ground theory .
it is doubled after each restart . * a formula
is considered small if its estimated grounding size is below @xmath497 atoms .
so far , we have described a lazy model expansion algorithm for function - free .
however , , the knowledge - base language of the system , supports a much richer input language . besides types which we use to initialize the domains it also supports ( partial ) functions , aggregates and arithmetic .
our current implementation ignores the latter extensions through a straightforward adaptation of ( algorithm [ fig : buildconstr ] ) : the case for literals is extended to return @xmath303when the literal is not part of the function - free language .
for example , given a rule @xmath498 , @xmath499 can be used in a justification but @xmath500 can not . for functions ,
there is also the option to replace them by graph predicates during the preprocessing step . as for the experiments of section [ ssec : aspcompexper ] , functions ,
if any , are given in the input structure and hence play no role .
it is part of future work to extend lazy grounding for these extensions , especially for functions .
techniques developed in smt and in constraint programming to handle ( ground ) atoms containing function symbols are useful to reduce the size of the grounding and improve search . in previous work ,
these techniques have been integrated in the system @xcite and it is certainly worthwhile to fully integrate them with lazy grounding . in , it is checked for each assigned literal whether it is defined in @xmath193and whether it violates any justifications . to implement this cheaply ,
our implementation maintains a mapping for literals in @xmath192 .
it states whether the literal is defined in @xmath193and also lists the justifications in which its negation occurs .
this mapping is extended whenever a new literal is added to @xmath192and maintained whenever justifications change .
the performance of the search loop is unaffected as long as literals are assigned for which the mapping is empty . in algorithm
[ fig : lazymx ] , we took the standard stopping criterion used in most search algorithms ( line [ stop - early ] ) : to stop in a conflict - free state where @xmath34is two - valued on all symbols of @xmath501 . in principle , we may stop earlier , with a partial structure @xmath34that admits a total justification for @xmath129 .
indeed , corollary [ col : acceptable ] tells us that such an @xmath40 can be expanded to a model .
this has a considerable impact on grounding size . indeed , assigning a truth value to an atom @xmath502 defined in
@xmath193that is irrelevant ( in effect , does not appear in the justification ) will trigger grounding of @xmath502 s definition , which in turn might introduce new literals defined in @xmath193 , causing a cascade of unnecessary groundings and assignments . our solver algorithm does not maintain a justification of @xmath192 , so it can not know exactly when a justification exists .
instead , the implemented algorithm only chooses literals that are watched by some formula / rule .
it stops with a partial structure in which unwatched literals may not be assigned .
it can be shown that this suffices to guarantee that @xmath34admits a justification .
hence it is safe to stop search . in some cases ,
can not find a valid justification for a large formula because a few literals are already false in @xmath34 .
for example for a formula @xmath503 , returns @xmath303if at least one atom of @xmath17 is false .
instead , we have adapted with a heuristic check on the number of expected violations .
if it is small enough , the justification is still returned . naturally , we are then required to check whether there are any real violations , by querying the justification formula over @xmath34 , and apply to them .
alviano , m. , calimeri , f. , charwat , g. , dao - tran , m. , dodaro , c. , ianni , g. , krennwallner , t. , kronegger , m. , oetsch , j. , pfandler , a. , phrer , j. , redl , c. , ricca , f. , schneider , p. , schwengerer , m. , spendier , l. k. , wallner , j. p. , xiao , g. 2013 .
the fourth answer set programming competition : preliminary reportin cabalar , p. son , t. c. , lpnmr , 8148 of lncs , 4253 .
springer .
bruttomesso , r. , cimatti , a. , franzn , a. , griggio , a. , hanna , z. , nadel , a. , palti , a. , sebastiani , r. 2007 . a lazy and layered smt(bv )
solver for hard industrial verification problemsin damm , w. hermanns , h. , cav , 4590 of lncs , 547560 .
springer .
dao - tran , m. , eiter , t. , fink , m. , weidinger , g. , weinzierl , a. 2012 .
omiga : an open minded grounding on - the - fly answer set solverin del cerro , l. f. , herzig , a. , mengin , j. , jelia , 7519 of lncs , 480483 .
springer . , b. , denecker , m. , stuckey , p. j. 2012 .
lazy model expansion by incremental groundingin dovier , a. costa , v. s. , iclp ( technical communications ) , 17 of lipics , 201211 .
schloss dagstuhl - leibniz - zentrum fuer informatik .
denecker , m. 2000 .
extending classical logic with inductive definitionsin lloyd , j. w. , dahl , v. , furbach , u. , kerber , m. , lau , k .- k . , palamidessi , c. , pereira , l. m. , sagiv , y. , stuckey , p. j. , cl , 1861 of lncs , 703717 .
springer .
denecker , m. , lierler , y. , truszczynski , m. , vennekens , j. 2012 . a tarskian informal semantics for answer set programmingin dovier , a. costa , v. s. , iclp ( technical communications ) , 17 of lipics , 277289 .
schloss dagstuhl - leibniz - zentrum fuer informatik .
drescher , c. walsh , t. 2012 .
answer set solving with lazy nogood generationin dovier , a. costa , v. s. , iclp ( technical communications ) , 17 of lipics , 188200 .
schloss dagstuhl - leibniz - zentrum fuer informatik .
eiter , t. , ianni , g. , schindlauer , r. , tompits , h. 2005 . a uniform integration of higher - order reasoning and external evaluations in answer - set programmingin kaelbling , l. p. saffiotti , a. , ijcai , 9096 . professional book center .
gebser , m. , kaminski , r. , kaufmann , b. , ostrowski , m. , schaub , t. , thiele , s. 2008 .
engineering an incremental asp solverin garca de la banda , m. pontelli , e. , iclp , 5366 of lncs , 190205 .
springer .
marek , v. w. truszczyski , m. 1999 .
stable models and an alternative logic programming paradigmin apt , k. r. , marek , v. w. , truszczyski , m. , warren , d. s. , the logic programming paradigm : a 25-year perspective , 375398 .
springer - verlag .
marin , m. , wittocx , j. , denecker , m. , bruynooghe , m. 2008 .
: satisfiability of propositional logic extended with inductive definitionsin kleine bning , h. zhao , x. , sat , 4996 of lncs , 211224 .
springer .
, j. p. , lynce , i. , malik , s. 2009 .
conflict - driven clause learning sat solversin biere , a. , heule , m. , van maaren , h. , walsh , t. , handbook of satisfiability , 185 of frontiers in artificial intelligence and applications , 131153 .
ios press .
son , t. c. , pontelli , e. , le , t. 2014 .
two applications of the asp - prolog system : decomposable programs and multi - context systemsin flatt , m. guo , h .- f . ,
padl , 8324 of lecture notes in computer science , 87103 . springer .
|
finding satisfying assignments for the variables involved in a set of constraints can be cast as a ( bounded ) _ model generation _ problem : search for ( bounded ) models of a theory in some logic .
the state - of - the - art approach for bounded model generation for rich knowledge representation languages like and and a csp modeling language such as zinc , is _ ground - and - solve _ : reduce the theory to a ground or propositional one and apply a search algorithm to the resulting theory .
an important bottleneck is the blow - up of the size of the theory caused by the grounding phase .
_ lazily grounding _ the theory during search is a way to overcome this bottleneck .
we present a theoretical framework and an implementation in the context of the knowledge representation language . instead of grounding all parts of a theory , _ justifications _
are derived for some parts of it . given a partial assignment for the grounded part of the theory and valid justifications for the formulas of the non - grounded part , the justifications provide a recipe to construct a complete assignment that satisfies the non - grounded part .
when a justification for a particular formula becomes invalid during search , a new one is derived ; if that fails , the formula is split in a part to be grounded and a part that can be justified .
experimental results illustrate the power and generality of this approach .
|
since the pioneering work of bohm and pines@xcite , the conduction electrons in a metal have been described as a three - dimensional ( 3d ) gas of electrons in a neutralizing uniform positive charge . the dynamical linear density - response function of a 3d electron gas ( 3deg ) was evaluated by lindhard@xcite in the so - called random - phase approximation ( rpa ) , in which each electron is assumed to move in the external field plus the induced field of all electrons .
the wave - vector and frequency dependent dynamical quadratic density - response function of a 3deg has also been evaluated@xcite , by going beyond linear response theory .
the knowledge of this quantity has been proved to be of great importance in discussing the properties of electrons in a variety of 3d systems@xcite , and , in particular , in explaining the experimentally observed difference between the electronic energy losses of protons and antiprotons moving through a solid@xcite .
the suggested existence of two - dimensional electron layers in metal - insulator - semiconductor structures and on the surface of liquid helium led several years ago to a great activity in the study of a two - dimensional electron gas@xcite , where the electrons are confined to a plane and neutralized by an inert uniform rigid positive plane background .
the 2d electron system has also been considered in discussing the physics of new - class materials such as copper - oxide planes of high - temperature superconductors@xcite .
it has been found that electrons confined to a 2d layer show sometimes interesting properties not shared by a 3d electron system .
for instance , for a 2d metal the plasma frequency goes to zero in the long - wavelength limit , in contrast to the 3d situation@xcite .
stern@xcite evaluated the dynamical linear density - response function of a 2deg in the rpa , and calculated the plasmon dispersion and the asymptotic screened coulomb potential .
also , much effort has gone into studying the ground state energy@xcite and the excitation spectrum of a 2deg@xcite .
the stopping power for a fast particle moving parallel to a 2deg was first evaluated in the rpa , within linear response theory , by horing _
et al_@xcite , and the effect on this quantity of finite temperature@xcite , local field corrections@xcite and recoil@xcite has also been considered .
nonlinear calculations of the stopping power for slow protons and antiprotons have been performed only very recently@xcite on the basis of a scattering theory approach , the scattering cross sections being calculated for a statically screened potential . in this paper we present results for the dynamical quadratic electronic density - response of a 2deg to longitudinal external fields of arbitrary wave - vector and frequency , which we evaluate on the same level of approximation as the rpa linear density - response function of stern@xcite .
in order to illustrate the usefulness of the knowledge of this quantity , we consider , as an example , the stopping power of a 2deg , and we use the quadratic density - response function to evaluate the @xmath0 nonlinear correction to the stopping power of a 2deg for particles of charge @xmath2 moving at a fixed distance from the 2d plasma .
we consider a uniform electron system of density @xmath3 at zero temperature .
linear and quadratic density - response functions of this system , @xmath4 and @xmath5 , with @xmath6 , may be introduced in connection with the response to the presence of a time - varying external influence , after an expansion of the induced electron density in powers of the external potential @xmath7 . according to time - dependent perturbation theory , the induced electron density is given , up to second order in @xmath7 , by ( we use atomic units throughout , i.e. , @xmath8 ) @xmath9 where @xmath10|\psi_0 > \label{1}\ ] ] and @xmath11 , \tilde\rho_h(x'')\right]|\psi_0>/2\cr + & & \theta(t - t'')\theta(t''-t')<\psi_0|\left[\left[\tilde\rho_h(x),\tilde\rho_h(x '' ) \right],\tilde\rho_h(x')\right]|\psi_0>/2 . \label{1}\end{aligned}\ ] ] here @xmath12 denotes the normalized ground state ,
@xmath13 is the density fluctuation operator @xmath14 , where @xmath15 is the exact heisenberg density operator in the unperturbed system , and @xmath16 is the heaviside step function accounting for causality . in a self - consistent field or random - phase approximation ( rpa ) , it is assumed that the electron density induced by an external potential can be replaced by the electron density induced in a non - interacting electron gas by the sum of the external potential , @xmath7 , and the potential created by the induced electron density itself , @xmath17 .
hence , in this approximation : @xmath18\cr + & & \int{\rm d}{x}'\int{\rm d}{x } '' y^0(x , x',x'')\left[v(x')+v^{ind}(x')\right]\left[v(x'')+v^{ind}(x'')\right].\end{aligned}\ ] ] here @xmath19 and @xmath20 are free - particle linear and quadratic density - response functions , and @xmath21 where @xmath22 represents the instantaneous coulomb interaction . introducing eq .
( 2.5 ) into eq . ( 2.4 ) and keeping in eq . (
2.4 ) only terms up to second order in @xmath7 , one finds the following integral equations for rpa linear and quadratic density - response functions : @xmath23 and @xmath24 where @xmath25 is the so - called inverse dielectric function : @xmath26 the integral equations ( 2.6 ) and ( 2.7 ) have , within many - body perturbation theory , the simple diagrammatic interpretation shown in fig .
1 , where empty and full bubbles represent non - interacting and interacting linear density - response functions , @xmath27 and @xmath28 , respectively . similarly ,
empty and full triangles represent non - interacting and interacting quadratic density - response functions , @xmath29 and @xmath30 , respectively , and dashed lines represent the coulomb interaction , @xmath22 .
thus , rpa linear and quadratic density - response functions are represented diagrammatically by summing over the infinite set of diagrams containing one ( see fig .
1a ) and three ( see fig .
1b ) strings of empty bubbles , respectively . in the case of a homogeneous 2deg ,
the electrons are free to move in two spatial dimensions , having their motion constrained in the third dimension .
thus , assuming time invariance , we define the fourier transforms @xmath31}\chi({\bf r}_1,t_1;{\bf r}_2,t_2)\ ] ] and @xmath32 } { \rm e}^{-{\rm i}\left[({\bf q}_1+{\bf q}_2)\cdot({\bf r}_2-{\bf r}_3)-(\omega_1+\omega_2)(t_2-t_3)\right]}\cr & & \times y({\bf r}_1,t_1;{\bf r}_2,t_2;{\bf r}_3,t_3),\end{aligned}\ ] ] where @xmath33 , @xmath34 and @xmath35 represent two - dimensional position - vectors in the 2d plane , and @xmath36 is the trimomentum @xmath37 .
hence , within the rpa we find : @xmath38 and @xmath39 where @xmath40 and @xmath41 . for a non - interacting fermi gas ,
the ground state is obtained by filling all the plane wave states inside the fermi sphere of radius @xmath42 , @xmath43 being the average interelectronic distance ( @xmath44 ) . as in the case of a 3deg@xcite
, we find non - interacting linear and quadratic density - response functions to be@xcite @xmath45 \label{34}\ ] ] and @xmath46 , \label{38 } \end{aligned}\ ] ] where @xmath47 , @xmath48 , and @xmath49 is a positive infinitesimal .
analytical evaluation of eq .
( 2.12 ) results in the non - interacting linear density - response function of stern@xcite . as for the non - interacting quadratic density - response function , we first sum occupation numbers in eq .
( 2.15 ) to find @xmath50 . \label{39}\end{aligned}\ ] ] for the real part , we find@xcite
: @xmath51=-\left[(i_{q , q_1}+i_{q , q_1}')+(i_{-q ,- q+q_1}+i_{-q ,- q+q_1}')+ ( i_{-q_1,q - q_1}-i_{-q_1,q - q_1}')\right],\label{9}\ ] ] where @xmath52 -{\rm sgn}a\arctan\left[{\sin\chi\sqrt{a^2-q^2_f}\over a_1-a\cos \chi}\right]\theta ( a^2-q^2_f)\cr & + & \arctan\left[{a_1\sin\chi\over a - a_1\cos\chi}\right ] -{\rm sgn}a_1\arctan\left[{\sin\chi\sqrt{a_1 ^ 2-q^2_f}\over a - a_1\cos \chi}\right]\theta ( a^2-q^2_f)\}\label{10}\end{aligned}\ ] ] and @xmath53 here @xmath54 , @xmath55 , and @xmath56 , @xmath57 being the angle between @xmath58 and @xmath59 . as for the imaginary part of @xmath60 we define@xcite the function @xmath61 , which can be represented in terms of a sum over hole and particle states and
gives the second order contribution to the so - called absorption probability , as demonstrated in ref .. we find , @xmath62=h_{q , q_1}+h_{q_1,q}+h_{q - q_1,-q_1},\label{14}\ ] ] where@xcite @xmath63\label{17}\ ] ] and @xmath64 in particular , at low frequencies an expansion of @xmath61 in powers of the frequency @xmath65 gives , after retaining only the first - order terms , @xmath66\sqrt{4q^2_f-|{\bf q}|^2}}\,q^0.\label{20}\ ] ] in the static limit ( @xmath67 ) both first and second order contributions to the absorption probability , @xmath68 and @xmath61 , are proportional to the frequency @xmath65 , as in the case of a 3deg .
we consider an ion of charge @xmath1 moving with constant velocity @xmath69 at a fixed distance @xmath70 from a 2deg of density @xmath3 .
the coulomb potential of this moving particle has the form @xmath71 where @xmath72 represents , as in eqs .
( 2.9 ) and ( 2.10 ) , a two - dimensional position - vector in the 2d plane , and @xmath73 denotes the coordinate normal to the 2deg which we consider to be located at @xmath74 .
hence , in order to obtain the induced potential , we introduce this time - varying external potential into eq .
( 2.1 ) , and eq . ( 2.1 ) into eq . ( 2.5 ) .
we note that density - response functions of a 2deg have their @xmath73 arguments localized to the 2d plane by positional @xmath75 functions , we fourier transform , and find , up to second order in @xmath1 : @xmath76}v_{{\bf q}}y_{{\bf q},\omega;-{\bf q}_1,-\omega_1}v_{{\bf q}_1}v_{{\bf q}-{\bf q}_1 } , \label{60}\end{aligned}\ ] ] where @xmath77 and @xmath78 .
the stopping power of the 2deg is simply the retarding force that the polarization charge distribution in the vicinity of the projectile exerts on the projectile itself@xcite , and is given by @xmath79 substituting eq .
( 3.2 ) into eq . ( 3.3 ) , we have@xcite @xmath80v_{\bf q}\cr -&&{z_1 ^ 3\over v}\int{{\rm d}^2{\bf q}\over(2\pi)^2}\omega\int{{\rm d}^2{\bf q}_1\over(2\pi)^2}{\rm e}^{-(|{\bf q}|+|{\bf q}_1|+|{\bf q}-{\bf q}_1|)h}v_{{\bf q}}{\rm im}\left[y_{{\bf q},\omega;-{\bf q}_1,-\omega_1}\right]v_{{\bf q}_1}v_{{\bf q}-{\bf q}_1}. \label{60}\end{aligned}\ ] ] in the rpa , @xmath81 and @xmath82 are obtained from eqs .
( 2.11 ) and ( 2.12 ) or , equivalently , from @xmath83 and @xmath84 where @xmath85 the linear contribution to the stopping power of eq . ( 3.4 ) , which is proportional to @xmath86 , was evaluated in the rpa by horing _
et al_@xcite at @xmath87 , and similar calculations were presented by bret and deutsch@xcite at finite temperature . on the other hand , the quadratic contribution to the stopping power of eq .
( 3.4 ) is , in the rpa and for a geometry with the ion - beam in - plane within the 2d electron layer ( @xmath88 ) , equivalent to the result derived in ref .
, within many - body perturbation theory , as the energy loss per unit path length of the projectile , the integration space being now in two dimensions instead of three dimensions .
this contribution to the stopping power , which is proportional to @xmath0 , discriminates between the energy loss of a proton and that of an antiproton , and appears as a consequence of losses to one- and two - step electronic excitations generated by both linearly and quadratically screened ion potentials , as discussed in ref .. in the case of slow intruders ( @xmath89 ) , only the low - frequency form of the response enters in the evaluation of eq .
( 3.4 ) . at zero frequencies
both linear and quadratic density - response functions are real , @xmath90 $ ] and @xmath61 being at low frequencies proportional to the frequency @xmath65 .
thus , retaining only the first - order terms in the frequencies , both @xmath86 and @xmath0 contributions to the stopping power are found to be proportional to the velocity of the projectile , as in a 3deg . for the rpa quadratic contribution to the stopping power we find , after insertion of eq .
( 2.23 ) into eq . (
3.4 ) , the following result : @xmath91^{(3)}=8 v z_1 ^ 3 \int_0^\infty{{\rm d}|{\bf q}|\over\sqrt{4q^2_f-|{\bf q}|^2}}\int_0^\infty{\rm d}|{\bf q}_1|\int^{\pi}_0 { \rm d}\chi{\rm e}^{-(|{\bf q}|+|{\bf q}_1|+|{\bf q}-{\bf q}_1|)h}{f^l_1+f^l_2\over|{\bf q}-{\bf q}_1|},\label{21}\ ] ] where @xmath92 and @xmath93
in the low - velocity limit ( @xmath89 ) , the stopping power can be evaluated to all orders in @xmath1 , on the basis of a free - electron picture , with the additional assumption of independent , individual , elastic electron scattering , and it is easily found to be , in both two and three dimensions , proportional to the velocity of the projectile@xcite . by using density - functional theory ( dft)@xcite to calculate the self - consistent potential generated by a static charge submerged in a 3deg , echenique _
et al_@xcite evaluated the nonlinear stopping power of a 3deg for slow ions .
nonlinear calculations for an in - plane projectile in 2d have been performed recently by using a nonlinearly screened scattering potential based on the sjlander - stott theory@xcite , with the use of a nonlinear version of the linearized fermi - thomas potential@xcite in which the screening constant is determined from the friedel sum rule , and on the basis of a self - consistent potential@xcite obtained within dft as in ref . for the 3d case .
the nonlinear calculations presented in ref . for the energy - loss of slow protons and antiprotons are very close to the calculations of ref . when exchange and correlation ( xc ) contributions to the dft scattering potential are excluded .
these calculations are correct to all orders in @xmath1 , and they represent , therefore , a good check for our quadratic response calculations , which should be exact in the high - density and/or low @xmath1 limits . nevertheless , the approaches of refs . and have the limitation of being restricted to very low ion velocities ( @xmath94 , @xmath95 being the fermi velocity ) , while our nonlinear @xmath0 corrections are valid for arbitrary non - relativistic velocities .
the low - velocity limit of the quadratic stopping power , as obtained from eq .
( 3.8 ) for an in - plane ( @xmath88 ) moving charge and divided by the velocity , is represented in fig . 2 by a solid line , as a function of the electron - density parameter @xmath43 .
stars and squares represent the full nonlinear contribution to the stopping power for antiprotons reported in refs . and , respectively , multiplied by a factor of @xmath96 , showing an excellent agreement , in the high - density limit , with our @xmath0 nonlinear contribution . for @xmath97 ,
higher order corrections become important , and for @xmath98 , @xmath86 contributions ( dashed line ) are smaller than the @xmath0 correction , indicating that the external potential can not be treated , for these electron densities , as a small perturbation . in the case of a projectile moving at a given @xmath70 distance above the 2d plasma the external perturbation is obviously diminished and , in particular , for @xmath99 the quadratic stopping power is smaller than the linear one for all electron densities ( see the inset of fig .
the full nonlinear contribution to the stopping power for protons reported in refs . and for @xmath88 is also represented in fig . 2 by circles and crosses , respectively , showing large differences with our @xmath0 calculations for all electron densities .
these differences appear as a consequence of perturbation theory failing to describe electronic states bound to the proton , which in 2d systems can be supported by arbitrarily weak attractive potentials@xcite .
as the velocity of the projectile ( @xmath100 ) and/or the distance from the 2d plasma ( @xmath70 ) increase , the ion potential becomes a relatively smaller perturbation and the @xmath0 contribution to the stopping power for antiprotons may , therefore , be expected to approximately describe the full nonlinear contribution to the stopping power for arbitrary values of @xmath100 and @xmath70 , as long as @xmath101 , and also for lower densities ( @xmath97 ) as the velocity and the @xmath70 distance increase .
substitution of the full rpa linear and quadratic 2deg density - response functions of eqs .
( 3.5 ) and ( 3.6 ) into eq . ( 3.4 ) results in the quadratic stopping power plotted by a solid line in fig .
3 , as a function of the impinging projectile velocity , for @xmath102 and @xmath88 , and also for @xmath102 and @xmath99 ( see the inset of fig .
3 ) . the dotted line represents the low - velocity limit , as obtained from eq .
( 3.8 ) , and dashed lines represent linear rpa contributions to the stopping power of eq .
( 3.4 ) .
the quadratic contribution to the stopping power of a 2deg presents properties not shared by the 3deg .
first , the range of validity of the linear velocity dependence of this contribution to the stopping power ( see fig .
3 ) persists only up to velocities much smaller than the fermi velocity , in contrast to the 3d situation in which the linear velocity dependence persists up to velocities approaching the fermi velocity@xcite .
second , at velocities around the plasmon threshold velocity , for which the projectile has enough energy to excite a plasmon@xcite , the ratio between @xmath0 and @xmath86 contributions to the stopping power increases , again in contrast to the 3d situation .
furthermore , we have found that the increase of this ratio at the plasmon threshold velocity becomes dramatic as the electron density decreases .
of course , the ratio between quadratic and linear contributions to the stopping power decreases very rapidly with @xmath70 ( see the inset of fig .
3 ) , but the relative , and for large @xmath43 dramatic , increase of this ratio at the plasmon threshold velocity persists for all values of @xmath70 that are not much larger than @xmath103 . for a 2deg the group velocity of the plasmon wave nearly coincides with the plasmon threshold velocity of the projectile ( see fig .
4 ) ; also , the inclusion of short - range correlations , which are ignored in the rpa , is known to have a substantial effect on the plasmon dispersion@xcite .
we interpret the anomalous enhancement of the ratio between quadratic and linear contributions to the stopping power at the plasmon threshold velocity and small electron densities ( @xmath104 ) as a result of neglecting , within the rpa , short - range correlations between the electrons of a 2d system , since these correlations are non - negligible for all values of @xmath70 as long as @xmath43 is not small .
we have presented an analytical evaluation of the wave - vector and frequency dependent non - interacting quadratic density - response function of a 2deg .
we have used this result to find , within a self - consistent field approximation , the @xmath0 correction to the stopping power of a 2deg for charged recoiless particles moving at a fixed distance from the 2d plasma sheet .
we have reproduced , in the high - density limit , previous full nonlinear calculations of the stopping power of a 2deg for slow antiprotons , and we have gone further to calculate the @xmath0 correction to the stopping power for a wide range of projectile velocities . we have found that the @xmath0 contribution to the stopping power of a 2deg presents properties not shared by the 3deg . on the one hand , the range of validity of the linear velocity dependence of the @xmath0 contribution to the stopping power persists only up to velocities much smaller than the fermi velocity , and , on the other hand , an anomalous enhancement of the ratio between quadratic and linear contributions to the stopping power at the plasmon threshold velocity is found , within the rpa , at small electron densities ( @xmath104 ) . also , our results indicate that linear response calculations are , for all projectile velocities , less reliable in 2d than in 3d .
d. bohm and d. pines , phys . rev . * 92 * , 609 ( 1953 ) .
j. lindhard , k. dan .
* 28 * ( 8) , 1 ( 1954 ) .
r. cenni and p. saracco , nucl .
a * 487 * , 279 ( 1988 ) . c. f.
richardson and n. w. ashcroft , phys .
b * 50 * , 8170 ( 1994 ) ; j. m. rommel and g. kalman , phys . rev .
b * 54 * , 3518 ( 1996 ) . j. m. pitarke , r. h. ritchie , p. m. echenique , and e. zaremba , europhys . lett . * 24 * , 613 ( 1993 ) ; j. m. pitarke , r. h. ritchie , and p. m. echenique , nucl .
methods b * 24 * , 613 ( 1993 ) .
j. m. pitarke , r. h. ritchie , and p. m. echenique , phys .
b * 52 * , 13882 ( 1995 ) .
t. ando , a. fowler , and f. stern , rev .
phys . * 54 * , 437 ( 1982 ) . j. r. engelbrecht and m. randeira , phys .
lett . * 65 * , 1032 ( 1990 ) ; p. w. anderson , phys .
lett . * 64 * , 1839 ( 1990 ) .
r. h. ritchie , phys . rev . * 106 * , 874 ( 1957 ) ; r. h. ritchie , surf .
sci . * 34 * , 1 ( 1973 ) . f. stern , phys .
. lett . * 18 * , 546 ( 1967 ) .
a. k. rajagopal and j. c. kimball , phys .
b * 15 * , 2819 ( 1977 ) .
a. v. chaplik , zh .
fiz . * 60 * , 1845 ( 1971 ) ; g.f .
giuliani and j.j .
quinn , phys . rev . *
b26 * , 4421 ( 1982 ) ; l. zheng and s. d. sarma , phys . rev .
b * 53 * , 9964 ( 1996 ) .
n. j. m. horing , h. c. tso , and g. gumbs , phys . rev . * b36 * , 1588 ( 1987 ) .
a. bret and c. deutsch , phys .
* e48 * , 2994 ( 1993 ) .
wang and t .- c .
ma , phys .
a * 200 * , 319 ( 1995 ) .
a. bergara , i. nagy , and p. m. echenique , phys .
b * 55 * , 12864 ( 1997 ) . a. krakovsky and j.k .
percus , phys .
b * 52 * , r2305 ( 1995 ) .
n . wang and t .- c .
ma , phys .
a * 55 * , 2087 ( 1997 ) . c. d. hu and e. zaremba , phys .
b * 37 * , 9268 ( 1988 ) ; a. bergara , i. campillo , j. m. pitarke , and p. m. echenique , phys .
b * 56 * , 15654 ( 1997 ) . differences between the retarded density - response functions of eqs . (
2.14 ) and ( 2.15 ) , on the one hand , and time - ordered response functions , as defined within many - body theory ( see , e.g. , g. d. mahan , _ many - particle physics _
( plenum , new york , 1981 ) ) , on the other hand , lie in @xmath105 , which should be replaced in eqs .
( 2.14 ) and ( 2.15 ) by @xmath106 [ @xmath107 being either @xmath65 , @xmath108 or ( @xmath109 ) ] in order to obtain their time - ordered counterparts . @xmath110 , @xmath111 and @xmath112 in eq . ( 2.17 ) should be multiplied by @xmath113 , @xmath114 and @xmath115 , respectively , to obtain the real part of the corresponding time - ordered quadratic response function .
similarly , the right - hand side of eq .
( 2.21 ) should be multiplied by @xmath116 in order to obtain the imaginary part of the time - ordered quadratic response function .
m. echenique , f. flores , and r. h. ritchie , solid state phys . * 43 * , 229 ( 1990 ) .
the real parts of @xmath81 and @xmath82 give no contribution to the integrals of eq .
( 3.4 ) , since @xmath117={\rm re}\left[\chi_{-q}\right]$ ] and @xmath118={\rm re}\left[y_{-q ,- q_1}\right]$ ] .
i. nagy , phys . rev . *
b51 * , 77 ( 1995 ) .
p. hohenberg and w. kohn , phys . rev .
* 136 * , b864 ( 1964 ) ; w. kohn and l. j. sham , _ ibid . _ * 140 * , a1133 ( 1965 ) .
p. m. echenique , r. m. nieminen , and r. h. ritchie , solid state commun .
* 37 * , 779 ( 1981 ) ; p. m. echenique , r. m. nieminen , j. c. ashley , and r. h. ritchie , phys .
a * 33 * , 897 ( 1986 ) .
e. zaremba , p. m. echenique , and i. nagy , to be published .
e. n. economou , _
green s functions in quantum physics _ ( springer - verlag , berlin , 1990 ) , pg
. 64 . at long wavelengths
the plasma frequency in the 2deg varies like @xmath119 , in contrast to the 3d case ( see , e. g. , a. l. fetter , ann .
( n.y . ) * 81 * , 367 ( 1973 ) ) . at short wavelengths ,
the group velocity of the 2d plasmon is nearly constant and approaches the plasmon threshold velocity , @xmath120 . for @xmath102 , @xmath121 ( @xmath122
is bohr s velocity ) .
a. gold and l. calmels , phys .
rev . * b48 * , 11622 ( 1993 ) .
|
the electronic response of a two - dimensional ( 2d ) electron system represents a key quantity in discussing one - electron properties of electrons in semiconductor heterojunctions , on the surface of liquid helium and in copper - oxide planes of high - temperature superconductors .
we here report an evaluation of the wave - vector and frequency dependent dynamical quadratic density - response function of a 2d electron gas ( 2deg ) , within a self - consistent field approximation .
we use this result to find the @xmath0 correction to the stopping power of a 2deg for charged particles moving at a fixed distance from the plane of the 2d sheet , @xmath1 being the projectile charge .
we reproduce , in the high - density limit , previous full nonlinear calculations of the stopping power of a 2deg for slow antiprotons , and we go further to calculate the @xmath0 correction to the stopping power of a 2deg for a wide range of projectile velocities .
our results indicate that linear response calculations are , for all projectile velocities , less reliable in two dimensions than in three dimensions .
|
star formation in the early universe has played a critical role in subsequent evolution of the universe .
first - generation stars may have substantially contributed to the cosmic reionization and metal pollution of the universe by their radiative / kinetic feedback effects .
these events are important for the formation and evolution of protogalaxies .
the studies of the initial collapse of primordial pre - galactic objects in the @xmath0cdm universe have been done in both analytical @xcite and numerical method @xcite .
these objects are formed with masses of the order of @xmath7 at redshifts @xmath8 .
theoretical studies suggested that in these low - mass halos , @xmath9 molecules are formed up to the level of @xmath10 , because the virial temperature and the central density of these halos are high enough to activate h@xmath4 formation @xcite so that the cooling time becomes short enough .
this small fraction of h@xmath4 is sufficient to cool the gas , which leads to the formation of first - generation stars in these halos .
early studies on the formation of primordial stars have been done in almost one - zone approximation @xcite . on the other hand , after 1990s , there have been a number of numerical studies of the formation of primordial stars .
these studies consistently suggested that first - generation stars are very massive ( @xmath11 ) . because of their extreme mass scale ,
first - generation stars emit copious amount of ionizing radiation , as well as a strong flux of @xmath9-dissociating lyman - werner ( lw ) band radiation .
therefore , the radiation from the first stars dramatically influences their surroundings , heating and ionizing the gas within a few kiloparsecs around the progenitor star .
if the primeval star is very massive ( @xmath12 ) , the ionization front from the star will break out of parent halos up to @xmath13 in mass .
half of the baryons in the halo will be swept up into a dense shell that grows to the virial radius of the halo ( @xmath14 ) by the end of the life of the star .
therefore , the gas density within the shell radius will drastically decrease to low uniform densities of @xmath15 prior to the supernova ( sn ) explosion @xcite .
the dark halos are also affected by the radiation from the first star in the neighboring halos .
the local radiative feedback effect is sensitive to the distance from the source star , as well as the mass and the evolutionary stage of the target halo .
the ionizing radiation from the star totally or partially photoevaporates them before the sn blast ever reaches them @xcite .
therefore , these halos will be less massive or be gone altogether by the time the sn shock strikes them .
it is also should be noted that the lw band radiation from the source star photodissociates h@xmath4 molecules in these halos in case the halos are located close enough to the source @xcite .
some of massive primordial stars end their lives as energetic sne .
@xcite suggested that the progenitor star of @xmath16 dies in type ii sn blasts while those of @xmath17 dies in pair instability sn . on the other hand , it is important to note that some pop iii stars lying in between the ranges indicated above may also explode as hypernovae @xcite .
the studies of first sn explosions have been performed in both sph @xcite and grid codes @xcite . in less massive halos ( @xmath18 ) ,
the ejecta first expands into a nearly uniform rarefied ionized medium , then interacts violently with the dense shell swept up in the progenitor h ii region .
part of the energy of the blast wave is reflected into the center as a reverse shock and the rest pushes the shell forward @xcite . in massive parent halos ( @xmath19 ) , however , the sn remnant ( snr ) will only expand @xmath20 into the halo and then recollapse @xcite , never reaching any nearby halos because the _ i_-front does not break out these parent halos prior to the sn explosion .
the shock wave from an sn that break away from the host halo impact and go through the neighboring halos those survived photoevaporation process .
the shock wave may blow off the gas component ; however , it also promote the formation of @xmath9 and hd molecules in the gas because of the enhanced electron abundance by the shock heating @xcite .
we also note that the relic h ii region including the photoevaporated halos recombines out of equilibrium , which is basically same physical condition of the postshock enhancement of h@xmath4/hd formation .
the sn shock can sweep up this h@xmath4/hd and carry it into the nearby halo .
therefore , these molecules in the fossil h ii region are important for the formation of secondary stars as well as those would be formed in the sn shock by collisional ionization .
star formation could be triggered by snrs in several ways .
primordial snrs may sweep up a dense shell of ambient gas or collide with the dense shell swept up by an h ii region , either of which becomes contaminated by metals in the ejecta , subsequently comes to be dominated by its self - gravity . as a result
, the shell is expected to fragment into smaller filaments / cores where populations of less massive stars are formed @xcite . on the other hand , in rather massive halos ( @xmath21 ) , the h ii region generated by the progenitor star is confined well inside the virial radius and the gas is kept neutral . in this case , first snrs expand in neutral halos , heavily mix their interiors with heavy elements , and then recollapse without escaping the halo @xcite .
the interaction of the first sn with neighboring halos in the early universe has not been investigated in detail so far .
recently , @xcite performed a cosmological simulation of first sn explosion .
they found that the shock from the sn can accelerate the star formation process in neighboring rather massive halos in which stars could be formed without feedback effects .
they also indicated that if the sn explosion occurs in an h ii region , the snr comes to pressure equilibrium at half of the radius of the relic h ii region because of its relatively large temperatures .
this limits the reach of the sn explosion to @xmath22 @xcite .
in addition , @xcite performed numerical simulation of snrs ram - pressure stripping cosmological halos .
they focus on how heavy elements from the remnant are mixed with the halo gas .
however , the number of halos studied by numerical simulations is restricted .
more analytical criteria are needed in order to obtain a systematic understanding of first sn feedback effects on neighboring dark matter halos . in this paper , we investigate feedback effects by first sn onto the nearby halos in the early universe , which is potentially important for the total star formation activities in the early universe .
we use analytic arguments in order to obtain the universal criteria for feedback effects by first sn on the neighboring dark halos .
the outline of this paper is as follows .
the initial setup and the description of snr evolution are given in section 2 . in section 3 and 4 ,
the sn feedback effects are listed and quantified .
section 5 and 6 are devoted to discussion and summary .
we consider sn explosions in the first collapsed halos ( @xmath23 ) .
the gravitational potential of these halos are so shallow that the ionizing radiation from the progenitor first stars can sweep out the gas of the host halos @xcite . as a result
, the subsequent sne can easily break away from the halo because of the decreased gas density by photoionization prior to the explosion @xcite . in order to understand the nature of the sn shock expanding into an essentially uniform and ionized intergalactic space ,
we assume spherical symmetry and initially uniform ambient gas density of averaged cosmological density , @xmath24 . strictly speaking
, this assumption is not valid since the gas density of the halo still slightly higher than @xmath24 even after the feedback of uv radiation .
this assumption should have some effects to increase the energy of the sn shock , but we use this for simplicity .
the energy loss mechanism from the snr is dominated by bremsstrahlung for a first few years , then line emission , and then inverse compton scattering becomes important @xcite .
we take into account all of the cooling rate stated above as well as @xmath9 cooling at low temperature .
we use the fitting formula for these rates from the compilation by @xcite and @xcite , respectively .
we consider the neighboring dark matter halos with total mass of @xmath25 , whose baryonic fraction in mass is assumed to be the cosmic mean value , @xmath26 .
we have another free parameter in our calculations , which is the distance from the sn center to the target halo .
additionally , we use two typical fixed redshift when the sn explode ( @xmath27 ) and when the nearby halos virialized ( @xmath28 ) .
we tested this issue with three different sn explosion energy @xmath29 , and @xmath30 . throughout the paper
, we work with the @xmath0cdm universe with @xmath31 , @xmath32 , @xmath33 , and @xmath34 .
we introduce three timescales @xmath35 , @xmath36 , and @xmath37 , that characterize the important physical processes .
they represent the sound crossing timescale of the gas component in the neighboring halo , the cooling timescale , and the free - fall timescale , respectively . the sound crossing timescale of the gas component in the neighboring halo is described as @xmath38 where @xmath39 denotes the length that we are interested and @xmath40 is the sound speed of the gas , respectively .
this timescale is also interpreted as the expansion timescale of the shock heated gas if we substitute the size of the shock - heated region for @xmath39 and use the sound speed of shocked gas as @xmath40 .
the cooling timescale is defined as @xmath41 where @xmath42 and @xmath43 are the temperature and the number density , @xmath44 and @xmath45 denote the boltzmann constant and the ratio of specific heats , respectively .
@xmath46 represents the fraction of @xmath9 , and @xmath0 ( @xmath47 ) denotes the cooling rate of gas .
the free - fall timescale is written as @xmath48 here , @xmath49 is the gravitational constant and @xmath50 is the virial density given by @xmath51 , where @xmath52 g @xmath53 .
the time evolution of an snr in intergalactic medium ( igm ) is mainly described by following four stages : ( 1 ) : : _ the free - expansion stage_. the free - expansion stage lasts until the sn ejecta sweeps up roughly the same amount of mass as their own in the surrounding medium . in this stage ,
the velocity of the sn ejecta decreases linearly with radius @xcite .
( 2 ) : : _ sedov taylor adiabatic expansion stage_. the expansion of the shock front is well approximated by the sedov
taylor solution .
( 3 ) : : _ pressure - driven expansion stage_. after the postshock gas is cooled via the radiative cooling , the snr depart from an adiabatic expansion . geometrically thin shell is formed just behind the shock front .
the expansion of the shocked shell is driven by the high pressure of the hot cavity .
( 4 ) : : _ momentum - driven expansion stage_. in this stage , the shocked shell expands conserving its momentum . in our calculation , we do not take into consideration the stage ( 1 ) because the duration of this stage is very short and it hardly affects the entire result .
we also do not consider the stage ( 4 ) in igm since the shock velocity at this stage is too small to activate h@xmath4 molecule formation or to evacuate the gas from the halo .
we consider the scenario , in which the snr gradually sweeps up mass and then collides with the neighbor halo .
in fact , the remnant first collides violently with the dense shell swept up in the progenitor h ii region , but we leave this effect for future works , since introducing such effect makes the analysis complicated .
the expansion of the shock front in the sedov
taylor adiabatic stage is described by the self - similar solution @xcite . the radius and the expansion velocity of the shock front
is written as @xmath54 @xmath55 where @xmath56 and @xmath57 represent the sn explosion energy and the ambient density , respectively .
the postshock temperature is derived from the sedov
taylor solution and the rankine
hugoniot relation as follows : @xmath58 where @xmath59 and @xmath60 denote the atomic mass unit and the mean molecular weight , respectively .
these solutions are based upon the assumption , that the gas is adiabatic , however , the postshock gas forms a dense shell which is cooled by the radiative cooling subsequently . after the cooling timescale of the shell becomes shorter than the expansion timescale of the shell ( @xmath61 ) , the snr move on to the pressure - driven expansion stage .
the shock front expands by the high pressure in the hot cavity .
we assume that the pressure inside the cavity decreases adiabatically as @xmath62 where @xmath63 denotes the pressure inside the cavity , and also @xmath64 and @xmath65 represent the radius of the shock front and the postshock pressure at the beginning of the pressure - driven expansion stage . combining this equation with the equation of motion of the shell
, we have the shock radius / velocity as follows @xcite .
@xmath66 we also obtain the postshock temperature as follows : @xmath67 in the present paper , we do not take into account the cooling rate due to heavy elements .
in fact , the shock - heated gas in the nearby halos is expected to be polluted by metals ejected from the sn , although the abundance of heavy elements in the shock - heated gas is hard to evaluate .
the ultra high resolution simulations would be indispensable to assess the degree of metal mixing . aside from such difficulty
, we can evaluate the metal abundance assuming complete mixing between the sn ejecta and the surrounding material .
based upon such assumption , the mean metallicity of the swept up mass by the sn shock is @xmath68 @xcite .
in addition , the metallicity of the high-_z _ igm observed by ly@xmath69 absorption systems is at a level of @xmath70 to @xmath71 @xcite .
thus , @xmath70 to @xmath71 could be a rough standard to assess the effects of metal cooling . at such low metallicity , the radiative cooling rate is hardly affected above @xmath72k @xcite , since it is dominated by h and he cooling .
the cooling rate below @xmath72 k is basically proportional to the amount of metals @xcite , however , again at such low @xmath73 , the metal cooling rate is comparable to h@xmath4 cooling , as long as @xmath74 @xcite .
therefore , radiative cooling by heavy elements in these halos do not play central role as long as complete mixing is assumed . on the other hand , metals in the interior of expanding snr
will mix with the shocked shell it sweeps up because of rayleigh taylor and kelvin helmholtz instabilities .
this will enrich the shocked region to much higher metallicities than @xmath75 , radiatively cool the shell much faster than any h@xmath4 it has formed or swept up , and then fragment clumps .
it may be a clump that collides with the neighbor halo , not an intact shell , but this is for future numerical studies to determine . before the shocked gas cools by radiative cooling .
after that , the shock propagates under the momentum conservation .
the intermediate case ( case(ii ) ) indicates that the shock shifts to the momentum - driven expansion phase in the halo . ]
in this section , we assess the effect of the sn shock onto neighboring dark halos by analytic arguments comparing timescales . in case
a neighboring halo is located very close to the center of the sn explosion , the gas in the halo would be evacuated by the shock momentum . as a result ,
subsequent star formation in the halo should be inhibited . assuming the momentum conservation
after the gas behind the shock cools , we can roughly evaluate the velocity of the gas in the halo after the impact of the shocked shell . in figure [
fig : fig1 ] , we represent three cases of gas evacuation by shock momentum .
first , we consider the simple momentum conserving case ( case(i ) in figure [ fig : fig1 ] ) , where @xmath76 ( i.e. , @xmath77 ) is satisfied . here , @xmath78 denotes the distance that the shock sweeps the halo before it enters the momentum conserving phase , i.e. , @xmath79 here , @xmath80 is the radius of the dark halo .
the equation of the momentum conservation is given as follows : @xmath81 where @xmath82 is the mass of the dark halo and @xmath83 is the mass of the gas shell colliding with the halo , which is given as @xmath84 here , @xmath85 is the mass of progenitor popiii star and we assume that the whole mass of this popiii star is released as ejecta .
we assume @xmath86 in the present calculation . on the other hand , in case of @xmath87 ( i.e. , @xmath88 , case(ii ) in figure [
fig : fig1 ] ) , the gas in the halo is cooled while the shock propagates in the halo by radiative cooling , and it evolves into the momentum conserving phase . since the gas pressure of the halo behind the shock front can keep pushing the shock before it cools , @xmath89 can not be estimated by simple momentum conservation equation .
it is difficult to assess this effect analytically .
thus , we assume that the shock velocity is not affected by the halo while the radiative cooling is still inefficient even after the shock enters the halo .
this assumption could be oversimplification , since even adiabatic shock will slow down to some extent when it enters the dense region . in order to quantify this effect ,
numerical study would be necessary , which is beyond the scope of present study .
therefore , we have to keep in mind that the shock heating effect is maximally taken into consideration in this model . in this intermediate case ( ( ii ) in figure [ fig : fig1 ] ) , we assume that the momentum conservation after shocked gas in the halo is cooled : @xmath90 in case the shock propagate through the entire halo within the cooling time of the shocked gas , that is the case of @xmath91 , the shock velocity in the halo is as large as @xmath92 during the shock propagation because of the inefficient cooling ( case ( iii ) in figure [ fig : fig1 ] )
. the velocity of the shocked gas shell , @xmath89 , is obtained from the equation ( [ eq : mon_cos ] ) as follows : @xmath93 this equation includes all of the three cases .
the cases of @xmath94 and @xmath91 are the limits of efficient / inefficient cooling . )
is plotted as a function of the distance from the sn center , @xmath95 , for @xmath96 and @xmath97 .
the dotted line represents the escape velocity of the halo with @xmath97 . ]
the gas will be still bounded in the halo even after the shock arrival , in case @xmath89 is smaller than the escape velocity @xmath98 of the halo . on the other hand ,
if @xmath99 , the gas is blown away from the halo potential , in which subsequent star formation in the halo is totally quenched .
thus , we regard the condition @xmath100 as a necessary condition for the triggered star formation in the neighboring dark halos . in figure
[ fig : fig2 ] , @xmath89 , as a function of the distance from the sn center assuming @xmath101 and @xmath102 . in this case
, @xmath89 is smaller than @xmath98 in the ranges @xmath103 and @xmath104 .
therefore , in case the neighboring halo is located at such distance , the gas in the halo is not evacuated by the shock momentum .
figure [ fig : fig3 ] shows the gas evacuation from the halos for @xmath96 .
the vertical axis denotes the mass of the dark matter halos ( @xmath82 ) , whereas the horizontal axis represents the distance from the sn center . in the hatched area denoted as @xmath100 ,
the gas in the dark halos are not lost by the shock momentum .
it is worth noting that the hatched area around @xmath105kpc corresponds to @xmath106k , where the ly@xmath69 cooling dominates the others . .
the horizontal axis shows the distance between the nearby halo and sn center , while the vertical axis denotes the mass of the halo . in the hatched area ( @xmath100 )
, the shocked gas is bounded inside the dark halo .
the unhatched region labeled as @xmath99 , the gas will be evacuated by the shock from the sn . ]
the halos that satisfies the condition @xmath100 are able to survive the disruption by the sn shock momentum . as a next step
, we consider whether those survived halos can collapse or not .
we set the collapse criteria as the condition where the gas temperature is decreased below @xmath107 , before the gas cloud expands by the thermal pressure , or bounces by adiabatic compression . in other words , if following conditions @xmath108 are both satisfied until they cool below the initial virial temperature , we regard that the gas in such halos can collapse to form stars . here , we consider the sound crossing timescale , @xmath109 , to be the expansion timescale of the shocked region . in general , the condition @xmath110 is regarded as the collapse criterion of gravitationally bound system ( e.g. , * ? ? ? * ) .
once this condition is satisfied , the system starts to collapse .
the cooling timescale in low - density gas is basically inversely proportional to the density , while the free - fall timescale proportional to the inverse of square root of density . as a result
, the ratio @xmath111 gets smaller as the collapse proceeds .
that s why the condition is regarded as the collapse criterion . in case
we consider the primordial star formation , the cooling timescale in most of the final run - away collapse phase is not proportional to the inverse of density , however , the absolute value of the cooling timescale is shorter than the other timescales outside the halo .
thus , the collapse of the cloud continues following the track along which @xmath112 is satisfied .
in addition , we have to consider the adiabatic expansion of the shocked gas , since the sound crossing timescale of the shocked gas could be very short in the present case .
thus , we need to add the inequality @xmath113 to the condition of the collapse criteria . in order to assess this condition
, we have to follow the thermal evolution of the shock heated gas in the dark halo .
each of the four panels in figure [ fig : fig4 ] shows the evolution of timescales after the shock heating for halos with @xmath114 located at @xmath115 , and @xmath116 .
thermal energy of the sn explosion is assumed as @xmath117 .
the horizontal axes denote the temperature of the gas on the way of cooling , whereas the vertical axes show timescales . in order to assess the cooling timescale for @xmath118k
, we have to take into account the h@xmath4 cooling , which is proportional to h@xmath4 fraction .
h@xmath4 fraction could be obtained by solving nonequilibrium chemical reaction equations , however , approximate values are already obtained .
we use @xmath119 in case the postshock temperature , @xmath120 , is less than @xmath72k @xcite , whereas @xmath121 is employed for @xmath122k @xcite .
if the halo with @xmath123 is located at @xmath124 ( upper - left panel ) , the shocked temperature is too high for the gas to remain the gas inside the halo potential .
in fact , expansion timescale is already shorter than the cooling timescale at @xmath125 .
thus , the gas in this halo is lost because of the shock heating . on the other hand , in the cases of @xmath126 ( upper - right panel ) , @xmath127 ( lower - left panel ) , and @xmath128 ( lower - right panel ) ,
the gas in the halo can start to cool because @xmath129 is smaller than @xmath130 at @xmath125 . among these examples
, @xmath129 is shorter than @xmath131 and @xmath130 as long as @xmath132 is satisfied in the case @xmath126 .
therefore , the gas can be cooled below @xmath107 before it expands , and the cooling phase could be followed by gravitational contraction .
in contrast , in the cases of @xmath127 and @xmath128 , the cooling process becomes inefficient before @xmath133 is achieved . in the case of @xmath134 ,
the postshock temperature exceeds @xmath72k .
thus , @xmath135 is achieved even for @xmath136k , whereas @xmath119 for @xmath128 .
therefore , the cooling condition is not simply determined by the efficiency of h@xmath4 formation , although it is necessary for the cooling below @xmath72k .
the allowed region for cooling / collapse of the shocked gas is shown in the next section .
now , we are ready to combine all conditions to find the criteria for positive / negative feedback effects by first sne on the neighboring dark halos .
three panels in figure [ fig : fig5 ] correspond to the results with @xmath137 , and @xmath138 , respectively .
extremely energetic cases with @xmath139 and @xmath138 can be interpreted as the hypernova or the pair - instability sn .
the hatched area bounded by dotted curves ( labeled as `` survive '' ) denotes the regions in which @xmath100 is satisfied .
the shaded region denoted as `` cool '' represents the condition in which the cooling rate is high enough for the gas in the halo to proceed further gravitational contraction . plane .
three panels denoted as ( a ) , ( b ) , ( c ) represent @xmath140 , and @xmath141 , respectively . in the hatched area labeled as `` survive '' ( @xmath100 ) , the shocked gas is bounded in the halo potential . on the other hand ,
in the shaded region denoted as `` cool '' , postshock gas cools below @xmath107 . ]
first of all , gas components in dark matter halos considered here ( @xmath142 ) are blown away by the shock momentum if the halos are close enough to the sn center .
the critical distance within which the gas is evacuated falls within the range @xmath6kpc depending on the sn energy and mass of the halo . in case
we assume normal core - collapse sn ( @xmath143erg ) , the critical distance is @xmath144 kpc , whereas it is @xmath145kpc for pair - instability sn ( @xmath146erg ) .
the mass dependence is not so strong , but basically the low - mass halos are more fragile than the massive halos . in addition , we also note that the boundaries of the survive regions have complicated structures at @xmath147kpc , reflecting the shape of the cooling function .
recent studies suggested that halos further than @xmath148 from the original star will not be reached by the blast . in the present paper , we include distances greater than this in our analysis , but we are doing so only for completeness .
the region where `` survive '' and `` cool '' are compatible with each other correspond to the case that the shock - heated gas successfully cooled and collapse in the dark halo .
however , it is clear that the halos which can survive the evacuation by the sn shock and collapse do not exist .
consequently , the sn feedback has basically negative effects on the star formation in surrounding halos .
the nearby dark halos come under the influence of the radiation from the sn progenitor star on the satellite halos prior to the explosion .
thus , many low - mass dark halos might be photoevaporated by the uv flux of progenitor star .
one - dimensional radiation hydrodynamics simulations are performed by @xcite on this issue .
they found that effects of radiative feedback on the gravitational contraction of the gas in low - mass halos ( @xmath149 ) are very complicated .
the feedback effects are qualitatively different depending on the distance from the source star , evolutionary stage of the halo at the onset of radiative feedback , and mass of the halo .
@xcite also have investigated on this problem by two - dimensional radiation hydrodynamics simulations , for a single halo with mass of @xmath150 , in which they take into account a detailed evolutionary stage of the halo .
they found that most of the gas in this halo is photoevaporated by the sn progenitor star prior to its death at very early evolutionary stages of the halo . on the other hand
, they indicated that the _
i_-front could not reach the core of the halo and this core survived quite well if central densities of the halo rose beyond @xmath151 .
@xcite also performed three - dimensional radiation hydrodynamics simulations with realistic cosmological density field .
they found the h ii region of 100@xmath152 popiii star extend to @xmath153kpc , within which most of low - mass halos are photoevaporated .
however , the number of survived halos through uv flux from progenitor star is still under debate , since we only have the results by multidimensional simulations with restricted parameter space ( mass of the halo , mass of the progenitor star ) . in any case
, it is worth to investigate the effects of sn shock on the nearby halos with various parameters by analytical calculations . in our study , we completely ignore mass loss in the neighbor halos due to photoevaporation by the progenitor star .
therefore , it must be noted that our results are taken to be a lower limit on the damage done by the expanding remnant to the halo because the halos the sn blast actually encounters will have less gas .
also the negative feedback effect we have found applies only to the one scenario , in which the snr increasingly sweeps up mass and then collides with the neighboring halo
. it could be that sn directly form more stars by other means than those they quench in nearby halos in our mechanism .
we have studied the feedback effects by first sne with @xmath154 , and @xmath138 on their neighboring dark matter halos .
consequently , the conclusion can be summarized as follows .
we find that the star formation activities in the neighboring dark matter halos ( @xmath5 ) are basically suppressed in case they are located close to the sn center , because of the gas evacuation effect .
the critical distance within which the gas is blown away falls within the range @xmath6kpc depending on the sn energy and the halo mass . in case
we assume normal core - collapse sn ( @xmath143erg ) , the critical distance is @xmath144 kpc , whereas it is @xmath155kpc for pair - instability sn ( @xmath146erg ) .
in addition , we find there is very little window in the parameter space where star formation activities in dark halos are induced or promoted by neighboring sn .
we thank the anonymous referee for critical comments to improve the paper .
we also thank t. kitayama , d. sato , m. umemura , & k. ohsuga for fruitful discussions and useful comments .
the analysis has been made with computational facilities at center for computational sciences in university of tsukuba and rikkyo university .
this work was supported in part by ministry of education , culture , sports , science , and technology ( mext ) , grants - in - aid , specially promoted research 16002003 .
|
the first - generation stars in the @xmath0cdm universe are considered to have formed in dark halos with total masses in the range @xmath1 at @xmath2 . these stars expected to be very massive and
in some cases they end their lives as the first supernovae ( sne ) .
we explore the problem of whether star formation in low mass dark halos ( @xmath3 ) was triggered or suppressed by the sn feedback from neighboring star - forming halos .
we take into consideration mainly two effects by the sn shock : one is the evacuation of gas components from the halos and the other is the promotion of h@xmath4 formation because of the enhanced ionization degree by shock heating . combining above effects
, we find that the star formation activities in the neighboring dark matter halos ( @xmath5 ) are basically suppressed in case they are located close to the sn center , because of the gas evacuation effect .
the critical distance within which the gas is blown away falls within the range @xmath6kpc depending on the sn energy and the halo mass .
in addition , we find there is very little window in the parameter space where star formation activities in dark halos are induced or promoted by neighboring sn .
|
dk lac was a moderately fast ( t@xmath3 = 32 d ) nova in 1950 , reaching v = 6 mag .
the nova light curve developed small eruptions midway down the decline ( e.g. ribbe 1951 ) , and the nebulae was spatially resolved ( barely ) by cohen ( 1985 ) and by slavin , obrien & dunlop ( 1995 ) ; otherwise dk lac has rather undistinguished nova properties ( duerbeck 1981 ) , and is not well - studied in its post - nova stage .
as far as we are aware there has not been an orbital period study ( spectroscopic or photometric ) for dk lac .
nova - like ( nl ) cataclysmic variables ( cvs ) sometimes fade by 1 - 4 magnitudes , remaining low from months to years ; these are known as vy scl stars .
the vy scl phenomenon is thought to be due to a cessation or diminution of the mass transfer from the secondary star to the white dwarf , temporarily robbing the system of its accretion luminosity . in 2001
the autonomous 0.41-m telescope at indiana university found that dk lac had entered a low state ( henden , freeland & honeycutt 2001 ) which lasted about 3 years . a preliminary report on this low state , based on partial data available at that time , appeared in hoffman et al . 2003 ) .
this current paper reports on nearly 20 years of sampled dk lac photometry surrounding this low state , plus continuous photometric sequences on numerous nights during both the low state and the subsequent return to the high state .
we also report on spectra acquired during the high state .
the dk lac low state studied here is of interest because it is one of the very few old nova to have displayed vy scl - like behavior .
the character of the low state therefore bears on how novae behave between nova outbursts and whether they have long intervals of hibernation ( shara 1989 ) between nova eruptions , without accretion luminosity .
our dk lac photometry is of two types .
first , we have long - term monitoring at a typical cadence of 2 - 3 observations per week , obtained mostly with the local indiana automated photometric telescopes .
this program , operating nearly continuously since 1990 , is mostly used to monitor @xmath0120 nl and old nova cvs for low states , both for what can be learned from the photometry itself ( e.g. honeycutt , cannizzo & robertson 1994 ; honeycutt & kafka 2004 ; kafka & honeycutt 2005 ) , and to trigger low state spectrographic studies ( e.g. kafka et al . 2005a ; kafka , honeycutt & howell 2006 ; kafka et al .
2008 ) , as well as trigger more intensive low - state photometry ( e.g. kafka et al . 2005a ; 2005b ; 2007 ; 2008 ) . to ensure good time coverage the indiana telescopes
are fully automated for autonomous operation , including open - up / close - down decisions , dynamic scheduling , liquid nitrogen autofills , acquisition of daily bias and flats , and data reduction .
no comprehensive desscription of these systems was ever published , but much of the technical details ( honeycutt et al .
1990 ; honeycutt 1992 ; honeycutt & turner 1992 ; honeycutt et al .
1994a ) as well as motivations and strategies ( honeycutt 1994 ; honeycutt et al .
1994b ; honeycutt , cannizzo & robertson 1994 ; robertson , honeycutt & turner 1995 ; honeycutt & kafka 2004 ; kafka & honeycutt 2005 ) are available .
the second variety of our dk lac photometry consists of continuous monitoring in sequences lasting 2 - 7 hours each .
these data were mostly acquired in attended mode using a variety of telescopes and observers .
the motivation for the monitoring was two - fold : 1 ) it was hoped that the orbital period might be revealed , and 2 ) the low - state photometric behavior of vy scl stars on time scales of minutes to tens of minutes is not well defined , with several interesting effects to be explored .
for example , in the low state of some cvs one sometimes sees 0.6 mag flares lasting 10 - 20 min , which have been attributed to either stellar flares on the m dwarf doner star , or mass transfer bursts ( e.g. honeycutt & kafka 2004 and references therein ) .
table 1 is a log of the photometric observations .
column 1 is a sequence designation ( provided only for continuous exposures on a given night ) .
column 2 gives the date ( or the date range ) , while column 3 provides similar information for the jd .
column 4 is the telescope , column 5 the filter , column 6 the exposure time in sec , column 7 the number of useable exposures , and column 8 the duration of the data stream .
the long - term photometry was acquired and processed in two different indiana university ( iu ) observing programs .
the first program is an unattended , autonomous 0.41-m telescope in central indiana , informally called roboscope .
this telescope collected v - band exposures of dk lac for about 14 years 1990 - 2004 ( the first entry in table 1 ) .
flats and other detector calibration data were automatically acquired and applied after each exposurre , followed by aperture photometry and field identification , all using custom software ( honeycutt & turner 1992 ) .
final photometric reductions were done using the incomplete ensemble technique contained in astrovar , which is a custom package based on the technique described in honeycutt ( 1992 ) , but with the addition of a graphical user interface .
the second program for the long - term photometry used an unattended , autonomous 1.25-m telescope at the same site as roboscope ( the second entry in table 2 ) .
dk lac exposures were acquired by this telescope for nearly 3 years 2006 - 2009 .
the real - time data reductions used for the roboscope data were tightly integrated into our old vms operating system , while the 1.25-m telescope is linux - based .
the 1.25-m data were therefore reduced using a batch pipeline ( not real - time ) consisting of iraf routines for detector calibrations , followed by the application of sextractor .
the light curves were then generated using astrovar .
the iu 1.25-m exposures of dk lac were acquired during commissioning of the instrumentation .
the nightly dome flats gave poor results and we did not have twilight flats on most nights .
therefore we used median sky flats constructed from all the exposures in a given filter ( typically 25 - 75 ) on a given ( or adjacent ) night .
we had significant dark current from the thermoelectically - cooled ccd that was in service on the iu 1.25-m at this time , which somewhat degraded the s / n .
the errors for the 1.25-m dk lac data are typically 0.025 mag , about 2@xmath1 the expected error if we had been able to use high s / n dome flats and if the dark current were neglibible .
the iu 1.25-m exposures of dk lac began with the v filter , but later ( after jd 2454330 ) switched to a clear ( c ) filter .
however , we have no secondary standards for the c filter so the v standards were used instead .
this has little effect on the differential light curve , but does introduce some additional uncertainly into the zeropoint .
the dk lac exposures which consist of continuous ( or nearly continuous ) sequences on a given night ( as opposed to the much more widely spaced automated long - term photometry ) originated from 3 different observing programs .
first we have v - band sequences on 3 nights in 2001 from the 1.0-m telescope at the u.s .
naval obervatory , flagstaff station ( usno / fs ) ( entries 01 - 1 to 02 - 1 in table 1 ) .
these images were reduced using iraf routines for both detector calibrations and aperture photometry .
next we have v - band sequences on 15 different nights 2001 - 2006 using the 0.91-m wiyn telescope at kitt peak ( entries 03 - 1 to 06 - 5 in table 1 ) .
most of these images were reduced using iraf for both detector calibrations and aperture photometry .
however a few sequences used cmuniwin , a pc ( windows)-based photometry package orginally developed by horch ( 1998 ) .
these 0.91-m wiyn sequences were often ( but not always ) implemented as back - up programs for use on non - photometric nights , and therefore sometimes have gaps due to passing clouds .
finally we have sequences on 8 nights 2007 - 2008 obtained with the iu 1.25-m telescope ( entries 07 - 1 to 08 - 3 in table 1 ) , using the same reduction pipeline described earlier for the long - term photometry of dk lac on this telescope .
we also have occasional dk lac exposures ( not parts of sequences ) from the usno / fs 1.0-m telescope , from the wiyn 0.91-m telescope , and from the tenagra observatory 0.8-m telescope in southern arizona ; these occasional measures have been incorporated into our long - term light curve .
the last reduction stage for all the dk lac data used astrovar , which provides photometric errors based on the repeatability of non - variable stars as a function of instrumental magnitude .
these errors are for differential magnitudes with respect to the ensemble , which for dk lac consisted of 86 nearby stars .
all of the dk lac photometry used the secondary standards found in henden & honeycutt ( 1997 ) to establish the zeropoint .
the error in this zeropoint is typically @xmath00.015 mag .
the roboscope data and the usno / fs data have been transformed to the standard ubv system using transformation coefficients evaluated from regular b , v observations of standard star fields . in applying this transformation
we assumed a color for dk lac of 0.0 because b measures were usually not available for dk lac ; this approximation will introduce some additional error into the zeropoint .
finally , no transformations were applied for the wiyn 0.91-m data or for the iu 1.25-m data ( neither v nor c ) , resulting in additional zeropoint error for that data .
overall we estimate that the zeropoint error is @xmath40.05 mag for all the data , and significantly less for most of the data .
dk lac has a faint companion @xmath05@xmath5 east . on a low state usno / fs 1.0-m image acquired 2001-dec-17 ut ( jd = 2452260.70278 ) in good seeing these two stars were measured separately using daophot psf photometry in iraf , with the following results : dk lac v = [email protected] , b - v = [email protected] ; companion v = [email protected] , b - v = [email protected] . in combined light , the companion contributes @xmath030 - 50% of the v - band low - state light ( depending on the variable low - state brightness of dk lac ) , and @xmath09% of the high - state light .
it was not practical to do psf photometry on all our images because the two stars are not resolved in poor seeing nor when the system is in the high state .
therefore we chose to use a numerical diaphragm for dk lac that includes both stars , and correct the resulting magnitude for the contribution of the companion .
this means that our high - state magnitudes are @xmath00.10 mag fainter than the magnitudes which other observers might report for the same high - state epoch , if they are unaware of the contribution from the companion .
figure 1 shows our full light curve ( with 2304 points plus 92 limits ) , where we have plotted the magnitudes corrected for the light of the companion .
the error bars ( which are not shown in figure 1 , but which are present on subsequent light curve plots ) are for differential magnitudes with respect to the ensemble .
table 2 lists all the magnitudes and errors ( for the differential magnitudes ) .
the complete version of table 2 is available only in electronic form .
the data will also be archived with the aavso . in the low state
dk lac is mostly just beyond the magnitude limit for roboscope , resulting in numerous upper limits ( which nevertheless help define the duration of the low state ) in figure 1 .
( when a limit is obtained , this means that the combined light of dk lac and the companion was not detected .
technically then , the limits are for a different circumstance than for the magnitudes , which are for dk lac only , and the limits are actually fainter ( by an unknown amount ) than shown ) . labels 01 - 08 in figure 1 designate continuous sequences ; see table 1 and the caption to figure 1 for further information . it is seen in figure 1 that dk lac experienced a 2 mag deep low state beginning 2001-jan and ending sometime between 2003-jul and 2004-jun , for a duration of 2.5 - 3.4 years
. there also appears to be an overall decline of @xmath00.2 mag in the high state brightness over our 19 years of data .
spectra of highly magnetic nl cvs ( polars ) in the low state typically have emission lines arising from the inner hemisphere of the doner star .
these lines are not single - peaked but have satellite components whose origin remains uncertain ( kafka et al .
2005a , 2005b , 2007 ; 2008 ) .
these satellites may arise from large - scale magnetic structures on the secondary star , or magnetic structures connecting the secondary star to the white dwarf .
a central question is whether this behavior is induced by the strong magnetic field of the white dwarf or is a property of `` hyperactivity '' on the secondary star .
it is therefore important to obtain low - state spectra of cvs which are disk systems ( such as dk lac ) rather than polars , in order to help distinguish these two possibilities .
our attempts at low - state spectroscopy of dk lac were unsuccessful because of system faintness . by the time we were able to obtain spectra using larger telescopes , dk lac had returned to the high state .
figure 2 is an expanded portion of the dk lac light curve in figure 1 , with the times of our spectra marked .
table 3 is a journal of our spectroscopic observations .
spectral sets s1 and s4 were obtained using the goldcam slit spectrograph on the kpno 2.1-m telescope .
grating 35 was used in first order , providing coverage @xmath05000 - 8100 at @xmath03 resolution .
spectral set s2 was obtained using the mos / hydra multiple object spectrograph on the wiyn telescope .
grating 600 was used in first order , providing coverage @xmath05300 - 8200 .
the `` red '' 2@xmath5 fiber bundle was employed , yielding @xmath03 resolution ; numerous other fibers were used for sky subtraction .
spectral set s3 used the 6.5-m mmt at mt .
hopkins , arizona .
the blue channel slit spectrograph was employed with a 500 line / mm grating , providing coverage @xmath05000 - 8100 at a resolution of @xmath03.6 .
for all spectra our detector calibrations used standard iraf procedures , and for spectral extractions and wavelength calibrations we used iraf s onedspec / twodspec packages .
no spectrophotometric calibrations were applied for any of the spectra , and the continua in the reduced spectra were nomalized to unity .
also , no corrections were made for telluric spectral features in the near - ir .
each spectral sequence contained 2 - 4 exposures ( see table 2 ) , which were combined after reductions .
figures 3 and 4 are representative plots of the light curves of the individual photometric sequences listed in table 1 .
figure 3 contains two nights of low - state data , demonstrating that some nights have little variation , while other have systematic 1 mag changes over several hours .
furthermore the mean low - state mag can change by 0.5 mag between nights .
figure 4 is an example of a night of high - state photometry , where we see flickering - type behavior with little change in mean mag over 7 hours .
the location of the sequences in the long - term light curve are labeled in figure 1 .
we discuss separately below the low - state and the high - state photometry .
livio & pringle ( 1994 ) proposed that vy scl low states are due to a large starspot drifting into the l1 region of the doner star .
by analogy to sunspots ( which are observed to have signifcantly lower gas densities than the surrounding photosphere ) it was argued that a starspot at l1 will greatly lower @xmath7 , reducing the accretion luminosity and producing a vy scl low state .
vy scl stars typically take a few tens of days to enter a low state . in the starspot hypothesis
the ingress and egress time scales of low states are not well - contrained because they depend on the poorly - known drift rate of star spots on cv doner stars .
honeycutt & kafka ( 2004 ) reported that the shapes of many vy scl ingress and egress events were dual - sloped , always being faster when fainter .
it was argued that this effect is consistent with the starspot scenario in a number of respects , if the two slopes are associated with the passage of the umbral and penumbral portions of the starspot drifting across l1 .
figure 2 shows a few data points on the decline of dk lac into the low state , especially the early stages .
however later portions of the decline have few points because roboscope does not reliably reach fainter than v@xmath018 .
the recovery from the low state was almost totally missed .
the low state had an amplitude of 2.0 mag and ( taking into account the missing - data intervals ) lasted between 2.5 and 3.4 years .
figure 5 shows an expanded portion of figure 1 where we have characterized the decline as two straight lines by appealing to the behavior of better - sampled high - state low - state transitions of vy scl stars in honeycutt & kafka ( 2004 ) . in that work
it was found that , for @xmath040% of the observed transitions , a pronounced slope change occurred during the transition to and from the low state .
that slope change was always in the sense that the transition was faster when fainter .
the speed of the transitions was characterized by an e - folding time , @xmath8 . for 17 transitions having dual slopes in honeycutt & kafka ( 2004 ) ,
it was found that @xmath9 ranged from 3 to 40 days ( with a mean of 15 days ) , while @xmath10 ranged from 4 to 160 days ( with a mean of 52 days ) . for the straight lines plotted in figure 5 we have for dk lac @xmath9 = 12 days and @xmath10 = 175 days ,
values within or near the range for other vy scl stars .
the photometric sequences acquired during the low state fall naturally into two sets . set 1 consists of 3 sequences obtained over a 21 night interval between 2001-dec-18 and 2002-jan-07 ( sequences 01 - 1 to 02 - 1 in table 1 ) , while set 2 consists of 7 sequences obtained 20 months later , over an 18 night interval in 2003-jun / jul ( sequences 03 - 1 to 03 - 7 ) .
figure 6 shows the mean magnitudes for the sequences from 2001 - 02 ( set 1 ) alongside a similar plot for the sequences from 2003 ( set 2 ) .
there does not appear to be any systematic change in the mean low - state brightness over these intervals .
the short error bars ( with caps ) are for the standard deviation of the mean ( sdm ) , showing that night - to - night variations substantially exceed the sdm of the individual sequences . the long error bars ( without caps ) are for the standard deviation of a single observation ( sdso ) , which should be independent of the number of points in the sequence , measuring instead how `` noisy '' or `` quiet '' is each photometeric sequence .
there is no apparent systematic change in sdso as the low state progresses , nor any correlation of sdso with low - state mean magnitude .
the mean value of sdso for the 10 low - state sequences is 0.16 magnitude , and the mean low - state magnitude is 19.0 .
our data provide little leverage to characterize the low - state night - to - night variations .
therefore we searched for periodicities in the expected orbital period range of 1 to 12 hours , using data which we pre - whitened by adjusting the level of each sequence to a common mean magnitude .
periodograms ( not shown ) were constructed for a variety of groupings of the low - state data . in these power spectra
there are no isolated peaks that stand out well from the noise and the alising background .
instead we see that the major power is distributed primarily between 2 and 3 hours for the 2001 - 2002 data , but primarily between 3 and 5 hours for the 2003 data .
we folded the data on the highest peaks , but the resulting light curves did not provide any additional insight into the nature of the variations .
when we constructed new periodograms by randomizing the magnitudes , keeping the same jds , it was concluded that the excess power at these characteristic frequencies is real , as is the change in the characteristic time scale .
periodograms of the high - state photometry were examined using pre - whitened data and a variety of data groupings . for the continuous sequences we find broad power from 2 to 3.5 hours , with somewhat isolated peaks at 0.11162(5 ) and 0.08431(3 ) day for some data groupings .. the light curves folded on these two periods have sinusoidal amplitudes of @xmath00.03 mag with superimposed scatter of @xmath00.2 mag .
it seems likely that all these `` periods '' are spurious , but we mention them for completeness .
the iu 0.41-m data of 1990 - 2002 consists of single exposures acquired at typical intervals of a few days .
this cadence could hardly be more different than that for the sequenced photometry , which typically consists of 20 - 80 exposures over a few hours of a given night ( see table 1 ) .
the nights containing sequences are typical of conventionally - scheduled observing runs , consisting of widely - separated sets of several adjacent or nearly - adjacent nights .
the combination of these two cadences would seem to be ideal for establishing any reliable periodicities or quasi - periodicities , especially since the errors for the sequences and for the sampled photometry are similar . however , there are no significant peaks in common between the periodograms of the sequenced photometry and the sampled photometry , suggesting that no stable periodicities are present in the high - state data .
even the concentrated power between 2 and 3.5 hours seen in the periodograms of the sequence photometry is largely missing in the periodogram of the sampled photometry . with regard to
the b - v color of dk lac in the high state , szkody ( 1994 ) reported 0.08 and ringwald et al .
( 1996 ) gave 0.1 .
ringwald et al .
also provide e@xmath11 = 0.45 , which implies an intrinsic color corresponding to a blackbody with t@xmath1250,0000@xmath13 .
as for the high - state spectra , figures 7 , 8 , 9 , and 10 show the combined spectra for sets s1-s4 .
the best s / n is found in s3 ( figure 9 ) , although it has some residual near - ir fringing longward of h@xmath2 .
in s3 we see in emission h@xmath2 , hei 5876 , and hei 6676 .
hei 6676 is also visible in s2 ( fig 8) and s4 ( fig 10 ) .
the nad doublet is in absorption with a total equivalent width of 1.8 .
distance , extinction , and m@xmath14 are not well known for dk lac .
duerbeck ( 1981 ) , using a calibration of light curve decay time with luminosity ( along with an adopted extinction of 1.2 mag ) estimates a distance of 1500@xmath6200 pc , while slavin et al .
( 1995 ) used the method of nebular expansion parallax to derive 3900@xmath6500 pc . in our spectrum
s3 the ew of na d1 is 0.83 . using the calibration found in munari and zwitter ( 1997 ) we find a@xmath14 = 1 - 3 mag .
this wide range is due to saturation effects along with uncertainties introduced by unresolved multiple line components . using the nebular expansion parallax distance and adopting a@xmath14 = [email protected] we find m@xmath15 = [email protected] , and m@xmath14 = [email protected] 59 years after the outburst , in 2009 .
these m@xmath16 values are consistent with results for other novae at maximum and for old novae decades after the nova ( warner 1995 ) .
the extinction is unlikely to be much higher than our adopted value because then the intrinsic color becomes too blue for a blackbody , requiring the assumption of strong emission lines or other non - thermal contributions to the b - v color .
interestingly , the color of dk lac assumed an even bluer value of -0.17 in the low state .
this result is difficult to understand .
most cvs become redder in the optical when accretion luminoisity falls , as the cool secondary star begins to dominate .
perhaps the secondary star in dk lac is of particularly low luminosity and/or the accretion luminosity did not dissappear completely .
this blue color in the low state is based on only a single measurement , but nevertheless appears secure ( see @xmath17 2.1 ) .
the inclination of the system can be estimated using the calibrations found in warner ( 1986 ) . using the correlation of h@xmath2 ew with orbital inclination
, we find that i = 0 - 50@xmath13 for dk lac .
the correlation of m@xmath14 with disk inclination is more confining , yielding i = 0 - 25@xmath13 .
it seems clear that the disk of dk lac is presented nearly face - on .
this may account for the lack of any orbital signature in the photometry reported in this paper .
the effect of various kinds of non - axisymmetic accretion structure(s ) on the light curve is minimized when the portions of the disk seen by the observer are always in view , as is the case at low orbital inclination .
in fact both the high - state and low - state photometry appear random in nature , though perhaps with preferred time scales .
for the sampled high - state photometry obtained prior to the low state , prewhitened by observing season , the rms scatter is 0.094 mag . for the sequenced high state photometry after the low state , prewhitened by sequence ,
the rms scatter is 0.085 mag .
let us take 0.09 mag rms as the characteristic scatter ( flickering ) , during the high state , to be compared to 0.16 mag rms in the low state .
vy scl stars do not often stay in the low state long enough to characterize their low state variations , but when they do ( e.g. mv lyr in honeycutt & kafka 2004 ) the light curve seems flat to within @xmath00.15 mag , with brief 0.5 mag flares superimposed .
this behavior is similar to that seen in polars , where the low states are more frequent ( see low - state light curves of am her , st lmi , and ar uma in kafka & honeycutt 2005 ) .
schmidtke et al .
( 2002 ) reports on sequenced photometry obtained as lq peg recovered from a low state , finding 0.09 mag variations , with no coherent signal between 1 and 6 hours ; these sequences were obtained roughly mid - way up the recovery to the high state .
schmidtke et al .
( 2002 ) found that as lq peg brightened from 15.8 to 14.5 the flickering decreased from 0.09 to 0.02 mag .
this is in the same sense as the change we find in dk lac .
however , as dk lac brightened from 19.0 to 17.0 , the amount of flickering declined from 0.16 to 0.09 mag , a less extreme change than seen in lq peg .
schmidke et al . also found that the flickering was a constant fraction of the intensity as the system recovered from the low state .
that is clearly not the case in dk lac , in which the high state rms variation expressed in intensity units is 3 - 4@xmath1 the low state rms variation in intensity units .
the hibernation scenario for cyclic nova episodes ( see shara 1989 for a review ) holds that at some point after the nova outburst accretion drops to a near zero , a state known as hibernation ; the system is thought to remain in hibernation for most of the interval between nova outbursts .
advantages of hibernation include reducing the mean accretion rate to levels needed to allow the next thermonuclear runaway , and providing consistency between nova rates and cataloged cvs because of the inconspicious nature of hibernating systems
. the scenario has received criticisms ( e.g. naylor et al . 1992 ) but remains a topic of considerable interest ( e.g. martin & tout 2005 ) . the fact that post - nova systems fade by 1 - 2 mag per century following the nova ( vogt 1990 ; duerbeck 1992 ) is consistent with the hibernation scenario , but it may also be the case that the onset of low states marks the transition into hibernation .
at least one other old novae has had a low state subsequent to the nova outburst .
v533 her ( nova her 1963 ) experienced two prominent low states in 1995-june and 1996-jan , each lasting 2 - 3 months , with depths of 1.5 mag ( honeycutt & kafka 2004 ) .
these v533 her low states were too brief for in - depth study .
the dk lac and v533 her low states occurred 51 years and 32 years respectively after the nova .
neither system was monitored regularly prior to 1990 , so other low states might have been missed .
honeycutt & kafka found that in 12 years of monitoring 65 old novae and nl cvs , @xmath015% showed at least one low state exceeding 1.5 mag .
statistics for the old novae portion of the sample are not as good as desired , but there is nevertheless no evidence that the rate of low states in old novae ( 2 of @xmath024 ) is any different from that for nls , at least when the old novae are examined 30 - 60 years after the nova .
one might expect that if low states mark the manner in which old novae approach hibernation , then low states would be more frequent and/or more prolonged as the old novae became nls .
our tentative conclusion is that low states are unrelated to hibernation , as supported by the fact that the low state in fig 1 seems to be photometrically independent of the slow high - state decline , appearing to simply be superimposed on the light curve as an independent phenomenon .
this is consistent with low states being due to a short - term effect rather than part of an evolutionary progression , the migration of starspots under the inner lagragian point ( livio & pringle 1994 ) being a well - regarded mechanism . ignoring the low state interruption ,
the brightness of dk lac is seen in figure 1 to decline rather smoothly by 0.2 mag over 19 years .
activity cycles on the secondary stars of cvs have long been suspected to modulate the mass transfer rate and hence change the brightness of cvs on time scales of decades ( e.g. warner 1988 ; bianchini 1990 ; richman , applegate & patterson 1994 ; ak et al .
. however , we do not think that this effect is responsible for the 0.2 mag decline , for two reasons .
first , there is no inflection in the decline as might be expected for cyclic behavior shorter than 3 decades .
second , the decline rate is consistent with the rate found for other old novae many decades after outburst . for nine
old novae duerbeck ( 1992 ) finds a mean rate of decline 50 years after outburst of 10@xmath63 millimag per year .
this is attributed to the slow decline of irradiation - induced mass transfer , and is concluded to be consistent with the hibernation hypothesis .
the rate in dk lac is 10.5 millimag per year , fully consistent with the behavior of other post - novae . over roughly the same interval in which the continuum of dk lac declines by 0.2 mag , the ew of h@xmath2 declines by @xmath02@xmath1 .
because the continuum is falling at the same time , this weakening of h@xmath2 is actually even more pronounced . in figure 11
we plot the ew of h@xmath2 vs jd , where we have added a 1991 data point from ringwald et al .
the suggestion of a linear decline in h@xmath2 ew over 16 years is intriging , but needs confirmation .
the initial fall in @xmath7 ( and therefore in system brightness ) following the nova eruption is thought to be due to declining irradiation of the doner star as the white dwarf cools following the nova eruption ( duerbeck 1992 and references therein ) .
the hibernation scenario ( shara 1989 ) holds that @xmath7 continues to fall even further , taking the system through the dwarf nova ( dn ) regime and finally into a detached binary state with little or no accretion luminosity . in general , the ew of dn emission lines is greater than that of emission lines in old novae , opposite to the trend in figure 11
. therefore we suggest that the emission lines in dk lac arise mostly from the illuminated secondary star rather than from the accretion disk .
the breadth of the emission lines in dk lac are small ( fwhm@xmath18 = 12 , fwhm@xmath19 = 8 ) compared to most old nova .
while much of the narrowness can be attributed to the low system inclination , it is also consistent with the emission arising from the irradiated inner hemisphere of the secondary star .
long - term photometry of nova lac 1950 ( dk lac ) from 1991 to 2008 shows a 2 magnitude deep vy scl - type low state beginning 2001-jan and lasting for between 2.5 and 3.4 years .
the shape of the ingress into the low state is similar to that of other vy scl stars , having a more rapid fall as the ingress proceeds .
there is also an overall decline of 0.2 magnitude in the high state brightness over our 19 years of data , consistent with the fading rate of other old novae when measured several decades after the nova eruption .
our photometry also includes 26 continuous sequences each lasting 2 to 7 hours .
we searched unsuccessfully for photometric periodicities in the range of 1 - 12 hours ( the range of expected orbital periods ) using both the long - term and the sequenced data , and both the high and low state data .
the lack of orbital - time - scale variations is attributed to the nearly face - on presentation of the disk .
high - state spectra are also presented and discussed .
we find that the equivalent width of h@xmath2 falls by @xmath02@xmath1 from 1991 to 2008 .
the fact that this low state occured in an old novae suggests that it might be connected to the entrance of dk lac into hibernation .
however no distinction is found between the low states of old nova and those of nl cvs ( which were presumably unrecorded novae hundreds of years ago ) .
instead we conclude that low states are unrelated to the possible entrance of a system into hibernation .
constantine deliyannis , caty pilachowski , and their students , had programs on the wiyn 0.91-m that required photometric weather .
when the weather was mostly clear but non - photometric , dk lac frequently became the back - up target .
schmidtke , p.c . ,
ciudin , g.a . ,
indlekofer , u.r . ,
johnson , d.r .
, fried , r.e . & honeycutt , r.k .
2002 , asp conf .
261 , the physics of cataclysmic variables and related objects , ed .
gansicke et al . , p. 539
lllcccrc & 1990-nov-12 to & 2448207 to & iu 0.41-m & v & 240 & 739 & 14.1 yr + & 2004-dec-04 & 2453343 & & & & & + & & & & & & & + & 2006-aug-23 to & 2453947 to & iu 1.25-m & v , c & 150 & 251 & 2.8 yr + & 2009-may-21 & 2454972 & & & & & + & & & & & & & + 01 - 1 & 2001-dec-18 & 2452261 & usno / fs 1.0-m & v & 120 & 54 & 3.4 hr + 01 - 2 & 2001-dec-20 & 2452263 & usno / fs 1.0-m & v & 120 & 50 & 3.1
hr + 02 - 1 & 2002-jan-07 & 2452281 & usno / fs 1.0-m & v & 120 & 60 & 2.7 hr + & & & & & & & + 03 - 1 & 2003-jun-29 & 2452819 & wiyn 0.91-m & v & 90 - 150 & 55 & 3.3 hr + 03 - 2 & 2003-jun-30 & 2452820 & wiyn 0.91-m & v & 120 - 240 & 63 & 4.6 hr + 03 - 3 & 2003-jul-02 & 2452822 & wiyn 0.91-m & v & 100 - 500 & 35 & 4.1 hr + 03 - 4 & 2003-jul-03 & 2452823 & wiyn 0.91-m & v & 100 & 87 & 4.2 hr + 03 - 5 & 2003-jul-10 & 2452830 & wiyn 0.91-m & v & 180 & 40 & 3.0 hr + 03 - 6 & 2003-jul-11 & 2452831 & wiyn 0.91-m & v & 180 & 33 & 3.4 hr + 03 - 7 & 2003-jul-16 & 2452836 & wiyn 0.91-m & v & 180 & 26 & 2.8 hr + & & & & & & & + 04 - 1 & 2004-jun-29 & 2453185 & wiy 0.91-m & v & 60 & 43 & 1.6 hr + 04 - 2 & 2004-jul-03 & 2453189 & wiy 0.91-m & v & 60 & 18 & 0.9
hr + 04 - 3 & 2004-jul-05 & 2453191 & wiy 0.91-m & v & 60 & 70 & 2.4 hr + & & & & & & & + 06 - 1 & 2006-sep-30 & 2454008 & wiy 0.91-m & v & 120 & 35 & 2.0 hr + 06 - 2 & 2006-oct-01 & 2454009 & wiy 0.91-m & v & 120 - 300 & 109 & 7.0 hr + 06 - 3 & 2006-sep-03 & 2454011 & wiy 0.91-m & v & 120 - 300 & 77 & 7.0 hr + 06 - 4 & 2006-oct-04 & 2454012 & wiy 0.91-m & v & 120 - 300 & 18 & 1.3 hr + 06 - 5 & 2006-oct-05 & 2454013 & wiy 0.91-m & v & 120 - 300 & 40 & 4.0 hr + & & & & & & & + 07 - 1 & 2007-aug-18 & 2454330 & iu 1.25-m & v & 180 & 70 & 4.6 hr + 07 - 2 & 2007-sep-04 & 2454347 & iu 1.25-m & c & 180 & 70 & 4.7 hr + 07 - 3 & 2007-sep-14 & 2454357 & iu 1.25-m & c & 180 & 68 & 4.6 hr + 07 - 4 & 2007-oct-07 & 2454380 & iu 1.25-m & c & 120 & 63 & 2.7 hr + 07 - 5 & 2007-oct-10 & 2454383 & iu 1.25-m & c & 120 & 65 & 3.9 hr + & & & & & & + 08 - 1 & 2008-oct-03 & 2454742 & iu 1.25-m & c & 150 & 46 & 4.5 hr + 08 - 2 & 2008-oct-04 & 2454743 & iu 1.25-m & c & 150 & 20 & 2.4 hr + 08 - 3 & 2008-oct-05 & 2454744 & iu 1.25-m & c & 150 & 41 & 4.3 hr + cccc 2448207.68948 & 16.943 & 0.034 & robos + 2448208.63269 & 16.841 & 0.031 & robos + 2448208.76882 & 16.862 & 0.044 & robos + 2448209.64547 & 16.938 & 0.031 & robos + 2448209.73172 & 16.923 & 0.040 & robos + 2448209.80370 & 16.899 & 0.048 & robos + 2448233.72259 & 17.054 & 0.063 & robos + 2448234.61401 & 16.919 & 0.044 & robos + 2448234.71393 & 16.872 & 0.052 & robos + 2448235.58032 & 16.907 & 0.039 & robos + ............. & ...... & ..... & ..... + ............. & ...... & ..... & ..... + ............. & ...... & ..... & ..... + 2454972.82035 & 17.130 & 0.029 & iu 1.25-m + cccccc s1 & 2005-oct-12 & 2453655 & kpno 2.1-m & 600 - 900 & 2 + s2 & 2005-oct-26 & 2453669 & wiyn 3.5-m & 600 & 3 + s3 & 2006-sep-16 & 2453994 & mmt 6.5-m & 900 & 2 + s4 & 2008-jun-07 & 2454624 & kpno 2.1-m & 900 & 4 +
|
we report on extensive photometry of dk lac obtained during the interval 1990 - 2009 , which includes a 2 mag low state during 2001 - 2003 .
much of the photometry consists of exposures obtained with a typical spacing of several days , but also includes 26 sequences of continuous photometry each lasting 2 to 7 hours .
we find no evidence for periodicities in our data .
we do find that the random variations in the low state are @xmath02@xmath1 those in the high state , when expressed in magnitudes .
the lack of orbital - time - scale variations is attributed to the nearly face - on presentation of the disk .
there is a 0.2 mag decline in the high state brightness of the system over 19 years , which is consistent with the behavior of other old novae in the decades following outburst .
high - state spectra are also presented and discussed .
we find that the equivalent width of h@xmath2 falls by @xmath02@xmath1 from 1991 to 2008 .
the photometric properties are discussed in the context of the hibernation scenario for the behavior of novae between outbursts , in which we conclude that low states in old novae are probably unrelated to their possible entrance into hibernation .
|
zinc oxide has been the subject of considerable experimental and theoretical investigation for many years.@xcite its bandgap is in the near ultraviolet , making it useful as a transparent conductor and as sunscreen .
its piezoelectricity opens up transduction applications .
the activity has intensified more recently because of the possibility that zno might be useful for spintronics or spin - based quantum computation .
it has been predicted to be a room - temperature ferromagnet when doped with mn.@xcite furthermore , its spin - orbit coupling is generally thought to be very weak compared with gaas .
the usual measure of the strength of spin - orbit coupling in semiconductors is the energy splitting at the top of the valence band .
it is said that the spin - orbit coupling is negligible in zno because the valence - band splitting is @xmath1 mev,@xcite as opposed to @xmath2 mev for gaas .
smaller spin - orbit coupling should lead to long spin relaxation times .
long relaxation times are required if spin information is to be transported over appreciable distances .
the spin relaxation time @xmath0 has been measured by ghosh et al .
@xcite to be about 20 ns from 0 to 20 k in optical orientation experiments .
@xmath0 is sometimes called @xmath3 even in the absence of an external field . since the data from ref . ( ) used in this paper were taken at zero field , the relaxation time is taken to be @xmath0 to avoid confusion with experiments conducted at finite field .
the data show two surprising features .
first , the relaxation times are actually somewhat shorter than the longest relaxation times in gaas , which are about 100 ns.@xcite one might expect the opposite given the relative strength of spin - orbit coupling in the two materials .
second , @xmath0 shows a non - monotonic temperature dependence , first increasing slightly and then rapidly decreasing - but increasing temperature usually promotes spin relaxation . in this paper
, we show that a theory previously developed for @xmath0 in gaas @xcite can account for these observations .
the theory must be modified to take account of the different impurity levels and binding energies of zno .
this is important , because , in spite of intensive investigation , the nature of the impurities that govern the electrical properties of zno remains controversial , and our analysis sheds some light on this issue .
even more interestingly , it turns out that the wurtzite crystal structure has very important consequences for the dyakonov - perel ( dp ) @xcite scattering that dominates the relaxation at higher temperatures .
thus the crystal structure must be taken into account fully .
the final message will be that the weak spin - orbit coupling of zno is not negligible for spin relaxation , and it does not lead to long relaxation times . in the next section
we give the background information for zno .
iii is devoted to a derivation of the equations of motion for the spins . in sec .
iv the computational method is described .
the determination of the parameters in the equations of motion is a separate task .
the most important of the parameters is that which controls that dp spin relaxation . since the calculation of this parameters is not straightforward , we devote sec .
v to that .
vi gives the results .
vii is the conclusions and puts the results into context .
in zno produced by the hydrothermal method , it is generally thought that there are two sets of impurity states , one shallow and quasi - hydrogenic , one deep and very well localized.@xcite their precise physical nature is not known . in the case of the deep impurity , it is believed that a lattice defect accompanies the chemical impurity .
the binding energies are in the range of a few 10s of mev for the shallow impurity and a few 100s of mev for the deep impurity .
we shall demonstrate below that the optical orientation data can put bounds on these numbers .
zno crystallizes in the wurtzite structure rather than the zincblende structure familiar from the iii - v compounds .
this has very important implications for the conduction band states .
the spin - orbit interaction lifts the spin degeneracy in the conduction band .
in zincblende structures crystal symmetry implies that the splitting is cubic in the magnitude of the wavevector @xmath4 , but in the wurtzite structure the splitting is linear.@xcite however , the spin relaxation time of the low - lying conduction band states depends mainly on the spin splitting near the conduction band minimum , and this is larger in zno than in gaas for small enough @xmath4 . in optical orientation experiments ,
electrons are excited from the valence band to the conduction band by circularly polarized light tuned close to the bandgap energy ( pump step ) .
the population of conduction electrons so created is spin - polarized.@xcite energy relaxation then occurs on a short time scale ( @xmath5 ns ) , but most of this relaxation is from spin - conserving processes , so there is a longer time scale ( or time scales ) on which the spin of the system relaxes . when an external magnetic field is present , the time scale to relax the transverse component of the _ net _ magnetization
is called @xmath0 .
it is measured using faraday rotation or the kerr effect ( probe step ) .
the important physical point is that the fast energy relaxation leads to a thermal charge distribution for the electrons by the time 1 ns has elapsed , but the spin distribution relaxes on longer time scales .
the thermal charge distribution means that the localized donor impurity states are mostly full at the relatively low temperatures of the experiment .
the spins of the localized electrons must be included along with the conduction electron spins .
the spins of localized and extended states can be interchanged by the exchange coupling , a process we call cross- relaxation .
this is often a rather fast process and is particularly important when the relaxation times of the localized and extended states are very different in magnitude . in gaas
this process is important in all the regimes of temperature , applied field , and impurity density that have been studied , and it is important in zno as well . in the following section we derive a set of modified bloch equations to describe the aforementioned spin dynamics .
we consider a conduction electron in the semiclassical approximation .
it moves as a wavepacket with a well - defined momentum and scatters from impurities and phonons at time intervals of average length @xmath6 where @xmath7 is the momentum relaxation time .
its spin operator is @xmath8 the spin - dependent part of its hamiltonian in the absence of an external magnetic field is : @xmath9 the first term , @xmath10 , is the exchange interaction with impurity spins @xmath11 located at an positions @xmath12 .
it is the same interaction that is responsible for the kondo effect , but the temperatures here are all much greater than the kondo temperature .
the range of the function @xmath13 is roughly @xmath14 where @xmath15 is the effective bohr radius .
the second term , @xmath16 , represents other spin relaxation mechanisms that we model as a small random classical field @xmath17 with a correlation time much shorter than @xmath0 .
an analogous hamiltonian @xmath18 can be written for a localized electron .
first , we concentrate on the spin dynamics resulting from the spin - spin term and ignore the second term . in the dilute limit ( @xmath19 ) , a conduction electron encounters impurities with randomly aligned spins if no short - range order is present in the impurity system . an effective field from the impurity spin affects the conduction electron when it is within @xmath20 of the impurity .
when @xmath21 , the conduction electron proceeds unhindered by the effective field .
this effective field is a result of the exchange potential .
an itinerant electron will spend an average time of @xmath22 within the range of the effective field where @xmath23 is the velocity of the electron .
thus the time between encounters@xcite is @xmath24 . in a semiclassical picture
the spin of the itinerant electron undergoes precession of magnitude @xmath25 through a random angle during each encounter with an impurity .
the spin of the impurity electron also precesses but with angle @xmath26 . since the sum of spins , @xmath27 , commutes with @xmath28 , the total spin in the system
must be conserved .
however the spin in each subsystem may shift between one another ; this is cross - relaxation .
it turns out for the parameters of the system under consideration that @xmath29 , and we then find that @xmath30 which implies that the spin is essentially randomized after one impurity encounter . if we consider an ensemble of conduction electrons with a net magnetization @xmath31 , this magnetization is exchanged at a rate of @xmath32 . as previously mentioned , any magnetization lost from the conduction electrons must be gained by the localized electrons and vice - versa . for clarity
we write @xmath33 and @xmath34 where @xmath35 .
we now examine the second term of the hamiltonian @xmath36 this hamiltonian relaxes the conduction electron spin . to extract a relaxation rate from this hamiltonian
, we use the equation of motion @xmath37\ ] ] where @xmath38 is the @xmath39 spin density matrix for an electron of a given momentum .
we assume that the total density matrix for the conduction electron factorizes ; we neglect off - diagonal terms that come from correlations . by iteration
, we can write this equation as @xmath40\right\rangle -\frac{1}{\hbar^{2}}\int_{0}^{t}\left\langle { \left [ \left [ \rho\left ( t^{\prime}\right ) , h_{2}^{c}\left ( t^{\prime } \right ) \right ] , h_{2}^{c}\left ( t\right ) \right ] } \right\rangle dt^{\prime}\ ] ] where the angular brackets indicate averaging over all orientations of @xmath41 . to simplify notation , from now on angular brackets
will be suppressed on the density matrix . since @xmath42 ,
the first term is zero .
we assume that different directions of @xmath43 are uncorrelated and ( since the external field is zero ) different direction are equivalent
. then we have @xmath44 therefore , eq . ( [ vonneumann ] ) reduces to @xmath45\sigma_{i}\langle b(t)b(t^{\prime})\rangle dt^{\prime}.\label{vonneumann2}\ ] ] the correlation function is assumed to be stationary in time so @xmath46.@xcite if the correlation time of the @xmath47-fluctuations , @xmath48 , is short , @xmath49 will not change on that timescale and @xmath50 will be nearly a @xmath51- function ( markov approximation ) . eq .
( [ vonneumann2 ] ) can then be written as @xmath52\sigma_{i}\int_{0}^{\infty}\left\langle b(t)b(t^{\prime})\right\rangle dt^{\prime}.\ ] ] the integral is approximated by @xmath53 .
define the relaxation time scale @xmath54 by @xmath55 giving @xmath56\sigma_{i}.\label{tc}\ ] ] the density matrix can be expanded in pauli spin matrices @xmath57 where @xmath58 is the @xmath39 identity matrix and @xmath59 is the expected value of the magnetization . inserting eq .
( [ pauli ] ) in eq .
( [ tc ] ) and matching coefficients of pauli matrices gives a set of equations for the dynamics of @xmath60 .
for instance for conduction electron magnetization @xmath31 in the @xmath61-direction , @xmath62 . as with @xmath10 ,
similar expressions for the localized magnetization @xmath63 can be found : @xmath64 . by combining the effects of @xmath65 and @xmath66 ,
the modified bloch equations for the magnetizations can be expressed as @xmath67 for two spin systems - itinerant and localized spins . @xmath54 and @xmath68 in eq .
( [ bloch0 ] ) are now described in terms of well known relaxation mechanisms which will be discussed in the next section .
this model was successfully applied to gaas.@xcite for zno , these bloch equations are easily extended to account for the multiple - type impurities present .
we now seek to write equations like those of eq .
( [ bloch0 ] ) with regard given to the two types of impurities in zno - shallow and deep .
as mentioned above , we find that the cross - relaxation is important to understand the data .
these rates come from the kondo - like @xmath69 interaction between an impurity spin @xmath70 and a conduction band spin @xmath71 .
an expression for @xmath72 in terms of tight - binding parameters can be derived using the schrieffer - wolf transformation.@xcite one expects that the cross - relaxation between conduction and shallow donor electrons to be much more rapid than the cross - relaxation between conduction and deep donor electrons because of the greater binding energy of the deep impurity and its larger on - site coulomb energy .
this is confirmed by the fit to the data .
in fact we find that terms involving cross - relaxation between the deep donors and either the conduction band electrons or the shallow donor electrons can be neglected . with these simplifications , for zno eq .
( [ bloch0 ] ) extends to @xmath73 in this equation , @xmath74 @xmath75 and @xmath76 stand for the magnetizations of the conduction electrons , the electrons on shallow impurities , and the electrons on deep impurities , respectively . the @xmath77 s denote the corresponding volume densities .
each of the populations has a relaxation time @xmath78 and @xmath79 from eq .
( [ eq : blochsimp ] ) , we can then find the magnetization as a function of time .
standard methods can be used to solve these differential equations .
the solutions yield a time dependence of the total magnetization , @xmath80 , to be a sum of three exponentials , @xmath81 , @xmath82 , and @xmath83 where @xmath84 with @xmath85 given by @xmath86 no net moment can exist on the deep donor sites since no moment is excited into the deep states on account of them being significantly below the conduction band , and no net moment cross relaxes into these states .
therefore @xmath87 can be ruled out as being the observed relaxation rate . in the regime
that @xmath88 , the rate @xmath89 simplifies to @xmath90 and is very rapid and the rate @xmath91 is slower , @xmath92 we fit the data with this equation and associate it with @xmath0 .
we see that the relaxation rate depends on two factors : the thermodynamic occupations of the shallow donors ( the deep donors are always nearly full in the temperature range studied here ) and form of the relaxation rates for the conduction and localized shallow states .
the densities can be computed using standard formulas from equilibrium statistical mechanics , since we deal only with time scales long compared to the fast energy relaxation scale . as a function of temperature @xmath93
, the ratio @xmath94 naturally increases rapidly as @xmath95 is so large that these states are always occupied at the experimental temperatures , which range from 5k to 80k .
@xmath54 is fairly complicated to calculate because there are several mechanisms that can relax the conduction electron spins .
the simplest such mechanism is the elliot - yafet ( ey ) process@xcite that arises from spin mixing in the wavefunctions .
when a conduction electron is scattered by a spin - independent potential from state @xmath96 to state @xmath97 , the initial and final states are not eigenstates of the spin projection operator @xmath98 so the process relaxes the spin .
the rate of relaxation due to the ey process is well known to be of the form : @xmath99 where @xmath100 is a material dependent parameter and @xmath7 is the momentum relaxation time.@xcite we estimate @xmath101 k@xmath102 .
the bir - aronov - pikus ( bap ) mechanism@xcite arises from the scattering of electron and holes .
this relaxation mechanism is commonly considered to be negligible in @xmath77-type materials like those under consideration here since the number of holes is small.@xcite the dyakonov - perel ( dp ) mechanism @xcite arises from the ordinary scattering of conduction - band states .
since this has not previously been calculated in a wurtzite structure , we devote the next section to it .
this calculation yields an expression for @xmath54 as a function of temperature .
@xmath103 and @xmath104 are due to non - spin - conserving anisotropic exchange ( dzyaloshinski - moriya ) interactions.@xcite the anisotropic exchange term is important . it arises from spin - orbit coupling and produces a term proportional to @xmath105 where @xmath106 is related to the interspin separation and the exchange integral between the wavefunction on sites 1 and 2 .
however , it is not possible to calculate it in detail when the nature of the impurities is not well known .
we estimate the rate as @xmath107 where @xmath108 and @xmath109 are the total impurity concentrations of the shallow and deep impurity respectively and @xmath110 has a weak temperature dependence that we neglect .
the main contribution comes from the the overlap of the shallow impurity wavefunctions , which we take to be hydrogenic , with the deep impurity wavefunctions , which we take to be well - localized on an atomic scale .
the details of how to estimate the resulting relaxation may be found in refs .
the numerical value we find from theory is @xmath111 @xmath112 ns@xmath113 .
when nuclei possess nonzero magnetic moments , the hyperfine interaction between electron and nuclear spin is a source of spin relaxation for localized electrons.@xcite however , zero nuclear spin isotopes of zn and o are 96% and 99.5% naturally abundant respectively .
therefore we rule out the hyperfine interaction from being an observed relaxation mechanism in ref .
the conduction band states undergo ordinary impurity and phonon scattering .
each scattering event give a change in the wavevector @xmath96 , which in turn changes the effective magnetic field on the spin that comes from spin - orbit coupling .
this fluctuating field relaxes the spin .
the effective field strength is proportional to the conduction band spin splitting .
bulk zincblende crystals have conduction band splittings cubic - in-@xmath4 due to bulk inversion asymmetry ( dresselhaus effect ) .@xcite in addition to cubic terms , bulk wurtzite conduction bands also possess spin splittings proportional to linear terms in @xmath4 due to the hexagonal @xmath114 axis which gives bulk wurtzite a reflection asymmetry similar to the rashba effect.@xcite we can write the spin - orbit hamiltonian to include both the rashba and dresselhaus terms : @xmath115\cdot\mathbf{\sigma } \label{eq : hamiltonian}\ ] ] where @xmath116 is linear - in-@xmath4 , @xmath117 is cubic - in-@xmath4 , @xmath118 are the pauli spin matrices , and @xmath119 , @xmath120 are spin splitting coefficients.@xcite the parameter @xmath47 is roughly equal to four for all wurtzite materials.@xcite note that there is no spin splitting along the hexagonal axis ( @xmath121 ) .
the linear - in-@xmath4 term dominates and we can determine the spin relaxation rate by following the treatment given by pikus and titkov in ref .
( ) which yields the following relaxation rates : @xmath122@xmath123 where the overbar denotes angular averaging , and @xmath124 denote the cartesian components of @xmath125 .
the momentum relaxation rate is defined as @xmath126 where @xmath127 is the scattering cross section and @xmath128 is the angle between initial and final @xmath96.@xcite in bulk wurtzite @xmath129 , @xmath130 , and in the unstrained crystal , @xmath131 for @xmath132 . from eq .
( [ eq : relaxation ] ) , we can write @xmath133 where @xmath134 is the electron effective mass and @xmath135 is the energy @xmath136 .
this result can be boltzmann averaged ( denoted by angle brackets ) to obtain @xmath137 where @xmath138 and @xmath139 .
the temperature dependent momentum relaxation time , @xmath140 , can be determined from electron mobility ( @xmath141 ) measurements from @xmath142 where @xmath143 is the charge of an electron .
@xmath144 has been calculated @xcite to be @xmath145 ev - nm which gives a theoretical value of @xmath146 k@xmath113 ns@xmath102 .
similarly , the cubic - in-@xmath4 term can be calculated to be @xmath147 where @xmath148 where the dimensionless quantity @xmath149 depends on the type of scattering and is of order unity .
@xmath120 has been calculated@xcite to be @xmath150 ev - nm@xmath151 which yields @xmath152 k@xmath153 ns@xmath102 .
the sample from which the momentum relaxation times @xmath140 were extracted @xcite was hydrothermally grown by the same company as ghosh et al .
sample in ref .
in fig.1 we show that temperature dependence of @xmath0 as measured in a bulk zno sample and our fit ( using eq .
( [ eq : eigenvalue ] ) ) to the data .
it is seen immediately that the temperature dependence is not monotonic and that this is well - reproduced by the theory .
the reason is simple . at low temperatures @xmath154 nearly all the electrons are in localized states .
these states relax by the temperature - independent dm mechanism : @xmath155 .
this mechanism alone determines the @xmath156 values .
when @xmath93 approaches @xmath157 the deep impurities are all occupied but the rest of the population is shared by shallow localized and conduction band states .
initially , the conduction band electrons have a longer spin lifetime because impurity scattering is frequent at low temperatures so the dp mechanism that relaxes them is not very effective
. however , the dp mechanism increases rapidly as @xmath93 increases and the @xmath0 curve turns around . at @xmath158
the shallow impurity level is empty and the relaxation is dominated by the dp mechanism in the conduction band : @xmath159 . at this point
it is necessary to point out why only the linear - in-@xmath93 dp mechanism is needed to explain the observed conduction spin relaxation .
the other two viable candidates ( cubic dp and ey ) for relaxation are much too weak to explain the observed relaxation times in zno .
we use the calculated values for @xmath160 and @xmath161 in the previous section to obtain the relative relaxation efficiencies between the linear and cubic dp mechanism terms : @xmath162 which demonstrates that the efficiency of the cubic - in-@xmath93 term does not become comparable to the linear - in-@xmath93 term at temperatures below @xmath163 k which is far above the temperature range investigated here .
in fact the cubic - in-@xmath93 term does not even reach one - tenth the efficiency of the linear - in-@xmath93 term in the temperature range investigated here .
for this reason we can confidently ignore the cubic - in-@xmath93 dp mechanism term in our fit .
the crystal structure of zno therefore makes its spin relaxation qualitatively different from spin relaxation in bulk @xmath77-gaas .
we also compare the efficiencies of the dp and ey mechanisms : @xmath164 even if the momentum relaxation time taken to be unrealistically low , say @xmath165 fs , the dp mechanism is still nearly two orders of magnitude more efficient at relaxing spins than the ey mechanism in the temperature range studied here . due to the drastic qualitative and quantitative differences between relaxation mechanisms
, we have unequivocally determined the relevant conduction electron spin relaxation mechanism in zno .
[ fig : plot ] the fit of theory to the experimental data is clearly very good .
we found that no reasonable fit was possible using only a single impurity level , though this worked very well for gaas @xcite , so we used two levels .
a good fit by this method was possible by adjusting the coefficients @xmath166 and @xmath167 and the binding energies @xmath168 and concentrations @xmath169 of the two donors , subject to the constraint that the room temperature carrier density should equal the measured@xcite value of @xmath170 cm@xmath153 .
qualitatively , one finds that @xmath172 and @xmath173 to get the right order of magnitude of the relaxation at low @xmath174 physically , the deep impurity spins are important because they relax the shallow impurity spins by the dm mechanism , and the strength of the low @xmath93 relaxation implies that the deep impurities must be quite numerous .
quantitatively , a least squares fit to the data of ref .
( ) yields @xmath175 k@xmath113 ns@xmath102 , @xmath176 ns@xmath113 , @xmath177 mev , @xmath178 mev , and @xmath179 @xmath180 @xmath181 is about four times larger than the theoretical value of @xmath160 given above , possibly due to strain effects .
we also note that the values of @xmath7 that we used were taken from a different sample .
if we take @xmath109 to be near the highest values measured for the deep donor ( see below ) then @xmath182 @xmath183ns@xmath113 is about one order of magnitude larger than the theoretical estimate @xmath184 given above . in view of the very poor understanding of the impurity wavefunctions , and the exponential dependence of @xmath110 on the overlaps ,
this is perhaps not too disturbing .
the presence of a shallow donor and a very deep donor has been seen in hydrothermally grown zno samples of the type investigated here.@xcite donor concentrations up to nearly @xmath185 cm@xmath153 ( @xmath109 ) have been measured for donors @xmath186 mev ( @xmath187 deep.@xcite donors as shallow as @xmath188 mev ( @xmath189 ) have been measured@xcite at lower concentrations @xmath190 cm@xmath153 ( @xmath191 . comparison with our values indicates that the parameters extracted from the fit are very reasonable for this material . from this analysis
, we predict that in zno samples with fewer deep impurities , the relaxation time at low temperatures can be increased .
as the impurities of zno vary greatly between different growth techniques@xcite , this prediction could be tested by further optical orientation experiments on different samples .
we have found that @xmath0 in bulk zno can be understood by invoking previously known spin relaxation mechanisms .
the dominant mechanisms in the material turn out to be the dp ( scattering ) relaxation of the conduction electron spins for @xmath192 k and the dm ( anisotropic exchange ) mechanism for the localized spins for @xmath193 k. in addition , it is very important to include the cross - relaxation between localized and conduction states previously proposed for gaas .
these physical ingredients explain quantitatively the relatively fast relaxation at low temperatures as being due mainly to the dm mechanism which in turn depends on having both deep and shallow impurity states . at high temperatures ,
the conduction states are dominant , and the dp mechanism gives an excellent fit to the data .
the combination explains the very surprising non - monotonic temperature dependence of @xmath0 . finally , there are two aspects of the data in ref .
( ) that we have not addressed here : the applied magnetic field dependences on the spin relaxation and the spin relaxation observed in zno epilayers . we plan on addressing the former issue in a future publication . as for the latter issue
, the epilayers are doped three to four orders of magnitude higher than in the bulk case . at such high dopings ,
spin glass effects become important and localized donor states coalesce to produce donor bands ; we do not expect our theory to be applicable in such a regime .
the theory has now been sufficiently developed that optical orientation experiments can actually serve as a characterization tool for doped semiconductors , giving information about the binding energies and concentrations of the electrically active impurities in @xmath77-type materials .
|
doped zno is a promising material for spintronics applications . for such applications , it is important to understand the spin dynamics and particularly the spin relaxation times of this ii - vi semiconductor .
the spin relaxation time @xmath0 has been measured by optical orientation experiments , and it shows a surprising non - monotonic behavior with temperature . we explain this behavior by invoking spin exchange between localized and extended states .
interestingly , the effects of spin - orbit coupling are by no means negligible , in spite of the relatively small valence band splitting .
this is due to the wurtzite crystal structure of zno
. detailed analysis allows us to characterize the impurity binding energies and densities , showing that optical orientation experiments can be used as a characterization tool for semiconductor samples .
|
pseudogap or partial gap in the electronic structure , affecting some regions of the fermi surface while leaving other unaffected , is one of the key signatures of the underdoped cuprates @xcite .
it is revealed through anomalous behaviors of the temperature - dependent resistivity , magnetization , nmr knight shift and relaxation rate , as well as in spectroscopic data @xcite .
pseudogap shows the same @xmath7-space distribution as the superconducting gap @xcite and is universally observed in hole- and electron- doped cuprates in the underdoped regime .
features consistent with pseudogap are also clearly found in hole doped feas - based materials ( see @xcite for a review ) . because the parent compounds of iron pnictides are metals ,
the pseudogap here is believed to arise from nesting instability @xcite . in ba(fe@xmath1t@xmath2)@xmath3as@xmath3 ( bat122 in the following)substitution of the transition metals into fe position
leads to electron doping .
nmr studies suggest the existence of a pseudogap in baco122 over the broad doping range including full domain of superconductivity , from magnetically ordered parent compound to non - superconducting metal .
existence of pseudogap leads to a temperature - dependent knight shift @xmath8 , well described by a formula @xmath9 , where the first term describes contribution of the metallic portion of the fermi surface and the second activated term allows determination of the @xmath10 as 560 k @xmath11150 k at optimal doping @xcite . at temperatures
@xmath12 this leads to temperature independent knight shift and a crossover to metallic temperature dependence in the inter - plane resistivity @xcite .
no discernible features are observed in the in - plane resistivity @xcite , which suggests that the areas of the fermi surface affected by pseudo - gap are rather small and belong to the most warped parts of the fermi surface , contributing mostly to inter - plane transport . in this article
we report a systematic study of the evolution of the inter - plane resistivity with doping by other transition metals inducing superconductivity in ba122 , @xmath4= rh , ni , pd .
we show that similar anomalies are observed in the temperature dependent inter - plane resistivity for all types of substitution , with the characteristic temperature of resistive crossover being suppressed with doping .
the rate of @xmath5 suppression with @xmath13 is however notably higher in ni and pd doped compositions , even with correction for a difference in number of added electrons .
the doping - dependence of the pseudogap feature suggests that it represents an independent energy scale in the problem , different from that of structural / magnetic transition and superconductivity .
single crystals of bafe@xmath3as@xmath3 doped with ni , pd and rh were grown as described in detail in previous communications @xcite .
crystals were thick platelets with large faces corresponding to the tetragonal ( 001 ) plane .
the actual content of transition metals , @xmath13 , was determined with wavelength dispersive spectroscopy ( wds ) electron probe microanalysis and is used in the following .
we used two - probe resistivity measurements , as justified by ultralow contact resistance of sn - soldered contacts @xcite .
the details of sample preparation , sample screening and selection are identical to those used in our studies of @xmath14-axis resistivity in co - doped material @xcite .
the absolute values of the inter - plane resistivity at room temperature for most compositions stays in the range 1 to 1.5 m@xmath15 cm , with doping it decreases to approximately 0.5 m@xmath15 cm . for several @xmath13 compositions we were not able to find crystals with resistivity values lower than 2 m@xmath15 cm , despite the facts that ( 1 ) the evolution of the temperature - dependent resistivity for these samples followed the general trend , ( 2 ) close in @xmath13 compositions show usual resistivity values .
this limits the accuracy of the absolute @xmath6 value determination by approximately a factor of two .
in the top panel of fig . [ rh ] we plot inter - plane resistivity of ba(fe@xmath1rh@xmath2)@xmath3as@xmath3 , using normalized scale @xmath16 . to avoid overlapping ,
the curves are offset progressively upwards for higher dopings .
the data for parent compound are reproduced from ref . .
several features of the temperature dependence are essentially the same as observed in compositions with co doping @xcite . for low doping @xmath17=0.012 , @xmath18 shows resistivity increase on cooling for @xmath19 ( marked with an arrow ) , which is the pseudogap feature in all compositions studied . on reaching a temperature of structural / magnetic transition @xmath20 ,
resistivity dives down , and decreases monotonically all the way to base temperature , showing some signatures of filamentary superconductivity at about 20 k due to strain @xcite .
this decrease of resistivity below @xmath20 is similarly observed in lightly co doped composition @xmath21= 0.012 @xcite . for higher dopings @xmath17=0.026 and @xmath17=0.39 resistivity shows increase below @xmath20 and drop to zero below superconducting @xmath22 , in complete accord to the behavior found in co - doped compositions . finally , for compositions in which structural / magnetic transition is completely suppressed , @xmath23 , position of the resistivity maximum shifts down in temperature and for an ultimate doping @xmath24= 0.171 , when superconductivity is completely suppressed , the resistivity shows shallow minimum , marked with cross arrow in fig . [ rh ] . the fact of the appearance of the minimum at high dopings and even the doping value at which it appears are very similar to those found in co - doped compounds .
we summarize our observations in the temperature - doping , @xmath4-@xmath13 , phase diagram in bottom panel of fig .
[ rh ] . , normalized by its value at room temperature @xmath25 , for samples of ba(fe@xmath1rh@xmath2)@xmath3as@xmath3 with @xmath26 ( slightly above the concentration boundary for the superconducting dome ) .
lines are offset , from bottom to top , @xmath17= 0 , 0.012 , 0.026 , 0.039 , 0.076 , 0.096 , 0.131 and 0.171 ( top panel ) .
arrows show a position of the resistivity maximum , @xmath5 , cross - arrow shows a position of the resistivity minimum @xmath27 .
bottom panel shows the @xmath28 phase diagram , in which lines of structural @xmath20 , magnetic @xmath29 and superconducting @xmath22 states are determined from in - plane resistivity and magnetization measurements , see ref . .
, title="fig:",width=151 ] , normalized by its value at room temperature @xmath25 , for samples of ba(fe@xmath1rh@xmath2)@xmath3as@xmath3 with @xmath26 ( slightly above the concentration boundary for the superconducting dome ) .
lines are offset , from bottom to top , @xmath17= 0 , 0.012 , 0.026 , 0.039 , 0.076 , 0.096 , 0.131 and 0.171 ( top panel ) .
arrows show a position of the resistivity maximum , @xmath5 , cross - arrow shows a position of the resistivity minimum @xmath27 .
bottom panel shows the @xmath28 phase diagram , in which lines of structural @xmath20 , magnetic @xmath29 and superconducting @xmath22 states are determined from in - plane resistivity and magnetization measurements , see ref . .
, title="fig:",width=151 ] in fig .
[ rhco ] we compare explicitly the phase diagrams of the pseudogap features in @xmath14-axis resistivity of co and rh doped compounds . within rather big error bars due to crossover character of features , the diagrams are overlapping .
this fact is remarkable , since despite electronic equivalence of co and rh doping , the @xmath14-axis lattice parameters are different in two materials @xcite , and thus it would be natural that this difference should affect characteristic energy scales of the electronic overlap in the interplane direction . , in @xmath4=rh ( solid symbols , this study ) and
@xmath4=co ( open symbols , ref . )
doped ba(fe@xmath1t@xmath2)@xmath3as@xmath3 .
pseudogap features for two types of doping overlap within error bars , similar to lines of structural @xmath20 , magnetic @xmath29 and superconducting @xmath22 transitions , ref . .
vertical lines separate composition ranges in which fermi surface topology changes in @xmath4=co @xcite .
, width=302 ] in fig .
[ ni ] and fig .
[ pd ] we show doping - evolution of the temperature - dependent interplane resistivity in samples doped with transition metals of group 10 of mendeleev periodic table , 3@xmath30 @xmath4=ni and 4@xmath30 @xmath4=pd .
these atoms donate two electrons on substituting fe , and thus substitution level required to induce superconductivity is two times lower than in cases of co and rh doping .
bottom panels of the fig .
[ ni ] and fig .
[ pd ] , show doping phase diagram of the pseudogap features in interplane resistivity .
the suppression of characteristic temperature of the resistivity maximum , @xmath5 , is much more rapid for cases of ni and pd doping , and @xmath31 line in the phase diagram suggests critical concentration very close to the edge of the superconducting dome .
moreover , for highest doping levels resistivity monotonically decreases with temperature and does not reveal a minimum at @xmath27 as in cases of co and rh doping . on the other hand ,
similar to the case of co and rh , the phase diagrams of ni and pd dopings coincide within error bars , see top panel of fig .
[ nipdrh ] . , normalized by its value at room temperature @xmath25 , for samples of ba(fe@xmath1ni@xmath2)@xmath3as@xmath3 with @xmath32 ( slightly above the concentration boundary for the superconducting dome ) .
lines are offset , from bottom to top , @xmath33= 0 , 0.0067 , 0.016 , 0.024 , 0.032 , 0.054 , and 0.072 ( top panel ) .
arrows show a position of the resistivity maximum , @xmath5 , used to plot the phase diagram , see fig .
[ nipdphased ] . ,
width=302 ] , normalized by its value at room temperature @xmath25 , for samples of ba(fe@xmath1pd@xmath2)@xmath3as@xmath3 with @xmath34 ( slightly above the concentration boundary for the superconducting dome ) .
lines are offset , from bottom to top , @xmath35= 0 , 0.012 , 0.021 , 0.027 , 0.030 , 0.053 , and 0.077 ( top panel ) .
arrows show a position of the resistivity maximum , @xmath5 , used to plot the phase diagram , see fig . [
nipdphased ] . , title="fig:",width=302 ] = ni ( solid symbols ) and @xmath4=pd ( open symbols ) doped ba(fe@xmath1t@xmath2)@xmath3as@xmath3 as determined from inter - plane resistivity measurements . lines of structural @xmath20 , magnetic @xmath29 and superconducting @xmath22 states are determined from in - plane resistivity and magnetization measurements , see ref . .
all characteristic features for two types of doping overlap within error bars .
, title="fig:",width=302 ]
in systematic study of doping phase diagrams for a variety of transition metal dopants : co , rh , ni , pd , cu and cu+co , it was found that the superconducting transition temperature @xmath22 scales with the number of doped electrons , @xmath36 , while the structural / magnetic transition temperatures scale with number of dopant atoms @xmath37 .
it is therefore interesting if pseudogap features follow either of these scaling relations . while the break - up of @xmath38 scaling is obvious from the comparison of fig .
[ rhco ] and figs .
[ ni ] , [ pd ] , in bottom panel of fig .
[ nipdrh ] we compare the phase diagrams of rh and pd doped bafe@xmath3as@xmath3 vs the number of extra electrons @xmath39 .
it is clear that the pseudogap temperature does not scale with @xmath39 , contrary to the supercoducting transition temperature . because of relatively big error bar in determination of the pseudogap features , in the bottom panel of fig .
[ nipdrh ] we compare pseudogap features for all four compounds studied .
the difference between the two columns is still clearly resolved , supporting the lack of the scaling of pseudogap features with @xmath39 .
the difference in doping dependence of pseudogap features pushes us to recognize pseugogap as an independent characteristic scale in the phase diagram of ba122 materials . .
bottom panel shows phase diagram of the pseudogap features for all studied transition metals .
resistivity maximum , @xmath5 , for rh ( blue solid up - triangles ) , co ( black solid diamonds ) , pd ( open down - triangles ) and ni ( red solid squares ) and resistivity minimum , @xmath27 , for rh ( solid pink star ) and co ( open black circles ) . , title="fig:",width=151 ] .
bottom panel shows phase diagram of the pseudogap features for all studied transition metals .
resistivity maximum , @xmath5 , for rh ( blue solid up - triangles ) , co ( black solid diamonds ) , pd ( open down - triangles ) and ni ( red solid squares ) and resistivity minimum , @xmath27 , for rh ( solid pink star ) and co ( open black circles ) . , title="fig:",width=151 ] the evolution of the fermi surface topology in baco122 with doping was studied using arpes and thermopower measurements @xcite .
a sequence of three lifshits transitions was found , with concentration boundaries at @xmath40=0.3 , @xmath41=0.11 and @xmath42=0.195 @xcite .
these are shown with grey lines in fig .
unexpectedly , doping evolution of the pseudogap temperature @xmath5 shows little correlation with characteristic features in the fermi surface evolution . in particular , hole pocket near @xmath43 point in the brillouin zone changes shape from cylindrical to ellipsoidal at @xmath41 , but merely any feature can be noticed in @xmath31 , see fig .
considering low anisotropy of electrical resistivity of the compounds , it is not clear which cylinder of the fermi surface is responsible for carrier activation , however , pseudogap affects most strongly the most warped portions .
comparison of the temperature - dependent inter - plane resistivities for heavily overdoped rh @xmath17=0.171 and pd @xmath35=0.077 shows an interesting difference .
the resistivity monotonically decreases with heating for @xmath35=0.077 , top curve in fig .
[ pd ] , decreasing from low temperatures to room temperature by a factor of 2 . very similar magnitude of decrease is observed in sample with @xmath33=0.072 , top curve fig .
the magnitude of decrease in @xmath17=0.171 , top curve in fig .
[ rh ] , is merely 10% , and the curve shows metallic resistivity increase on heating above @xmath44200 k. this difference suggests that the intact metallic part of the fermi surface , contributing to the interplane transport , is notably smaller for ni and pd doping , than in cases of co and rh doping .
it is interesting if this difference can be found in nmr studies as well .
in conclusion , our systematic study puts constraints on the possible origin of the anomalous resistivity behavior in transition metal doped bafe@xmath3as@xmath3 .
characteristic temperatures of the pseudogap do not follow inter - plane distance , in the samples of the same column of mendeleev table , as would be naturally expected for variation of the inter - plane transfer integrals .
they neither follow the electron count as would be expected if they were in relation with the fermi surface volume / topology ( in contrast with the superconducting @xmath22 which scales with @xmath39 ) , nor doping evolution of the structural/ magnetic transitions .
this pushes us to recognize pseudogap as yet another energy scale in the fascinating complexity of the iron pnictide superconductors .
its origin remains as obscure at the moment as the origin of mysterious pseudogap phase in the cuprates .
we thank a. kreyssig for useful discussions .
work at the ames laboratory was supported by the department of energy - basic energy sciences under contract no .
de - ac02 - 07ch11358 .
slb was supported in part by the state of iowa through the iowa state university .
ning , k. ahilan , t. imai , a. s. sefat , r. jin , m. a. mcguire , b. c. sales , and d. mandrus j. phys .
soc . jpn . *
77 * , 103705 ( 2008 ) .
ning , k. ahilan , t. imai , a. s. sefat , r. jin , m. a. mcguire , b. c. sales , and d. mandrus , j. phys .
jpn . * 78 * , 013711 ( 2009 ) . f. l. ning , k. ahilan , t. imai , a. s. sefat , m. a. mcguire , b. c. sales , d. mandrus , p. cheng , b. shen , and h .-
h wen , phys .
rev . lett . * 104 * , 037001 ( 2010 ) .
n. ni , a. thaler , j. q. yan , a. kracher , e. colombier , s. l. budko , p. c. canfield , and s. t. hannahs , phys . rev .
b * 82 * , 024519 ( 2010 ) . p. c. canfield , s. l. budko , n. ni , j. q. yan , a. and kracher , phys .
b * 80 * , 060501(r ) ( 2010 ) .
eun deok mun , sergey l. budko , ni ni , alex n. thaler , and paul c. canfield , phys . rev .
b * 80 * , 054517 ( 2009 ) . c. liu , t. kondo , r. m. fernandes , a. d. palczewski , e. d. mun , n. ni , a. n. thaler , a. bostwick , e. rotenberg , j. schmalian , s. l. budko , p. c. canfield , and a. kaminski , nature physics * 6 * , 419 ( 2010 )
|
temperature - dependent inter - plane resistivity , @xmath0 , was measured systematically as a function of transition metal substitution in the iron - arsenide superconductors ba(fe@xmath1t@xmath2)@xmath3as@xmath3 , @xmath4= ni , pd , rh .
the data are compared with the behavior found in ba(fe@xmath1co@xmath2)@xmath3as@xmath3 , revealing resistive signatures of pseudogap . in all compounds we find resistivity crossover at a characteristic pseudogap temperature @xmath5 from non - metallic to metallic temperature dependence on cooling .
suppression of @xmath5 proceeds very similar in cases of ni and pd doping and much faster than in similar cases of co and rh doping . in cases of co and rh doping
an additional minimum in the temperature - dependent @xmath6 emerges for high dopings , when superconductivity is completely suppressed .
these features are consistent with the existence of a charge gap covering part of the fermi surface .
the part of the fermi surface affected by this gap is notably larger for ni and pd doped compositions than in co and rh doped compounds .
|
the prototype of the highly granular hadronic calorimeter for the international linear collider ( calice hcal @xcite ) consists of finely segmented arrays of scintillators .
each tile is read out by a geiger mode avalanche photodiode@xcite ( the sipm ) produced by mephi ( moscow engineering and physics institute ) , whose specific design matches the required performances needed for its application in the calorimeter @xcite . the small size ( 1 mm@xmath0 ) , high gain ( @xmath1 ) and relatively low bias voltage ( 30 - 70 v ) of the sipm made it the natural candidate for the high granularity calorimeter ; the photodetector could be directly installed in the tile . however , the sipm is a green sensitive photodetector ; the coupling with the blue emitting scintillator is obtained via a green wavelength shifting fibre , installed in a groove on the scintillator itself
. a simplification of this elementary structure would be highly desirable , in order to extend the concept to a large scale detector .
the new generation of geiger mode avalanche photodiodes , produced by hamamatsu , shows a better optical sensitivity in the 420 nm spectral region , making it possible to investigate the direct readout of the scintillation tile . in this
study the response to a minimum ionizing particle ( mip ) of a plastic organic scintillator read out by the most recent mppc@xcite , with both 400 and 1600 pixels , was analyzed .
the mip signal is in fact a benchmark for the high granularity hadron calorimeter response : it has to be well separated from thermally induced noise .
the direct readout option was compared with the established green wavelength shifting mediated design .
level.,title="fig : " ] [ dr ] the study is based on 10 samples of mppcs , 5 with 400 pixels and 5 with 1600 pixels .
they are protected by a special plastic package .
the suggested operation voltage is respectively 76 v and 78.1 v ( estimated @xmath22.5 v and @xmath21.4 v above the breakdown voltage ) , with a spread of 0.1 v. in this measurement two @xmath3 @xmath4 plastic organic scintillator tiles ( produced by _
uniplast _ enterprise in vladimir , russia ) are used . in one
a 1 mm diameter green wavelength shifting fiber ( kuraray multicladding wls fiber y11(200 ) ) is installed .
the scintillators are wrapped in a super - radiant vn2000 foil ( 3 m ) .
a special setup is designed in order to guarantee the reproducibility of the measurement ( fig .
the scintillator is installed in a fixed housing , in the middle of a robust main structure .
the mppc is secured at one end of a plastic bar .
two grooves are carved in the main structure , one starting from the middle of the side of the tile , one from the center and they host the mppc holder .
this allows the mppc to be coupled to the scintillator in two positions . in the direct readout setup , a window of @xmath5 mm@xmath0 is open in the reflective coating , in front of the mppc .
its holder is then completely guided in the casing and the mppc plastic coverage is in contact with the scintillator ; this improves the coupling between scintillator and photodetector .
the distance between the sensitive region of the mppc and the surface of the scintillator is @xmath2 2 mm . in the green wavelength
shifter mediated design , the scintillator equipped with fiber is installed in the dedicated housing and the green light is directly read out by the mppc , held by the same bar . in both designs ,
no specific optical coupling is used .
the response of the system to a @xmath6 source ( ru@xmath7 ) is investigated .
a trigger plate , consisting of a @xmath8 @xmath4 organic plastic scintillator , read out by a traditional photomultiplier tube , is installed behind the scintillator / mppc system .
the signal of the mppc is amplified by the wide - band voltage amplifier phillips scientific 6954 .
the integration is performed by the qdc lecroy 1182 , in a gate of 80 ns , produced in coincidence with the trigger .
the setup allows a reproduction of the measurement to within a systematic error of @xmath9 .
a characterization of the devices is initially performed .
the 1600 pixels mppc show a typical gain of ( @xmath10)@xmath11 . the dark rate at a threshold of 0.5 pixels is ( @xmath12 ) khz and the cross talk ( @xmath13)@xmath14 .
these values are measured at 2.5 v over the breakdown voltage .
the 400 pixels mppc exhibits a gain of ( @xmath15)@xmath11 , a dark rate at a threshold of 0.5 pixels of ( @xmath16 ) khz and cross talk of ( @xmath17)@xmath14 , operated at 1.4 v over the breakdown voltage .
the values are found to be in agreement with the data sheet performances of the photodetector .
the aim of the experiment is to extract the most probable value ( mpv ) of the number of photoelectrons produced in the mppc when a minimum ionizing particle crosses the scintillator .
the typical signal is shown in figure [ signal ] .
each peak corresponds to a certain number of pixels firing in the mppc .
the good separation of the peaks indicates a good uniformity of the device . the signal is fit with a multi - gaussian function ( fig .
[ signal].a ) .
the extracted area of each peak is then plotted versus the number of photoelectrons ( fig .
[ signal].b ) .
the areas follow a landau distribution , smeared by the poisson photo - statistics .
the maximum of this distribution is the searched mpv , which is estimated with a gaussian fit around the peak .
the results of these measurements for the green and blue light coupling for both 1600 pixels and 400 pixels mppc are shown in figure 3 , as a function of the voltage .
the efficiency of the 1600 pixels mppc is found to be about half of that of the 400 pixels device , in agreement with the photo - detection efficiency declared in the producer data sheet .
the mpv is in agreement within @xmath9 , in both cases when the mppc is either located centrally on the edge or near the corner .
the two measurements agree within the systematic error quoted above .
in the calorimeter , mip signals are used as a reference for the calibration : the mpv of the signal sets the energy scale of each channel .
an amplitude threshold defines the discrimination between the noise and the physics signal .
the normalized integral of the mip signal above the noise threshold is the mip detection efficiency .
all signals , with amplitude above threshold , constitute a hit in the calorimeter .
the procedure is shown in figure [ calor ] .
the threshold is fixed considering the pedestal spectrum ( fig .
[ calor].a ) such that the noise rate above threshold is 3 khz .
this requirement corresponds to an occupancy of about 8 accidental hits , as observed in the present 8000 channel prototype with @xmath18 ns integration time . for the ilc detector
, the occupancy of @xmath19 during one beam crossing interval ( @xmath20 ns ) translates into a sharper requirement of 300 hz noise above threshold for a single photo - sensor . the efficiency of the mip signal is then calculated using the determined threshold ( [ calor].b ) . for a given occupancy requirement ,
the dark noise of the sensor , the inter - pixel cross - talk and the most probable number of pixels corresponding to a mip determine the mip detection efficiency .
the first defines the minimum allowed threshold , the latter determines position and poisson width of the mip signal distribution and thus the integral above threshold . in the prototype of the hadron calorimeter ,
the mephi sipms provide @xmath21 photoelectrons at the mpv of the mip signal .
the efficiency of the signal collection , as measured in the prototype , is 95@xmath14 .
the efficiency of the mip signal collection of mppc , as a function of voltages , is shown in figure [ efficiency ] .
the measurements with direct coupling and with wavelength shifting fiber mediated readout are presented for both 1600 and 400 pixels mppc .
the mppc have a very low dark rate and cross talk , and the requested noise cut can be put at 1.5 - 2 pixels . the 1600 pixels device , in the direct coupling option , can be used at 2.5 - 3.5 v over the breakdown , providing 6 - 7 photoelectrons at the mpv , with a collection efficiency of @xmath22 @xmath14 .
the performance of this solution is then comparable with the one used in the actual prototype , and allows even an improvement in the dynamic range .
if the green wavelength shifting fiber is used , the mppc are definitely competitive with the sipms in use in the present prototype : when operated at 3 - 3.5 v above the breakdown voltage , they provide a better signal efficiency ( @xmath23 ) and a larger dynamic range .
the 400 pixels mppc , in both direct and wavelength shifter mediated readout , shows a very high mip signal collection efficiency , but the reduced number of pixels imposes strict bounds on the dynamic range .
it has to be noted that in both mppcs the dark rate drops rapidly as the threshold is raised , such that for the tighter requirements of the ilc a threshold of 2 - 4 pixels would be enough to keep the occupancy small . in this case
it is possible to operate at a overvoltage which preserves the mip efficiency above @xmath24 .
if thinner scintillator , e.g. 3 mm instead of 5 mm thickness with correspondingly smaller light yield , is to be used , a better coupling or a larger sensitive area of the photodetector have to be investigated .
a possible solution could be the application of the new @xmath5 mm@xmath0 mppc , but further systematical studies are needed .
the direct readout of a @xmath25 cm@xmath0 plastic organic scintillator by a multi pixel photon counter is possible , and can be an elegant solution for the future prototype of a hadron calorimeter .
the mppcs are suitable also for the green wavelength shifter fiber readout , chosen in the actual calorimeter prototype for the ilc , providing potentially even better performances than the ones in use at the moment , in term of dynamic range and mip signal collection efficiency .
the 1600 pixels devices are preferred for this particular application , as the linearity of the signal would be extended to a wider range of light intensity .
the uniformity of the light collection with respect to the particle impact position on the tile has still to be studied as well as the bounds on the dynamic range imposed by the physics observed at the linear collider .
this work is supported by desy , by the university of hamburg and by the helmholtz - nachwuchsgruppen fund vh - ng-206 .
we thank v. korbel , v. saveliev and f. sefkow for their useful suggestions and comments .
we thank hamamatsu , which kindly provided us the tested samples of mppc .
we thank also peter smirnov for his professional technical support .
we finally thank all the students , from the university and the high school who contributed to this project , in particular franziska klingberg and britta riechmann from the hamburg university , who performed an excellent laboratory practice at desy .
|
a new geiger mode avalanche photodiode , the micro pixel photon counter ( mppc ) , was recently released by hamamatsu .
it has a high photo - detection efficiency in the 420 nm spectral region .
this product can represent an elegant candidate for the design of a high granularity scintillator based hadron calorimeter for the international linear collider .
in fact , the direct readout of the blue scintillation photons with a mppc is a feasible techological solution .
the readout of a plastic scintillator by a mppc , both mediated by the traditional wavelength shifting fiber , and directly coupled , has been systematically studied .
desy 07 - 196 , ,
|
constraint satisfaction problems ( csp ) form a rich class of algorithmic problems with applications in many areas of computer science .
we only mention database systems , where csps appear in the guise of the conjunctive query containment problem and the closely related problem of evaluating conjunctive queries .
it has been observed by feder and vardi @xcite that as abstract problems , csps are homomorphism problems for relational structures .
algorithms for and the complexity of constraint satisfaction problems have been intensely studied ( e.g. @xcite ) , not only for the standard decision problems but also optimization versions ( e.g. @xcite ) and counting versions ( e.g. @xcite ) of csps . in this paper
we study the _ csp enumeration problem _ , that is , problem of computing all solutions for a given csp instance . more specifically , we are interested in the question which structural restrictions on csp instances guarantee tractable enumeration problems
. `` structural restrictions '' are restrictions on the structure induced by the constraints on the variables .
example of structural restrictions is `` every variable occurs in at most 5 constraints '' or `` the constraints form an acyclic hypergraph . ''
this can most easily be made precise if we view csps as homomorphism problems : given two relational structures @xmath0 , decide if there is a homomorphism from @xmath1 to @xmath2 . here
the elements of the structure @xmath1 correspond to the variables of the csp and the elements of the structure @xmath2 correspond to the possible values .
structural restrictions are restrictions on the structure @xmath1 .
if @xmath3 is a class of structures , then @xmath4 denotes the restriction of the general csp ( or homomorphism problem ) where the `` left hand side '' input structure @xmath1 is taken from the class @xmath5 .
@xmath6 denotes the corresponding enumeration problem : given two relational structures @xmath7 and @xmath2 , compute the set of all homomorphisms from @xmath1 to @xmath2 .
the enumeration problem is of particular interest in the database context , where we are usually not only interested in the question of whether the answer to a query is nonempty , but want to compute all tuples in the answer . we will also briefly discuss the corresponding _ search _ problem :
find a solution if one exists , denoted @xmath8 .
it has been shown in @xcite that @xmath6 can be solved in polynomial time if and only if the number of solutions ( that is , homomorphisms ) for all instances is polynomially bounded in terms of the input size and that this is the case if and only if the structures in the class @xmath3 have bounded fractional edge cover number .
however , usually we can not expect the number of solutions to be polynomial . in this case , we may ask which conditions on @xmath3 guarantee that @xmath6 has a polynomial delay algorithm . a _ polynomial delay algorithm _ for an enumeration problem is required to produce the first solution in polynomial time and then iteratively compute all solutions ( each solution only once ) , leaving only polynomial time between two successive solutions .
in particular , this guarantees that the algorithms computes all solutions in _ polynomial total time _ , that is , in time polynomial in the input size plus output size .
it is easy to see that @xmath6 has a polynomial delay algorithm if the class @xmath3 has bounded tree width .
it is also easy to see that there are classes @xmath3 of unbounded tree width such that @xmath6 has a polynomial delay algorithm .
it follows from our results that examples of such classes are the class of all grids or the class of all complete graphs with a loop on every vertex .
it is known that the decision problem @xmath9 is in polynomial time if and only if the cores of the structures in @xmath3 have bounded tree width @xcite ( provided the arity of the constraints is bounded , and under some reasonable complexity theoretic assumptions ) .
a _ core _ of a relational structure @xmath3 is a minimal substructure @xmath10 such that there is a homomorphism from @xmath3 to @xmath11 ; minimality is with respect to inclusion .
it is easy to see that all cores of a structure are isomorphic .
hence we usually speak of `` the '' core of a structure .
note that the core of a grid ( and of any other bipartite graph with at least one edge ) is a single edge , and the core of a complete graph with all loops present ( and of any other graph with a loop ) is a single vertex with a loop on it .
the core of a complete graph with no loops is the graph itself .
as a polynomial delay algorithm for an enumeration algorithms yields a polynomial time algorithm for the corresponding decision problem , it follows that @xmath6 can only have a polynomial delay algorithm if the cores of the structures in @xmath3 have bounded tree width .
unfortunately , there are examples of classes @xmath3 that have cores of bounded tree width , but for which @xmath6 has no polynomial delay algorithm unless @xmath12 ( see example [ exa : second - solution ] ) .
our main algorithmic results show that @xmath6 has a polynomial delay algorithm if the cores of the structures in @xmath5 have bounded tree width and if , in addition , they can be reached in a sequence of `` small steps . ''
an _ endomorphism _ of a structure is a homomorphism of a structure to itself .
retraction _ is an endomorphism that is the identity mapping on its image .
every structure has a retraction to its core .
however , in general , the only way to map a structure to its core may be by collapsing the whole structure at once . as an example , consider a path with a loop on both endpoints .
the core consists of a single vertex with a loop .
( more precisely , the two cores are the two endpoints with their loops . ) the only endomorphism of this structure to a proper substructure maps the whole structure to its core .
compare this with a path that only has a loop on one endpoint .
again , the core is a single vertex with a loop , but now we can reach the core by a sequence of retractions , mapping a path of length @xmath13 to a subpath of length @xmath14 and then to a subpath of length @xmath15 et cetera . we prove that if @xmath3 is a class of structures whose cores have bounded tree width and can be reached by a sequence of retractions each of which only moves a bounded number of vertices , then @xmath6 has a polynomial delay algorithm .
we also consider more general sequences of retractions or endomorphism from a structure to its core .
we say that a sequence of endomorphisms from a structure @xmath16 to a substructure @xmath17 , from @xmath18 to a substructure @xmath19 , , to a structure @xmath20 has _ bounded width _ if @xmath20 and , for each @xmath21 , the `` difference between @xmath22 and @xmath23 '' has bounded tree width . we prove that if we are given a sequence of endomorphisms of bounded width together with the input structure @xmath1 , then we can compute all solutions by a polynomial delay algorithm .
unfortunately , in general we can not compute such a sequence of endomorphisms efficiently .
we prove that even for width @xmath24 it is np - complete to decide whether such a sequence exists .
finally , we remark that our results are far from giving a complete classification of the classes @xmath3 for which @xmath25 has a polynomial delay algorithm and those classes for which it does not . indeed , we show that it will be difficult to obtain such a classification , because such a classification would imply a solution to the notoriously open _ csp dichotomy conjecture _ of feder and vardi @xcite ( see section [ sec : tract ] for details ) . due to space restrictions several proofs are omitted .
[ [ relational - structures . ] ] * relational structures . * + + + + + + + + + + + + + + + + + + + + + + + + a _ vocabulary _
@xmath26 is a finite set of _ relation symbols _ of specified arities .
relational structure _ @xmath1 over @xmath26 consists of a finite set @xmath27 called the _ universe _ of @xmath1 and for each relation symbol @xmath28 , say , of arity @xmath29 , an @xmath29-ary relation @xmath30 . note that we require vocabularies and structures to be finite .
a structure @xmath1 is a _ substructure _ of a structure @xmath2 if @xmath31 and @xmath32 for all @xmath33 .
we write @xmath34 to denote that @xmath1 is a substructure of @xmath2 and @xmath35 to denote that @xmath1 is a _ proper _ substructure of @xmath2 , that is , @xmath36 and @xmath37 .
a substructure @xmath38 is _ induced _ if for all @xmath28 , say , of arity @xmath29 , we have @xmath39 . for a subset @xmath31 , we write @xmath40 $ ] to denote the induced substructure of @xmath2 with universe @xmath27 . [
[ homomorphisms . ] ] * homomorphisms .
* + + + + + + + + + + + + + + + + we often abbreviate tuples @xmath41 by @xmath42 . if @xmath43 is a mapping whose domain contains @xmath44 we write @xmath45 to abbreviate @xmath46 .
a _ homomorphism _ from a relational structure @xmath1 to a relational structure @xmath2 is a mapping @xmath47 such that for all @xmath28 and all tuples @xmath48 we have @xmath49 .
a _ partial homomorphism _ on @xmath50 to @xmath2 is a homomorphism of @xmath51 $ ] to @xmath2 .
it is sometimes useful when designing examples to exclude certain homomorphisms or endomorphisms . the simplest way to do
that is to use unary relations .
for example , if @xmath52 is a unary relation and @xmath53 we say that @xmath54 _ has color _ @xmath52 . now
if @xmath55 does not have color @xmath52 then no homomorphism from @xmath1 to @xmath2 maps @xmath54 to @xmath56 .
two structures @xmath1 and @xmath2 are _ homomorphically equivalent _ if there is a homomorphism from @xmath1 to @xmath2 and also a homomorphism from @xmath2 to @xmath1 . note that if structures @xmath1 and @xmath57 are homomorphically equivalent , then for every structure @xmath2 there is a homomorphism from @xmath1 to @xmath2 if and only if there is a homomorphism from @xmath57 to @xmath2 ; in other words : the instances @xmath58 and @xmath59 of the decision csp are equivalent . however , the two instances may have vastly different sizes , and the complexity of solving the search and enumeration problems for them can also be quite different .
homomorphic equivalence is closely related to the concept of the core of a structure : a structure @xmath1 is a _ core _ if there is no homomorphism from @xmath1 to a proper substructure of @xmath1 .
a core of a structure @xmath1 is a substructure @xmath60 such that there is a homomorphism from @xmath1 to @xmath57 and @xmath57 is a core . obviously , every core of a structure is homomorphically equivalent to the structure .
we observe another basic fact about cores : [ obs : cores ] let @xmath1 and @xmath2 be homomorphically equivalent structures , and let @xmath57 and @xmath61 be cores of @xmath1 and @xmath2 , respectively .
then @xmath57 and @xmath61 are isomorphic . in particular , all cores of a structure @xmath1 are isomorphic
therefore , we often speak of _ the _ core of @xmath1 .
it is easy to see that it is np - hard to decide , given structures @xmath36 , whether @xmath1 is isomorphic to the core of @xmath2 .
( for an arbitrary graph @xmath62 , let @xmath1 be a triangle and @xmath2 the disjoint union of @xmath62 with @xmath1 .
then @xmath1 is a core of @xmath2 if and only if @xmath62 is 3-colorable . )
hell and neetil @xcite proved that it is co - np - complete to decide whether a graph is a core .
[ [ tree - decompositions . ] ] * tree decompositions .
* + + + + + + + + + + + + + + + + + + + + + + a _ tree decomposition _ of a graph @xmath62 is a pair @xmath63 , where @xmath64 is a tree and @xmath65 is a mapping that associates with every node @xmath66 a set @xmath67 such that ( 1 ) for every @xmath68 the set @xmath69 is connected in @xmath64 , and ( 2 ) for every @xmath70 there is a @xmath66 such that @xmath71 .
the sets @xmath72 , for @xmath66 , are called the _ bags _ of the decomposition .
it is sometimes convenient to have the tree @xmath64 in a tree decomposition rooted ; we always assume it is .
the _ width _ of a tree decomposition @xmath63 is @xmath73 .
the _ tree width _ of a graph @xmath62 , denoted by @xmath74 , is the minimum of the widths of all tree decompositions of @xmath62 .
we need to transfer some of the notions of graph theory to arbitrary relational structures . the _ gaifman graph _ ( also known as _ primal graph _ ) of a relational structure @xmath1 with vocabulary @xmath26 is the graph @xmath75 with vertex set @xmath1 and an edge between @xmath54 and @xmath56 if @xmath76 and there is a relation symbol @xmath28 , say , of arity @xmath29 , and a tuple @xmath77 such that @xmath78 .
we can now transfer graph - theoretic notions to relational structures .
in particular , a subset @xmath79 is _ connected _ in a structure @xmath1 if it is connected in @xmath75 .
tree decomposition _ of a structure @xmath1 can simply be defined to be a tree - decomposition of @xmath75 .
equivalently , a tree decomposition of @xmath1 can be defined directly by replacing the second condition in the definition of tree decompositions of graphs by ( 2 ) for every @xmath28 and @xmath77 there is a @xmath80 such that @xmath81 .
a class @xmath82 of structures has _ bounded tree width _ if there is a @xmath83 such that @xmath84 for all @xmath85 .
a class @xmath82 of structures has _ bounded tree width modulo homomorphic equivalence _ if there is a @xmath86 such that every @xmath85 is homomorphically equivalent to a structure of tree width at most w. [ obs : width - cores ] a structure @xmath1 is homomorphically equivalent to a structure of tree width at most @xmath87 if and only if the core of @xmath1 has tree width at most @xmath87 .
[ [ the - constraint - satisfaction - problem . ] ] * the constraint satisfaction problem . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for two classes @xmath88 and @xmath89 of structures , the _ constraint satisfaction problem _ , @xmath90 , is the following problem : the csp is a decision problem .
the variation of it we study in this paper is the following enumeration problem : we shall also refer to the search problem , @xmath91 , in which the goal is to find one solution to a csp - instance or output ` no ' if a solution does not exists .
if one of the classes @xmath88 , @xmath89 is the class of all finite structures , then we denote the corresponding csps by @xmath92 , @xmath93 ( respectively , @xmath94 , @xmath95 , @xmath96 , @xmath97 ) . the decision csp has been intensely studied . if a class @xmath82 of structures has bounded arity then @xmath98 is solvable in polynomial time if and only if @xmath82 has bounded tree width modulo homomorphic equivalence @xcite .
if the arity of @xmath82 is not bounded , several quite general conditions on a class of structures have been identified that guarantee polynomial time solvability of @xmath98 , see , e.g.@xcite .
problems of the form @xmath99 have been studied mostly in the case when @xmath82 is 1-element .
problems of this type are sometimes referred to as _ non - uniform_. it is conjectured that every non - uniform problem is either solvable in polynomial time or np - complete ( the so - called _ dichotomy conjecture _ ) @xcite .
although this conjecture is proved in several particular cases @xcite , in its general form it is believed to be very difficult .
a search csp is clearly no easier than the corresponding decision problem . while any non - uniform search problem @xmath100 is polynomial time reducible to its decision version @xmath99 @xcite , nothing
is known about the complexity of search problems @xmath101 except the result we state in section [ sec : tract ] .
paper @xcite provides some initial results on the complexity of non - uniform enumerating problems .
since even an easy csp may have exponentially many solutions , the model of choice for ` easy ' enumeration problems is algorithms with polynomial delay @xcite .
an algorithm alg is said to solve a csp _ with polynomial delay _
( _ wpd _ for short ) if there is a polynomial @xmath102 such that , for every instance of size @xmath13 , alg outputs ` no ' in a time bounded by @xmath102 if there is no solution , otherwise it generates all solutions to the instance such that no solution is output twice , the first solution is output after at most @xmath102 steps after the computation starts , and time between outputting two consequent solutions does not exceed @xmath102 .
if a class of relational structures @xmath82 has bounded arity , the aforementioned result of grohe @xcite imposes strong restrictions on enumeration problems solvable wpd .
[ obs : bw - core ] if a class of relational structures @xmath82 with bounded arity does not have bounded tree width modulo homomorphic equivalence , then @xmath103 is not wpd , unless p@xmath104np . unlike for the decision version ,
the converse is not true : bounded tree width modulo homomorphic equivalence does not imply enumerability wpd .
[ exa : second - solution ] let @xmath105 be the disjoint union of a @xmath106-clique and a loop and let @xmath107 . clearly , the core of each graph in @xmath88 has bounded tree width ( in fact , it is a single element ) , hence @xmath92 is polynomial - time solvable . for an arbitrary graph @xmath2 without loops , let @xmath61 be the disjoint union of @xmath2 and a loop .
it is clear that there is always a trivial homomorphism from @xmath105 ( for any @xmath108 ) to @xmath61 that maps everything into the loop .
there exist homomorphisms different from the trivial one if and only if @xmath2 contains a @xmath106-clique . thus if we are able to check in polynomial time whether there is a second homomorphism , then we are able to test if @xmath2 has a @xmath106-clique
. therefore , although @xmath92 and @xmath96 are polynomial - time solvable , a wpd enumeration algorithm for @xmath94 would imply @xmath109 .
it is not difficult to show that @xmath103 is enumerable wpd if @xmath82 has bounded tree width . for space restrictions
we do not include a direct proof and instead we derive it from a more general result in section [ sec : sequence ] .
thus enumerability wpd has a different tractability criterion than the decision version , and this criterion lies somewhere between bounded tree width and bounded tree width modulo homomorphic equivalence . thus in order to ensure that the solutions can be enumerated wpd , we have to make further restrictions on the way the structure can be mapped to its bounded tree width core .
the main new definition of the paper requires that the core is reached by `` small steps '' : let @xmath1 be a relational structure with universe @xmath27 .
we say that @xmath1 _ has a sequence of endomorphisms of width @xmath106 _ if there are subsets @xmath110 and homomorphisms @xmath111 such that 1 .
@xmath112 is a homomorphism from @xmath113 $ ] to @xmath114 $ ] , 2 .
@xmath115 for @xmath116 ; 3 .
if @xmath62 is the primal graph of @xmath1 , then the tree width of @xmath117 $ ] is at most @xmath106 for every @xmath118 ; 4 .
the structure induced by @xmath119 has tree width at most @xmath106 . in section [ sec : sequence ] , we show that enumeration for @xmath120 can be done wpd if a sequence of bounded width endomorphisms for @xmath1 is given in the input . unfortunately , we can not claim that @xmath94 can be done wpd if every structure in @xmath88 has such a sequence , since we do not know how to find such sequences efficiently .
in fact , as we show in section [ sec : hardness ] , it is hard to check if a width-1 sequence exists for a given structure . furthermore , we show a class @xmath88 where every structure has a width-2 sequence , but @xmath94 can not be done wpd , unless @xmath109 .
this means that it is not possible to get around the problem of not being able to find the sequences ( for example , by finding sequences with somewhat larger width or by constructing the sequence during the enumeration ) . thus having a bounded width sequence of endomorphisms is not the right tractability criterion .
we then investigate a more restrictive notion , where the bound is not on the tree width of the difference of the layers but on the number of elements in the differences .
however , in the rest of the section , we give evidence that enumeration problems solvable wpd can not be characterized in simple terms relying on tree width .
for instance , a description of search problems solvable in polynomial time would imply a description of non - uniform decision problems solvable in polynomial time .
this is shown via an analogous result for the search version of the problem , which might be of independent interest . by @xmath121
we denote the disjoint union of relational structures @xmath1 and @xmath2 .
[ lem : enum - vs - non - uniform ] let @xmath2 be a relational structure , which is a core , and let @xmath122 be @xmath123 . then @xmath124 is solvable in polynomial time if and only if so is the problem @xmath125 .
if the decision problem @xmath124 is solvable in polynomial time we can construct an algorithm that given an instance @xmath126 of @xmath127 computes a solution in polynomial time .
indeed , as @xmath124 is solvable in polynomial time by the aforementioned result of @xcite it is also polynomial time to find a homomorphism from a given structure to @xmath2 provided one exists .
if @xmath128 such a homomorphism @xmath129 exists by the definition of @xmath122 .
so our algorithms , first , finds some homomorphism @xmath129 .
then it decides by brute force whether or not there exists a homomorphism @xmath130 from @xmath2 to @xmath131 ( note that this can be done in polynomial time for every fixed @xmath2 ) .
if such a homomorphism does not exist then we can certainly guarantee that there is no homomorphism from @xmath1 to @xmath131 .
otherwise we obtain a required homomorphism @xmath132 as follows : let @xmath133 for @xmath134 , and @xmath135 for @xmath136 .
conversely , assume that we have an algorithm alg that finds a solution of any instance of @xmath127 in polynomial time , say , @xmath102 .
we construct from it an algorithm that solves @xmath124 .
given an instance @xmath120 of @xmath124 we call algorithm alg with input @xmath121 and @xmath2 .
additionally we count the number of steps performed by alg in such a way that we stop if alg has not finished in @xmath102 steps .
if alg produces a correct answer then we have to be able to obtain from it a homomorphism from @xmath1 to @xmath2 . if alg s answer is not correct or the clock reaches @xmath102 steps we know that alg failed .
the only possible reason for that is that @xmath121 does not belong to @xmath122 , which implies that @xmath1 is not homomorphic to @xmath2 . in what follows
we transfer this result to enumeration problems .
let @xmath88 be a class of relational structures .
the class @xmath137 consists of all structures built as follows : take @xmath138 and add to it @xmath139 independent vertices .
[ lem : search - enumeration ] let @xmath88 be a class of relational structures . then @xmath96 is solvable in polynomial time if and only if @xmath140 is solvable wpd .
if @xmath94 is enumerable wpd , then for any structure @xmath141 it takes time polynomial in @xmath142 to find the first solution . since @xmath57 is only twice of the size of the corresponding structure @xmath1 , it takes only polynomial time to solve @xmath96 .
conversely , given a structure @xmath143 , where @xmath138 and @xmath144 is the set of independent elements , and any structure @xmath2 .
the first homomorphism from @xmath57 to @xmath2 can be found in polynomial time , since @xmath96 is polynomial time solvable and the independent vertices can be mapped arbitrarily . let the restriction of this homomorphism onto @xmath1 be @xmath129 . then while enumerating all possible @xmath145 extensions of @xmath129 we buy enough time to enumerate all homomorphisms from @xmath1 to @xmath2 using brute force .
in this section we show that for every fixed @xmath106 , all the homomorphisms from @xmath1 to @xmath2 can be enumerated with polynomial delay if a sequence of width @xmath106 endomorphisms of @xmath1 is given in the input . given a sequence @xmath146 and @xmath147 , @xmath148 , @xmath149 as in the definition of a sequence of width @xmath106 endomorphisms ,
we denote @xmath114 $ ] by @xmath22 .
we will enumerate the homomorphisms from @xmath1 to @xmath2 by first enumerating the homomorphisms from @xmath20 , @xmath150 , @xmath148 to @xmath2 and then transforming them to homomorphisms from @xmath1 to @xmath2 using the homomorphisms @xmath112 .
we obtain the homomorphisms from @xmath22 by extending the homomorphism from @xmath151 to the set @xmath152 ; lemma [ lem : chain - extention ] below will be useful for this purpose . in order to avoid producing a homomorphism multiple times , we need a delicate classification ( see definitions of elementary homomorphisms and of the index of a homomorphism ) . [
lem : chain - extention ] let @xmath0 be relational structures and @xmath153 subsets , and let @xmath154 be a homomorphism from @xmath155 $ ] to @xmath65 .
for every fixed @xmath106 , there is a polynomial - time algorithm homomorphism - ext@xmath156 that decides whether @xmath154 can be extended to a homomorphism from @xmath157 $ ] to @xmath2 , if the tree width of induced subgraph @xmath158 $ ] of the gaifman graph of @xmath1 is at most @xmath106 .
the _ index _ of a homomorphism @xmath129 from @xmath1 to @xmath2 is the largest @xmath159 such that @xmath129 can be written as @xmath160 for some homomorphism @xmath132 from @xmath161 to @xmath2 . in particular ,
if @xmath129 can not be written as @xmath162 , then the index of @xmath129 is 0 .
observe that if the index of @xmath129 is at least @xmath159 , then there is a unique @xmath132 such that @xmath163 : this follows from the fact that @xmath164 is a surjective mapping from @xmath27 to @xmath165 , thus if @xmath166 and @xmath167 differ on @xmath165 , then @xmath168 and @xmath169 differ on @xmath27 .
a homomorphism @xmath132 from @xmath161 to @xmath2 is _ elementary _ , if it can not be written as @xmath170 .
a homomorphism is _ reducible _ if it is not elementary .
[ lem : elementary - index ] if a homomorphism @xmath132 from @xmath161 to @xmath2 is elementary , then @xmath160 has index exactly @xmath159 .
conversely , if homomorphism @xmath129 from @xmath1 to @xmath2 has index @xmath159 and can be written as @xmath163 , then the homomorphism @xmath132 from @xmath161 to @xmath2 is elementary .
lemma [ lem : elementary - index ] suggests a way of enumerating all the homomorphisms from @xmath1 to @xmath2 : for @xmath171 , we enumerate all the elementary homomorphisms from @xmath161 to @xmath2 , and for each such homomorphism @xmath132 , we compute @xmath172 . to this end , we need the following characterization of elementary homomorphisms : [ lem : hom - elementary ] a homomorphism @xmath132 from @xmath161 to @xmath2 is reducible if and only if 1 .
@xmath173 for every @xmath174 with @xmath175 , i.e. , for every @xmath176 , @xmath177 has the same value @xmath178 for every @xmath179 with @xmath180 , and 2 .
the mapping defined by @xmath181 is a homomorphism from @xmath182 to @xmath2 .
lemma [ lem : hom - elementary ] gives a way of testing in polynomial time whether a given homomorphism @xmath132 is elementary : we have to test whether one of the two conditions are violated .
we state this in a more general form : we can test in polynomial time whether a partial mapping @xmath154 can be extended to an elementary homomorphism @xmath132 , if the structure induced by the elements where @xmath154 is not defined has bounded tree width .
we fix values every possible way in which the conditions of lemma [ lem : hom - elementary ] can be violated and use homomorphism - ext to check whether there is an extension compatible with this choice . in order to efficiently enumerate all the possible violations of the second condition , the following definition is needed : given a relation @xmath183 of arity @xmath29 , a
_ bad prefix _ is a tuple @xmath184 with @xmath185 such that 1 .
there is no tuple @xmath186 for any @xmath187 , and 2 .
there is a tuple @xmath188 for some @xmath189 .
if @xmath190 , then there is a unique @xmath191 such that the tuple @xmath192 is a bad prefix : there has to be an @xmath193 such that @xmath194 can not be extended to a tuple of @xmath183 , but @xmath195 can .
[ lem : few - bad - prefixes ] the relation @xmath183 has at most @xmath196 bad prefixes , where @xmath29 is the arity of the relation .
[ lem : elementary - ext ] let @xmath197 be a subset of @xmath165 and let @xmath154 be a mapping from @xmath197 to @xmath65 .
for every fixed @xmath106 , there is a polynomial - time algorithm elementary - ext@xmath198 that decides whether @xmath154 can be extended to an elementary homomorphism from @xmath161 to @xmath65 , if the tree width of the structure induced by @xmath199 is at most @xmath106 .
we enumerate the elementary homomorphisms in a specific order defined by the following precedence relation .
let @xmath129 be an elementary homomorphism from @xmath22 to @xmath2 and let @xmath132 be an elementary homomorphism from @xmath200 to @xmath2 for some @xmath201 .
homomorphism @xmath132 is the _ parent _ of @xmath129 ( @xmath129 is a _ child _ of @xmath132 ) if @xmath129 restricted to @xmath202 can be written as @xmath203 . _
ancestor _ and _ descendant _ relations are defined as the reflexive transitive closure of the parent and child relations , respectively .
note that an elementary homomorphism from @xmath22 to @xmath2 has exactly one parent for @xmath204 and a homomorphism from @xmath20 to @xmath2 has no parent .
fix an arbitrary ordering of the elements of @xmath27 . for
@xmath205 and @xmath206 , let @xmath207 be the union of @xmath202 and the first @xmath208 elements of @xmath152 .
note that @xmath209 and @xmath210 .
[ lem : elementary - enum ] let @xmath132 be a mapping from @xmath207 to @xmath2 that can be extended to an elementary homomorphism from @xmath22 to @xmath2 .
assume that a sequence of width @xmath106 endomorphisms is given for @xmath1 .
for every fixed @xmath106 , there is a polynomial - delay , polynomial - space algorithm elementary - enum@xmath211 that enumerates all the elementary homomorphisms of @xmath22 that extends @xmath132 and all the descendants of these homomorphisms . by calling elementary - enum@xmath212 ( where @xmath154 is a trivial mapping from @xmath213 to @xmath2 )
, we can enumerate all the elementary homomorphisms . by the observation in lemma [ lem : elementary - index ] , this means that we can enumerate all the homomorphisms from @xmath1 to @xmath2 .
[ the : width - k - chain ] for every fixed @xmath106 , there is a polynomial - delay , polynomial - space algorithm that , given structures @xmath1 , @xmath2 , and a sequence of width @xmath106 endomorphisms of @xmath1 , enumerates all the homomorphisms from @xmath1 to @xmath2 .
theorem [ the : width - k - chain ] does not provide a complete description of classes of structures solvable wpd .
[ cor : unbounded - width ] there is a class @xmath88 of relational structures such that not all structures from @xmath88 have a sequence of width @xmath106 endomorphisms and @xmath94 is solvable wpd . let @xmath88 be the class of structures that are the disjoint union of a loop and a core .
obviously , @xmath96 is polynomial time solvable .
therefore , by lemma [ lem : search - enumeration ] , @xmath140 is solvable with polynomial delay .
however , it is not hard to see that @xmath137 does not have a sequence of endomorphisms of bounded tree width .
furthermore , as we will see in the next section it is hard , in general , to find a sequence of bounded width endomorphims . still , we can find a sequence of endomorphisms for a structure @xmath1 if we impose two more restrictions on such a sequence .
a retraction @xmath129 of a structure @xmath1 is called a _ @xmath106-retraction
_ if at most @xmath106 nodes change their value according to @xmath129 .
a structure is a _
@xmath106-core _ if the only @xmath106-retraction is the identity .
a @xmath106-core of a structure is any @xmath106-core obtained by a sequence of @xmath106-retractions .
[ lem : k - cores - isomorphic ] all @xmath106-cores of a structure @xmath1 are isomorphic .
lemma [ lem : k - cores - isomorphic ] amounts to say that when searching for a sequence of @xmath106-retractions converging to a @xmath106-core we can use the greedy approach and include , as the next member of such a sequence , any @xmath106-retraction with required properties . with this in hands we now can apply theorem [ the : width - k - chain ] .
the : finite - retractions ] let @xmath214 be a positive integer and let @xmath82 be a class of structures such that the @xmath106-core of every structure in @xmath82 has tree width at most @xmath106 .
then , the enumeration problem @xmath103 is solvable wpd .
if @xmath82 is a class of structures of bounded tree width then @xmath103 is solvable wpd .
the first result of this section shows that finding a sequence of endomorphisms of bounded width can be difficult even in simplest cases . [ the : hardness ]
it is np - complete to decide if a structure has a sequence of 1-width retractions to the core .
the second result shows that @xmath94 can be hard even if every structure in @xmath88 has a sequence of width-2 endomorphisms .
note that this result is incomparable with theorem [ the : hardness ] , since an enumeration algorithm ( in theory ) does not necessarily have to compute an sequence of endomorphisms .
we need the following lemma : [ lem : outerplanar ] if @xmath62 is a planar graph , then it is possible to find a partition @xmath215 of its vertices in polynomial time such that @xmath216 $ ] and @xmath217 $ ] have tree width at most @xmath218 .
[ pro : fpt - not - poly ] there is a class @xmath88 of relational structures such that every structure from @xmath88 has a sequence of width 2 endomorphisms to the core , and such that the problem @xmath94 is not solvable wpd , unless @xmath219 .
let @xmath88 be a class of graphs built in the following way .
take a 3-colorable planar graph @xmath62 and its partition @xmath215 according to lemma [ lem : outerplanar ] . using colorings
we can ensure that @xmath62 is a core .
then we take a disjoint union of this graph with a triangle @xmath64 having all the colors and a copy @xmath220 of @xmath216 $ ] .
let @xmath1 denote the resulting structure .
@xmath1 has a sequence of width-2 endomorphisms .
+ let @xmath132 be a 3-coloring of @xmath62 that is a homomorphism into the triangle , and @xmath166 the bijective mapping from @xmath220 to @xmath216 $ ] .
then @xmath147 is defined to act as @xmath132 on @xmath62 , as @xmath166 on @xmath221 and identically on @xmath64 .
endomorphism @xmath222 is just the 3-coloring of @xmath223 induced by @xmath132 .
the images of @xmath147 and @xmath222 are @xmath224 $ ] and @xmath64 , respectively , so all the conditions on a sequence of width-2 homomorphisms are easily checkable . claim 2 .
the planar graph 3-coloring problem is turing reducible to @xmath94 . + given a planar graph @xmath62
we find its partition @xmath215 and create a structure @xmath1 , as described above .
then we apply an algorithm that enumerates solutions to @xmath94 we may assume that such an algorithm stops with some time bound regardless whether @xmath62 is 3-colorable or not . if the algorithm succeeds we can now produce a 3-coloring of @xmath62 .
when making a query to a database one usually needs to obtain values of only those variables ( attributes ) ( s)he is interested in . in terms of homomorphisms
this can be translated as follows : for relational structures @xmath1 , @xmath2 , and a subset @xmath225 , we aim to list those mappings from @xmath226 to @xmath65 which can be extended to a full homomorphism from @xmath1 to @xmath2 . in other words , we would like to enumerate all the mappings from @xmath226 to @xmath65 that arise as the restriction of some homomorphism from @xmath1 to @xmath2 .
clearly , this problem significantly differs from the regular enumeration problem .
a mapping from @xmath226 to @xmath65 can be extendible to a homomorphism in many ways , possibly superpolynomially many , and an enumeration algorithm would list all of them . in the worst case scenario it would list them before turning to the next partial mapping .
if this happens it may destroy polynomiality of the delay between outputting consecutive solutions .
we present two results , first one of them shows that the problem @xmath227 is wpd when @xmath88 is a class of structures of bounded tree width , the second one claims that , modulo some complexity assumptions , in contrast to enumeration problems this can not be generalized to structures with @xmath106-cores of bounded tree width for @xmath228 . _ step 1 _ * set * @xmath231 , @xmath232 , @xmath233 , @xmath234 $ ] , complete@xmath235*false * + _ step 2 _ not complete * do * + _ step 2.1 _ * if * @xmath236 * then do * + _ step 2.1.1 _ * search * @xmath237 until a @xmath238 is found such that there exists a homomorphism extending + @xmath239 and * remove * all members of @xmath237 preceding @xmath56 inclusive + _ step 2.1.2 _ such a @xmath56 exists * then set * @xmath240 , @xmath241 + _ step 2.1.3 _ + _ step 2.1.3.1_ * if * @xmath242 * then set * @xmath243 and @xmath244 , @xmath245 + _ step 2.1.3.2_ complete:=*true * + _ step 2.2 _ + _ step 2.2.1 _ @xmath129 + _ step 2.2.2 _ @xmath246 , @xmath247 + [ exa : multi - col - clique ] recall that the multicolored clique problem ( cf .
@xcite ) is formulated as follows : given a number @xmath106 and a vertex @xmath106-colored graph , decide if the graph contains a @xmath106-clique all vertices of which are colored different colors .
this problem is @xmath248$]-complete , i.e. , has no time @xmath249 algorithm for any function @xmath43 and constant @xmath250 , unless fpt@xmath251 $ ] .
we reduce this problem to @xmath227 where @xmath88 is the class of structures whose 2-cores are 2-element described below .
let us consider relational structures with two binary and two unary relations .
this structure can be thought of as a graph whose vertices and edges have one of the two colors , say , red and blue , accordingly to which of the two binary / unary relations they belong to .
let @xmath105 be the relational structure with universe @xmath252 , where @xmath44 are red while @xmath253 are blue .
then @xmath254 induces a red clique , that is every @xmath255 ( @xmath256 are not necessarily different ) are connected with a red edge , and each @xmath257 is connected to @xmath258 with a blue edge .
it is not hard to see that every pair of a red and blue vertices induces a 2-core of this structure .
set @xmath259 .
the reduction of the multicolored clique problem to @xmath227 goes as follows .
given a @xmath106-colored graph @xmath260 whose coloring induces a partition of @xmath261 into classes @xmath262 .
then we define structures @xmath0 and a set @xmath225 .
we set @xmath263 , @xmath264 .
then let @xmath265 , the elements of @xmath261 are colored red and the induced substructure @xmath266 $ ] is the graph @xmath62 ( without coloring ) whose edges are colored also red .
finally , @xmath267 are made blue and each @xmath268 is connected with a blue edge with every vertex from @xmath269 .
it is not hard to see that any homomorphism maps @xmath254 to @xmath261 and @xmath226 to @xmath270 , and that the number of homomorphisms that do not agree on @xmath226 does not exceed @xmath271 .
moreover , @xmath62 contains a @xmath106-colored clique if and only if there is a homomorphism from @xmath1 to @xmath2 that maps @xmath226 onto @xmath270 .
if there existed an algorithm solving @xmath227 wpd , say , time needed to compute the first and every consequent solution is bounded by a polynomial @xmath102 , then time needed to list all solutions is at most @xmath272 .
this means that multicolored clique is fpt , a contradiction .
|
the homomorphism problem for relational structures is an abstract way of formulating constraint satisfaction problems ( csp ) and various problems in database theory .
the decision version of the homomorphism problem received a lot of attention in literature ; in particular , the way the graph - theoretical structure of the variables and constraints influences the complexity of the problem is intensively studied .
here we study the problem of enumerating all the solutions with polynomial delay from a similar point of view .
it turns out that the enumeration problem behaves very differently from the decision version .
we give evidence that it is unlikely that a characterization result similar to the decision version can be obtained .
nevertheless , we show nontrivial cases where enumeration can be done with polynomial delay .
= = = = = = = = andrei a. bulatov vctor dalmau martin grohe dniel marx
|
in @xcite , vaughan jones introduced the notion of index for type @xmath2 subfactors . to any finite index subfactor @xmath3 one can associate a tower of @xmath4 factors @xmath5 .
the standard invariant of the subfactor is then given by the grid of finite dimensional algebras of relative commutants ( see @xcite , @xcite , @xcite , @xcite ) @xmath6 sorin popa in @xcite studied the question of which families @xmath7 of finite - dimensional @xmath8-algebras could arise as the tower of relative commutants of an extremal finite - index subfactor , that is , when does there exist such a subfactor @xmath9 such that @xmath10 .
he obtained a beautiful algebraic axiomatization of such families , which he called @xmath11__-lattices__. ocneanu gave a combinatorial description of the standard invariant as so called _ paragroups _ ( see @xcite ) .
subsequently , jones gave a geometric reformulation of the standard invariant , which he called _
planar algebras _
( see @xcite ) .
jones then introduced the notion of ` modules over a planar algebra ' in @xcite and computed the irreducible modules over the temperley - lieb planar algebras for index greater than @xmath12 .
planar algebras became a powerful tool to construct subfactors of index less than @xmath12 .
in particular , a new construction of the subfactors with principal graph , @xmath13 and @xmath14 could be given ( see @xcite ) .
the author ( see @xcite ) established a one - to - one correspondence of all modules over the group planar algebra , that is , the planar algebra associated to the subfactor arising from the action of a finite group , and the representations of a non - trivial quotient of the quantum double of the group over a certain ideal .
the reason for the appearance of a quotient of the quantum double instead of just the quantum double was allowing rotation of internal discs in the definition of the modules over a planar algebra .
similar results also appeared in the field of tqfts .
kevin walker and michael freedman proved that the representations of the _ annularization _ of a tensor category satisfying suitable conditions that allow one to perform the reshetikhin - turaev construction of tqft , is equivalent to the representations of the quantum double of the category .
the author ( in @xcite ) also showed that the radius of convergence of the dimension of a module is at least as big as the inverse - square of the modulus in the case of group planar algebras and thus answering a question in @xcite .
subfactors have been extensively studied from the point of view of the associated bicategory of @xmath15 , @xmath16 , @xmath17 , @xmath18 bimodules ( see for example , @xcite , @xcite , @xcite , @xcite , @xcite ) .
it is natural to expect a correspondence between the bicategory and the planar algebra associated to the subfactor .
one of the main objectives of this paper is to construct a planar algebra directly from a bicategory . from @xcite
, it follows that if the modules over a planar algebra are defined with rigid internal disc then they are more interesting because of the connection with quantum double in the case of group planar algebras .
another objective is to find such modules ( called _ affine representations _ ) and prove finiteness results of affine representation for finite depth planar algebras .
next , we give a section - wise summary of the paper ; all results in this paper appeared in a phd thesis ( 2006 ) of the author submitted in university of new hampshire .
in the first section , we discuss the preliminaries from basic category theory .
the first subsection recalls the definition of multicategories and maps between them from @xcite .
we introduce the notion of the stucture of _ empty objects _ in a multicategory ; the trivial example , namely , the multicategory of sets or vector spaces possess the structure of empty objects . in the second subsection , we discuss basics of bicategory theory and several structures related to a bicategory , namely functors , transformation between functors and rigidity .
we construct a new example of a multicategory with the structure of empty objects which we call _ planar tangle multicategory _ in the second section .
we re - define jones s planar algebra simply as a map of multicategories from the planar tangle multicategory to the multicategory of vector spaces ; in fact , this was motivated by jones s idea of putting the planar algebra as well as its dual in the definition itself . in the end , we discuss more structures ( modulus , connectedness , local finiteness , @xmath8- structure , etc . ) on a planar algebra . in the third section , we start with fixing a @xmath0-cell in a pivotal strict @xmath1-category and construct a planar algebra .
some of the techniques used here are similar to jones s construction of a planar algebra from a subfactor .
however , we would like to mention that this construction is totally algebraic and heavily depends on the _ graphical calculus _ of the @xmath1-cells and the pivotal structure plays a key role here .
motivated with the connection of annular representation of the group planar algebra with the representations of a certain quotient of the quantum double of the group , we considered _ affine representations _ of a planar algebra in the fourth section ; this was introduced by jones and reznikoff in @xcite and graham and lehrer in @xcite .
we also discuss the general theory of such representations . in the fifth and the final section
, we discuss affine representations of a planar algebra associated to a finite depth subfactor .
we find a bound on the weights of these representations which is dependent on the depth of the planar algebra .
we also prove that at each weight , the number of isomorphism classes of irreducible affine representations is finite .
we answer jones s question on the radius of convergence of the dimension of affine representations for finite depth subfactor planar algebras .
0.5 cm * acknowledgement * : the author is grateful to professor dmitri nikshych for teaching him all he knows about bicategories and multicategories , and many useful discussions . the author would like to thank professors v. s. sunder and vijay kodiyalam for their constant help in working on ` modules over planar algebras ' which lead to the results in section [ fdpa ] , and also ved prakash gupta for numerous suggestions and corrections .
in this subsection , we revisit the definition of _ multicategory _ and an _ algebra for a multicategory _ ( introduced in @xcite ) .
we introduce the structure of _ empty object _ in a multicategory which will be useful in the subsequent sections .
a multicategory @xmath19 consists of : + ( i ) a class @xmath20 whose elements are called _ objects _ of @xmath19 , + ( ii ) for all @xmath21 , there exists a class @xmath22 whose elements are called _ morphisms _ or _ arrows _ from @xmath23 to @xmath24 , together with a distinguished arrow @xmath25 called _ identity morphism _ for @xmath24 , + ( iii ) for all @xmath26 where @xmath27 , there exists a _ composition map _
@xmath28 denoted in the following way : @xmath29 + where @xmath30 satisfies the following conditions : \(a ) _ associativity axiom _ : @xmath31 + @xmath32 whenever the composites make sense , \(b ) _ identity axiom _ : @xmath33 for all @xmath34 .
the associativity and identity axioms are easier to understand with pictorial notation of arrows ( see @xcite ) . the collection of sets @xmath35 ( resp .
vector spaces @xmath36 ) forms a multicategory where arrows are given by maps from cartesian product of finite collection of sets to another set ( resp .
multilinear maps from a finite collection of vector spaces to another vector space ) .
any tensor category @xmath19 has an inbuilt multicategory structure in the obvious way by setting @xmath37 set of objects of @xmath19 and @xmath38 .
let @xmath19 and @xmath39 be multicategories .
a map of multicategories @xmath40 consists of a map @xmath41 together with another map @xmath42 such that composition of arrows and identities are preserved .
( if @xmath19 and @xmath39 are multicategories with each morphism space being vector space and composition being multilinear , then we will assume that the map of multicategories is linear between the morphism spaces . )
let @xmath19 be a multicategory .
a @xmath19-algebra is simply a map of multicategories from @xmath19 to @xmath35 .
( if @xmath19 is a multicategory with each morphism space being vector space and composition being multilinear , then we will consider a @xmath19-algebra to be a map of multicategories from @xmath19 to @xmath36 . )
a multicategory @xmath19 is said to be _ symmetric _ if the following conditions hold : * for all @xmath43 , @xmath44 , @xmath45 , @xmath46 , there exists a map @xmath47 ( where @xmath48)satisfying : * @xmath49 and @xmath50 for all @xmath43 , @xmath44 , @xmath45 , @xmath51,@xmath52 @xmath53 , * @xmath54 + for all @xmath55,@xmath56 , @xmath45 , @xmath57 , @xmath58 , @xmath53 , @xmath59 , @xmath46 , @xmath60 , for @xmath61 , where @xmath62 and @xmath63 are permutations in @xmath64 defined by : + @xmath65 + for all @xmath61 , @xmath66 assuming @xmath67 .
it will be easier to understand the axioms of symmetricity in pictorial notation as in @xcite .
clearly , the multicategories @xmath35 , @xmath36 and the one arising from a symmetric tensor category are symmetric .
a multicategory @xmath19 is said to have the structure of _ empty object _ if for all @xmath45 , there exists a class @xmath68 such that the composition in @xmath19 extends in the following way : + for all @xmath43 , @xmath69 , @xmath45 , @xmath57 , @xmath70 , @xmath71 , and + for all @xmath56 , @xmath58 , @xmath59 where @xmath72 , + @xmath73 + such that this composition map is associative and @xmath74 for all @xmath75 .
both @xmath35 and @xmath36 indeed have the structure of empty object ; for instance , @xmath76 for any set @xmath77 .
we demand that a map of multicategories both having the structure of empty object , should preserve this structure . in this subsection
, we will recall the definition of _ bicategories _ and various other notions related to bicategories which will be useful in section 3 .
most of the materials in this section can be found in any standard textbook on bicategories .
a bicategory @xmath78 consists of : * a class @xmath79 whose elements are called objects or @xmath80-cells , * for each @xmath81 , @xmath82 , there exists a category @xmath83 whose objects @xmath84 are called @xmath0-cells of @xmath78 and denoted by @xmath85 and whose morphisms @xmath86 are called @xmath1-cells of @xmath78 and denoted by @xmath87 where @xmath88 , @xmath89 are @xmath0-cells in @xmath90 , * for each @xmath81 , @xmath91 , @xmath92 , there exists a functor @xmath93 , * identity object : for each @xmath94 , there exists an object @xmath95 ( the identity on @xmath81 ) , * associativity constraint : for each triple @xmath96 , @xmath97 , @xmath98 of @xmath0-cells , there exists an isomorphism @xmath99 in @xmath100 , * unit constraints : for each @xmath0-cell @xmath96 , there exist isomorphisms @xmath101 and @xmath102 in @xmath103 such that @xmath104 , @xmath105 and @xmath106 are natural in @xmath107 , @xmath108 , @xmath84 , and satisfy the pentagon and the triangle axioms ( which are exactly similar to the ones in the definition of a tensor category ) . on a bicategory @xmath78
, one can perform the operation @xmath109 ( resp .
@xmath110 ) and obtain a new bicategory @xmath111 ( resp .
@xmath112 ) by setting ( i ) @xmath113 , ( ii ) @xmath114 as categories ( where @xmath109 of a category is basically reversing the directions of the morphisms ) . a bicategory will be called a _ strict _ @xmath1__-category _ _ if the associativity and the unit constraints are identities
. an _ abelian _ ( resp .
_ semisimple _ ) bicategory @xmath78 is a bicategory such that @xmath83 is an abelian ( resp .
semisimple ) category for every @xmath81 , @xmath115 and the functor @xmath116 is additive .
@xmath117 is a tensor category and @xmath83 is a @xmath118-bimodule category for @xmath80-cells @xmath81 , @xmath91 .
( see @xcite , @xcite for definition of module category . )
a bicategory with only one @xmath80-cell is simply a tensor category .
a bicategory can be obtained by taking rings as @xmath80-cells , @xmath0-cells @xmath119 being @xmath120-bimodules and @xmath1-cells being bimodule maps .
the tensor functor is given by the obvious tensor product over a ring .
let @xmath78 , @xmath121 be bicategories .
a _ weak functor _
@xmath122 consists of : * a function @xmath123 , * for all @xmath81,@xmath115 , there exists a functor @xmath124 written simply as @xmath125 , * for all @xmath81 , @xmath91 , @xmath92 , there exists a natural isomorphism @xmath126 written simply as @xmath127 ( where @xmath116 and @xmath128 are the tensor products of @xmath78 and@xmath129 respectively ) , * for all @xmath94 , there exists an invertible ( with respect to composition ) @xmath1-cell @xmath130 , satisfying commutativity of certain diagrams ( consisting of @xmath1-cells ) which are analogous to the hexagonal and rectangular diagrams appearing in the definition of a tensor functor .
let @xmath131 , @xmath132 be weak functors .
a weak transformation @xmath133 consists of : * for all @xmath94 , there exists a @xmath0-cell @xmath134 , * for all @xmath81 , @xmath115 , there exists a natural transformation @xmath135 written simply as @xmath51 ( where @xmath136 , @xmath137 are functors defined in the obvious way ) , satisfying the follwoing : * for all @xmath138 , @xmath139 where @xmath81 , @xmath91 , @xmath92 , the following diagrams commute:@xmath140 @xmath141 where @xmath142 , @xmath143 are the left and right unit constraints of @xmath121 .
composition of weak functors and weak transformations follows exactly from composition of functors and natural transformations in categories .
one can also extend the notion of natural isomorphisms in categories to weak isomorphism in bicategories .
( coherence theorem for bicategories ) let @xmath78 be a bicategory . then there exists a strict @xmath1-category @xmath121 and functors @xmath144 , @xmath145 such that @xmath146 ( resp .
@xmath147 ) is weakly isomorphic to @xmath148 ( resp .
@xmath149 ) .
see @xcite for a proof .
let @xmath96 be a @xmath0-cell in a bicategory @xmath150 .
a _ right _
_ left _ ) _ dual of _
@xmath84 is an @xmath0-cell @xmath151 ( resp .
@xmath152 ) such that there exists @xmath1-cells @xmath153 and @xmath154 ( resp .
@xmath155 and @xmath156 ) such that the following identities ( ignoring the associativity and unit contraints ) are satisfied : @xmath157 ( here @xmath158 stands for evaluation and @xmath159 stands for coevaluation . ) one can show that two right ( resp .
left ) duals are isomorphic via an isomorphism which is compatible with the evaluation and coevaluation maps .
a bicategory is said to be _ rigid _ if right and left dual exists for every @xmath0 -cell .
further , in a rigid bicategory @xmath78 , one can consider right dual as a weak funtor @xmath160 in the following way : * for each @xmath0-cell @xmath84 , we fix a triple @xmath161 so that when @xmath162 where @xmath94 , then @xmath163 , @xmath164 ( @xmath165 , see @xcite for proof ) , @xmath166 , * @xmath167 induces identity map on @xmath79 , * for all @xmath81 , @xmath115 , @xmath84 , @xmath168 and @xmath1-cell @xmath169 , define the contravariant functor @xmath170 by @xmath171 and @xmath172 denoted by @xmath173 , is given by the composition of the following @xmath1-cells @xmath174 * for all @xmath81 , @xmath91 , @xmath92 , the natural isomorphism @xmath175 is defined by : + for @xmath176 , @xmath177 , the invertible @xmath1-cell @xmath178 is given by the composition of the following @xmath1 -cells + @xmath179 ignoring the associativity and the unit constraints necessary to make sense of the composition , * for all @xmath94 , the invertible @xmath1-cell @xmath180 is given by identity morphism on @xmath181 . similarly , one can define a left dual functor in a rigid bicategory .
in this section , we will introduce a new example of a symmetric multicategory , namely , the _ planar tangle multicategory _ ( @xmath182 ) which possesses the additional structure of empty object__. _ _ any _ planar algebra _ , in the sense of @xcite , turns out to be a @xmath183-algebra . in the end
, we also exhibit some examples and define more structures on a planar algebra .
let us first define _ planar tangles _ which are the building blocks of the planar tangle multicategory .
fix @xmath184 and @xmath185 [ def of pln tang ] a @xmath186-planar tangle is an isotopy class of pictures containing : * an external disc @xmath187 on the euclidean plane @xmath188 with @xmath189 distinct points on the boundary numbered clockwise , * finitely many ( possibly zero ) non - intersecting internal discs @xmath190 , @xmath191 , lying in the interior of @xmath187 with @xmath192 distinct points on the boundary of @xmath193 numbered clockwise where @xmath194 for @xmath61 , * a collection @xmath195 of smooth non - intersecting oriented curves ( called strings ) on @xmath196 $ ] such that : + \(a ) each marked point on the boundaries of @xmath197 is connected to exactly one string , + \(b ) each string either has no end - points or has exactly two end - points on the marked points , + \(c ) the orientations induced on each connected component of @xmath198 $ ] by different bounding strings should be same , * the orientation induced in the connected component of @xmath198 $ ] , adjacent to the first and the last marked point on the boundary of @xmath187 , should have orientation positive ( anti - clockwise ) or negative ( clockwise ) according to the sign of @xmath199 . for each @xmath200 , we can assign @xmath201 to the internal disc @xmath193 depending on the orientation of the connected component of @xmath202 $ ] , adjacent to the first and the last marked points on the boundary of @xmath193 .
@xmath203 will be called the colour of @xmath193 and @xmath204 will be the colour of @xmath205 . sometimes , instead of numbering each marked point on the boundary of a disc with colour @xmath204 , we will write @xmath206 very close to the boundary of the disc and in the connected component adjacent to the first and the last points .
the orientation of the strings is equivalent to putting checkerboard shading on the connected components such that all components with negative orientation get shaded .
-planar tangle with two internal discs @xmath190 , @xmath207 with colours @xmath208 , @xmath209 respectively . ]
+ let @xmath210 be the set of @xmath204-planar tangles with @xmath55 internal discs @xmath211 with colours @xmath212 , @xmath213 , @xmath214 , @xmath215 respectively , @xmath216 be the set of @xmath204-planar tangles with no internal disc and @xmath217 be the set of all @xmath204-planar tangles .
the composition of two tangles @xmath218 and @xmath219 ( resp .
@xmath220 ) , denoted by @xmath221 ( resp .
@xmath223 ) , is obtained by gluing the external boundary of @xmath224 with the boundary of the @xmath225 internal disc of @xmath226 preserving the marked points on either of them with the help of isotopy , and then erasing the common boundary .
the _ planar tangle multicategory _ , denoted by @xmath183 , is defined as : * _ objects _ : @xmath227 , * _ morphisms _ : @xmath228 ( resp .
@xmath229 ) is the vector space generated by @xmath210 ( resp .
@xmath216 ) as a basis , * composition of morphisms is given by the multilinear extension of the composition of tangles as described above , * the identity morphism @xmath230 is given by the @xmath231-planar tangle with exactly one internal disc with colour @xmath204 , containing precisely @xmath232 strings such that @xmath225 point on the internal disc is connected to the @xmath225 point on the external disc by a string for @xmath233 . ]
we leave the checking of associativity and identity axioms to the reader . a moment
s observation also reveals that @xmath183 is symmetric and has the structure of empty object .
a planar algebra @xmath234 is a @xmath183-algebra , that is , a map of multicategories from @xmath183 to @xmath36 .
the first natural example of planar algebra is the @xmath183-algebra which takes the object @xmath231 to @xmath235 , and morphisms @xmath226 to the multilinear map given by left - compostion of @xmath226 .
this is called the _ universal planar algebra _ in @xcite . for a planar algebra @xmath234
, the collection of vector spaces @xmath236 forms a unital filtered algebra where @xmath237 .
the multiplication of @xmath238 , inclusion of @xmath239 inside @xmath240 and identity of @xmath238 are induced by the following tangles : -8 mm respectively .
we will now define more structures on a planar algebra .
a planar algebra @xmath234 is said to be _ connected _ ( resp .
_ locally finite _ ) if @xmath241 ( resp .
@xmath242 for all @xmath204 ) .
a planar algebra @xmath234 is said to have _ modulus _
@xmath243 if @xmath244 ( resp .
@xmath245 ) where @xmath226 is a planar tangle with a contractible loop oriented clockwise ( resp . anti - clockwise ) and @xmath246 is the tangle @xmath226 with the loop removed .
a connected planar algebra @xmath234 is called _ spherical _ if two tangles @xmath247 and @xmath248 induce the same multilinear functional by expressing the images of @xmath249 and @xmath250 as scalar multiples of the identities of @xmath251 and @xmath252 respectively whenever one can obtain @xmath246 from @xmath253 after embedding them on the unit sphere and using spherical isotopy .
if @xmath254 is a planar algebra in the sense of jones ( @xcite ) with modulus @xmath255 ( where @xmath256 denotes the action of a tangle @xmath226 necessarily with positive colors for all discs in it ) , then one can define a planar algebra @xmath257 via : \(i ) @xmath258 + where the tangle @xmath259 is given in figure [ lcek ] , \(ii ) for @xmath260 , first define @xmath261 where @xmath262 , and @xmath263 and @xmath264 are given by the tangles in figure [ leftcapcup ] .
note that @xmath265 is a tangle with positive colors on each of its internal discs .
set @xmath266 .
it is routine to check that @xmath234 preserves composition and identity .
the definition of @xmath234 as a map of multicategories is motivated by jones s definition of dual planar algebra ( see @xcite ) . if @xmath267 ( resp .
@xmath268 ) , then @xmath269 ( resp .
@xmath270 ) is defined as the tangle obtained by reflecting @xmath226 about any straight line not intersecting @xmath226 , and the first point of an internal ( resp .
external ) disc of the reflected @xmath226 is taken to be the reflected point of the last point of the corresponding internal ( resp .
external ) disc in @xmath226 such that the reflection preserves the colour of each disc .
for example , the @xmath271 of the tangle in figure [ egtang ] is given by the tangle in figure [ stareg ] where we reflect the figure [ egtang ] about a vertical line .
we extend the map @xmath272 conjugate linearly to @xmath273 .
it is clear that @xmath167 is an involution .
this makes @xmath274 into a unital filtered @xmath275-algebra for @xmath237 .
@xmath234 is said to be a @xmath167 _ -planar algebra _
@xmath8__-planar algebra _
_ ) if @xmath234 is a planar algebra , @xmath238 is a @xmath167-algebra ( resp .
@xmath276-algebra ) for each colour @xmath204 and the map @xmath234 is @xmath275 preserving in the sense : + if @xmath277 ( resp .
@xmath278 ) and @xmath279 for @xmath61 , then @xmath280 .
a locally finite spherical @xmath281-planar algebra is called _ subfactor - planar algebra_. [ jonthm ] ( jones ) any extremal subfactor with finite index gives rise to a subfactor - planar algebra .
conversely , any subfactor planar algebra gives rise to an extremal subfactor with finite index .
jones proved the first part of theorem [ jonthm ] ( in @xcite ) by prescribing an action of tangles on the standard invariant of a subfactor .
however , jones proved the converse using popa s result on @xmath282-lattices ( @xcite ) . very recently , another proof of the converse using planar algebra techniques appeared in @xcite and then in @xcite .
in this section , we will show how one can construct a planar algebra from a @xmath0-cell of an abelian ` pivotal ' strict @xmath1-category with exactly two @xmath80- cells .
the techniques used in this construction are motivated by jones s construction of planar algebra from a subfactor ( in @xcite ) . before we proceed towards the construction , we will first state or deduce some useful results and set up some notations .
a bicategory @xmath78 is called _ pivotal _ if @xmath78 is rigid and there exists a weak transformation @xmath283 such that @xmath284 for all @xmath285 , where @xmath286 is the right dual functor and @xmath287 is the weak functor @xmath288 . from now on
, we will consider only strict @xmath1-category instead of general bicategories unless otherwise mentioned ; however all results modified with appropriate associativity and unit constraints , will hold even in the absence of the ` strict ' assumption by the coherence theorem for bicategories .
r40 mm we next set up some pictorial notation to denote @xmath1-cells which is analogous to the graphical calculus of morphisms in a tensor category ( see @xcite , @xcite ) .
let @xmath78 be a pivotal strict @xmath1-category as defined above .
we denote a @xmath1-cell @xmath289 by a rectangle labelled with @xmath84 , placed on @xmath188 so that one of the sides is parallel to the @xmath77-axis and a vertical line segment labelled with @xmath290 ( resp .
@xmath291 ) is attached to the top ( resp .
bottom ) side of the rectangle .
sometimes we will not label the strings attached to a recatangle labelled with a @xmath1-cell ; the @xmath1-cell itself will induce the obvious labelling to the strings .
we list below pictorial notations of several other @xmath1-cells which will be the main constituents of the construction without describing them meticulously in words like the way we described @xmath84 above . where @xmath290 , @xmath291 are @xmath0-cells and @xmath84 , @xmath108 are @xmath1-cells .
to each local maximum or minimum of a string with an orientation marked at the maximum or minimum and labelled with a @xmath0-cell @xmath292 , we associate a @xmath1-cell in the following way : we will next exhibit some easy consequences in terms of the pictorial notation . [ basic lemma](i ) for any @xmath0-cell @xmath290 , 1.6 cm ( ii ) for any @xmath1-cell @xmath289 , and 1.6 cm ( iii ) @xmath293 and @xmath294 1.6 cm ( iv ) @xmath295 and @xmath296 1.6 cm ( v ) @xmath297 for all @xmath0- cells @xmath290 and @xmath291 .
\(i ) follows from the definition of @xmath298 being the right dual of @xmath290 and @xmath24 being invertible .
first part of ( ii ) follows from the definition of @xmath299 and naturality of @xmath24 and the second part easily follows from the first one .
\(iii ) and ( iv ) follow from the way the weak functors @xmath167 and @xmath300 are defined .
definition of the pivotal structure @xmath24 implies ( v ) . parts ( iii ) and ( iv ) of the above lemma do not use the pivotal structure @xmath24 at all .
however , with the help of pivotal structure , especially part ( v ) of the above lemma , one may also prove the following : + @xmath301 and @xmath295 1.6 cm using the above graphical calculus , we immediately obtain the following relation which will be useful later . [ ax * ] @xmath302 for all @xmath0-cell @xmath77 .
[ rotation invariance ] for any @xmath1-cell @xmath303 , the following identities hold : @xmath304 2.4 cm it is enough to show one of the identities ( because applying the reverse rotation and using lemma [ basic lemma ] ( i ) , one can deduce the other identity ) .
we will sketch the proof of the first identity . for the case
@xmath305 , the result follows trivially from the naturality of @xmath24 .
. then lhs of the first identity + @xmath307 @xmath307 2.4 cm ( using lemma [ basic lemma ] ( v ) ) + @xmath307 @xmath307 2.5 cm ( using lemma [ basic lemma ] ( i ) ) . for @xmath308 , an analogous result ( with @xmath309 replaced by @xmath310 )
can be deduced by applying the above result recursively . after working on the rest of the curves ( emanating from the top of
the rectangle labelled with @xmath84 ) in the same way as above , the lhs of the first identity + @xmath307 @xmath311 ( using naturality of @xmath24 ) .
we now construct a planar algebra from a bicategory .
let @xmath78 be a pivotal @xmath312-linear strict @xmath1-category with @xmath313 as the set of @xmath80-cells and fix @xmath314 .
for each colour @xmath204 , set @xmath315 if @xmath316 and @xmath317 .
define @xmath318 .
now , for a @xmath204-planar tangle @xmath319 we wish to define a multilinear map @xmath320 . for this
we extensively use the graphical calculus of the @xmath1-cells of @xmath78 . for the ease of dealing with @xmath1-cells replaced by labelled rectangles , we will consider the planar tangle @xmath321 as an isotopy class of pictures where each disc ( internal or external ) is replaced by a rectangle with first half of the strings being attached to one of the side ( called the _ top side _ ) and the remaining half of the strings attached to the opposite side ( called the _ bottom side _ ) .
next , in the isotopy class of @xmath321 , we fix a picture @xmath322 placed on @xmath188 with the bottom side of the external rectangle being parallel to the @xmath77-axis , satisfying the following properties : * the collection of strings in @xmath322 must have finitely many local maximas and minimas , * each internal rectangle is aligned in such a way that the top side of the external rectangle is parallel and also nearer to the top side of the internal rectangle than its bottom side , * the projections of the maxima , minima and one of the vertical sides of each internal rectangle ( that is , the sides other than the top and bottom ones ) on the vertical sides of the external rectangle of @xmath322 are disjoint
. we will say that an element @xmath322 in the isotopy class of @xmath321 is in _ standard form _ if @xmath322 satisfies the above conditions__. _
_ for example , a standard form representation of the tangle in figure [ egtang ] will be the following diagram let @xmath322 be an element in standard form of the isotopy class of @xmath323 .
we now cut @xmath322 into horizontal stripes so that every stripe should have at most one local maxima , minima or internal rectangle .
each component of every string in a horizontal stripe is labelled with @xmath77 or @xmath324 according as the orientation of the string is from the bottom side to the top side of the horizontal stripe or reverse respectively ; each local maxima or minima is labelled with @xmath77 and the orientation is induced by the orientation of the actual string in @xmath322 .
for example , + will be replaced by 1.92 cm to define @xmath325 , we fix @xmath1-cells @xmath326 for @xmath61 .
we label the @xmath225 internal rectangle ( contained in some horizontal stripe ) with @xmath327 .
now , each horizontal stripe makes sense as a @xmath1-cell according to the notation already set up .
we define @xmath328 as the composition of these @xmath1-cells .
it is easy to easy that @xmath329 is a multilinear map from @xmath330 to @xmath331 .
natural question to ask will be why @xmath332 is independent of the choice of @xmath322 in the isotopy class of @xmath321 . for this
, first note that one standard form representative of a tangle can be obtained from another applying finitely many moves of the following three types : \(i ) _ sliding move : _ @xmath333 1.6 cm ( ii ) _ rotation move : _ 1.6 cm ( iii ) _ wiggling move : _ 1.6 cm where @xmath84 and @xmath108 are @xmath1-cells . to show @xmath332 is well - defined , it is enough to show that two standard form representatives @xmath334 and @xmath335 of @xmath336 labelled with @xmath337 , differing by any of the above three moves , will assign identical @xmath1-cell .
invariance under sliding moves hold from the functoriality of @xmath116 and rotation moves follow from proposition [ rotation invariance ] and corollary [ ax * ] ; finally , wiggling moves are justified by lemma [ basic lemma ] ( i ) .
thus , we have a well - defined map @xmath338 ( resp .
@xmath339 ) . finally , define the planar algebra @xmath340 via : * @xmath341 , * the linear map @xmath342 ( resp .
@xmath343 ) is defined by extending the map @xmath344 ( resp .
@xmath339 ) linearly . clearly , @xmath345 for all @xmath231 . to check @xmath234 preserves composition of morphisms ,
let us consider two tangles @xmath226 and @xmath224 such that the first internal disc @xmath346 of @xmath226 has color @xmath231 same as that of @xmath224 .
choose @xmath347 and @xmath348 as standard form representatives of @xmath226 and @xmath224 respectively such that dimension of the external disc of @xmath348 along with the marked points match with that of @xmath346 in @xmath347 .
let @xmath349 denote the picture obtained by replacing @xmath346 by @xmath348 and then erasing the external boundary of @xmath348 .
note that @xmath349 is a standard form representative of @xmath350 .
now , we consider @xmath1-cells for each internal discs ( except @xmath205 ) in @xmath226 and @xmath224 coming from the appropriate vector spaces and label the corresponding rectangles in @xmath347 , @xmath348 , and @xmath349 with them .
if we slice @xmath349 as described while defining the action of @xmath234 on the morphism spaces and induce the slicing of @xmath349 on @xmath347 and @xmath348 , then the @xmath1-cells corresponding to the slices appearing in @xmath349 are the same as those for @xmath347 with the slice containing @xmath346 being replaced by the slices coming from @xmath348
. thus @xmath234 must preserve composition .
this completes the construction of the planar algebra .
in this section , we will introduce the notion of an _ affine representation of a planar algebra _ which is a generalization of the concept of the _ hilbert space representation of annular temperley - lieb _ by vaughan jones and sarah reznikoff ( @xcite ) ; one can also treat this as an _
annular representations of a planar algebra _ with rigid boundaries .
we then discuss some general theory of the affine representations following exactly the way jones developed the theory for annular representations in @xcite . before going into the definition of affine representations
, we will first introduce the _ affine category over a planar algebra_. an @xmath351-affine tangle is an isotopy class of pictures consisting of : * the annulus @xmath352 , * the set of points @xmath353 ( resp .
@xmath354 ) are numbered clockwise starting from @xmath1 ( resp .
@xmath0 ) as the first points , * @xmath355 consists of internal discs @xmath190 , @xmath207,@xmath214 , @xmath356 with colour @xmath212 , @xmath213,@xmath214,@xmath357 respectively and non - intersecting oriented strings ( just like in an ordinary planar tangle described in definition [ def of pln tang ] ) so that the inner ( resp
. outer ) boundary of @xmath355 gets the colour @xmath358 ( resp .
@xmath359 ) , * any isotopy should keep the boundary of @xmath355 fixed
. -affine tangle ] let @xmath234 be a planar algebra .
an @xmath351-affine tangle is said to be @xmath234__-labelled _ _ if , to each internal disc @xmath205 of @xmath355 with colour @xmath360 , an element of @xmath361 is assigned .
let @xmath362 denote the set of all @xmath363-affine tangles and @xmath364 denote the set of all @xmath234-labelled @xmath363-affine tangles .
if @xmath365 and @xmath366 then we can define @xmath367 ( @xmath368 ) as the affine tangle obtained by considering the picture @xmath369 .
we might have to smoothen out the strings which are attached with the inner boundary of @xmath370 and outer boundary of @xmath91 ; this can also be avoided by requiring the strings to meet the inner and the outer boundaries radially in the definition of an affine tangle .
we now set up a convenient way of sketching an affine tangle ; instead of marking the points on the inner ( resp .
outer ) boundary at the roots of unity ( resp .
twice the roots of unity ) , we will mark them close to each other on the top with @xmath0 as the leftmost point .
further , with the help of isotopy , every @xmath371 can be expressed as : for some @xmath372 .
note that @xmath226 and @xmath373 are not unique .
for example , the affine tangle in figure [ aff tang eg ] can be expressed as the above annular tangle for @xmath374 , @xmath375 , @xmath376 and @xmath377 1.92 cm and @xmath378 .
let @xmath379 be the vector space with @xmath380 as a basis , @xmath381 be the set of all @xmath234__-labelled _ _ @xmath204-planar tangles , @xmath382 ( resp .
@xmath383 ) be the vector space with @xmath217 ( resp .
@xmath381 ) as a basis and @xmath384 be the annular tangle .
2.16 cm observe that @xmath385 induces a linear map @xmath386 which also lifts to the level of the @xmath234-labelled ones .
moreover , for any @xmath387 , there exists @xmath388 and @xmath389 such that @xmath390 .
set @xmath391 where we use the linear map @xmath392 induced by the map of multicategories @xmath234 .
it is a fact that @xmath393 is a vector subspace of @xmath379 .
for instance , if @xmath394 and @xmath395 , @xmath396 such that @xmath397 , @xmath398 for @xmath399 , @xmath1 and @xmath400 , then one can obtain @xmath401 such that @xmath402 by wiggling back and forth a string emanating from either of the vertical sides of @xmath246 around the inner disc of @xmath403 until the total number of strings around the inner disc of @xmath403 increases from @xmath404 to @xmath405 ; finally , @xmath406 .
define the category @xmath407 by : * @xmath408 * @xmath409 the quotient vector space of @xmath410 over @xmath393 ( also denoted by @xmath411 ) , * the composition of affine tangles is linearly extended for @xmath412 s ; one can easily verify that @xmath367 @xmath413 whenever @xmath414 and @xmath415 , or @xmath416 and @xmath417 ; this implies the composition is induced in the level of quotient vector spaces as well , * the identity of @xmath204 denoted by @xmath418 , is given by an @xmath419-affine tangle obtained by joining the @xmath225 point of the inner boundary with the @xmath225 point of the outer boundary by a straight string for all @xmath420 .
we will refer the category @xmath407 as _ affine category over _ @xmath234 .
an additive functor @xmath421 is said to be an affine representation of @xmath234 .
the functor induced by @xmath234 itself gives an affine representation of @xmath234 ; this is called the ` trivial ' affine representation .
[ aff rep basic lemma]if @xmath125 is an affine representation of @xmath422 then \(a ) @xmath423 , \(b ) @xmath424 is isomorphic to @xmath425for all colours @xmath204 . the inclusion in part ( a ) is given by considering the @xmath125-image of the inclusion tangle .
for part ( b ) , consider the rotation tangle @xmath426 obtained by joining the points @xmath427 and @xmath428 on the boundary of @xmath355 by a string which does not make a full round about the inner disc .
@xmath429 gives the desired isomorphism in ( b ) it may seem so that @xmath430 is the identity @xmath419-affine tangle ( that is , the tangle obtained by joining the points @xmath427 and @xmath431 by a straight line ) , but this is not true because of the restriction of the isotopy being identity on boundary of @xmath355 .
this is the main difference between the annular representations of @xmath234 ( in ann rep and @xcite ) and the affine representations .
the _ weight _ of an affine representation @xmath125 denoted by @xmath432 , is given by the smallest integer @xmath433 such that @xmath434 .
the @xmath432 is well - defined by lemma [ aff rep basic lemma ] .
an affine representation @xmath125 will be called _ locally finite _ if @xmath424 is finite dimensional for all colours @xmath204
. the _ dimension _ of an affine representation @xmath125 is defined as a pair of formal power series @xmath435 where@xmath436 * * question : * * _ _ if _ _
_ is a planar algebra with modulus _
_ @xmath437 _ _ , is the radius of convergence of the dimension of an affine representation greater than or equal to _ _
_ _ the above question appeared in @xcite for annular representations of a planar algebra .
the question for annular representations was answered in affirmative for the temperley - lieb planar algebras by jones ( in @xcite ) and for the group planar algebras by ghosh ( in @xcite ) .
we will show the same for affine representations of any finite depth planar algebra in the section [ fdpa ] .
let @xmath234 be a @xmath167- or a @xmath8-planar algebra .
then , @xmath383 becomes a @xmath167-algebra where @xmath167 of a labelled tangle is given by @xmath167 of the unlabelled tangle whose internal discs are labelled with @xmath167 of the labels .
one can define @xmath167 of an affine tangle by relecting it around a circle concentric to inner or outer boundary and then isotopically stretch or shirnk to fit into the annulus @xmath355 such that the first point of inner or outer boundary after reflection remains the same whereas the first point of any internal disc after reflection is given by the reflection of the last point and colours of all discs are preserved ; this can be induced in the @xmath234-labelled ones by labelling the internals discs of the reflected tangle with @xmath167 of the labels .
note that @xmath167 is an involution .
extending @xmath167 conjugate linearly , we can define the map @xmath439 for all colours @xmath440 , @xmath441 .
it is easy to check that @xmath442 .
this makes the category @xmath407 a @xmath167-category .
an additive functor @xmath443 is said to be an _ affine _
@xmath167__-representation if _ _
@xmath125 is @xmath167 preserving , that is , @xmath444 for all @xmath445 where @xmath446 denotes the category of hilbert spaces . note that if @xmath125 is an affine @xmath167-representation , then @xmath447 for all @xmath448 , @xmath449 , @xmath450 .
the category of affine representations of a planar algebra @xmath234 with natural transformations as morphism space , forms an abelian category and the dimension is additive with respect to direct sum .
one can further talk about _ irreducibilty _ and _ indecomposability _ of an affine representation ( see @xcite for details ) .
for example , the trivial affine representation of @xmath234 is irreducible .
however , if we restrict ourselves to the case of a locally finite , non - degenerate @xmath8-planar algebra @xmath234 and the category of locally finite affine @xmath167-representations , the notions of irreducibility and indecomposability coincide . in this case
, one can also talk about _ orthogonality _ of affine representations .
these treatments for annular representations can be found in more details in @xcite .
jones indicated a procedure of finding annular representations of a locally finite @xmath8-planar algebra @xmath234 with modulus @xmath451 in @xcite ; the same works for the affine ones as well .
for this , we need to consider a subspace of the morphism space @xmath452 , namely,@xmath453it is easy to see that @xmath454 is an ideal in @xmath455 .
we list some common properties shared by affine @xmath167-representations and annular @xmath167-representations of @xmath234 ; the proofs can be found in @xcite .
\(i ) an affine representation @xmath125 is irreducible iff @xmath424 is irreducible as an @xmath456-module for all colours @xmath457 .
\(ii ) if @xmath458 is an irreducible @xmath456-submodule of @xmath424 for some colour @xmath457 , then @xmath458 generates an irreducible subrepresentation of @xmath125 .
\(iii ) orthogonal @xmath456-submodules of @xmath424 for some colour @xmath457 , generate orthogonal subrepresentations of @xmath125 .
\(iv ) if @xmath125 and @xmath459 are representations with @xmath125 being irreducible and if @xmath460 is a non - zero @xmath452-linear homomorphism for some colour @xmath457 , then @xmath321 extends to an injective homomorphism from @xmath461 to @xmath459 , that is , an injective natural transformation from @xmath125 to @xmath459 .
\(v ) if @xmath462 @xmath463 , then @xmath464 from ( v ) , we can conclude that for an affine @xmath167-representation @xmath125 with weight @xmath433 , we have @xmath465 since @xmath466 turns out to be zero and hence @xmath424 forms a module over the quotient @xmath467 .
we denote this quotient algebra by @xmath468 ( lowest weight algebra at @xmath204 ) . by ( i ) , if f is an irreducible affine @xmath167-representation with weight @xmath433 , then @xmath424 is an irreducible module over @xmath469 . in order to find the irreducible affine @xmath167-representations of @xmath234
, it suffices to do the following : \(i ) find the irreducible representations of @xmath469 , \(ii ) find which irreducible representation of @xmath469 gives rise to an irreducible affine @xmath167-representation of the planar algebra .
we will use this method to deduce some results on the irreducible affine @xmath167-representations of a _ finite depth planar algebra _ in the next section .
in this section , we will recall the notion of the _ depth _ of a planar algebra which is motivated from the _ depth of a finite index subfactor_. we then prove some finiteness results for the category of affine representation of subfactor - planar algebras . finally , we answer the question mentioned in section [ affrep ] for subfactor - planar algebras with finite depth .
let @xmath234 be a planar algebra with modulus @xmath243 .
we first define below a tangle called _ jones projections_. + @xmath470{pictures
/ fdpa / e_k.jpg } \right ) \in p_{(k+1,\varepsilon ) } $ ] where @xmath471 and @xmath472 .
+ note that @xmath473 the proof of being right ideal easily follows by considering the tangle 1.28 cm and noting that the range of the @xmath234 action of this tangles is inside @xmath482 .
proof of left ideal follows from the same tangle with upside down .
consider the tangles @xmath490 @xmath491 1.2 cm and @xmath492 @xmath493 1.28 cm where @xmath494 of course depends on the @xmath495 which will be automatically determined from the context .
note that @xmath483 if and only if @xmath496 ( @xmath307 span of the image of @xmath497 ) .
a planar algebra @xmath234 is said to have finite depth if @xmath500 for some @xmath501 , @xmath502 and in that case , the depth of @xmath234 will be a pair of natural numbers @xmath503 such that @xmath504 is the smallest natural number such that @xmath505 . from lemma [ s lemma ] , one can deduce that if @xmath506 denotes the depth of @xmath234 and @xmath504 is even ( resp .
odd ) , then @xmath507 ( resp .
@xmath508 ) , that is , either both @xmath509 and @xmath510 are same or they are consecutive natural numbers with the larger one being even .
suppose the statement of the lemma is true for all @xmath520 .
to show the same for @xmath521 , we consider the tangle @xmath522 . clearly , @xmath523 since @xmath524 .
again , for even @xmath55 ( @xmath525 say ) , the tangle @xmath526 isotopically looks like : note that the tangle bounded by the dotted line , denoted by @xmath226 , is a @xmath204-planar tangle and thus @xmath527 where @xmath226 sits in the @xmath528 position ; this implies @xmath529 \subset range\left [ p\left ( s_{\left ( k,\varepsilon \right ) } ^{(n+1)}\right ) \right ] \subset p\left ( k+n,\varepsilon \right ) \text{.}\ ] ] hence , @xmath530 .
similar arguments can be used to prove the same for odd @xmath55 .
[ affine rank]if @xmath234 is a finite depth planar algebra with @xmath515 as its depth , then @xmath531 for all colours @xmath532 , @xmath533 and @xmath534 where @xmath535 $ ] .
( @xmath536 $ ] denotes the greatest integer function . )
if either of @xmath537 and @xmath538 is less than or equal to @xmath539 , then the equality can easily be established by wiggling a string sufficiently and then decomposing the affine tangle .
one can also assume @xmath540 because the case when they are different can be deduced using rotation tangles . without loss of generality ,
let @xmath541 , @xmath542 and @xmath543 .
let @xmath544 .
then , @xmath81 can be expressed as the equivalence class of the affine tangle @xmath545 such that the internal rectangle is labelled with an element of @xmath546 where @xmath547 can be chosen to exceed @xmath373 ( using wiggling around the inner disc ) . by lemma [ general s lemma ]
, @xmath81 is a linear combination ( l.c . ) of equivalence class ( eq . cl . ) of labelled @xmath548{pictures / fdpa / wt_reduction1.jpg } \right)$ ] .
now , we consider two cases .
+ * case 1 * : @xmath373 is odd , that is , @xmath549 .
we can isotopically move the internal rectangles attached to left side of the above tangle around the inner disc and bring them to the to the right side . in this way
, we express @xmath81 as : l.c . of eq .
cl . of @xmath550{pictures / fdpa / wt_reduction2.jpg } \right)$ ] + @xmath307 l.c . of eq .
cl . of @xmath551{pictures
/ fdpa / wt_reduction3.jpg } \right ) $ ] . identifying the two vertical sides of the last tangle we get an affine tangle which we cut along the dotted line ;
this cutting induces a decomposition of @xmath81 .
note that the dotted line intersects exactly @xmath552 strings .
thus @xmath553 . +
* case 2 * : @xmath373 is even , that is , @xmath554 .
using similar arguments as in case 1 , we can conclude that @xmath81 is a l.c . of eq .
cl . of @xmath550{pictures
/ fdpa / wt_reduction4.jpg } \right ) $ ] . cutting the tangle along the dotted line just like in case 1
, we can decompose @xmath81 and get @xmath555 .
corollary [ affine ideal ] implies that the lowest weight algebra @xmath560 whenever @xmath561 .
thus , from the discussion of finding irreducible affine representations in section @xmath12 , all irreducible affine representations have weight atmost @xmath539 . to prove the same for non - irreducible ones , note that taking direct sums never increases the weight .
if @xmath234 is a finite depth subfactor - planar algebra with modulus @xmath562 , then every irreducible affine @xmath167-representation of @xmath234 is locally finite and the radius of convergence of its dimension is at most @xmath563 .
moreover , the number of irreducibles at each weight is finite .
let @xmath125 be an irreducible affine @xmath167-representation with weight @xmath433 .
so , @xmath564 is an irreducible module of @xmath565 .
irreducibility of @xmath125 says that @xmath125 induces an surjective linear map from @xmath566 to @xmath567 .
therefore , we have @xmath568 .
we look back once again into the two cases in the proof of proposition [ affine rank ] .
let @xmath373 and @xmath539 be as in proposition [ affine rank ] for the rest of the proof .
a careful observation on the two cases will say that there exists a surjective linear map from @xmath569 ( resp .
@xmath570 ) to @xmath571 when @xmath373 is odd ( resp . even ) . therefore , @xmath572 since @xmath234 is locally finite .
the lowest weight algebras become finite dimensional and hence there are finitely many irreducibles at each weight .
this also implies @xmath564 has finite dimension .
thus @xmath125 is locally finite
. tangle if @xmath373 is odd ( resp .
even ) . 2.16 cm by lemma [ general s lemma ] and proof of proposition [ affine rank ] , eq . cl .
of such labelled tangles generate @xmath575 .
therefore , @xmath576 so , @xmath577 .
now , we try to find the limit of @xmath578 as @xmath537 tends to infinity .
note that @xmath579 is constant .
next , @xmath580 norm of the principal graph @xmath307 the index of the finite depth subfactor corresponding to the planar algebra . by jones theorem , index of the subfactor is square of the modulus .
hence , @xmath581 which implies radius of convergence of @xmath582 is at least @xmath563 .
this ends the proof .
d. bisch , _
bimodules , higher relative commutants and the fusion algebra associated to a subfactor _ , operator algebras and their applications , 13 - 63 , fields inst .
commun . , 13 , amer .
soc . , providence , ri , ( 1997 )
|
we define jones s planar algebra as a map of multicategories and constuct a planar algebra starting from a @xmath0-cell in a pivotal strict @xmath1-category .
we prove finiteness results for the affine representations of finite depth planar algebras .
we also show that the radius of convergence of the dimension of an affine representation of the planar algebra associated to a finite depth subfactor is at least as big as the inverse - square of the modulus .
|
break junction experiments have allowed to perform transport studies on nanoscale metallic contacts in which the mean free path of the electrons is much larger than the junction length .
the observation of quantized conductance in such systems is a hallmark of ballistic transport and opened new vistas to study the scaling of electronic devices down to the atomic length scale.@xcite a drawback of such experiments is the limited control of the microscopic arrangement in the junction which hinders a straight forward interpretation of the data and makes a comparison with theoretical calculations difficult.@xcite in this respect , a great advantage is given by the use of scanning tunneling microscopy ( stm ) experiments , in which a tip can approach and contact single atoms or molecules on a surface.@xcite in such experiments , it has been possible to measure the conductance as a function of tip - sample distance from the tunneling to the contact regime . due to the promise of spintronic devices for future applications with low power consumption and high speed ,
a recent focus of such contact measurements has been magnetic systems , e.g. spin - valve behavior has been observed in single magnetic molecules or atoms on surfaces @xcite and the occurrence of the kondo effect has been found in ferromagnetic atomic contacts.@xcite it has been emphasized that the low coordination of the contact atoms in nanoscale junctions leads to an enhanced tendency towards magnetism , e.g. magnetic moments are formed in systems of otherwise non - magnetic materials.@xcite naturally , transport phenomena in such magnetic low - dimensional systems have raised a lot of attention and triggered many theoretical studies , which mainly focused on systems with collinear magnetic order , considering also the effect of magnetoresistance.@xcite it was also recently realized that , if the magnetization direction of the two electrodes is opposite , a domain wall can form in the contact between them and the non - collinear order in the domain strongly affects the conductance and the magnetoresistance.@xcite finally , the effect of spin - orbit coupling on the conductance needs to be considered@xcite which leads to novel transport phenomena such as the ballistic anisotropic magnetoresistance@xcite or the tunneling anisotropic magnetoresistance.@xcite recently , the transition regime from tunneling to contact in a spin - polarized stm geometry has been studied based on density functional theory in order to explain e.g. the conductance of a single magnetic atom,@xcite and to analyze the contribution from different conduction channels.@xcite as a magnetic stm tip approaches a single magnetic atom on a surface an exchange interaction with the tip apex atom occurs . in principle , it is possible to switch the magnetic moment of the adatom in such a way.@xcite if the magnetic moment of the adatom is exchange coupled to the substrate ( as in ref . ) there is a competition of exchange interactions which can result in a canting of the spins close to contact .
non - collinear spin alignment in such an atomic contact can also occur if the adatom spin is canted due to exchange coupling on a substrate with a spin spiral structure as in refs . and .
the effect of such a non - collinearity in the spin direction of the tip apex and the adatom on the conductance is the focus of the present work .
we introduce an approach to calculate the conductance in magnetic nanojunctions with non - collinear spin structure from first - principles , employing the methodology of non - collinear wannier functions ( wfs ) , which we describe in detail . in order to start from an accurate description of the electronic and magnetic structure of the system we use the full - potential linearized augmented plane wave ( flapw ) method based on density functional theory .
we map the electronic structure of a system in a non - collinear magnetic state from the flapw description to a tight - binding like hamiltonian via wfs .
finally , we calculate the conductance within the landauer approach with the technique of green s functions . as a model system ,
we consider two co monowires to each of which a single apex mn atom is attached .
we vary the distance between the two mn atoms in order to calculate the conductance from the tunneling to the contact regime . the magnetization direction of the two co electrodes is chosen either parallel ( p ) or antiparallel ( ap ) which allows us to obtain the distance - dependent magnetoresistance . in the tunneling regime
the conductance is dominated by states of @xmath3-@xmath4-orbital character and only in the contact regime there is an additional contribution due to @xmath8-states .
as the latter conduction channel is suppressed in the ap alignment the magnetoresistance displays a large rise close to contact .
when the two mn atoms approach in the p electrode alignment a competition of the exchange interactions between the two mn spins and the mn spins with the co electrodes occurs . while the mn spins couple ferromagnetically to the co electrodes , they couple antiferromagnetically with each other . as a result
a non - collinear arrangement becomes the magnetic ground state and the angle between the two mn spins changes gradually from zero to about @xmath9 at the closest separation we considered .
the conductance displays a characteristic dip as the non - collinear state forms which is also apparent in the distance - dependent magnetoresistance .
we explain this reduction of the conductance due to non - collinear spin states from the spin - dependent hybridization of @xmath8-states between the two mn atoms which depends on the angle between their spin moments and partly suppresses the conduction in this channel .
the paper is organized as follows . in sec .
[ sec : method ] we introduce our method to calculate the conductance of a one - dimensional nanoscale junction with a non - collinear spin structure .
we discuss the extension of wannier functions to systems with non - collinear order ( sec .
[ sec : nocowfs ] ) , the implementation within the flapw method ( sec .
[ sec : flapw_to_wfs ] ) , and the incorporation into our transport code ( sec .
[ sec : nocotrans ] ) . in sec .
[ sec : co - mn - mn - co ] we introduce our model system consisting of two co monowires to each of which a single mn atom is attached .
first , we analyze the magnetic and transport properties of collinear spin states from tunneling to contact ( sec .
[ subsec : t - to - c ] ) before we address the occurrence of non - collinear spin states in the contact regime ( sec .
[ subsec : origin_noco ] ) .
we analyze the ballistic conductance of such spin states ( sec .
[ subsec : noco_junction ] ) and show that a characteristic fingerprint is observed in the distance - dependent conductance and the magnetoresistance ( sec .
[ subsec : t - to - c_rev ] ) .
we end with a summary in sec .
[ sec : summary ] .
the density functional theory ( dft)@xcite states that the energy functional of a general magnetic system @xmath10 $ ] is uniquely determined by the charge density @xmath11 and the magnetization density @xmath12 .
the most common approximation made to a general magnetic system is to assume a collinear magnetization density , i.e. , @xmath13 , where @xmath14 is an arbitrary direction . within this collinear approximation
the energy is a unique functional of the charge density @xmath15 and the scalar magnetization density @xmath16 . due to
decoupled spin and real space , the spin - channels can be treated independently .
however , it is known that relaxing the collinear approximation and allowing for non - collinearity of the magnetization density in real space in the dft setup leads to an ability of reliably treating whole classes of new phenomena , which rely on the properties of complex magnetic states.@xcite within the dft formulation for non - collinear magnetic systems one solves the kohn - sham equations@xcite @xmath17 where @xmath18 is the @xmath19 unity matrix , @xmath20 is the potential matrix which also mixes the spin channels , @xmath21 is the band index and @xmath22 is the spinor bloch function with spin - up and spin - down components @xmath23 and @xmath24 , respectively . for the @xmath25 converged spinor kohn - sham orbitals @xmath26 on a uniform mesh of @xmath27 @xmath28-points , the orthonormal set of wannier functions can be obtained via the transformation@xcite @xmath29 where the number of wfs @xmath30 is smaller than or equal to @xmath25 and the matrices @xmath31 represent the gauge freedom of the wfs . in case when @xmath32 and the group of bands we are extracting the wfs from is isolated from other bands , the @xmath31-matrices are unitary at each @xmath33-point . imposing the constraint of maximal localization of wfs in real space determines the set of @xmath31-matrices up to a common global phase , and the corresponding wfs
are called maximally - localized wannier functions ( mlwfs).@xcite like the kohn - sham orbitals , the mlwfs are spinors and can be written as @xmath34 in terms of their spin - up and spin - down components @xmath35 and @xmath36 , respectively . for the construction of mlwfs within dft electronic structure codes , the matrix elements @xmath37 and @xmath38 need to be computed , where @xmath39 is a localized orbital , which defines the starting point of the iterative procedure of determining the mlwfs .
@xcite since spin - up and spin - down are coupled in noncollinear calculations , these matrix elements involve a summation over the spin @xmath40 .
the matrix elements @xmath41 for the wfs tight - binding hamiltonian @xmath42 are given by@xcite @xmath43 even though the wannier and bloch functions are spinor valued , the transformation of the hamiltonian from bloch into wannier representation is fully determined by the matrices @xmath44 and the eigenvalues @xmath45 as in the collinear case .
the treatment of noncollinear magnetism within the flapw method as implemented in the jlich dft code ` fleur`@xcite neglects the effect of intra - atomic non - collinearity .
space is partitioned into the muffin - tin ( mt ) and interstitial regions ( ir ) .
the spin density @xmath46 in the ir is treated without shape approximation as a continuous vector field . in the mt - sphere @xmath47 of atom
@xmath48 only the projection of the spin density onto the direction @xmath49 of the average spin moment is used for the generation of the exchange - correlation potential .
the explicit one- and two - dimensional implementations also contain a third region , the vacuum region ( vr ) , which can be treated analogously to the ir.@xcite thus , the self - consistent spin density is approximated as @xmath50 a part of the intra - atomic non - collinearity can still be described within this hybrid approach by decreasing the mt radii . the radial solutions @xmath51 of angular momentum @xmath52 of the scalar - relativistic schrdinger equation in @xmath47 and their energy derivatives @xmath53 are calculated for the two spins @xmath54 and used for the expansion of basis functions and bloch functions .
the spin quantum number @xmath54 refers to the local spin quantization axis @xmath49 .
the expansion coefficients of the eigenspinors of the local spin quantization axis in terms of the eigenspinors of the global spin quantization axis , which is the @xmath55 axis , are given by @xmath56 where @xmath57 and @xmath58 are azimuthal and polar angles of the spin direction of @xmath47 with respect to the global frame @xmath59 .
within the mts the wave function @xmath60 is thus given by : @xmath61 y_{l}(\hat{\mathbf r}))\chi^{\alpha g}_{\sigma^\alpha } \end{array}\ ] ] where @xmath62 denotes the angular momentum quantum numbers and @xmath21 is the band index .
using functions @xmath39 which are restricted each to a single mt - sphere has been found to result in a very good starting point for the iterative optimization of collinear wfs .
@xcite due to the approximate intra - atomic collinearity , it is reasonable to choose in the noncollinear case the localized orbitals @xmath39 to be eigenstates of the projection of the spin - operator onto the local spin - quantization axis @xmath49 : @xmath63 here , @xmath64 are expansion coefficients , @xmath65 is the index of the atom for which @xmath66 is non - zero , @xmath67 is the spin associated with this trial orbital @xmath68 and @xmath69 is the radial part of the trial orbital .
thus we obtain : @xmath70^*\int u_{l\sigma(n)}^{\alpha(n)}(r)\tilde{u}_{nl}(r)r^2\,dr\\&\\ & + [ b^\alpha_{ml\sigma(n)}(\mathbf{k})]^*\int \dot u_{l\sigma(n)}^{\alpha ( n)}(r)\tilde{u}_{nl}(r)r^2\,dr\}. \end{array}\ ] ] the mt contribution to the @xmath71 matrix may be written as @xmath72^ { * } e^{-i\mathbf{b}\cdot\mathbf{r } } \psi_{\mathbf{k}+\mathbf{b}n\sigma^{\alpha}}(\mathbf{r}),\ ] ] with @xmath73 given by @xmath74 .
\end{aligned}\ ] ] the computation of the @xmath71 matrix for noncollinear systems reduces therefore to integrals for which explicit expressions have been given for the flapw method .
@xcite the extension of the collinear scheme for ballistic transport , which we described in detail in our previous publication , ref . , to a non - collinear setup is now rather straightforward . given the minimal wfs hamiltonian @xmath75 ( eq . ( [ h_wfs ] ) ) for a non - collinear system , we are able to construct the tight - binding hamiltonian of the nanojunction in accordance to our transport method,@xcite which employs the partitioning of space into the scattering region @xmath76 , as well as left @xmath77 and right @xmath78 leads : @xmath79 compared to a collinear calculation we have to deal with twice as many wannier functions due to the inseparable spin - channels . two calculations using the locking technique@xcite are required for eq .
( [ opensys ] ) .
the hamiltonian matrix @xmath80 and the matrices @xmath81 , describing the coupling to the leads , are obtained from a super - cell flapw calculation .
the green s functions for the leads @xmath82^{-1}$ ] can be brought to finite sized surface green s functions @xmath83 by constructing @xmath84 based on principal layers @xmath85 and the coupling matrices @xmath86.@xcite those matrices are obtained from a separate calculation of a perfect periodic lead . following our landauer - bttiker method
the non - collinear ballistic transport can be calculated with the green s function of the scattering region : @xmath87^{-1}.\ ] ] the interaction between scattering region and the leads , and the resulting level broadenings are described by broadening matrices @xmath88 @xmath89,\ ] ] where @xmath90 are the self - energies of the leads : @xmath91 finally the ballistic transport process is described by the transmission function @xmath92 @xmath93.\ ] ] resulting in the conductance through the junction @xmath94 with the conductance quantum @xmath95 . in the non - collinear case
the trace operation of eq .
( [ transmission ] ) has to be additionally performed over the spin @xmath40 .
the spin - channel information is therefore lost for general non - collinear systems .
we tested non - collinear wannier functions on freestanding non - collinear magnetic mn chains and found them to reproduce the flapw electronic structure with any given accuracy .
for the performed transport calculations a @xmath96 nearest - neighbor ( nn ) tight - binding like hamiltonian is sufficient due to the excellent correspondence of flapw and wfs electronic structure in the vicinity of the fermi level in absence of @xmath3-@xmath4 band edges in that particular region , which usually would require to consider more neighbors @xcite . for the system , considered in the following
, the orbitals participating in transport can be arranged according to the symmetry into the @xmath97 ( @xmath3 and @xmath98 orbitals ) , the @xmath99 ( @xmath100 ) and the @xmath101 ( @xmath102 ) groups .
in the following section , we investigate the ballistic transport properties of collinear and non - collinear magnetic configurations of a @xmath103-@xmath104-@xmath104-@xmath103 junction , consisting of semi - infinite ferromagnetic co monowires with magnetic mn `` tip '' atoms , see sketch of the structure in fig .
[ fig : config ] .
we will discuss the effect of non - collinear magnetism on ballistic transport through such a junction specifically keeping in mind tunneling - to - contact stm@xcite and mechanically - controllable break - junctions@xcite experiments .
in particular , we investigate the changes in the transport properties upon changing the distance between the two mn atoms , while keeping all other interatomic distances fixed at their equilibrium `` semi - infinite '' values .
we will show , that upon bringing the leads together the non - collinearity in this system emerges as a result of competing mn - mn and mn - co exchange interactions .
we then explore the influence of non - collinear magnetism on ballistic transport for various collinear and non - collinear configurations .
the nomenclature for the magnetic states in the @xmath103-@xmath104-@xmath104-@xmath103 junction includes the alignment of the magnetization directions of the leads , parallel ( p ) or anti - parallel ( ap ) to each other , and the directions of the two mn spins . without loss of generality
, these directions are denoted with respect to the left lead which has magnetization `` up '' .
the mn spins can point `` up '' ( @xmath105 ) , `` down '' ( @xmath106 ) or in a direction which makes an angle @xmath48 with the direction `` up '' , see fig .
[ fig : config ] . in the latter case
we consider the symmetric configuration , denoted as p@xmath48 , in which the spins of the mn atoms make an angle of @xmath107 between each other .
for all considered non - collinear p@xmath48 states we fixed the direction of all co atoms either up or down , depending on the magnetization direction of the corresponding lead .
the energy differences between different magnetic states are given per mn atom .
-@xmath104-@xmath104-@xmath103 junction .
the magnetization of the left lead always points up ( @xmath108 ) .
the magnetization of the right lead can be either parallel ( p ) or anti - parallel ( ap ) to it .
the direction of the mn spins is marked with respect to the left lead : pointing up ( @xmath108 ) , down ( @xmath109 ) , or along a direction at an angle @xmath48 with the magnetization of the left lead .
special configurations of interest which are discussed in the text are shown : ( a ) p@xmath108@xmath108 , ( b ) symmetric p@xmath48 state with @xmath110 , p@xmath9 , ( c ) p@xmath109@xmath109 ( d ) p@xmath109@xmath108 and ( e ) ap@xmath108@xmath109 . in the text
, we refer to the separation between the leads in terms of the distance between the mn atoms , @xmath111 .
co atoms are displayed as orange spheres , mn atoms are displayed as red spheres . ,
scaledwidth=45.0% ] for all collinear and non - collinear electronic structure calculations we used density functional theory within generalized gradient approximation ( gga ) to the exchange - correlation potential,@xcite as implemented in the flapw jlich code ` fleur`.@xcite the wires were calculated in three - dimensional super - cells , with an interchain separation in the @xmath112-@xmath113@xmath114plane of 13 bohr .
the super - cell setup along the chain s axes ( @xmath55-direction ) is described in detail below .
the brillouin zone ( bz ) was sampled by 12 or 24 @xmath33-points along the @xmath55-axis , depending on the size of the super - cell .
all calculations were performed with an lapw basis cut - off parameter @xmath115 of 3.7 bohr@xmath116 , resulting in approximately 625 lapw basis functions per atom .
the parallel magnetic configuration ( p ) of @xmath103-@xmath104-@xmath104-@xmath103 junctions was investigated in an 8-atom super - cell along the chain direction , consisting of six co atoms with an equilibrium interatomic distance of the co infinite monowire of @xmath117 bohr , and two attached mn atoms , see fig . [
fig : config ] . for
all considered magnetic configurations , with parallel or antiparallel alignment of the magnetization of the leads , as well as non - collinear magnetic states , irrespective of the separation between the leads , we fixed the co - mn distance @xmath118 to 4.48 bohr , which corresponds to the equilibrium distance between the ferromagnetic co and mn atoms at a very large separation between the leads .
for the p@xmath108@xmath108 and p@xmath109@xmath109-states of the junction we considered the inter - mn separation of @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 , @xmath126 and @xmath127 bohr .
the anti - parallel magnetic configuration ( ap ) of @xmath103-@xmath104-@xmath104-@xmath103 junctions was calculated in a 16-atom super - cell , consisting of six @xmath108-co atoms and two mn atoms on one side , and six @xmath109-co atoms and two mn atoms at another end of the junction . in this case
@xmath111 was set to 4.5 , 5.0 , 5.5 , 7.0 and 8.5 bohr . for the conductance calculations we applied the locking technique to a perfect monowire to describe the semi - infinite leads , as described in detail in ref . .
in all cases the wannier functions were generated on a @xmath128 @xmath33-point grid in the bz .
for the collinear cases the wfs were generated from one @xmath129- and five @xmath130-orbitals per atom for each spin separately , which were constructed from the radial solutions for the flapw potential . in non - collinear calculations
the spin channels are mixed , and two @xmath129- and ten @xmath130-orbitals per atom were used to construct the wfs per atom .
the energy bands were disentangled using the procedure described in ref . .
for the collinear calculations the lowest 58 eigenvalues per @xmath33-point were used to obtain 48 wfs for the 8 atom super - cell and the lowest 104 eigenvalues per @xmath33-point for 96 wfs for the 16 atom super - cell calculations . with non - collinearity of
the magnetization included the lowest @xmath131 eigenvalues per @xmath33-point were used to obtain 96 wfs for the 8 atom unit cell . for testing purposes , for several non - collinear configurations we compared the electronic structure of the system calculated with ` fleur ` and with corresponding wfs , finding that a very good description of the electronic structure can be achieved with wfs within the 3rd nearest - neighbor approximation , while for wfs calculations of the transmission in the vicinity of the fermi level already the 2nd nearest - neighbor approximation to the wfs hamiltonian provides very reliable results .
we start the investigation of the @xmath103-@xmath104-@xmath104-@xmath103 junction with both leads positioned far away from each other . to mimic a tip - sample approach , we decrease the mn - mn distance @xmath111 , and calculate the energies of the collinear states p@xmath132 , p@xmath133 and p@xmath134 , showing the results in fig . [
fig : stretch](a ) .
the energy difference between p@xmath108@xmath108 and p@xmath109@xmath109 states when the distance is varied in the tunneling regime from @xmath135 bohr down to @xmath2 bohr remains relatively constant and constitutes around @xmath136 mev per atom , indicating weak interaction between both sides of the junction and a weak ferromagnetic coupling between the mn atom and its nearest co neighbor ( nn co ) .
after a small reduction of the energy difference between the p@xmath108@xmath108 and p@xmath109@xmath109 states around @xmath137 bohr , the ferromagnetic ( fm ) mn - co coupling becomes more stable for decreasing @xmath111 , expressed in an increasing energy difference . in the contact regime
we find a slight decrease in the energy difference from 93.5 mev at @xmath119 bohr down to @xmath138 mev per atom at @xmath139 bohr .
this decrease in energy can be correlated with strong changes in the mn and nn co spin moments , @xmath140 and @xmath140 , respectively , upon decreasing the distance , see fig .
[ fig : stretch](b ) and ( c ) ( see also discussion in the next section ) .
@xmath109 state ( red diamonds , dashed line ) and p@xmath109@xmath108 state ( green triangles , dotted line ) with respect to the energy of the p@xmath108@xmath108 state , as a function of the mn - mn interatomic distance @xmath111 .
spin moment of the mn atoms , @xmath140 , ( b ) , and nn co atoms , @xmath141 , ( c ) , for p@xmath108@xmath108 ( black circles , solid line ) and p@xmath109@xmath109 ( red diamonds , dashed line ) configurations are given as a function of @xmath111 .
( d ) total conductance at the fermi level , @xmath142 ( upper line ) and @xmath97 conductance @xmath143 ( lower line ) on a logarithmic scale for the p@xmath108@xmath108-state .
gray shaded area is associated with the @xmath97 conductance , while cyan shaded area with the @xmath99 conductance .
, scaledwidth=48.0% ] while in all cases the spin moments of the co atoms , not neighboring the mn atoms directly ( @xmath144 ) are very similar to the spin moments of the co atom in an infinite lead ( @xmath145 ) , the spin moments of mn atoms and the nn co atoms can be strongly affected by @xmath111 at close contact and the spin configuration of the junction .
namely , for the p@xmath109@xmath109 state @xmath140 decreases from 4.3 @xmath146 to 3.9 @xmath146 , while nn @xmath141 increases from @xmath147 to @xmath148 , as @xmath111 is varied from 5.5 to 4.5 bohr . on the other hand ,
if the mn spin moment exhibits a similar variation as a function of distance for the p@xmath108@xmath108 state , the spin moment of the nn co atoms remains relatively constant ( @xmath149 ) .
this interplay between structure and magnetism already indicates that the intra - atomic as well as inter - atomic exchange , given by the stoner parameter @xmath150 and the heisenberg exchange constants @xmath151 , respectively , may be of importance for further understanding of the magnetic properties of this system . the change from fm coupling at larger interatomic distances to an antiferromagnetic ( afm ) coupling at smaller @xmath111 in an infinite mn chain
has been previously predicted based on dft calculations.@xcite in the vicinity of this crossover point the mn spins favor non - collinear magnetic order.@xcite to demonstrate a strong tendency of mn spin moments to afm coupling at smaller values of @xmath111 we plot the energy difference between the p@xmath108@xmath108 and p@xmath109@xmath108 states in fig . [
fig : stretch](a ) . reversing one of the mn spin moments in the p@xmath108@xmath108 configuration is clearly energetically more favorable than the p@xmath108@xmath108 state when the distance between the mn atoms is below @xmath152 bohr . in this case
the gain in energy due to switch of the mn spin moment can be explained only by the strong afm coupling of the two mn atoms for this regime of interatomic distance , since the coupling of the mn atom with its nn co atom is ferromagnetic .
the spin moments in the p@xmath109@xmath108 state at @xmath139 bohr constitute @xmath153 for mn and @xmath154 for its nn co on the afm side , and @xmath155 for mn and @xmath156 for the n.n co on the fm side . for larger @xmath111 values
the p@xmath108@xmath108 state is the lowest in energy as compared to all possible collinear states of the junction in which the magnetization direction of the left and right leads is the same , which is indicative of the fm mn - mn coupling for larger distances . in fig .
[ fig : stretch](d ) we present the results of our calculations for the evolution of the ballistic conductance of a p@xmath108@xmath108 junction when going from the tunneling to the contact regime .
the main contribution to the conductance at large mn - mn distances is coming solely from the @xmath97 channel , owing to the overlap between the @xmath3-@xmath4 orbitals of the neighboring mn atoms across the barrier . within our approach ,
the expected exponential behavior of the conductance at very large distances is very nicely reproduced . at a distance of @xmath157 bohr
the conductance approaches the magnitude of the conductance quantum , reaching saturation upon further decreasing the distance .
for distances in the contact regime below 7 bohr more localized @xmath158-orbitals of @xmath99 symmetry start contributing to the total conductance , as can be seen in fig .
[ fig : stretch](d ) .
the @xmath99 contribution to the conductance increases with decreasing distance . as we shall see in the following the details of hybridization between the @xmath99 orbitals are very sensitive to the magnetic state of the junction . on the other hand , in all considered cases the @xmath158-states of @xmath101 symmetry do not contribute to the conductance due to an energetic mismatch between the states of this symmetry of nn co and mn atoms , see discussion in section [ subsec : noco_junction ] . according to the findings presented above
, we expect that mn spin moments in the contact regime will experience a frustration when the magnetizations of the leads are parallel to each other . in this case , when mn atoms are close enough , fm coupling of mn spins with nn co atoms and afm mn - mn coupling can possibly lead to a stable non - collinear magnetic state . in order to consider this situation
, we introduce an angle @xmath48 between the spin moments of the mn and the nn co atoms , rotating the first mn spin moment by @xmath159 and the second one by @xmath160 , while keeping the moments of the co atoms fixed , see fig . [
fig : config](b ) .
this is what we call a symmetric p@xmath48-state .
we choose a distance of 4.5 bohr between the mn atoms as a representative of the contact regime at which the mn - mn coupling is strongly antiferromagnetic .
the results of our calculations for the total energy of the p@xmath48 state , @xmath161 , in relation to the energy of the p@xmath108@xmath108 state are shown in fig .
[ fig : rotate](a ) as a function of the angle @xmath48 . from this plot
we observe that the minimum of the total energy is acquired for the non - collinear p@xmath162 state , which is @xmath163 mev lower in energy than the corresponding collinear p@xmath108@xmath108 state .
the failure of a straightforward description of the energy landscape @xmath161 in terms of a simple heisenberg model which assumes just the nearest - neighbor co - mn and mn - mn exchange coupling , given by antiferromagnetic @xmath164 and ferromagnetic @xmath165 , respectively , can be understood from noticing that the expression for the energy within this approximation , given by @xmath166 acquires a minimum for angles @xmath48 below @xmath167 , in contradiction to our calculations .
state with respect to the p@xmath108@xmath108 state as a function of angle @xmath48 ( solid line , filled circles ) . with open black squares
we show the total energy of the p@xmath48 state from which the stoner energy due to creation of the nn co spin moments is subtracted .
the corresponding fit of the corrected energy to the heisenberg model is shown with a dashed blue line . with the red dotted line
we show the total energy of the p@xmath108@xmath109 state with only the @xmath106-mn spin rotated by an angle @xmath48 , calcualted within the heisenberg model and incorporating also the stoner exchange .
( b ) mn spin moment as a function of angle @xmath48 .
( c ) nn co spin moment as a function of angle @xmath48 .
( d ) ballistic conductance at the fermi level for the p@xmath48 state as a function of angle @xmath48 : total conductance @xmath142 ( upper line ) , the @xmath99 conductance contribution ( cyan shaded area ) and and @xmath97 conductance @xmath143 ( lower line , gray shaded area ) . , scaledwidth=47.0% ] the solution to this deficiency of the heisenberg model can be given by lifting the assumption that the exchange interaction between the mn and co spins , given by @xmath168 , is ferromagnetic .
as we can see from fig .
[ fig : rotate](b ) and ( c ) , while the mn spin moment remains relatively constant upon changing @xmath48 , @xmath141 for values of @xmath48 below 60@xmath169 is by as much as @xmath170 larger than for @xmath171 .
owing to the intra - atomic stoner exchange , the non - collinear states with small @xmath48 therefore acquire a negative contribution to the total energy in addition to that proportional to @xmath168 , as compared to larger angles .
if we account for the energy gain due to creation of the nn co spin moments by a stoner parameter of co , @xmath172 mev , @xcite and subtract the energy gain @xmath173 from the calculated dft dispersion , we arrive at the energy dispersion ( squares in fig .
[ fig : rotate](a ) ) , which reflects only exchange interactions between the atoms .
if we fit this curve according to eq.([heisenberg ] ) ( dashed line in fig .
[ fig : rotate](a ) ) , we obtain the `` non - renormalized '' heisenberg exchange constants of @xmath174 mev and @xmath175 mev .
it becomes clear now , that , although the `` pure '' exchange coupling between the mn and co spins is expectedly antiferromagnetic , the larger spin moment of co in the parallel spin alignment with mn , tips the balance in favor of ferromagnetic coupling between the spins , which can be observed for a large range of distances @xmath111 , c.f .
[ fig : stretch](a ) . in fig .
[ fig : stretch](a ) we also observe that , judging from the energies , in the close contact regime the collinear p@xmath109@xmath108-state is competing with the non - collinear p@xmath48-state for the global ground state of the system . indeed , our calculations show that at the @xmath111 of 4.5 bohr the p@xmath109@xmath108-configuration is by a tiny value of 5 mev lower in energy than the p105@xmath169 solution .
we argue , however , that the p@xmath109@xmath108-state is not very likely to appear in experiments , given that the co electrodes are identical . in this case , the adiabatic rise of the intrinsically asymmetric p@xmath109@xmath108-configuration via symmetric non - collinear states can not happen , as the electrodes , initially being in the p@xmath108@xmath108-state when very far from each other , are brought together ( see also discussion at the beginning of section [ subsec : t - to - c_rev ] ) . nevertheless , it seems plausible , that such state , if observed in experiment , is created via a rapid flip of one of the mn atoms in the contact regime , during , e.g. , a reformation of the lead geometry , or an inelastic current - induced spin - flip process .
our calculations , shown with a dotted line in fig . [ fig : rotate](a ) , based on the heisenberg model extended by the stoner term of intra - atomic exchange of the co moments , indicate , that once the system enters the p@xmath109@xmath108-state , it is effectively `` trapped '' there , since the @xmath109-mn is energetically quite stable versus deviations in the angle its spin makes with the rest of the spins in the system .
thus , we do not consider any non - collinear states associated with the p@xmath109@xmath108-state in the following . in this section
we perform a detailed analysis of the ballistic conductance @xmath176 of the p@xmath48 state at fixed distance between mn atoms of 4.5 bohr . at this distance
, we calculate @xmath176 as a function of angle @xmath48 and present the results in fig .
[ fig : rotate](d ) . in this plot
we observe that the conductance exhibits a very non - trivial dependence on @xmath48 , originating mainly from the @xmath99-orbitals ( @xmath8 ) , while the @xmath97 contribution ( @xmath177 ) to the conductance , @xmath178 , remains almost perfectly constant .
surprisingly , the @xmath99-conductance almost vanishes for @xmath48 of about 70@xmath169 , away from any high - symmetry spin configuration in the junction , suggesting , that the dependence of the details of hybridization and electronic structure on the angle between the mn spins can be rather delicate . in order to analyze this dependence in more detail , as a function of @xmath48 , we plot the energy - dependent conductance , @xmath179 , versus the local densities of states ( ldos ) of mn and nn co atoms resolved into spin - up and spin - down contributions with respect to the global spin quantization @xmath55-axis , fig . [
fig : t_vs_dos ] .
mainly we focus on the @xmath99-contribution to the conductance and the ldos , and only in the upper ( p@xmath108@xmath108 ) and lower ( p@xmath109@xmath109 ) panels of fig .
[ fig : t_vs_dos ] we show also the total ldos of the atoms .
the conductance at a given energy @xmath180 depends on the presence of available states in the ldos of the atoms at @xmath180 , and on the coupling between these states across the junction @xmath114 both of which depend on the orientation of the spins with respect to each other . by looking at the ldos of the atoms presented in fig .
[ fig : t_vs_dos ] for @xmath181 we can explain the absence of the @xmath101 ( @xmath182 ) contribution to the conductance : the localized @xmath101 states of the co atoms , which can be seen as pronounced peaks in the ldos marked with the dashed line in fig .
[ fig : t_vs_dos ] , are positioned at about @xmath1142 ev for spin - up channel and directly at the fermi energy for spin - down channel , while the corresponding mn @xmath101 states are positioned below @xmath1142.5 and above @xmath1831 ev , prohibiting thus the hybridization between the co and mn orbitals of @xmath101 symmetry across the junction .
noticeably , the ldos of both atoms for the up - spin in a wide region of energies around @xmath184 is absent , leading to a negligible @xmath108-conductance . here , it is important to remark , that the ldos of the nn co atoms around the fermi energy overall resembles quite well the ldos of a co atom in a co monowire , see e.g. fig .
13(a ) in ref . , or even of a co atom deposited on a noble - metal surfaces , see e.g. ref .
this means that our results should be rather stable with respect to the geometry of the co leads , manifesting that the main influence on the @xmath99 conductance at @xmath184 would come from the hybridization of the mn and nn co states .
turning now to the comparatively delocalized @xmath99-states ( solid line ) on both co and mn atoms , we observe for p@xmath132 ( upper panels of fig .
[ fig : t_vs_dos ] ) that they hybridize directly at the fermi energy , which leads to a significant @xmath99 contribution to the conductance .
specifically , while the @xmath185 subband of co spreads from @xmath1141.8 to @xmath1831 ev , the @xmath99-down states of mn atoms are distinctly split into wide bonding ( `` b '' ) states at the fermi energy and narrow anti - bonding ( `` a '' ) states at @xmath1831.8 ev . very importantly for the transport properties of the system , the hybridization of the @xmath185 band of co with the @xmath185 states of mn is non - trivial .
( i ) the @xmath185 states of co exhibit a dip at the position of the maximal density of bonding states of mn due to the fact that these mn states are localized mainly in between the mn atoms prohibiting strong overlap with the co states .
( ii ) the upper , antibonding , part of the co @xmath185 band hybridizes stronger with the bonding states of mn , since the antibonding states of co atoms have a larger overlap with the mn orbitals , which results in a larger @xmath99-conductance above @xmath184 .
( iii ) analogously , for energies below @xmath184 the conductance is suppressed , since the bonding - like @xmath185 co states have smaller overlap with the mn bonding states .
let us now follow the evolution of the electronic structure upon increasing the angle between the mn spins .
two trends in the ldos can be clearly observed in fig .
[ fig : t_vs_dos ] .
firstly , with increasing @xmath48 the splitting between the bonding and antibonding mn states decreases owing to the mixed spin character of the states . at the angle of 90@xmath169 ,
when mn spins are anti - parallel to each other , both types of states transform into degenerate @xmath99-orbitals of the `` isolated '' mn atoms , since the hybridization between the mn states of the same spin is almost absent due to large exchange splitting . on the other hand ,
the dip in the @xmath109-ldos of the nn co atoms follows the position of the bonding state of the mn dimer , moving from the fermi energy at @xmath181 to @xmath1830.2 ev for @xmath186 ( indicated by filled triangles in fig .
[ fig : t_vs_dos ] ) .
overall , such redistribution of the ldos of the atoms combined with the effect of decreasing ldos of mn atoms for spin - down channel at the fermi energy when the angle @xmath48 is varied , results first in a decrease of the conductance at @xmath184 for @xmath187 , followed by a consequent increase with increasing angle . when the angle @xmath48 increases further beyond 90@xmath169 , the bonding and anti - bonding mn states eventually acquire their initial splitting at @xmath188 ( p@xmath109@xmath109-state ) , when the mn spins are collinear again . simultaneously , with increasing angle
, we observe that the mn states around the fermi energy become sharper , since the hybridization with the co leads decreases as the mn states become predominantly spin - up in character .
interestingly , while for @xmath189 a large value of the @xmath99-conductance is due to a significant amount of delocalized co and bonding mn states at the fermi energy in the spin - down channel , for larger angles the value of @xmath190 is due to a sharp resonant co state in the spin - up channel at the fermi energy , coupled to a bonding mn state at @xmath184 . when further increasing @xmath48 above 170@xmath169 , this resonance becomes more localized and decoupled from the states in the leads , while the mn ldos at the @xmath184 in the minority spin - channel vanishes , causing a sharp drop in the @xmath99-conductance . by looking at the total ldos of the nn co atom in the p@xmath109@xmath109-state
we observe that it remains basically unaffected , as compared to the p@xmath108@xmath108-configuration , while the mn states become pronouncedly decoupled from the states of the nn co owing to the energetical mismatch for both spin channels . finally , we investigate the evolution of the conductance and the magnetoresistance of different magnetic states of the junction within the contact regime mimicking a typical stm or break junction experiment . here , we are partly motivated by the fact that a non - trivial behavior of magnetoresistance when going from tunelling to contact has been recently observed in stm experiments , see e.g. ref . .
at a very large separation between the leads ( or , the tip and the sample in the stm language ) , owing to the fm coupling of the mn atom to the co chain , one can imagine only two possible magnetic configurations @xmath114
p@xmath108@xmath108 and ap@xmath108@xmath109 .
the conductance of these two magnetic states in the tunneling regime , arising mainly from the @xmath3-orbitals , is orders of magnitude smaller than in the contact regime , for which the dependence of @xmath176 on the distance can be non - trivial due to the large contribution of the @xmath158-states . in the case of the ap@xmath108@xmath109 configuration
, the starting collinear arrangement of the spins will survive over the whole range of the separation between the leads , since in the contact regime , when the mn atoms are close to each other , both exchange preferences of the mn spins , that is , fm coupling to the nn co spins and afm coupling among each other , are fulfilled .
small possible deviations from the collinear arrangement of the mn spins , which can affect the details of the distribution of the @xmath99-states and their coupling to the leads , would not manifest in a conductance measurement , owing to the antiparallel magnetizations of the leads , and corresponding complete dominance of the @xmath97 channel for conductance at @xmath184 in this case , fig .
[ fig : g ] . as we can see from this figure
, @xmath176 lies in between 0.5 and 1.0@xmath191 , when the distance between the mn atoms is varied from 8.5 to 4.5 bohr .
this is very similar to the behavior of the conductance at the fermi energy of pure ap co leads without mn atoms , see e.g. fig .
11(b ) of ref . .
owing to the magnetic frustration of the mn spins of the junction in the contact regime , for the p@xmath108@xmath108 initial configuration , we consider the p@xmath48 and p@xmath109@xmath108 states in addition to the p@xmath108@xmath108 state when the mn - mn distance is relatively small . here , as we have seen in the preceding section , the conductance at the fermi energy can be very strongly influenced by the details of hybridization between the @xmath99-orbitals . on the other hand , since very often transport measurements serve as the only experimental insight into the magnetic structure of a system , it is very important to coin each of the possible magnetic states with a unique fingerprint which can be related to the experimental data .
below , we suggest that indeed three distinct spin states in a @xmath103-@xmath104-@xmath104-@xmath103 junction @xmath114 which can occur in an experiment due to various reasons such as structural details , temperature fluctuations , external magnetic field , etc .
@xmath114 lead to different transport signatures . as already shown in fig .
[ fig : stretch](d ) , the conductance of the collinear p@xmath108@xmath108 state rapidly rises towards a value of 1.8@xmath191 as the distance between the leads is decreased .
compared to other possible magnetic configurations of the junction , @xmath192 is significantly larger in value , see fig .
[ fig : g ] , because of the alignment of the minority spin @xmath3- and @xmath158-states of the co electrodes and the mn atoms at the fermi energy , which ideally favors perfect transmission .
in contrast , the conductance of the collinear p@xmath109@xmath108-state is significantly suppressed , reaching only 1.0@xmath191 at the separation of 4.5 bohr , due to the large exchange splitting of the @xmath99 states of the mn atoms with antiparallel spin moments which hinders the @xmath99 conductance .
the conductance of the non - collinear ground - state p@xmath48-state lies in between the values for both limiting collinear configurations . in the close contact regime , at @xmath111 of 4.5 bohr , the conductance of the p@xmath48-state of 1.4@xmath191 is exactly in between the values of @xmath192 and @xmath193 .
clearly , the difference of 0.4@xmath191 , stemming from the variation in the @xmath99-conductance with the spin state , can be easily detected in experiment , allowing for a way to distinguish between different possible magnetic configurations . at the distance of 5.0
bohr the ground state among the p@xmath48 states is the p90@xmath169 state , while at larger distances above 5.5 bohr the system converges to a collinear configuration .
the angle @xmath48 in the lowest in energy p@xmath48 state decreases smoothly with increasing the separation , and we speculate , that owing to the non - monotonous behavior of the conductance as a function of @xmath48 , seen in fig .
[ fig : rotate](d ) , the conductance as a function of @xmath194 , can exhibit several features similar to that at @xmath194 of 5.0 bohr , although we did not perform the calculations to support this statement owing to the required computational effort .
-@xmath104-@xmath104-@xmath103 junction as a function of the distance between the mn atoms . following magnetic states
are considered : ap@xmath109@xmath108 ( filled circles , solid line ) , p@xmath108@xmath108 ( open circles , dashed line ) , p@xmath109@xmath108 ( diamonds , dash - dash - dotted line ) , and p@xmath48 ( squares , dot - dashed line ) . the @xmath97 contribution for the p@xmath108@xmath108-state is shown with a thin solid line and grey shaded area , while the @xmath99 part is shaded in cyan . for the p@xmath48 states the state which is lowest in energy among all possible angles @xmath48 at a fixed distance is considered . in the inset
the values of the magnetoresistance for different p - states are shown .
, scaledwidth=43.0% ] according to recent experiments,@xcite the conductance of the junction with the parallel ( p ) and antiparallel ( ap ) orientation of the lead s magnetization can be related to each other via measuring the magnetoresistance ( mr ) . from the values presented in fig .
[ fig : g ] we calculate the mr of the junction , defined as : @xmath195 and present the mr as a function of separation between the electrodes in the inset of fig . [ fig : g ] , where we choose @xmath196 for @xmath197 , and values of @xmath192 , @xmath193 and @xmath198 for @xmath199 .
the overall smaller ap@xmath109@xmath108 conductance as compared the p - configurations results in positive magnetoresistance values .
the mr curves as a function of the distance generally resemble those of the conductance , with the values of the mr of 22 , 62 and 105% at the distance of 4.5 bohr for the p@xmath109@xmath108 , p@xmath48 and p@xmath108@xmath108-states , respectively .
much more pronounced in the mr is the feature characteristic to the p@xmath109@xmath108 and p@xmath48-configurations @xmath114 a dip around the distance of 5.0 bohr , also present in the conductance curves . as can be seen from fig .
[ fig : g ] , at this distance , the mr almost completely vanishes when the mn spins exhibit a different from fm configuration .
overall , we conclude , that the pronounced difference in the shape and magnitude of the mr curves can be also used in experiments to shed light onto the complex magnetism in this type of systems .
in this work we presented the realization of a first principles scheme for calculating the ballistic transport properties of magnetically complex one - dimensional systems employing the technique of non - collinear wannier functions .
we use the flapw method in order to calculate the electronic structure of the system with high accuracy and use the wannier functions to transfer it to our transport calculations performed within the landauer approach . as spin - orbit interaction can be naturally included into the consideration within this technique , c.f
. ref .
, the method introduced here can be used to explore the rich field of transport phenomena in systems such as nano- or atomic - sized contacts , break junctions , or stm experiments for which both effects , spin - orbit coupling and frustrated exchange interactions , can be prominent . as a first application of our approach , we consider the ballistic transport properties of a single - atom junction formed by two semi - infinite co electrodes with a single apex mn atom .
we study the conductance as a function of the separation between the two mn atoms from the tunneling to the contact regime taking into account the complex magnetic interaction in the junction .
as we demonstrate , even such a simple setup allows to draw some general conclusions concerning the interplay of structure and magnetism for the transport through such atomic - sized contacts which are in the focus of today s research .
we analyze the ballistic conductance of the junction with lead magnetizations in parallel and antiparallel alignment .
we consider separately the tunneling ( separation larger than about 5@xmath200 ) and contact ( below 5@xmath200 ) regimes of the junction , and we demonstrate that in the tunneling regime the conductance @xmath201 is solely coming from the overlap between the @xmath97 ( @xmath3-@xmath4-orbitals ) of the contacts . in this case
the mn spins prefer to order ferromagnetically with respect to the magnetization of the leads . on the other hand , upon reaching the close contact regime ( below 3.5@xmath200 ) , when the magnitude of the conductance reaches 1@xmath191 , the hybridization between the @xmath99 ( @xmath202-orbitals ) states of the junction starts to provide a sizable contribution to @xmath201 . in the close contact regime ,
when the hybridization between the mn atoms is significant , mn spins experience a frustration due to the fm coupling with the leads and an afm mn - mn coupling .
the competition between the two gives rise to a stable non - collinear solution which can be characterized by a tilting angle of the spins , @xmath48 .
general for this type of junction is the sensitivity of the @xmath158-orbital conductance on the angle @xmath48 , which is due to a delicate interplay between the hybridization details of the mn and co states at the fermi energy , as well as spin - asymmetry in their distribution .
this gives rise to a non - trivial @xmath48-dependence of the conductance of the @xmath158-states .
we show that the complicated @xmath99-channel conductance arising on the background of almost constant @xmath97 contribution can be used in order to distinguish between different magnetic states of the contact via either a direct conductance measurement , or via measuring the magnetoresistance , which , according to our calculations , can vary in the contact regime between 20 and 100% , depending on the spin arrangement .
finally , we would like to comment on our approximation for the geometry of the junction we have assumed in this work .
albeit being very simple , it allows to capture the key features which govern the transport properties of the system , while keeping the computational burden reasonable .
namely , within this geometry : ( i ) the transition from tunneling to contact can be naturally studied ; ( ii ) the magnetic frustration of the spins in the junction , and ( iii ) the delicate details of the hybridization of the adatom with the lead reservoirs are taken into account ; ( iv ) the sensitive dependence of the spin moments on the magnetic configuration in the nano - contact is included into our considerations .
of course , in order to achieve a quantitative agreement of the calculated values to the experimentally measured ones in this type of junction beyond the major trends , all details of the structure and structural reformation upon approaching should be ideally accounted for .
such a challenging study lies , however , outside of the scope of the current work , and we leave it for future studies .
we acknowledge helpful discussions with stefan blgel .
funding by the dfg within the sfb677 is gratefully acknowledged .
thanks the dfg for financial support under he3292/8 - 1 . y.m . and f.f .
gratefully acknowledge the jlich supercomputing centre for computing time and funding under the hgf - yig programme vh - ng-513 .
|
we present a first - principles computational scheme for investigating the ballistic transport properties of one - dimensional nanostructures with non - collinear magnetic order .
the electronic structure is obtained within density functional theory as implemented in the full - potential linearized augmented plane - wave ( flapw ) method and mapped to a tight - binding like transport hamiltonian via non - collinear wannier functions .
the conductance is then computed based on the landauer formula using the green s function method . as a first application we study the conductance between two ferromagnetic co monowires terminated by single mn apex atoms as a function of mn - mn separation .
we vary the mn - mn separation from the contact ( about @xmath0 to @xmath1 ) to the far tunneling regime ( @xmath1 to @xmath2 ) .
the magnetization direction of the co electrodes is chosen either in parallel or antiparallel alignment and we allow for different spin configurations of the two mn spins . in the tunneling and into the contact regime
the conductance is dominated by @xmath3-@xmath4-states . in the close contact regime ( below @xmath5 ) there is an additional contribution for a parallel magnetization alignment from the @xmath6- and @xmath7-states which give rise to an increase of the magnetoresistance as it is absent for antiparallel magnetization .
if we allow the mn spins to relax a non - collinear spin state is formed close to contact due to the competition of ferromagnetic coupling between mn and co and antiferromagnetic coupling between the mn spins .
we demonstrate that the transition from a collinear to such a non - collinear spin structure as the two mn atoms approach leaves a characteristic fingerprint in the distance - dependent conductance and magnetoresistance of the junction .
we explain the effect of the non - collinear spin state on the conductance based on the spin - dependent hybridization between the @xmath8-states of the mn spins and their coupling to the co electrodes .
|
the dramatic increase in mobile applications and the number of devices demanding for wireless connectivity poses a challenge in today s wireless networks @xcite , and calls for new networking mechanisms .
one of the promising solutions to address the increasing data and connectivity demand is device - to - device ( d2d ) networking . as illustrated in fig .
[ fig : intro_example_new](a ) , the default operation in current wireless networks is to connect each device to the internet via its cellular or wi - fi interface . the d2d connectivity idea , which is illustrated in fig .
[ fig : intro_example_new](b ) , breaks this assumption : it advocates that two or more devices in close proximity can be directly connected , _
i.e. , _ without traversing through auxiliary devices such as a base station or access point .
d2d networking , that can be formed by exploiting d2d connections such as wi - fi direct @xcite , is a promising solution to the ever increasing number and diversity of applications and devices . in this context
, it is crucial to identify scarce resources and effectively utilize them to fully exploit the potential of d2d networking .
although bandwidth is traditionally considered as the primary scarce resource in wireless networks , in d2d networks , thanks to close proximity among devices and the developments in communication theory , the main bottleneck shifts from bandwidth to other scarce resources including processing power and energy .
next , we present our pilot study demonstrating that processing power can be more pronounced as a bottleneck than bandwidth in d2d networks . [
fig : intro_example_new ] we developed a prototype for this pilot study as shown in fig .
[ fig : complexityvsrate_new](a ) , where a mobile device @xmath0 receives data from another device @xmath1 over a wi - fi direct link .
we use android operating system @xcite based nexus 7 tablets @xcite as mobile devices . in this experiment , after receiving the packets , the mobile device @xmath0 performs operations with complexities of @xmath2 , @xmath3 , and @xmath4 above the transport layer ( tcp ) , where @xmath5 is the packet size , and the operations we perform are counting the bytes in the packets .
in particular , @xmath2 , @xmath3 , and @xmath4 correspond to ( i ) no counting , ( ii ) counting every byte in a packet once , and ( iii ) counting every byte in a packet @xmath5 times , respectively .
we demonstrate in fig .
[ fig : complexityvsrate_new](c ) the received rate at the mobile device @xmath0 ( note that this is the rate we measure at the mobile device @xmath0 after performing computations ) versus time .
this figure demonstrates that the received rate decreases significantly when the complexity increases .
our pilot study shows that even if actual bandwidth is high and not a bottleneck , processing power could become a bottleneck in d2d networks .
similar observations are made for the energy bottleneck as detailed in appendix a. furthermore , with the advances in communication theory , _
millimeter wave communication @xcite , it is expected that data rates among devices in close proximity will increase significantly , which will make processing power and energy more pronounced as bottlenecks .
however , existing applications , algorithms , and protocols are mainly designed by assuming that bandwidth is the main bottleneck .
thus , it is crucial to develop new networking mechanisms when bandwidth is not the primary bottleneck , but processing power and energy are .
[ fig : complexityvsrate_new ] thus , in this paper , our goal is to create group of devices that help each other cooperatively by exploiting high rate d2d connections to overcome the processing power and energy bottlenecks .
the next example demonstrates our approach .
[ ex1 ] let us consider fig .
[ fig : intro_example_new](a ) again , where device @xmath1 would like to receive a file from a remote resource via using its cellular or wi - fi connection .
assume that the cellular ( or wi - fi ) rates of all devices are 1mbps , but device @xmath1 can receive data with 500kbps rate due to processing power bottleneck , _
device @xmath1 has limited processing power ( similar to our pilot study we presented earlier ) . in a traditional system
, @xmath1 will behave as a single end point , so its receiving rate will be limited to 500kbps . on the other hand ,
if devices @xmath1 , @xmath0 , and @xmath7 will behave as a group and cooperate , then devices @xmath0 and @xmath7 can also receive and process 500kbps portions of data , and transmit the processed data to device @xmath1 over d2d connections .
this increases the receiving rate of device @xmath1 to 1.5mbps from 500kbps , which is a significant improvement .
this example could be extended for scenarios when energy ( battery of mobile devices ) is limited .
for example , if device @xmath0 s battery level is too low , its participation to the group activity should be limited .
@xmath6 the scenario in the above motivating example could arise in different practical applications from health , education , entertainment , and transportation systems .
the following are some example applications .
_ health : _ a person may own a number of health monitoring devices ( activity monitoring , hearth monitoring , etc . ) which may need updates from the core network . these updates - potentially coded for error correction , compression , and security reasons - should be processed ( decoded ) by these devices .
processing takes time , which may lead to late reaction to the update ( which may require timely response ) and energy consumption . on the other hand , by grouping mobile devices , the person s smartphone or tablet could receive the update , process , and pass the processed data to the health monitoring devices via high rate d2d links .
_ education & entertainment : _ a group of students may want to watch the video of a lecture from an online education system ( or an entertainment video ) while sitting together and using several mobile devices . in this setup , one of the devices can download a base layer of a video and decode , while the other devices could download enhancement layers and decode .
the decoded video layers could be exchanged among these mobile devices via high rate d2d links . as in the motivating example , if one device s download and decoding rate is limited to 500kbps , it could be improved to 1.5mbps with the help of other devices .
note that the processing overhead in these applications could be due to any computationally intensive task related to data transmission .
for example , for video transmission applications , h.264/avc decoders introduce higher computational complexity when higher quality guarantees are needed @xcite .
another example could be network coding ; for example , data could be network coded at the source to improve throughput , error correction , packet randomization potential of network coding @xcite .
however , most of the network coding schemes introduce high computational complexity at the receiver side ; @xmath8 , @xcite , which limits the transmission rate .
encryption could be another example that introduces processing overhead @xcite .
thus , there exists several applications and scenarios where bandwidth and energy could be bottlenecks , while bandwidth is not the bottleneck .
this makes our approach demonstrated in example [ ex1 ] crucial .
in particular , in this paper , we develop an _ energy - aware cooperative computation _ framework for mobile devices . in this setup , a group of cooperative mobile devices , within proximity of each other , ( i ) use their cellular or wi - fi ( 802.11 ) links as their primary networking interfaces , and ( ii ) exploit their d2d connections ( wi - fi direct ) for cooperative computation .
our approach is grounded on a network utility maximization ( num ) formulation of the problem and its solution @xcite .
the solution decomposes into several parts with an intuitive interpretation , such as flow control , computation control , energy control , and cooperation & scheduling .
based on the structure of the decomposed solution , we develop a stochastic algorithm ; _ energy - aware cooperative computation_. the following are the key contributions of this work : * we consider a group of cooperative mobile devices within proximity of each other . in this scenario ,
we first investigate the impact of processing power to transmission rate .
then , we develop an energy - aware cooperative computation model , where devices depending on their energy constraints could cooperate to get benefit of aggregate processing power in a group of cooperative devices .
* we characterize our problem in a num framework by taking into account processing power , energy , and bandwidth constraints .
we solve the num problem , and use the solution to develop our stochastic algorithm ; _ energy - aware cooperative computation ( eacc)_. we show that eacc provides stability and optimality guarantees . * an integral part of our work is to understand the performance of eacc in practice . towards this goal
, we develop a testbed consisting of nexus 5 smartphones and nexus 7 tablets .
all devices uses android 5.1.1 as their operation systems .
we implement _ eacc _ in this testbed , and evaluate it .
the experimental results show that our algorithm brings significant performance benefits .
the structure of the rest of the paper is as follows .
section [ sec : related ] presents related work .
section [ sec : system ] gives an overview of the system model .
section [ sec : num ] presents the num formulation of our cooperative computation scheme .
section [ sec : coopcomp ] presents our stochastic algorithm ; _ eacc_. section [ sec : performance ] evaluates the performance of our scheme in a real testbed .
section [ sec : conclusion ] concludes the paper .
this work combines ideas from d2d networking , network utility maximization , and stochastic network control .
the idea of d2d networking is very promising to efficiently utilize resources , so it has found several applications in the literature . in particular ,
d2d connections are often used to form cooperative groups for data streaming applications , and for the purpose of ( i ) content dissemination among mobile devices @xcite , ( ii ) cooperative video streaming over mobile devices @xcite , and ( iii ) creating multiple paths and providing better connectivity by using multiple interfaces simultaneously @xcite . as compared to this line of work
, we investigate the impact of processing power and energy in d2d networks , and develop mechanisms to effectively utilize these scarce resources .
d2d networking is often used for the purpose of offloading cellular networks .
for example , previous work @xcite disseminate the content to mobile devices by taking advantage of d2d connections to relive the load on cellular networks . instead of offloading cellular networks ,
our goal is to create energy - aware cooperation framework to overcome the processing power and energy bottlenecks of mobile devices .
there is an increasing interest in computing using mobile devices by exploiting connectivity among mobile devices @xcite .
this approach suggests that if devices in close proximity are capable of processing tasks cooperatively , then these devices could be used together to process a task as it is a cheaper alternative to remote clouds .
this approach , sparking a lot of interest , led to some very interesting work in the area @xcite , @xcite , @xcite .
as compared to this line of work , we focus on processing power and energy bottlenecks in mobile devices and address the problem by developing energy - aware cooperative computation mechanism .
an integral part of our proposed work in this task is to develop efficient resource allocation mechanisms . in that sense ,
our approach is similar to the line of work emerged after the pioneering work in @xcite , @xcite , @xcite .
however , our focus is on energy - aware cooperative computation .
we consider a cooperative system setup with @xmath9 mobile devices , where @xmath10 is the set of the mobile devices .
our system model for three nodes are illustrated in fig .
[ fig : systemmodel](a ) .
the source in fig .
[ fig : systemmodel](a ) represents the core network and base stations ( access points ) .
this kind of abstraction helps us focus on the bottlenecks of the system ; processing power , energy of mobile devices , and downlink / uplink data rates . in this setup ,
mobile devices communicate via d2d connections such as wi - fi direct , while the source communicates with mobile devices via cellular or wi - fi links .
we consider in our analysis that time is slotted and @xmath11 refers to the beginning of slot @xmath11 .
[ fig : systemmodel ] * connecting devices together : * the total flow rate towards device @xmath5 in fig .
[ fig : systemmodel](a ) ( and also explained in fig .
[ fig : systemmodel](b ) ) is @xmath12 , where @xmath13 is the transmission rate of the packets from the source towards device @xmath5 , and these packets will be used by device @xmath5 .
note that @xmath14 is the transmission rate of the packets from the source towards device @xmath5 , and these packets will be processed by device @xmath5 and forwarded to device @xmath15 . on the other hand , @xmath16 is the total flow rates targeting device @xmath5 as demonstrated in fig .
[ fig : systemmodel](b ) .
the source constructs a queue @xmath17 for the packets that will be transmitted to the mobile device @xmath5 .
the evolution of @xmath18 based on @xmath16 and @xmath19 is expressed as @xmath20 + y_{n}(t ) , \end{aligned}\ ] ] where the inequality comes from the fact that there may be less than @xmath21 packets arriving into @xmath17 at time @xmath11 in practice ( _ e.g. , _ in real time applications , the number of available packets for transmission could be limited ) .
the flow rate @xmath16 is coupled with a utility function @xmath22 , which we assume to be strictly concave function of @xmath16 .
this requirement is necessary to ensure stability and utility optimality of our algorithms .
the ultimate goal in our resource allocation problem is to determine the flow rates ; @xmath16 which maximize the sum utility @xmath23 .
finally , flow rate over d2d connection between device @xmath5 and @xmath15 is @xmath24 , @xmath25 .
note that @xmath24 is to help node @xmath15 using node @xmath5 as a processing device .
* inside a mobile device : * in each device , we develop different modules depending on where data is arriving from ( as shown in fig .
[ fig : systemmodel](c ) ) ; _ i.e. , _ from the source via cellular or wifi interface , or other mobile devices via d2d interfaces .
when data is arriving from a d2d interface , it is directly passed to the application layer , as this data is already processed by another device . on the other hand , when data is arriving from the source via cellular or wifi interfaces , packets go through multiple queues as shown in fig .
[ fig : systemmodel](c ) , where @xmath26 , @xmath27 , and @xmath28 represent three different queues constructed at mobile device @xmath5 for the purpose of helping node @xmath15 .
incoming packets via cellular or wifi links are stored in @xmath26 , which then forwards the packets to _ computation _ block with rate @xmath29 .
the computation block processes the packets , and pass them to queue @xmath27 . note that the output rate from computation block is @xmath30 , where @xmath31 is a positive real value .
this value captures any possible rate changes at the computation block , _
@xmath31 is a rate shaper .
for example , if the computation block is h.264/avc decoder or transcoder , we expect that the rate at the output of the computation block should be higher than the input .
thus , @xmath31 captures this fact for any @xmath32 . on the other hand , if there is no rate change after the processing , then @xmath33 .
the processed ( and possibly rate shaped ) packets are queued at @xmath34 and passed to _
energy filter_. the energy filter is coupled to the energy source , which determines the amount of energy that can be spent to support the tasks at each slot .
the amount of energy is determined according to _
energy credits_. in particular , the energy source , depending on the battery level as well as the estimate on the expected battery consumption in the near future , calculates the number of packets that can be supported by the mobile device , and the same number of energy credits enter the energy filter .
( note that both energy filter , energy source , and energy credits are not real , but virtual entities , so they can be modeled by using a few counters in practice . ) thus , at each transmission slot , packets are transmitted from @xmath34 to @xmath35 if there are energy credits in the energy filter .
finally , packets from @xmath35 are transmitted to application if device @xmath5 is the destination of the data ( _ i.e. , _
@xmath36 ) , or they are transmitted to the original destination via d2d interface with rate @xmath24 . the computation and energy filter blocks in fig .
[ fig : systemmodel](c ) model the processing and energy bottlenecks of the mobile device , respectively .
if packets in @xmath26 increases too much , this means that the computation block , hence processing power , is the bottleneck , so node @xmath5 should not receive much packets from the source .
similarly , if @xmath27 increases too much , this means that energy filter is the bottleneck , so again node @xmath5 should not receive much packets . note that there could be also some buildup in @xmath28 if the link between node @xmath5 and @xmath15 is the bottleneck of the system , and it should be taken into account when the energy - aware cooperative computation framework is developed . also , it is crucial in our system model to put energy filter after the computation block , because if device @xmath5 will help device @xmath15 , the actual amount of packets that are supposed to be transmitted are the processed packets , which will cause energy consumption ( _ i.e. , _ not the packets before processing ) .
based on the above intuitions and observations , we will develop our resource allocation problem and algorithm in the next sections .
the evolution of the queues @xmath37 , @xmath34 , and @xmath35 are provided in table [ table : queueevolution ] .
.evolution of queues @xmath37 , @xmath34 , and @xmath35 . [ cols="^",options="header " , ] [ table : queueevolution ] * links : * in our system model , we consider two scenarios : ( i ) cellular + wi - fi direct , and ( ii ) wi - fi + wi - fi direct . in both cases , the d2d links between mobile devices are wi - direct . in the first case , _ i.e. , _ in cellular + wi - fi direct ,
the links between the source and mobile devices are cellular , while they are wi - fi in the second case , _ i.e. , _ in wi - fi + wi - fi direct .
these two scenarios are different from each other , because in the first scenario , cellular and wi - fi direct links could operate simultaneously as they use different parts of the spectrum . on the other hand , in the second scenario , both wi - fi and wi - fi direct use the same spectrum , so they time share the available resources .
our model and energy - aware cooperative computation framework are designed to operate in both scenarios .
next , we provide details about our link models . , we relax the link model assumptions we made in this section , and evaluate our algorithms on real devices and using real links . ] in the system model in fig .
[ fig : systemmodel](a ) , each mobile device @xmath38 is connected to the internet via its cellular or wi - fi link . at slot @xmath11
, @xmath39 is the channel state vector of these links , where @xmath40 @xmath41 @xmath42 @xmath41 @xmath43 .
we assume that @xmath44 is the state of the link between the source and mobile device @xmath5 , and it takes `` on '' and `` off '' values depending on the state of the channel . without loss of generality ,
if mobile device @xmath5 does not have internet connection , then @xmath44 is always at `` off '' state , which means there is no cellular or wi - fi connection . since we consider that mobile devices are in close proximity and transmission range , they form a fully connected clique topology . at slot @xmath11
, @xmath45 is the channel state vector of the d2d links , where @xmath45 @xmath46 @xmath47 @xmath41 @xmath48 @xmath41 @xmath49 .
we assume that @xmath50 is the state of the d2d link between node @xmath5 and @xmath15 .
we consider protocol model in our formulations @xcite , where each mobile device can either transmit or receive at the same time at the same frequency . assuming that @xmath51 is the channel state vector of the system including both the links between the source and mobile devices as well as among mobile devices
, @xmath52 denotes the set of the link transmission rates feasible at time slot @xmath11 depending on our protocol model . in particular , for cellular + wi - fi direct setup , @xmath52 is the set that allows more links to be operated at the same time , while for the wi - fi + wi - fi direct setup , @xmath52 is more limited set to take into account the interference among the links .
in this section , we characterize the stability region of the energy - aware cooperative computation problem , and formulate network utility maximization ( num ) framework .
the solution of the num framework provides us insights for developing the stochastic control algorithms in the next section .. by abuse of notation , we use a variable , _
e.g. , _ @xmath53 as the average value of @xmath54 in our num formulation if both @xmath53 and @xmath54 refers to the same parameter . ]
we provide the stability region of the cooperative computation system for both cellular + wi - fi direct and w - fi + wi - fi direct setups .
first , the flow conservation constraint at the source should be @xmath55 to stabilize the system .
this constraint requires that the total outgoing rate from the source , _
@xmath56 should be larger than the generated rate @xmath57 .
furthermore , the following flow conservation constraints inside a mobile device should be satisfied for stability ; @xmath58 , @xmath59 , and @xmath60 .
these constraints are necessary for the stability of queues @xmath26 , @xmath27 , and @xmath28 , respectively .
finally , the transmission rates over the links should be feasible , _
i.e. , _ @xmath61 .
thus , we define the stability region as @xmath62 @xmath63 @xmath57 , @xmath64 , @xmath65 , @xmath66 , @xmath67 @xmath68 , @xmath69 , @xmath55 , @xmath58 , @xmath59 , @xmath60 , @xmath70 . now , we characterize our num problem .
@xmath71 the objective of the num problem in ( [ opt : eq1 ] ) is to determine @xmath72 for @xmath69 which maximize the total utility @xmath73 .
lagrangian relaxation of the flow conservation constraints that characterize the stability region @xmath74 gives the following lagrange function : @xmath75 where @xmath76 , @xmath77 , @xmath78 , and @xmath79 are the lagrange multipliers .
note that we will convert these lagrange multipliers to queues @xmath80 , @xmath26 , @xmath27 , and @xmath28 when we design our stochastic algorithm in the next section .
the lagrange function in ( [ relax : eq1 ] ) is decomposed into sub - problems such as flow , computation , and energy controls as well as cooperation and scheduling .
the solutions of ( [ relax : eq1 ] ) for @xmath57 , @xmath65 , @xmath66 , @xmath64 , and @xmath67 are expressed as : @xmath81 \nonumber \\ & \mbox{s.t .
} \mbox { } \ { x_{n , k } , h_{n , k } \}_{\forall
n \in { \mathcal{n } } , k \in { \mathcal{n } } } \in \gamma_{\boldsymbol c}\end{aligned}\ ] ] next , we design a stochastic algorithm ; energy - aware cooperative computation inspired by the num solutions in ( [ eq : decsol1 ] ) , ( [ eq : decsol2 ] ) , ( [ eq : decsol3 ] ) , ( [ eq : decsol4 ] ) .
now , we provide our energy - aware cooperative computation algorithm which includes _ flow control _ , _ computation control _ , _ energy control _ , and _ cooperation & scheduling_. * _ flow control : _ at every time slot @xmath11 , @xmath82 is determined by maximizing @xmath83 @xmath84 $ ] subject to @xmath85 , where @xmath86 is a positive constant larger than the transmission rate from the source , and @xmath87 is a large positive constant .
note that @xmath88 is the queue size at the source of flow and stores packets that are supposed to be transmitted to mobile device @xmath5 .
after @xmath82 is determined , @xmath82 packets are inserted in queue @xmath18 ( as illustrated in fig .
[ fig : systemmodel](a ) ) .
* _ computation control : _ at every time slot @xmath11 , the computation control algorithm at device @xmath5 determines @xmath29 by optimizing @xmath89 \nonumber \\ \mbox{s.t . } & \sum_{k \in { \mathcal{n } } } d_{n , k}(t ) \leq d_{n}^{\max}\end{aligned}\ ] ] where @xmath90 is a positive constant larger than the processing rate of the computation block in device @xmath5 dedicated to help device @xmath15 .
the interpretation of ( [ eq : decoder ] ) is that at every time slot @xmath11 , @xmath91 packets are passed to the computation block ( in fig .
[ fig : systemmodel](b ) ) if @xmath92 , where @xmath93 is the mobile device that maximizes ( [ eq : decoder ] ) .
otherwise , no packets are sent to the computation block .
the packets that are being processed by the computation block are passed to @xmath34 .
note that some computation blocks may require to receive a group of packets to be able to process them .
in that case , @xmath90 is arranged accordingly ( _ i.e. , _ it can be increased to transfer a group of packets ) .
* _ energy control : _ at every time slot @xmath11 , the energy control algorithm at device @xmath5 determines @xmath94 by optimizing @xmath95 \nonumber \\ \mbox{s.t . } & \sum_{k \in { \mathcal{n } } } e_{n , k}(t ) \leq e_{n , k}^{\max}\end{aligned}\ ] ] where @xmath96 is a positive constant larger than the energy capacity of device @xmath5 dedicated to help device @xmath15 .
the interpretation of ( [ eq : energy ] ) is that at every time slot @xmath11 , @xmath97 packets are passed to the energy filter ( as illustrated in fig .
[ fig : systemmodel](b ) ) if @xmath98 , where @xmath93 is the mobile device that maximizes ( [ eq : energy ] ) .
otherwise , no packets are sent to the energy filter .
the packets passing through the energy filter are inserted in @xmath35 . * _ scheduling & cooperation : _ at every time slot @xmath11 ,
the scheduling and cooperation algorithm determines transmission rates over links , _
@xmath14 and @xmath24 by maximizing @xmath99 \nonumber \\ \mbox{s.t . } & { \boldsymbol x , \boldsymbol h } \in \gamma_{\boldsymbol c(t)}\end{aligned}\ ] ] for cellular + wi - fi direct system , ( [ eq : schedulingcoop ] ) is decomposed into two terms : maximizing @xmath100 @xmath101 @xmath102 @xmath103 and @xmath100 @xmath104 , because cellular and wi - fi direct transmissions operate simultaneously and transmission over one link does not affect the other . on the other hand , for wi - fi + wi - fi direct setup , the joint optimization in ( [ eq : schedulingcoop ] ) should be solved . +
note that transmissions over all links are unicast transmissions in our work , where unicast is dominantly used in practice over cellular , wi - fi , and wi - fi direct links . also , it is straightforward to extend our framework for broadcast transmissions .
[ eec_theorem1 ] if channel states are i.i.d . over time slots , and the arrival rates
@xmath105 = a_n , \forall n \in { \mathcal{n}}$ ] are interior of the stability region @xmath74 , then energy - aware cooperative computation stabilizes the network and the total average queue sizes are bounded .
furthermore , if the channel states are i.i.d . over time slots , and the traffic arrival rates are controlled by the flow control algorithm of energy - aware cooperative computation , then the admitted flow rates converge to the utility optimal operating point with increasing @xmath87 . _
proof : _ the proof is provided in appendix b. @xmath106 our energy - aware cooperative computation framework has several advantages : ( i ) distributed , ( ii ) takes into account scarce resources such as processing power and energy in addition to bandwidth to make control decisions , and ( iii ) utilizes available resources ; processing power , energy , and bandwidth in a utility optimal manner .
theorem [ eec_theorem1 ] shows the theoretical performance guarantees of our framework , while we focus on its performance in a practical setup in the next section .
in this section , we evaluate our _ energy - aware cooperative computation _ ( eacc ) scheme using a tested that consists of android based smartphones and tablets .
the evaluation results show that our scheme significantly improves throughput as compared to ( i ) no - cooperation , where each device receives its content from the source without cooperating with other devices , and ( ii ) cooperation , where multiple mobile devices cooperate , but the cooperating devices do not do computation and energy control for other devices ( mobile devices just receive packets from the source , and relay them to other mobile devices without processing and energy control ) .
next , we present testbed setup and results in detail .
[ fig : rate_vs_time_results ] we implemented a testbed of the setup shown in fig .
[ fig : systemmodel](a ) using real mobile devices , specifically android 5.1.1 based nexus 5 smartphones and nexus 7 tablets .
we classify devices as ( i ) a source device , which acts as the source in fig .
[ fig : systemmodel](a ) , ( ii ) helper devices , which receive data from the source , process it , and transmit to other devices ( receivers ) to help them , and ( iii ) receiver devices , which receive data from both the source device and the helpers .
a receiver device processes data arriving from the source , but it does not process the data arriving from helpers as the helpers send already processed data .
note that a device could be both receiver and a helper device depending on the configuration .
_ integration to the protocol stack : _ we implemented our energy - aware cooperative computation ( eacc ) framework as a slim layer between transport and application layers .
in other words , we implemented our framework on top of tcp .
this kind of implementation is advantageous , because ( i ) mobile devices do not need rooting , and ( ii ) our framework and codes could be easily transferred to mobile devices using other operating systems such as ios .
_ source configuration and eacc implementation : _ we implemented the source node in fig .
[ fig : systemmodel ] using a nexus 5 smartphone .
basically , multiple files ; file@xmath107 , file@xmath108 requested by devices @xmath5 and @xmath15 are read by using the public java class _ bufferedinputstream _ according to the flow control algorithm described in section [ sec : coopcomp ] and shown in fig .
[ fig : systemmodel](b ) .
the bytestream is packetized by setting each packet to @xmath109 , and packets are inserted into source buffers ; @xmath18 , @xmath110 .
we set the flow control parameters as ; @xmath111 , @xmath112 , and slot duration is @xmath113 .
we used @xmath114 function as our utility function . in this setup , reading files , converting bytestream into packets , and inserting packets into the input queues are done by multiple threads , _
i.e. , _ a thread runs for each file ; file@xmath107 in fig .
[ fig : systemmodel](b ) .
the other set of threads at the source device make packet transmission decisions from the source device to receiver and helper devices .
in particular , the source node collects @xmath37 information from all mobile devices . at each time slot , the source node checks @xmath115 , and if @xmath116 , then @xmath117 packets are transmitted from @xmath110 to the tcp socket at the source device for transmission to mobile device @xmath5 .
_ eacc operation on mobile devices : _ all mobile devices ( including helper or receiver+helper devices ) implement all the building blocks illustrated fig .
[ fig : systemmodel](c ) .
multiple threads are used to make these blocks operating simultaneously . the first thread at mobile device @xmath5 receives packets that are transmitted by the source node , and inserts these packets in @xmath26 .
the second thread has two tasks .
first , it transfers packets from @xmath26 to @xmath27 according to the computation control algorithm in ( [ eq : decoder ] ) , where @xmath118 packets and the slot duration is @xmath113 .
we set @xmath119 in our experiments as our applications do not change the rate as explained later in this section .
the second task of this thread to actually do the computation tasks related to the application . in our experiments , the computation block counts the bytes in the packets . in particular , similar to the pilot study in the introduction , @xmath2 , @xmath3 , and @xmath4 correspond to ( i ) no counting , ( ii ) counting every byte in a packet once , and ( iii ) counting every byte in a packet @xmath5 times , respectively .
the third thread transfers packets from @xmath27 to @xmath28 using the energy control algorithm in ( [ eq : energy ] ) , where we set @xmath120 depending on the battery level of the device .
for example , if the battery level is below some threshold @xmath120 is limited .
we evaluated different configurations in our experiments as we explain layer .
the slot duration is again set to @xmath113 .
the final thread transfers packets from @xmath28 to application layer if @xmath36 , or transmits to node @xmath15 if @xmath121 . in the second case , _
i.e. , _ if @xmath121 , the number of packets in tcp socket is checked at every time slot , where the time slot duration is @xmath113 .
if it is below a threshold of @xmath122 packets , then @xmath117 packets are removed from @xmath28 and inserted to the tcp socket to be transmitted to node @xmath15 .
when node @xmath5 receives packets from node @xmath15 , it directly passes the packets to the application layer as illustrated in fig .
[ fig : systemmodel](c ) , because these packets are the ones that are already processed by node @xmath15 .
if node @xmath5 is both a helper and a receiver device , it runs all the threads explained above in addition to the receiving thread from node @xmath15 ( illustrated in fig .
[ fig : systemmodel](c ) ) .
_ information exchange : _ our implementation is lightweight in the sense that it limits control information exchange among mobile devices .
the only control information that is transmitted in the system is @xmath26 from each mobile device to the source node .
each mobile device @xmath5 collects @xmath26 , @xmath123 , and transmits this information to the source node periodically , where we set the periods @xmath124 .
_ connections : _ all the devices in the system including the source device , helpers , receivers , and helper+receiver devices are connected to each other using wi - fi direct connections in our testbed .
the source node is configured as the group owner of the wi - fi direct group .
we note that cooperation in this setup does not bring any benefit in terms of bandwidth utilization as all the links use the same transmission channel in a wi - fi direct group . however , as we demonstrate later in this section , it brings benefit due to cooperative processing power and energy utilization , which is our main focus in this paper . therefore , this setup ( where all the devices are connected to each other using wi - fi direct links ) well suits to our evaluation purposes . _
test environment : _ we conducted our experiments using our testbed in a lab environment where several other wi - fi networks were operating in the background .
we located all the devices in close proximity of each other , and we have evaluated eacc for varying levels of computational complexity , number of receivers , and number of helpers .
next , we present our evaluation results . [
fig : rate_vs_helpers_one_receiver_results ] we first consider a setup as shown in fig .
[ fig : rate_vs_time_results](a ) which consists of a source device , one receiver ( @xmath125 ) , and one helper ( @xmath126 ) . fig .
[ fig : rate_vs_time_results](b ) shows the average rate versus time graph for the setup shown in fig . [
fig : rate_vs_time_results](a ) when all three devices are android based nexus 7 tablets .
the average rate is calculated as the average over 10 trials ( with different seeds ) .
the computation under consideration in this experiment is @xmath127 , which basically counts the number of bytes in a packet @xmath5 times , @xmath5 is the packet size . as can be seen ,
if there is no cooperation , the rate measured at @xmath125 is on the order of 1.5mbps . on the other hand , eacc increases the rate to almost 3mbps .
this means that helper device @xmath126 helps the receiver device @xmath125 to process the packets in eacc . in this setup
, eacc doubles the rate as compared to no - cooperation , which is a significant improvement . for the same setup in fig .
[ fig : rate_vs_time_results](a ) , we also evaluate the effect of energy control part of eacc on the average rate performance .
in particular , fig .
[ fig : rate_vs_time_results](c ) shows the average rate versus battery level at the receiver device @xmath125 . in these results
, we used android based nexus 5 smartphones .
the average rate is calculated as the average over 10 trials ( with different seeds ) .
the computation under consideration in this experiment is @xmath127 , which basically counts the number of bytes in a packet for each byte in the packet .
we consider that if the battery level of a device reduces below 40% threshold , then energy credits are not generated for the processing of the received packets .
this makes @xmath27 large over time , and after some point no packets are transmitted to that device for the processing task . in fig .
[ fig : rate_vs_time_results](c ) , when the battery level of @xmath125 reduces below 40% , then it stops receiving packets for processing .
if there is no cooperation , then the rate towards @xmath125 reduces to 0 . on the other hand , with eacc ,
the rate is still higher than 0 thanks to having helper .
the helper device with larger energy level ( for the sake of this experiment ) , receives packets from the source , process them , and forwards them to @xmath125 , which receives already processed data .
after 40% threshold , both eacc and no - cooperation improve , because @xmath125 starts processing packets .
this result shows the importance of energy - awareness in our cooperative computation setup .
now , we consider the effect of the number of helpers to overall rate performance . in particular , we develop a setup shown in fig .
[ fig : rate_vs_helpers_one_receiver_results](a ) , where there is one source , one receiver , and varying number of helpers . in this setup , the source device , receiver , and the first two helper devices
are nexus 5 smartphones , while the other helpers are nexus 7 tablets .
[ fig : rate_vs_helpers_one_receiver_results](b ) shows the average rate ( averaged over 10 seeds ) when eacc is employed versus the number of helpers for different computational complexities such as @xmath128 , @xmath129 , and @xmath127 , where the processing task is counting the number of bytes in a packet .
as expected , when complexity increases , the rate decreases .
more interestingly , the increasing number of helpers increases the rates of all complexity levels .
there are two reasons for this behavior .
first , even if complexity level is low , _ e.g. , _ @xmath128 ,
processing power is still a bottleneck , and it can be solved by increasing the number of helpers . note that after the number of helpers exceeds a value , the achievable rates saturate , which means that processing power is not a bottleneck anymore , but bandwidth is .
the second reason is that receiving data over multiple interfaces increases diversity . in other words ,
when the channel condition over one interface ( _ e.g. , _ between source and the mobile device ) degrades , the other interface ( _ e.g. , _ between two mobile devices ) can still have a better channel condition . in order to understand the real impact of processing power in a cooperative system , we tested both eacc and cooperation ( without computation and energy control ) in the setup shown in fig .
[ fig : rate_vs_helpers_one_receiver_results](a ) .
the results are provided in fig .
[ fig : rate_vs_helpers_one_receiver_results](c ) when the complexity is @xmath127 . as can be seen , while eacc significantly increases the rate with increasing number of helpers , cooperation slightly increases the rate ( due to diversity ) .
the improvement of eacc over cooperation is as high as 83% , which is significant .
[ fig : rate_vs_helpers_two_receiver_results ] finally , we consider a scenario that there are multiple receivers interested in different files .
[ fig : rate_vs_helpers_two_receiver_results](a ) shows the system model with one source , two receivers , and multiple helpers . in this setup ,
the source , two receivers , and the first helper is android based nexus 5 smartphone , while the rest of the helpers are nexus 7 tablets .
[ fig : rate_vs_helpers_two_receiver_results](b ) and ( c ) show the average rate ( averaged over 10 seeds ) measured at @xmath125 and @xmath130 when eacc is employed with respect to the increasing number of helpers , respectively .
similar to previous setups , @xmath128 , @xmath129 , and @xmath127 correspond to different computational complexities , where the processing task is counting the number of bytes in a packet . as can be seen , the measured rate at both @xmath125 and @xmath130 increases with increasing number of helpers .
this shows that our eacc algorithm successfully accommodates multiple flows and receivers .
we considered that a group of cooperative mobile devices , within proximity of each other , ( i ) use their cellular or wi - fi ( 802.11 ) links as their primary networking interfaces , and ( ii ) exploit their d2d connections ( wi - fi direct ) for cooperative computation .
we showed that if mobile devices cooperate to utilize their aggregate processing power , it significantly improves transmission rates .
thus , for this scenario , we developed an _ energy - aware cooperative computation _ framework to effectively utilize processing power and energy .
this framework provides a set of algorithms including flow , computation and energy controls as well as cooperation and scheduling .
we implemented these algorithms in a testbed which consists of real mobile devices .
the experiments in the testbed show that our _ energy - aware cooperative computation _ framework brings significant performance benefits .
`` cisco visual networking index : global mobile data traffic forecast update , '' 2014 - 2019 . `` ericsson mobility report , '' february 2015 .
`` wi - fi direct , '' http://www.wi-fi.org/discover-and-learn/wi-fi-direct .
`` android developer , '' http://developer.android.com/develop/index.html .
`` nexus tech specs , '' https://support.google.com/nexus/answer/6102470?hl=en .
t. rappaport , s. sun , r. mayzus , h. zhao , y. azar , k. wang , g. wong , j. schulz , m. samimi , and f. gutierrez , `` millimeter wave mobile communications for 5 g cellular : it will work ! '' _
access , ieee _ , vol .
1 , pp . 335349 , 2013 .
j. ostermann , j. bormans , p. list , d. marpe , m. narroschke , f. pereira , t. stockhammer , and t. wedi , `` video coding with h.264/avc : tools , performance , and complexity , '' _ circuits and systems magazine , ieee _ , vol .
1 , pp . 728 , first 2004 . m. horowitz , a. joch , f. kossentini , and a. hallapuro , `` h.264/avc baseline profile decoder complexity analysis , '' _ circuits and systems for video technology , ieee transactions on _ , vol .
13 , no . 7 , pp . 704716 , july 2003 .
j. k. sundararajan , d. shah , m. jakubczak , m. mitzenmacher , and j. barros , `` network coding meets tcp : theory and implementation , '' _ proceedings of the ieee _ , pp .
490512 , march 2011 .
p. vingelmann , p. zanaty , f. fitzek , and h. charaf , `` implementation of random linear network coding on opengl - enabled graphics cards , '' _ in wireless conference _ , 2009 .
european , may 2009 , pp .
118 123 .
h. shojania , b. li , and x. wang , `` nuclei : gpu - accelerated many - core network coding , '' _ in infocom 2009 , ieee _ , april 2009 , pp . 459467 .
s. arora and b. barak , _ computational complexity : a modern approach_. cambridge university press , 2009 .
m. chiang , s. t. low , a. r. calderbank , j. c. doyle , `` layering as optimization decomposition : a mathematical theory of network architectures , '' _ in proceedings of the ieee _
95(1 ) , jan . 2007 .
d. syrivelis , g. iosifidis , d. delimpasis , k. chounos , t. korakis , and l. tassiulas , `` bits and coins : supporting collaborative consumption of mobile internet , '' _ in proc .
ieee infocom _ , hong kong , april 2015 .
a. asadi , q. wang , and v. mancuso , `` a survey on device - to - device communication in cellular networks , '' april 2014 , technical report - arxiv:1310.0720v6[cs.gt ] .
p. hui , j. crowcroft , e. yoneki , `` bubble rap : social - based forwarding in delay tolerant networks , '' _ in proc . of acm mobihoc _ ,
hong kong , may 2008 . c. boldrini , m. conti , a. passarella , `` exploiting users social relations to forward data in opportunistic networks : the hibop solution , '' _ in proc . of pervasive and mobile computing _ , oct .
l. keller , a. le , b. cici , h. seferoglu , c. fragouli , a. markopoulou , `` microcast : cooperative video streaming on smartphones '' _ in proc . of acm mobisys _ ,
low wood bay , lake district , uk , june 2012 .
h. seferoglu , l. keller , b. cici , a. le , a. markopoulou,``cooperative video streaming on smartphones '' _ in proc . of allerton _ , 2011 .
m. ramadan , l. el zein , z. dawy , `` implementation and evaluation of cooperative video streaming for mobile devices , '' _ in proc . of ieee pimrc _ , cannes , france , sept . 2008 .
s. li and s. chan , `` bopper : wireless video broadcasting with peer - to - peer error recovery , '' _ in proc . of ieee icme _ ,
beijing , china , july 2007 .
j. chesterfield , r. chakravorty , i. pratt , s. banerjee , and p. rodriguez , `` exploiting diversity to enhance multimedia streaming over cellular links , '' _ in proc . of ieee infocom _ , march 2005 .
j. chesterfield , r. chakravorty , i. pratt , s. banerjee , and p. rodriguez , `` a system for peer - to - peer video streaming in resource constrained mobile environments , '' _ in proc . of acm
u - net _ , december 2009 .
s. ioannidis , a. chaintreau , l. massoulie , `` optimal and scalable distribution of content updates over a mobile social network , '' _ in proc . of infocom _ ,
rio de janeiro , brazil , apr .
b. han , p. hui , v. a. kumar , m. v. marathe , g. pei , a. srinivasan , `` cellular traffic offloading through opportunistic communications : a case study , '' _ in proc . of acm workshop on challenged networks ( chants )
_ , chicago , il , sept . 2010 .
r. k. lomotey and r. deters , `` architectural designs from mobile cloud computing to ubiquitous cloud computing - survey , '' _ in proc .
ieee services _ , anchorage , alaska , june 2014 .
t. penner , a. johnson , b. v. slyke , m. guirguis , and q. gu , `` transient clouds : assignment and collaborative execution of tasks on mobile devices , '' _ , in proc .
ieee globecom _ ,
austin , tx , december 2014 .
m. satyanarayanan , s. smaldone , b. gilbert , j. harkes , and l. iftode , `` bringing the cloud down to earth : transient pcs everywhere , '' _ in mobicase10 _ , 2010 , pp . 315322 .
e. miluzzo , r. caceres , and y. chen , `` vision : mclouds - computing on clouds of mobile devices , '' _ in acm workshop on mobile cloud computing and services _ , low wodd bay , lake district , uk , june 2012 .
l. tassiulas , a. ephremides , `` stability properties of constrained queueing systems and scheduling policies for maximum throughput in mul- tihop radio networks , '' _ in ieee trans . on automatic control _ , vol .
37(12 ) , dec . 1992 .
l. tassiulas and a. ephremides , `` dynamic server allocation to parallel queues with randomly varying connectivity , '' _ in ieee trans . on information theory _ ,
39(2 ) , march 1993 .
m. j. neely , e. modiano , and c. li , `` fairness and optimal stochastic control for heterogeneous networks , '' _ in ieee / acm trans . on networking _ ,
16(2 ) , april 2008 . p. gupta , p. r. kumar , `` the capacity of wireless networks , '' _ in ieee trans .
on information theory _ , vol .
34(5 ) , 2000 .
[ fig : energy_bottleneck ] in this section , we consider how energy can create a bottleneck .
we consider the same setup as in fig .
[ fig : complexityvsrate_new](a ) , where data is transmitted from mobile device @xmath1 to another mobile device @xmath0 . in this setup
, @xmath1 transmits data after processing packets with computational complexities @xmath2 , @xmath3 , and @xmath4 , where @xmath5 is the packet size , which is set to @xmath109 .
we have two scenarios ; `` energy constraint '' and `` no energy constraint '' . in the energy constrained scenario ,
the assumption is that the battery level of @xmath1 is low , so it limits its transmission rate to @xmath131 . on the other hand , in the unconstrained scenario , @xmath1 can transmit at the maximum possible transmission rates , which are reported in fig . [
fig : complexityvsrate_new](b ) . fig .
[ fig : energy_bottleneck ] shows energy consumption of the two scenarios over time for different computational complexities .
as seen , the energy consumption of the energy constrained version is much lower in all computational complexity levels , and the gap between the two scenarios increases when the computational complexity increases . from these results
, we can conclude that devices can limit their transmission rates when their battery levels are limited .
this will reduce energy consumption , but for the cost of limited transmission rate .
thus , energy levels of devices can become bottlenecks , even if bandwidth is not a bottleneck .
_ stability : _ let @xmath132 , where @xmath133 , @xmath134 , @xmath135 , and @xmath136 .
let the lyapunov function be ; @xmath137 the lyapunov drift is ; @xmath138\end{aligned}\ ] ] which is expressed as ; @xmath139 @xmath140\end{aligned}\ ] ] considering the fact that @xmath141+a)^{2 } \leq q^{2 } + a^{2 } + b^{2 } + 2q(a - b)$ ] , ( [ eq : appa_3 ] ) is expressed as ; @xmath142\end{aligned}\ ] ] which is expressed as @xmath143\end{aligned}\ ] ] there always exists a finite positive constant @xmath144 satisfying @xmath145\end{aligned}\ ] ] because the maximum values of @xmath146 , @xmath14 , @xmath29 , @xmath24 , @xmath94 , and @xmath31 terms are bounded by finite positive constants by our eacc algorithm .
the minimization of the right hand side of the drift inequality in ( [ eq : appa_6 ] ) corresponds to the decoder control in ( [ eq : decoder ] ) , energy control in ( [ eq : energy ] ) , and scheduling & cooperation in ( [ eq : schedulingcoop ] ) .
if the arrival rates satisfy @xmath148 = a_{n}$ ] and @xmath149 is inside the stability region @xmath74 , then there exists a randomized policy with solution ; @xmath150 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 , satisfying @xmath155 \leq -\delta_1 , \forall n \in { \mathcal{n}}\nonumber \\ & -e[{{\mathop{d}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{x}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) ] \leq -\delta_2 , \forall n \in { \mathcal{n } } , k \in { \mathcal{n}}\nonumber \\ & -e[{{\mathop{e}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{d}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t){\alpha}_{n , k}(t ) ] \leq -\delta_3 , \forall n \in { \mathcal{n } } , k \in { \mathcal{n}}\nonumber \\ & -e[{{\mathop{h}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{e}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) ] \leq -\delta_4 , \forall n \in { \mathcal{n } } , k \in { \mathcal{n}}\end{aligned}\ ] ] where @xmath156 are positive small constants .
since our eacc algorithm minimizes the right hand side of ( [ eq : appa_6 ] ) , the following inequalities satisfy : ( i ) @xmath157 \leq -e[\sum_{k \in { \mathcal{n } } } { { \mathop{x}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{k , n}(t ) - { { \mathop{x}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_n(t ) ] \leq -\delta_1 $ ] , ( ii ) @xmath158 \leq -e[{{\mathop{d}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{x}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) ] \leq -\delta_2 $ ] , ( iii ) @xmath159
\leq -e[{{\mathop{e}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{d}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t){\alpha}_{n , k}(t ) ] \leq -\delta_3 $ ] , and ( iv ) @xmath160 \leq -e[{{\mathop{h}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) - { { \mathop{e}\limits^{\vbox to -.5\ex@{\kern-\tw@\ex@ \hbox{\scriptsize * } \vss}}}}_{n , k}(t ) ] \leq -\delta_4 $ ] .
thus , the following inequality satisfy @xmath161 _ optimality : _ let us define a drift - plus - penalty function as @xmath166 $ ] which is , considering the bound in ( [ eq : appa_6 ] ) , expressed as @xmath167 \leq b - 2e\biggl [ \sum_{n \in { \mathcal{n } }
} \nonumber \\ & \biggl ( s_n(t ) \bigl ( \sum_{k \in { \mathcal{n } } } x_{k , n}(t ) - x_n(t ) \bigr ) \biggr ) + \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \biggl ( u_{n , k}(t ) \bigl ( d_{n , k}(t ) \nonumber \\ & - x_{n , k}(t ) \bigr ) \biggr ) + \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \biggl ( q_{n , k}(t ) \bigl ( e_{n , k}(t ) - d_{n , k}(t ) \alpha_{n , k}(t ) \bigr ) \biggr ) \nonumber \\ & + \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \biggl ( z_{n , k}(t ) \bigl ( h_{n , k}(t ) - e_{n , k}(t ) \bigr ) \biggr ) | \boldsymbol h(t)\biggr ] - \sum_{n \in { \mathcal{n } } } m \nonumber \\ & e\biggl [ g_{n}({x}_{n}(t ) ) | \boldsymbol h(t ) \biggr ] \end{aligned}\ ] ] note that the minimization of the right hand side of the drift inequality in ( [ eq : appa_11 ] ) corresponds the flow control part of eacc in section [ sec : coopcomp ] as well as the decoder control in ( [ eq : decoder ] ) , energy control in ( [ eq : energy ] ) , and scheduling & cooperation in ( [ eq : schedulingcoop ] ) .
since there exists a randomized policy as discussed in the stability part above , the right hand side of ( [ eq : appa_11 ] ) is bounded as @xmath167 \leq b - 2 \sum_{n \in { \mathcal{n } } } s_n(t ) \nonumber \\ & \delta - 2 \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \bigl ( u_{n , k}(t ) + q_{n , k}(t ) + z_{n , k}(t ) \bigr ) \delta - \sum_{n \in { \mathcal{n } } } m \nonumber \\ & e[g_{n}(a_{n}+\delta)]\end{aligned}\ ] ] where @xmath168 is the maximum time average of the sum utility function that can be achieved by any control policy that stabilizes the system .
we can rewrite ( [ eq : appa_11 ] ) as @xmath169 \biggr ] \leq \limsup_{t \rightarrow \infty } \nonumber \\ & \frac{1}{t } \sum_{\tau = 0}^{t-1 } \biggl [ b - 2 \sum_{n \in { \mathcal{n } } } s_n(t ) \delta - 2 \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \bigl ( u_{n , k}(t ) + q_{n , k}(t ) + \nonumber \\ & z_{n , k}(t ) \bigr ) \delta - \sum_{n \in { \mathcal{n } } } m u_{n}(a_{n } + \delta ) \biggr]\end{aligned}\ ] ] let us first consider the stability of the queues .
if both sides of ( [ eq : appa_12 ] ) are divided by @xmath170 and the terms are arranged , we have @xmath171 \biggr ] - \nonumber \\ & \sum_{n \in { \mathcal{n } } } \frac{mg_{n}(a_{n}+\delta)}{\delta}. \end{aligned}\ ] ] this concludes that eacc stabilizes the queues in the system when the flow control algorithm in section [ sec : coopcomp ] is employed .
next , we consider the optimality of eacc . if both sides of ( [ eq : appa_12 ] )
are divided by @xmath87 , we have @xmath172 \biggr ] \leq \limsup_{t \rightarrow \infty } \frac{1}{t } \sum_{\tau = 0}^{t-1 } \biggl [ \frac{b}{m } - \nonumber \\ & 2 \sum_{n \in { \mathcal{n } } } s_n(t ) \frac{\delta}{m } - 2 \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \bigl ( u_{n , k}(t ) + q_{n , k}(t ) + z_{n , k}(t ) \bigr ) \nonumber \\ & \frac{\delta}{m } - \sum_{n \in { \mathcal{n } } } g_{n}(a_{n } + \delta ) \biggr ] \end{aligned}\ ] ] which is expressed as @xmath173 \biggr ] \geq
\limsup_{t \rightarrow \infty } \frac{1}{t } \sum_{\tau = 0}^{t-1 } \biggl [ \sum_{n \in { \mathcal{n } } } \nonumber \\ & g_{n}(a_{n } + \delta ) - \frac{b}{m } + 2 \sum_{n \in { \mathcal{n } } } s_n(t ) \frac{\delta}{m } + 2 \sum_{n \in { \mathcal{n } } } \sum_{k \in { \mathcal{n } } } \bigl ( u_{n , k}(t ) \nonumber \\ & + q_{n , k}(t ) + z_{n , k}(t ) \bigr ) \frac{\delta}{m } \biggr ] \end{aligned}\ ] ] since @xmath174 \geq 0 $ ] , the following inequality holds @xmath175 \biggr ] \geq
\sum_{n \in { \mathcal{n } } } g_{n}(a_{n } + \delta ) - \frac{b}{m}. \end{aligned}\ ] ] this proves that the flow rates achieved by eacc converge to the utility optimal operating point with increasing @xmath87 .
this concludes the optimality part of the proof .
|
new data intensive applications , which are continuously emerging in daily routines of mobile devices , significantly increase the demand for data , and pose a challenge for current wireless networks due to scarce resources .
although bandwidth is traditionally considered as the primary scarce resource in wireless networks , the developments in communication theory shifts the focus from bandwidth to other scarce resources including processing power and energy .
especially , in device - to - device networks , where data rates are increasing rapidly , processing power and energy are becoming the primary bottlenecks of the network .
thus , it is crucial to develop new networking mechanisms by taking into account the processing power and energy as bottlenecks . in this paper
, we develop an energy - aware cooperative computation framework for mobile devices . in this setup , a group of cooperative mobile devices , within proximity of each other , ( i ) use their cellular or wi - fi ( 802.11 ) links as their primary networking interfaces , and ( ii ) exploit their device - to - device connections ( _ e.g. , _ wi - fi direct ) to overcome processing power and energy bottlenecks .
we evaluate our energy - aware cooperative computation framework on a testbed consisting of smartphones and tablets , and we show that it brings significant performance benefits .
|
in this work we model an asteroid s apparent visual magnitude @xmath3 ( log flux ) as @xmath4,\ ] ] where @xmath5 is the absolute magnitude ( a constant ) , @xmath6 is a periodic variability term due to rotation ( _ e.g. _ , if the object is spinning and has some asymmetry in shape or albedo ) , @xmath7 and @xmath8 are the heliocentric and geocentric distances ( in au ) , and @xmath9 is the _ phase function _ , which varies with the solar phase angle @xmath10 ( the sun - asteroid - earth angle ) . when @xmath11 ( _ i.e. _ , at opposition ) , @xmath12 by definition , while in general @xmath13 for @xmath14 ( with @xmath15 decreasing as @xmath10 increases ) .
a key feature of our approach is the simultaneous fitting of both the phase function @xmath15 and the rotation term @xmath6 .
the detailed forms of @xmath15 and @xmath6 , as well as the algorithm underlying our fitting procedure , are motivated by a variety of prior work in this area , as described in the following sections .
building upon the work of @xcite , @xcite discuss the inversion of asteroid lightcurve data taken over several oppositions to obtain a 3d shape solution .
the form of @xmath6 ( cf .
equation [ 1 ] ) in this case consists of a large number of free parameters ( several tens to hundreds ) .
results from inversion agree well with those from stellar occultations , adaptive optics imaging , and in - situ spacecraft imagery @xcite .
knowledge of the detailed irregular shapes of asteroids improves our ability to constrain models of their internal structure , as well the magnitude and timescale of spin and orbital evolution due to solar - radiation and thermal emission , including the yarkovsky and yorp effects ( see @xcite and references therein ) . a simpler model for @xmath6suitable for fitting to data sparser than that required for most inversion methods is a jacobi ellipsoid @xcite in its principal - axis spin state .
the lightcurve of such an ellipsoid is a double - peaked sinusoid , given by a simple expression depending solely ( assuming constant surface albedo ) on the axes ratio , and angle between the line of sight and spin axis .
the fitted amplitude thus yields a lower - bound elongation estimate for the asteroid .
the predicted distribution of the rotation frequencies of a collisionally - equilibrated system of particles has long been claimed to be a maxwellian function @xcite , which as reviewed by @xcite very well approximates the observed distribution of several hundred of the brightest ( @xmath040-km or larger ) asteroids , but breaks down for smaller objects , among which an excess of slow and fast rotators appear to exist .
@xcite more recently argue that collision instead leads to a lvy distribution , and that a significant primordial spin component remains in the present observed population . some studies that have examined
the spin distribution of small objects are @xcite , @xcite , the thousand asteroid lightcurve survey @xcite , and two brief observing runs conducted within the ptf survey ( @xcite ; @xcite ) .
@xcite describe the lightcurve database ( lcdb ) , which compiles several thousand densely - sampled lightcurves of asteroids targeted by dedicated observing teams .
lightcurves in the lcdb have the following features : 1 .
lcdb lightcurves dense sampling generally permits fitting of fourier series with many harmonic terms , 2 .
lcdb lightcurves are often sampled over the shortest time window necessary to measure the period , and therefore generally do not require large or uncertain corrections due to phase angle effects , 3 .
lcdb lightcurves fitted periods are assigned integer quality codes by a human reviewer ( from 1 = poor to 3 = confident ) .
all three of the above features are either impractical or infeasible when the set of lightcurves is very large and the data sparsely sampled , as is the case for ptf . in this work
we adopt the following modified approaches when fitting lightcurves : 1 .
we truncate the rotation curve s fourier - series fit after the 2nd harmonic , a simplification broadly justified by @xcite and the assumption of an ellipsoidal shape ( cf .
section 3.1.2 ) , 2 .
we simultaneously fit a phase - function model with the rotational part , 3 .
we use a machine - learned classifier to objectively aid in estimating the validity of each fitted period .
the classifier is trained using all fitted lightcurves that have previously ( and confidently ) measured lcdb periods and takes into account the accuracy with which the true period was retrieved along with 20 lightcurve metrics ( fitted period , amplitude , ratio of peaks , @xmath16 per degree of freedom of fit , number of data points , and more ) .
use of a machine classifier in asteroid lightcurve period quality assessment is entirely novel and inspired in part by work done by ptf collaborators in extragalactic transient science @xcite and variable star science ( @xcite ; @xcite ) , as well as waszczak et al .
( in prep ) s work on detection techniques for streaking neos . among the advantages of using a machine - classified quality score
is that , via cross - validation with the known - period sample , one estimates the completeness and contamination , _ i.e. _ , the true - positive and false - positive rates with respect to identifying an accurately - fit period , as a function of , _
e.g. _ , the period , amplitude , etc .
the resulting true- and false - positive rates may then be used to de - bias the classifier - filtered period distribution .
models ( cf . section 3.2 ) . ]
the analytic phase function of an ideal lambertian - scattering sphere fits well to featureless , atmospheric planets like venus , but quite poorly to airless bodies ( see figure 3.9 of @xcite for a comparison ) . in later sections we describe several @xmath15 models that have been derived for ( or empirically fit to ) asteroids .
qualitatively , asteroids show an approximately linearly decreasing @xmath15 out to @xmath17 deg , modified by a surge ( increase in slope ) at low phase angles ( @xmath18 deg ) , known as the _ opposition effect _ ( see figure 1 )
. early work ( _ e.g. _ @xcite and refs . therein ) on a small sample of well - observed asteroids , suggested that different asteroid spectral types display distinct behavior in @xmath15 .
figure 1 compares example phase curve data for d , c , s and e types ) visible reflectance spectra .
] , incorporating photometry from various sources .
we emphasize the fact that all of the data points in figure 1 have been corrected for rotational modulation ( the @xmath6 in equation [ 1 ] ) through dense sampling of each asteroid s lightcurve at each phase angle ( equivalently , each epoch ) . using a large corpus of low - precision photometry from the mpc ] , oszkiewicz et al .
( @xcite , @xcite ) showed that a fitted parameter of one particular @xmath15 model correlates well with an asteroid s sdss visible color . while they were unable to correct for rotational variation ( @xmath6-term in equation [ 1 ] ) , the oszkiewicz et al .
work nevertheless demonstrates a solid trend between @xmath15 and a compositional attribute ( color ) .
these prior works motivate several defining aspects of this work s phase - function analysis : 1 .
we fit multiple phase function models to each lightcurve , both for compatibility with the literature and to explore how the fitted parameters are related , 2 .
we simultaneously fit the rotational component with the phase - function part , 3 .
we introduce a single colorimetric index for quantifying c - type vs. s - type taxonomic classification , based on the compilation of several visible - band - color asteroid datasets ( see appendix ) , and examine the variation in phase - function parameters as a function of this color index .
the _ palomar transient factory ] _ ( ptf ) is a synoptic survey designed primarily to discover extragalactic transients ( @xcite ; @xcite ) .
the ptf camera , mounted on palomar observatory s 1.2-m oschin schmidt telescope , uses 11 ccds ( each 2k @xmath19 4k ) to image 7.3 deg@xmath20 of sky at a time at @xmath21/pixel resolution .
most exposures ( @xmath085% ) use a mould-@xmath22 filter filter is very similar to the sdss-@xmath7 filter ; see @xcite for its transmission curve . ]
( hereafter `` @xmath22 '' ) .
the remaining broadband images acquired use a gunn @xmath23-band filter .
nearly all broadband ptf images are 60-second integrations , regardless of filter . about 15% of nights ( near full moon )
are devoted to a narrowband ( h@xmath10 ) imaging survey of the full northern sky .
science operations began in march 2009 , with a nominal one- to five - day cadence for supernova discovery and typical twice - per - night imaging of fields .
median seeing is @xmath24 with a limiting magnitude @xmath25 ( for 5@xmath26 point - source detections ) , while dark conditions routinely yield @xmath27 @xcite .
the ptf survey is ongoing and expected to continue through mid-2016 . in january 2013
the ptf project formally entered a second phase called the _ intermediate ptf _
( ` iptf ' ; @xcite ) . in this paper
we simply use ` ptf ' to mean the entire survey , from 2009 through the present ( 2015 ) .
the iptf program accommodates more varied ` sub - surveys ' as opposed to a predominantly extragalactic program , including variable star and solar system science .
images are still acquired with the same telescope / camera / filters with 60s exposures , and are processed by the same reduction pipeline .
@xcite describe the ptf data reduction and archiving pipelines , hosted at the infrared processing and analysis center ( ipac ) at caltech .
processing at ipac includes bias and flat - field corrections , astrometric calibration against ucac3 @xcite , astrometric verification against 2mass @xcite , creation of source catalogs with source extractor @xcite , and production of reference images ( stacks of @xmath02030 ptf images that reach @xmath28 ) .
ofek et al .
( @xcite , @xcite ) describe the ptf survey s absolute photometric calibration method , which relies on source matching with sdss dr7 @xcite , and thus requires at least partial overlap of ptf with sdss each night .
a separate , _ relative _ photometric calibration ( based on lightcurves of non - variable field stars ) also exists for ptf data and is described by @xcite and in the appendix of @xcite . in this work
we utilize all @xmath22-band and @xmath23-band ptf data accumulated from the survey s start ( march 2009 ) through july 2014 .
the asteroid magnitudes reported in this work use relative photometric zeropoints when available ( which as of this writing applies to @xmath085% of ptf images ) and absolute photometric zeropoints otherwise .
the ptf s robotic survey program and processing pipeline , as well as our data aggregation and analysis in this work , make use of many functions from the matlab package for astronomy and astrophysics @xcite .
abazajian , k. n. et al . , 2009 . the seventh data release of the sloan digital sky survey .
http://dx.doi.org/10.1088/0067-0049/182/2/543 [ ] .
belskaya , i.n . ,
shevchenko , v.g . , 2000 .
opposition effect of asteroids .
http://dx.doi.org/10.1006/icar.2000.6410 [ ] .
bertin , e. , arnouts , s. , 1996 .
sextractor : software for source extraction .
http://dx.doi.org/10.1051/aas:1996164 [ ] .
bezdec , j.c . , 1981 .
pattern recognition with fuzzy objective function algorithms .
plenum press , new york .
bloom , j. , richards , j.w . ,
nugent , p.e . ,
quimby , r.m . ,
kasliwal , m.m . ,
starr , d.l . ,
poznanski , d. , ofek .
cenko , s.b . ,
butler , n.r . ,
kulkarni , s.r .
, gal - yam , a. , law , n. , 2012 . automating discovery and classification of transients and variable stars in the synoptic survey era .
http://dx.doi.org/10.1086/668468 [ ] .
blomme , j. , sarro , l.m . , odonovan ,
f.t . , et al . , 2011 . improved methodology for the automated classification of periodic variable stars .
bodewitts , d. , kelley , m. s. , li , j.y . ,
landsman , w. b. , besse , s. , ahearn , m. f. , 2011 .
collisional excavation of asteroid 596 scheila .
bottke , w.f .
et al . , 2013 .
the yarkovsky and yorp effects : implications for asteroid dynamics .
http://dx.doi.org/10.1146/annurev.earth.34.031405.125154 [ ] .
bowell , e. , chapman , c.r . ,
gradie , j.c . ,
morrison , d. , zellner , b. , 1978 .
taxonomy of asteroids .
http://dx.doi.org/10.1016/0019-1035(78)90085-4 [ ] .
bowell , e. , hapke , b. , domingue , d. , lumme , k. , peltoniemi , j. , harris , a. , 1989 .
application of photometric models to asteroids . in : http://adsabs.harvard.edu/abs/1989aste.conf..524b [ ] .
breiman , l. , friedman , j.h . , olshen , r.a . , stone , c.i . , 1984 .
classification and regression trees ( belmont , calif . : wadsworth ) .
breiman , l. , 2001 .
random forests .
http://dx.doi.org/10.1023/a:1010933404324 [ ] .
breiman , l. , cutler , a. , 2004 . random forests@xmath30 .
http://oz.berkeley.edu/~breiman/randomforests/ [ ] .
brett , d.r . , west , r.g . ,
wheatley , p.j .
the automated classification of astronomical light curves using kohonen self - organizing maps .
http://dx.doi.org/10.1111/j.1365-2966.2004.08093.x [ ] .
burns , j.a . , safronov , v.s . ,
asteroid nutation angles .
http://adsabs.harvard.edu/abs/1973mnras.165..403b [ ] .
bus , s.j . ,
binzel , r.p . ,
phase ii of the small main - belt asteroid spectroscopic survey : the observations . in : http://dx.doi.org/10.1006/icar.2002.6857
[ ] . bus , s.j . ,
binzel , r.p .
, 2003 . small main - belt asteroid spectroscopic survey , phase ii .
ear - a - i0028 - 4-sbn0001/smassii - v1.0 http://sbn.psi.edu/pds/resource/smass2.html [ ] .
bus , s.j . ,
vilas , f. , barucci , m. , 2002 .
visible - wavelength spectroscopy of asteroids . in : http://adsabs.harvard.edu/abs/2002aste.conf..169b [ ] .
carbognani , a. , 2010 . statistical analysis of c and s main - belt asteroids .
http://dx.doi.org/10.1016/j.icarus.2009.08.001 [ ] .
carry , b. , 2012 .
density of asteroids .
http://dx.doi.org/10.1016/j.pss.2012.03.009 [ ] .
cellino , a. , hestroffer , d. , tanga , p. , mottola , s. , delloro , a. , 2009 . genetic inversion of sparse disk - integrated photometric data of asteroids : application to hipparcos data .
chandrasekhar , s. 1969 .
ellipsoidal figures of equilibrium .
new haven : yale university press .
chang , c. , ip , w. , lin , h. , cheng , y. , ngeow , c. , yang , t , waszczak , a. , kulkarni , s.r . , levitan , d. , sesar , b. , laher , r. , surace , j. , prince , t.a .
313 new asteroid rotation periods from palomar transient factory observations .
http://dx.doi.org/10.1088/0004-637x/788/1/17 [ ] .
chang , c. , waszczak , a. , lin , h. , ip , w. , prince , t.a .
, kulkarni , s.r . , laher , r. , surace , j. , 2014 . a new large super - fast rotator : ( 335433 ) 2005 uw163 .
http://dx.doi.org/10.1088/2041-8205/791/2/l35 [ ] .
chapman , c.r . ,
gaffey , m.,1979 .
reflectance spectra for 277 asteroids . in : http://adsabs.harvard.edu/abs/1979aste.book..655c [ ] .
chapman , c.r . , gaffey , m. , and mcfadden , l. , 1993 .
24-color asteroid survey .
ear - a - dbp-3-rdr-24color - v2.1 .
http://sbn.psi.edu/pds/resource/24color.html [ ] .
chiu , s. , 1994 .
fuzzy model identification based on cluster estimation .
http://dx.doi.org/10.3233/ifs-1994-2306 [ ] .
cikota , s. , ortiz , j.l . ,
cikota , a. , morales , n. , tancredi , g. , 2014 . a photometric search for active main belt asteroids .
deb , s. , & singh , h.p .
, 2009 . light curve analysis of variable stars using fourier decomposition and principal component analysis .
http://dx.doi.org/10.1051/0004-6361/200912851 [ ] .
gehrels , t. , 1956 .
photometric studies of asteroids .
v. the light - curve and phase function of 20 massalia .
giorgini , j.d .
et al . , 1996 .
jpl s on - line solar system data service . in : http://adsabs.harvard.edu/abs/1996dps....28.2504g [ ] .
grav , t. et al . , 2012 .
wise / neowise observations of the jovian trojan population : taxonomy .
http://dx.doi.org/10.1088/0004-637x/759/1/49 [ ] .
hanu , j. , urech , j. , 2012 .
the potential of sparse photometric data in asteroid shape modeling .
http://dx.doi.org/10.1016/j.pss.2012.01.014 [ ] .
hanu , j. et al .
asteroids physical models from combined dense and sparse photometry and scaling of the yorp effect by the observed obliquity distribution .
http://dx.doi.org/10.1051/0004-6361/201220701 [ ] .
hapke , b. , 2012 .
theory of reflectance and emittance spectroscopy , second ed .
cambridge university press , cambridge , uk .
harris , a.w . , burns , j.a . , 1979 .
asteroid rotation : i. tabulation and analysis of rates , pole positions and shapes .
http://dx.doi.org/10.1016/0019-1035(79)90058-7 [ ] .
harris , a.w . , young , j.w . ,
1988 . observations of asteroid phase relations . in : http://adsabs.harvard.edu/abs/1988baas...20..865h [ ] .
harris , a.w . ,
young , j.w .
asteroid lightcurve observations from 19791981 .
http://dx.doi.org/10.1016/0019-1035(89)90056-0 [ ] .
harris , a.w .
et al . , 1989a .
photoelectric observations of asteroids 3 , 24 , 60 , 261 , and 863 .
http://dx.doi.org/10.1016/0019-1035(89)90015-8 [ ] .
harris , a.w .
phase relations of high albedo asteroids - the unusual opposition brightening of 44 nysa and 64 angelina .
http://dx.doi.org/10.1016/0019-1035(89)90057-2 [ ] .
harris , a.w . ,
young , j.w . , dockweiler , t. , gibson , j. , 1992 . asteroid lightcurve observations from 1981 .
http://dx.doi.org/10.1016/0019-1035(92)90195-d [ ] .
harris , a.w .
et al . , 2014 . on the maximum amplitude of harmonics of an asteroid lightcurve .
http://dx.doi.org/10.1016/j.icarus.2014.03.004 [ ] .
helfenstein , p. , veverka , j. , 1989 . physical characterization of asteroid surfaces from photometric analysis . in : http://adsabs.harvard.edu/abs/1989aste.conf..557h [ ] .
hogg , d.w . , bovy , j. , lang , d. , 2010 .
data analysis recipes : fitting a model to data .
ivezi , .
color confirmation of asteroid families .
http://dx.doi.org/10.1086/344077 [ ] .
ivezi , z. , juric , m. , lupton , r.h . ,
tabachnik , s. , quinn , t. , and the sdss collaboration , 2010 .
sdss moving object catalog v3.0 .
ear - a - i0035 - 3-sdssmoc - v3.0 .
http://sbn.psi.edu/pds/resource/sdssmoc.html[nasa planetary data system ] .
kaasalainen , m. , 2004 .
physical models of large number of asteroids from calibrated photometry sparse in time .
http://dx.doi.org/10.1051/0004-6361:20048003 [ ] .
kaasalainen , m. , torppa , j. , muinonen , k. , 2001 .
optimization methods for asteroid lightcurve inversion ii .
the complete inverse problem .
http://dx.doi.org/10.1006/icar.2001.6674 [ ] .
kneevi , z. , milani , a. , 2012 .
asteroids dynamic site astdys .
http://adsabs.harvard.edu/abs/2012iaujd...7p..18k [ ] .
kron , r.g . , 1980 .
photometry of a complete sample of faint galaxies .
http://adsabs.harvard.edu/abs/1980apjs...43..305k [ ] .
kulkarni , s.r .
the intermediate palomar transient factory ( iptf ) begins .
http://adsabs.harvard.edu/abs/2013atel.4807....1k [ ] .
laher , r.r . ,
surace , j. , grillmair , c.j . ,
ofek , e.o . ,
levitan , d. , sesar , b. , van eyken , j. , law , n. , helou , g. , hamam , n. , masci , f. , mattingly , s. , jackson , e. , hacopeans , e. , mi , w. , groom , s. , teplitz , h. , desai , v. , hale , d. , smith , r. , walters , r. , quimby , r. , kasliwal , m. , horesh , a. , bellm , e. , barlow , t. , waszczak , a. , prince , t.a . , kulkarni , s.r .
ipac image processing and data archiving for the palomar transient factory .
http://dx.doi.org/10.1086/677351 [ ] .
law , n. et al .
the palomar transient factory : system overview , performance , and first results .
http://dx.doi.org/10.1086/648598 [ ] .
law , n. et al . , 2010 .
the palomar transient factory survey camera : 1st year performance and results .
http://dx.doi.org/10.1117/12.857400 [ ] .
lazarro , d. et al . , 2004 .
s@xmath31os@xmath20 : the visible spectroscopic survey of 820 asteroids .
http://dx.doi.org/10.1016/j.icarus.2004.06.006 [ ] .
levitan , d. et al , 2011 .
ptf1 j071912.13 + 485834.0 : an outbursting am cvn system discovered by a synoptic survey . http://dx.doi.org/10.1088/0004-637x/739/2/68 [ ] .
lupishko , d.f . ,
tupieva , f.a . ,
velichko , f.p . ,
kiselev , n.n . ,
chernova , g.p . , 1981 .
ubv photometry of the asteroids 19 fortuna and 29 amphitrite . http://adsabs.harvard.edu/abs/1981avest..15...25l [ ] .
lowry , s.c .
, fitzsimmons , a. , pravec , p. , vokrouhlick , d. , boehnhardt , h. , taylor , p.a . , margot , j. , gald , a. , irwin , m. , irwin , j. , kusnirk , p. , 2007 .
direct detection of the asteroidal yorp effect .
http://dx.doi.org/10.1126/science.1139040 [ ] .
mahabal , a. , djorgovski , s.g . ,
turmon , m. , et al . , 2008 .
automated probabilistic classification of transients and variables .
http://dx.doi.org/10.1002/asna.200710943 [ ] .
marsden , b.g . , 1986
. notes from the iau general assembly .
http://www.minorplanetcenter.net/iau/ecs/mpcarchive/1985/mpc_19851227.pdf [ ] .
martin , d.c .
et al . , 2005 .
the _ galaxy evolution explorer _ : a space ultraviolet survey mission .
http://dx.doi.org/10.1086/426387 [ ] .
masci , f.j . ,
hoffman , d.i .
, grillmair , c.j . ,
cutri , r.m . , 2014 .
automated classification of periodic variable stars detected by the wide - field infrared survey explorer .
http://dx.doi.org/10.1088/0004-6256/148/1/21 [ ] .
masiero , j. , jedicke , r. , urech , j. , gwym , s. , denneau , l , larsen , j. , 2009 .
the thousand asteroid light curve survey .
http://dx.doi.org/10.1016/j.icarus.2009.06.012 [ ] .
masiero , j. et al . , 2011 .
main - belt asteroids with wise / neowise
. i. preliminary albedos and diameters . http://dx.doi.org/10.1088/0004-637x/741/2/68 [ ] .
masiero , j. et al . , 2014 .
main - belt asteroids with wise / neowise : near - infrared albedos .
http://dx.doi.org/10.1088/0004-637x/791/2/121 [ ] .
massey , f.j .
the kolmogorov - smirnov test for goodness of fit .
http://dx.doi.org/10.1080/01621459.1951.10500769 [ ] .
matson , d.l . ,
veeder , g.j . ,
tedesco , e.f .
, lebofsky , l.a . ,
walker , r.g . , 1986 . the iras survey of asteroids .
miller , a. , bloom , j.s . , richards , j. , et al . 2014 .
a machine learning method to infer fundamental stellar parameters from photometric light curves .
morissey , p. et al . , 2005 .
the on - orbit performance of the _ galaxy evolution explorer_. http://dx.doi.org/10.1086/424734 [ ] .
morissey , p. et al . , 2007 .
the calibration and data products of _ galex_. http://dx.doi.org/10.1086/520512 [ ] .
muinonen , k. , belskaya , i.n . ,
cellino , a. , delb , m. , levasseur - regourd , a. , penttil , a. , tedesco , e.f . , 2010 .
a three - parameter magnitude phase function for asteroids .
http://dx.doi.org/10.1016/j.icarus.2010.04.003 [ ] .
ofek , e.o .
et al . , 2011 .
a very large array search for 5 ghz radio transients and variables at low galactic latitudes .
http://dx.doi.org/10.1088/0004-637x/740/2/65 [ ] .
ofek , e.o .
the palomar transient factory photometric calibration .
http://dx.doi.org/10.1086/664065 [ ] .
ofek , e.o .
, 2012b . the palomar transient factory photometric catalog 1.0 .
http://dx.doi.org/10.1086/666978 [ ] .
ofek , e.o . ,
matlab package for astronomy and astrophysics .
http://adsabs.harvard.edu/abs/2014ascl.soft07005o [ ] .
oszkiewicz , d.a .
et al . , 2011 .
online multi - parameter phase - curve fitting and application to a large corpus of asteroid photometric data .
http://dx.doi.org/10.1016/j.jqsrt.2011.03.003 [ ] .
oszkiewicz , d.a .
et al . , 2012 .
asteroid taxonomic signatures from photometric phase curves .
http://dx.doi.org/10.1016/j.icarus.2012.02.028 [ ] .
park , m. , oh , h .- s . ,
kim , d. , 2013 .
classification of variable stars using thick - pen transform method .
http://dx.doi.org/10.1086/670671 [ ] .
parker , a. , ivezi , . ,
juri , m. , lupton , r. , sekora , m.d . ,
kowalski , a. , 2008 .
the size distributions of asteroid families in the sdss moving object catalog 4 .
http://dx.doi.org/10.1016/j.icarus.2008.07.002 [ ] .
polishook , d. , borsch , n. , 2009 .
photometry and spin rate distribution of small - sized main belt asteroids .
http://dx.doi.org/10.1016/j.icarus.2008.10.020 [ ] .
polishook , d. et al . , 2012 .
asteroid rotation periods from the palomar transient factory survey .
http://dx.doi.org/10.1111/j.1365-2966.2012.20462.x [ ] .
pravec , p. , harris , a.w . ,
vokrouhlicky , d. , et al . , 2008 .
spin rate distribution of small asteroids .
http://adsabs.harvard.edu/abs/2008icar..197..497p [ ] .
pravec , p. , harris , a.w . ,
michalowski , t. , 2002 .
asteroid rotations . in : http://adsabs.harvard.edu/abs/2002aste.conf..113p
rau , a. et al . ,
2009 . exploring the optical transient sky with the palomar transient factory .
http://dx.doi.org/10.1086/605911 [ ] .
salo , h. , 1987 .
numerical simulations of collisions between rotating particles .
http://dx.doi.org/10.1016/0019-1035(87)90073-x [ ] .
sanchez , j.a . ,
reddy , v. , nathues , a. , cloutis , e.a . ,
mann , p. , hiesinger , h. , 2012 .
phase reddening on near - earth asteroids : implications for mineralogical analysis , space weathering and taxonomic classification .
http://dx.doi.org/10.1016/j.icarus.2012.04.008 [ ] .
seager , s. , 2010 .
http://adsabs.harvard.edu/abs/2010eapp.book.....s [ ] .
princeton , nj : princeton univ . press . pp .
shevchenko , v.g . , 1997 .
analysis of asteroid brightness - phase relations .
http://adsabs.harvard.edu/abs/1997sosyr..31..219s [ ] .
shevchenko , v.g .
et al . , 2012 . opposition effect of trojan asteroids .
http://dx.doi.org/10.1016/j.icarus.2011.11.001 [ ] .
shevchenko , v.g .
et al . , 2013 .
influence on magnitude phase angle dependence of asteroid 596 scheila .
http://adsabs.harvard.edu/abs/2013lpi....44.2706s [ ] .
skrutskie , m.f . , et al . , 2006 .
the two micron all sky survey ( 2mass ) .
http://dx.doi.org/10.1086/498708 [ ] .
steinberg , e. , sari , r. , 2015 .
spins of large asteroids : a hint of a primordial distribution in their spin rates .
http://dx.doi.org/10.1088/0004-6256/149/4/124 [ ] .
stetson , p.b . , 1996 .
on the automatic determination of light - curve parameters for cepheid variables .
http://dx.doi.org/10.1086/133808 [ ] .
sykes , m.v .
et al . , 2000 .
the 2mass asteroid and comet survey .
http://dx.doi.org/10.1006/icar.2000.6366 [ ] .
sykes , m.v .
, cutri , r , m . ,
skrutskie , m.f .
, fowler , j.w . ,
tholen , d.j . , painter , p.e .
, nelson , b. , kirkpatrick , d.j . , 2010 .
2mass asteroid and comet survey v2.0 .
ear - a - i0054/i0055 - 5 - 2mass - v2.0 .
http://sbn.psi.edu/pds/resource/2mass.html [ ] .
tedesco , e.f .
ubv mean asteroid colors .
ear - a-5-ddr - ubv - mean - values - v1.2 .
http://sbn.psi.edu/pds/resource/ubvmean.html [ ] .
tedesco , e.f .
, noah , p.v . , noah , m. , price , s.d . , 2002 .
the supplemental iras minor planet survey .
http://dx.doi.org/10.1086/338320 [ ] .
tedesco , e.f.,egan , m.p . , price , s.d . , 2002 .
the midcourse space experiment infrared minor planet survey .
http://dx.doi.org/10.1086/340960 [ ] .
usui , f. et al . , 2011 .
asteroid catalog using akari : akari / irc mid - infrared asteroid survey .
http://dx.doi.org/10.1093/pasj/63.5.1117 [ ] .
usui , f. et al . , 2014 . a comparative study of infrared asteroid surveys : iras , akari , and wise .
walsh , k.j .
, morbidelli , a. , raymond , s.n . ,
obrien , d.p . ,
mandell , a.m. , 2011 . a low mass for mars from jupiter s early gas - driven migration .
http://dx.doi.org/10.1038/nature10201 [ ] .
warner , b. , harris , a.w .
, pravec , p. , 2009 .
the asteroid lightcurve database .
http://dx.doi.org/10.1016/j.icarus.2009.02.003 [ ] .
, a. et al . , 2013 .
main - belt comets in the palomar transient factory survey : i. the search for extendedness .
http://dx.doi.org/10.1093/mnras/stt951 [ ] .
, g.v . , 2012 .
minor planet astrophotometry .
http://adsabs.harvard.edu/abs/2012phdt.........7w [ ] .
woniak , p.r . ,
williams , s.j . , vestrand , w.t . ,
gupta , v. , 2004 .
identifying red variables in the northern sky variability survey .
http://dx.doi.org/10.1086/425526 [ ] .
wright , e.l .
et al . , 2010 .
the wide - field infrared survey explorer ( wise ) : mission description and initial on - orbit performance .
http://dx.doi.org/10.1088/0004-6256/140/6/1868 [ ] .
xu , s.,binzel , r.p . ,
burbine , t.h . , bus , s.j . , 1995 .
small main - belt asteroid spectroscopic survey .
http://dx.doi.org/10.1006/icar.1995.1075 [ ] .
xu , s.,binzel , r.p . ,
burbine , t.h . , bus , s.j . , 1996 .
small main - belt asteroid survey .
ear - a - m3spec-3-rdr - smass - v2.1 .
http://sbn.psi.edu/pds/resource/smass1.html [ ] .
york , d.g .
et al , 2000 .
the sloan digital sky survey : technical summary .
http://dx.doi.org/10.1086/301513 [ ] .
zacharias , n. , 2010 .
ucac3 pixel processing .
http://dx.doi.org/10.1088/0004-6256/139/6/2208 [ ] .
zellner , b. , tholen , d.j . , tedesco , e.f .
, 1998 . the eight - color asteroid survey - results for 589 minor planets .
http://dx.doi.org/10.1016/0019-1035(85)90133-2 [ ] .
zellner , b. , tholen , d.j . , tedesco , e.f .
eight color asteroid survey .
ear - a-2cp-3-rdr - ecas - v4.0 .
http://sbn.psi.edu/pds/resource/ecas.html [ ] .
|
we fit 54,296 sparsely - sampled asteroid lightcurves in the palomar transient factory ( ptf ) survey to a combined rotation plus phase - function model .
each lightcurve consists of 20 or more observations acquired in a single opposition . using 805 asteroids in our sample that have reference periods in the literature
, we find the reliability of our fitted periods is a complicated function of the period , amplitude , apparent magnitude and other lightcurve attributes . using the 805-asteroid ground - truth sample , we train an automated classifier to estimate ( along with manual inspection ) the validity of the remaining @xmath053,000 fitted periods . by this method
we find 9,033 of our lightcurves ( of @xmath08,300 unique asteroids ) have ` reliable ' periods .
subsequent consideration of asteroids with multiple lightcurve fits indicate a 4% contamination in these ` reliable ' periods . for 3,902 lightcurves with sufficient phase - angle coverage and either a reliably - fit period or low amplitude ,
we examine the distribution of several phase - function parameters , none of which are bimodal though all correlate with the bond albedo and with visible - band colors . comparing the theoretical maximal spin rate of a fluid body with our amplitude versus spin - rate distribution suggests that , if held together only by self - gravity , most asteroids are in general less dense than @xmath02 g/@xmath1 , while c types have a lower limit of between 1 and 2 g/@xmath1 .
these results are in agreement with previous density estimates . for 520 km
diameters , s types rotate faster and have lower amplitudes than c types
. if both populations share the same angular momentum , this may indicate the two types differing ability to deform under rotational stress .
lastly , we compare our absolute magnitudes ( and apparent - magnitude residuals ) to those of the minor planet center s nominal ( @xmath2 , rotation - neglecting ) model ; our phase - function plus fourier - series fitting reduces asteroid photometric rms scatter by a factor @xmath03 .
|
the tensor force plays important roles in nuclear structure .
the study of nuclear matter using the brueckner theory showed that the tensor force is closely related to the binding mechanism and the saturation property of nuclear matter @xcite .
the almost exact calculations with very large model spaces exhibit a large attractive energy comes from the tensor force @xcite .
the tensor force is inferred to be responsible for about a half of the single - particle spin - orbit splitting in light nuclei @xcite . recently due to the development of experimental techniques , we can access to various kinds of unstable nuclei experimentally .
those experiments have been revealing that the shell structures of unstable nuclei may change from those of stable nuclei @xcite .
considering the importance of the tensor force in nuclear structure , the tensor force probably has an effect on such structure changes of nuclei @xcite .
therefore , the study of the effect of the tensor force in neutron - rich nuclei is interesting and important . to study the effect of the tensor force in a relatively large mass region in the nuclear chart including unstable nuclei , we have developed a framework based on a mean - field - type model @xcite .
one of the most important tensor correlations in closed - shell nuclei is a 2-particle2-hole ( 2p2h ) correlation . in a usual hartree - fock calculation ,
the correlation induced by the tensor force can be treated restrictively and the 2p2h correlation is hard to be handled .
the effect of the tensor force is thought to be included in other kinds of forces like the central , ls , and density - dependent forces in the usual hartree - fock calculations . to treat the tensor force directly
, we introduce a single - particle state with the parity and charge mixing considering the pseudoscalar and isovector characters of the pion , which mediates the tensor force @xcite . because a total wave function made from such single - particle states with the parity and charge mixing does not have a good parity and a definite charge number , the parity and charge number projections are performed before variation @xcite .
we call this method the charge- and parity - projected hartree - fock ( cpphf ) method . in the previous studies we applied the cpphf method to the alpha particle and showed that the tensor correlation can be exploited in the cpphf method .
the cpphf method was also applied to @xmath5be to study the effect of the tensor force on alpha clustering @xcite .
there are other attempts to treat the tensor correlation by expanding usual model spaces like a mean - field model @xcite , a shell model @xcite , the antisymmetrized molecular dynamics ( amd ) @xcite . those studies including ours showed the importance of the 2p2h configuration mixing and high - momentum components in single - particle states , which are not treated in usual model space calculations .
neff and his collaborators took a different approach to treat the tensor correlation @xcite .
they used the unitary correlation operator method ( ucom ) to make an effective interaction in a moderate model space . in their effective interaction ,
the attractive correlation by the tensor force is included in other forces like the central and ls forces .
they performed the calculations up to the second order perturbation based on hartree - fock calculations using their effective interaction and obtained a nice agreement of binding energies over the whole mass region @xcite .
otsuka and his collaborators @xcite showed that a particle - hole ( p - h ) correlation by the tensor force is also important and changes single - particle spin - orbit splittings of neutron ( proton ) orbits with proton ( neutron ) numbers .
the effect was inferred in old hartree - fock calculations @xcite . in the present paper
, we apply the cpphf method to sub - closed - shell oxygen isotopes , @xmath0o , @xmath1o , @xmath2o , @xmath3o , and @xmath4o , which are assumed to have sub - closed - shell structures for neutron orbits up to @xmath6 , @xmath7 , @xmath8 , @xmath9 , and @xmath10 respectively , to see the dependence of the correlation induced by the tensor force on neutron numbers .
we extend the cpphf method to treat a three - body force , which is needed to reproduce the saturation property of nuclei with relatively large mass numbers . in section [ sec : cpphf ] we explain the cpphf method with a three - body force . in section [ sec : result ] the results of the cpphf method are presented . in section [ sec : summary ]
we summarize the paper .
in this section we formulate the charge- and parity - projected hartree - fock ( cpphf ) method in the case where a three - body force exists .
a hamiltonian for an a - body system with the two - body and three - body forces can be written as @xmath11 where @xmath12 , @xmath13 , and @xmath14 are one - body , two - body , and three - body operators respectively .
@xmath15 s are coordinates including spin and isospin . in the cpphf method
, we assume as single - particle states the ones with the parity and charge mixing .
it means each single - particle wave function has both positive - parity and negative - parity components and both proton and neutron components .
the single - particle wave function which consists of these four components has the following form , @xmath16 in the above equation , @xmath17 denotes parities , @xmath18 for positive parity and @xmath19 for negative parity , and @xmath20 denotes isospins , @xmath21 for proton and @xmath22 for neutron . in the cpphf method , we take as a wave function of an a - body system a slater determinant which consists of the single - particle states with the parity and charge mixing , @xmath23 here , @xmath24 is the antisymmetrization operator .
because @xmath25 does not have a good parity and a definite charge number , we need to perform the projection operators of parity ( @xmath26 ) and charge number ( @xmath27 ) on @xmath25 to obtain the wave function with a good parity and a definite charge number ; @xmath28 here , @xmath29 is the parity - projection operator , where @xmath30 projects out the positive parity state and @xmath31 projects out the negative parity one .
@xmath32 is the charge - number - projection operator , which projects out the wave function with a charge number @xmath27 .
therefore , @xmath33 has a good parity ( @xmath26 ) and a definite charge number ( @xmath27 ) .
the parity projection operator @xmath29 is defined as @xmath34 where the total parity operator @xmath35 is the product of the parity operator @xmath36 for each single - particle state .
the charge projection operator @xmath32 is defined as @xmath37 where @xmath38 is the charge number operator , which is the sum of the single - particle proton projection operator @xmath39 , and the charge - rotation operator is defined as @xmath40 .
we take the expectation value for the hamiltonian @xmath41 with the projected wave function and obtain the energy functional , @xmath42 the denominator in the right - hand side of the above equation is the normalization of the total wave function , @xmath43 here , @xmath44 is the determinant of the norm matrix between the original single - particle wave functions @xmath45 and the charge - rotated single - particle wave functions @xmath46 . @xmath47
is the determinant of the norm matrix between the original single - particle wave functions @xmath45 and the parity - inverted and charge - rotated single - particle wave functions @xmath48 . @xmath49
the charge - rotated wave function @xmath50 and the parity - inverted and charge - rotated wave function @xmath51 are defined as @xmath52 where @xmath53 is the single - particle parity operator in ( [ eq : parityop ] ) and @xmath54 is the single - particle proton projection operator in ( [ eq : chargeop ] ) .
the numerator in the right - hand side of ( [ eq : etot ] ) is the unnormalized total energy , @xmath55 @xmath56 in the right - hand side of ( [ eq : unnorme ] ) has a similar form as a simple hartree - fock energy but the single - particle wave functions in the ket are modified by the charge rotation , @xmath57 here , @xmath58 is the superposition of @xmath59 weighted by the inverse of the charge - rotated norm matrix @xmath60 , @xmath61 this summation for @xmath62 comes from the antisymmetrization of the total wave function .
the hats in the last two terms represent the antisymmetrization as @xmath63 @xmath64 reduces to a simple hartree - fock energy .
@xmath65 in the right - hand side of ( [ eq : unnorme ] ) has a similar form as @xmath56 but @xmath66 s are replaced by @xmath67 s , @xmath68 here , @xmath69 is the sum of @xmath70 weighted by the inverse of the parity - inverted and charge - rotated norm matrix @xmath71 , @xmath72
we then take the variation of @xmath73 with respect to a single - particle wave function @xmath45 , @xmath74 the lagrange multiplier @xmath75 is introduced to guarantee the ortho - normalization of single - particle wave functions , @xmath76 . as the result
, we obtain the following hartree - fock - like equation with the charge and parity projections ( the cpphf equation ) for each @xmath45 , @xmath77 \notag\\ & = n^{(\pm;z ) } \sum_{b=1}^a \epsilon_{ab } \psi_{\alpha_b}(x_a ) , \label{eq : cphf}\end{aligned}\ ] ] where @xmath78 . here ,
@xmath79 and @xmath80 are defined as follows , @xmath81 @xmath82 , @xmath83 , and @xmath84 are the single - particle operators originated from the one - body , two - body , and three - body operators , which are defined as following , @xmath85 the expressions of @xmath86 , @xmath87 , and @xmath88 are obtained from those of @xmath82 , @xmath83 , and @xmath84 by replacing @xmath89 with @xmath90 .
the notations for the integration of the two - body matrix elements , @xmath91 and for that of the three - body matrix elements , @xmath92 are introduced .
the system of the coupled equations ( [ eq : cphf ] ) for @xmath93 are solved selfconsistently .
we give here the expressions for the expectation value of the kinetic energy @xmath94 with the center of mass correction , that of the two - body potential energy @xmath95 , and that of three - body potential energy @xmath96 for the later convenience .
@xmath97,\end{aligned}\ ] ] @xmath98 @xmath99
in this section , we apply the charge- and parity - projected hartree - fock ( cpphf ) method formulated in the last section to the ground states of the sub - closed oxygen isotopes , @xmath0o , @xmath1o , @xmath2o , @xmath3o and @xmath4o . we assume the parities of these nuclei in the ground states are positive .
those nuclei are assumed to have the closed shell up to @xmath7 for proton and sub - closed or closed shells up to @xmath6 ( @xmath0o ) , @xmath7 ( @xmath1o ) , @xmath8 ( @xmath2o ) , @xmath9 ( @xmath3o ) and @xmath10 ( @xmath4o ) for neutron .
although @xmath4o is know to be unbound from the experiment , we calculate the nucleus to study the shell - configuration dependence of the contribution from the tensor force theoretically .
we assume the spherical symmetry . in this case , only the total angular momentum @xmath100 is a good quantum number of a single - particle state , because parities and charges are mixed in intrinsic single - particle states .
an intrinsic wave function can be written in the following form , @xmath101 a single - particle wave function @xmath102 is composed of four components , proton and positive parity , proton and negative parity , neutron and positive parity , and neutron and negative parity , @xmath103 here , @xmath104 is the eigenfunction of the total angular momentum @xmath105 , @xmath106 is the isospin wave function with @xmath107 for proton and @xmath108 for neutron . @xmath109 and @xmath110 are the orbital angular momentum with positive parity and negative parity respectively . for example @xmath111 and @xmath112 for @xmath113 , @xmath114 and @xmath112 for @xmath115 , and so on .
( [ eq : spwf ] ) indicates that in the present calculation assuming the spherical symmetry only the correlations among the same @xmath100 orbits can be treated .
it is a limitation of the cpphf method with the spherical symmetry . for the calculation of @xmath1o four single - particle states with @xmath113 and two states with @xmath115
are included .
they correspond to @xmath116 , @xmath117 , @xmath118 and @xmath119 for @xmath113 and @xmath120 and @xmath121 for @xmath115 .
because parities and charges are mixed in single - particle states such a classification is approximately valid . for the calculation of @xmath0o one state with @xmath122
is subtracted and for the calculation of @xmath2o , @xmath3o and @xmath4o new states with @xmath123 , @xmath113 and @xmath115 are added one by one . in the present study we use the modified volkov force no . 1 ( mv1 force )
@xcite for the central potential and the g3rs force @xcite for the non - central forces .
the mv1 force is the modified version of the volkov force no .
1 @xcite and includes the @xmath124-function - type three - body force , @xmath125 the majorana parameter in the mv1 force is fixed to 0.6 .
the g3rs is determined from the nucleon - nucleon scattering data .
the effect of the tensor force is effectively included in the mv1 force because the mv1 force is determined so as to reproduce the binding energy of @xmath1o in the absence of the tensor force .
the effect of the tensor force is thought to appear in the @xmath126e channel of the central force as attraction .
therefore we multiply the attraction part in the @xmath126e channel of the central force by @xmath127 .
we also multiply the three - body - force part by @xmath128 .
the @xmath124-function - type three - body force in eq .
( [ eq : delta3b ] ) reduces to the density - dependent two - body force , @xmath129 , for the wave functions of even - even nuclei with time - reversal symmetry in the hartree - fock level @xcite .
@xmath130 is the spin - exchange operator and @xmath131 is a single - particle density . in this case
the density - dependent force only acts on the @xmath126e channel . in the mv1 force @xmath132 is positive and the density - dependent force has the repulsive effect on the @xmath126e channel .
the ls forces determined from the nn scattering data is usually weak to be used in the mean filed ( hartree - fock ) calculation .
hence , we multiply the ls force by 2 . in this case
the strength of the ls force is comparable to those adopted in the skyrme forces and the gogny forces @xcite .
we also multiply the @xmath133 part of the tensor force a numerical factor @xmath134 as in the previous study @xcite , because the cpphf method is a mean - field - type calculation and can only take into account the correlations induced by limited couplings among single - particle states . actually , in the spherical symmetry the 2-particle2-hole ( 2p2h ) correlation which can be treated in the cpphf method
is like ( @xmath135 ) with @xmath136 and @xmath137 .
however 2p-2h configurations with @xmath138 and @xmath139 are also important @xcite .
furthermore , other effects may enhance the tensor correlation as mentioned in our previous paper @xcite . to take into account such effects effectively , we take @xmath134=1.5 ( the strong tensor force case ) in addition to @xmath134=1.0 ( the normal tensor force case ) . @xmath127 and @xmath128 are determined to reproduce the binding energy and the charge radius of @xmath1o for each @xmath134 .
we expand single - particle wave functions in the gaussian basis as in the previous study @xcite .
the number of the gaussian basis used is 10 for each orbit with the minimum range 0.5 fm and the maximum range 10 fm .
the cpphf equation ( [ eq : cphf ] ) is solved by the gradient or the damped gradient method @xcite .
the convergence of the calculation is quite slow for the case with a large difference between a proton number ( @xmath27 ) and a neutron number ( @xmath140 ) . as in @xmath4o . to remedy it , the quadratic constraint potential term @xcite for @xmath27 , @xmath141 ) .
the addition of the constrained potential makes the convergence faster .
the value of @xmath142 is taken as 1000 mev .
we first take @xmath1o as a typical example and show the effect of the tensor force in the cpphf method .
.[table : 16o ] results for @xmath1o in the hf and cpphf method .
@xmath134 is a numerical factor multiplied to the @xmath133 part of the tensor force . @xmath143 and @xmath144 are the total energy and the total kinetic energy respectively .
@xmath145 , @xmath146 , @xmath147 , @xmath148 and @xmath149 are the potential energies form the central , the three - body force , the tensor force , the ls force , and the coulomb force respectively .
those are give in the unit of mev .
@xmath150 is the root - mean - square charge radius in the unit of fm .
[ tbl : 16o ] [ cols="^,^,^,^,^,^,^,^",options="header " , ] in table [ tbl : 16o ] the results for @xmath1o in the hartree - fock ( hf ) and the charge- and parity - projected hartree - fock ( cpphf ) schemes are shown . in the hf scheme we fix @xmath134 , @xmath127 , and @xmath128 to 1.0 . in the cpphf scheme @xmath127 and @xmath128
are 1.025 and 1.25 for the normal tensor force ( @xmath151 ) case , and 1.040 and 1.55 for the strong tensor force ( @xmath152 ) case .
the root - mean - square charge radius @xmath153 is calculated from the proton root - mean - square radius @xmath154 as @xmath155 .
this approximation for @xmath153 corresponds to assume the charge radius of proton as 0.80 fm . in the hf calculation
the expectation value for the potential energy from the tensor force @xmath147 is negligibly small .
if we perform the charge and parity projection before variation ( the cpphf case ) , @xmath147 comes out to be a sizable value .
it becomes about 10 mev for the normal tensor force case and about 40 mev for the strong tensor force case .
this result indicates that the cpphf method is effective to treat the correlation from the tensor force . in the cpphf cases the kinetic energy @xmath144 becomes larger than in the hf case . in the cpphf scheme , to gain the tensor correlation energy the opposite - parity components compared to the simple shell - model picture have to be mixed into single - particle states .
this mixing causes an over shell correlation and , as the result , the kinetic energy becomes larger .
a similar tendency is also observed in the alpha particle case @xcite .
( color online ) intrinsic single - particle wave function of the @xmath113 state with an @xmath156 proton component as a dominant one in @xmath1o as a function of the radial distance @xmath157 .
the solid , dashed , dotted and dashed - dotted curves correspond @xmath156 proton , @xmath158 proton , @xmath156 neutron , and @xmath158 neutron respectively . @xmath159 and @xmath160 are the mixing probabilities of the negative - party and the neutron components respectively . ]
( color online ) intrinsic single - particle wave function of the @xmath115 state with a @xmath161 proton component as a dominant one in @xmath1o as a function of the radial distance @xmath157 .
the solid , dashed , dotted and dashed - dotted curves correspond @xmath162 proton , @xmath161 proton , @xmath162 neutron , and @xmath161 neutron respectively . @xmath163 and @xmath160 are the mixing probabilities of the positive - party and the neutron components respectively . ]
( color online ) intrinsic single - particle wave function of the @xmath113 state with a @xmath158 proton component as a dominant one in @xmath1o as a function of the radial distance @xmath157 .
the solid , dashed , dotted and dashed - dotted curves correspond @xmath156 proton , @xmath158 proton , @xmath156 neutron , and @xmath158 neutron respectively . @xmath163 and @xmath160 are the mixing probabilities of the positive - party and the neutron components respectively . ] in figs .
[ fig : o16wfs1ht15 ] , [ fig : o16wfp3ht15 ] and [ fig : o16wfp1ht15 ] the intrinsic single - particle wave functions in @xmath1o with the strong tensor force ( @xmath164 ) are plotted .
the wave functions plotted have proton components as dominant ones .
the wave function in fig .
[ fig : o16wfs1ht15 ] has a proton @xmath156 component as a dominant component .
the mixing probabilities for negative parity ( @xmath158 ) and neutron are 19.5% and 27.4% respectively . from the figure
you can see that the spread of the @xmath158 components are smaller than the @xmath156 ones .
it indicates that the opposite - parity components induced by the tensor force have high - momentum components . the wave function which has a proton @xmath161 component as a dominant one
is plotted in fig .
[ fig : o16wfp3ht15 ] and the shrinkage of @xmath162 components compared to @xmath161 ones are clearly seen also . in the wave function which has a proton @xmath158 component as a dominant one , which is plotted in fig .
[ fig : o16wfp1ht15 ] , the shrinkage is not so clear compared to the previous two cases , probably because of the orthogonality condition to the wave function to the first @xmath113 state in fig .
[ fig : o16wfs1ht15 ] . in the alpha particle case ,
@xmath158 components mixing into @xmath156 ones are also compact in size @xcite .
the importance of this shrinkage is confirmed in a shell model calculation @xcite and the antisymmetrized molecular dynamics ( amd ) calculation @xcite , too .
the present result infer that the shrinkage of mixing single - particle wave functions is generally important in heavier - mass region .
there are also wave functions with a neutron component as a dominant one .
the general tendency just mentioned above is almost the same if proton and neutron are interchanged .
( color online ) densities for @xmath1o in the hartree - fock ( hf ) and the charge- and parity - projected hartree - fock ( cpphf ) methods as a function of the radial distance ( @xmath157 ) .
the solid , the dashed , and the dotted curves correspond to the hf calculation , the cpphf calculation with the normal tensor force ( @xmath165 ) , and the cpphf calculation with the strong tensor force ( @xmath164 ) . ]
( color online ) absolute values of the charge form factor for @xmath1o in the hartree - fock ( hf ) and the charge- and parity - projected hartree - fock ( cpphf ) methods as a function of the momentum squared ( @xmath166 ) .
the solid , the dashed , and the dotted curves correspond to the hf calculation , the cpphf calculation with the normal tensor force ( @xmath165 ) , and the cpphf calculation with the strong tensor force ( @xmath164 ) . ] in fig .
[ fig : o16density ] the densities for @xmath1o in the hf and cpphf calculations are shown . because the tensor force induces the opposite - parity components with narrower widths in single - particle wave functions ,
the densities are depleted in the middle in the cpphf calculations compared to the one in the hf calculation .
this effect is larger for the case with the strong tensor force as expected . to see the effect of the tensor correlation more clearly , in fig .
[ fig : o16cff ] the charge form factors are plotted as a function of the momentum squared .
from the figure , higher - momentum components appear in the cpphf calculation .
it indicates that the tensor force enhances the charge form factor in a high - momentum region .
the short - range correlation , which is not treated properly here , should have a contribution to the charge form factor in the high - momentum region .
hence , we need further investigation to compare the present result of the charge form factor in the cpphf method with the experimental data . the enhancement of the charge form factor in a high - momentum region is also found in @xmath167he in the calculation with the charge- and parity - projected relativistic mean field model @xcite . in this subsection
we show the results for the sub - closed - shell oxygen isotopes .
( color online ) binding energies per particle with minus sign for sub - closed - shell oxygen isotopes .
the horizontal line indicates mass numbers .
the circle symbols correspond to the hartree - fock ( hf ) calculation , the triangle ones to the charge- and parity projected hartree - fock ( cpphf ) calculation with the normal tensor force ( @xmath165 ) , and the diamond ones to the cpphf calculation with the strong tensor force ( @xmath164 ) .
the square symbols indicate the experimental data @xcite . ] in fig .
[ fig : be ] , the results for the binding energies per particle in the hartree - fock ( hf ) calculation ( the circle symbols ) , the charge- and parity - projected hartree - fock ( cpphf ) calculation with the normal tensor force ( @xmath134=1.0 ) ( the triangle symbols ) and the cpphf calculation with the strong tensor force ( @xmath134=1.5 ) ( the diamond symbols ) are shown .
the experimental data ( the square symbols ) @xcite are also plotted.the general tendency is reproduced in our result , although the agreement with the experimental data is not as good as the hartree - fock - type calculations @xcite .
the hf calculation with the mv1 force underestimates the binding energy of @xmath0o and @xmath2o a little bit largely .
it indicates that if we adopt our calculation on the central and the density - dependent forces in the recent sophisticated effective interaction , the agreement with the experimental results should become better .
there is an ambiguity in the treatment of the density - dependent force when we perform the parity and charge projections , because the density - dependent force can not be written in a simple two - body operator form .
it causes a difficulty when we use the density - dependent force in the cpphf calculation .
the optimization of the central force and the management of the density - dependent force will be our future problems .
the agreement with the experimental data is good for @xmath0o in the cpphf method with the strong tensor force . in this case
quite a large attractive potential energy comes from the tensor force as shown below .
( color online ) root - mean - square matter radii for sub - closed - shell oxygen isotopes .
the horizontal line indicates mass numbers .
the circle symbols correspond to the hartree - fock ( hf ) calculation , the triangle ones to the charge- and parity projected hartree - fock ( cpphf ) calculation with the normal tensor force ( @xmath165 ) , and the diamond ones to the cpphf calculation with the strong tensor force ( @xmath164 ) .
the square symbols indicate the experimental data with error bars @xcite . ] in fig .
[ fig : rm ] , the results for the root - mean - square matter radii ( @xmath168 ) are plotted with the experimental data @xcite .
we use the same symbol for each case in fig .
[ fig : be ] . except for @xmath0o
the results for all the three cases are almost the same and reproduce the experimental data well . for @xmath0o the cpphf calculation with the strong tensor force
gives a smaller matter radius compared to the other two calculations and the agreement with the experimental data becomes better in the strong tensor force case .
the reduction of the radius is caused by the shrinkage of the opposite parity component in a single - particle wave function induced by the tensor correlation .
such an effect can not be treated in simple hartree - fock calculations .
( color online ) kinetic energy ( @xmath144 ) per particle and the sum of the potential energies from the central force @xmath145 , the coulomb force @xmath149 , and the three - body force @xmath146 divided by mass numbers .
the horizontal axis indicates mass numbers .
the dashed lines correspond to the kinetic energy and the solid lines to the sum of the potential energies .
the circle , triangle , and diamond symbols correspond to the hartree - fock ( hf ) calculation , the charge- and parity - projected hartree - fock ( cpphf ) calculation with the normal tensor force ( @xmath165 ) , and the cpphf calculation with the strong tensor force ( @xmath164 ) . ]
( color online ) potential energies from the tensor force ( @xmath147 ) and the ls force ( @xmath148 ) divided by mass numbers .
the horizontal axis indicates mass numbers .
the solid lines correspond to the potential energy from the tensor force .
the dashed line corresponds to the potential energy from the ls force in the cpphf calculation with the normal tensor force .
other results for the ls force are almost unchanged
. the meanings of the symbols are the same as in fig .
[ fig : tandvc ] . ] to see the effect of the tensor force on the binding mechanism in the oxygen isotopes in fig .
[ fig : tandvc ] the results for the total kinetic energy @xmath144 ( the dashed lines ) and the sum of the potential energies from the central , the three - body and the coulomb forces @xmath169 ( the solid lines ) are plotted .
we also show in fig .
[ fig : vtandvls ] the result for the potential energies from the tensor force @xmath147 ( the solid lines ) and the ls force @xmath148 ( the dashed line ) .
the same symbols are used for the hf calculation and the cpphf calculations with the normal and strong tensor forces as in fig .
[ fig : be ] .
all values are divided by mass numbers to make a isotope dependence clear .
the total kinetic energy and the sum of the potential energies from the central , three - body and coulomb forces show a volume - like behavior . the kinetic energy for @xmath0o in the cpphf calculation with the strong tensor force is larger than the other two cases .
the increase of the kinetic energy is also caused by the strong tensor correlation in @xmath0o , because to gain the correlation energy from the tensor force the opposite - parity components must mix into single - particle states and the opposite - parity components have larger kinetic energy .
the potential energies from the ls force behave in almost the same manner for the three cases .
it becomes attractive if neutron shells are jj - closed and negligibly small if neutron shells are ls - closed .
the potential energy from the tensor force becomes weakly repulsive in the hf calculation for all the oxygen isotopes . in the cpphf calculations
the potential energies from the tensor force become attractive for all the oxygen isotopes .
in contrast to the ls potential energies , the tensor potential energies have sizable values in ls - closed shell nuclei like @xmath1o and @xmath4o . the attraction from the tensor force
is the same order that from the ls force even in the case with the normal tensor force . in cpphf case @xmath147
becomes maximum in @xmath0o and decreases with the mass number .
for the strong tensor force case , the attractive energy from the tensor force becomes larger as expected .
the attractive energy is quite large for @xmath0o in the cpphf calculation with the strong tensor force . for other isotopes
the attractive energies from the tensor force are small and do not change so much with the mass number .
probabilities of the mixing of the opposite parity components for single - particle states with neutron components as dominant ones in the charge- and parity projected hartree - fock calculation with the normal tensor force .
the horizontal axis indicates mass numbers .
the square , circle , triangle , asterisk , diamond , and star symbols correspond to the first @xmath100=1/2 ( @xmath170 ; @xmath156 dominant ) , the first @xmath100=3/2 ( 0@xmath171 ; @xmath161 dominant ) , the second @xmath100=1/2 ( @xmath172 ; @xmath158 dominant ) , the first @xmath100=5/2 ( @xmath173 ; @xmath174 dominant ) , the third @xmath100=1/2 ( @xmath175 ; @xmath156 dominant ) , and the second @xmath100=3/2 ( @xmath176 ; @xmath162 dominant ) , respectively . ] in fig .
[ fig : pmixnt10 ] the probabilities of the mixing of the opposite parity components @xmath177 in the single - particle states with neutron components as the dominant ones in the result of the cpphf calculation with the normal tensor force are shown .
the result for the strong tensor case shows almost the same tendency . in the usual shell - model classification
the first @xmath100=1/2 state ( @xmath156 dominant ) , the first @xmath100=3/2 state ( @xmath161 dominant ) , the second @xmath100=1/2 state ( @xmath158 dominant ) , the first @xmath100=5/2 state ( @xmath174 dominant ) , the third @xmath100=1/2 state ( @xmath156 dominant ) , and the second @xmath100=3/2 state ( @xmath162 dominant ) correspond to @xmath178 , @xmath6 , @xmath7 , @xmath8 , @xmath9 , and @xmath10 respectively .
all those states are mixed states of positive and negative parities as in eq .
( [ eq : spwf ] ) .
we add a prime in the following to indicate each single - particle state has both positive - parity and negative - parity components .
for example , a @xmath170 state has @xmath156 ( positive - parity ) and @xmath158 ( negative parity ) components with an @xmath156 component as a dominant one . for @xmath0o
the neutron orbits are filled up to @xmath179 .
the mixing probability of the opposite parity components for @xmath170 is larger than 20% and that for @xmath179 is a few % in @xmath0o . in the cpphf method
the tensor correlation energy is gained by parity mixing and , therefore , @xmath177 is a measure to indicate how each single - particle orbit contributes to the tensor correlation .
the large @xmath177 for @xmath170 indicates that this orbit is largely affected by the tensor correlation . in @xmath1o
, the @xmath172 neutron orbit is added newly .
the addition of the @xmath172 orbit reduces the parity mixing of the @xmath170 because they have the same total angular momentum @xmath100=1/2 . as the result ,
the tensor correlation energy becomes smaller .
this effect is more significant for the strong tensor force case as seen in fig .
[ fig : vtandvls ] . in @xmath2o
the neutron @xmath173 is filled .
because there are no @xmath100=5/2 below , @xmath177 s for the states filled already in @xmath1o do not change largely . in @xmath3o
the @xmath175 orbit is newly occupied .
the occupation of @xmath175 reduces @xmath177 s for the states with @xmath100=1/2 , @xmath170 and @xmath172 .
the tensor correlation energy is enhanced , although the amount of the change is small .
finally , in @xmath4o the neutron @xmath176 is filled .
s for the previously filled orbits change by this addition but the changes are not so large .
the change of @xmath177 as shown above indicates that the large change of the tensor correlation energy from @xmath0o to @xmath1o in the strong tensor force case is caused by the blocking effect for the @xmath100=1/2 orbits .
the blocking effect of this kind is shown to be important in the single - particle @xmath180-splitting in @xmath181he @xcite .
the effect of the blocking is less significant for the excess neutron orbits .
the effect of the blocking on binding energies in neutron excess oxygen isotopes seems to be small but the blocking may affect single - particle natures or collectivity in neutron excess oxygen isotopes , because the mixing probability is affected by the blocking effect .
the negligible @xmath177 for the @xmath173 orbit indicates that this orbit does not contribute to the tensor correlation although there is no other occupied orbits which have @xmath100=5/2 .
the main part of the tensor correlation comes from the @xmath182 channel .
because the proton @xmath100=5/2 orbit is not filled in neutron - excess oxygen isotopes , the @xmath173 orbit is hard to contribute to the tensor correlation and the mixing probability of it becomes small .
we have studied the effect of the tensor force in the sub - closed - shell oxygen isotopes using the charge- and parity - projected hartree - fock ( cpphf ) method .
we have extended the cpphf method to the cases with hamiltonians including three - body forces , although the extension is straightforward . in the cpphf method
the parity and charge - number projections are performed before variation . by applying the cpphf method to the oxygen isotopes actually
, we have found that a sizable potential energy from the tensor force is obtained in the cpphf method while in the hartree - fock calculation quite a small potential energy from the tensor force is obtained .
we have investigated @xmath1o in some details .
the correlation energy from the tensor force is about 10 mev for the normal tensor force case and about 40 mev for the strong tensor force case . in the strong tensor force case the strength of the @xmath133 channel in the tensor force
is multiplied by 1.5 .
the opposite parity components induced in single - particle states by the tensor force have compact sizes as compared to the normal parity components .
this indicates that for the tensor correlation high - momentum components are important as already found in the alpha particle case @xcite .
the present result infers that the importance of the high - momentum component for the tensor correlation is valid in heavier mass nuclei .
we have also shown the density and the charge form factor calculated in the cpphf method .
the density in the cpphf method is reduced around the center and is pulled in to the inside region .
this is caused by the parity mixing and the shrinkage of single - particle wave functions of opposite parities .
the effect of the shrinkage appears in the charge form factor as a tail in a high - momentum region , because the shrinkage of the single - particle wave functions induces high - momentum components in the density . in the results for the oxygen isotopes ,
the general tendencies for the binding energies and the matter radii are reproduced in the cpphf calculation with the effective interaction adopted here , while the agreement of the binding energy with the experimental data is not so good compared to the hartree - fock - type calculations with recent effective interactions .
the root - mean - square matter radii are well reproduced within error bars except for @xmath0o with both the normal tensor force and the strong tensor force . in the strong tensor force case ,
the matter radius of @xmath0o becomes smaller and close to the experimental data .
the reduction of the matter radius in @xmath0o with the strong tensor force is due to the shrinkage of single - particle wave function by the strong tensor correlation , which is large in @xmath0o .
because the @xmath7 neutron orbit is not occupied in @xmath0o , there are no blocking states for the @xmath178 proton orbit in the tensor correlation and , therefore , the tensor correlation becomes large .
actually , the correlation energy from the tensor force per particle amounts to more than 5 mev in this case .
for all the oxygen isotopes the calculated potential energy from the tensor force is in the same order as that from the ls force for the normal tensor force case . in the strong tensor force case
it becomes about two times larger .
in contrast to the potential energy from the ls force , the potential energy from the tensor force in an ls - closed - shell nucleus does not become close to zero in the cpphf calculation . in the hartree - fock calculation
it becomes negligibly small because both the total spin and the total orbital angular momenta is almost zero in an ls - closed - shell nucleus .
the sudden decrease of the potential energy from the tensor force from @xmath0o to @xmath1o is attributed to the blocking effect of the @xmath100=1/2 orbits .
the blocking effect is also seen in the neutron - excess oxygen isotopes but does not affect the binding energy largely . in the present study we have applied the cpphf method to the ground states of the sub - closed - shell oxygen isotopes .
the application to odd - mass nuclei and open - shell nuclei to study the effect of the tensor force on single - particle natures and the change of collectivity by the tensor correlation in a neutron excess region are interesting because the tensor force changes the spin , the orbital angular momenta and the isospin of nucleon orbits simultaneously , which is realized in the cpphf method by the parity and charge mixing . as for an effective interaction ,
we combine the available effective interaction and the tensor force in the free space with some modifications .
we need to use effective interactions which have the connection with the realistic nuclear forces to reveal the relation between nuclear structure and the underlying nuclear force and have the consistency between the tensor force and other forces like the central and ls forces .
the study in such a direction is also important and now under progress .
we acknowledge fruitful discussions with prof .
h. horiuchi on the role of the tensor force in light nuclei .
this work is supported by the grant - in - aid for the 21st century coe `` center for diversity and universality in physics '' from the ministry of education , culture , sports , science and technology ( mext ) of japan .
this work is partially performed in the research project for the study of unstable nuclei from nuclear cluster aspects sponsored by the institute of physical and chemical research ( riken ) and a part of the calculation of the present study was performed on the riken super combined cluster system ( rscc ) .
bethe , annu .
* 21 * , 93 ( 1971 ) .
y. akaishi , in cluster models and other topics , edited by t.t.s . kuo and e.
osnes ( world scientific , singapore , 1986 ) , p. 259 .
h. kamada , a. nogga , w. glckle , e. hiyama , m. kamimura , k. varga , y. suzuki , m. viviani , a. kievsky , s. rosati , j. carlson , s.c .
pieper , r.b .
wiringa , p. navrtil , b.r .
barrett , n. barnea , w. leidemann , and g. orlandini , phys .
c * 64 * , 044001 ( 2001 ) .
s. takagi , w. watari , and m. yasuno , prog .
* 22 * , 549 ( 1959 ) ; t. terasawa , prog .
* 23 * , 87 ( 1960 ) ; a. arima and t. terasawa , prog .
* 23 * , 115 ( 1960 ) .
k. and and h. band , prog .
66 * , 227 ( 1980 ) .
t. myo , k. kat , and k. ikeda , prog .
* 113 * , 763 ( 2005 ) .
a. ozawa , t. kobayashi , t. suzuki3 , k. yoshida , and i. tanihata , phys .
* 84 * , 5493 ( 2000 ) .
h. grawe , springer lect .
notes in phys . *
651 * , 33 ( 2004 ) .
t. otsuka , t. suzuki , r. fujimoto , h. grawe , and y. akaishi phys .
lett . * 95 * , 232502 ( 2005 ) h. toki , s. sugimoto , and k. ikeda , prog .
phys . * 108 * 903 ( 2002 ) .
y. ogawa , h. toki , s. tamenaga , h. shen , a. hosaka , s. sugimoto , and k. ikeda , prog .
theor . phys . *
111 * , 75 ( 2004 ) .
s. sugimoto , k. ikeda , and h. toki , nucl .
a * 740 * , 77 ( 2004 ) .
y. ogawa , h. toki , s. tamenaga , s. sugimoto , and k. ikeda , phys .
rev c * 73 * , 034301 ( 2006 ) .
s. sugimoto , k. ikeda , and h. toki , nucl - th/0511087 .
y. akaishi , nucl .
a * 738 * , 80 ( 2004 ) .
a. dot , y. kanada - enyo , h. horiuchi , y. akaishi , k. ikeda , prog .
* 115 * , 1069 ( 2006 ) .
t. neff and h. feldmeier , nucl .
a * 713 * , 311 ( 2004 ) .
r. roth , p. papakonstantinou , n. paar , h. hergert , t. neff , and h. feldmeier , phys .
c 73 , 044312 ( 2006 ) .
tarbutton and k.t.r .
davies , nucl phys .
a * 120*,1 ( 1968 ) .
a. bouyssy , j .- f .
mathiot , n. van giai , and s. marcos , phys .
c * 36 * , 380 ( 1987 ) .
t. ando , k. ikeda , and a. tohsaki - suzuki , prog .
64 * , 1608 ( 1980 ) .
a. b. volkov , nucl . phys .
* 74 * , 33 ( 1965 ) .
r. tamagaki , prog .
. phys . * 39 * , 91 ( 1968 ) .
d. vautherin and d.m .
brink , phys .
c * 5 * 626 ( 1972 ) .
z. patyk , a. baran , j.f .
berger , j. decharg j. dobaczewski , p. ring , and a. sobiczewski , phys .
c * 59 * , 704 ( 1999 ) .
t. myo et al . in preparation .
reinhard , in _ cmpulationl nuclear phyiscs 1 _ ( edited by k. langanke , j.a .
maruhn , and s.e .
koonin , springer - verlag , berlin , 1991 ) chapter 2 .
h. flocard , p. quentin , a.k .
kerman , and d. vautherin , nucl .
phys . a * 203 * , 433 ( 1973 ) .
g. audi , a. h. wapstra and c. thibault , nucl .
a * 729 * , 337 ( 2003 ) .
a. ozawa , t. suzuki and i. tanihata , nucl .
a * 693 * , 32 ( 2001 ) .
m. bender , p .- h .
heenen , and p .-
reinhard , rev .
phys . * 75 * , 121 ( 2003 ) and the references therein .
d. vretenar , a.v .
afanasjev , g.a .
lalazissis , and p. ring , phys .
* 409 * , 101 ( 2005 ) and the references therein .
h. nakada and m. sato , nucl .
a * 699 * , 511 ( 2002 ) .
h. nakada , nucl .
a * 764 * , 117 ( 2006 ) .
|
recently , we developed a mean - field - type framework which treats the correlation induced by the tensor force . to exploit the tensor correlation we introduce single - particle states with the parity and charge mixing . to make a total wave function have a definite charge number and a good parity , the charge number and parity projections are performed .
taking a variation of the projected wave function with respect to single - particle states a hartree - fock - like equation , the charge- and parity - projected hartree - fock equation , is obtained . in the charge- and parity - projected hartree - fock method ,
we solve the equation selfconsistently .
in this paper we extend the charge- and parity - projected hartree - fock method to include a three - body force , which is important to reproduce the saturation property of nuclei in mean - field frameworks .
we apply the charge- and parity - projected hartree - fock method to sub - closed - shell oxygen isotopes ( @xmath0o , @xmath1o , @xmath2o , @xmath3o , and @xmath4o ) to study the effect of the tenor correlation and its dependence on neutron numbers .
we obtain reasonable binding energies and matter radii for these nuclei .
it is found that relatively large energy gains come from the tensor force in these isotopes and there is the blocking effect by occupied neutron orbits on the tensor correlation .
|
hordesat @xcite is a modular massively parallel sat solver which allows the integration of any sequential cdcl - based sat solver in its core .
this enables the transfer of advancements in cdcl sat solving to a parallel setting .
experiments showed that hordesat can achieve superlinear average speedup on hard benchmarks .
the logic of quantified boolean formulas ( qbfs ) extends sat by explicit quantification of propositional variables .
problems in complexity classes beyond np , particularly pspace - complete problems in domains like , e.g. , formal verification , reactive synthesis , or planning , can naturally be encoded as qbfs .
qbf solvers based on qcdcl , the qbf - specific variant of cdcl , apply techniques similar to cdcl sat solvers . thanks to this fact , it is possible to replace the sat solver in the core of hordesat by _ any _ qcdcl qbf solver .
thereby , it is not necessary to change the framework of hordesat which controls the sharing of learned information and the execution of the core solver instances .
we integrated the latest public version 5.0 of the qcdcl - based solver depqbf @xcite in hordesat to obtain the massively parallel qbf solver hordeqbf .
we present the implementation of hordeqbf , which is not tailored towards the use of depqbf as a core solver , and evaluate its scalability on a computer cluster with 1024 processor cores .
experiments using the application benchmarks of the 2014 qbf gallery show that hordeqbf achieves superlinear average and median speedup for hard instances .
we consider closed qbfs @xmath0 in _ prenex cnf ( pcnf ) _ consisting of a quantifier - free cnf @xmath1 over a set @xmath2 of variables and a _ quantifier prefix _ @xmath3 in which @xmath4 and @xmath5 .
qbf solving with clause and cube learning ( _ qcdcl _ ) @xcite , also called _ constraint learning _ , is a generalization of _ conflict - driven clause learning ( cdcl ) _ for sat .
the variables in a pcnf @xmath6 are assigned by _ decision making _ , _ unit propagation _ , and _ pure literal detection_. assignments by decision making have to follow the prefix ordering from left to right . if a clause is falsified under the current assignment @xmath7 , then a _ learned clause _
@xmath8 is derived from @xmath6 by _ q - resolution _
@xcite and added conjunctively to @xmath6 .
if all clauses are satisfied under @xmath7 , then a _ learned cube _ is constructed from @xmath7 and added disjunctively to @xmath6 .
learned cubes may also be derived by _ term resolution _
@xcite , a variant of q - resolution applied to previously learned cubes . after a new clause or cube
has been learned , assignments are retracted during _
backtracking_. qcdcl terminates if and only if the _ empty clause ( resp .
cube ) _ is derived during learning , indicating that @xmath6 is unsatisfiable ( resp .
satisfiable ) .
approaches to parallel qbf solving are based on shared and distributed memory architectures .
pqsolve @xcite is an early parallel dpll @xcite solver without knowledge sharing .
it comes with a dynamic master / slave framework implemented using the message passing interface ( mpi ) @xcite .
search space is partitioned among master and slaves by variable assignments .
qmiraxt @xcite is a multithreaded qcdcl solver with search space partitioning .
paqube @xcite is an mpi - based parallel variant of the qcdcl solver qube @xcite .
clause and cube sharing in paqube can be adapted dynamically at run time .
search space is partitioned like in the sat solver psato based on _ guiding paths _
@xcite . the mpi - based solver mpidepqbf @xcite implements a master / worker architecture without knowledge sharing .
a worker consists of an instance of the qcdcl solver depqbf @xcite .
the master balances the workload by generating subproblems defined by variable assignments ( assumptions ) , which are solved by the workers .
parallel solving approaches have also been presented for quantified csps @xcite and non - pcnf qbfs @xcite .
hordeqbf is a parallel portfolio solver with clause and cube sharing .
whereas sequential portfolio solvers like aqme @xcite include different qbf solvers , hordeqbf integrates instances of the same qcdcl solver ( i.e. , depqbf ) . unlike mpidepqbf , hordeqbf
does not rely on search space partitioning .
instead , the parallel instances of depqbf are diversified by different parameter settings .
hordesat is a portfolio sat solver with clause sharing @xcite .
it can be viewed as a multithreaded program running several instances of a sequential sat solver and communicating via mpi with other instances of the same program .
the parallelization framework has three main tasks : to ensure that the core solvers are diversified , to handle the clause exchange , and to stop all the solvers when one of them has solved the problem . to communicate with the core solvers it uses an api which is described in detail in the hordesat paper @xcite .
since the hordeqbf interface is identical , we only briefly list the most relevant methods : this method tells the core solver to diversify its settings .
the specifics of diversification are left to the solver . the description for depqbf is given in the following section .
this method is used to import learned clauses ( and cubes ) received from other solvers of the portfolio .
this method sets a callback class that will process the clauses ( and cubes ) shared by this solver .
in parallel qcdcl - based qbf solving , learned cubes may be shared among the solver instances in addition to learned clauses . although hordesat does not provide api functions dedicated to cube sharing , its available api readily supports it .
we describe the integration of the qcdcl - based qbf solver depqbf in hordeqbf , which applies to any qcdcl - based qbf solver .
we rely on version 5.0 of depqbf which comes with a dynamic variant of _ blocked clause elimination ( qbce ) _
@xcite for advanced cube learning .
qbce allows to eliminate redundant clauses from a pcnf @xcite .
dynamic qbce is applied frequently during the solving process .
if all clauses in the pcnf are satisfied under the current assignment or removed by qbce , then a cube is learned .
depqbf features a sophisticated analysis of variable dependencies in a pcnf @xcite to relax the linear ordering of variables in the prefix . for the experiments in this paper , however , we disabled dependency analysis for both hordeqbf and the sequential variant of depqbf since the use of dependency information causes run time overhead ( during clause / cube learning ) in addition to overhead already caused by dynamic blocked clause elimination ( qbce ) @xcite .
we modified depqbf as follows to integrate it in hordeqbf
. learned constraints are exported to the master process right after they have been learned .
the master does not distinguish between learned clauses and cubes but treats them as sorted lists of literals .
we add special marker literals to learned clauses and cubes to distinguish between them at the time when the master provides the workers with sets of shared learned constraints . in depqbf
we check whether shared constraints are available for import after a restart has been carried out . to this end , we modified the restart policy of depqbf to always backtrack to decision level zero .
this is different from the original restart policy of depqbf @xcite , where the solver backtracks to higher decision levels depending on the current assignment . after a restart ,
available shared constraints are imported , the watched data structures are updated , and qcdcl continues by propagating unit literals resulting from imported constraints .
every instance of depqbf receives a random seed from the master and diversifies the solving process as follows .
the values of variables in the _ assignment cache _
@xcite are initialized at random . in general ,
decision variables are assigned to the cached value ( if any ) .
the assignment cache is updated with values assigned by unit propagation and pure literal detection . as an effect of random initialization ,
the first value assigned to a decision variable is always a random value .
parameters of _ variable activity scaling _ are set at random .
depqbf implements variable activities similar to minisat @xcite . additionally , the amount ( percentage ) of learned constraints that are removed periodically is initialized at random .
depqbf stores learned clauses and cubes in separate lists with certain capacities . if a list has been filled during learning then less frequently used constraints are removed and the capacity of the list is increased .
depqbf implements a _ nested restart scheme _ similar to picosat @xcite , the parameters of which are randomly selected .
variants of _ dynamic qbce _
@xcite are enabled at random , including switching off dynamic qbce at all , or applying qbce only as a preprocessing or inprocessing step .
finally , applications of _ long - distance resolution _
@xcite , an extension of traditional q - resolution @xcite used to derive learned constraints , are toggled at random .
to examine our portfolio - based parallel qbf solver hordeqbf we performed experiments using all the 735 benchmark problems from the application track of the 2014 qbf gallery @xcite .
we compared hordeqbf with depqbf , which is the qbf solver in the core of hordeqbf .
.the speedup of hordeqbf configurations relative to depqbf .
the second column is the number of instances solved by hordeqbf , the third is the number of instances solved by both depqbf ( in 50000s ) and the hordeqbf ( in 900s ) .
the following six columns contain the average , total , and median speedups for either all the instances solved by hordeqbf or only big instances ( solved after 10@xmath9#cores seconds by depqbf ) .
the last column is the parallel efficiency ( median speedup/#cores ) .
[ cols= " > , > , > , > , > , > , > , > , > , > " , ] # cores seconds by depqbf the data corresponding to columns 79 of table [ tab - speedup ] ) .
] the experiments were run on a cluster with nodes having two octa - core 2.6 ghz intel xeon e5 - 2670 processors ( sandy bridge ) and 64 gb of main memory .
each node has 16 cores and we used 64 nodes which amounts in the total of 1024 cores .
the nodes communicate using an infiniband 4x qdr interconnect and use the suse linux enterprise server 11 ( x86_64 ) ( patch level 3 ) operating system .
hordeqbf was compiled using the icpc compiler version 15.0.2 .
the complete source code and detailed experimental results are available at http://baldur.iti.kit.edu / hordesat/. we ran experiments using 2 , 4 , , 64 cluster nodes . on each node we ran four processes with four threads each , which amounts to 16 core solver ( depqbf ) instances per node .
the results are summarized in figure [ fig_cact ] using cactus plots .
we can observe that increasing the number of cores is beneficial for both sat and unsat instances since the number of solved instances steadily increases and runtimes are reduced .
however , it is not easy to see from a cactus plot whether the additional performance is a reasonable return on the invested hardware resources . therefore we include table [ tab - speedup ] in order to quantify the overall scalability of hordeqbf .
we compute speedups for all the instances solved by the parallel solver .
we ran depqbf with a time limit @xmath10 and for the instances it did not solve we use the runtime of @xmath11 in speedup calculation .
the parallel configurations have a time limit of @xmath12 .
columns 4 , 5 , and 6 of table [ tab - speedup ] show the average , total ( sum of sequential runtimes divided by the sum of parallel runtimes ) and median speedup values respectively . while the average and total speedup values are high , the median speedup is below one .
nevertheless , these figures treat hordeqbf unfairly since the majority of the benchmarks is easy ( solvable under a minute by depqbf ) and it makes no sense to use large computer clusters to solve them . in parallel computing , it is usual to analyze the performance on many processors using _
weak scaling _ where one increases the amount of work involved in the considered instances proportionally to the number of processors . therefore in columns 79 we restrict ourselves to `` big instances '' where depqbf needs at least 10@xmath9(the number of cores used by hordeqbf ) seconds to solve them .
the average , total and median speedup values get significantly larger and in fact we obtain highly superlinear average and median speedups .
figure [ fig_speedup ] shows the distribution of speedups for these instances , it also reveals how many instances ( x - axis ) qualify as `` big instances '' .
we showed that qbf solving can be successfully parallelized using the same techniques as for massively parallel sat solving .
our parallel qbf solver hordeqbf achieved superlinear total and median speedups for hard instances , i.e. , instances where parallelization makes sense . as future work
it would be interesting to consider further variants of q - resolution systems @xcite ( apart from traditional @xcite and long - distance resolution @xcite ) as a means of diversification in hordeqbf , which would amount to a combination of qbf proof systems with different power .
further , it may be promising to equip hordeqbf with search space partitioning as in feldmann , r. , monien , b. , schamberger , s. : a distributed algorithm to evaluate quantified boolean formulae . in : proc . of the 17th nat .
conference on artificial intelligence ( aaai 2000 ) .
aaai ( 2000 )
|
the recently developed massively parallel satisfiability ( sat ) solver hordesat was designed in a modular way to allow the integration of any sequential cdcl - based sat solver in its core .
we integrated the qcdcl - based quantified boolean formula ( qbf ) solver depqbf in hordesat to obtain a massively parallel qbf solver hordeqbf . in this paper
we describe the details of this integration and report on results of the experimental evaluation of hordeqbf s performance .
hordeqbf achieves superlinear average and median speedup on the hard application instances of the 2014 qbf gallery .
|
in @xcite , it was argued that , with some very mild assumptions about genericity , we can characterize small field inflation quite simply .
first , it was argued that the effective theory should exhibit an approximate ( global ) supersymmetry in order that there be fields light on the scale of the hubble constant during inflation , @xmath3 .
then , assuming @xmath4 : 1 .
the inflaton is a pseudomodulus , labeling a set of approximate ground states with spontaneously broken supersymmetry .
2 . the effective theory should obey a discrete @xmath5 symmetry in order that the cosmological constant ( c.c . )
be approximately zero at the end of inflation .
3 . at the end of inflation
, the inflaton must couple through relevant or marginal operators to fields which are light with respect to the scale of the energy density during inflation , in order that the cosmological constant be small at the end of inflation .
in particular , it was stressed that inflation typically ends , in the hybrid case , _ before the inflaton reaches the waterfall region_. so - called models of hybrid inflation@xcite have in common the last feature above ; in @xcite it was argued that this full set of conditions should be taken as the definition of hybrid inflation . within such models , these authors noted general features : 1 .
the ( approximate ) goldstino may or may not lie in a multiplet with the inflaton .
2 . the effective theory exhibits an approximate , continuous @xmath5 symmetry .
terms allowed by the discrete symmetry break the accidental continuous global symmetry and spoil inflation , unless the inflationary scale ( the square of the goldstino decay constant ) is sufficiently small .
4 . there are further requirements on the potential in order to obtain slow roll inflation with adequate @xmath6-foldings .
this sets an _
irreducible _ minimum amount of fine tuning necessary to achieve acceptable inflation .
this tuning grows in severity with the number of hubble mass fields .
5 . in order that inflation ends with small c.c . , the inflaton must couple , as noted above , to other light degrees of freedom , or must have appreciable self - couplings in the final ground state .
the coupling to this extra field , or the self couplings , are fixed by the density perturbations @xmath7 and the inflationary scale . in the case of extra fields ,
the resulting structure is necessarily what is called
hybrid inflation"@xcite .
the spectral index , quite generally , is less than one . in @xcite , it was noted that for a broad range of parameters , @xmath0 was typical ; this is widely considered a general result of hybrid models .
recently , considering the planck cmb temperature data supplemented by the wmap large - scale polarization data , the planck collaboration has reported a value @xcite : n_s = 0.9603 0.0073 . and
indeed , the authors of the planck papers argued that their data excludes hybrid inflation . within the definition outlined above ,
it is interesting to look more carefully at the range of allowed values of @xmath8 . in this paper , we systematically consider various planck scale corrections to the simplest version of hybrid inflation .
we explain why ( parametrically ) the most important are the quartic corrections to the potential , and certain power law corrections to the superpotential .
the former must be suppressed by an amount of order @xmath9 , where @xmath10 is the number of @xmath6-foldings . the latter lead to an approximately zero c.c . , supersymmetric minimum for large fields ; in turn this means that the potential has a local maximum ( saddle ) .
this gives rise to a variant of
hilltop inflation"@xcite ; we will see that the initial conditions need not be substantially tuned in order that one obtain adequate @xmath6-foldings and @xmath11 .
if the superpotential has coefficient scaled by a suitable power of @xmath12 and a dimensionless coefficient of order one , one obtains a _ prediction _ of the scale of inflation .
the scale depends on the index @xmath13 of a @xmath14 @xmath5 symmetry , and ranges from about @xmath15 gev to @xmath16 gev . in the next section ,
we review the simplest hybrid model , and recall the prediction @xmath0 . in section [ hierarchy ] , we classify the various planck scale corrections to the simplest hybrid model . in section [ wr ] ,
we consider the implications of the leading superpotential corrections for inflation , explaining why one obtains the structure of hilltop inflation . in section
[ numerical ] , we present numerical results for these models . in section [ susybreaking ]
, we suggest that predictions might arise if inflation is connected with supersymmetry breaking . in section [ conclusions ] , we conclude by considering possible observable consequences of this picture .
the simplest model of hybrid inflation contains two chiral superfields , @xmath17 and @xmath18 , with superpotential w = s(^2 - ^2 ) .
[ simplestmodel ] if one imposes , as is usually done , a _ continuous _
@xmath5 symmetry under which the charges of @xmath17 and @xmath18 are respectively 2 and zero , this superpotential is the most general permitted by symmetries .
classically , the theory has a moduli space , s ^2 > | ^2 | , on which v = v_0 = ^4 . at one loop , the potential receives corrections . in the global limit : v = v_0 ( 1 + ^2 8 ^2 s ) .
[ quantum ] if one considers only this term , one has , for the number of @xmath6-foldings : = 1 2 8 ^2 ^2 s ^2 . in this simple model ,
the @xmath19 parameter is negligible , and = -1 2 n. this yields n_s = 1- 1 n. this is the origin of the prediction that @xmath20 . in this model , @xmath21
is related to @xmath22 by the fluctuation spectrum : = 0.17 ( 10 ^ 15 gev ) ^2 = 7.1 10 ^ 5 ( m_p ) ^2 .
[ kappaequation ]
this treatment , however , is oversimplified . already , in @xcite ,
the role of higher order terms in the potential was considered . more recently , in @xcite , the effects of a linear term in the potential for @xmath17 , arising from the constant term in the superpotential ( needed to account for the small cosmological constant of the present universe ) has been considered . in @xcite , this particular contribution
was treated as small , but a number of other effects were considered .
so it is first worthwhile to consider the various possible corrections in powers of @xmath23 , and their relative importance .
first , it is generally believed that theories of gravity should not exhibit continuous global symmetries ; in string theories , this is a theorem .
replacing the continuous @xmath5 symmetry by a discrete @xmath14 symmetry allows corrections of the form w_r= 2(n+1 ) s^n+1 m_p^n-2 .
more generally , our viewpoint will be that all terms allowed in the effective action below @xmath12 should appear with order one coefficients ; we will assume that smaller coefficients represent a fine tuning " of parameters .
we can systematically consider types of corrections , ordered in powers of @xmath23 : 1 .
potential corrections : @xmath24 , @xmath25 .
superpotential corrections : in addition to @xmath26 ( and higher powers of @xmath17 , other fields ) , at some level there must be a constant in the superpotential , @xmath27 , to account for the smallness of the cosmological constant now .
3 . supersymmetry breaking effects . the term k = m_p^2(s^s)^2 has been noted already in @xcite . in @xcite ,
precise limits on @xmath28 ( of order @xmath9 , where @xmath10 is the number of @xmath6-foldings ) were discussed .
it was noted that the quantum corrections of eqn . only dominate over this potential correction for sufficiently small @xmath17 .
in fact , as we will review shortly , for the simplest model , the quantum corrections _ never _ dominate unless @xmath22 is quite small .
terms of sixth order or higher in @xmath17 in the potential are irrelevant .
they lead to highly suppressed contributions to @xmath29 and @xmath19 , for example
. we will be more quantitative about this question when we turn to models that can reproduce the planck value of @xmath8 .
now we turn to the various superpotential corrections .
our definition of hybrid inflation is motivated by the hypothesis that the scale of inflation is large compare to scales of supersymmetry breaking .
this means , in particular , that m_3/2=w_0 m_p^2 h_i with @xmath30 the hubble scale during inflation . as a result ,
terms in the potential arising from @xmath27 can be neglected during inflation .
if , in fact , the actual value of @xmath31 is comparable to @xmath3 , then this term , and terms associated with supersymmetry breaking , would be important .
even for @xmath32 tev , this corresponds to an inflationary energy scale well below @xmath33 gev .
so finally we turn , again , to @xmath26 .
the presence of this term in the superpotential gives rise to a supersymmetric minimum of the potential at @xmath17 large but parametrically smaller than @xmath12 .
this is unlike the case , for example , of higher order corrections to the potential . as a result
, this term qualitatively alters the behavior of the system , for large but not planck scale fields . in @xcite this term was used to constrain features of inflation . requiring that it was not important during inflation constrained the scale of inflation , and lead to the prediction @xmath20 . to be compatible with the results from planck , however , it is clearly necessary that inflation occur in a region near the local maximum ( as in
hilltop inflation"@xcite )
. we will explore this in the next section .
including @xmath26 , it is first important that the system not flow towards the supersymmetric minimum . indeed , for an intermediate range of field values , there are corrections to the potential of the form v_r = ^2 s^n m_p^n-2 + c.c . for negative @xmath34 ,
this leads to a maximum , for s^n m_p^n-2 . to obtain suitable inflation ,
it is necessary that @xmath17 be smaller than this at the beginning .
but , given eqn .
, except for very large @xmath13 , @xmath17 is smaller than the waterfall value " , s_w = . as a result of these considerations ,
the simplest ( and rather standard ) model of hybrid inflation ( allowing for @xmath26 ) does not appear suitable . in @xcite ,
a simple modification was suggested with two fields , @xmath17 and @xmath35 , with couplings at the renormalizable level : w = s(^2 - ^2 ) + i ^+ [ modifiedhybrid ] the theory , classically , has two flat directions , one with large @xmath17 , one with large @xmath35 . as in the previous model , in order that inflation occur , the potential must be tuned so that at least one of the fields @xmath17 or @xmath35 , has mass small compared to the hubble constant during inflation , @xmath30 . to obtain a workable model , we require that @xmath35 be the light field
. this amounts to requiring that in the potential term k = @xmath28 should be close to unity .
the waterfall regime is now at smaller value of the inflaton field @xmath35 , @xmath36 , and hybrid inflation can be driven by the quantum and discrete symmetry corrections . assuming a discrete @xmath5 symmetry , there are a variety of possible higher dimension terms which might appear in @xmath37 depending on the transformation properties of the fields .
we will consider a term of the form w = - . [ wri ] alternatively , a term proportional to @xmath38 , for example , corrects the potential for @xmath35 if there is a term in the potential k = .
the allowed values of @xmath39 depend on the discrete charge assignments of the fields .
if @xmath39 is not too large , its effects are dramatic .
such terms , again , lead to a supersymmetric minimum of the potential at large @xmath35 ( with @xmath40 ) , and again give rise , for positive @xmath41 , to a _ maximum _ of the potential for @xmath35 at field strength generically large compared to @xmath22 but small compared to @xmath12 . proceeding as before , using the superpotential and corrections in eqns .
, we can compute the number of @xmath6-foldings and the slow roll parameters ( and hence @xmath8 ) .
the potential for @xmath35 is now , approximately , [ potn ] v(i ) = ^4 ( 1 + ^2 16 ^2 ( i^i ) - ( -1))-^2 m_p^2 ( i m_p ) ^n + c.c .. the fluctuation spectrum relates @xmath21 and @xmath22 , as before . for a given value of @xmath22 , the initial value of the field at @xmath42 @xmath6-foldings is fixed .
so , then , is @xmath8 . to get a rough sense of scalings
, we can suppose that @xmath35 starts very near the maximum of the potential , and that @xmath43 ( in order to achieve @xmath44 ) . because @xmath45 at the hilltop , we will simply use the formula for normal hybrid inflation in our estimate ; shortly we will check the accuracy of this numerically , and see that this leads to an order one error .
then one finds that = ( ( .02 n)^n(6.4 10 ^ 9)^2-n ( n)^-2 ) ^1 4n -12 . for particular values of @xmath13 , we can compute @xmath22 and @xmath21 : taking @xmath46 and @xmath47 , this gives @xmath48 gev and @xmath49 .
for @xmath50 , one obtains @xmath51 gev , and @xmath52 .
the scale @xmath22 grows slowly with @xmath13 , reaching @xmath53 gev at @xmath54 and @xmath16 gev for @xmath55 . in general ,
these results scale with @xmath41 as : ^-1 2(n-3).we discuss numerical studies of this problem in the next section .
but the lesson here is that , for fixed values of @xmath41 , and for a given @xmath13 , the scale of inflation , @xmath22 , is fixed to a narrow range .
denoting the real part of the field @xmath35 by @xmath56 , the potential in eqn . becomes [ potnsigma ] v ( ) = ^4 ( 1 + ^2 16 ^2 ( ^2 ) - ( -1))-^2 m_p^2 ( m_p ) ^n , where we have included in @xmath41 the numerical factor @xmath57 coming from the field redefinition .
it will be handy to denote the hilltop position by @xmath58 , and to investigate how close @xmath56 has to be to @xmath58 in order to successfully have @xmath595060 @xmath6-foldings of inflation . for a given @xmath13 ,
the parameters of the two field model are readily enumerated : @xmath22 , @xmath21 , @xmath28 , and @xmath41 . given knowledge of these , we can compute the observable predictions of the inflationary model , to be compared with the planck collaboration results @xcite : 1 .
the number of @xmath6-foldings @xmath60 . to solve the horizon and flatness problems
, it must be @xmath61 . in our numerical treatment
, we will assume the range of @xmath595060 @xmath6-foldings .
2 . the slow roll parameters @xmath62 , which result in the spectral index @xmath63 .
the measured value by the planck collaboration is @xmath64 3 .
the density perturbation spectrum @xmath7 , whose amplitude is a function of @xmath65 .
planck measurements translate to @xmath66 .
we can , in principle , compute the tensor to scalar ratio @xmath67 , but in all such models this will be unobservably small . in general , as said in the previous section , @xmath68 quantifies the correction independent from the discrete symmetry , and is already required to be small , while the dependence on @xmath41 is weak . in the following
we will set @xmath69 , @xmath70 .
given the potential , the expression for the number of @xmath6-foldings @xmath60 involves an integral that can be computed numerically . with a @xmath71 analysis , for each given @xmath13 , we set the three remaining parameters @xmath72 by fitting the experimental values of @xmath73 .
for example , in fig .
[ n4allcontours ] , where we set @xmath47 and @xmath21 to its best - fit value , we show how the allowed ranges for each experimental quantity intersect at specific values of @xmath22 and @xmath56 .
( dashed red ) , the density perturbation @xmath65 ( solid black ) and the number of @xmath6-foldings @xmath60 ( solid blue ) , for @xmath47 .
the coupling @xmath21 is kept fixed at its best fit value of @xmath74 .
the shaded zones indicate the 1-sigma regions allowed by the planck results for @xmath8 and @xmath65 , and the range of 5060 @xmath6-foldings .
the @xmath71 is minimized where the bands intersect each other . for each value of @xmath21 a specific range of @xmath22 and @xmath56
is allowed .
as @xmath21 varies , each variable changes independently and the allowed region moves and shrinks , until the three bands do not intersect .
, height=377 ] in table [ numtable ] , we give the best - fit values of @xmath75 , with the corresponding uncertainties for @xmath13 from 4 to 12 . there is no fine - tuning associated with the inflaton being close to the hilltop value , as the allowed values for @xmath76 are in the range @xmath77@xmath78 . for small @xmath13 , the coupling @xmath21 is tuned to be small . in the last column , we show how close to unity @xmath28 has to be for the correction not to overcome the discrete symmetry correction . as @xmath79 is already fine - tuned to be of order @xmath80 in order not to spoil inflation , we conclude that there is another mild tuning which operates to keep @xmath28 close to 1 .
@xmath81 for @xmath55 , the initial value of the field is @xmath82 gev , just a factor of 10 below @xmath12 . for larger @xmath13 ,
it is not possible to accomodate @xmath44 within the framework of small field inflation . even for this large value of the field ,
the tensor - to - scalar ratio is predicted to be small : r=0.12 210 ^ -6
the picture of small field inflation we have developed up to now assumes that the scale of inflation is large compared to the scale of supersymmetry breaking , i.e. that @xmath4 .
this is the origin of the requirement that the superpotential should vanish and supersymmetry be unbroken , to a good approximation , at the end of inflation .
but one might consider the possibility that @xmath83 . a higher scale of @xmath31 is suggested by the observed higgs mass and supersymmetry exclusions . in addition , for small values of @xmath13 , we have obtained small values of @xmath84 .
so it is interesting to consider the possibility that the the scale of inflation is comparable to @xmath31 . for example we can modify the models we have studied , to give them an oraifeartaigh like structure , adding to the superpotential of eqn .
a coupling m . provided m^2
> ^2 supersymmetry is broken , in a state with @xmath85 .
it is interesting that in this case , inflation ends without ever passing into a waterfall " regime .
as we have stressed , the so - called waterfall is indeed not the distinguishing feature of hybrid inflation .
a different approach has been pursued in @xcite .
again , it is assumed that the scale of inflation is not too much different than the scale of supersymmetry breaking .
one writes a theory of a single field , @xmath18 , and does not require an unbroken @xmath5 symmetry at the end of inflation .
instead , one assumes that the negative contribution to the cosmological constant arising from the vev of the superpotential is cancelled by some supersymmetry breaking dynamics . to constrain the form of the superpotential ,
one still assumes a discrete @xmath5 symmetry .
it is necessary , as in hybrid inflation , to tune the potential so that the @xmath86 term is small .
the superpotential takes the form : w ( ) = v^2 - g n+1 ^n+1 , while the quartic term in the potential must be quite small .
the resulting model is of the hilltop type .
the potential exhibits a local maximum at the origin , and the initial value of the field must lie quite close to the maximum ( compared to the distance of the origin from the minimum ) . inflation occurs in a region very close to the origin in field space ( defined by an unbroken @xmath5 symmetry ) .
the field then settles into a minimum with small cosmological constant and broken supersymmetry and @xmath5 symmetry .
the model can produce the requisite number of @xmath6-foldings and fluctuation spectrum , without introducing an extremely small number analogous to @xmath21 of eqn . .
however , it predicts too _ small _ a value of @xmath8 , n_s = 0.94 . to obtain a spectral index consistent with planck ,
it is necessary to introduce a small and well - tuned constant in the superpotential , which the authors denote @xmath87 , and is of order @xmath88 ( in planck units ) .
there are other issues , such as a possible gravitino problem and overproduction of dark matter , but these can readily be solved by introducing additional matter coupled to the inflaton
. both approaches , then , seem viable , and have the potential to relate supersymmetry breaking dynamics to inflationary dynamics .
each requires certain tunings .
the results from planck pose challenges for models of small field inflation .
it has been said that they rule out
hybrid inflation . " here , following @xcite , we have carefully defined models of hybrid inflation as models in which inflation occurs on a pseudomoduli space , with supersymmetry and an @xmath5 symmetry approximately restored at the end of inflation . we have assumed a _ discrete _ @xmath5 symmetry , and have considered the importance of corrections to the superpotential and potential . for initial values of the field far from the local maximum of the potential , one predicts a spectral index inconsistent with planck . to obtain @xmath89 ,
it is necessary that the field start near the local maximum , though this condition is not severely tuned . for @xmath14 symmetry with @xmath47 ,
the scale of inflation is rather low , and we considered the possibility that @xmath90 . in this case , the dynamics of inflation might be closely tied to the scale of supersymmetry breaking , and there is some chance that aspects of the physics of inflation could be studied in accelerator experiments . we have noted that , in this case , the assumption of an unbroken @xmath5 symmetry and unbroken supersymmetry at the end of inflation might be relaxed , and compared the hybrid models with those of @xcite .
each of these models can reproduce the data , and involves very small parameters and tunings .
the fact that many models with such features can reproduce the basic data of inflation raises , as always , the question of whether there is any way they might be testable or falsifiable .
we would argue that the best hope is connecting inflation with the dynamics responsible for supersymmetry breaking
. it will be particularly interesting to explore dynamical supersymmetry breaking ( and generation of scales ) in this framework .
|
it is sometimes stated that @xmath0 in hybrid inflation ; sometimes that it predicts @xmath1 . a number of authors have consider aspects of planck scale corrections and argued that they affect these predictions .
here we consider these systematically , describing the situations which can yield @xmath2 , and the extent to which this result requires additional tuning .
scipp 13/15 + 1.2 cm * small field inflation and the spectral index * 1.4 cm milton bose , michael dine , angelo monteux and laurel stephenson haskins + 0.4 cm _
santa cruz institute for particle physics and + department of physics , santa cruz ca 95064 _ + 1.5 cm 1.0 cm
|
the relation between quantum spectra , on the one hand , and the dynamics of the corresponding classical systems , on the other , is a problem of fundamental importance in physics , which has attracted attention ever since the early days of quantum mechanics .
a key to understanding this relation in chaotic systems was gutzwiller s discovery ( cf .
@xcite ) that the semiclassical density of states in these systems can be written as a sum over all ( isolated ) periodic orbits of the classical system .
however , practical applications of gutzwiller s trace formula , like other periodic orbit sums , are greatly impeded by the fact that the sums usually diverge in the domain where the physical eigenvalues are located , mainly as a consequence of the rapid proliferation of periodic orbits with growing period .
various techniques were developed over the years to overcome the convergence problem of the periodic orbit sums @xcite , but most of them depend on special properties of the systems , such as ergodicity or the existence of complete symbolic dynamics . as an alternative
, the technique of harmonic inversion of semiclassical trace formulae has been introduced as a very general method for extracting eigenvalues from periodic orbit sums for both chaotic and regular systems @xcite .
harmonic inversion circumvents the convergence problems , and allows one to obtain the semiclassical eigenvalues from a relatively small number of periodic orbits .
moreover , the method does not require any special properties of the systems , and therefore is applicable also to the particularly challenging case of systems with strong pruning , i.e. , systems with a highly incomplete symbolic coding of the periodic orbits . in this paper
we demonstrate the power of the harmonic inversion technique in the semiclassical quantization of the closed three - disk billiard system , for which all other semiclassical quantization methods to date have failed in determining more than the few lowest eigenvalues , chiefly because of the extremely rapid proliferation of periodic orbits with increasing action and the strong pruning typical of this system . from a technical point of view
our study will show that harmonic inversion is numerically stable even when handling huge ( on the order of several million ) periodic orbit sets .
the three - disk billiard system consists of three equally spaced hard disks of unit radius .
the existence or nonexistence of periodic orbits and the behavior of the periodic orbit parameters in the three - disk system sensitively depend on the distance @xmath0 between the centers of the disks .
large disk separations , especially @xmath1 , have served as a test case for periodic orbit quantization of chaotic systems in many investigations in recent years . in particular , the system has been employed as prototype example for the usefulness of periodic orbit quantization by cycle expansion techniques @xcite . for large disk separations , the assumptions of the cycle expansion , namely that the contributions from long orbits are shadowed by those of short orbits , are almost exactly fulfilled .
this is no longer true for small disk separations .
as the disks approach each other , the convergence of the cycle expansion becomes slower and slower , until it finally breaks down .
periodic orbit quantization by harmonic inversion does not depend on specific properties , such as shadowing of orbits etc .
, and therefore is expected to work well also for small disk separations .
in fact , the harmonic inversion technique has successfully been applied to the `` standard '' literature disk separation @xmath1 as well as to the small distance @xmath2 , where the convergence of the cycle expansion is already slow @xcite . in this paper
we concentrate on the limiting case of mutually touching disks , @xmath3 , where the system turns into a bound system , and the requirements of the cycle expansion and other semiclassical methods are no longer fulfilled at all . for the closed three - disk system ,
the lowest exact quantum eigenvalues were first calculated by tanner et al .
@xcite and by scherer @xcite .
a remarkable step towards the semiclassical quantization of the closed three - disk system was achieved by combining the conventional cycle expansion with a functional equation @xcite .
based on periodic orbit corrections to the mean density of states , approximations to the very lowest eigenvalues could be obtained from a small set of periodic orbits up to cycle length @xmath4 , where the symbolic dynamics is still complete ( including formally a zero length orbit ) . for the very lowest eigenvalues ,
the results were found to be in good agreement with the exact quantum values . however , the method failed for higher eigenvalues , where only the mean density of states could be reproduced . since the problems with the method arose from the strong pruning of orbits in this system , they are of a fundamental nature , and
could not be overcome by including more orbits .
no other semiclassical method has been successful in obtaining higher eigenvalues in the closed three - disk billiard system yet . in this paper
, we will demonstrate that harmonic inversion passes even this acid test of semiclassical quantization . in sec .
[ 3disk - po+qm ] we discuss the construction of the periodic orbit signal for the closed three - disk system , which is a nontrivial task because of , firstly , the extremely rapid proliferation of periodic orbits and , secondly , the strong pruning of the symbolic dynamics .
once the signal has been constructed , we refer in sec .
[ hi ] to the harmonic inversion technique for the extraction of the semiclassical eigenvalues .
the numerical recipe of the harmonic inversion method has already been established in the literature @xcite , however , to make the article self - consistent we recapitulate the important steps in sec .
[ method ] .
the semiclassical results are discussed in secs .
[ d=2_hi ] and [ d=2_crosscorr ] .
we start by summarizing the basic properties of the three - disk system relevant for periodic orbit quantization .
furthermore , the pruning of orbits and the distribution of the orbit parameters are investigated , and the resulting strategies for the numerical search for periodic orbits are discussed .
the periodic orbits of the three - disk system can be labeled by a ternary symbolic code , which is complete for sufficiently large disk separations .
the classical dynamics of the three - disk system has been studied , among others , by gaspard and rice @xcite . if the disks are labeled by the numbers 1 , 2 , 3 , each periodic orbit is characterized by a sequence of these numbers , indicating the disks the particle collides with during one period of the orbit . for sufficiently separated disks
, there is a one - to - one correspondence between the symbolic code and the periodic orbits of the system : for every sequence there exists one unique periodic orbit ( with the restriction that consecutive repetitions of the same symbol are forbidden and circular shifts of a sequence describe the same orbit ) . as was first pointed out by hansen @xcite , at disk separation @xmath5
`` pruning '' sets in , i.e. , part of the periodic orbits become unphysical as they begin to run through one of the disks , and thus the symbolic code is no longer complete .
the number of pruned orbits rapidly becomes larger and larger if the disks continue to approach each other . in the limiting case of touching disks , @xmath3 ,
the system exhibits strong pruning .
examples of pruned orbits will be discussed in section [ numsearch ] . in our calculations , we make use of the symmetry reduction introduced by cvitanovi and eckhardt in ref .
@xcite : the three - disk system is invariant under the symmetry operations of the group @xmath6 , i.e. , reflections at three symmetry lines and rotations by @xmath7 and @xmath8 .
the periodic orbits fall into three classes of distinct symmetry : orbits invariant under reflections at one of the symmetry lines ( multiplicity 3 ) , orbits invariant under rotations by @xmath7 and @xmath8 ( multiplicity 2 ) , and orbits with no symmetry ( multiplicity 6 ) .
the quantum states are grouped in the three irreducible subspaces @xmath9 , @xmath10 , and @xmath11 , where the states of the @xmath9 ( @xmath10 ) subspace are symmetric ( antisymmetric ) under reflection at the symmetry lines , respectively , and the states of the @xmath11 subspace are invariant under rotations by @xmath7 and @xmath8 . following refs .
@xcite , one can map the system onto a fundamental domain , which consists of a one - sixth slice of the full system , with the symmetry axes acting as straight mirror walls ( see fig .
[ fig1 ] ) .
the periodic orbits of the full system can be described completely in terms of the periodic orbits in the fundamental domain .
the symmetry reduced periodic orbits are labeled by a binary code , where the symbol ` 0 ' represents backscattering ( or change between clockwise and anti - clockwise scattering ) and the symbol ` 1 ' stands for scattering to the third disk in the original full domain picture . for example , the shortest full domain orbit 12 maps onto the 0 orbit , and the 123 orbit maps onto the 1 orbit .
the quantum resonances of the three different subspaces @xmath9 , @xmath10 and @xmath11 can be obtained separately from the periodic orbits in the fundamental domain by introducing appropriate weight factors for the orbits in gutzwiller s trace formula , as was shown by cvitanovi and eckhardt @xcite .
we concentrate on the @xmath9 subspace , for which each orbit has a weight factor equal to 1 . as for all billiard systems ,
the shape of the periodic orbits is independent of the wave number @xmath12 , and the action scales as @xmath13 where the scaled action @xmath14 is equal to the physical length of the orbit .
we consider the density of states as a function of the wave number @xmath15 with a scaled response function @xmath16 . applying the gutzwiller trace formula to the closed three - disk system yields for the @xmath9 subspace @xmath17 where @xmath18 and @xmath19 are the monodromy matrix and the maslov index of the orbit , respectively ,
@xmath20 is the symbol length , @xmath21 is the scaled action , and @xmath22 denotes the expanding stability eigenvalue of the orbit ( i.e. , the eigenvalue with an absolute value larger than one ) .
the sum runs over all symmetry reduced periodic orbits including multiple traversals .
here , @xmath23 denotes the repetition number with respect to the corresponding primitive orbit . in practice ,
only the primitive periodic orbits have to be determined .
the parameters of the @xmath24 repetition of the primitive orbit ( here characterized by the index 0 ) are then given by @xmath25 , @xmath26 and @xmath27 .
for extracting the quantum resonances of the system from eq .
( [ gw3disk ] ) by harmonic inversion , all periodic orbits up to a maximum scaled action have to be included .
the parameters of the periodic orbits scaled action and stability eigenvalues have to be determined numerically .
we calculate the primitive periodic orbits using the symbolic code as input . for simplicity ,
the calculations are carried out in the full domain , and the results are then translated back into the symmetry reduced system .
the disks are `` connected '' according to the code , starting with arbitrary reflection points on the disks as initial condition .
the reflection points are then varied in such a way that the total length ( i.e. , the action ) of the orbit reaches a minimum .
all orbit parameters can then be calculated from the reflection points .
the scaled action is given by the length of the ( symmetry reduced ) orbit , and the stability eigenvalue @xmath28 can be determined by an algorithm proposed by bogomolny @xcite .
in addition to action and stability , one can also determine averages of different classical quantities ( distance from the center of the system , angular momentum , etc . ) , which are needed for the cross - correlation technique ( see section [ hi ] ) . for disk separations smaller than the pruning limit @xmath5
, it has to be checked whether or not the orbits are physical , i.e. , whether or not they stay completely outside the disks . in our numerical calculations , we found two different classes of pruned orbits .
an example of the first class is given in figure [ fig2 ] : as the disks approach each other , a section of the orbit connecting two of the disks gets inside the third disk .
the pruned orbit still corresponds to a unique minimum of the total length when the reflection points on the disks are varied for given symbolic code .
this is not the case for the second class , an example of which is shown in figure [ fig3 ] : in this case , as the disks come closer , the reflection angle at one specific reflection point approaches @xmath29 . as the value @xmath29
is reached , the minimum of the action with respect to this specific reflection point splits into two equal minima and one local maximum .
accordingly , the orbit splits into three unphysical orbits with the same symbolic code . in one case , the orbit is now reflected on the inside of the disk ( see fig . [
fig3]c ) , which corresponds to a reflection angle larger than @xmath29 ( i.e. , the orbit has penetrated into the disk in the vicinity of the reflection point ) .
this orbit does not correspond to a minimum of the total length but to a saddle point . in the other two cases , figs .
[ fig3]d and [ fig3]e , the reflection point moves along the disk in such a way that the reflection law is no longer fulfilled but the orbit just passes the `` reflection point '' in a straight line and penetrates into the disk .
these two cases correspond to two equally deep minima of the total length .
the two orbits have the same shape but only differ as to which penetration point is considered as `` reflection point '' , indicated by the squares in figs .
[ fig3]d and [ fig3]e .
in fact , the shape of the orbit is equal to that of a pruned orbit with a ( symmetry reduced ) symbol length that is shorter by 1 and which belongs to the first class of pruned orbits as defined above .
( in this orbit the reflection point in question is simply missing , i.e. , none of the penetration points is considered as reflection point . ) besides the existence or nonexistence of periodic orbits , the distance between the disks also strongly influences the distribution of the periodic orbit parameters .
this is illustrated in figure [ fig4 ] , which compares the orbit parameters of the shortest primitive periodic orbits found for the open three - disk system with disk separation @xmath1 and for the case of touching disks , @xmath3 .
for @xmath1 , the action and stability of the orbits are mainly determined by the symbol length , the orbit parameters are nicely bunched in easily distinguishable intervals , and the number of orbits up to a given action is relatively small . however , the picture changes completely when the disks approach each other : as all orbits become shorter , the total number of orbits up to a given action increases rapidly .
the parameters of the orbits are no longer simply determined by the symbol length , and their overall behavior becomes much more complicated , as is evident from figure [ fig4 ] . for @xmath3 , the orbits can be grouped in `` channels '' with the same `` tail '' ( end figures ) but growing number of leading 0 s in the code .
( a sequence of @xmath30 leading 0 s in the code will be denoted by @xmath31 in the following ) .
these orbits have the same basic shape but run deeper and deeper into the corner formed by two touching disks , bouncing back and forth between the two disks ( see fig . [ fig5 ] ) . in each channel
, the action of the orbits grows very slowly with increasing symbol length , while the expanding stability eigenvalue increases relatively fast ( in fact exponentially ) with every additional collision . since adding a leading ` 0 ' to the symbolic code
does not change the action considerably , there is a huge number of orbits with very long symbol lengths but relatively small , and comparable , actions .
in fact , most channels break off because of pruning . e.g. , the @xmath32 series , labeled ( a ) in fig . [ fig4]b , breaks off after six orbits , and the @xmath33 series , marked ( b ) , already breaks off after three orbits .
our analysis strongly suggests that there exist only two infinite channels : the @xmath34 and the @xmath35 series , dubbed ( c ) in fig .
[ fig4]b ( these channels are only shown up to 20 leading 0 s in the code ) .
the shapes and actions of these orbits converge very slowly towards the shape and action of the limiting orbit which starts exactly in the touching point of two disks and is directly reflected back by the opposite disk ( cf .
[ fig5 ] ) . although all other channels break off because of pruning , the number of orbits in a channel up to a given action often becomes very large ( or , in the two cases mentioned , even infinite ) .
note that the 0 orbit , bouncing back and forth between two disks , does not exist in the closed three - disk system .
while for large disk separations finding all orbits up to a given action simply corresponds to calculating all orbits up to a certain maximum symbol length , the complicated distribution of periodic orbit parameters in the closed three - disk system renders the search for the relevant periodic orbits up to a given action a nontrivial task . due to the extremely rapid proliferation of periodic orbits with increasing length @xmath14 , we were not able to determine all orbits up to a given length @xmath36 , but we had to introduce restrictions for the orbits to be included . as the amplitude in the gutzwiller formula depends on the inverse of the stability parameter ,
a reasonable cut - off criterion is the size of the stability of the orbit . on the other hand , since in each channel the absolute value of the expanding stability eigenvalue grows exponentially fast , the main contributions of each channel to the periodic orbit sum come from the orbits with small symbol lengths .
furthermore , the maslov indices of adjacent orbits in each channel differ by two , and therefore the periodic orbit amplitudes alternate in sign , which leads to approximate cancellations of terms in the semiclassical trace formula . as a second criterion
, we therefore restricted the number of consecutive symbols ` 0 ' in the symbolic code , thus selecting the most relevant orbits of each channel .
the orbits were calculated channel by channel , starting from the shortest orbit in the channel and adding more and more leading zeros to the code .
the calculation of the channel was broken off if one of the following conditions was fulfilled : * the action exceeds the maximum action @xmath36 .
* the absolute value of the stability eigenvalue exceeds the given value @xmath37 . * the maximum number of consecutive symbols ` 0 ' is reached . *
the orbits become pruned . for the set of periodic orbits shown in figure [ fig4]b , we searched for orbits with physical length @xmath38 . the maximum for the absolute value of the stability eigenvalue
was chosen to be @xmath39 .
the maximum number of successive zeros in the code was restricted to 20 for orbits with length @xmath40 and to 12 for orbits with @xmath41 .
the fundamental problem , and challenge , now is to extract semiclassical eigenvalues in a numerically stable way from the huge periodic orbit set presented in fig .
[ fig4]b . to this end
, we resort to harmonic inversion of periodic orbit signals @xcite . here , we review the basic ideas and present an extension of the method introduced in @xcite to harmonic inversion of cross - correlated periodic orbit sums . the starting point is to introduce a weighted density of states in terms of @xmath42 @xmath43 with @xmath44 where @xmath45 is the eigenvalue of the wave number of eigenstate @xmath46 and @xmath47 are the diagonal matrix elements of a chosen set of @xmath48 linearly independent operators @xmath49 , @xmath50 .
the fourier transform of ( [ g_ab_qm ] ) yields a @xmath51 cross - correlated signal @xmath52 a semiclassical approximation to the weighted quantum response function ( [ g_ab_qm ] ) is given by the semiclassical response function ( [ scgutz ] ) weighted with the classical averages @xmath53 with @xmath54 the wigner transform of the operator @xmath49 @xcite . by fourier transformation
we obtain the semiclassical approximation to the quantum signal ( [ c_ab_qm ] ) , which reads @xmath55 where @xmath23 is the repetition number counting the traversals of the primitive orbit , and @xmath18 and @xmath19 are the monodromy matrix and maslov index of the orbit , respectively .
semiclassical approximations to the eigenvalues @xmath45 , and eventually also to the diagonal matrix elements @xmath56 , are obtained by adjusting the semiclassical cross - correlated periodic orbit signal ( [ c_ab_sc ] ) to the functional form of the quantum signal ( [ c_ab_qm ] ) .
this can be done with the filter - diagonalization method for the signal processing of cross - correlation functions @xcite . here , we resort to an alternative method introduced in ref .
@xcite , which will now be extended for the harmonic inversion of cross - correlated periodic orbit sums . the special form of the periodic orbit signal ( [ c_ab_sc ] ) as a sum of @xmath57 functions allows for an even simpler procedure than filter - diagonalization , viz .
analytical filtering . in the following
we will apply a rectangular filter , i.e. , @xmath58 for frequencies @xmath59 $ ] , and @xmath60 outside the window . applying this filter to the semiclassical signal @xmath61 in ( [ c_ab_sc ] )
we obtain the band - limited ( bl ) cross - correlated periodic orbit signal , @xmath62}\over \pi(s - s_{\rm po } ) } { \rm e}^{{\rm i}s_{\rm po}k_0 } \ ; . \label{c_sc_bl}\end{aligned}\ ] ] the introduction of @xmath63 into the arguments of the exponential functions in ( [ c_sc_bl ] ) causes a shift of frequencies by @xmath64 in the frequency domain .
note that @xmath65 is a smooth function and can be easily evaluated on an arbitrary grid of points @xmath66 provided the periodic orbit data are known for the set of orbits with classical action @xmath67 .
applying now the same filter as used for the semiclassical periodic orbit signal to the quantum one , we obtain the band - limited quantum signal @xmath68 in contrast to the signal @xmath69 in eq .
( [ c_ab_qm ] ) , the band - limited quantum signal consists of a _
finite _ number of frequencies @xmath45 , @xmath70 , where in practical applications @xmath71 can be of the order of @xmath72 ( 50 - 200 ) for an appropriately chosen frequency window , @xmath73 .
the problem of adjusting the band - limited semiclassical signal in eq .
( [ c_sc_bl ] ) to its quantum mechanical analogue in eq .
( [ c_qm_bl ] ) can now be written as a set of nonlinear equations @xmath74 for the unknown variables , viz . the shifted frequencies , @xmath75 , and the parameters @xmath76 , @xmath70 .
note that @xmath71 is related to the signal length @xmath77 and the dimension @xmath48 of the cross - correlation matrix by @xmath78 .
the signal ( [ c_bld ] ) can be evaluated on an equidistant grid , @xmath79 , with relatively large step size @xmath80 , and the discrete signal points @xmath81 are called the `` band - limited '' cross - correlated periodic orbit signal .
we now wish to solve the nonlinear system , eq .
( [ c_bld ] ) , which can be written as @xmath82 where @xmath83 and @xmath76 are , generally complex , variational parameters .
we assume that the number of frequencies in the signal is relatively small ( @xmath84 to @xmath85 ) .
although the system of nonlinear equations is , in general , still ill - conditioned , frequency filtering reduces the number of signal points , and hence the number of equations . in ref .
@xcite three different methods , viz .
linear predictor ( lp ) , pad approximant ( pa ) , and signal diagonalization ( sd ) have been employed to solve eq .
( [ c_n : eq ] ) for a one - dimensional signal , i.e. , @xmath86 .
the sd method can be generalized in a straightforward manner to the harmonic inversion of cross - correlated periodic orbit sums .
the problem of solving the nonlinear set of equations ( [ c_n : eq ] ) can be recast in the form of the generalized eigenvalue problem @xcite @xmath87 where the elements of the @xmath88 operator matrix @xmath89 and overlap matrix @xmath90 depend trivially upon the signal points : @xmath91 the matrices @xmath89 and @xmath90 in eq .
( [ geneval ] ) are complex symmetric ( i.e. , non - hermitian ) , and the eigenvectors @xmath92 are orthogonal with respect to the overlap matrix @xmath90 , @xmath93 where the brackets define a complex symmetric inner product @xmath94 , i.e. , no complex conjugation of either @xmath95 or @xmath96 the overlap matrix @xmath90 is not usually positive definite and therefore the @xmath97 s are , in general complex , normalization parameters .
an eigenvector @xmath92 can not be normalized for @xmath98 .
the parameters @xmath76 in eq .
( [ c_n : eq ] ) are obtained from the eigenvectors @xmath92 via @xmath99 the parameters @xmath100 in eq .
( [ c_n : eq ] ) are given as the eigenvalues of the generalized eigenvalue problem ( [ geneval ] ) , and are simply related to the frequencies @xmath101 in eq .
( [ c_bld ] ) via @xmath102 .
the resolution of the results depends on the signal length . for a one - dimensional signal ,
the method requires a signal length of @xmath103 , with @xmath104 the mean density of states , to resolve the frequencies @xcite .
this means that all periodic orbits up to the scaled action @xmath36 have to be included .
the advantage of using the cross - correlation approach is based on the insight that the total amount of independent information contained in the @xmath51 signal is @xmath105 multiplied by the length of the signal , while the total number of unknowns ( here @xmath76 and @xmath45 ) is @xmath106 times the total number of poles @xmath45 .
therefore the informational content of the @xmath51 signal per unknown parameter is increased ( compared to the one - dimensional signal ) by roughly a factor of @xmath48 , and the cross - correlation approach should lead to a significant improvement of the resolution .
the power of this method for periodic orbit quantization has already been proven for the example of an integrable system @xcite .
we will now demonstrate how the method works in the case of three touching disks , as an example of a chaotic system with extremely rapid proliferation of orbits and strong pruning .
we first investigate a one - dimensional signal ( @xmath107 ) by simply choosing @xmath108 , i.e. , the unity operator .
the signal is constructed from the set of periodic orbits shown in fig .
[ fig4]b , which contains about 5 million primitive orbits .
we analyzed the signal up to length @xmath109 using the decimated signal diagonalization method @xcite .
the frequencies obtained by harmonic inversion of the signal are the semiclassical approximations to the eigenvalues @xmath110 . fig .
[ fig6 ] shows the results from harmonic inversion ( solid lines ) , compared with the exact quantum results ( dashed lines ) , both presented in terms of the energy @xmath111 .
the exact quantum eigenvalues were taken from scherer @xcite and wirzba @xcite .
the harmonic inversion results clearly reproduce the quantum eigenvalues up to @xmath112 with only small deviations . however , the signal length was not sufficient to resolve the eigenvalues in the region @xmath113 , where the density of states with respect to the wave number @xmath42 ( which grows @xmath114 ) becomes too large . apart from the values shown in fig .
[ fig6 ] , the harmonic inversion of the signal also yielded a number of unconverged frequencies .
converged frequencies were identified by having an imaginary part close to zero ( since the eigenvalues of the wave number must be real for bound systems ) and an amplitude close to the theoretical value @xmath115 .
the whole set of frequencies obtained is shown in figure [ fig7 ] together with the real part of their amplitudes .
the solid vertical lines mark the positions of the exact quantum eigenvalues .
the semiclassical values included in fig .
[ fig6 ] are represented by filled circles in fig .
the amplitudes are not as well converged as the frequencies ( i.e. the eigenvalues ) but partly show larger deviations from the theoretical value @xmath115 .
[ this can be explained by the observation that when the decimated signal diagonalization method is used for harmonic inversion the frequencies usually converge faster than the amplitudes .
] however , up to @xmath112 , the eigenvalues can be clearly identified .
the small deviations of the eigenvalues obtained by harmonic inversion from the exact quantum values may have different reasons .
especially for the lowest eigenvalues , the major part is probably due to the semiclassical error . on the other hand ,
as explained above , we did not include all orbits in the signal but left out a large number of relatively unstable orbits .
although each of these orbits only gives a negligible small contribution to the periodic orbit sum , the number of excluded orbits may be so large that their contributions sum up in a way so as to have a visible effect on the density of states .
finally , even in the low - lying part of the spectrum , deviations may arise from the fact that the signal length was very short .
the influence of the missing orbits and the relatively short signal length are reflected in the relatively poor convergence of the amplitudes . in order to test the stability of the results with respect to the signal parameters , and to obtain an estimate of random errors
, we performed the same calculation with various sets of different signal parameters .
figure [ fig8 ] shows the harmonic inversion results for the eigenvalues as a function of the signal length @xmath36 . from all frequencies
obtained the values with @xmath116 and with amplitudes @xmath117 were singled out , respectively .
the filled circles mark amplitudes with @xmath118 .
it can clearly be seen how the maximum eigenvalue up to which the spectrum can be resolved depends on the signal length .
[ fig8 ] shows that the results for the low - lying eigenvalues are quite stable with respect to variation of the signal length .
compared with the results obtained by tanner et al .
@xcite using the extended cycle expansion method , the analysis of a single signal by harmonic inversion has yielded about the same number of eigenvalues for the closed three - disk system . for the resolution of higher eigenvalues ,
the signal length was not sufficient .
extending the signal to significantly larger scaled actions is in practice not possible because of the extremely rapid increase of the number of orbits . it should be stressed , however , that , in contrast to the cycle expansion method , which runs into severe problems because of the pruning of orbits in this system , it is at least _ in principle _ possible to improve the harmonic inversion results by including more orbits . on the other hand , as discussed before , the harmonic inversion method offers the possibility to significantly reduce the signal length required for the resolution of eigenvalues by the construction and analysis of cross - correlated periodic orbit sums .
to increase the accuracy of the semiclassical eigenvalues , we now apply the cross - correlation technique to the closed three - disk system . instead of the one - dimensional signal investigated in the previous section ,
we choose a set of three operators and construct a @xmath119 cross - correlated signal according to eq .
( [ c_ab_sc ] ) . as input
, we again use the set of periodic orbits presented in fig . [ fig4 ] together with the averages of different classical quantities over the orbits .
[ fig9 ] shows the results from the harmonic inversion of a @xmath119 signal of length @xmath120 ( solid lines ) , which was constructed using the operators @xmath121 ( unity ) , @xmath122 and @xmath123 , where @xmath23 and @xmath124 denote the distance from the center of the system and the absolute value of the angular momentum , respectively .
the eigenvalues are again presented in terms of the energy @xmath111 . for comparison ,
the dashed lines indicate the positions of the exact quantum eigenvalues .
the converged eigenvalues for the wave number @xmath42 have been singled out from the whole set of frequencies obtained by the condition that they should have an imaginary part close to zero and an amplitude close to the theoretical value @xmath115 .
the amplitudes and the imaginary parts of all frequencies obtained are shown in figure [ fig10 ] by circles . in particular , the filled circles denote the frequencies included in fig .
[ fig9 ] . in fig .
[ fig10]a , the positions of the exact quantum eigenvalues are given by the solid vertical lines . with the cross - correlated signal , one can now clearly identify eigenvalues up to @xmath125 .
in addition , the convergence of the lowest eigenvalues is improved as compared to the results from the single signal obtained in section [ d=2_hi ] , as can be seen from the amplitudes .
it is not possible to determine to what extent the accuracy of the semiclassical eigenvalues has improved since the results can only be compared with the exact quantum eigenvalues and the size of the semiclassical error is unknown .
again , we have performed the same calculation for various sets of parameters in the harmonic inversion scheme .
figure [ fig11 ] shows the results from a @xmath119 and from a @xmath126 signal as functions of the signal length .
the @xmath119 signal was constructed from the operators @xmath121 ( unity ) , @xmath122 and @xmath123 , and the @xmath126 signal contained the operators @xmath121 , @xmath127 , @xmath122 and @xmath123 . in both diagrams , only the frequencies with an imaginary part @xmath116 and an amplitude @xmath118 were included .
the filled circles mark values with @xmath128 .
the positions of the exact quantum eigenvalues are again marked by the solid vertical lines .
as with the one - dimensional signal analyzed in the previous section , it can be observed that the results for the lowest eigenvalues are very stable with respect to the variation of the signal length as long as the signal is not too short .
however , the very lowest frequencies show a tendency to split into two or more , which was a frequent observation in our calculations when the signal length and the matrix dimensions were chosen too large . the results for the higher eigenvalues depend more sensitively on the signal length as the density of states approaches the limiting resolution that can be achieved with the signal . in this region
, the @xmath126 signal shows a better resolution than the @xmath119 signal .
with the @xmath119 signal and the two largest values of the signal length considered , the lowest eigenvalue was not obtained .
the reason for this probably lies in inaccuracies at the end of the signal , as the signal length approaches the value @xmath129 : in the decimated signal diagonalization scheme , which we used for the harmonic inversion of the signal , the @xmath57 functions in the original signal ( [ c_ab_sc ] ) are replaced with sinc functions @xcite [ @xmath130 .
therefore , every periodic orbit contributes to the signal not only exactly at its scaled action @xmath131 , but also in a range @xmath132 around this point .
if the signal length is close to @xmath129 , also orbits with actions slightly larger than this limit may give non - negligible contributions to the very end of the signal .
since we did only include orbits with @xmath133 , these contributions are missing in our signal .
some eigenvalues between @xmath134 and @xmath135 seem to be particularly hard to obtain from the set of orbits used .
possibly , these eigenvalues are related to the orbits running very deep into the corners between the disks , which were not included in the signal .
we conjecture that the missing orbits as well as the short signal length are again responsible for the deviations especially of the results for higher eigenvalues from the exact quantum values .
we have successfully applied the technique of harmonic inversion to the semiclassical quantization of a system with an extremely rapid proliferation of periodic orbits and strong pruning . with the harmonic inversion of cross - correlated periodic orbit sums
we were able to calculate accurate semiclassical energy eigenvalues of the closed three - disk system up to the region @xmath125 ( a total of 29 levels ) .
this is the first time that such high - lying eigenvalues have been correctly calculated by periodic orbit theory in this system .
our results demonstrate that the harmonic inversion of cross - correlated periodic orbit sums is indeed a powerful method of periodic orbit quantization and allows the stable handling even of huge periodic orbit sets .
in contrast to other methods , which depend on the existence of a complete symbolic code , the harmonic inversion technique is not affected by the strong pruning of orbits . in the closed three - disk system ,
the only restriction for the practical application of the harmonic inversion method lies in the extremely large number of periodic orbits , which is a special feature of this system .
it should also be noted that the semiclassical eigenvalues were obtained without resorting to the mean density of states , which , including all necessary correction terms , is nontrivial to calculate even in the case of the closed three - disk system @xcite .
the present results may stimulate future work on other challenging systems without special structural information on the quantum or classical level .
one example with strong pruning of the symbolic dynamics is the three - dimensional generalization of the closed three - disk billiard , i.e. , a system consisting of four touching spheres at the corners of a regular tetrahedron .
by contrast to the closed three - disk system this is an open system where no functional equation can be applied .
experiments on chaotic light - scattering from the four spheres have recently attracted much attention @xcite .
we thank a. wirzba for communicating to us quantum mechanical data for the closed three - disk system .
this work was supported by deutsche forschungsgemeinschaft and deutscher akademischer austauschdienst .
|
classical chaotic systems with symbolic dynamics but strong pruning present a particular challenge for the application of semiclassical quantization methods . in the present study
we show that the technique of periodic orbit quantization by harmonic inversion of trace formulae , which does not rely on the existence of a complete symbolic dynamics or other specific properties , lends itself ideally to calculating semiclassical eigenvalues from periodic orbit data even in strongly pruned systems . as the number of periodic orbits proliferates exponentially in chaotic systems , we apply the harmonic inversion technique to cross - correlated periodic orbit sums , which allows us to reduce the required number of orbits .
the power of the method is demonstrated for the closed three - disk billiard as a prime example of a classically chaotic bound system with strong pruning .
|
the interpretation of the atmospheric muon neutrino deficit in the superkamiokande is that muon neutrinos oscillate to tau neutrinos with nearly maximal mixing @xcite .
experiments such as icanoe , opera and monolith will be able to test the oscillation hypothesis by searching for tau neutrino conversion to taus @xcite after the accelerator beam of muon neutrinos has passed over the long - baseline between cern and the gran sasso laboratory .
the minos experiment run at lower energies @xcite will look for the characteristic muon neutrino deficit . ultimately , one would like to know the neutrino - nucleon charged current cross section to the level of a few percent for precision measurements of oscillation mixing angles .
mass corrections , from both the out - going charged lepton and from the target nucleon , are important at this level . here , we report on lepton and target mass corrections in the collinear limit to neutrino - nucleon charged current interactions in deep inelastic scattering .
next - to - leading order qcd corrections and charm production are included .
this is a portion of the total cross section which includes quasi - elastic scattering and few pion production at the energies of interest @xcite . at the precision level , an anomaly in the interaction of neutrinos of higher @xmath5 energies
has recently been observed by the fermilab nutev experiment @xcite . for this paschos - wolfenstein - like @xcite measurement of the weak mixing angle @xmath3
, deep inelastic scattering is the dominant interaction channel .
nlo qcd , charm mass ( @xmath6 ) and target mass corrections are potentially important ingredients in the interpretation of the experimental result . in the next section ,
we review the deep inelastic scattering ( dis ) formalism , concentrating on charged current scattering ( cc ) where two additional structure functions @xmath7 and @xmath8 come into play for energies not too large compared to the outgoing lepton mass .
we show our results for the structure functions with nlo qcd corrections , and @xmath6 and nucleon mass @xmath9 corrections in the collinear limit , in section 3 .
our cross section results for @xmath1 cc scattering with isoscalar nucleons @xmath10 and @xmath11 cc scattering are shown in section 4 .
conclusions appear in section 5 .
more complete references and further details of this work appear in ref .
@xcite , and extensions to include neutral current interactions as well as target mass corrections through the operator product expansion approach of georgi and politzer @xcite , through nlo qcd , will be presented in ref .
the qcd dynamical part of the evaluation of neutrino interactions is given through the hadronic tensor @xmath12 sandwiched between the current operators in the first line of eq .
[ eq : wmunu0 ] , an inclusive sum over a complete set of states has been performed : @xmath13 in which @xmath14 our focus here is on the dis contribution , however , there is an issue of double counting if one does the inclusive sum and adds to separate exclusive processes important in the few gev energy region .
one way to approximately avoid double counting @xcite is to impose a cut on the hadronic energy @xmath15 for @xmath16-boson momentum @xmath17 .
in addition , in ref .
@xcite we evaluate the contributions from a range of @xmath18 to assess the importance of non - perturbative or higher - twist contributions to the cross section .
a general hadronic tensor has six structure functions @xmath19 which can be rescaled into the more conventional @xmath20 .
the structure function @xmath21 which multiplies the antisymmetric tensor with @xmath22 and @xmath17 never appears because the tensor combination contracts to zero with the leptonic equivalent of @xmath23 in the cross section .
the full differential cross section for neutrino cc scattering , including the lepton mass ( for definiteness set to @xmath24 ) is @xmath25 f_2^{w^\pm } \\
\nonumber & \pm & \left[x y ( 1-\frac{y}{2})-\frac{m_{\tau}^2 y}{4 e_{\nu } m_n}\right ] f_3^{w^\pm } \\
\label{eq : nusig } & + & \frac{m_{\tau}^2(m_{\tau}^2+q^2)}{4 e_{\nu}^2 m_n^2 x } f_4^{w^\pm } \\
\nonumber & - & \left .
\frac{m_{\tau}^2}{e_{\nu } m_n } f_5^{w^\pm } \right\}\ .\end{aligned}\ ] ] where @xmath26 are the standard dis kinematic variables related through @xmath27 and where we have neglected factors of @xmath28 which come from the @xmath29 part of the massive boson propagator .
they are negligible both at low and at high neutrino energies so do not enter our numerics .
they can be included by replacing : @xmath30 with @xmath31}{m_w^4\ [ 4(y-1)e_{\nu}^2+m_{\tau}^2+q^2 ] } \\
\epsilon_3 & = & 0 \\ \nonumber \epsilon_4 & = & \frac{q^2\ ( q^2 + 2 m_w^2)}{m_w^4 } \\ \nonumber \epsilon_5 & = & \frac{q^2}{m_w^2}+\frac{(m_w^2+q^2)\ ( m_{\tau}^2+q^2)\ y}{2 m_w^4}\end{aligned}\ ] ] in eq .
[ eq : nusig ] , the extra structure functions @xmath7 and @xmath8 appear with powers of the charged lepton mass , and so are usually neglected . at leading order ,
they are related to the usual structure functions by the albright - jarlskog relations @xcite , generalizations of the callan - gross relations : @xmath32 these relations are violated by kinematic target mass corrections and at nlo in qcd when quark masses are retained .
more specifically , eq . [ eq : aj1 ] is violated by @xmath33 and @xmath34 corrections , but not by the charm mass @xmath6 when the strange quark mass vanishes .
[ eq : aj2 ] is violated by non - zero @xmath9 and @xmath6 , but for the massless case , this albright - jarlskog relation holds to any order in @xmath35 .
the complete lepton mass correction comes from the above equation together with the limits of integration to get the full cross section .
details of the lepton mass dependence of the integration limits appear in ref .
@xcite .
target mass corrections appear in eq .
[ eq : nusig ] , and they also appear implicitly in the @xmath20 s . for neutrino scattering , it is conventional to rewrite @xmath20 in terms of @xmath36 which are written at leading order , neglecting the target and charm quark masses , as @xmath37 in terms of quark distribution functions @xmath38 .
target mass corrections appear in the collinear ( @xmath39 ) limit due to the fact that the _ small _ nucleon light cone momentum @xmath40 is not a simple rescaling of the massless parton momentum @xmath41 when we have @xmath42 for the _ large _ components .
when one includes @xmath9 in the collinear limit and @xmath6 , where @xmath6 is replaced by the parameter @xmath43 , one gets for the case of cc charm production : @xmath44 here , @xmath45 a further target mass correction come in through the nachtmann variable @xcite @xmath46 for massless quark production and @xmath47 for charm production . for the result shown below @xmath48 expressions for the @xmath49 corrections appear in detail in ref .
the light quark contributions are obviously understood to be included in the @xmath50 limit .
in figs . [ fig : f4 ] , [ fig : f5 ] we quantify the violations of the albright - jarlskog relations discussed above .
the violations are concentrated at low-@xmath51 which is not the most relevant kinematic region for present neutrino energies .
the naive albright - jarlskog relations therefore provide reasonable approximations in the evaluation of integrated cross sections . the target mass corrections shown in figs .
[ fig : f4],[fig : f5 ] are in the collinear limit .
these corrections differ by at most 10% from the target mass corrected structure functions evaluated in the operator product expansion approach @xcite in the range of @xmath52 gev@xmath53 @xcite .
the differences are smaller than 10% where the structure functions are largest .
the integrated cross sections for tau and muon ( anti-)neutrino dis are plotted in fig .
[ fig : fig6 ] .
the uncertainty in the cross sections from varying the parton distribution functions within their bounds is typically about a few percent : the results in the figures are based on cteq6 @xcite .
we have also compared to grv98 @xcite . in terms of a purely dis evaluation of the neutrino cross - section ,
@xmath2 production is a natural observable because the tau mass cuts off non - dis interaction to a good extent .
the general nlo features of our results for @xmath55 dis are , therefore , equally valid for @xmath56 dis with kinematic dis cuts . in figs .
[ fig : mu ] and [ fig : kfac2 ] we show two typical features of the inclusion of nlo effects : scale dependence and the size of the k - factor .
the scale dependence is substantially reduced compared to a lo calculation and the k factor hints at the importance of nlo corrections towards low neutrino energies .
different from @xmath2 production , the deep inelastic component becomes small in @xmath56 scattering at lower energies as shown in fig .
[ fig : nuint1 ] . since the tau threshold in @xmath55 dis extends to large energies one may wonder about the impact of muon mass terms in low energy @xmath56 scattering .
[ fig : nuint2 ] looks at such corrections to the dis component .
the results strongly depend on the kinematic cuts imposed .
this observation , taken together with the smallness of the _ bona fide _ deep inelastic component at such energies do not allow to draw a conclusion from the perturbative framework which we consider here . rather , one will have to look into the equivalent corrections to elastic and low - multiplicity form factors .
the @xmath58 cross section is an important ingredient in current and future atmospheric and neutrino factory experiments .
our evaluation @xcite of the nlo corrections for @xmath1n cc interactions including charm mass corrections and an estimate of target mass effects is part of a larger theoretical program to understand the inelastic @xmath58 cross section over the full energy range relevant to experiments .
apart from oscillation searches , the recent observation of an anomaly in the ratios of neutral to charged current neutrino ineractions also necessitate a solid theory for neutrino interactions with nuclei .
research along these lines is progressing and further results are under completion @xcite .
y. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys . lett .
b 436 ( 1998 ) 33 [ arxiv : hep - ex/9805006 ] ; y. fukuda _ et al . _
[ super - kamiokande collaboration ] , phys .
lett . 81 ( 1998 ) 1562 [ arxiv : hep - ex/9807003 ] .
a. rubbia , nucl .
suppl 91 ( 2000 ) 223 [ arxiv : hep - ex/0008071 ] ; f. arneodo _ et al . _
[ icarus and noe collaboration ] , `` icanoe : imaging and calorimetric neutrino oscillation experiment , '' lngs - p21/99 , infn / ae-99 - 17 , cern / spsc 99 - 25 , spsc / p314 ; m. guler _ et al _ [ opera collaboration ] cern / spsc 2000 - 028 , spsc / p318 , lngs - p25/2000 ; f. terranova [ monolith collaboration ] , int .
j. mod .
phys . a 16s1b ( 2001 ) 736 ; v. paolone , nucl .
. suppl . 100 ( 2001 ) 197 .
|
understanding neutrino interactions is an important task in searches for neutrino oscillations ; e.g. the @xmath0 oscillation hypothesis will be tested through @xmath1 production of @xmath2 in long - baseline experiments as well as underground neutrino telescopes .
an anomaly in the deep inelastic interaction of neutrinos has recently been observed by the nutev collaboration resulting in a measured weak mixing angle @xmath3 that differs by @xmath4 from the standard model expectation . in this contribution to the proceedings of nuint02 , we summarize results on the nlo neutrino structure functions and cross sections in which charm quark mass and target mass effects in the collinear approximation are included .
|
we will be using probabilities very heavily , so it will help to set some notations .
capital letters @xmath139 will used to denote random variables .
events are denoted by calligraphic letters @xmath140 .
the probability of an event @xmath141 is denoted by @xmath142 .
the expectation of a random variable @xmath143 is denoted by @xmath144 $ ] .
the random variable @xmath143 conditioned on event @xmath141 is denoted by @xmath145 .
* graphs : * we will always deal with directed graphs @xmath146 .
the set @xmath3 will be @xmath147 $ ] , the set of positive integers upto @xmath1 .
for every @xmath148 , @xmath149 ( resp .
@xmath150 ) denotes the set of out ( resp . in )
neighbors of @xmath5 in @xmath23 .
we use @xmath151 to denote the outdegree of @xmath5 ( similarly , define @xmath152 ) .
+ + * boolean strings and functions : * we will use the set @xmath153 ( instead of @xmath154 ) to denote bits , thereby aligning ourselves with the theory of boolean functions .
the @xmath1-dimensional boolean hypercube is denoted by @xmath155 .
this is also a representation for all @xmath1 bit strings .
boolean function _ is a function @xmath156 .
for any two elements @xmath157 , @xmath158 is the hamming distance between @xmath30 and @xmath124 . for @xmath159
, @xmath33 is the unique element of @xmath160 which is the same as @xmath30 on all coordinates except for the @xmath5th coordinate .
we use @xmath87 to denote the @xmath5th coordinate of @xmath30 .
we use @xmath161 ( where @xmath141 is some event ) to denote the probability of @xmath141 over the uniform distribution of @xmath30 ( which means we choose an @xmath30 uniformly at random from @xmath160 and check the probability of @xmath141 ) .
similarly , @xmath162 $ ] denotes the expectation over a random uniform @xmath30 .
+ + * boolean networks : * a boolean network @xmath0 consists of a directed graph @xmath23 and a set of transfer functions @xmath163 .
the set @xmath163 has a transfer function @xmath164 , for each @xmath148 .
the state of a boolean network is just an assignment of @xmath153 to each vertex in @xmath3 .
this can be represented as an @xmath1-bit string , or alternately , an element @xmath30 of the @xmath1-dimensional boolean hypercube @xmath155 .
suppose @xmath0 starts at the state @xmath30 .
the state at vertex @xmath5 after @xmath8 steps of @xmath0 is denoted by the boolean function @xmath165 .
the function @xmath166 is state of @xmath0 after @xmath8-steps ( so this is an @xmath1-dimensional vector having @xmath165 as its @xmath5th coordinate ) .
+ + * assumptions on boolean network : * we will analyze boolean networks that arise from a particular distribution .
first , the graph @xmath23 is chosen from a random distribution .
we will assume some fixed indegree distribution @xmath20 .
( this is simply a distribution on positive integers . ) for each vertex @xmath5 , we _
independently _ choose @xmath152 from @xmath20 .
then , we choose @xmath152 uniform random vertices ( without replacement ) to be the in - neighborhood ( @xmath150 ) of @xmath5 .
we denote the average indegree by @xmath14 , and the maximum possible indegree by @xmath21 .
next , we assume that there is a distribution @xmath13 on transfer functions . formally , this is a union of distributions @xmath34 , where this family only contains boolean functions that take @xmath35 inputs . for each vertex @xmath5 , we first choose an independent function @xmath167 from @xmath168 .
sort @xmath150 ( according to its label ) to get @xmath169 .
assign the vertex @xmath170 to input @xmath171 of @xmath172 .
this gives us the transfer function for vertex @xmath5 .
we will assume that the family @xmath13 has _ either _ of the following properties : * full independence : a random function in @xmath13 is generated by taking an empty truth table , and filling in each entry independently with the same distribution .
* balance on average : a uniform random member of @xmath13 evaluated on a uniform random input outputs @xmath24 with probability @xmath58 .
formally , @xmath173 $ ] @xmath27 @xmath174 $ ] we now discuss some of our main definitions .
the following is one of the most important concepts .
[ def : inf ] for a boolean function @xmath29 and coordinate @xmath5 , the _ influence _ of @xmath5 on @xmath29 , denoted by @xmath31 is @xmath175 $ ] . the _ total influence _ of @xmath29 is @xmath176 and the _ average influence _ of @xmath29 is @xmath177 . for a distribution @xmath178 of transfer functions , the influence of the distribution @xmath34
is denoted by @xmath36 .
formally , @xmath179 $ ] . often , when the defintion of @xmath13 is unambiguous , we write this as @xmath180 .
[ def : inf - net ] the _ influence of @xmath5 at time @xmath8 _ on @xmath0 , denoted by @xmath181 , is @xmath182 $ ] . the _ average influence at time @xmath8 _ is @xmath183 .
[ clm : avg - inf ] the average influence of @xmath0 at time @xmath8 can be expressed as @xmath184 .
let us focus on @xmath185 .
we choose a uniform random @xmath30 and evolve @xmath0 from the states @xmath30 and @xmath33 .
let @xmath186 be the indicator random variable for the event that @xmath187 .
note that @xmath188 = { { \rm{inf}}}_i(f_{t , j})$ ] . by the definition of hamming distance and linearity of expectation ,
@xmath189 \\ & = & { \hbox{\bf e}}_x[\sum_j \chi(t , j ) ] = \sum_j { \hbox{\bf e}}_x[\chi(t , j ) ] = \sum_j { { \rm{inf}}}_i(f_{t , j})\end{aligned}\ ] ] averaging this equality over all @xmath5 completes the proof .
@xmath190 + we will need the following simple facts about influences .
( this is a restatement of proposition [ prop : bool ] .
[ prop : bool2 ] * consider a function @xmath191 .
an edge of the boolean hypercube @xmath52 is called _ bichromatic _ if one endpoint is labelled @xmath24 and the other is labelled @xmath25 . then @xmath192 @xmath193 .
+ consider a distribution @xmath13 over functions @xmath54 .
@xmath194 = ( 1/2^{d-1 } ) \sum_{e : \textrm{edge } \
\in { { \cal b}}^d } \pr[\textrm{$e$ is bichromatic}]\ ] ] * for any boolean function @xmath29 and input index @xmath5 , @xmath195 \\ & = & \pr_x[(f(x ) = -1 ) \wedge ( f(x ) \neq f({x^{(i ) } } ) ) ] = { { \rm{inf}}}_i(f)/2 \end{aligned}\ ] ] consider all pairs @xmath196 , where the @xmath5th bit of @xmath124 is @xmath53 .
these pairs form a partition of the hypercube and are actually all edges of the hypercube parallel to the @xmath5th dimension .
the influence @xmath31 is exactly the probability that a uniformly random @xmath30 belongs to a bichromatic pair .
let @xmath197 be the number of bichromatic pairs .
noting that the total number of edges parallel to the @xmath5th dimension is @xmath198 , @xmath199 .
summing over all @xmath5 , we get that @xmath192 @xmath193 . to deal with a distribution ,
we simply apply linear of expectation to this bound .
now for the second part .
when does the event @xmath200 happen ?
this happens when @xmath30 belongs to a bichromatic pair , and @xmath201 .
exactly half the member of bichromatic pairs have value @xmath53 ( or @xmath25 ) .
hence , @xmath202 = { { \rm{inf}}}_i(f)/2 $ ] .
we restate theorem [ thm : avg - inf ] for convenience .
[ thm : avg - inf2 ] the average influence at time @xmath44 for the boolean network @xmath0 lies in the range @xmath203 .
before diving into the gory details of the proof of our main result , we offer a high level intuition of the overall proof .
the proof essentially consists of two central steps , lemma [ lem : tree ] and claim [ clm : g - tree ] , discussed in sections [ sec : trees ] and [ sec : top ] respectively .
lemma [ lem : tree ] considers an idealized situation where the underlying graph of a boolean network is simply a root - directed tree .
any modification of the state of a leaf travels up the tree , possibly affecting the root . in this case
, we can give an exact expression of the influence of leaves on the root .
our key next step is claim [ clm : g - tree ] , which proves that for our distribution of graph topologies ( which is closely related to the configuration model ) , the logarithmic - distance neighborhood of most nodes looks like a root - directed tree . note that this does not imply that the graph itself decomposes into disjoint tree , since the trees rooted at each node are interconnected in complicated ways .
these two steps are combined together for the final proof in section [ sec : final ] .
this proof makes heavy use of the linearity of expectation and some conditional probability arguments .
it enables us to perform exact short - run analysis of the boolean network by only considering local neighborhoods of an average " node . at the core of
the proof of lemma [ lem : tree ] is a straightforward induction argument .
consider a tree network , where we change the state at some leaf .
the catch is that induction requires the family of transfer functions to satisfy the technical conditions of balance or full independence .
this is one of the major insights of this work , since these conditions on transfer function families have generally been implicit in previous results .
the proof forces us to make these conditions explicit .
section [ sec : trees ] has the details .
the proof of claim [ clm : g - tree ] consists of a combinatorial probability calculation .
we are generating our graph through the randomized process of choosing the ( immediate ) neighborhood for each node independently ( and uniformly ) at random .
we show that the probability that the short - distance neighborhood of a node contains a cycle is extremely small . the formal proof is given in section [ sec : top ] .
in this section , we will make some arguments about tree networks .
let @xmath163 be a directed tree ( so all edges point towards the root ) with all leaves at the same depth @xmath204 .
we are interested in the influence of the leaf variables on the root @xmath205 .
let the function giving the state of the root @xmath205 at time @xmath8 be @xmath206 .
note that @xmath207 is _ only _ a function of the leaves , since there is no feedback in this graph .
note that we are not particularly bothered with what happens in the leaves are time step @xmath53 ( since those values are not even defined ) .
we are merely interested in how the values at the leaves will propagate up the tree . for each @xmath208 ,
the transfer function @xmath209 is chosen from @xmath13 ( technically , from @xmath210 ) .
[ clm : tree ] let @xmath163 be a tree of depth @xmath204
. for a leaf @xmath211 , let the path to the root be @xmath212 .
then , @xmath213 = \prod_{i = 1}^{h } { \hbox{\bf e}}_{{{\cal t}}}[{{\rm{inf}}}_{v_{i-1}}(\tau_{v_{i}})]$ ] .
( we remind the reader that @xmath13 is either balanced on average or fully independent . )
we prove by induction on the depth @xmath204 .
when @xmath13 is balanced on average , we will also show that @xmath214 = 1/2 $ ] .
for the base case , set @xmath215 . hence , all the leaves are directly connected to the root , and the set @xmath13 has only one function @xmath216 ( for the root @xmath205 ) . the probability ( over @xmath30 ) that @xmath217 is exactly @xmath218 .
suppose @xmath13 is balanced on average . since @xmath219 , @xmath220 = { { \rm pr}}_{{{\cal t}},x}[\tau_r(x)= 1 ] = 1/2 $ ] .
now for the induction step .
assume the claim is true for trees of depth @xmath221 .
we will denote the indegree of @xmath205 by @xmath35 .
the root @xmath205 is connected to a series of subtrees @xmath222 of depth @xmath221 .
the roots of each of these @xmath223 are the children of @xmath205 . for convenience ,
assume that @xmath224 .
note that for @xmath225 and @xmath226 , @xmath227 . in the final step ,
the function evaluated is @xmath228 .
first , let us assume that @xmath13 is fully independent ( the proof is much easier in this case ) .
the probability that @xmath229 is , by the induction hypothesis , @xmath230 $ ] . conditioned on this ,
what is the probability that @xmath231 @xmath232 ? since each entry in the truth table of @xmath209 is chosen independently , this probability is exactly @xmath233 $ ] .
multiplying , we get that @xmath213 = \prod_{i = 1}^{h } { \hbox{\bf e}}_{{{\cal t}}}[{{\rm{inf}}}_{v_{i-1}}(\tau_{v_{i}})]$ ] .
this completes the proof for this case .
now , we assume that @xmath13 is balanced on average
. for convenience , set random variable @xmath234 , and @xmath235 to be @xmath236 . for a bit @xmath237 ,
let @xmath238 denote the event that @xmath239 .
we set @xmath240 to denote the event that @xmath241 .
we use @xmath242 as shorthand for a vector @xmath243 of bits .
the indicator @xmath244 is @xmath53 when @xmath245 .
@xmath246 \\ & = & { \hbox{\bf e}}_{{\cal t}}[\pr_x[\tau_v(f_{h-1,r_1}(x ) , f_{h-1,r_2}(x ) , \ldots , f_{h-1,r_d}(x ) ) \\ & & \neq \tau_v(f_{h-1,r_1}({x^{(\ell ) } } ) , f_{h-1,r_2}({x^{(\ell ) } } ) , \ldots , f_{h-1,r_d}(x ) ) ] ] \\ & = & { \hbox{\bf e}}_{{\cal t}}[\pr_x[\tau_v(x_1,x_2,\ldots , x_d ) \neq \tau_v(x'_1,x_2,\ldots , x_d ) ] ] \\ & = & { \hbox{\bf e}}_{{\cal t}}[\sum_{\bar{b } } \chi(\bar{b } ) \pr_x[\bigwedge_i { { \cal e}}_i(b_i ) \wedge { { \cal f } } ] ] \\ & = & { \hbox{\bf e}}_{{\cal t}}[\sum_{\bar{b } } \chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f } } ] \pr_x[\prod_{i=2}^d { { \cal e}}_i(b_i ) ] ] \\ & = & \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f } } ] \pr_x[\prod_{i=2}^d { { \cal e}}_i(b_i ) ] ] \\ & = & \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f } } ] ] \pr_x[\prod_{i=2}^d { \hbox{\bf e}}_{{\cal t}}[{{\cal e}}_i(b_i ) ] ] \\ & = & \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f } } ] ] \prod_{i=2}^d { \hbox{\bf e}}_{{\cal t}}[\pr_x[f_{h-1,r_i } = b_i ] ] \\ & = & ( 1/2)^{d-1 } \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f}}]]\end{aligned}\ ] ] the final step uses the induction hypothesis .
now , we use proposition [ prop : bool ] to deal with @xmath247 $ ] @xmath248 $ ] .
let @xmath249 be the distribution of transfer functions excluding @xmath216 .
@xmath250 & = & ( 1/2)^{d-1 } \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) \pr_x[{{\cal e}}_1(b_1 ) \wedge { { \cal f}}]]\\ & = & ( 1/2)^d \sum_{\bar{b } } { \hbox{\bf e}}_{{\cal t}}[\chi(\bar{b } ) { { \rm{inf}}}_\ell(f_{h-1,r_1 } ) ] \\ & = & { \hbox{\bf e}}_{{{\cal t } } ' } [ { { \rm{inf}}}_\ell(f_{h-1,r_1 } ) ] ( 1/2)^d \sum_{\bar{b } } { \hbox{\bf e}}_{{{\cal t}}_r}[\chi(\bar{b } ) ] \\ & = & \prod_{i=1}^{h-1 } { \hbox{\bf e}}_{{{\cal t}}}[{{\rm{inf}}}_{v_{i-1}}(\tau_{v_{i } } ) ] { \hbox{\bf e}}_{{{\cal t } } } [ \pr_{\bar{b } } [ \chi(\bar{b } ) ] ] \\ & = & \prod_{i=1}^{h-1 } { \hbox{\bf e}}_{{{\cal t}}}[{{\rm{inf}}}_{v_{i-1}}(\tau_{v_{i } } ) ] { \hbox{\bf e}}_{{{\cal t } } } [ { { \rm{inf}}}_1(\tau_r ) ] \\ & = & \prod_{i=1}^h { \hbox{\bf e}}_{{{\cal t}}}[{{\rm{inf}}}_{v_{i-1}}(\tau_{v_{i}})]\end{aligned}\ ] ] @xmath190 consider a tree @xmath163 where all leaves have fixed depth @xmath204 .
set @xmath251 to be the set of all leaves of @xmath163 .
define @xmath252 . for a leaf @xmath211 that is a descendant of some vertex @xmath208 , suppose the path between them is @xmath253 .
define @xmath254 .
[ lem : inf - tree ] let @xmath255 be the children of the root @xmath205 .
let @xmath256 be the subtree rooted at @xmath257 .
then , @xmath258 = \sum_i { \hbox{\bf e}}_{{{\cal t}}_r } [ { { \rm{inf}}}_{r_i}(\tau_r ) ] { \hbox{\bf e}}_{{\cal t}}[{{\rm{inf}}}(t_i)]$ ] .
define @xmath259 to be the set of leaves of @xmath256 .
@xmath260 & = & { \hbox{\bf e}}_{{\cal t}}[\sum_{\ell \in l } { { \rm{inf}}}_\ell(f_{h , r } ) ] \\ & = & \sum_i { \hbox{\bf e}}_{{\cal t}}[\sum_{\ell \in l_i } { { \rm{inf}}}_\ell(f_{h , r } ) ] \\ & = & \sum_i \sum_{\ell \in l_i } { \hbox{\bf e}}_{{\cal t}}[{{\rm{infprod}}_{\ell , r } } ] \\ & = & \sum_i \sum_{\ell \in l_i } { \hbox{\bf e}}_{{{\cal t}}_r}[{{\rm{inf}}}_{r_i}(\tau_r ) ] { \hbox{\bf e}}_{{\cal t}}[{{\rm{infprod}}_{\ell , r_i } } ] \\ & = & \sum_i { \hbox{\bf e}}_{{{\cal t}}_r}[{{\rm{inf}}}_{r_i}(\tau_r ) ]
\sum_{\ell \in l_i } { \hbox{\bf e}}_{{\cal t}}[{{\rm{infprod}}_{\ell , r_i } } ] \\ & = & \sum_i { \hbox{\bf e}}_{{{\cal t}}_r}[{{\rm{inf}}}_{r_i}(\tau_r ) ] { \hbox{\bf e}}_{{\cal t}}[{{\rm{inf}}}(t_i)]\end{aligned}\ ] ] @xmath190 consider the following randomized tree construction .
first , we have a probability @xmath261 , for all positive integers @xmath35 .
we define this construction recursively .
trees of height @xmath262 are just singleton vertices . to construct a tree of height @xmath263
, we first have a root @xmath205 .
we choose the number of children to be @xmath35 with probability @xmath261 .
then , for each child , we recursively construct a tree of height @xmath221 .
[ lem : tree ] consider a completely balanced tree @xmath163 with root @xmath205 and height @xmath204 generated by the randomized procedure described above . furthermore , let the transfer functions be chosen as described above
. then @xmath264 = $ ] @xmath265 . for any vertex @xmath208 ,
let @xmath266 be the subtree rooted at @xmath208 .
we will show by induction on @xmath204 , that for any vertex at height @xmath204 , @xmath267 = $ ] @xmath265 . for vertex @xmath208 with @xmath268 , we trivially have @xmath269
. now consider @xmath208 at some height @xmath204 .
suppose @xmath208 has @xmath35 children @xmath270 .
for each child @xmath67 of @xmath208 , @xmath271 . by the induction hypothesis , @xmath272 $ ] ( which are identical for all @xmath5 ) has value exactly @xmath273 . by lemma
[ lem : inf - tree ] , conditioned on @xmath208 having @xmath35 children @xmath267 $ ] @xmath274 { \hbox{\bf e}}[{{\rm{inf}}}(t_{v_i})]=$ ] @xmath275 . by noting that the probability that @xmath208 has @xmath35 children is @xmath261 , @xmath267 = $ ] @xmath265 .
we now prove some topological properties of the random graphs ( which are effectively directed erds - rnyi graphs ) .
we remind that reader that the indegrees for all vertices are chosen independently from the same distribution .
[ def : nbd ] the _ distance @xmath8 in - neighborhood of @xmath5 _ is denoted by @xmath276 .
this is the set of all vertices whose shortest path distance ( along directed paths ) to @xmath5 is exactly @xmath8 .
we set @xmath277 .
we define @xmath41 .
[ clm : g - tree ] fix a vertex @xmath5 and let @xmath44 .
let @xmath278 denote the event that the subgraph induced by @xmath279 is a directed tree with edges directed towards root @xmath5 .
then , @xmath280 and @xmath281 .
let us start with an empty graph , and slowly add random edges in a prescribed order .
we begin with @xmath5 , and then choose the incoming edges .
this gives the set @xmath282 .
then , we choose all the in - neighbors of @xmath282 .
this is done by iterating over all vertices in @xmath282 , and for each such vertex , selecting every vertex as a neighbor with probability @xmath283 .
this gives us @xmath284 .
proceeding this way , we incrementally build up @xmath279 .
note that @xmath285 .
consider the construction of @xmath286 .
every new element added to this set is a uniform random element from @xmath147 $ ] .
consider a random sequence of @xmath287 elements chosen uniformly at random ( with replacement ) from @xmath147 $ ] .
the probability that no element is repeated at least @xmath288 if no element in @xmath286 is repeated , then the subgraph induced by @xmath286 is a directed tree . @xmath190
we are now ready to prove our main result .
( of theorem [ thm : avg - inf ] ) by claim [ clm : avg - inf ] , the average influence at time @xmath8 is @xmath289 . for a fixed vertex @xmath5 ,
let us compute @xmath290 .
denote this quantity by @xmath143 .
we apply bayes rule to split @xmath144 $ ] into conditional expectations .
@xmath291 = { { \rm pr}}({{\cal e}}_i ) { \hbox{\bf e}}[x | { { \cal e}}_i ] + { { \rm pr}}(\overline{{{\cal e}}_i } ) { \hbox{\bf e}}[x | \overline{{{\cal e}}_i}]\ ] ] observing that @xmath143 is always positive and applying claim [ clm : g - tree ] , we get @xmath144 \geq ( 1 - 1/\sqrt{n } ) { \hbox{\bf e}}[x | { { \cal e}}_i]$ ] .
note that since @xmath292 , @xmath293 . hence @xmath144 \geq { \hbox{\bf e}}[x | { { \cal e}}_i ] - n^{-1/4}$ ]
. we now obtain an upper bound applying claim [ clm : g - tree ] again .
@xmath294 & \leq & { { \rm pr}}({{\cal e}}_i ) { \hbox{\bf e}}[x | { { \cal e}}_i ] + ( 1/\sqrt{n } ) { \hbox{\bf e}}[x | \overline{{{\cal e}}_i } ] \\ & \leq & { \hbox{\bf e}}[x | { { \cal e}}_i ] + n^{-1/4}\end{aligned}\ ] ] it only remains to determine @xmath295 $ ] . conditioned on @xmath278 , the induced subgraph on @xmath296 is a directed tree .
we apply lemma [ lem : tree ] to complete the proof .
|
we present a rigorous mathematical framework for analyzing dynamics of a broad class of boolean network models .
we use this framework to provide the first formal proof of many of the standard critical transition results in boolean network analysis , and offer analogous characterizations for novel classes of random boolean networks .
we precisely connect the short - run dynamic behavior of a boolean network to the average influence of the transfer functions .
we show that some of the assumptions traditionally made in the more common mean - field analysis of boolean networks do not hold in general .
for example , we offer some evidence that imbalance , or expected internal inhomogeneity , of transfer functions is a crucial feature that tends to drive quiescent behavior far more strongly than previously observed .
[ [ introduction . ] ] introduction .
+ + + + + + + + + + + + + complex systems can usually be represented as a network of interdependent functional units .
boolean networks were proposed by kauffman as models of genetic regulatory networks @xcite and have received considerable attention across several scientific disciplines .
they model a variety of complex phenomena , particularly in theoretical biology and physics @xcite .
a boolean network @xmath0 with @xmath1 nodes can be described by a directed graph @xmath2 and a set of _ transfer functions_. we use @xmath3 and @xmath4 to denote the sets of nodes and edges respectively , and denote the indegree of node @xmath5 by @xmath6 .
each node @xmath5 is assigned a @xmath6-ary boolean function @xmath7 , termed _
transfer function_. if the state of node @xmath5 at time @xmath8 is @xmath9 , its state at time @xmath10 is described by @xmath11 the state of @xmath0 at time @xmath8 is just the vector @xmath12 .
boolean networks are studied by positing a distribution of graph topologies and boolean functions from which independent random draws are made .
we denote the distribution of transfer functions by @xmath13 .
an early observation was that when the indegree of a network is fixed at @xmath14 and each transfer function is chosen uniformly randomly from the set of all @xmath14-input possibilities , the network dynamics undergo a critical transition at @xmath15 , such that for @xmath16 the network behavior is quiescent and small perturbations die out , while for @xmath17 it exhibits chaotic features @xcite .
this result has been generalized to non - homogeneous distributions of transfer functions , when the output bit is set to 1 with probability @xmath18 ( called _ bias _ ) independently for every possible input string @xcite .
the resulting critical boundary is described by the equation @xmath19 .
all analysis of boolean networks to date uses mean - field approximations , an annealed approximation @xcite , simulation studies @xcite , or combinations of these , to understand the dynamic behavior .
many previous studies rely solely on short - run characteristics ( e.g. , derrida plots that consider only a very short , often only a single - step , horizon @xcite ) and extrapolate to understand long - term dynamics . hamming distance between boolean network states that diverges exponentially over time for small perturbations to initial state suggests sensitivity to initial conditions typically associated with chaotic dynamical systems . nonetheless , the connection between short - run and long - run sensitivity is not a foregone conclusion @xcite and remains an open question .
we provide a formal mathematical framework to analyze the behavior of booleam networks over a logarithmic ( in the size of the graph ) number of discrete time steps , and give conditions for exponential divergence in hamming distance in terms of the indegree distribution and influence of transfer functions in @xmath13 .
[ [ assumptions . ] ] assumptions .
+ + + + + + + + + + + + we assume that the boolean network @xmath0 is constructed as follows .
first , we specify an indegree distribution @xmath20 with a maximum possible indegree @xmath21 , and for each node @xmath5 independently draw its indegree @xmath22 .
we then construct @xmath23 by choosing each of the @xmath6 neighbors of every node @xmath5 uniformly at random from all @xmath1 nodes .
next , for each node @xmath5 we independently choose a @xmath6-input transfer function according to @xmath13 .
we assume that the family @xmath13 has _ either _ of the following properties : * * full independence * : each entry in the truth table of a transfer function is i.i.d . , _ or _ * * balanced on average * : transfer functions drawn from @xmath13 have , on average , an equal number of @xmath24 and @xmath25 output entries in the truth table .
formally , @xmath26 $ ] @xmath27 , where @xmath28 denotes the probability of an event when @xmath29 is drawn from @xmath13 , and input @xmath30 for @xmath29 is chosen uniformly at random .
[ [ influence . ] ] influence .
+ + + + + + + + + + the notion of _ influence _ of variables on boolean functions was defined by kahn _
et al . _
@xcite and introduced to the study of boolean networks by shmulevich and kauffman @xcite . the _ influence _ of input @xmath5 on a boolean function @xmath29 , denoted by @xmath31 , is @xmath32 , \ ] ] where @xmath33 is the same as @xmath30 in all coordinates except @xmath5 . given a distribution @xmath13 of transfer functions , let @xmath34 denote the induced distribution over @xmath35-input transfer functions .
the expected total influence under @xmath34 , denoted by @xmath36 , is @xmath37 $ ] .
when @xmath34 is clear from the context we write this simply as @xmath38 .
suppose that we have an indegree distribution where @xmath39 is the probability that indegree is @xmath35 .
we show that the quantity that characterizes the dynamic behavior of boolean networks is @xmath40 [ [ main - result . ] ] main result .
+ + + + + + + + + + + + we present our main result that characterizes dynamic behavior of boolean networks under the assumptions stated above .
define @xmath41 .
the following theorem tracks the evolution of hamming distance up to time @xmath42 , starting with a small ( single - bit ) perturbation .
we note that our theorem applies for any distribution of indegrees with a maximum bounded by @xmath21 , though increasing density ( @xmath21 ) shortens the effective horizon @xmath42 .
[ thm : avg - inf ] choose a random boolean network @xmath0 having a random graph @xmath23 with @xmath1 nodes and a distribution of transfer functions @xmath13 .
evolve @xmath0 in parallel from a uniform random starting state @xmath30 and its flip perturbation @xmath43 ( with a uniform random @xmath5 ) .
the expected hamming distance between the respective states of @xmath0 at time @xmath44 lies in the range @xmath45 .
the proof of this theorem is non - trivial and is provided in the supplement .
it shows that the effects of flip perturbations vanish when @xmath46 while perturbations diverge exponentially when @xmath47 .
thus , criticality of the system is equivalent to @xmath48 .
much of the past work assumed ( or explicitly stated ) that it suffices to consider the expected influence value @xmath49 for the _ mean _ indegree @xmath14 .
a direct consequence of theorem [ thm : avg - inf ] is that @xmath49 characterizes a critical transition _
iff @xmath38 is affine_. to see this , observe that @xmath50 iff @xmath51 this is true if and only if @xmath38 is affine .
[ [ applications . ] ] applications .
+ + + + + + + + + + + + + in this section we use theorem [ thm : avg - inf ] to recover most of the characterizations of critical indegree thresholds to date and prove results for new natural classes of transfer functions .
we show that our assumptions are crucial in obtaining the observed results .
an important step in applying the theorem is computing the quantity @xmath38 for a given class of transfer functions @xmath13 .
the following proposition ( proven in the online supplement ) facilitates this process .
let @xmath52 denote a @xmath35-dimensional boolean hypercube .
the edges of @xmath52 connect pairs of elements with hamming distance @xmath53 .
a function @xmath54 can be represented by labeling element @xmath55 by @xmath56 .
an edge of @xmath52 is called _
@xmath29-bichromatic _ if one endpoint is labeled @xmath24 and the other @xmath25 .
[ prop : bool ] consider a distribution @xmath34 over @xmath35-input functions
. then @xmath57}{2^{d-1}}.\ ] ] * uniform random transfer functions . *
we begin with the classical model of random boolean networks in which each entry in the truth table of a transfer function is chosen to be @xmath24 and @xmath25 with equal probability .
it has previously been observed that the critical transition occurs at mean indegree @xmath15 @xcite .
we now demonstrate that it is a simple corollary of our theorem .
first , we need to compute @xmath38 using proposition [ prop : bool ] . in this model , the probability that an edge is @xmath29-bichromatic is exactly @xmath58 . hence , @xmath59 .
since the total number of edges ( of @xmath52 ) is @xmath60 , we obtain @xmath61 .
notice that @xmath62 is linear in this case , and , consequently , considering @xmath63 suffices for any distribution @xmath39 .
applying theorem [ thm : avg - inf ] then gives us the well - known critical transition at @xmath64 .
* transfer functions with a bias @xmath18 .
* a simple generalization of uniform random transfer functions is to introduce a bias , that is , a probability @xmath18 that an entry in the truth table is @xmath24 ( but still filling in the truth table with i.i.d .
entries ) @xcite . in this case , the probability that an edge is @xmath29-bichromatic is @xmath65 and therefore @xmath66 .
since @xmath38 is linear , we can characterize the critical transition in this case at @xmath19 for any indegree distribution with mean @xmath14 .
* canalizing functions . * kauffman @xcite and others have observed that since uniform random transfer functions are typically chaotic , they are unlikely to represent a distribution of transfer functions that accurately models real phenomena , such as genetic regulatory networks .
biased transfer functions only partially resolve this , as they still tend to fall easily into a chaotic regime for a rather broad range of @xmath18 @xcite . empirical studies of genetic networks suggest another class of transfer functions called _
a canalizing function has at least one input , @xmath5 , such that there is some value of that input , @xmath67 , that determines the value of the boolean function .
shmulevich and kauffman @xcite show heuristically that canalizing functions have @xmath68 and thus exhibit a critical transition at @xmath69 .
we now show that this is a corollary of our theorem , using proposition [ prop : bool ] to obtain @xmath38 . to compute @xmath38 , fix ( without loss of generality ) the canalizing input index to be @xmath53 and the canalizing input and output values to @xmath24 .
consider the distribution of functions conditional on these properties . by symmetry ,
the expected number of bichromatic edges conditional on this is the same as the overall expectation .
hence , we can focus on choosing @xmath29 from this conditional distribution .
split the hypercube @xmath52 into the @xmath70-dimensional sub - hypercubes @xmath71 and @xmath72 such that @xmath71 has all inputs with @xmath73 and @xmath72 has all inputs that have @xmath74 .
edges can be partitioned into three groups @xmath75 .
the set of edges @xmath76 ( resp .
@xmath77 ) are those that are internal to @xmath71 ( resp .
@xmath72 ) .
the set of edges @xmath78 have endpoints in both @xmath79 and @xmath71 .
note that @xmath80 , and @xmath81 .
because the function is canalizing , the edges in @xmath76 are all @xmath29-monochromatic , and all other edges are @xmath29-bichromatic with probability @xmath58 .
hence , the expected number of bichromatic edges is @xmath82 by proposition [ prop : bool ] , we then have @xmath83 . since this is affine in @xmath35
, we can conclude that @xmath84 characterizes the short - run dynamic behavior for any indegree distribution with mean @xmath14 .
* threshold functions . *
a threshold function @xmath56 with @xmath35 inputs has the form @xmath85 $ ] with @xmath86 where @xmath87 is the value of input @xmath5 , @xmath88 is its weight , which has a natural interpretation of an input being inhibiting ( @xmath89 ) or excitatory ( @xmath90 ) in regulatory networks , and @xmath91 is a real number in @xmath92 $ ] representing an inhibiting / excitatory threshold for @xmath29 .
such 2-input threshold functions have been studied by greil and drossel @xcite and szejka et al .
@xcite and are classified as biologically meaningful by raeymaekers @xcite .
we now use theorem [ thm : avg - inf ] to show that random threshold functions lead to criticality for any indegree distribution .
consider @xmath13 in which the value of @xmath93 for each input @xmath5 , as well as @xmath91 , are chosen uniformly at random .
to compute @xmath62 , consider a threshold function with threshold @xmath91 and an edge @xmath94 .
this edge is bichromatic exactly when the @xmath91 lies between @xmath56 and @xmath95 .
note that @xmath96 , regardless of the values @xmath97 .
since the range of @xmath91 has size @xmath98 , the probability that this happens is @xmath99 .
so @xmath100 .
since it is independent of @xmath35 , the result follows immediately by theorem [ thm : avg - inf ] .
* majority function . *
an important specific threshold function is a majority function , which has @xmath101 for all inputs @xmath5 and @xmath102 .
suppose @xmath13 consists exclusively of majority functions .
we demonstrate that the quiescence - chaos transition properties of this class are very different from those of general threshold functions .
one detail that needs to be specified for @xmath13 is what to do when the number of positive and negative inputs is exactly balanced . to satisfy the condition that @xmath13 is balanced in expectation ,
we let the output be @xmath24 or @xmath25 with equal probability in such an instance ( for a specific majority function this choice is determined , but it is randomized for any majority function generated from @xmath13 ) .
given this @xmath13 , we now show that @xmath103 when @xmath35 is odd , bichromatic edges are those that connect the @xmath104-level to the @xmath105-level . for @xmath35 even , these are the edges connecting the @xmath106-level to the @xmath107-level ( or the @xmath108-level ) . in either case
, the number of these edges is @xmath109 , giving @xmath38 as above .
consequently , when @xmath110 or @xmath98 , @xmath111 , while for @xmath112 , @xmath113 .
thus , if a boolean network has a fixed indegree @xmath14 , it is critical for @xmath114 and chaotic for @xmath17 .
* strong majority function .
* we now show an interesting and natural class of functions where the expected average influence goes _ down _ as the indegree @xmath35 increases .
consider threshold functions where @xmath101 for all inputs @xmath5 and the threshold is either @xmath91 or @xmath115 with equal probability for some fixed @xmath116 $ ] .
for example , when @xmath117 , the function returns @xmath24 iff a 2/3 majority of inputs have value @xmath24 . for this class of functions ,
bichromatic edges are those that connect the @xmath118-level to the @xmath119-level , where @xmath120 .
thus , the expected number of bichromatic edges for a fixed @xmath35 is @xmath121 and , consequently , @xmath122 in figure [ fig : str - maj ] we plot @xmath49 , where @xmath14 is a fixed indegree , for different values of @xmath123 .
there are two rather remarkable observations to be made about this class of transfer functions : first , the sawtooth behavior of @xmath49 , and second , that the boolean network actually becomes _ more quiescent _ with increasing @xmath14 .
to our knowledge , this is the first example in which there is no single critical transition from order to chaos , and increasing connectivity leads to greater order .
-axis is @xmath14 .
the @xmath124-axis gives the average influence parameter @xmath49 .
we show the cases where @xmath125 . for larger @xmath123 we reach quiescent behavior more rapidly with increasing @xmath14 .
] we show that for @xmath35 large enough , @xmath38 tends to 0 . for convenience ,
assume that @xmath35 is an even integer and @xmath126 is non - integral . by tail bounds on binomial coefficients , @xmath127 for some constant @xmath128 .
( this can be proven using a chernoff bound , such as theorem 4.1 in @xcite . )
hence @xmath129 for large enough @xmath35 , and tends to zero as @xmath35 increases .
we had previously noted that it is commonly assumed that @xmath38 is linear in @xmath35 .
strong majority transfer functions feature @xmath38 that is clearly non - linear , and we therefore expect this assumption to be consequential . to illustrate , consider two network structures : one with a fixed @xmath130 , and another where the indegree distribution follows a power law with mean @xmath130 . using @xmath117 ,
in the former , we get @xmath131 , while in the latter ( with @xmath132 ) , @xmath133 . thus , while a fixed @xmath14 yields decidedly chaotic dynamics , using a power law distribution with the same mean indegree produces quiescence .
* the importance of graph structure . * our results rely fundamentally on the fact that the inputs into each node are chosen independently .
the fact that the size of the neighborhood at distance @xmath8 grows exponentially with @xmath8 is crucial for our proofs .
furthermore ( for the random graphs we sample from ) , this neighborhood is a root directed tree , when @xmath134 .
when graphs exhibit only polynomial local growth , we do not expect chaotic dynamic behavior even when other conditions for it are met .
we illustrate this point in figure [ f : balance ] ( left ) , which compares a random network with @xmath130 to a grid ( a bidirectional square lattice that also has @xmath130 ) .
while both initially appear to be in a chaotic regime , the hamming distance stops diverging for a grid , but diverges exponentially in the random network . * the importance of being balanced . * the assumption that @xmath13 is balanced is crucial .
balance has previously been noted to play an important role in determining the order to chaos transition , but entirely under the assumption that each truth table entry is i.i.d .
@xcite .
it has been pointed out that much of the resulting space of parameter values gives rise to chaotic dynamics @xcite .
what we now demonstrate is that this observation is largely an artifact of independence , and when truth table entries are not independently distributed , even a slight deviation from balance ( homogeneity ) may push boolean network dynamics to quiescence .
consider networks in which every transfer function is a strong majority ( with @xmath135 being a simple majority ) .
we get a balanced distribution of transfer function by choosing between @xmath91 and @xmath115 with equal probability .
an _ imbalanced distribution _ is obtained by choosing only one of them .
figure [ f : balance ] ( middle ) shows several examples of how the hamming distance evolves for different values of @xmath91 , and contrasts the balanced and unbalanced settings
. the difference could hardly be more dramatic : even a slight deviation from simple majority ( @xmath136 ) is a difference between chaos and quiescence ; indeed , it is instructive to see the initial increase in hamming distance for the imbalanced strong majority with @xmath137 , only to be ultimately suppressed .
similarly , we can compare the balanced and unbalanced versions of strong majorities with @xmath117 : the balanced version is clearly chaotic , while in the network with the unbalanced analogue , initial perturbation effects erode within two iterations .
a similar picture emerges when we consider nested canalizing functions , previously offered as an explanation of robustness in genetic regulatory networks @xcite .
classes of these are generated by a parameter @xmath138 that governs the fraction of 1 s in the transfer function truth table , with larger values of @xmath138 leading to greater imbalance .
figure [ f : balance ] ( right ) compares evolution of networks with nested canalizing functions , as well as with transfer functions following an empirical distribution of transfer functions based on regulatory networks @xcite .
we see that the main driver of quiescence appears to be the internal inhomogeneity of transfer functions , rather than canalizing properties .
_ sandia is a multiprogram laboratory operated by sandia corporation , a wholly owned subsidiary of lockheed martin corporation , for the u.s .
department of energy under contract de - ac04 - 94al85000 .
_ 15ifxundefined [ 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 @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' in @noop _
_ ( ) pp .
@noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop _ _ ( , ) * supplemental material * the following supplemental material for the paper includes all formalization and proof details .
|
collisions of relativistic heavy - ions aim to create a unique state of matter , where quarks and gluons can move over large volumes in comparison to the typical size of a hadron .
this state is known as the quark gluon plasma , qgp @xcite .
measurements of jet quenching have provided strong evidence such a state is created in heavy - ion collisions at the relativistic heavy ion collider , rhic @xcite .
these measurements show a suppression of hadron production at high-@xmath0 relative to p+p collisions , scaled by the number binary collisions , and the disappearance of correlated back to back hadron pairs in certain kinematic regions . how the chemistry within the jet is altered in heavy - ion collisions
may provide further information on the nature of jet quenching .
in particular , there are two studies which predict jet chemistry will be significantly altered in the presence of a medium .
the first , by sapeta and wiedemann @xcite , shows by increasing the in - medium parton splitting probabilities in the mlla scheme , the @xmath6 and @xmath14 ratios may increase by a factor of 2 for in - medium jets .
the increase in parton splitting probabilities simultaneously leads to a suppression of hadron production in the high @xmath15 @xmath16 region of the fragmentation function relative to vacuum jets .
the second study , by liu and fries @xcite , shows that flavor conversions may alter the final state jet hadro - chemistry significantly . in this scheme ,
hard scattered partons can exchange flavor while traversing the medium resulting in @xmath3 production being less suppressed at high-@xmath17 relative to @xmath1 production , while @xmath2 production is similarly suppressed at high-@xmath17 relative to @xmath1 production .
the suppression is characterized by the nuclear modification factor , @xmath9 which is defined as : @xmath18 where @xmath19 is the mean number of binary collisions occurring in heavy - ion collisions . in the absence of jet quenching ,
one naively expects @xmath20 , while if jet quenching occurs in heavy - ion collisions , @xmath21 for the applicable @xmath17 region . in these proceedings , we report precision measurements of inclusive hadron production at high-@xmath17 for a variety of particle species in p+p and au+au @xmath5 gev collisions .
these measurements extend previously published spectra @xcite beyond @xmath22 gev / c .
we first show the extended p+p spectra and compare our measurements to various next to leading order calculations ( nlo ) , in search of a suitable theoretical baseline to describe hadron production in p+p collisions .
we then show the @xmath6 ratios in central heavy - ion collisions with a comparison to p+p , and present @xmath9 for various particle species for central collisions .
finally , we show @xmath9 as a function of system size for the @xmath1 and @xmath23 particles .
[ cols= " < , < " , ] the data presented from p+p @xmath5 gev collisions were collected in 2005 . in order to maximize the reach of the p+p spectra ,
the data are from calorimeter triggered events which select events with high @xmath0 particles .
two trigger setups were used : jet - patch , which triggers on events where a large amount of neutral energy is deposited over many calorimeter towers , and high - tower , which triggers on events where a large amount of neutral energy is deposited in a single tower .
since the triggers only select a subset of events , trigger efficiency corrections are applied to present an unbiased invariant spectra .
the star time projection chamber , tpc , is used to detect charged particles .
the @xmath24 , @xmath25 , and @xmath26 particles are identified via energy loss , @xmath27 , measurements in the tpc , while the @xmath28 particles are identified via their weak decay products ( @xmath29 ) .
the @xmath4 particles , like the @xmath23 particles , are also obtained via reconstruction of the respective decay products in the tpc .
the charged and @xmath4 particles were obtained from jet patch trigger events , while the @xmath28 particles were obtained from high tower events .
all spectra are corrected for tracking inefficiencies and acceptance .
the corrections are determined by embedding monte carlo particles into real events , then counting the number which are reconstructed under various cuts which are placed on the real data .
systematic uncertainties on the measurements are due to uncertainties in the trigger efficiency corrections , uncertainties in the tpc tracking efficiency / acceptance corrections , uncertainties in the raw yield extraction , and uncertainties due to the effects of finite momentum resolution .
the au+au @xmath5 gev data were collected in 2004 .
two trigger setups were used ; an online central trigger which obtained @xmath30 million 0 - 12% central events , and a minbias trigger which obtained @xmath30 million au+au events independent of centrality .
the techniques for extracting the various spectra are the same as for the p+p data . uncertainties from the trigger efficiency corrections do not apply to the data from au+au collisions , since no trigger efficiency corrections are applied .
the left panel in figure [ fig1 ] shows the extended mid - rapidity @xmath3 spectra in p+p @xmath5 gev collisions .
the open black symbols show previously published results @xcite , and it is clear the new spectra are consistent within the overlapping @xmath17 range .
we also show two nlo calculations ; one from the albino - kniehl - kramer ( akk ) @xcite , the other from the deflorian - sassot - stratmann ( dss ) @xcite .
although both sets of calculations appear to reproduce the shape of the various @xmath3 spectra , there are deviations at @xmath31 gev / c . since both calculations use fits from measured @xmath3 fragmentation functions in @xmath32 collisions , these deviations maybe due to poor constraints on those fits from the data
. the data in figure [ fig1 ] in conjunction with independent measurements of @xmath3 fragmentation functions ( at rhic ) should provide better constraints on models in the future .
the right panel shows the mid - rapidity @xmath6 ratio as a function of @xmath17 in p+p @xmath5 gev collisions .
the solid black points show previously published data @xcite and again we observe a consistency with the new data . we also observe both akk and dss again fail to describe the data at high-@xmath0 .
clearly , an improved theoretically baseline is needed .
the left panel in figure [ fig2 ] shows the mid - rapidity @xmath6 ratio as a function of @xmath17 in au+au @xmath33 @xmath5 gev collisions .
the p+p data are also shown for comparison .
it is clear for @xmath34 gev / c , the @xmath6 ratio is enhanced in heavy - ion collisions relative to p+p . in particular
, the enhancement occurs at @xmath13 gev / c and persists until @xmath8 gev / c which is a clear sign jet - chemistry is indeed altered in heavy - ion collisions .
this is qualitatively consistent with expectations from the sapeta and wiedemann model @xcite , where the @xmath6 ratio is roughly a factor of 2 higher in medium jets over all hadron @xmath0 .
it must be stated however the authors have yet to make specific predictions for rhic inclusive spectrum measurements .
their current predictions are for lhc jets with energies above 50 gev .
the right panel in figure [ fig2 ] shows @xmath9 for various particles species .
the au+au centrality is again @xmath33 for @xmath5 gev collisions .
there are a number of things to note .
firstly , both @xmath10 and @xmath11 sit significantly above @xmath12 at @xmath35 gev / c .
we speculate the underlying physics behind this does not relate to the higher mass of the @xmath3 or @xmath2 particles since we observe @xmath36 .
it has also been shown that @xmath37 in the same @xmath0 region @xcite .
we also show predictions for the @xmath23 particles with and without the previously mentioned jet flavor conversions @xcite .
the @xmath3 data are quantitatively consistent with the conversion scenario .
however , the same scheme predicts @xmath38 which is quantitatively inconsistent with our data .
finally , it is trivial to show @xmath39 results from a higher @xmath14 in heavy - ion collisions relative to p+p .
again , this is qualitatively consistent with expectations in the sapeta and wiedemann model @xcite . in figure [ fig3 ]
, we show the centrality dependance of integrated @xmath40 and @xmath12 in au+au @xmath5 gev collisions where @xmath41 gev / c .
as expected , we observe @xmath42 for central au+au collisions , however we also note this difference persists in peripheral collisions
. the @xmath6 ratio is therefore higher in peripheral collisions relative to p+p at high-@xmath0 .
this raises the question whether jet chemistry is also altered in the smaller systems , or whether there is some other soft a+a production mechanism contributing in this @xmath17 range for all centralities .
measurements at higher @xmath0 from future higher statistics rhic a+a runs may help in addressing these questions .
in summary , we have shown extended high-@xmath0 spectra for various particle species in p+p and au+au @xmath5 gev collisions .
we have found that the @xmath3 spectra are poorly described by current next to leading order calculations for @xmath35 gev / c .
we also present experimental evidence for modification of jet chemistry in heavy - ion collisions relative to p+p .
this is characterized by a higher @xmath6 ratio for inclusive spectrum measurements at @xmath43 gev / c in central au+au collisions , which persists until @xmath8 gev / c .
the nuclear modification factor also displays a species dependance for central au+au collisions with @xmath11 and @xmath10 being higher than @xmath12 at @xmath35 gev / c .
we observe the integrated @xmath10 ( @xmath44 gev / c ) is higher than @xmath12 for peripheral collisions .
two models which predict changes in heavy - ion jet chemistry describe various aspects our data , however a complete quantitative description remains .
9 j. adams et al .
( star collaboration ) , nucl .
phys . a 757 ( 2005 )
102 j. adams et al .
( star collaboration ) phys
91(2003 ) 072304 , j. adams et al .
( star collaboration ) phys .
97 ( 2006 ) 162301 , s. s.adler et al .
( phenix collaboration ) phys .
rev.c 69 ( 2004 ) 034910 s. sapeta and u.a .
wiedemann , eur .
j. c 55 ( 2008 ) 293 w liu and r l fries , phys . rev .
c 77 ( 2008 ) 054902 j. adams et al .
( star collaboration ) phys lett b 637 ( 2006 ) 161 , j. adams et al .
( star collaboration ) phys .
b 616 ( 2005 ) 8 b.i .
abelev et al .
( star collaboration ) phys .
97 ( 2006 ) 152301 , b.i .
abelev et al .
( star collaboration ) phys .
b 655 ( 2007 ) 104 s. albino , b.a . kniehl and g. kramer , nucl .
b 803 ( 2008 ) 42 daniel deflorian , rodolfo sassot and marco stratmann , _ private communication _ s. s. adler ( phenix collaboration ) , phys . rev .
c 75 ( 2007 ) 024909
|
we present high transverse momentum , @xmath0 , pion ( @xmath1 ) , proton ( @xmath2 ) , kaon ( @xmath3 ) , and rho ( @xmath4 ) spectra measured with the star experiment from p+p and au+au collisions with @xmath5 gev .
we find the @xmath6 ratio to be enhanced in au+au @xmath5 gev collisions relative to p+p @xmath5 gev collisions at @xmath7 gev / c .
the enhancement persists until @xmath8 gev / c for central au+au 200 gev collisions .
we also show the nuclear modification factor , @xmath9 , measured at the same center of mass energy , and find @xmath10 and @xmath11 to be higher than @xmath12 at @xmath13 gev / c .
implications for medium induced modifications of jet chemistry is discussed .
|
the interest in the form factor @xmath3 for transition processes @xmath4 and @xmath5 , where @xmath6 and @xmath7 are photon momenta , has again increased recently . experimentally , the data on the form factor @xmath8 for small virtuality of one of the photons , @xmath9 , with the virtuality of the other photon being scanned up to @xmath10 gev@xmath11 , are known from cello @xcite and cleo @xcite collaborations .
theoretically , at zero virtualities , the form factor @xmath12 is related to the axial anomaly . at
asymptotically large photon virtualities , the behavior is predicted by perturbative qcd ( pqcd ) @xcite-@xcite ( see @xcite for recent discussions ) and depends crucially on the internal pion dynamics that is parametrized by the nonperturbative pion distribution amplitude ( da ) , @xmath13 , with @xmath14 being the fraction of the pion momentum , @xmath15 , carried by a quark .
moreover , the knowledge of the off - shell structure of the form factor enables one to significantly reduce the uncertainty in the evaluation of the hadronic light - by - light scattering contribution to the muon @xmath16
@xcite , which is relevant for the current experiment e821 at bnl @xcite . in the following ,
it is convenient to parametrize photon virtualities as @xmath17 , @xmath18 , where @xmath19 and @xmath20 are , respectively , the total virtuality of the photons and the asymmetry in their distribution : @xmath21 the experimental data from cleo @xcite for the process @xmath22 @xmath23 can be fitted by a monopole form factor : @xmath24 where @xmath25 is the two - photon pion decay constant . in the lowest order of pqcd , by using the light - cone operator product expansion ( ope ) , the high @xmath26 behavior of the form factor is predicted @xcite as @xmath27 with the asymptotic coefficient given by @xmath28 where @xmath29 mev is the weak pion decay constant and the leading - twist pion light - cone da is normalized by @xmath30 . since the pion da reflects the internal nonperturbative pion dynamics , the prediction of the value of @xmath31 is rather a nontrivial task , and its accurate measurement would provide quite valuable information .
it is important to note that for the transition process considered , the leading asymptotic term of pqcd expansion ( [ amplasympt ] ) is not suppressed by the strong coupling constant @xmath32 . hence , the pqcd prediction ( [ amplasympt ] ) may become reasonable . ] at the highest of the presently accessible momenta @xmath33 gev@xmath34 . at asymptotically
high @xmath26 , the da evolves to @xmath35 @xmath36 and @xmath37 .
the fit of cleo data ( [ fpiggfit ] ) corresponds to @xmath38 indicating that already at moderately high momenta this value is not too far from its asymptotic limit .
however , since the pqcd evolution of the da reaches the asymptotic regime very slowly , its exact form at moderately high @xmath19 may not coincide with @xmath35 . at lower @xmath26 , the power corrections to the form factor
become important .
thus , to study the behavior of the transition form factor , at experimentally accessible @xmath26 , is the subject of nonperturbative dynamics , where the same type of the leading high @xmath26 behavior , as in eq .
( [ amplasympt ] ) , was obtained by different methods .
so , the theoretical determination of the transition form factor is still challenging , and it is desirable to obtain direct calculations of @xmath39 , without any _ a priori _ assumptions about the shape of the pion da .
the transition form factor in the symmetric kinematics , @xmath40 , at high virtualities was considered in @xcite by using the local ope with the result @xmath41 for the asymptotic coefficient , which is in agreement with prediction from ( [ j ] ) at @xmath42 . within the local ope
, one can represent @xmath43 as an expansion in powers of @xmath44 , with the coefficients of expansion given by the moments of the pion da : @xmath45 . in @xcite
( see also @xcite ) , it was shown that the local ope is well convergent in the kinematic region , when the photon virtualities are close to each other : @xmath46 . in this kinematics , the result for
the asymptotic coefficient is still close to 4/3 .
however , in these references it was pointed out that for @xmath47 potentially large corrections to the first term of the local ope @xcite and also to the light - cone pqcd @xcite predictions exist at any finite @xmath26 . with increasing @xmath48 , the number of terms of ope with higher - dimension vacuum expectation values growths rapidly , but it is practically a hopeless task to find more than few terms of the local expansion .
therefore , much more detailed information about the nonperturbative qcd vacuum is necessary to have control over the operator expansion . in ref .
@xcite , some progress was achieved , by using a refined technique based on the ope with nonlocal condensates @xcite , which is equivalent to inclusion of the whole series of power corrections . by using the qcd sum rules with nonlocal condensates
, it was shown that this approach works in almost the whole kinematic region @xmath49 , and that for high values of the asymmetry parameter @xmath50 , the pion transition form factor is very sensitive to the nonlocal structure of the qcd vacuum .
the latter is characterized by the average quark virtuality in the vacuum @xcite , @xmath51 , and , within the instanton model @xcite , may be expressed through the average instanton size , @xmath52 , as @xmath53 @xcite . in @xcite , the form factor @xmath54
was directly calculated from a qcd sum rule for the three - point function , leading to the estimate @xmath55 .
the covariant nonlocal low - energy models ( see , _
e.g. _ , @xcite ) , based on the schwinger - dyson ( sd ) approach to dynamics of quarks and gluons , have many attractive features , as the approach is consistent with the low - energy theorems . in particular ,
the abelian axial anomaly is within this approach , and the standard result for @xmath56 is reproduced exactly . within this nonperturbative model of quark - pion interaction , both the small mass and composite structure of the pion are realistically described .
furthermore , the intrinsic nonlocal structure of the model may be motivated by fundamental qcd processes like the instanton and gluon exchanges . in @xcite
the transition form factor @xmath57 was considered at low @xmath19 and agreement with data was obtained .
there , it was observed that the results are very sensitive to the value of constituent quark mass . in this letter , within the covariant nonlocal low - energy model of the quark - pion interaction we study the high @xmath19 behavior of the pion transition form factor @xmath58 in general kinematics .
we show that the asymptotic coefficient @xmath59 , as demonstrated in qcd sum rules @xcite , depends on the kinematics of the transition process and on the internal pion dynamics induced by the nonlocal structure of the qcd vacuum .
the dynamic dependence of @xmath60 is governed by the so - called diluteness parameter @xmath61 , where @xmath62 is the constituent quark mass .
when considering the model dependence of the asymptotic coefficient @xmath60 , experimental data can be very useful to distinguish between different assumptions made on nonperturbative dynamics of the qcd vacuum . within the nonlocal quark - pion model the expression for the asymptotic coefficient @xmath60
is found in the whole kinematic region of @xmath20 .
moreover , from this dependence , the pion da is reconstructed in terms of the quark - pion vertex function .
the effective quark - pion dynamics can be summarized in the covariant nonlocal action given by @xmath64 \bar{q}(x+y/2)\ [ m_{q}+g_{\pi \bar{q}q}i\gamma _ { 5}\tau ^{a}\pi ^{a}(x)]\ q(x - y/2 ) , \label{s_q_pi}\ ] ] where the dynamic vertex @xmath65 $ ] with nonlocality size @xmath66 depends on the coordinates of the quark and antiquark ; @xmath67 and @xmath68 are , respectively , the quark and pion fields .
the nonlocal vertex characterizes the coordinate dependence of order parameter for spontaneous chiral - symmetry breaking and can be expressed in terms of the nonlocal quark condensates . in the following calculations ,
we restrict ourselves to the approximation ( see , e.g. @xcite ) @xmath69 \to f(y^2 , \lambda^{-2 } ) , \label{separable}\ ] ] when the dynamic quark - pion vertex depends only on the relative coordinate of the quark and antiquark squared , @xmath70 , if neglecting the dependence of the vertex on angular variable @xmath71 .
the fourier transform of the vertex function in the minkowski space is defined as @xmath72 with normalization @xmath73 , and we assume that it rapidly decreases in the euclidean region ( @xmath74 ) .
we also approximate the momentum - dependent quark self - energy in the quark propagator @xmath75 by a constant mass @xcite and neglect small effects of the pion mass .
we have to note that the approximations used are not fully consistent .
in particular , due to neglecting the momentum dependence of the quark mass , some low - energy theorems are violated .
further , as we show below , the choice of the model for the quark - pion vertex ( [ separable ] ) depending only on the relative coordinate induces a certain artifact in the @xmath14 behavior of da ( see below ) .
however , these deficiencies of the approximation chosen are not essential for the present purposes and do not lead to essential numeric errors .
the quark - pion coupling is given by the compositeness condition @xcite @xmath76 and the pion weak decay constant is expressed by @xmath77 we have rescaled the integration variable by the quark mass squared and introduced the parameter @xmath78 that characterizes the diluteness of the qcd vacuum .
within the instanton vacuum model , the size of nonlocality of the nonperturbative gluon field , @xmath79 , is much smaller than the quark compton length @xmath80 ; thus , @xmath81 is a small parameter @xcite .
let us consider the contribution to the @xmath82 invariant amplitude as calculated from the triangle diagrams : @xmath83 where @xmath84 are the photon polarization vectors , and @xmath85{\hat { e}}_{1}s(k+p/2)\}.\ ] ] if the tensor @xmath86 is factorized from this amplitude , the form factor can be expressed as @xmath87 where the feynman integral @xmath88 is given by @xmath89 [ m_{q}^{2}-(k - p/2)^{2}-i\varepsilon ] [ m_{q}^{2}-(k-(q_{1}-q_{2})/2)^{2}-i\varepsilon ] } .
\label{ipigg}\ ] ] within the constant - mass approximation , the low - energy relation following from the adler - bell - jackiw ( abj ) axial anomaly [ @xmath90 is well reproduced numerically with a relative error smaller than @xmath91 @xcite . in the formal limit of a very dilute vacuum medium @xmath92 ,
the results are consistent with the abj anomaly and the goldberger - treiman relation .
let us note that the integral ( [ ipigg ] ) is similar in structure to the integral arising in the lowest order of pqcd treating the quark - photon interaction perturbatively . in the latter case ,
its asymptotic behavior is due to the subprocess @xmath93 with @xmath14 ( @xmath94 ) being the fraction of the pion momentum @xmath15 carried by the quark produced at the @xmath95 ( @xmath96 photon vertex .
the relevant diagram is similar to the handbag diagram for hard exclusive processes , with the main difference that one should use , as a nonperturbative input , the quark - pion vertex instead of the pion da .
as we see below , this similarity allows one to translate the form of the quark - pion vertex into a specific shape of the pion da .
in this section , we estimate the asymptotics of the transition form factor . to this end , we rewrite the expression for integral ( [ ipigg ] ) in the form that is obtained after rotating to the euclidean space [ @xmath98 @xmath99 @xmath100 , by using the feynman @xmath101parameterization for the denominators and integrating over the angular variables .
then , the corresponding integral @xmath102 is given by @xmath103 , \label{igpigg3}\ ] ] where @xmath104 in this way , the expression ( [ igpigg3 ] ) can be safely analyzed in the asymptotic limit of high total virtuality of the photons @xmath105 .
moreover , the integral over @xmath106 can be taken analytically , leading , in the chiral limit @xmath107 , to the asymptotic expression given by eq .
( [ amplasympt ] ) , where @xmath108 } \right\ } \label{igpigg2 } \\ & & n_\pi = \left [ { \ \int_{0}^{\infty}duu \frac{\tilde f(-u;\chi ^{-2 } ) } { \left ( 1+u\right ) ^{2}}}\right ] ^{-1}. \nonumber\end{aligned}\ ] ] the integrand in the numerator of ( [ igpigg2 ] ) is quite different from the integrand defining the decay constant @xmath109 , given in eq .
( [ f_pi ] ) . from eq .
( [ igpigg2 ] ) it is clear , that the prediction of the nonperturbative approach about the asymptotic coefficient is rather sensitive to the product @xmath81 of the value of the constituent mass @xmath62 and the size of nonlocality @xmath66 of the vertex @xmath110 and to the relative distribution of the total virtuality among photons , @xmath111 . in particular , for the off - shell process @xmath112 in the kinematic case of symmetric distribution of photon virtualities , @xmath113 ( @xmath114 ) , the result obtained from eq .
( [ igpigg2 ] ) is @xmath115 in agreement with the ope prediction @xcite .
let us note , that we use an approximation to the model with constant constituent quark masses for all three quark lines in the diagram of the process .
however , the asymptotic result ( [ igpigg2 ] ) is independent of the value of the mass parameter in the quark propagator with hard momentum flow , as it should be .
the other two quark lines remain soft during the process ; thus , the mass parameter @xmath62 may be considered as given on a certain characteristic soft scale in the momentum - dependent case @xmath116 .
it means that the dynamic and kinematic dependence of @xmath117 found in ( [ igpigg2 ] ) will be unchanged , even if one includes the momentum dependence of the quark mass and considers the dressed quark - photon vertex which goes into the bare one , @xmath118 , as one of the squared quark momenta becomes infinite .
both the expressions for @xmath60 derived within the nonlocal quark - pion model ( [ igpigg2 ] ) and from the light - cone ope ( [ j ] ) can be put into the common form @xmath119 \label{r}\ ] ] with @xmath120 and similar expressions for @xmath121 . equating both contributions
, we find the pion da in terms of the vertex function on a certain low - energy scale @xmath122 @xmath123 thus , we show that eq .
( [ igpigg2 ] ) obtained within the nonlocal quark - pion model is equivalent to the standard lowest - order pqcd result ( [ j ] ) , with the only difference that the nonperturbative information accumulated in the pion da @xmath124 is represented by the quark - pion vertex function @xmath125 .
we have to note that an explicit form of the asymptotic coefficient ( [ igpigg2 ] ) and the relation between the da and the vertex function depend on the model of quark - pion interaction ( [ s_q_pi ] ) . in particular , the expression ( [ wf_vf ] ) is obtained within the approximation ( [ separable ] ) , when the quark - pion vertex depends only on the relative coordinate .
this approximation results in the artificial dependence of da on the modulo function of @xmath14 and leads to the nonsmooth behavior of the distribution at @xmath126 ( see fig .
1 ) . these peculiarities disappear if the angular dependence of the vertex motivated by , _
e.g. _ , the instanton model is recovered .
let us estimate a realistic value for the diluteness parameter @xmath127 and check if the model under consideration is consistent with cleo data .
the vertex function @xmath128 phenomenologically describes the nonlocal structure of the nonperturbative qcd vacuum and may be modeled within the instanton vacuum model @xcite . for the present purpose , the vertex function can be well approximated by the gaussian form @xmath129 .
the inverse size of the vertex nonlocality , @xmath130 , is naturally related to the average virtuality of quarks that flow through the vacuum , @xmath51 , @xcite @xmath131 where @xmath132 is the covariant derivative with respect to the strong gauge field , and the quark condensate is expressed as @xmath133 the value of @xmath51 is known from the qcd sum rule analysis , @xmath51 @xmath134 @xcite . for the gaussian vertex
, one has @xmath135 if @xmath136 .
the quark mass parameter @xmath62 is given by the goldberger - treiman relation @xmath137 , with the quark - pion coupling being fixed by the compositeness condition ( [ gpiqq ] ) , and its value is consistent with @xmath138 mev .
varying the model parameters within the intervals @xmath139 @xmath140 and @xmath141 mev , we have @xmath142 @xmath140 , @xmath143 and @xmath144 .
when taking into account the error in the experimental fit , this estimate is in agreement with the cleo data
. it also agrees with the new estimate @xmath145 made in @xcite by the qcd sum rules with nonlocal condensates .
it is instructive to consider some extreme cases , depending on the physics under consideration .
if the qcd vacuum were a very dilute vacuum @xmath146 , then the vertex function @xmath147 is a very slowly decaying function .
this corresponds to the local quark - pion vertex . in that case
, the coefficient @xmath148 goes logarithmically to infinity and the da also becomes flat @xmath149 .
the latter can be seen from eq .
( [ wf_vfdif ] ) , where the derivative of da with respect to @xmath14 is very small , since in this limit , the normalization factor @xmath150 becomes small . from ( [ wf_vf ] )
it is clear that the integration region is effectively restricted from above by the scale @xmath151 and from below by @xmath152 .
these scales are well separated in the region @xmath153 , where @xmath154 .
it means that the da is suppressed at the edges of the kinematic interval @xmath155 , where quarks are soft .
as it was pointed out earlier , the incorporation of nonperturbative effects results in the intrinsic transverse structure of hadronic wave functions @xcite , as well as the sudakov perturbative factor @xcite modifies the hard scattering picture of exclusive reactions and essentially improves it . as a result ,
perturbative qcd calculations of the hadron form factors extend the kinematic region of self - consistency from asymptotic values of @xmath156 to the region starting from @xmath157 . in the opposite extreme case of a very dense medium ( heavy quark limit , @xmath158 ) , @xmath159 ,
as it is predicted by the first term in the ope result @xcite . in that case , the limit @xmath160 is small , and the integrand in ( [ wf_vf ] ) is concentrated in the vicinity of @xmath126 .
thus , the da becomes @xmath161 , as it is expected .
as we shown above , a realistic situation is in - between these two extremes .
these different situations are illustrated in terms of the pion da , ( [ wf_vf ] ) , in fig.1 .
as it is clear from the figure , the model pion da , under the realistic choice of the parameter @xmath162 , is close to the asymptotic da . as noticed in the introduction , by considering the actual accessible data
, the nonperturbative dynamics may dominate .
therefore , the data turn out to be quite restrictive and uniquely indicate that the dilute regime is realized in the qcd vacuum . in fig .
2 , for the process @xmath163 ( @xmath164 ) , we plot the asymptotic coefficient @xmath165 as a function of the dynamical diluteness parameter squared @xmath166 . in this figure , we indicate the values of @xmath165 obtained from cleo data and model predictions . in fig .
3 , the asymptotic coefficient @xmath167 is plotted as a function of the kinematic asymmetry parameter @xmath111 , at @xmath168 and @xmath169 .
the first value of @xmath170 corresponds to the model estimate ; and the second one , to the central point of the cleo data fit . to get further interpretation of eq .
( [ wf_vf ] ) , we can express the da as the transverse momentum integral of the pion light - cone wave function @xcite @xmath171 rewriting the r.h.s . of ( [ wf_vf ] ) via the original variable @xmath172 and then substituting @xmath173 by the light - cone combination @xmath174 , that is the invariant mass of the @xmath175 pair squared , we identify the pion wave function as @xmath176 the vertex function @xmath177 in our model of the pion wave function plays a similar role as the sharp @xmath178-function in the local duality " wave function @xcite @xmath179 , with @xmath180 being the duality interval .
note that numerically @xmath181 gev@xmath34 is close to the value of the nonlocality parameter @xmath182 .
as in the case of ( [ wf_vf ] ) , the pion light - cone wave function ( [ psi ] ) displays non - analytic dependence on @xmath14 that disappears if a more realistic quark - pion vertex with the angular dependence is considered .
recently , in @xcite , it was claimed that the schwinger - dyson approach predicts the same asymptotic coefficient @xmath183 for _ all _ nonlocal quark - photon vertices @xmath184 $ ] which go into the bare ones , @xmath118 , as soon as one of the squared momenta ( @xmath185 or @xmath186 ) becomes infinite ( as in the curtis - pennington @xcite form of the vertex ) . in @xcite ,
the quark propagator that depends on the photon momenta was approximated , at large @xmath26 , by its asymptotic form @xmath187 ^{-1}\rightarrow 2/q^{2}$ ] .
after this change , the integral ( [ ipigg ] ) attains the same form as the integral in ( [ f_pi ] ) defining @xmath109 . by taking into account the coefficients in front of the integrals of equations ( [ f_pi ] ) and ( [ gpiqq ] ) , one immediately reaches the conclusion of @xcite ( see also @xcite ) about the asymptotic coefficient ( @xmath188 ) .
as we show above , such a quick asymptotic estimation is rather naive and does not lead to the accurate result .
the approximation made in @xcite is justified only in the formal limit @xmath189 .
our analysis is based on the consideration of a triangle diagram in which the quark propagator and quark - pion vertex are determined nonperturbatively . in this respect ,
our approach is close to earlier work @xcite .
however , in @xcite , the approximations in the calculation of the triangle diagram were used that simplify the dynamics of the process .
it turns out that these approximations are not justified in the kinematic region of large @xmath48 . as a result
the expression was obtained for the asymptotic coefficient that is independent of the internal nonlocal structure of the pion . in conclusion , within the covariant nonlocal model describing the quark - pion dynamics , we obtain the @xmath190 transition form factor at moderately high momentum transfers squared , where the perturbative qcd evolution does not yet reach the asymptotic regime . from the model calculations it is possible to find the absolute normalization of the asymptotic @xmath191 term .
the asymptotic normalization coefficient @xmath192 , given in ( [ igpigg2 ] ) , depends on the ratio of the constituent quark mass on a certain soft scale to the characteristic size of qcd vacuum fluctuations and also on the kinematics of the process .
this result does not confirm the statement about the universality of the asymptotic coefficient given in @xcite .
when considering the dependence of the asymptotic coefficient on the internal dynamics , the cleo data are consistent with a small value of the diluteness parameter , which confirms the hypothesis about the small density of the instanton liquid vacuum @xcite . from the comparison of the kinematic dependence of the asymptotic coefficient of the transition pion form factor , given by pqcd and the nonperturbative model , the new relation eq .
( [ wf_vf ] ) between the pion distribution amplitude and the dynamical quark - pion vertex function is derived . in the specific case of symmetric kinematics
, our result agrees with the one obtained by ope @xcite .
the present results are in accordance with the conclusions made in @xcite within the qcd sum rules .
a more complete analysis of the light pseudo - scalar meson transition form factors will be done later , where effects of the finite hadron masses , nonlocality of the quark - photon vertex , etc .
, will be considered .
we are thankful to n.i .
kochelev for discussions on the subject of the paper .
our special thanks are to s.v .
mikhailov for stimulating and helpful discussions .
the referees remarks were highly useful .
one of us ( a.e.d . ) thanks the colleagues from instituto de fsica terica , unesp , ( so paulo ) for their hospitality and interest in the work .
was partially supported by st .
- petersburg center for fundamental research grant : 97 - 0 - 6.2 - 28 .
thanks partial support from conselho nacional de desenvolvimento
cientfico e tecnolgico do brasil ( cnpq ) and , in particular , to the `` funda o de amparo pesquisa do estado de so paulo ( fapesp ) '' to provide the essential support for this collaboration .
f. del aguila , m.k .
chase , nucl.phys .
* b193 * ( 1981 ) 517 ; e. braaten , phys . rev . , * d 28 * ( 1983 ) 524 ; e.p .
kadantseva , s.v .
mikhailov , a.v .
radyushkin , sov .
* 44 * ( 1986 ) 326 ; p. gosdzinsky , n. kivel , nucl.phys . *
b521 * ( 1998 ) 274 .
n. isgur , c.h .
llewellyn - smith nucl
b 317 * ( 1989 ) 526 ; phys .
lett . * 52 * 1080 ( 1984 ) ; a.v .
radyushkin , nucl . phys . * a 532 * ( 1991 ) 141 ; v.a .
nesterenko , a.v .
radyushkin , phys.lett . *
b115 * 410 ( 1982 ) ; _ ibid . _ * b128 * 439 ( 1983 ) ; a.v .
radyushkin , phys.rev .
* d58 * 114008 ( 1998 ) ; m. diehl , t. feldmann , r. jacob and p. kroll , eur .
j. * c 8 * ( 1999 ) 409 .
d. klabuar and d. kekez , fizika b ( zagreb ) * 8 * ( 1999 ) 303 ( hep - ph/9905251 ) ; p. tandy , fizika b ( zagreb ) * 8 * ( 1999 ) 295 ( hep - ph/9902459 ) ; c. d. roberts , fizika b ( zagreb ) * 8 * ( 1999 ) 285 ( hep - ph/9901091 ) .
|
the behavior of the transition pion form factor for processes @xmath0 and @xmath1 at large values of space - like photon momenta is estimated within the nonlocal covariant quark - pion model .
it is shown that , in general , the coefficient of the leading asymptotic term depends dynamically on the ratio of the constituent quark mass and the average virtuality of quarks in the vacuum and kinematically on the ratio of photon virtualities .
the kinematic dependence of the transition form factor allows us to obtain the relation between the pion light - cone distribution amplitude and the quark - pion vertex function .
the dynamic dependence indicates that the transition form factor @xmath2 at high momentum transfers is very sensitive to the nonlocality size of nonperturbative fluctuations in the qcd vacuum .
keywords : nonperturbative calculations , pion form factors , nonlocal theories , and models
|
m85 optical transient 2006 - 1 ( m85ot2006 - 1 ; [email protected] ) was discovered on 2006 jan 6 by the lick observatory supernova search team ( filippenko et al .
2001bait / kait.html ] ) as a faint , @xmath11 mag transient in the galaxy m85 ( ngc4382 ) , which is at a distance of @xmath12 mpc ( mei et al . 2007 ) .
subsequent spectroscopy , as well as visible light and infra - red ( ir ) photometry , presented in kulkarni et al .
( 2007 ) , showed that m85ot2006 - 1 has a recession velocity of @xmath13 km s@xmath14 , and is therefore associated with m85 .
moreover , we showed that the temporal and spectral properties of this object are unlike those of supernovae , novae , or luminous blue variables .
m85ot2006 - 1 peaked at absolute @xmath15-band magnitude of about @xmath16 .
the light curve settled into a @xmath17 day plateau , followed by a decrease in bolometric luminosity during which the black - body emission peak shifted toward near - ir wavelengths .
the early spectrum of m85ot2006 - 1 , obtained six weeks after discovery , resembles that of a @xmath18 k black body , with h@xmath19 and h@xmath20 narrow emission lines ( full width at half maximum of @xmath21kms@xmath14 ) , along with several other unidentified emission lines .
spitzer ir observations obtained about six months after the discovery revealed a @xmath22k black body spectral energy distribution ( rau et al . 2007 ) . the spectral and temporal properties of this object resemble those of m31-rv ( discovered by rich et al .
1989 ; e.g. , mould et al .
1990 ; bryan & royer 1992 ) , v838 mon ( discovered by brown 2002 ; e.g. , kimeswenger et al .
2002 ; bond et al . 2003 ; corradi & munari 2007 ) , and possibly the less studied object v4332 sgr ( martini et al . 1999 ) . however , the m85 transient is the most luminous member of the v838 mon class .
the favored model for this emerging class of v838 mon - like objects ( also known as luminous red novae ) is that they are the result of stellar mergers ( e.g. , soker & tylenda 2006 ) .
however , other models have been suggested to explain these objects ( e.g. , retter & marom 2003 ; lawlor 2005 ) .
the nature of these events , with their energetics lying between the realms of supernovae and novae , remains uncertain . in this paper , we present hubble space telescope ( hst)- advanced camera for surveys ( acs)/wide field camera ( wfc ) and near infrared camera and multi - object spectrometer ( nicmos ) observations , as well as palomar 5 m spectroscopy , of the environment of m85ot2006 - 1 .
the observations are used to characterize the environment of the transient and to set a limit on the mass of the progenitor .
m85 was observed using hst / acs on 2003 as part of the hst - acs virgo cluster survey ( ct et al .
subsequently , the transient was observed serendipitously with acs / wfc and nicmos / nic2 in 2006 ( go-10515 ) as a follow - up study to peng et al .
the acs observations on 2006 were obtained 18days after the discovery of the transient .
the log of observations , the measured magnitude of the m85 transient or the @xmath23 upper limit at the ot location , as derived from the hst images taken on 2003 and 2006 , are listed in table [ tab - mag ] .
the hst images of the galaxy and the transient environment are presented in figures [ f1 ] and [ f2 ] .
lclccc 2003 feb 01 & 750 & @xmath24 & & @xmath25 & @xmath26 + 2003 feb 01 & 1120 & @xmath27 & & @xmath28 & @xmath29 + 2006 jan 24 & 2204 & @xmath24 & @xmath30 & & + 2006 jan 24 & 2224 & @xmath31 & @xmath32 & @xmath33 & @xmath34 + 2006 feb 28 & 500 & @xmath35 ( h ) & @xmath36 & @xmath37 & @xmath38 + [ tab - mag ] on 2007 january 20 , after the m85ot2006 - 1 faded away , we obtained a spectrum at the location of m85ot2006 - 1 .
the spectrum ( figure [ m85_2d_spec ] ) consist of @xmath39s exposures with the double beam spectrograph mounted on the palomar 5 m telescope .
we used the @xmath40 lines / mm grating blazed at @xmath41 in the red arm . the @xmath42 slit was centered on the nucleus of m85 , at a position angle of @xmath43deg .
the position angle was chosen such that the location of the transient will be included in the slit .
offset from the galaxy center along the slit ) , obtained about one year after the transient discovery , covering the h@xmath19 wavelength region ( marked by ellipse ) .
no h@xmath19 emission is seen in the vicinity of the transient .
the spectrum is shown before sky subtraction .
[ m85_2d_spec],width=321 ] on 25 june 2005 , m85 was observed by the spitzer space telescope with the multiband imaging photometer for spitzer ( mips ) .
the 70 micron image , with exposure time of 670 s , is shown in fig .
[ spitzer_mips70 ] .
v838 mon - like objects are found in both young regions ( e.g. , v838mon ; afar & bond 2007 ) and old stellar populations ( e.g. , m31rv ; bond & siegel 2006 ) .
m85ot2006 - 1 took place in an early - type galaxy . therefore , as we explain below , it can be used to set an upper limit on the minimal progenitor mass that can produce v838 mon - like eruptions . from the 2-dimensional spectrum , shown in fig .
[ m85_2d_spec ] , we can set an upper limit on the flux of the h@xmath19 emission at the location of m85ot2006 - 1 of @xmath44ergs@xmath45s@xmath14 , at the @xmath46 level ( equivalent to luminosity of @xmath47 ergs s@xmath14 ) .
this corresponds to a star formation rate smaller than about @xmath48 m@xmath7 yr@xmath14 ( kennicutt 1998 ) in a radius of @xmath49 pc around the transient location . for comparison ,
the h@xmath19 luminosity of the orion nebula is about @xmath50 ergs s@xmath14 ( haffner et al .
2003 ; assuming a distance of 392 pc ; jeffries 2007 ) . therefore , our observations rule out the presence of a prominent star forming region in this location .
moreover , based on the far - ir flux in the region of the transient , obtained from the spitzer / mips 70 micron image shown in fig .
[ spitzer_mips70 ] , we can set an upper limit on the star formation rate in this region to be less than @xmath51 m@xmath7 yr@xmath14 ( kennicutt 1998 ) .
, width=321 ] the absence of regions in m85 rule out the possibility that the progenitor had a delay ( from birth to outburst ) of @xmath52 myr ( corresponds to @xmath53 m@xmath7 ) , which is the typical life time of regions ( e.g. , mayya 1995 ) .
an independent limit on the age and mass of the progenitor can be inferred from the absence of stars brighter than @xmath2-band absolute magnitude @xmath54 in the transient environment . from the lejeune & schaerer ( 2001 ) stellar tracks
, we find that stars older than 50 myr ( and therefore , more massive than 7 m@xmath7 ) do not reach @xmath2-band absolute magnitudes brighter than @xmath54 mag ( at the red - supergiant stage ) .
we note that the @xmath2-band stellar track magnitudes were obtained by interpolation of the @xmath15 and @xmath55-bands .
therefore , we can set a lower limit on the age of the most massive stars in the transient environment to be @xmath56 myr , which corresponds to mass @xmath6 m@xmath7 ( assuming solar metallicity ) . otherwise , we were likely to detect individual stars in this region .
we note that we can limit the extinction in the transient location to @xmath57 mag , based on the balmer lines ratio , assuming case - b recombination ( kulkarni et al .
2007 ) .
ferrarese et al .
( 2006 ) reported possible faint wisps and patches of dust in m85 .
moreover , schweizer & seitzer ( 1992 ) reported that m85 is somewhat bluer than typical s0 galaxies , therefore possibly younger .
this claim is supported by terlevich & forbes ( 2002 ) who estimated the age and metallicity based on line indices . they have found a mean luminosity - weighted age of @xmath58gyr and metallicity of [ fe / h]@xmath59 and [ mg / fe]@xmath60 .
we note that the actual mean age is probably higher than that indicated by line indices given that younger populations have higher weight than old population .
if the mean age is representative of the progenitor of m85ot2006 - 1 , then we can set a lower limit of about @xmath61gyr on the age of m85ot2006 - 1 progenitor / s .
this further suggests that the mass of the progenitor / s is probably below @xmath9
m@xmath7 ( the life time of solar metallicity @xmath62 m@xmath7 stars is @xmath63 gyr ; lejeune & schaerer 2001 ) .
this limit is based on the mean stellar age of this galaxy .
however , stars younger than 1 gyr may be present in this galaxy in relatively small numbers .
although several models exist for v838 mon - like objects ( e.g. , soker & tylenda 2003 ; lawlor 2005 ) , in the absence of detailed simulations , the nature of these objects remain elusive . a clue to their origin
can be derived from their environment , luminosity function and rate .
given that only a small number of these objects are known , and they were found serendipitously in various searches , the luminosity function and rate are not well constrained .
however , the fact that at least two events were observed in our galaxy ( i.e. , v838 mon and v4332 sgr ) in the last @xmath64 years suggests that they have a higher rate than sne .
we can set a lower limit on their rate , of @xmath65yr@xmath66l@xmath67 , at the @xmath68 confidence level , where l@xmath69 is the milky way luminosity .
now we discuss the implications of our observations for a specific model for v838 mon - like objects .
soker & tylenda ( 2006 ) presented a model for violent stellar mergers in which , prior to the merger , the spins and orbital frequencies of the binary star are losing synchronization due to the darwin instability ( e.g. , eggleton & kiseleva - eggleton 2001 ) .
they found that for a given primary mass , the maximal energy production obtained for a binary mass ratio of @xmath70 , is @xmath71 , where @xmath72 is the gravitational constant , and @xmath73 and @xmath74 are the mass and radius of the primary star .
given the upper limit on the progenitor mass , based on the mean stellar age in m85 , @xmath75 m@xmath7 , and assuming a main - sequence mass - radius relation , @xmath76 , the maximum available energy in their model is short by a factor of three in the total energy production , as compared to the radiated energy of m85ot2006 - 1 in the first two months , @xmath77 ergs ( assuming a distance of @xmath12 mpc to m85 ; mei et al .
moreover , it is expected that a large fraction of the energy will go into lifting the outer region of the star rather than radiated away .
furthermore , if the primary is an evolved star , then its radius will be larger than the radius of a main sequence star with the same mass , and the extracted energy will be even smaller .
this suggests that either more detailed modeling of violent stellar mergers is required , or that this event is not the result of a violent stellar merger .
another possible solution is that the mass of the progenitor is somewhat larger
. a larger progenitor mass will still be consistent with our upper limit of @xmath78 m@xmath7 which is based on the absence of stars brighter than @xmath79 mag .
for example , according to soker & tylenda ( 2006 ) model , a @xmath78 m@xmath7 progenitor can yield @xmath80 times more energy than a @xmath9 m@xmath7 progenitor and may explain the discrepancy .
we note , however , that other kinds of instabilities can lead to stellar mergers ( e.g. , in triple systems ) and that the above comparison is valid only for the specific case discussed by soker & tylenda ( 2006 ) .
existing hydrodynamical simulations of the common envelope phase in stellar mergers ( and also star @xmath10 neutron star mergers ) predict that the total dissipated energy is of the order of that observed in v838mon and m85ot2006 - 1 ( e.g. , taam & bodenheimer 1989 ; terman et al . 1995 ; terman & taam 1996 ) .
moreover , simulations of the common envelope phase predicts that most of the envelope will be ejected in the equatorial plane ( e.g. , taam & ricker 2006 ) .
indeed , in rau et al .
( 2007 ) we reported evidence suggesting that the expansion of m85ot2006 - 1 is asymmetric .
however , more detailed hydrodynamical simulations of the vast parameter space available for stellar mergers are needed in order to understand these processes and to test if v838mon - like objects are indeed the results of stellar mergers . to summarize ,
we show that , in contrast to v838mon , but similarly to m31rv , m85ot2006 - 1 was probably produced by members of an old stellar population ( @xmath81gyr ) , and that its progenitor / s mass was probably @xmath82m@xmath7 .
these constraints narrow down the allowed venue of stellar models for the nature of this event .
|
m85 optical transient 2006 - 1 ( m85ot2006 - 1 ) is the most luminous member of the small family of v838 mon - like objects , whose nature is still a mystery .
this event took place in the virgo cluster of galaxies and peaked at an absolute magnitude of @xmath0 . here
we present hubble space telescope images of m85ot2006 - 1 and its environment , taken before and after the eruption , along with a spectrum of the host galaxy at the transient location .
we find that the progenitor of m85ot2006 - 1 was not associated with any star forming region .
the @xmath1 and @xmath2-band absolute magnitudes of the progenitor were fainter than about @xmath3 and @xmath4 mag , respectively .
therefore , we can set a lower limit of @xmath5myr on the age of the youngest stars at the location of the progenitor that corresponds to a mass of @xmath6 m@xmath7 .
previously published line indices suggest that m85 has a mean stellar age of @xmath8 gyr .
if this mean age is representative of the progenitor of m85ot2006 - 1 , then we can further constrain its mass to be less than @xmath9 m@xmath7 .
we compare the energetics and mass limit derived for the m85ot2006 - 1 progenitor with those expected from a simple model of violent stellar mergers .
combined with further modeling , these new clues may ultimately reveal the true nature of these puzzling events .
|
since the result of andersson @xcite and friedman & morsink @xcite , numerous works have dealt with inertial modes of rotating relativistic stars ( see reviews by friedman & lockitch @xcite , andersson @xcite and kokkotas & ruoff @xcite ) . in perfect fluids ,
these modes have indeed been proven to satisfy the chandrasekhar - friedman - schutz ( cfs ; @xcite and @xcite ) criterion for instability whatever the angular velocity of the star is .
hence , due to their coupling to the gravitational field , these oscillations make neutron stars ( nss ) very promising sources of gravitational waves ( gw ) . yet , as nss are complex relativistic objects , far from perfect fluids , the astrophysical relevance of the instability of inertial modes in rotating nss is still an open issue .
the final answer to this question will depend on the many physical phenomena that take place within nss : their initial state at birth @xcite , existence of differential rotation @xcite , of a crust ( @xcite , @xcite , @xcite , @xcite ) , of a huge magnetic field @xcite , of superfluid states of nucleons ( @xcite , @xcite , @xcite ) , of exotic particles , _ etc_. understanding better that potential instability is important for two main reasons .
first , if it were proven to be relevant , it could have a great impact on our idea about the evolution of nss and might explain the relatively slow angular velocity of observed pulsars .
moreover , in order for the data collected by gw detectors like virgo or ligo to be useful , they have to be compared with predicted signals , such as the signal that results from the instability of inertial modes .
since theoretical predictions depend on the models used for the inner structure of nss , this confrontation could possibly be as instructive for nuclear physics at high density as was helioseismology for neutrino physics .
+ in a previous article @xcite ( referred to here as paper i ) , we have presented a spectral hydrodynamical code , which uses spherical coordinates , with the aim of studying time evolution of inertial modes in slowly rotating neutron stars . having discussed the motivation for the slow rotation approximation in this previous article ( it is still good even for the fastest known pulsars , see paper i ) , we keep working with it in the present paper . thus , the next step in our project was to improve the linear study of relativistic inertial modes by trying to take into account in a more realistic way the microphysical conditions that occur inside actual nss . among them , we retained the stratification of the star .
it was quite easy to implement , and it is well - known that inertial modes change drastically depending on whether the equation of state ( eos ) that describes the ns is barotropic or nonbarotropic depends only on one variable , for instance the baryonic number density @xmath1 . ] , while the status of purely axial modes of relativistic nonbarotropic nss is still under debate ( @xcite,@xcite ) .
in addition , it is easy to show that , for the most basic model of a star built with the so - called `` @xmath2 '' matter , the assumption of a barotropic eos , even if easier to use , is wrong as soon as the star is cold enough and the matter not in a stationary state .
hence , here we consider the time evolution of inertial modes in slowly rotating relativistic nss composed of neutrons , protons and electrons , with an `` out - of - equilibrium '' eos .
it should be kept in mind that in the following , we shall call `` nonbarotropic '' a peculiar type of nonbarotropic eoss : the situation of @xmath2 matter with frozen composition . + the plan of the paper is as follows . in section [ sec : bg ] , we start with an overview of the basic assumptions made in the study of the structure of cold neutron stars , with emphasis on the beta equilibrium between neutrons , protons and electrons , which constitute , for not too massive nss , the most important part for studies of their oscillations : their core .
we end this section with a description of our calculation of stationary rotating nss and of the eos used to get configurations to perturb in the following sections : the prakash - ainsworth - lattimer ( pal ) eos @xcite . in section [ sec : pertmicro ] , a short summary of the microphysics of `` perturbed '' nss composed of @xmath2 matter is given . in this section ,
we justify the assumption of frozen composition , which is then used in section [ sec : perthydro ] that describes the basis of the linear hydrodynamical study .
the main feature of a nonbarotropic star is the existence of a nondegenerate spectrum for its polar modes , even when it is not rotating . as frozen composition results in a nonbarotropic effective eos ,
these so - called gravity modes ( _ g_-modes ) are studied for the pal eos in section [ sec : gmod ] .
after which , in section [ sec : rotmod ] , we describe inertial modes of rotating nss with the pal eos but without the ( correct ) hypothesis of frozen composition , before ending with gravity and inertial modes in rotating nss taking into account the frozen composition .
finally , the results are discussed in the conclusion .
the outer layers of cold nss are known to play key roles in several physical phenomena , as the early and late cooling of pulsars @xcite , the existence of glitches ( @xcite , @xcite , @xcite ) and the spectra of x - ray bursts @xcite . yet , since more than @xmath3 of the mass of cold nss is included in their core , outer layers are not relevant at the most basic levels of studies of both the structure of rotating relativistic stars and their oscillations .
thus , in the following , a ns is modeled by its core , and the outer layers are only discussed as time independent boundary conditions for the oscillations ( see section [ sec : gmod ] ) .
+ the minimal nontrivial description of a core of ns is a mixture of neutron , protons , and electrons at a temperature that is sufficiently low for the matter to be degenerate but sufficiently high for the nucleons not to be superfluid .
such conditions , which imply that thermal contributions to the eos can be neglected , are satisfied for @xmath4 , at densities above one tenth of normal nuclear density @xmath5 .
moreover , when a ns has cooled down enough for these assumptions to be valid , the matter has become transparent for neutrinos , which therefore do not contribute to the eos either . + another physical constraint to take into account for the study of both unperturbed and perturbed stars is the electrical charge neutrality , which comes from the very short characteristic time scale of electromagnetic processes . in degenerate
@xmath2 matter , the so - called plasma frequency of electrons , whose inverse give the typical time for the plasma to maintain charge neutrality , is @xmath6 where @xmath7 is the chemical potential of electrons similar to @xmath8 mev ( see below ) , with the electron density @xmath9 .
this leads to an electromagnetic relaxation time , which implies that , for our purpose ( phenomena with typical time scales of some ms ) , matter can always be considered as neutral , and there is equality between the electron and proton densities @xmath10 .
this relation also makes thermodynamical quantities functions of nucleon ( baryon ) density @xmath11 and proton fraction @xmath12 only .
furthermore , at baryon densities of the order of @xmath13 , nucleons form a strongly interacting fermi system and matter is very rich in neutrons , while we can replace nucleon masses ( which are assumed to be equal ) by the neutron mass @xmath14 mev .
if we neglect now the coulomb interactions between charged constituents of matter and treat the electrons as a free ultrarelativistic fermi gas , the energy per nucleon ( excluding rest energy of nucleons ) can be written @xmath15 where @xmath16 is the nucleon contribution , and the electron term @xmath17 is @xmath18 with @xmath19 .
+ extensive many - body calculations with realistic two - nucleon and three - nucleon forces show that in the whole range of @xmath20 , the energy per nucleon is ( to a very good approximation ) quadratic in the neutron excess parameter @xmath21 ( @xcite , @xcite ) . at @xmath22 , characteristic value for terrestrial nuclear physics ,
this is just the consequence of the charge symmetry of strong - interaction . at @xmath23 ,
this is a numerical fact .
therefore , the @xmath24-dependence factors out , and @xmath16 is given by the formula @xmath25 here , @xmath26 is the energy per nucleon of the symmetric nuclear matter ( @xmath27 ) , and @xmath28 is the _ symmetry energy _ at nucleon density @xmath1 .
many possibilities exist for these two functions , among them those that form the pal eos @xcite .
this eos , which is retained in this study , will be described in more detail with the presentation of the obtained background stars used for our modes calculations ( section [ sec : bgm ] ) .
however , to be able to calculate these unperturbed stars , another relation is needed to close the system of equations .
this relation comes from the hypothesis of beta equilibrium between neutrons , protons and electrons .
for matter transparent to neutrinos , the equilibrium with respect to the weak - interaction beta - processes @xmath29 implies a relation between the chemical potentials of the matter constituents , @xmath30 , with ( n_b , x_p)=n_b(m_nc^2+e ) the total energy density [ energy per baryon @xmath31 that comes from eq.([e : e.ne ] ) plus nucleon rest energies ] , and @xmath32 . for matter transparent to neutrinos , this relation reads @xmath33 with our quadratic form of the @xmath24-dependence of @xmath16 , eq.([e : en.w0.s ] ) , the beta equilibrium condition involves only the nuclear symmetry term .
we have @xmath34 the resulting fraction of protons for matter in beta equilibrium , denoted by @xmath35 , is determined from @xmath36 or [ e : beteqy ] y^3+y-=0 , where @xmath37 and @xmath38 .
+ the last equation admits analytical solutions ( _ cf .
_ cardan s method ) among which the only real one is [ e : xbeta ] x_p^eq(n_b)- with @xmath39 .
+ then , the barotropic eos in beta equilibrium , which is used for the calculation of unperturbed stars , is obtained from [ e : rho.bar.eos ] _
eq(n_b)=(n_b,(n_b ) ) , with the total pressure [ e : p.bar.eos ] p_eq(n_b)=n_b^2 .
|_x_p . note that when the chemical potential of the electrons is above the rest mass of muons , , taking into account the appearance of the latter via the equilibrium between the processes e^-^-+_e+| _ ^-e^-+|_e+ _ , would have implied higher proton fractions .
but as a rule , the role of muons is practically irrelevant , as far as the eos is concerned , and we shall always neglect them .
the numerical solution of the einstein s equations for a cold stationary rotating relativistic star with a barotropic eos , @xmath40 [ reached by eq.([e : rho.bar.eos ] ) and ( [ e : p.bar.eos ] ) ] , has now become a classical problem ( see @xcite for a review ) whose detail we will not enter into here .
let us just remind the reader of the main assumptions : - : : there are two killing vectors @xmath41 ( timelike at spacelike infinity ) for stationarity and @xmath42 ( with closed orbits , spacelike everywhere except on the so - called rotation axis where it vanishes ) for axisymmetry ; - : : spacetime is asymptotically flat , the metric @xmath43 `` tends toward '' the flat minkowski metric @xmath44 at infinity ; - : : matter is a perfect fluid whose energy - momentum tensor is [ e : perfflu ] * t*(+p ) uu+p*g * , where @xmath45 is the @xmath46-velocity of the fluid .
obviously , @xmath47 and @xmath48 are measured in the frame comoving with the fluid and thus are the equilibrium functions obtained in the previous section ; - : : the velocity field corresponds to rigid axisymmetric rotation and can be written [ e : dec_u ] u = (+ ) , where @xmath49 is by definition the angular velocity of the star and @xmath50 a scalar field such as the norm of @xmath45 is @xmath51 ( with our convention for the signature ) . with these assumptions ,
the energy - momentum tensor satisfies the condition of circularity @xcite and the _
maximal slicing quasi - isotropic _ coordinates @xcite can be adopted to write the metric as [ e : met_msqi ] s^2 g _
x^ x^ - (n^2-n_n^ ) t^2 - 2n _ t + (r^2+r^2 ^ 2+b^4r^2 ^2 ^2 ) , where we have introduced the notation of the @xmath52 formalism : @xmath53 is the lapse and @xmath54 the third component of the shift @xmath55-vector , with the covariant @xmath56-component of the latter .
+ because of the existence of two killing vectors all the functions only depend on ( @xmath57 ) , and this choice of coordinates reduces the einstein s equations to a system of elliptic equations that we solve with the fully relativistic spectral code presented in bonazzola _
this code also provides us with what will be the velocity field to perturb in the study of oscillations ( section [ sec : perthydro ] ) . indeed the @xmath50 scalar field , which normalizes the velocity ,
can be shown to satisfy [ e : lbd ] = -(|ps .
|)^- , where @xmath58 denotes the scalar product associated with the metric . in order to study inertial modes of nss more realistic than simple barotrops ,
we need an eos for the nuclear matter that is also valid when the fluid is not in beta equilibrium ( [ e : beta.proc ] ) .
indeed , as explained in the introduction and as calculations will show in section [ sec : pertmicro ] , a lump of matter transported out of its equilibrium position by an oscillation will not adjust its composition to its new environment instantaneously .
this will require some time , which has to be taken into account in the dynamics since it allows new modes ( see section [ sec : perthydro ] ) to exist .
+ among the numerous eoss that describe nuclear matter in nss , we decided to use the equation proposed by prakash _
@xcite which is among the few which are valid also out of beta equilibrium .
moreover , this analytical eos based partly on the experimental properties of nuclear matter , and partly on the results of many - body calculations of asymmetric nucleon matter , is in fact a set of eoss .
indeed , several possibilities are given for the choice of the symmetry energy function @xmath28 , while the energy per nucleon of the symmetric nuclear matter @xmath26 depends on some parameters .
thus , following prakash _ et al .
_ , we have [ e : pal1 ] w_0(n_b)e_f^(0)u^ + au+ + 3_i=1,2c_i()^3 (- ) , where - : : @xmath59 is the free nucleon gas fermi energy at saturation density ( @xmath60 mev ) ; - : : the dimensionless variable @xmath61 is @xmath62 , where is the saturation density of symmetric nuclear matter ; - : : the second and third terms , that depend on the constants @xmath63 and @xmath64 , reproduce the static interactions between nucleons and are chosen in such a way that the eos is causal ; - : : the last two terms , that depend on the momentum , mimic the dynamical part of the strong interaction . in those terms , the @xmath65 parameters are characteristic lengths associated to coupling constants @xmath66 .
the first term ( such as @xmath67 and @xmath68 ) is attractive at large distances , whereas the second ( @xmath69 with @xmath70 ) reproduces the repulsive behavior of the strong interaction at short scales . in order to separate the kinetic and potential contributions in the symmetry energy , prakash _ et al .
_ @xcite write s(n_b)=(2 ^ -1)e_f^(0 ) (u^-f(u))+s_0f(u ) , where @xmath71 is a function which describes the ( poorly known ) density dependence of @xmath72 . to get
, one has @xmath73 .
+ playing with the form of the functions @xmath26 and @xmath28 enables one to change the values of the compression modulus and the symmetry energy .
however , this analytical eos , that fits known results around the saturation density , has the drawback that , at low densities , it is no longer physical .
indeed , to be realistic at low densities , any eos has to take into account the appearance of nuclei .
hence , for several choices of the key quantities ( such as the value of the compression modulus and the symmetry energy ) within the proposal of prakash _ et al .
_ , pathological features like negative pressure or convective instability appear .
moreover , even for the eoss that do not present these awkward characteristics , the low density part is unphysical and some prescription has to be adopted for the study of modes . + among the functions and the set of parameters that prakash _ et al . _ propose , we retain @xmath74 and take values of the parameters that give a compression modulus of either 180 mev ( model a ) or 240 mev ( model b ) .
in addition , to prevent the troubles linked to low densities , our choice is to _ cut the star _ below a given threshold density . as discussed in section [ sec : bg ] , since the mass of the crust is quite small , this cutoff does not really affect either the metric or the main properties of the star . moreover , adding some boundary conditions on the surface described by the value of the density chosen for the cutoff is an easy way to mimic the phase transition that occurs between the outer - core and the inner - crust .
this point will be discussed in more detail in the sections that deal with modes ( sections [ sec : perthydro ] , [ sec : gmod ] , [ sec : rotmod ] ) but it was noticed now to clarify table [ tab : bg ] . indeed , in this table are summarized the background configurations that we use .
but in order to calibrate the effect of the cutoff on the modes , we tried several values for the density threshold .
hence , in table [ tab : bg ] , we show structural properties of stars for the two eoss that we keep with different central densities and the values that we take for the cutoff density .
this range is consistent with realistic evaluations of the density at crust - core interface ( see @xcite and @xcite ) ] : @xmath75 or @xmath76 baryon.@xmath77 . as a first validation of the cutoff procedure
, we can check that for given eos and central density , the cutoffs imply differences in the gravitational mass of some few percent in the worst case .
finally , note also that the central ( maximal ) proton fraction is always below @xmath78 , which forbids direct urca processes for @xmath2 matter ( see also section [ sec : eosandnon - eq.proc ] ) .
[ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] the first thing that we shall look at is the influence of our cutoff density on inertial modes . to explore this , we did several time evolutions with the newtonian @xmath0 _ r-_mode as initial data ( see paper i ) .
this purely axial mode was evolved in all backgrounds with different angular velocities of the star , and a typical result is illustrated by fig.[fig : specrmcutvt ] .
this figure depicts the spectra of the @xmath79 component of the velocity for a star of model a(21 ) with an angular velocity of @xmath80 rad.s@xmath81 .
three different values for the cutoff density are used : 0.053 , 0.08 or , that is to say more or less @xmath82 , @xmath83 or @xmath84 of the saturation density @xmath13 .
moreover , in order to test the possible appearance of a continuous spectrum , we draw the spectra calculated at 2 different positions within the star .
thus , the first obvious result is that for this model at that precise angular velocity , we see a discrete spectrum , with the inertial mode ( the higher peak in the spectrum ) that seems to have a frequency around @xmath85 hz .
this value corresponds to a ratio between the relativistic and newtonian frequencies in the rotating frame which is around @xmath86 , in quite good agreement with the results of lockitch _
@xcite for that compactness .
in addition , various inertial modes can be seen that can be identified as the relativistic counterparts of the inertial modes found by lockitch & friedman @xcite , but since it was shown @xcite that none of these modes are relevant for gw with the cfs instability , we shall not describe them in detail .
+ moreover , as can be seen , the influence of the cutoff density on the spectrum of inertial modes is even smaller than for gravity modes .
this can probably be understood from the fact that starting with purely axial initial data we reach a radial velocity smaller than in the case of the _ g-_modes .
this last statement is supported by the comparison between figures [ fig : speccut ] and [ fig : rma21vtvrs ] .
both of these figures depicts spectra for time evolutions of @xmath87 modes in a star of model a(21 ) .
however , for the first one , the spectra of the @xmath79 and @xmath88 components of the velocity correspond to gravity modes in a nonrotating star [ initial data identical to the eq.([eq : denspert2 ] ) and still without velocity ] , while for the second they correspond to the newtonian inertial mode in a barotropic a(21 ) star with . notice that for both calculations the cutoff density is @xmath76 @xmath77 .
what can be seen is that the orders of magnitude of the @xmath79 component and of the radial component for the main peak are almost the same for gravity modes ( fig.[fig : speccut ] ) , whereas in the case of inertial modes ( fig.[fig : rma21vtvrs ] ) the radial velocity is more than one order of magnitude smaller than the @xmath79 component .
this feature can of course also be explained by the fact that the spectra of relativistic inertial modes mainly depend on the compactness of the star as we shall further verify .
notice in addition that , as expected , this inertial mode is not a purely axial mode .
+ the last curve of fig.[fig : rma21vtvrs ] is the spectrum of one of the two independent components of the current quadrupole tensor ( see paper i for the exact definition ) . since this tensor dominates the post - newtonian reaction force @xcite , we shall use it as an indicator of the fastest growing modes , for those that verify the cfs criterion .
this indicator tells us that the already mentioned inertial mode actually seems to be the most unstable one
. however , even if fewer modes can be seen here than in the spectra of the velocity s components , several other modes appear . among them , a peak around @xmath89 hz but also a quite hidden one around @xmath90 hz that is probably the axial - led @xmath91 mode whose stability was discussed by @xcite : it has indeed a frequency in the rotating frame that is some @xmath92 of the frequency of the @xmath0 mode , while in the newtonian case , lockitch _
_ @xcite found a ratio of @xmath93 .
notice that this argument relies on the assumption of `` almost similarity '' of the spectra calculated in the rotating frame for different stars . formulated in another way ,
it just means that the spectra in the rotating frames mainly depend on one global physical parameter , the compactness .
this assumption will be tested in the following .
anyway , this mode was shown not to be relevant for gw , which is supported by its very low presence ( compared with the main peak ) in the @xmath94 spectrum .
it was more noticeable in the velocity s spectra .
+ now that we have analysed the effect of the cutoff density , we shall be able to investigate the inertial modes for all our models . in fig.[fig
: rmallvth ] , one can see the same kind of spectra for the @xmath79 component of velocity using all our background models and changing the angular velocity from @xmath95 to @xmath96 rad.s@xmath81 , while in fig.[fig : rm80vth ] only the curves for @xmath97 rad.s@xmath81 are depicted in order to better see the differences between different stars .
thus , the first comment is that the influence of the background star on the spectra is quite small .
the higher resolution makes that this is better seen in fig.[fig : rm80vth ] , but already in fig.[fig : rmallvth ] one can observe a very small dispersion of the frequencies . moreover looking carefully , one can discern a cross inside the highest square , _ i.e. _ the square denoting the @xmath98 inertial mode for model b(24 ) .
thus , models a(26 ) and b(24 ) seem to have the same frequency for this mode , which supports the fact that more than the eos , the compactness of the star plays a key - role in the difference between newtonian and relativistic frequencies . indeed , as can be seen in tab .
[ tab : bg ] , those stars have almost exactly the same compactness : 0.189 and 0.191 respectively .
+ next , to test the dependence with respect to @xmath49 of the frequencies , we shall divide all of them by the angular velocity of the star .
since it is pointless for gravity modes , we shall restrict this analysis to the @xmath98 inertial mode , whose dimensionless frequency is illustrated versus the angular velocity for all compactness in fig.[fig : lm2.baro ] . nevertheless , before doing more comments about what turns out from this figure , it is worth to stress its limitation .
indeed , the version of the code used in this study works with a given time step of , and the total duration of the evolutions is always of the order of 1s , whilst the angular velocity changes .
hence , the error bars on frequencies expressed in hz are of about 1 hz , but for frequencies normalised with the angular velocity @xmath49 , they are equal to @xmath99 .
thus , for , this is @xmath100 and this is less than @xmath101 for .
all of this explains why we did not put the data for that were completely meaningless .
+ on this other hand , higher angular velocities of the star imply higher frequencies that give better resolution , which support the compactness dependence of the frequency as shown by fig.[fig : lm2.baro ] .
however , for all angular velocities , the frequencies of the @xmath0 mode for models with compactness @xmath102 and @xmath103 are the same . yet , our main purpose was not to study inertial modes of barotropic stars , but to try to have an idea of the influence of the microphysical conditions , which implies nonbarotropicity , on the spectra .
this is the reason why instead of working with time step and duration scaled by the inverse of @xmath49 ( as we did in paper i ) , we use here only `` physical scales '' that we shall also keep in the next section for the study of the modes of stratified rotating nss .
the first mentioned effect of rotation was the splitting of noninertial modes .
our investigation of the pulsations of rotating stratified relativistic nss will begin with the illustration of this phenomenon within the study of density perturbations in rotating stratified nss . to simplify , we shall first restrict ourselves to the model a(21 ) with various angular velocities , but keeping the cutoff density at @xmath76 @xmath77 .
in fact , since we have verified in the previous sections its low influence , we shall use in the following only that value of the cutoff density .
+ fig.[fig : dn.21.dn.om ] depicts the spectra of the lagrangian perturbation of density .
the initial data consist of the same @xmath87 perturbation of density [ see eq.([eq : denspert2 ] ) ] without any velocity as in section [ sec : gmod2 ] .
as expected , the splitting of the _ g-_modes can be seen , but in addition , we can observe the emergence of some probable inertial modes : see the peaks around and for . the main reasons why these modes can be thought to be inertial modes are that both their frequencies and their relative importance in the spectra grow with the angular velocity . yet , to better describe them , it is worth to have a look at the spectra of the velocity in figs.[fig : vth.21.dn.om ] and [ fig : vr.21.dn.om ] , where we show spectra of the @xmath79 and radial components of the velocity , respectively .
+ first of all , fig.[fig : vth.21.dn.om ] directly shows that not all modes were visible in the spectra of the density .
thus , in addition to the splitted gravity modes and to the already mentioned candidates inertial modes , there are many others that do not correspond to peaks in the density spectra .
we shall now see how we are lead to classify them as axial - led inertial modes , following the classification of lockitch _
_ @xcite . obviously , they are inertial modes since their frequencies increase with the angular velocity .
but our view of classifying them as axial - led is supported by fig.[fig : vr.21.dn.om ] , which shows that the radial velocity for those modes is very small comparatively to the radial velocity of others .
moreover , one of them even seems not to have any radial counterpart ( at this scale of precision ) : the mode with a frequency of for , that is to say the mode with the lowest frequency among those axial - led inertial modes .
+ interestingly , fig.[fig : sij.21.dn.om ] , which depicts the spectra for the current quadrupole tensor , shows that this mode is the mode with the largest contribution to this tensor .
notice anyway that its amplitude is very low ( logarithmic scale ) due to the initial data .
nevertheless , for all the preceding reasons , we can now say that this is the @xmath0 inertial mode , and that table [ tab : bornes ] tells us that its frequency is inside the range of the continuous spectrum .
indeed , its dimensionless frequency in the inertial frame is @xmath104 whereas for model a(21 ) the range is @xmath105 $ ] .
moreover , with the already displayed spectra of velocity , the spectra of a global quantity ( the current quadrupole ) , plus various spectra of velocity calculated in several positions within the star for the barotropic case . ] , we verified that this is a discrete eigenvalue .
+ finally , before we try to investigate the @xmath0 inertial mode in more detail by changing the initial data , let us introduce some axial and polar `` kinetic energies '' that shall help us to study the coupling between polar and axial parts of the modes .
what we shall call `` kinetic energy '' in the following was already introduced in paper i and is defined as _ star ^3
v ^2 , where @xmath106}^2\,a[r]$ ] [ see eq.([e : lin_bar_cons ] ) ] , @xmath107 is the three dimensional vector field ( the eulerian perturbation of velocity ) defined in eq.([e:4v_tot ] ) and @xmath108 is the flat elementary volume .
+ with the anelastic approximation , this quantity can be shown to be conserved for a barotropic eos and a vanishing density at the surface , conditions that were verified in paper i. but for nonbarotropic eoss with a cutoff density , two ways to violate this conservation exist . the first one , linked to the eos , corresponds to the existence of a `` chemical '' part of the physical energy , while the second is linked to fluxes through the surface .
to calibrate the violation of the conservation due to fluxes , we did time evolutions of this kinetic energy in barotropic stars with cutoff densities @xmath109 , @xmath75 or @xmath76 baryon.@xmath77 .
we found that for these values , the higher was the cutoff , the better was the conservation , which means that our usual cutoff at @xmath76 baryon.@xmath77 is the best of them .
notice that in all cases we saw temporary violations of the conservation , but the temporal mean value of the energy was always conserved .
moreover , even in the worst cases , the violation were never higher than @xmath110 .
on the other hand , testing the influence of the `` chemical energy '' , we verified that for nonbarotropic eoss with a cutoff at @xmath76 baryon.@xmath77 , the mean value was no longer conserved , but the violation was only of the order of some @xmath55 percents .
for all these reasons , we shall work in the following with this ( kinetic ) energy as a not too bad indicator of the energetic behavior of the oscillations , at least as far as the coupling between polar and axial parts is concerned .
+ indeed , we remind that our algorithm for the solution of euler s equations ( see paper i ) relies on the use of the radial velocity and two scalar potentials whose angular divergence and curl enable us to recover the usual spherical components . by using these variables , the separation of the polar and axial parts of the velocity field is done very easily , which naturally leads to the polar and axial kinetic energies . thus , fig.[fig : en.21.dn.om ] that displays the time evolution of the ratio between the polar kinetic energy and the total kinetic energy give a first insight on the energy flows between the polar and axial parts of the modes .
+ in this figure , time evolutions of the ratio are shown , whose duration is only @xmath111 ms in order to make visible the `` quasistationnarity '' that is reached after a while . as can be seen , starting with a purely polar perturbation , the proportion of the energy stored in polar modes is indeed decreasing down to a fraction that is between @xmath96 and @xmath112 , depending on the angular velocity of the star : the faster it rotates , the more energy flows to axial modes .
notice anyway that since the initial data are a perturbation of density without any velocity , at the very beginning , there is , rigorously speaking , no kinetic energy .
but in fact , from the very first time steps the perturbation of density implies a velocity which is initially purely polar .
moreover , we encountered quite similar results when we did some tests with as initial data a purely polar velocity field without any density perturbation . only the ratio was slightly changed , whilst the time scales were identical .
nevertheless , the relevant issue concerning energy flows will be how barotropicity affects this evolution .
this will be one of the topics of the next section .
we shall now study the time evolution of oscillating out - of - equilibrium nss with different initial data : the @xmath113 inertial mode instead of a density perturbation . in this way
, the differences with the barotropic case shall be easier to notice , while we should also be able to see how much our previous results for nonbarotropic stars depend on the initial data . to start with , fig.[fig : dn.21.rm.om ]
depicts the spectra of the lagrangian perturbation of density for this case , with the results already presented in fig.[fig : dn.21.dn.om ] for comparison .
the first thing to comment on is that even for low angular velocity , gravity modes appear , with , as expected , an amplitude that grows with @xmath49 .
but it shall also be pointed out that the @xmath98 and @xmath91 inertial modes are visible , even if we consider the density perturbation spectra , and that their amplitudes stay the same for all values of @xmath49 .
but the coupling between axial and polar parts of the velocity field should be better investigated with the spectra of the velocity components .
these are displayed in figs.[fig : vth.21.rm.om ] and [ fig : vr.21.rm.om ] .
+ fig.[fig : vth.21.rm.om ] shows that the @xmath79 component of the velocity is hardly affected by the fact that the star is a barotrop .
the main difference between the spectra is indeed only that , for nonbarotropic stars , gravity modes appear , with amplitudes that become higher for faster rotating stars .
even the frequency of the @xmath0 inertial mode is almost exactly the same .
but on the other hand , the radial part of the velocity ( fig.[fig : vr.21.rm.om ] ) is completely changed and quite strongly depends on the angular velocity , which was predictable .
thus , for slowly rotating stars , gravity modes dominate and the radial velocity of the inertial mode is some two orders of magnitude lower than in the case of a barotropic star . yet , for increasing angular velocity , the amplitude and the frequency of the radial part of the @xmath0 inertial mode become more and more similar to what they are in a barotropic star . for , there is less than one order of magnitude difference between the barotropic and the nonbarotropic cases for the @xmath98 mode ( whose frequency in the inertial frame is in this case ) , whereas there was two orders for . on the other hand ,
it should be pointed out that also the amplitude of the _ g-_modes grows with @xmath49 ( probably meaning that their `` inertial '' part is on the way to dominate buoyancy ) while many of the inertial modes that existed for barotropic stars have disappeared .
+ if we now look at fig.[fig : sij.21.rm.om ] that depicts the spectra for the current quadrupole tensor , we shall observe that from the point of view of gw emission , barotropicity seems not to have a major impact , at least in the spectrum . indeed
, the spectra of this tensor do not show remarkable differences , with only a small shift in the frequency of the probably most unstable mode that can be pointed out .
we shall now investigate in more detail how this shift in the frequency depends on the eos , after a brief comment on the time evolution of the ratio between polar and axial kinetic energies . + indeed , while all previous spectra were very comparable , the time evolutions of the polar and axial parts of the kinetic energy are quite different for barotropic and nonbarotropic stars .
the latter situation is illustrated by fig.[fig : en.21.rm.om ] ( out - of - equilibrium case ) , and by fig.[fig : en.21.baro.om ] ( barotropic case ) .
thus , one observes that , for a barotropic star , energy is fast flowing back and forth from the axial part to the polar part , leading to a kind of `` stationary state '' with a medium value of the ratio around @xmath101 . on the other hand , for nonbarotropic stars , the coupling between polar and axial parts seems to depend much more on the angular velocity .
hence , for @xmath114 or , even after @xmath115 ms of evolution , the `` final state '' is still not reached and the ratio still grows .
but the curves for higher angular velocities suggest that they should not get too high since the `` stationary situations '' obtained in those cases are characterized by ratios lower than for barotropic stars ( always less than @xmath116 ) . + the difference between the typical times is quite easy to understand .
indeed , energy flows should have for typical time scales the periods of the modes measured in the rotating frame . for nonbarotropic stars , whatever the angular velocity , some _ g_-modes exist with periods of some ms , which are then responsible for the quite fast flows .
however , in the barotropic case , only inertial modes can be observed .
since they have periods proportional to the inverse of the angular velocity , this is the explanation of the two features of the barotropic case : the self - similarity ( for different values of @xmath117 ) of the time evolutions and the acceleration of the flows with @xmath117 .
+ on the other hand , with some caution due to our numerous approximations ( and also to our definition of `` kinetic energy '' ) , we could say that the result concerning amplitude of the ratios can be seen as an indication than stratification probably helps the unstable inertial mode ( the @xmath0 _ r-_mode ) to be driven to instability by gw since its coupling to polar modes ( which are not driven to instability by the current quadrupole ) is smaller . yet , the answer to that question requires a proper treatment of the emission of gravitational waves without the cowling approximation , while different viscosities for polar and axial modes should also complicate the problem .
+ to conclude this work , we shall give an overview of how the frequency of the @xmath0 inertial mode depends on compactness , barotropicity and angular velocity .
the latter will be the easier : as already discussed , due to the error bars for the normalized frequencies , we are not really able to see any difference .
moreover , since our study is linear in @xmath49 , we expect that the normalized frequency of the inertial modes should only depend on the compactness of the stars but not on @xmath49 .
this view is supported in the barotropic case and in the nonbarotropic case by ( respectively ) both figs.[fig : lm2.baro ] and [ fig : lm2.nonbaro ] .
thus , the more compact a ns is , the higher will be its dimensionless frequency as measured by an inertial observer .
however , the comparison between these two figures shows that taking into account the fact that @xmath2 matter should keep a frozen composition , which implies a nonbarotropic eos , leads to lower frequencies .
moreover , it is worth pointing out that we verified that , for all of our models , the @xmath0 inertial modes had discrete spectra , even if fig.[fig : lm2.nonbaro ] and tab .
[ tab : bornes ] demonstrate that all these modes were found within the range of the continuous spectra .
nevertheless , as already mentioned and as can be seen in fig.[fig : vr.21.rm.om ] , they are modes with a radial velocity , which means that they are not purely axial . for very slowly rotating stars , this radial component is quite small , but it becomes as large as what it is for nonpurely axial inertial modes of barotropic stars when the angular velocity increases . finally , notice that even if models a(26 ) and b(24 ) have the same compactness and then the same frequency for the @xmath0 inertial mode , due to the mixing with gravity modes that depends on the detail of the microphysics , their full spectra are quite different .
this issue is illustrated by fig.[fig : spec.vr.comp ] that depicts the spectra of the radial component of the velocity for those models with the out - of - equilibrium hypothesis .
hence , this probably means that more precise studies including more coupling between modes ( nonlinear terms in @xmath49 or nonlinear corrections to the euler equations ) should enable us to make some differences between stars of the same compactness , and then to start to probe the inner composition of nss when gravitational waves emitted by their instabilities are detected .
inertial modes and gravity modes of slowly rotating relativistic stratified neutron stars have been investigated in the linear regime .
our study is based on a spectral three dimensional evolutionary code which has , at the moment of writing , the restrictions of using the cowling and the anelastic approximations .
however , the effect of the latter approximation is mainly to kill high frequency acoustic modes that should not be strongly coupled to inertial and gravity modes for slow rotation .
additionally , the slow rotation approximation is reasonable for observed pulsars as , even for the fastest known pulsar , the pertinent dimensionless parameter is of the order of some percents .
the validity of the cowling approximation is more subtle since without perturbations of space - time we do not directly include gravitational waves in our calculations , whereas they are one of the main motivations for this work .
nevertheless , previous studies have proven that the properties of inertial modes were not too much affected by its use @xcite .
+ another major feature of our work is that we tried to simulate as well as possible the microphysical conditions that should occur inside not too massive and not too cold neutron stars .
thus , by using a quite realistic equation of state for nuclear matter valid even when the beta equilibrium is broken , we did the first three dimensional time evolution of inertial modes in stratified relativistic stars , taking into account the frozen composition of every perturbed lump of @xmath2 matter . with this approach
, we were able to show that the coupling between polar and axial modes quite strongly increases with the angular velocity .
this happens in such a way that , in nonbarotropic stars , initially axial inertial modes develop a radial part that can be as large as the polar part of the general inertial modes of barotropic stars .
yet , several of the polar - lead inertial modes of barotropic stars do no longer exist in nonbarotropic stars .
moreover , we found that the characteristic time scale for the exchanges of energy between polar and axial modes is different for barotropic and nonbarotropic stars .
thus , the coupling of inertial modes with gravity modes in nonbarotropic stars seems to give birth to fast energy interchange between the polar and axial parts of the fluid motion .
this phenomenon could have some implications on the way viscosity or gravitational waves emission act , but it would of course need further studies to draw robust conclusions .
+ however , the comparison of inertial modes in stars with and without the assumption of frozen composition has already lead us to the conclusion that nonbarotropicity decreases the influence of compactness on the frequency of the _
r-_mode , making in addition this frequency slightly lower in a nonbarotropic star .
if further studies with different equations of state and composition support this result , it could mean that the measurement of the global parameters of a neutron star ( compactness , mass and radius if absorptions lines are observed , binary system parameters , _ etc .
_ ) in conjunction with the observation of gravitational waves emitted by some instabilities of the inertial modes would be very instructive about the inner structure of neutron stars .
but such a grail would probably need to first improve our understanding of the physics with wider explorations of the physical parameters : centrifugal terms and deformation of the star , magnetic field , differential rotation , nonlinear coupling , _ etc .
_ among them , superfluidity is one of the key phenomena , since several studies proved that gravity modes are suppressed by its existence . as a consequence ,
even if superfluidity in old neutron stars is already strongly supported by both theory and observation ( glitches ) , the detailed analysis of the gravitational wave spectra of oscillating neutron stars could be the possibility to definitely demonstrate the superfluid nature of the neutron stars content by the lack of well understood features linked with gravity modes .
we are indebted with j.a .
pons for his very critical reading of the manuscript and for his numerous comments , but it is also a real pleasure to thank n. andersson and a. nagar for many useful suggestions . finally , we would like to thank the computer department of the paris - meudon observatory for the technical assistance . this work has been supported by the eu programme ` improving the human research potential and the socio - economic knowledge base ' ( research training network contract hprn - ct-2000 - 00137 ) .
ph was partially supported within the program of european associated laboratory `` astrophysics poland - france '' ( astro - pf ) , and lv benefited from the jumelage pan - cnrs astronomy france - poland .
in this appendix , the usual basic way to derive the newtonian dispersion relation for _
g-_modes is summarized , and we briefly discuss the effect of the divergence free and anelastic approximations . the easiest way to reach the dispersion relation for gravity restored modes is based on the perturbative lagrangian approach . if we restrict ourselves to linearized equations , and write @xmath118 the displacement vector , @xmath119 the lagrangian perturbations and @xmath120 the eulerian ( linked by the symbolic relation @xmath121 ) , we have for the euler equation [ e : ee ] _ t^2=-p-- and for the mass conservation equation [ e : cons_n ] n =- nps . , where @xmath47 is the pressure , @xmath122 the gravitational potential and @xmath123 indicates the usual euclidean scalar product .
+ if we now consider only spherical nonrotating stars ( only radial dependence for @xmath122 ) and neglect the eulerian perturbation of the gravitational field ( cowling approximation ) , the nonradial part of eq.([e : ee ] ) and eq.([e : cons_n ] ) give us n = n(_r(r^2_r)- ) , in which we assumed a decomposition in spherical harmonics of the scalar variables [ @xmath124 is the index of the spherical harmonic @xmath125 and a harmonic time dependency of @xmath126 : _ t^2
-w^2 . introducing now the equilibrium @xmath127 and frozen composition ( adiabatic ) @xmath128 speeds of sound , |p|c_eq^2|n| and pc_f^2n , with the brunt - visl frequency defined as ^2||^2(c_^-2-c_f^-2 ) , we obtained the following system of equations @xmath129 with the lamb frequency defined as @xmath130 with @xmath131 the local gravitational acceleration .
+ the new variables xr^2_re^a , ype^-a , with a - r , enable to write the system as ) . a wkb - like approximation , _
r v = ikv , gives the dispersion relation [ e : disp ] k^2=(n^2-w^2)(-1 ) .
since for neutron stars we have c_fk , eq.([e : disp ] ) leads to the two asymptotic and separated branches for the _ p-_modes [ e : disp_p ] w^2=^2((kr)^2+l(l+1 ) ) and the _ g-_modes [ e : disp_g ] w^2=n^2 .
the two approximations that we shall now discuss differ by the way they treat the continuity equation ( [ e : cons_n ] ) .
the divergence free approximation consists , as indicated by its name , in the replacement of it by the condition ps .
= 0 , while the anelastic approximation neglects the time variation of the eulerian perturbation of density , which gives ps .
(n)=0 . in the case of the anelastic approximation ,
some algebra shows that the system ( [ e : sys ] ) is replaced with which leads to the relation ( [ e : disp_g ] ) as the unique solution : the _ p-_modes have been `` killed '' by the anelastic approximation .
+ on the other hand , it turns out that the same exercise with the divergence free approximation also gives the relation ( [ e : disp_g ] ) as the unique solution , but the vector on which the wkb approximation is done has to be defined as v ( ) . this small difference in the definition of the second component of the vector between the divergence free approximation and the `` exact solution '' seems to indicate that the anelastic approximation ( for which there is not such a difference ) is closer to the exact solution than the divergence free approximation is .
k. d. kokkotas and j. ruoff , review in _ 2001 : a relativistic spacetime odyssey _ , proceedings of the 25th johns hopkins workshop on current problems in particle theory , edited by i. ciufolini , d. dominici and l. lusana ( world scientific , new jersey ) , 325 ( 2003 ) .
|
with ( nonbarotropic ) equations of state valid even when the neutron , proton and electron content of neutron star cores is not in beta equilibrium , we study inertial and composition gravity modes of relativistic rotating neutron stars .
we solve the relativistic euler equations in the time domain with a three dimensional numerical code based on spectral methods , in the slow rotation , relativistic cowling and anelastic approximations .
principally , after a short description of the gravity modes due to smooth composition gradients , we focus our analysis on the question of how the inertial modes are affected by nonbarotropicity of the nuclear matter . in our study , the deviation with respect to barotropicity results from the frozen composition of nonsuperfluid matter composed of neutrons , protons and electrons , when beta equilibrium is broken by millisecond oscillations .
we show that already for moderately fast rotating stars the increasing coupling between polar and axial modes makes those two cases less different than for very slowly rotating stars .
in addition , as we directly solve the euler equations , without coupling only a few number of spherical harmonics , we always find , for the models that we use , a discrete spectrum for the @xmath0 inertial mode .
finally , we conclude that , for nonbarotropic stars , the frequency of this mode , which is our main focus , decreases in a non - negligible way , whereas the time dependence of the energy transfer between polar and axial modes is substantially different due to the existence of low - frequencies gravity modes .
* u * ps . * * ( ( ) )
|
the _ suffix array _ ( ) , the _ longest common prefix array _ ( ) and the _ burrows - wheeler transform _
( ) are data structures with many important applications in stringology @xcite .
the array of a string contains the lengths of the longest common prefixes of the suffixes pointed to by adjacent elements of the suffix array of the string @xcite .
the most immediate utility of the is to speed up suffix array algorithms and to simulate the more powerful , but more resource consuming , suffix tree .
indeed , the array , combined with or , simplifies the algorithms for some applications such as the rapid search for maximal exact matches , shortest unique substrings and shortest absent words @xcite .
real world situations may present us with datasets that are not a single string but a large collection of strings , examples being corpora of web pages or the data coming from `` next - generation '' dna sequencing ( ngs ) technologies .
it is common for the latter in particular to generate aggregations of hundreds of millions of dna strings ( sometimes called `` reads '' ) , leading to a need for algorithms that deal not only with the collective nature of the data but also with its large overall size . in this paper
we give an algorithm for the construction of array that addresses both these requirements .
the importance of the array has meant that its construction has been well - studied in the literature .
for instance , there are algorithms that work in linear time and in internal memory ( cf .
@xcite ) . other algorithms work in semi - external memory ( see for instance @xcite ) or directly via ( see @xcite ) . in @xcite , the first algorithm establishing that array can be computed in @xmath0 complexity in the external memory model ( i.e.
, the complexity of sorting @xmath1 integers in external memory ) is presented .
recently , in @xcite an external memory algorithm to construct the suffix array of a string based on the induced sorting principle is provided .
such an algorithm can be augmented to also construct the array .
the overhead in time and i / o operations for this extended algorithm over plain suffix array construction is roughly two . in another recent paper @xcite
, the authors introduced an external memory array construction algorithm for a string .
the main idea in @xcite is to divide the string into blocks that are small enough to fit in ram and then scan the rest of the string once for each block .
such a strategy needs @xmath2 bytes of disk space , where @xmath1 is the length of the string .
one s initial strategy for building the array of a collection might therefore be to concatenate its members into a single string and then apply one of the above methods .
however , some stumbling blocks become apparent , and it is not immediate how these single - string approaches can be adapted to circumvent them .
first , many existing algorithms for computing the array require data structures of size proportional to the input data to be held in ram , which has made it impractical to compute the array of a very large string .
moreover , by definition , the values in array should not exceed the lengths of the strings , so one would ideally employ distinct end - marker symbols to act as separators between the strings of the collection
. however , assigning a different end - marker to each string is not feasible when the number of strings in the collection is very large . on the other hand
, the use of the same end - marker symbol throughout the collection could give rise to values with the undesirable properties of sometimes exceeding the lengths of the strings and having dependencies on the order in which the strings are concatenated .
in the literature , the problem of the computation of array for a collection of strings has been considered in @xcite and in @xcite . moreover , a preliminary version of the results presented in this paper is in @xcite . in particular , this paper includes full proofs of theorems from @xcite and more detailed examples .
furthermore , additional experimental results are also described .
finally , in this paper we introduce a simple solution for dealing with strings having different lengths , allowing to work on any collection of strings . defining @xmath3 and @xmath4 as the sum of the lengths of all strings and the length of the longest string in the collection respectively
, the approach in @xcite requires @xmath5 time , but the @xmath6 bits of internal memory needed to store the collection and its in internal memory prevents the method from scaling to massive data .
the external memory algorithm ( called ) in @xcite builds both suffix and arrays for a collection of strings .
such interesting strategy has an overhead in working disk space , so it seems that can not be used for huge collections of strings .
the methodology presented in this paper attempts to overcome the above mentioned limitations .
in fact , the aim of our paper is to design a lightweight algorithm for the computation , at the same time , of the longest common prefix array and the burrows - wheeler transform of a very large collection of strings having different or same length .
the computation is realized by performing disk data accesses only via sequential scans .
the total disk space usage never needs more than twice the output size , excluding the disk space required for the input . in our approach , we directly compute both data structures , without needing to concatenate the strings of the collection and without requiring pre - computed auxiliary information such as the suffix array of the collection . in order to do this ,
our algorithm is built upon the approach introduced in @xcite related to an extension of the _ burrows - wheeler transform _ to a collection of strings defined in @xcite .
in particular , our algorithm ( called ) adds to the strategy in @xcite ( called ) some lightweight data structures , and allows the simultaneous computation of both the longest common prefix and the burrows - wheeler transform of a collection of @xmath7 strings .
such a computation is performed in @xmath8 bits of memory , with a worst - case time complexity of @xmath9 , where @xmath10 is the time taken to sort @xmath7 integers , @xmath11 is the size of the alphabet , @xmath3 is the sum of the lengths of all strings and @xmath4 is the length of the longest string .
note that needs at most @xmath12 bits of disk space and requires @xmath13 disk i / o operations , where @xmath14 is the disk block size .
the low memory requirement enables our algorithm to scale to the size of dataset encountered in human whole genome sequencing datasets : in our experiments , we compute the and of collections as large as @xmath15 million strings having length @xmath16 .
moreover , allows to compute also the suffix array of the strings of a collection ( called _ generalized suffix array _ ) .
such a further computation increases the number of i / o operations , but it does not need further data structures in internal memory . section [ sec : prel ] gives preliminaries that we will use throughout the paper , in section [ sec : ads_collection ] we define the main data structures for a collection of strings , section [ bcrstrategy ] describes an incremental strategy for computing the of a collection of strings of any length .
section [ sec : algorithm ] shows the sequential computation of the array for the collection .
we present details on the efficient implementation of the algorithm and its complexity in sections [ sec : implementation ] and [ sec : bcrlcp_complexity ] , respectively .
computational results on real data are described in section [ sec : experiments ] .
section [ sec : conclusion ] is devoted to some conclusions .
let @xmath17 be a finite ordered alphabet with @xmath18 , where @xmath19 denotes the standard lexicographic order .
we append to a finite string @xmath20 an end - marker symbol @xmath21 that satisfies @xmath22 .
we denote its characters by @xmath23 , w[2],\ldots , w[k]$ ] , where @xmath24 is the _ length _ of @xmath25 , denoted by @xmath26 .
note that , for @xmath27 , @xmath28\in \sigma$ ] and @xmath29=\$ \notin \sigma$ ] .
a _ substring _ of a string @xmath25 is written as @xmath30 = w[i ] \cdots w[j]$ ] , with a substring @xmath31 $ ] being called a _ prefix _ , while a substring @xmath32 $ ] is referred to as a _
suffix_. a range is delimited by a square bracket if the correspondent endpoint is included , whereas the parenthesis means that the endpoint of the range is excluded .
we denote by @xmath33 the collection of @xmath7 strings of length at most @xmath4 .
we suppose that to each string @xmath34 is appended an end - marker symbol @xmath35 smaller than @xmath36 , and @xmath37 if @xmath38 .
let us denote by @xmath3 the sum of the lengths of all strings in @xmath39 . for @xmath40
, we refer to the suffix @xmath41 $ ] of a string @xmath34 as its _
@xmath42-suffix _ ; the _ 0-suffix _ of @xmath34 contains @xmath35 alone .
the length of a @xmath42-suffix is equal to @xmath42 ( up to considering the end - marker ) .
let us denote by @xmath43 the collection of the @xmath42-suffixes of all the strings of @xmath39 .
in our algorithm presented in this paper we use a unique end - marker @xmath44 for all strings in @xmath39 , because we set @xmath45 < w_t[|w_t|]$ ] if and only if @xmath46 , so that if two strings @xmath47 and @xmath48 share the @xmath42-suffix , then @xmath49 < w_t[|w_t|-j,|w_t|]$ ] if and only if @xmath46 .
however , to ease the presentation distinct end - markers are shown .
we say that the symbol @xmath50 $ ] is _ associated with _
the @xmath42-suffix of @xmath34 for @xmath51 , because @xmath50 $ ] precedes the @xmath42-suffix of @xmath34 , i.e. the suffix @xmath41 $ ] .
moreover , we assume that @xmath52=\$_i$ ] is associated with the @xmath53-suffix of @xmath34 , i.e. @xmath54 $ ] .
suffix array , longest common prefix array and burrows - wheeler transform are all commonly defined with reference to a single string .
this section describes the extension of such notions to a collection of strings .
the _ suffix array _ of a string @xmath25 is an array containing the permutation of the integers @xmath55 that arranges the starting positions of the suffixes of @xmath25 into lexicographical order .
there exist some natural extensions of the suffix array to a collection of strings ( see @xcite ) .
we define the _ generalized suffix array _ @xmath56 of the collection @xmath33 as the array of @xmath3 pairs of integers @xmath57 , corresponding to the lexicographically sorted suffixes @xmath58 $ ] , where @xmath59 and @xmath60 .
in particular , @xmath61=(t , j)$ ] is the pair corresponding to the @xmath62-th smallest suffix of the strings in @xmath39 , i.e. to the suffix @xmath58 $ ] .
the _ longest common prefix array _ of a string contains the lengths of the longest common prefixes of the suffixes pointed to by adjacent elements of of the string @xcite .
the _ longest common prefix array _ of a collection @xmath39 of strings , denoted by @xmath63 , is an array storing the length of the longest common prefixes between two consecutive suffixes of @xmath39 in the lexicographic order . for every @xmath64 ,
if @xmath65=(p_1,p_2)$ ] and @xmath66=(q_1,q_2)$ ] , @xmath67 $ ] is the length of the longest common prefix of suffixes starting at positions @xmath68 and @xmath69 of the words @xmath70 and @xmath71 , respectively .
we set @xmath72=0 $ ] . for @xmath73 , a _ range minimum query _
@xmath74 on the interval @xmath75 $ ] in the array returns an index @xmath24 such that @xmath76 = \min \{lcp[l ] : i \leq l \leq j\}$ ] . if @xmath77=(p_1,p_2)$ ] and @xmath78=(q_1,q_2)$ ] , it is not difficult to show that the length of the longest common prefix between the suffixes starting at positions @xmath68 and @xmath69 of the words @xmath70 and @xmath71 corresponds to @xmath79 $ ] .
the suffix array of a string is related to the _ burrows - wheeler transform _ introduced in @xcite .
the original burrows and wheeler transform ( ) on a string is described as follows : given a word @xmath20 , the output of is the pair @xmath80 obtained by lexicographically sorting the list of the conjugates of @xmath25 . in particular , @xmath81 is the word obtained by concatenating the last symbol of each conjugate in the sorted list and @xmath82 is the position of @xmath25 in such a list .
for instance , if @xmath83 then @xmath84 is not appended to the input string @xmath25 .
] . actually , in several implementations of the , in order to improve the efficiency of the computation , one can consider a variant of the by the sorting the suffixes of @xmath85 rather than the conjugates of @xmath25 . to ensure the reversibility of the transform
, one needs to append the symbol @xmath21 at the end of the input string @xmath86\cdots w[k-1]$ ] , where @xmath28 \in \sigma$ ] , @xmath87 and @xmath88 .
hence the @xmath21 symbol is at the position @xmath24 of @xmath25 , so that @xmath86\cdots w[k-1 ] w[k]$ ] . in this variant ,
the output @xmath89 is a permutation of @xmath25 , obtained as concatenation of the letters that ( circularly ) precede the first symbol of the suffix in the lexicographically sorted list of its suffixes : for @xmath90 , @xmath91=w[sa[i]-1]$ ] ; when @xmath92=1 $ ] , then @xmath91=\$$ ] ( it wraps around ) . in other words , the @xmath93-th symbol of
the is the symbol just before the @xmath93-th suffix .
for instance , if @xmath94 then @xmath95 . note that , in this case , the second output of the ( the index @xmath82 ) is not useful , because one can use the position of the @xmath21-symbol for recovering the input string . the burrows - wheeler transform can be extended to a collection of strings . in its original definition @xcite
( see also @xcite ) , such a reversible transformation ( called ) produces a string that is a permutation of the characters of all strings in @xmath39 and it does not make use of any end - marker . the of a collection @xmath39 is a word ( denoted by @xmath96 ) ) obtained by letter permutation of the words in @xmath39 together a set of indexes ( denoted by @xmath97 ) used to recover the original collection . in particular , @xmath96 is obtained by concatenating the last symbol of each element in the sorted list of the conjugates of the words in @xmath39 .
the sorting exploits an order relation defined by using lexicographic order between infinite words .
for instance , if @xmath98 , the output of of @xmath39 is the couple @xmath99 . in this paper we use a more efficient variant of of a collection of strings , that needs to append a different end - marker to each string of @xmath39 . in this case the word obtained as output , denoted by @xmath100 , is obtained by concatenating the symbols just preceding each suffix of the list of the suffixes of the words in @xmath39 in according with the lexicographic order .
the output @xmath100 can be also defined in terms of the generalized suffix array of @xmath39 . in particular , if @xmath101=(t , j)$ ] then @xmath102 = w_j[(t-1)]$ ] ; when @xmath103 , then @xmath102=\$_j$ ] . by using the above example , @xmath104 .
the external memory methods for computing the output @xmath100 are given in @xcite .
furthermore , in practice , such methods use a unique end - marker rather than @xmath7 different end - markers . note that @xmath100 differs , for at least @xmath7 symbols , from @xmath105 applied to the string obtained by concatenating all strings in @xmath39 .
our approach for computing the array is built upon the algorithm introduced in @xcite to compute @xmath100 , where @xmath39 is a collection of strings .
note that in @xcite is described for collections of strings of fixed length . in this section
we focus on the description of on collections of strings of any length .
the algorithm @xcite computes the @xmath100 of the collection of strings @xmath106 without concatenating the strings . in the sequel
, we assume that @xmath107 .
note that we consider the symbol @xmath35 appended to each string @xmath34 .
we suppose that @xmath4 is the maximal length ( including the end - markers ) of the strings in @xmath39 and @xmath3 the sum of their lengths ( including the end - markers ) .
we assume that @xmath108 . in subsection [ subsec : bcr_algo ]
we describe how the algorithm works on a collection of strings of any length , subsection [ subsec : bcr_ds ] is devoted to detail the involved data structures needed to allow a lightweight implementation and to reduce the i / o operations .
the basic idea of the algorithm is to scan all the strings @xmath109 in the collection @xmath39 from right to left at the same time .
this means that , at each iteration , it considers a `` slice '' of ( at most ) @xmath7 characters from the strings in the collection
. builds incrementally , via @xmath4 iterations , the burrows - wheeler transform of @xmath39 by simulating , step by step , the insertion of all suffixes having the same length in the list of sorted suffixes . at the end of each iteration @xmath110 ,
algorithm builds a partial @xmath100 ( denoted by @xmath111 ) .
it is referred to as partial , because if one inserts the @xmath21-characters in their correct position ( rather than the symbols that precede the @xmath42-suffixes ) , then one immediately obtains the @xmath112 of all @xmath42-suffixes of @xmath39 ( if the length of some string is greater than @xmath42 ) .
for instance , if @xmath113 ( including the distinct end - markers ) , @xmath114 corresponds to the computation of @xmath115 when the end - markers are inserted .
the key point of this strategy is to establish the positions where the new symbols associated with the @xmath42-suffixes must be inserted in @xmath116 in order to obtain @xmath111 . in other words
, we have to find the position of each @xmath42-suffix in the list of sorted @xmath117-suffixes with @xmath118 , without explicitly computing such a list . in particular , at the step @xmath119
, we have to consider the symbols associated with the @xmath120-suffixes and establish how they must be concatenated in order to obtain @xmath121 . since we use ( implicit ) distinct end - markers and @xmath122 if @xmath38 , then it is easy to verify that @xmath121 is obtained by considering the last symbol of each string @xmath34 and by concatenating them in the same order as the strings appear in the collection : @xmath123w_1[|w_1| - 1 ] \cdots w_{m-1}[|w_{m-1}| - 1]$ ] . at the iteration @xmath124
, we need to retain @xmath116 , keep track of the positions within it of the symbols @xmath125 $ ] associated with the @xmath126-suffixes of @xmath39 and build @xmath111 by inserting at most @xmath7 new symbols into @xmath116 , i.e. @xmath50 $ ] for @xmath127 or @xmath21 for @xmath128 .
such operations simulate the insertion of the @xmath42-suffixes into the list of sorted suffixes computed in the previous step . note that if @xmath129 then the string @xmath34 is not considered .
the process ends at the step @xmath130 , when considers the @xmath42-suffixes of the strings @xmath34 with @xmath131 and inserts the end - markers of such strings into @xmath132 , in order to obtain @xmath133 , i.e. the @xmath100 . in order to find , at each step @xmath134 , the positions where the new ( at most ) @xmath7 symbols must be inserted into @xmath116 , we use the notions related to backward search , such as the table @xmath135 and the @xmath136 function that have been extensively used in fm - index ( see @xcite ) .
formally , given a string @xmath25 , @xmath137 $ ] is a table of @xmath11 integers that , for each character @xmath138 , contains the number of occurrences of lexicographically smaller characters in the string @xmath25 .
moreover , given a symbol @xmath139 , an integer @xmath140 and a string @xmath25 , the function @xmath141 returns the number of occurrences of character @xmath139 in the prefix @xmath142 $ ] . in our context
we suppose that , at the step @xmath143 , we have inserted the symbol @xmath144 $ ] at the position @xmath140 of @xmath116 . at the step @xmath134 , for each @xmath93 with @xmath145 ,
the position @xmath146 , where we have to insert the new symbol associated with the @xmath42-suffix @xmath147 $ ] into @xmath116 , is computed in the following way : @xmath148 + \rank(x , r , \bwts{j-1 } ) + 1.\ ] ] note that such a formula corresponds to the computation of the so - called @xmath149-mapping , a fundamental operation of the fm - index .
in order to decrease the i / o operations , does not use the table @xmath135 and computes the function on segments of @xmath116 rather than on the entire @xmath116 .
more precisely , considers @xmath111 as the concatenation of @xmath150 segments @xmath151 , where the symbols in @xmath152 are the characters preceding the lexicographically sorted @xmath120-suffixes of @xmath43 ( such suffixes consist in only the end - marker symbols ) and the symbols in @xmath153 , with @xmath154 , are the characters preceding the lexicographically sorted suffixes of @xmath43 starting with @xmath155 .
it is easy to see that , @xmath156 and the segments @xmath157 ( for @xmath158 ) does not contain any symbols . now , we can omit the table @xmath135 , because it is only useful for determining the segment where the suffix must be inserted .
indeed , the table @xmath135 simply allows us to establish that if the @xmath42-suffix starts with the symbol @xmath159 $ ] then it is larger than the suffixes starting with a symbol smaller than @xmath160 .
this is equivalent to say that the @xmath42-suffix must be inserted in the block containing the suffixes starting with the symbol @xmath160 , i.e. in @xmath161 .
one can verify that the position @xmath140 in @xmath116 corresponds to a position @xmath162 in @xmath163 where @xmath164 is the first symbol of @xmath126-suffix of @xmath34 , i.e @xmath165 $ ] .
now , the new symbol @xmath50 $ ] ( or the end - marker @xmath21 for the last step ) must be inserted in the position @xmath140 of @xmath161 , where @xmath140 is obtained by computing the number of occurrences of @xmath166 $ ] in @xmath167 and in @xmath168 $ ] . moreover
, the computation of @xmath168 $ ] corresponds to the computation of the function @xmath169 .
note that , in order to compute the occurrences in @xmath167 , a table of @xmath170 bits of memory can be used .
we remark that for each step @xmath42 , our strategy computes , for each @xmath42-suffix @xmath171 , its definitive position in the sorted list of the elements of @xmath43 , regardless of the fact that some @xmath42-suffixes lexicographically smaller than @xmath171 have not been considered yet in the current step @xmath42 .
actually , this means that the algorithm computes the definitive position ( called _ absolute position _ ) of the symbol associated to @xmath171 in the correspondent segment @xmath172 . during each iteration @xmath42
, we use the arrays @xmath173 , @xmath3 , @xmath174 , @xmath175 containing at most @xmath7 elements each .
the values in such arrays are updated during each step .
for ease of presentation we denote by @xmath176 , @xmath177 , @xmath178 , @xmath179 the arrays at the step @xmath42 , described as follows . for each @xmath180 : * @xmath181 is an array that stores all the symbols , if they exist , located at the position @xmath182 from the right .
more formally , @xmath183=w_q[|w_q|-j-1]$ ] if @xmath184 , @xmath183=\$_q$ ] if @xmath185 , @xmath183=\#$ ] if @xmath186 . note that @xmath187 is a symbol that does not appear in any string of the collection and it is not involved in the computation of @xmath111 .
the array @xmath181 takes @xmath188 bits of workspace .
* @xmath189 is an array of integers such that @xmath190=i$ ] if the @xmath42-suffix of the string @xmath191 ( with @xmath192 ) is the @xmath62-th @xmath42-suffix in the lexicographic order .
it uses @xmath193 bits of workspace .
* @xmath194 is an array of integers such that @xmath195 $ ] is the absolute position of the symbol @xmath196 $ ] ( or the end - marker @xmath35 ) , associated with the @xmath42-suffix of @xmath34 ( with @xmath192 ) , in @xmath161 , where @xmath197 $ ] and @xmath160 is the first symbol of the @xmath42-suffix of @xmath34 , i.e. @xmath198 $ ] .
it needs @xmath199 bits of workspace . *
@xmath200 is an array of integers such that @xmath201 $ ] stores the index @xmath202 where @xmath166 $ ] where @xmath190=i$ ] , i.e. the first symbol of the @xmath42-suffix of @xmath34 ( with @xmath192 ) .
it uses @xmath188 bits of workspace .
note that , at each step @xmath42 ranging from @xmath120 to @xmath130 , the algorithm considers only the first @xmath117 values of arrays @xmath178 , @xmath179 and @xmath177 where @xmath117 is the number of the strings of the collection having length greater than or equal to @xmath42 . at the start of the iteration @xmath42 ,
we compute the new values in @xmath200 , @xmath194 , @xmath177 in the following way .
we suppose that the @xmath126-suffix of @xmath34 is the @xmath117-th @xmath126-suffix in lexicographic order , i.e. @xmath203 $ ] .
hence , we know that the first symbol of the @xmath42-suffix of @xmath34 is @xmath198 $ ] and has been inserted in the position @xmath204=s$ ] in @xmath163 , where @xmath205=v$ ] and @xmath164 is the first symbol of @xmath126-suffix of @xmath34 , i.e @xmath206 $ ] .
the position @xmath140 of the new symbol @xmath50 $ ] ( or the end - marker @xmath35 for @xmath207 ) that must be inserted in @xmath161 is obtained by computing the number of occurrences of @xmath166 $ ] in @xmath167 and in @xmath168 $ ] , where @xmath208 $ ] is the position of @xmath160 in @xmath163 . note
that during the iteration @xmath42 , we store the values in @xmath178 , @xmath179 , @xmath177 and @xmath176 by replacing the corresponding values in @xmath209 , @xmath210 , @xmath211 and @xmath212 , so @xmath213=z$ ] , @xmath214=r$ ] and @xmath215=i$ ] .
finally , we sort @xmath200 , @xmath194 , @xmath177 where the first and the second keys of the sorting are the values in @xmath179 and @xmath178 , respectively .
so that we can insert , in sequential way , the symbols @xmath216 $ ] ( for @xmath217 ) , into each segment @xmath218 for @xmath219 .
[ ex : ebwt ] we suppose that @xmath220 ( including the distinct end - markers ) .
the algorithm starts , at the step @xmath119 , by computing the @xmath121 .
the @xmath120-suffixes are @xmath221 , and the new symbols that we have to insert in @xmath222 ( the segment associated with the suffixes starting with the end - marker ) are @xmath135 and @xmath135 .
we set @xmath223 $ ] , @xmath224 $ ] and @xmath225 $ ] .
since , we use ( implicit ) distinct end - markers and @xmath226 , we set @xmath227 $ ] .
so , @xmath228
. then , we consider @xmath229 $ ] .
both symbols in @xmath230 should be inserted into @xmath231 because both the associated suffixes ( @xmath232 and @xmath233 ) start with the symbol @xmath135 .
so , we set @xmath234 $ ] and @xmath235 $ ] . the position in @xmath231 of the symbol @xmath236 associated with
the @xmath237-suffix of @xmath238 is equal to @xmath237 in the segment @xmath231 , indeed @xmath232 is the smallest suffix in @xmath239 starting with the letter @xmath135 .
the position in @xmath231 of the symbol @xmath240 associated with the @xmath237-suffix of @xmath241 is equal to @xmath242 in the segment @xmath231 , because the symbol @xmath135 associated with @xmath243 follows the symbol @xmath135 associated with @xmath244 .
so we set @xmath245 $ ] and obtain @xmath231 by inserting @xmath236 and @xmath240
. then @xmath246 . during the third step , the array @xmath247 $ ]
is considered .
since the last inserted symbol of @xmath238 is @xmath236 then the new symbol @xmath236 associated with @xmath248 must be inserted in the segment @xmath249 , whereas since the last inserted symbol of @xmath241 is @xmath240 then the new symbol @xmath135 associated with @xmath250 must be inserted in the segment @xmath251 .
so we set @xmath252 $ ] , @xmath253 $ ] .
since the number of occurrences of @xmath236 in @xmath254 and in @xmath255 $ ] is @xmath237 then we have to insert the symbol @xmath236 at the position @xmath237 in @xmath249 .
since the number of occurrences of @xmath240 in @xmath254 and in @xmath256 $ ] is @xmath237 then we have to insert the symbol @xmath135 at the position @xmath237 in @xmath251 .
since @xmath248 and @xmath250 are the smallest suffixes starting with @xmath236 and @xmath240 respectively , we set @xmath257 $ ] . the first three iterations are depicted in figure [ fig : threeiterations ] .
the process continues via left extensions of suffixes until all symbols have been inserted .
the main goal of this section consists in the description of the strategy for computing , by using the , the array of a massive collection of strings via sequential scans of the disk data .
in particular , the main theorem of the section enables the simultaneous computation of both and of a string collection @xmath33 of maximum length @xmath4 .
we recall that the last symbol of each string @xmath34 is the ( implicit ) end - marker @xmath35 .
our method follows the algorithm in the sense that it scans all the strings from right - to - left in @xmath4 steps and simulates , at the step @xmath42 , the insertion of suffixes of length @xmath42 into the sorted list of suffixes .
this time , however , we wish to compute both the and .
so , at the step @xmath42 , the longest common prefix array ( denoted by @xmath259 ) of the collection @xmath43 of the suffixes having length at most @xmath42 is computed alongside @xmath111 . as well as computing an value for the inserted suffix
, we must also modify the value for the suffix that comes after it to reflect the longest prefix common to it and the inserted suffix .
our goal in this section is to frame these calculations in terms of @xmath260s on sets of intervals within @xmath259 , serving as a preliminary to the next section , where we show how these computations may be arranged to proceed via sequential scans of the data .
it is easy to see that when @xmath261 , @xmath259 coincides with @xmath63 , the array of @xmath39 .
since all @xmath7 end - markers are considered distinct , the longest common prefix of any pair of @xmath120-suffixes is @xmath120 , so the first @xmath7 positions into @xmath259 are @xmath120 for any @xmath262 . the general idea behind our method
is described in the following and depicted in figure [ fig : idealcp ] . at each step
@xmath263 , the value of a generic element @xmath264 $ ] ( with @xmath265 ) have to be computed by taking into account the @xmath42-suffix @xmath171 that is placed at the position @xmath140 of the sorted suffixes of @xmath43 .
such a value depends on the @xmath266-suffix @xmath267 placed at the position @xmath268 in the sorted list and , moreover , could lead an updating of the value @xmath269 $ ] ( if it exists ) corresponding to the @xmath270-suffix @xmath271 , as shown in figure [ fig : idealcp ] . by using our method ,
the computation of @xmath264 $ ] can be realized by using the values of the arrays @xmath272 and @xmath116 .
in fact , if @xmath273 ( where @xmath274 is a symbol and @xmath275 is the @xmath126-suffix placed at a certain position @xmath117 in the sorted list of the suffixes of @xmath276 ) , then @xmath277 $ ] .
let us denote @xmath278 and @xmath279 , where @xmath280 and @xmath281 are symbols and @xmath282 and @xmath283 are , respectively , the @xmath284-suffix and the @xmath285-suffix of @xmath276 placed at the positions @xmath286 and @xmath287 .
one can see that @xmath264 $ ] is equal to @xmath120 if @xmath288 , otherwise it is equal to @xmath237 plus the longest common prefix between @xmath282 and @xmath275 computed in the array @xmath272 .
moreover @xmath269 $ ] is equal to @xmath120 if @xmath289 , otherwise it is equal to @xmath237 plus the longest common prefix between @xmath275 and @xmath283 computed in the array @xmath272 . in this section
we show how such a computation can be sequentially performed .
@xmath290 note that @xmath259 can be considered the concatenation of @xmath150 arrays @xmath291 where , for @xmath158 , the array @xmath292 contains the values corresponding to the lengths of the longest common prefix of the suffixes of @xmath43 that start with @xmath155 , while @xmath293 ( corresponding to the @xmath120-suffixes ) is an array of @xmath7 zeroes .
it is easy to see that @xmath294 and that @xmath295 is empty for @xmath296 .
we note that , for each @xmath297 , @xmath298=0 $ ] and @xmath299\geq 1 $ ] for @xmath300 , because the suffixes associated with such values share at the least the symbol @xmath155 .
as shown in section [ bcrstrategy ] , @xmath111 can be partitioned in an analogous way into segments @xmath151 . given the segments @xmath153 and @xmath292 , @xmath219 , for the symbol @xmath139 occurring at position @xmath140 of @xmath153 we define the @xmath301__-lcp current interval _ _ of @xmath139 in @xmath140 ( denoted by @xmath302 ) as the range @xmath303 $ ] in @xmath304 ( so we set @xmath305 $ ] ) , where @xmath286 is the greatest position smaller than @xmath140 of the symbol @xmath139 in @xmath218 , if such a position exists .
if such a position does not exist , we define @xmath306 $ ] .
analogously , we define for the symbol @xmath139 the @xmath301__-lcp successive interval _ _ of @xmath139 in @xmath140 ( denoted by @xmath307 ) as the range @xmath308 $ ] in @xmath304 ( so we set @xmath309 $ ] ) , where @xmath287 is the smallest position greater than @xmath140 of the symbol @xmath139 in @xmath218 , if it exists .
if such a position does not exist we define @xmath310 $ ] . in order to compute the values @xmath286 and @xmath287 we use the function @xmath311 that together with @xmath136 function
play a fundamental role in fm - index .
in particular , @xmath312 takes in input a symbol @xmath274 , an integer @xmath140 and a string @xmath81 and finds the position of the @xmath140-th occurrence of @xmath274 in @xmath81 . in our context ,
if @xmath286 and @xmath287 exist , then @xmath313 \neq x$ ] for @xmath314 and for @xmath315 , so it is easy to verify that @xmath316 and @xmath317 .
we observe that the computation of the minimum value into @xmath302 and @xmath307 is equivalent to the computation of @xmath318 and @xmath319 , respectively .
we can not directly compute these values , because we build each @xmath153 and @xmath292 in sequential way , so we do not know the left extreme of @xmath302 and the right extreme of @xmath307 of all symbols that we have to insert in each @xmath153 , @xmath158 . the following theorem , related to ( * ? ? ?
* lemma 4.1 ) , shows how to compute the segments @xmath292 , with @xmath320 , by using @xmath321 and @xmath322 for any @xmath323 .
[ th : lcp_case_allsuffix ] let @xmath324 be the set of the positions where the symbols associated with the @xmath42-suffixes starting with the letter @xmath160 must be inserted into @xmath325 . for each position @xmath326 ( @xmath327 ) , @xmath328=\left\{\begin{array}{ll } 0 & \mbox { if $ r_b=1 $ } \\ 1 & \mbox { if $ r_b>1 $ and $ lci_{j-1}^v(c_z , t)=l_{j-1}(v)[t]$ } \\ \min lci_{j-1}^v(c_z , t ) + 1 & \mbox{otherwise } \end{array}\right.\ ] ] where @xmath164 is the first character of the @xmath126-suffix of @xmath34 , and @xmath117 is the position in @xmath163 of symbol @xmath160 preceding the @xmath126-suffix of @xmath34 . + for each position @xmath329 ( where @xmath326 and @xmath327 ) , then @xmath330=\left\{\begin{array}{ll } 1 & \mbox { if $ lsi_{j-1}^v(c_z , t)=l_{j-1}(v)[t]$ } \\ \min lsi_{j-1}^v(c_z , t ) + 1 & \mbox{otherwise } \end{array}\right.\ ] ] for each position @xmath162 , where @xmath331 ( for @xmath332 ) , @xmath333 ( for @xmath334 ) , @xmath335 ( for @xmath336 ) then @xmath337=l_j(z)[s - b]\ ] ] we consider a generic position @xmath338 corresponding to the position where the new symbol @xmath50 $ ] ( or @xmath52=\$_i$ ] ) must be inserted into @xmath161 and the corresponding value must be inserted into @xmath339 . the new symbol @xmath50 $ ] precedes the @xmath42-suffix @xmath41 $ ] .
such a suffix is obtained by concatenating the symbol @xmath166 $ ] with its @xmath126-suffix .
let us suppose that the symbol @xmath160 associated with @xmath126-suffix starting with the symbol @xmath164 is in position @xmath117 in @xmath163 .
hence , @xmath117 is also the position of the @xmath126-suffix in the lexicographic order among the suffixes of @xmath39 of length at most @xmath126 starting with @xmath164 . in order to find @xmath340 $ ]
, we have to distinguish two cases . in the first case
, the @xmath42-suffix is the smallest suffix ( in the lexicographic order ) of length at most @xmath42 starting with the symbol @xmath160 , i.e. @xmath341 .
this means that @xmath160 does not exist in any segment @xmath322 , @xmath342 and there does not exist in @xmath343 $ ] . in this case
@xmath344 $ ] and @xmath345=0 $ ] . in the second case
, there exists some suffix starting with @xmath160 of length at most @xmath42 that is lexicographically smaller than the @xmath42-suffix , i.e. @xmath346 . recall that @xmath340 $ ] represents the length of the longest common prefix between the @xmath42-suffix and a @xmath266-suffix of a string @xmath347 ( for some @xmath348 ) , with @xmath349 , starting with the symbol @xmath160 , that immediately precedes the @xmath42-suffix in the lexicographic order .
if the longest common prefix between the @xmath266-suffix of @xmath350 and the @xmath42-suffix of @xmath34 is @xmath160 then @xmath344 $ ] and @xmath346 , so @xmath340=1 $ ] .
this means that the symbol @xmath160 associated to the @xmath284-suffix of @xmath347 is not contained in the segment @xmath163 , but it is contained in some @xmath322 , @xmath342 .
if the longest common prefix between the @xmath266-suffix of @xmath350 and the @xmath42-suffix of @xmath34 is longer than @xmath160 then both the @xmath284-suffix of @xmath347 and the @xmath126-suffix of @xmath34 start with the same symbol @xmath164 .
so , we can suppose that the symbol associated with the @xmath284-suffix of @xmath350 is at the position @xmath286 in the segment @xmath163 .
remark that symbols in position @xmath286 and @xmath117 are equal to @xmath160 . then @xmath351 $ ] and @xmath340 = \min ( lci_{j-1}^v(c_z , t ) ) + 1 $ ] .
recall that @xmath352 $ ] represents the length of the longest common prefix between the @xmath42-suffix and a @xmath270-suffix of a string @xmath353 ( for some @xmath354 ) , with @xmath355 , starting with the symbol @xmath160 , that immediately follows the @xmath42-suffix in the lexicographic order .
if the longest common prefix between the @xmath42-suffix of @xmath34 and @xmath270-suffix of @xmath356 is @xmath160 then @xmath357 $ ] , so @xmath352=1 $ ] .
this means that the symbol @xmath160 associated to the @xmath285-suffix of @xmath353 is not contained in the segment @xmath163 ( but it is contained in some @xmath322 , @xmath358 ) .
if the longest common prefix between the @xmath42-suffix of @xmath34 and the @xmath270-suffix of @xmath356 is longer than @xmath160 then both the @xmath126-suffix of @xmath34 and the @xmath285-suffix of @xmath353 start with the same symbol @xmath164 .
so , we can suppose that the symbol associated with the @xmath285-suffix of @xmath356 is at the position @xmath287 in the segment @xmath163 .
remark that symbols in position @xmath140 and @xmath287 are equal to @xmath160 . then @xmath359 $ ] and @xmath352 = \min(lsi_{j-1}^v(c_z , t ) ) + 1 $ ] .
note that the position @xmath360 in @xmath339 does not exist when the the @xmath42-suffix is the greatest suffix ( in the lexicographic order ) of length at most @xmath42 starting with the symbol @xmath160 .
the suffix that , eventually , immediately follows the @xmath42-suffix in lexicographic order is involved in a segment @xmath218 , for some @xmath361 , hence such suffix starts with a symbol greater than @xmath160 .
a consequence of the theorem is that the segments @xmath362 and @xmath363 can be constructed sequentially and stored in external files .
this fact will be used in the next section .
based on the strategy described in the previous section , here we propose an algorithm ( named ) that simultaneously computes the and the of a collection of strings @xmath39 .
memory use is minimized by reading data sequentially from files held in external memory : only a small proportion of the symbols of @xmath39 need to be held in internal memory .
we could also add the computation of the generalized suffix array of @xmath39 without adding further data structures .
as in the previous sections , we assume that the collection @xmath39 comprises @xmath7 strings of length at most @xmath4 , that @xmath364 , @xmath365 , @xmath366 , that @xmath367 and @xmath368 . when @xmath261 , @xmath369 and @xmath370 .
we also assume that @xmath371 .
note that , at each iteration @xmath42 , both the segments @xmath362 and @xmath363 , initially empty , are stored in different external files that replace the files used in the previous iteration .
consequently , both @xmath111 and @xmath259 are updated accordingly .
the main part of the algorithm consists of @xmath4 consecutive iterations . at iteration @xmath42
, we consider all the @xmath42-suffixes of @xmath39 and simulate their insertion in the sorted suffixes list .
for each symbol associated with the @xmath42-suffix that we have to insert at the position @xmath140 into @xmath153 , we also have to insert the new values related to the longest common prefix at position @xmath140 and @xmath360 into @xmath292 , where @xmath372 is the first symbol of the considered @xmath42-suffix . in order to compute @xmath111 and @xmath259 ,
the algorithm needs to hold six arrays of @xmath7 integers in internal memory .
four of these ( @xmath174 , @xmath175 , @xmath3 and @xmath173 ) are as employed by the algorithm ( see section [ bcrstrategy ] ) and further two arrays ( @xmath135 and @xmath373 ) are needed to compute and update the values of the longest common prefixes at each iteration . as for @xmath174 , @xmath175 , @xmath3 and @xmath173 arrays ( see section [ bcrstrategy ] ) , for ease of presentation we denote by @xmath374 and @xmath375 the arrays at the @xmath42-th iteration .
they contain exactly one integer for each string , i.e. they use @xmath376 bits of workspace .
they are sequentially computed by using other auxiliary data structures described in subsection [ subsec : bcrlcp_ds ] . more formally , if @xmath377 $ ] then @xmath378 $ ] stores the length of the longest common prefix between the @xmath42-suffix of @xmath34 and the previous suffix ( if it exists ) in the list of sorted suffixes with respect to the lexicographic order of all the suffixes of @xmath39 of length at most @xmath42 , whereas @xmath379 $ ] contains the length of the longest common prefix between the @xmath42-suffix of @xmath34 and the successive suffix @xmath171 in the list of sorted suffixes ( if it exists ) . such values will be computed at the iteration @xmath380 according to theorem [ th : lcp_case_allsuffix ] .
note that @xmath381 $ ] is used when the suffix @xmath171 exists and @xmath195 + 1 \neq p_{j}[q+1]$ ] .
example [ exlcp ] provides an idea of the running of the algorithm and shows how the involved arrays are used .
[ exlcp ] figure [ fig : compute ] illustrates an execution of our method on the collection @xmath220 at two consecutive iterations .
note that we have appended different end - marker to each string ( @xmath244 and @xmath243 , respectively ) .
in particular , we suppose that at the iteration @xmath382 , we have computed the arrays @xmath383 , @xmath384 and @xmath385 .
we recall that we have computed @xmath386 and @xmath387 in the previous iteration . at the end of the first phase of the iteration @xmath388 we have @xmath389 $ ] ,
@xmath390 $ ] , @xmath391 $ ] , @xmath392 $ ] , @xmath393 $ ] , @xmath394 $ ] .
note that they are sorted by using the first and the second keys the values in @xmath384 and @xmath383 respectively . in this way the new @xmath395-segments ( on the left in the figure )
have been constructed by adding the bold characters .
we can compute @xmath396 and @xmath397 and @xmath398 and @xmath399 useful to terminate the iteration @xmath388 .
we obtain that @xmath396 corresponds to the range @xmath400 $ ] in @xmath401 .
so the minimum value is @xmath242 and the value in @xmath402 associated with the @xmath403-suffix @xmath171 ( i.e. @xmath404 ) of @xmath241 is @xmath405 ( this value is stored into @xmath406 ) .
now we have to compute the value of the longest common prefix between the suffix @xmath171 and the suffix that immediately follows @xmath171 in the lexicographic order and to store this value into @xmath407 .
since the symbol @xmath408 does not appear in the range @xmath409 $ ] , it means that there are not suffixes starting with @xmath410 lexicographically greater than @xmath171 , so such value will be less than @xmath242 . because the symbol @xmath408 does appear at least once in @xmath395 with @xmath411 ( in this case @xmath412=g$ ] ) ,
it means that there exists at least a suffix starting with @xmath408 lexicographically greater than @xmath171 ( in this case the suffix @xmath413 ) .
so the value in @xmath414 of the suffix that follows @xmath171 must be updated to @xmath237 ( i.e. we store this value in @xmath407 ) .
similarly , since @xmath415 $ ] , the minimum value is @xmath237 and so the value in @xmath402 for the @xmath403-suffix @xmath416 ( i.e. @xmath417 ) of @xmath238 is @xmath242 ( i.e. we insert this value in @xmath406 ) .
moreover , @xmath418 $ ] , the minimum value is @xmath242 and hence the value in @xmath402 of the suffix that follows @xmath416 must be updated to @xmath405 ( i.e. we insert this value in @xmath407 ) . at the iteration @xmath419 , we compute the arrays @xmath420 , @xmath421 and @xmath422 , whereas the arrays @xmath406 and @xmath407 have been computed in the previous iteration .
so we have @xmath423 $ ] , @xmath424 $ ] , @xmath425 $ ] , @xmath426 $ ] , @xmath427 $ ] , @xmath428 $ ] .
while the new @xmath429-segments ( on the right in the figure ) are being constructed , we can sequentially insert and update the new values in @xmath430 and @xmath414 and compute the new values @xmath431 and @xmath432 .
@xmath433 in this subsection we show how , at the generic iteration @xmath42 of the algorithm , @xmath259 is sequentially computed and updated by using the data structures previously described . at the first iteration ,
@xmath434=0 $ ] and @xmath435=0 $ ] for each @xmath371 , because the @xmath7 end - markers are considered distinct .
moreover , the algorithm initializes the segments @xmath436 and @xmath437 in the following way : @xmath438w_1[|w_1|-1]\cdots w_{m-1}[|w_{m-1}|-1]$ ] and @xmath439=0 $ ] , for each @xmath62 . consequently , the arrays are initialized by setting @xmath440=q-1 $ ] , @xmath441=q$ ] , @xmath442=0 $ ] , @xmath443=1 $ ] and @xmath444=1 $ ] , for each @xmath62 .
each iteration @xmath263 can be divided into two consecutive phases . during the first phase we only read the segments @xmath445 in order to compute the arrays @xmath178 , @xmath179 @xmath177 and @xmath176 . then we sort @xmath200 , @xmath194 , @xmath177 , @xmath446 , @xmath375 , where the first and the second keys of the sorting are the values in @xmath179 and @xmath178 respectively .
we omit the description of the first phase , because it can be found in section [ bcrstrategy ] , so we focus on the second phase . in the second phase , the segments @xmath445 and @xmath447 are read once sequentially both for the construction of new segments @xmath362 and @xmath363 and for the computation of the arrays @xmath448 and @xmath449 , as they will be used in the next iteration . moreover , the computation of the segments @xmath450 is performed by using the arrays @xmath446 and @xmath451 constructed during the previous step .
since the identical elements in @xmath179 are consecutive , we open the pair files @xmath322 and @xmath321 ( for @xmath219 ) at most once .
each of these files can be sequentially read , because the positions in @xmath178 are sorted in according with @xmath179 . in the sequel ,
we focus on a single segment @xmath325 , for @xmath452 , by assuming that @xmath453=z$ ] for each @xmath454 , with @xmath455 and @xmath456 , i.e. we are considering the elements in @xmath325 and @xmath457 associated with the suffixes starting with @xmath160 .
since @xmath458 < \ldots < p_{j}[l']$ ] , we can sequentially build @xmath325 and @xmath457 by copying the old values from @xmath459 and @xmath460 respectively , by inserting each new symbol @xmath461 $ ] into @xmath462 $ ] and @xmath463 $ ] into @xmath464 $ ] and by updating the value @xmath465 + 1]$ ] with the value @xmath466 $ ] , if the position @xmath467
+ 1\neq p_{j}[p+1]$ ] exists .
the crucial point consists in computing , at the same time , each value @xmath468 $ ] and @xmath469 $ ] ( required for the next iteration ) related to the @xmath470-suffix of all string @xmath34 , with @xmath471 $ ] and @xmath192 , without knowing the left extreme @xmath286 of @xmath302 and the right extreme @xmath287 of @xmath307 , where @xmath472 $ ] and @xmath473 $ ] for each @xmath146 . in the sequel , we say that the left extreme is the _ opening position _ and the right extreme is the _ closing position _ of some symbol @xmath139 . a pseudo - code that realizes the second phase of a generic iteration @xmath263 of the algorithm can be found in figure [ algo : cs_computation ] and it uses the following additional arrays of @xmath11 elements , defined as follows : * isminlciop and isminlsiop are arrays of flags , where isminlciop@xmath474 $ ] and isminlsiop@xmath474 $ ] indicate the opening of the @xmath475 and @xmath476 associated with @xmath164 , respectively . *
minlci and minlsi are arrays of integers , where minlci@xmath474 $ ] and minlsi@xmath474 $ ] store the minimum value among the values in @xmath457 from the opening position of the intervals associated with @xmath164 and the current position @xmath162 .
* minlsinseq is an array of integers , where minlsinseq@xmath474 $ ] contains the index @xmath270 of the position in which minlsi@xmath474 $ ] will be stored in @xmath449 .
this is useful , because when we close the @xmath476 associated with the symbol @xmath164 at some position @xmath162 , we have to store minlsi@xmath474 $ ] in some position @xmath270 of @xmath449 , such that @xmath477<s$ ] .
the routines insertnewsymbol , updatelci and updatelsi are described in figure [ algo : insertnewsymbol ] and figure [ algo : updatemin ] .
insertnewsymbol(@xmath139 ) routine inserts at position @xmath478 $ ] in @xmath325 each new symbol @xmath479 $ ] . by theorem [ th : lcp_case_allsuffix ] , it follows that @xmath464=0 $ ] if @xmath467=1 $ ] or @xmath464=c_j[p]$ ] otherwise .
moreover , the position @xmath480 $ ] is surely : * the closing position of @xmath481)$ ] .
if isminlciop@xmath474=0 $ ] , then @xmath482 $ ] is the position of the first occurrence of @xmath164 in @xmath161 , hence @xmath481)=l_j(z)[p_j[p]]$ ] and we set @xmath468=1 $ ] according to theorem [ th : lcp_case_allsuffix ] .
otherwise , isminlciop@xmath474 $ ] has been set to @xmath237 in some position @xmath483 $ ] , so @xmath484)=l_j(z)(d_1,p_j[p]]$ ] and we set @xmath468 = \min(lci_{j}^z(c_v , p_j[p]))+1 $ ] .
such minimum value is stored into minlci@xmath474 $ ] .
moreover , we set isminlciop@xmath474 = 0 $ ] .
* the opening position of @xmath485)$ ] .
so , we set isminlsiop@xmath474=1 $ ] , the value minlsi@xmath474 $ ] is updated and minlsinseq@xmath474 = p$ ] .
we observe that if the position @xmath482 $ ] is the last occurrence of @xmath164 in @xmath161 ( this fact is discovered when the end of the file is reached ) , it means that @xmath485 ) = l_j(z)[p_j[p]]$ ] , i.e. we set @xmath469=1 $ ] . note that in order to compute the values @xmath468 $ ] and @xmath469 $ ] for each @xmath146 , we do not need to know the exact opening position @xmath286 of @xmath486,p_{j}[p])$ ] and the exact closing position @xmath287 of @xmath487,p_{j}[p])$ ] , but we only need to compute the minimum values in these ranges . when at any position @xmath162 in @xmath325 we insert the symbol @xmath164 ( from @xmath176 or from @xmath459 ) , then the position @xmath162 is assumed to be : * the opening position of @xmath488 , if another occurrence of @xmath164 will be inserted , as new symbol , at some next position @xmath489 $ ] , for some @xmath266 such that @xmath490>s$ ] .
so , we set isminlciop@xmath474 = 1 $ ] and the value minlci@xmath474 $ ] is updated ( see updatelci(@xmath139 ) routine in figure [ algo : updatemin ] ) .
* the closing position of @xmath491)$ ] , if another occurrence of @xmath164 has been inserted , as new symbol , at some previous position @xmath492 $ ] , for some @xmath270 such that @xmath492 < s$ ] . in this case ,
@xmath491)=l_j(z)(p_j[g ] , s]$ ] ( because minlsinseq@xmath474=g$ ] ) and we set @xmath493 = \min ( lsi_{j}^z(c_v , p_j[g]))+1 $ ] according to theorem [ th : lcp_case_allsuffix ] .
such a minimum value is stored into minlsi@xmath474 $ ] .
we set isminlsiop@xmath474 = 0 $ ] ( see updatelsi(@xmath139 ) routine in figure [ algo : updatemin ] ) . when @xmath325 is entirely built , the closing position of some @xmath491)$ ] for some @xmath494 could remain not found .
so , we could have some value in isminlsiop equal to @xmath237 . for each @xmath164
such that isminlsiop@xmath474=1 $ ] , the last occurrence of @xmath164 appears at position @xmath492 $ ] ( we recall that we have set minlsinseq@xmath474 $ ] to @xmath270 ) . in this case ,
@xmath491)=l_j(z)[p_j[g]]$ ] and we set @xmath493 = 1 $ ] according to theorem [ th : lcp_case_allsuffix ] ( see the for loop in figure [ algo : cs_computation ] )
. one can verify that these steps work in a sequential way .
moreover , one can deduce that , while the same segment is considered , for each symbol @xmath495 at most one @xmath496)$ ] for some @xmath497 and at most one @xmath498)$ ] for some @xmath494 will have not their closing position .
the complexity of algorithm depends mainly on @xmath4 , i.e. the length of the longest string in the collection , because the algorithm works in @xmath4 passes and for each step it needs to build @xmath325 and @xmath457 from @xmath459 and @xmath460 for each @xmath499 . at each step
@xmath42 , the used internal memory depends on the number of strings in @xmath39 of length greater than @xmath42 .
such a value is upper bounded by @xmath7 for each step .
note that , at each step , the total size of @xmath500 files containing the partial @xmath100 is increased by at most @xmath7 symbols .
analogously the total size of @xmath500 files containing the partial @xmath63 is increased by at most @xmath7 values .
so , the used disk space mainly depends on @xmath7 for each step . since our algorithm accesses disk data only by sequential scans , we analyze it counting the number of disk passes as in the standard external memory model ( see @xcite ) .
we denote by @xmath14 the disk block size and we assume that both the ram size and @xmath14 are measured in units of @xmath501-bit words . from the size of the data structures and from the description of the phases of the algorithm given in previous sections , we can state the following theorem . given a collection @xmath39 of @xmath7 strings over an alphabet of size @xmath11 where @xmath4 is the maximal length ( including the end - markers ) and @xmath3 is the sum of their length , _ _ algorithm simultaneously compute the and the array of @xmath39 by using @xmath13 disk i / o operations and @xmath8 bits of ram in @xmath502 cpu time , where @xmath10 is the time taken to sort @xmath7 integers in internal memory .
moreover , _ _ needs at most @xmath12 bits of disk space . at each iteration @xmath42 ,
the main data structures used by are @xmath176 of size @xmath503 bits , @xmath179 of size at most @xmath503 bits , @xmath177 of size at most @xmath504 bits , @xmath178 of size at most @xmath505 bits , @xmath374 and @xmath375 of size at most @xmath506 bits .
moreover , we need @xmath507 bits for computing the number of occurrences of each symbol in each segment @xmath153 , for @xmath158 . the additional arrays take @xmath508 ( assuming that @xmath108 ) .
so , the workspace is @xmath8 bits .
as we sort @xmath200 , @xmath194 , @xmath177 , @xmath446 , @xmath375 where the first and the second keys of the sorting are the values in @xmath179 and @xmath178 , we need @xmath509 cpu time , where @xmath10 is the time taken to sort @xmath7 integers in internal memory .
the total cpu time is @xmath502 .
we observe that the input is split into @xmath4 files , so that each file contains @xmath7 symbols , one for each string . at each step , such symbols will read and store into @xmath176 array .
the amount of i / o operations for handling the input file is @xmath510 . during the first phase of each iteration @xmath263
, we need to read at most @xmath511 bits for computing @xmath179 , @xmath178 , @xmath177 arrays . in the second phase
we need to read at most @xmath511 bits from @xmath445 and @xmath447 segments in order to obtain @xmath448 , @xmath449 arrays and @xmath362 and @xmath363 segments by writing @xmath512 bits of disk space .
the disk i / o operations for obtaining the output of each iteration is @xmath513 .
the total number of i / o operations is @xmath514 .
the internal memory of can be reduced significantly by observing that rather than storing @xmath174 , @xmath175 , @xmath3 and @xmath135 to internal memory , they could reside on disk because these arrays are sequentially processed . in this way , the algorithm needs to store in internal memory the array @xmath373 , isminlciop , isminlsiop , minlci , minlsi and minlsinseq that require random accesses .
clearly in this case one should use an external memory sorting algorithm .
[ rm : gsa ] it is easy to verify that the algorithm can be also adapted for computing the generalized suffix array of @xmath39 together the .
such further computation increases the number of i / o operations , but one do not need of further data structures in internal memory .
note that , for each @xmath93 , @xmath61=(t , j)$ ] is the pair corresponding to the @xmath62-th smallest suffix of the strings in @xmath39 , i.e. to the suffix @xmath58 $ ] . since we know at each iteration of the values @xmath117 and @xmath42 , because they are store into @xmath178 and @xmath177 arrays respectively , it is enough to modify the code by adding the instructions for storing such values in external memory .
our algorithm represents the first lightweight method that simultaneously computes , via sequential scans , the array and the of a vast collection of strings .
we developed an implementation of the algorithm described in section [ sec : implementation ] , which is available upon request from the authors .
our primary goal has been to analyze the additional overhead in runtime and memory consumption of simultaneously computing both and via compared with the cost of using ( cf .
@xcite ) to compute only the . for this goal
, we used a publicly available collection of human genome strings from the sequence read archive @xcite at ftp://ftp.sra.ebi.ac.uk/vol1/era015/era015743/srf/ and created subsets containing @xmath16 , @xmath515 and @xmath15 million reads , each read being 100 bases in length on the alphabet @xmath516 .
.the input string collections were generated on an illumina gaiix sequencer , all strings are @xmath16 bases long .
size is the input size in gib , wall clock time the amount of time that elapsed from the start to the completion of the instance is given as microseconds per input base , and memory denotes the maximal amount of memory ( in gib ) used during execution .
the efficiency column states the cpu efficiency values , i.e. the proportion of time for which the cpu was occupied and not waiting for i / o operations to finish , as taken from the output of the `
/usr / bin / time ` command . [ cols="<,>,<,<,<,<,<",options="header " , ] actually , we were not able to compare the performance of and on very large datasets
. indeed , current implementation of shows an overhead in disk space usage that prevented us to make tests on very large string collections .
for instance , for a collection of @xmath517 millions of strings of length @xmath16 needs @xmath518 bytes per symbol , whereas uses @xmath519 byte per symbol .
so , the datasets in table [ tab : inputinstances ] are too demanding to be managed .
however , experiments have been conducted using smaller real collections ( see table [ tab : datasets ] ) : the dataset @xmath520 ( long jump library of human chromosome 14 ) , the datasets @xmath521 and @xmath522 ( human genome sequences from sequence read archive ) are collections of dna strings on the alphabet @xmath516 . moreover , in order to evaluate and when the dataset contains strings over alphabets with more than five symbols , we have created a new collection of proteins ( called @xmath523 ) , obtained from uniprotkb / trembl dataset , by truncating the strings longer than @xmath16 aminoacids , so that it has the same maximum string length as the datasets @xmath520 , @xmath521 , @xmath522 . for each dataset ,
is compared with using two different values of : an higher value is comparable with the size of memory required from the data structures used by ; a lower value have been chosen to produce a number of subsets less than @xmath524 .
we have not set other parameters of code .
the results of our tests are described in table [ tab : compare_egsa ] .
the first three experiments , in which we consider strings over the same alphabet , show that the bigger the string collection is , the better is the overall performance of with respect to .
in fact , the wall clock time values of and become comparable , but the total disk usage for significantly increases .
however , the fourth experiment in table [ tab : compare_egsa ] shows that the alphabet size is a parameter that seems to have a significant impact on the wall clock time value for rather than for .
finally , we observe that , as pointed in remark [ rm : gsa ] , could produce as additional output the generalized suffix array .
although our algorithm is not optimized for this purpose , our tests have shown that the wall clock time is about twice because the i / o operations are doubled .
however , the computation of the gsa does not produce an increase in the internal memory .
in this paper , we proposed which is a lightweight algorithm to construct , at the same time , the array and the of a collection of strings . actually , the and are two of the three data structures needed to build a compressed suffix tree ( cst ) @xcite of a string .
the strategy proposed in this paper could enable the lightweight construction of csts of string collections for comparing , indexing and assembling vast datasets of strings when memory is the main bottleneck .
the algorithm is a tool parameter - free designed for very large collection , indeed table [ tab : inputinstances ] shows that it also works with collections of @xmath525gib and tables [ tab : varlenres ] and [ tab : compare_egsa ] show that it can be also used on large collections with varying string lengths .
the experimental results show that our algorithm is a competitive tool for the lightweight simultaneous computation of and on string collections .
our current prototype can be further optimized in terms of memory by performing the sorting step in external memory .
further saving of the working space could be obtained if we embody our strategy in or ( see @xcite ) .
these methods , although slower than , need to store only a constant and ( for the dna alphabet ) negligibly small number of integers in ram regardless of the size of the input data .
g. rosone and m. sciortino are partially supported by the project miur - sir cmacbioseq ( `` combinatorial methods for analysis and compression of biological sequences '' ) grant n. rbsi146r5l and by `` gruppo nazionale per il calcolo scientifico ( gncs - indam ) '' .
j. cox is an employee of illumina cambridge ltd .
+ the authors are very grateful to the anonymous referees for their helpful remarks and constructive comments .
louza , _ um algoritmo para a construo de vetores de sufixo generalizados em memria externa _ , master s dissertation in cincias de computao e matemtica computacional , so carlos .
instituto de cincias matemticas e de computao , university of so paulo , 2013 .
louza , g.p .
telles , and c.d.a ciferri , _ external memory generalized suffix and lcp arrays construction _ , cpm ( johannes fischer and peter sanders , eds . ) , lcns , vol .
7922 , springer berlin heidelberg , 2013 , pp .
|
the longest common prefix array is a very advantageous data structure that , combined with the suffix array and the burrows - wheeler transform , allows to efficiently compute some combinatorial properties of a string useful in several applications , especially in biological contexts .
nowadays , the input data for many problems are big collections of strings , for instance the data coming from `` next - generation '' dna sequencing ( ngs ) technologies . in this paper
we present the first lightweight algorithm ( called ` extlcp ` ) for the simultaneous computation of the longest common prefix array and the burrows - wheeler transform of a very large collection of strings having any length .
the computation is realized by performing disk data accesses only via sequential scans , and the total disk space usage never needs more than twice the output size , excluding the disk space required for the input . moreover , ` extlcp ` allows to compute also the suffix array of the strings of the collection , without any other further data structure is needed .
finally , we test our algorithm on real data and compare our results with another tool capable to work in external memory on large collections of strings .
( # 1)`bwt`(#1 ) ( # 1)`c - ebwt`(#1 ) ( # 1)`bwt`(#1 ) # 1_#1 ( ) # 1_#1 ( ) # 1_#1 ( ) # 1_#1(w ) # 1_#1(w ) # 1(#1 )
|
the information theory realm , it is well - known that given the second - order moment ( or variance ) , a gaussian density function maximizes the differential entropy .
similarly , given the second - order moment , the gaussian density function minimizes the fisher information , a result which is referred to as the cramr - rao inequality in the signal processing literature .
surprisingly , the proofs proposed in the literature for these fundamental results are relatively quite diverse , and no unifying feature exists . since differential entropy or fisher information is a functional with respect to a probability density function , the most natural way to establish these results is by approaching them from the perspective of functional analysis .
this paper presents a unifying variational framework to address these results as well as numerous other fundamental information theoretic results .
a challenging information theoretic inequality , referred to as the extremal entropy inequality ( eei ) @xcite , can be dealt with successfully in the proposed framework of functionals .
furthermore , the proposed variational calculus perspective is useful in establishing other novel results , applications and extensions of the existing information theoretic inequalities .
the main theme of this paper is to illustrate how some tools from calculus of variations can be used successfully to prove some of the fundamental information theoretic inequalities , which have been widely used in information theory and other fields , and to establish some applications .
the proposed variational approach provides alternative proofs for some of the fundamental information theoretic inequalities and enables finding novel extensions of the existing results .
this statement is strengthened by the fact that the proposed variational framework is quite general and powerful , and it allows easy integration of various linear and inequality constraints into the functional that is to be optimized .
therefore , we believe that a large number of applications could benefit of these tools .
the proposed variational approach offers also a potential guideline for finding the optimal solution for many open problems .
variational calculus techniques have been used with great success in solving important problems in image processing and computer vision @xcite such as image reconstruction ( denoising , deblurring ) , inverse problems , and image segmentation . recently
, variational techniques were also advocated for optimization of multiuser communication systems @xcite , for deriving analytical wireless channel models using the maximum entropy principle when only limited information about the environment is available @xcite , and for designing optimal training sequences for radar and sonar applications @xcite-@xcite .
maximum entropy principle found also applications in spectral estimation ( e.g. , burg s maximum entropy spectral density estimator @xcite ) and bayesian statistics @xcite .
the major results of this paper are enumerated as follows .
first , using calculus of variations , the maximizing differential entropy and minimizing fisher information theorems are proved under the classical ( standard ) assumptions found in the literature as well as under a different set of assumptions .
it is shown that a gaussian density function maximizes the differential entropy but it minimizes the fisher information , given the second - order moment .
it is also shown that a half normal density function maximizes the differential entropy over the set of non - negative random variables , given the second - order moment .
furthermore , it is shown that a half normal density function minimizes the fisher information over the set of non - negative random variables , provided that the regularity condition and [ ef_thm6 ] . ]
is ignored and the second - order moment is given .
it is also shown that a chi density function minimizes the fisher information over the set of non - negative random variables , under the assumption that the regularity condition holds and the second - order moment is given .
second , a novel proof of the worst additive noise lemma @xcite is provided in the proposed functional framework .
previous proofs of the worst additive noise lemma were based on jensen s inequality or data processing inequality @xcite , @xcite . unlike the previous proofs , our approach is purely based on calculus of variations techniques , and the vector version of the lemma is treated .
third , eei is studied from the perspective of a functional problem .
the main advantage of the proposed new proof is that neither the channel enhancement technique and the entropy power inequality ( epi ) , adopted in @xcite , nor the equality condition of data processing inequality and the technique based on the moment generating functions , used in @xcite , are required . using a technique based on calculus of variations , an alternative proof of eei
is provided .
finally , several applications and extensions of the proposed results are discussed .
the rest of this paper is organized as follows .
some variational calculus preliminary results and their corollaries are first reviewed in section [ cal ] . maximizing differential entropy theorem and minimizing fisher information theorem ( cramr - rao inequality )
are proved in section [ ef ] . in section [ wa ] ,
the worst additive noise lemma is introduced and proved based on variational arguments .
eei is proved in section [ ei ] . in section [ app ] ,
some additional applications of the proposed variational techniques are briefly mentioned .
finally , section [ con ] concludes this paper .
in this section , we will review some of the fundamental results from variational calculus , and establish the concepts , notations and results that will be used constantly throughout the rest of the paper .
these results are standard and therefore will be described briefly without further details .
additional details can be found in calculus of variations books such as @xcite-@xcite .
[ cal_def1 ] a functional @xmath0 $ ] might be defined as @xmath1 = \int_{a}^{b } k(x , f(x ) , f'(x ) ) dx,\end{aligned}\ ] ] which is defined on the set of continuous functions @xmath2 with continuous first - order derivatives @xmath3 on the interval @xmath4 $ ] .
the function @xmath5 is assumed to satisfy the boundary conditions @xmath6 and @xmath7 . the functional @xmath8 is also assumed to have continuous first - order and second - order ( partial ) derivatives with respect to ( wrt ) all of its arguments . [ cal_def2 ] the increment of a functional @xmath0 $ ] is defined as @xmath9 = u[f + t ] - u[f],\end{aligned}\ ] ] where the function @xmath10 , that satisfies the boundary conditions @xmath11 , represents the admissible increment of @xmath5 , and it is assumed independent of the function @xmath5 and twice differentiable . [ cal_def3 ] suppose that given @xmath5 , the increment in ( [ cal_eq2_1 ] ) is expressed as @xmath12 = \varphi\left[t\right ] + \epsilon \|t\|,\end{aligned}\ ] ] where @xmath13 $ ] is a linear functional , @xmath14 goes to zero as @xmath15 approaches zero , and @xmath16 denotes a norm defined in the case of a function @xmath5 as : @xmath17 where @xmath18 are assumed to exist and be continuous for @xmath19 on the interval @xmath4 $ ] , and the summation upper index @xmath20 might vary depending on the normed linear space considered ( e.g. , if the normed linear space consists of all continuous functions @xmath21 , which have continuous first - order derivative on the interval @xmath22 $ ] , @xmath23 , and in this case @xmath24 ; see e.g. , @xcite for further details ) . under the above assumptions , the functional @xmath25 $ ] is said to be differentiable , and the major part of the increment @xmath13 $ ] is called the ( first - order ) variation of the functional @xmath25 $ ] and it is expressed as @xmath26 $ ] .
based on definitions [ cal_def1 ] , [ cal_def2 ] , [ cal_def3 ] and taylor s theorem ( see e.g. , @xcite-@xcite for additional justifications ) , the first - order and the second - order variations of a functional @xmath25 $ ] can be expressed as @xmath27 \hspace*{-3 mm } & = & \hspace*{-4mm}\int \hspace*{-1 mm } \left[k'_{f } \left(x , f , f ' \right ) t ( x)+ k'_{f ' } \left(x , f , f ' \right ) t ' ( x)\right ] dx\\ \label{cal_eq5_2 } \delta^2 u\left[f \right ] \hspace*{-3 mm } & = & \hspace*{-3mm}\frac{1}{2 } \int \big[k''_{f f } \left(x , f , f ' \right ) { t ( x)}^2 + 2 k''_{f f ' } \left(x , f , f ' \right ) t ( x ) t ' ( x ) + k''_{f ' f ' } \left(x , f , f ' \right){t ' ( x)}^{2 } \big ] dx\nonumber\\ \hspace*{-3 mm } & = & \hspace*{-3mm}\frac{1}{2 } \int \left[k''_{f ' f ' } { t ' } ^{2 } + \left ( k''_{f f } - \frac{d}{dx } k''_{f f ' } \right ) { t } ^2 \right ] dx,\end{aligned}\ ] ] where @xmath28 and @xmath29 stand for the first - order partial derivatives wrt @xmath30 and @xmath31 , respectively , @xmath32 denotes the second - order partial derivative wrt @xmath30 and @xmath31 , @xmath33 represents the second - order partial derivative wrt @xmath30 , and @xmath34 is the second - order partial derivative wrt @xmath31 . throughout the paper to simplify the exposition , the arguments of functionals or functions are omitted unless the arguments are ambiguous or confusing . also , the range of integration in various integrals will not be explicitly marked unless the range is ambiguous .
[ cal_thm1 ] a necessary condition for the functional @xmath35 $ ] in ( [ cal_eq1_1 ] ) to have an extremum ( or local optimum ) for a given function @xmath36 is that its first variation vanishes at @xmath36 : @xmath37 = 0,\end{aligned}\ ] ] for all admissible increments.this implies @xmath38 a result which is known as euler s equation .
when the functional in ( [ cal_eq1_1 ] ) includes multiple functions ( e.g. , @xmath39 ) and multiple integrals wrt @xmath40 , i.e. , @xmath41 then euler s equation in ( [ cal_eq7_1 ] ) takes the form of the system of equations : @xmath42 in particular , when the functional does not depend on the first - order derivative of the functions @xmath39 , the equations in ( [ cal_eq7_4 ] ) reduce to @xmath43 details of the proof of this theorem can be found , e.g. , in @xcite-@xcite .
[ cal_thm2 ] a necessary condition for the functional @xmath0 $ ] in ( [ cal_eq1_1 ] ) to have a minimum for a given @xmath36 is that the second variation of functional @xmath0 $ ] be nonnegative : @xmath44 \geq 0,\end{aligned}\ ] ] for all admissible increments .
this implies @xmath45 in particular , when the functional in ( [ cal_eq1_1 ] ) does not depend on the first - order derivative of the function @xmath46 , ( [ cal_eq8_3 ] ) simplifies to @xmath47 when the functional in ( [ cal_eq1_1 ] ) includes multiple functions ( e.g. , @xmath39 ) and multiple integrals wrt @xmath40 ,
i.e. , @xmath48 then the condition in ( [ cal_eq8_5 ] ) is expressed in terms of the positive semi - definiteness of the matrix : @xmath49 \succeq 0 .\end{aligned}\ ] ] the inequality in ( [ cal_eq8_5 ] ) is easily derived from the inequality in ( [ cal_eq8_3 ] ) since @xmath50 and @xmath51 are vanishing in ( [ cal_eq5_2 ] ) when the functional in ( [ cal_eq1_1 ] ) does not depend on the first - order derivative of the function @xmath52 .
the remaining details of the proof can be tracked in @xcite .
[ cal_thm3 ] given the functional @xmath53 = \int_{a}^{b } k(x , f_{1 } , f_{2},f'_{1 } , f'_{2 } ) dx,\end{aligned}\ ] ] assume that the admissible functions satisfy the following boundary conditions : @xmath54 = \int_{a}^{b } \tilde{l}(x , f_{1 } , f_{2 } , f'_{1 } , f'_{2 } ) dx = l,\end{aligned}\ ] ] where @xmath55 , @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 , and @xmath61 are constants , @xmath62 is a functional wrt @xmath63 and @xmath64 , and @xmath65 $ ] is assumed to have an extremum for @xmath66 and @xmath67 .
if @xmath68 and @xmath69 are not extremals of @xmath70 $ ] , or @xmath71 and @xmath72 do not vanish simultaneously at any point in ( [ cal_eq10_2 ] ) , there exist a constant @xmath73 and a function @xmath74 such that @xmath68 and @xmath69 are extremals of the functional @xmath75 based on theorem [ cal_thm3 ] , the following corollary is derived .
[ cal_cor1 ] given the functional @xmath76 = \int_{a}^{b } \int_{a}^{b } k(x , y , \fx , \fy ) dxdy,\end{aligned}\ ] ] assume that the admissible functions satisfy the following boundary conditions : @xmath77 = \int_{a}^{b } \int_{a}^{b } \tilde{l}(x , y,\fx , \fy ) dx dy= l , \label{cal_eq13_1}\end{aligned}\ ] ] where @xmath55 , @xmath56 , @xmath78 , @xmath79 , @xmath80 , and @xmath81 stand for some constants , @xmath82 is a function of @xmath83 , @xmath84 is a function of @xmath85 , @xmath86 is a function of @xmath87 , and @xmath88 is a function of @xmath89 .
the functional @xmath90 $ ] is assumed to have an extremum at @xmath91 and @xmath92 . unless @xmath93 and @xmath94 are extremals of @xmath95 $ ] , or @xmath96 and @xmath97 simultaneously vanish at any point of @xmath98 , there exist a constant @xmath73 and a function @xmath99 such that @xmath100 and @xmath92 is an extremal of the functional @xmath101 + \lambda(y)g(y,\fy ) \big\}dy . \label{cal_eq15_1}\end{aligned}\ ] ] this corollary is a simple extension of theorem [ cal_thm3 ] for multiple integrals .
therefore , the detailed proof is omitted .
based on theorems [ cal_thm1 ] , [ cal_thm2 ] and corollary [ cal_cor1 ] , we can derive the following corollary , which will be repeatedly used throughout this paper .
[ cal_cor2 ] based on the functional defined in ( [ cal_eq15_1 ] ) , the following necessary conditions are derived for the optimal solutions @xmath93 and @xmath94 : @xmath102 and the matrix @xmath103,\end{aligned}\ ] ] is positive semi - definite .
the functional @xmath104 is defined as @xmath105 and @xmath106 is a ( arbitrary but fixed ) function which satisfies @xmath107 , and it is introduced to homogenize the functional in ( [ cal_eq15_1 ] ) .
the equations in ( [ cal_eq16_1 ] ) and ( [ cal_eq16_2 ] ) are derived from the first - order variation condition in theorem [ cal_thm1 ] .
the equations in ( [ cal_eq16_1 ] ) and ( [ cal_eq16_2 ] ) are euler s equations for multiple integrals .
the positive semi - definiteness of the matrix in ( [ cal_eq17_1 ] ) is derived from the second - order variation condition in theorem [ cal_thm2 ] .
this condition is the same as the one in ( [ cal_eq8_7 ] ) .
the details of the proof are omitted here .
this simple but significant result given the second - order moment ( or variance ) of a random vector , a gaussian random vector maximizes the differential entropy is well - known . in this section , a completely rigorous and general derivation of the distribution achieving the maximum entropy will be first provided .
this proof sets up the variational framework for establishing a second important result in this section , namely the cramr - rao bound , which states that for a given mean and correlation matrix , a normally distributed random vector minimizes the fisher information matrix .
[ ef_thm2 ] given ( a mean vector @xmath108 and ) a correlation matrix @xmath109 , a gaussian random vector @xmath110 with the correlation matrix @xmath109 ( and the mean vector @xmath108 ) maximizes the differential entropy , i.e. , @xmath111 where @xmath112 denotes differential entropy , @xmath113 is an arbitrary ( but fixed ) random vector with the correlation matrix @xmath109 .
we first construct a functional , which represents the inequality in ( [ ef_eq8_1 ] ) and required constraints , as follows : @xmath114 using theorem [ cal_thm3 ] , the functional in ( [ ef_eq30_1 ] ) is expressed as @xmath115,\end{aligned}\ ] ] where @xmath116=$ ] @xmath117 @xmath118 @xmath119 , @xmath120 is the lagrange multiplier associated with the constraint ( [ ef_eq30_2 ] ) , and @xmath121 stand for the lagrange multipliers corresponding to the constraints ( [ ef_eq30_3 ] ) .
based on theorem [ cal_thm1 ] , by checking the first - order variation condition , we can find the optimal solution @xmath122 as follows : @xmath123 with the matrix @xmath124 $ ] , @xmath125 . considering the constraints in ( [ ef_eq30_2 ] ) and ( [ ef_eq30_3 ] ) , from ( [ ef_eq32_1 ] ) it turns out that @xmath126 where @xmath127 two remarks are now in order .
first , the correlation matrix @xmath109 is assumed to be invertible .
when the correlation matrix is non - invertible , similar to the method shown in @xcite , we can equivalently re - write the functional problem in ( [ ef_eq30_1 ] ) as @xmath128 where @xmath129 , and in the spectral factorization @xmath130 , @xmath131 is an orthogonal matrix , @xmath132 @xmath133 and @xmath134 denotes a diagonal matrix . let @xmath135^{\scriptscriptstyle t}$ ] , where the dimensions of @xmath136 and @xmath137 are @xmath138 and @xmath139 , respectively . since the correlation matrix ( or covariance matrix ) of @xmath140 , @xmath141 , is equal to the diagonal matrix @xmath142 , @xmath136 and @xmath137 are statistically independent .
therefore , @xmath143 .
since the correlation of @xmath144 is a zero matrix , we can consider @xmath144 as a deterministic vector , and we only need to consider @xmath145 .
therefore , the equation in ( [ ef_add_eq1 ] ) and constraints in ( [ ef_eq30_2])-([ef_eq30_3 ] ) are equivalently re - written as @xmath146 where @xmath147 is a positive - definite matrix .
therefore , without loss of generality , we assume that the correlation matrix @xmath109 is invertible .
second , if an additional constraint , related to the mean vector of @xmath113 , @xmath108 , is enforced , the optimal solution is a multi - variate gaussian density function with mean @xmath108 , instead of the multi - variate gaussian density function with zero mean ( [ ef_eq33_1 ] ) .
based on theorem [ cal_thm2 ] , since @xmath148 the second - order variation @xmath149 $ ] is positive , and the optimal solution @xmath150 is a minimal solution for the variational problem in ( [ ef_eq30_1 ] ) .
therefore , the negative of differential entropy @xmath151 is minimized ( or equivalently @xmath152 is maximized ) when @xmath113 is a multi - variate gaussian random vector with zero mean and covariance matrix @xmath153 .
even though theorems [ cal_thm1 ] , [ cal_thm2 ] are necessary conditions for the minimum , in this case , a multi - variate gaussian density function is the actual solution since there is only one solution , namely the multi - variate gaussian density function , in the feasible set .
an alternative justification of global optimality of multi - variate gaussian pdf can be achieved by exploiting the convexity of @xmath154 wrt @xmath89 .
the proposed proof is different from the ones mentioned in @xcite , @xcite in the sense that the proposed proof relies only on variational calculus techniques , and it turns out that the constraints related to the first - order moment or non - singularity of correlation matrix are not necessary . the maximum entropy result can be extended in various ways .
a simple variation consists in considering only non - negative random variables .
then it turns out that a gaussian random variable is not anymore the solution which maximizes the differential entropy . the following theorem can be easily established and states that a half - normal random variable maximizes the differential entropy over the set of non - negative random variables .
[ ef_thm3 ] within the class of non - negative random variables with given second - order moment @xmath155 , a half - normal random variable @xmath156 maximizes the differential entropy , i.e. , @xmath157 where @xmath158 is an arbitrary ( but fixed ) non - negative random variable with the second - order moment @xmath155 , and @xmath112 denotes differential entropy . the proof is omitted since it can be established following similar steps to the proof of theorem [ ef_thm2 ] . adopting a similar variational approach to the one in theorem [ ef_thm2 ]
, we can also determine the probability density function that minimizes the fisher information matrix as shown by the following theorem .
[ ef_thm5 ] given a mean vector @xmath108 and a correlation matrix @xmath109 , the gaussian density function with the mean vector @xmath108 and the correlation matrix @xmath109 minimizes the fisher information matrix , i.e. , @xmath159 where @xmath113 and @xmath110 stand for an arbitrary ( but fixed ) random vector and gaussian random vector , respectively , with given mean @xmath108 and correlation matrix @xmath109 , and @xmath160 denotes the fisher information matrix : @xmath161,\end{aligned}\ ] ] @xmath162 the following regularity condition is assumed to be satisfied : @xmath163 where @xmath164 denotes the gradient , and the covariance matrix is supposed to be non - singular .
we first represent the inequality in ( [ ef_eq25_1 ] ) as a functional with the required constraints as follows : @xmath165 where @xmath166 is an arbitrary but fixed non - zero vector , and it is defined as @xmath167^{\scriptscriptstyle t}$ ] .
using theorem [ cal_thm3 ] , the functional problem in ( [ ef_eq36_1 ] ) is expressed as @xmath168,\end{aligned}\ ] ] where @xmath116= \int k(\mathbf{x},f_{\scriptscriptstyle x } , \nabla f_{\scriptscriptstyle x } ) d\mathbf{x}$ ] , @xmath169 , and @xmath120 , @xmath170 , and @xmath121 are the lagrange multipliers corresponding to the three constraints in ( [ ef_eq36_2 ] ) .
based on theorem [ cal_thm1 ] , by confirming the first - order variation condition , i.e. , @xmath171=0 $ ] , we can find the optimal solution @xmath172 as follows : @xmath173 where @xmath174 therefore , the left - hand side of the equation in ( [ ef_eq38_1 ] ) is expressed as @xmath175 unlike theorem [ ef_thm2 ] , we can not directly calculate @xmath122 from ( [ ef_eq38_1 ] ) .
fortunately , the first two parts in equation ( [ ef_eq40_1 ] ) are expressed as quadratic forms when @xmath122 is a multi - variate gaussian density function , and therefore , the multi - variate gaussian density function satisfies the equality in ( [ ef_eq40_2 ] ) .
when @xmath122 is a multi - variate gaussian density function : @xmath176 with @xmath177 and @xmath178,\end{aligned}\ ] ] its partial derivatives can be expressed as follows : @xmath179 by substituting ( [ ef_eq43_2 ] ) into ( [ ef_eq40_1 ] ) , it turns out that @xmath180\nonumber\\ % & & + \sum_{i=1}^n \sum_{j=1}^n \left(\sigma_{\scriptscriptstyle x_{ij}}^2 + \sigma_{\scriptscriptstyle x_{ji}}^2\right ) \xi_i \xi_j + \alpha + \sum_{i=1}^n \zeta_i x_i + \sum_{i=1}^n \sum_{j=1}^n
\lambda_{ij}x_i x_j\nonumber\\ % & = & \sum_{l=1}^n \sum_{m=1}^n \left[\left(x_l-\mu_{\scriptscriptstyle x_l}\right)\left(x_m - \mu_{\scriptscriptstyle x_m}\right)\boldsymbol{\xi}^{\scriptscriptstyle t}\boldsymbol{\sigma}_{\scriptscriptstyle x_{lm } } \boldsymbol{\xi } \right]+ \sum_{i=1}^n \sum_{j=1}^n \left(\sigma_{\scriptscriptstyle x_{ij}}^2 + \sigma_{\scriptscriptstyle x_{ji}}^2\right ) \xi_i
\xi_j \nonumber\\ % & & + \alpha + \sum_{i=1}^n \zeta_i x_i + \sum_{i=1}^n \sum_{j=1}^n \lambda_{ij}x_i x_j\nonumber\\ \hspace*{-2 mm } & = & \hspace*{-2 mm } \sum_{l=1}^n \sum_{m=1}^n \omega_{lm}\left(x_l-\mu_{\scriptscriptstyle x_l}\right)\left(x_m-\mu_{\scriptscriptstyle x_m}\right ) + \sum_{i=1}^n \sum_{j=1}^n \left(\sigma_{\scriptscriptstyle x_{ij}}^2\sigma_{\scriptscriptstyle x_{ji}}^2\right ) \xi_i \xi_j+\alpha + \sum_{i=1}^n
\zeta_i x_i + \sum_{i=1}^n \sum_{j=1}^n \lambda_{ij}x_i x_j\nonumber\\ % & = & \left(\mathbf{x } - \boldsymbol{\mu}_{\scriptscriptstyle x}\right)^{\scriptscriptstyle t } \boldsymbol{\omega } \left(\mathbf{x}- \boldsymbol{\mu}_{\scriptscriptstyle x}\right)+ \boldsymbol{\xi}^{\scriptscriptstyle t } \boldsymbol{\psi }
\boldsymbol{\xi } + \alpha + \boldsymbol{\zeta}^{\scriptscriptstyle t } \mathbf{x } + \mathbf{x}^{\scriptscriptstyle t}\boldsymbol{\lambda } \mathbf{x}\nonumber\\ \hspace*{-2 mm } & = & \hspace*{-2 mm } \mathbf{x}^{\scriptscriptstyle t } \boldsymbol{\omega } \mathbf{x } + \mathbf{x}^{\scriptscriptstyle t}\boldsymbol{\lambda } \mathbf{x}+\boldsymbol{\zeta}^{\scriptscriptstyle t } \mathbf{x}- 2\boldsymbol{\mu}_{\scriptscriptstyle x}^{\scriptscriptstyle t } \boldsymbol{\omega } \mathbf{x}+ \boldsymbol{\mu}_{\scriptscriptstyle x}^{\scriptscriptstyle t } \boldsymbol{\omega } \boldsymbol{\mu}_{\scriptscriptstyle x } + \boldsymbol{\xi}^{\scriptscriptstyle t } \boldsymbol{\psi } \boldsymbol{\xi } + \alpha \nonumber \\
\hspace*{-2 mm } & = & \hspace*{-2 mm } 0,\end{aligned}\ ] ] where @xmath181 , \ ; \boldsymbol{\lambda } \hspace*{-1mm}=\hspace*{-1mm}\left[\begin{array}{ccc } \lambda_{11 } & \cdots & \lambda_{1n}\\ \vdots & \ddots & \vdots \\ \lambda_{n1 } & \cdots & \lambda_{nn } \end{array}\right ] , \nonumber\\ \boldsymbol{\psi } \hspace*{-4mm}&=&\hspace*{-4 mm } \left[\begin{array}{ccc } \psi_{11 } & \cdots & \psi_{1n}\\ \vdots & \ddots & \vdots \\ \psi_{n1 } & \cdots & \psi_{nn } \end{array}\right ] , \ ; \boldsymbol{\omega } \hspace*{-1mm}=\hspace*{-1 mm } \left[\begin{array}{ccc } \omega_{11 } & \cdots & \omega_{1n}\\ \vdots & \ddots & \vdots \\ \omega_{n1 } & \cdots & \omega_{nn } \end{array}\right]\nonumber\\ \sigma_{\scriptscriptstyle x_{ij}}^{\scriptscriptstyle lm } \hspace*{-3 mm } & = & \hspace*{-3 mm } \frac{1}{4 } \left(\sigma_{\scriptscriptstyle x_{il}}^2+\sigma_{\scriptscriptstyle x_{li}}^2\right)\left(\sigma_{\scriptscriptstyle x_{jm}}^2+\sigma_{\scriptscriptstyle x_{mj}}^2\right)\nonumber\\ \hspace*{-3 mm } & = & \hspace*{-3 mm } \sigma_{\scriptscriptstyle x_{li}}^2 \sigma_{\scriptscriptstyle x_{jm}}^2,\ ; i , j=1,\ldots , n,\quad l , m=1,\ldots , n \nonumber\\ \psi_{ij } \hspace*{-3 mm } & = & \hspace*{-3 mm } 2\sigma_{\scriptscriptstyle x_{ij}}^2 , \ ; i , j=1,\ldots , n \nonumber\\ \omega_{lm } \hspace*{-3 mm } & = & \hspace*{-3 mm } \boldsymbol{\xi}^{\scriptscriptstyle t}\boldsymbol{\sigma}_{\scriptscriptstyle x_{lm } } \boldsymbol{\xi } , \ ; l , m=1,\ldots , n.\end{aligned}\ ] ] therefore , the lagrange multipliers @xmath120 and @xmath121 must be selected as @xmath182 since the second - order variation is positive : @xmath183 based on theorem [ cal_thm2 ] , the gaussian distribution @xmath122 is necessary optimal for the variational problem in ( [ ef_eq36_1 ] ) .
even though theorems [ cal_thm1 ] and [ cal_thm2 ] are necessary conditions for the minimum , in this case , the multi - variate gaussian density function is sufficiently the global minimum solution since this is a convex optimization problem ( the objective function is strictly convex and its constraint set is convex ) . using similar variational arguments
, one can show that a half - normal and a chi density function minimize the fisher information over the set of non - negative random variables as shown by the following two theorems .
[ ef_thm6 ] within the class of non - negative continuous random variables with fixed second - order moment @xmath155 , the fisher information is minimized by a half - normal random variable @xmath156 : @xmath184 where @xmath158 is an arbitrary ( but fixed ) non - negative random variable with the second - order moment @xmath155 , and @xmath185 denotes the fisher information .
theorem [ ef_thm6 ] does not assume the following regularity condition : @xmath186 for the fisher information .
the following result establishes the counterpart of theorem [ ef_thm6 ] for the class of non - negative random variables with fixed second order moment and whose distribution satisfies the regularity condition in ( [ ef_eq28_1_1 ] ) .
[ ef_thm7 ] within the class of non - negative continuous random variables @xmath158 with fixed second - order moment and whose distributions satisfy the regularity condition in ( [ ef_eq28_1_1 ] ) , the fisher information is minimized by a chi - distributed random variable @xmath187 : @xmath188 where @xmath185 stands for the fisher information . unlike the proof in @xcite , by considering the first - order and the second - order moments instead of variance
, we construct a variational problem and address the problem using the first - order and second - order necessary conditions , as well as the convexity property of the problem .
the details of the proof are omitted because of the similar steps to those encountered in the proof of theorem [ ef_thm5 ] .
worst additive noise lemma was introduced and exploited in several references @xcite , @xcite , @xcite , and it has been widely used in numerous other applications .
one of the main applications of the worst additive noise lemma pertains to the capacity calculation of a wireless communication channel subject to different constraints such as gaussian mimo broadcasting , gaussian mimo wire - tap , etc . in this section , the worst additive noise lemma for random vectors
will be proved solely based on variational arguments .
[ wa_thm2 ] assume @xmath113 is an arbitrary but fixed random vector and @xmath110 is a gaussian random vector , whose correlation matrix is identical to that of @xmath113 , denoted as @xmath109 .
given a gaussian random vector @xmath189 , assumed independent of both @xmath113 and @xmath110 and with the correlation matrix @xmath190 , then the following relation holds : @xmath191 our novel proof is entirely anchored in the variational calculus framework .
a summary of our proof runs as follows .
first , we construct a variational problem , which represents the inequality in ( [ wa_eq14_1 ] ) and required constraints in a functional form .
second , using the first - order variation condition , we find necessary optimal solutions , which satisfy euler s equation .
third , using the second - order variation condition , we show that the optimal solutions are necessarily local minima .
finally , we justify that the local minimum is also global . by setting @xmath192 , where @xmath113 and @xmath189 are independent of each other , in ( [ wa_eq14_1 ] )
, the mutual information @xmath193 can be expressed as @xmath194 then , we consider the functional : @xmath195 the density function @xmath196 and conditional density function @xmath197 are expressed as @xmath198 respectively . therefore , by substituting @xmath196 for @xmath199 and @xmath200 for @xmath197 , respectively , and appropriately changing the constrains in ( [ wa_eq15_2 ] ) , the variational problem in ( [ wa_eq15_1 ] ) is expressed as @xmath201 d\mathbf{x}d\mathbf{y } \\ \label{wa_eq17_3_1 } \hspace*{-4mm}&&\hspace*{-4mm}\text{s
. t. } \int \hspace*{-2 mm } \int \hspace*{-2 mm } f_{\scriptscriptstyle x}(\mathbf{x})f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x } ) d\mathbf{x } d\mathbf{y } = 1,\\ \label{wa_eq17_3_2 } \hspace*{-3 mm } & & \hspace*{+2mm}\int \hspace*{-2 mm } \int \hspace*{-1mm}\mathbf{x } f_{\scriptscriptstyle x}(\mathbf{x})f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x } ) d\mathbf{x } d\mathbf{y } = \boldsymbol{\mu}_{\scriptscriptstyle x},\\ \label{wa_eq17_3_3 } \hspace*{-3 mm } & & \hspace*{+2mm}\int \hspace*{-2 mm } \int \hspace*{-1 mm } \mathbf{x}\mathbf{x}^{\scriptscriptstyle t } f_{\scriptscriptstyle x}(\mathbf{x})f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x } ) d\mathbf{x } d\mathbf{y } = \boldsymbol{\omega}_{\scriptscriptstyle x},\\ \label{wa_eq17_3_4 } \hspace*{-3 mm } & & \hspace*{+2 mm } \int f_{\scriptscriptstyle y}(\mathbf{y } ) d\mathbf{y } = 1,\\ \label{wa_eq17_3_5 } \hspace*{-3 mm } & & \hspace*{+2 mm } \int \mathbf{y } f_{\scriptscriptstyle y}(\mathbf{y } ) d\mathbf{y } = \boldsymbol{\mu}_{\scriptscriptstyle y},\\ \label{wa_eq17_3_6 } \hspace*{-3 mm } & & \hspace*{+2mm}\int \mathbf{y}\mathbf{y}^{\scriptscriptstyle t } f_{\scriptscriptstyle y}(\mathbf{x } ) d\mathbf{y } = \boldsymbol{\omega}_{\scriptscriptstyle y},\\ \label{wa_eq17_3_7 } \hspace*{-3 mm } & & \hspace*{+2mm}f_{\scriptscriptstyle y}(\mathbf{y } ) = \int f_{\scriptscriptstyle x}(\mathbf{x})f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x}.\end{aligned}\ ] ] the functional problem in ( [ wa_eq17_1 ] ) can be re - cast into the following equivalent form : @xmath202 d\mathbf{x})\nonumber\\ \hspace*{-4mm}&+&\hspace*{-4 mm } f_{\scriptscriptstyle y}(\mathbf{y})\left[\alpha_1 + \sum_{i=1}^n \eta_{i } y_i + \sum_{i=1}^n \sum_{j=1}^n \theta_{ij } y_i y_j + \lambda(\mathbf{y})\right]d\mathbf{y},\end{aligned}\ ] ] where @xmath203 $ ] , @xmath204 $ ] , and @xmath205 , @xmath206 , @xmath207 , @xmath208 , @xmath209 , @xmath210 , and @xmath211 stand for the lagrange multipliers corresponding to the constraints ( [ wa_eq17_3_1 ] ) , ( [ wa_eq17_3_2 ] ) , ( [ wa_eq17_3_3 ] ) , ( [ wa_eq17_3_4 ] ) , ( [ wa_eq17_3_5 ] ) , ( [ wa_eq17_3_6 ] ) , and ( [ wa_eq17_3_7 ] ) , respectively .
define now the functional @xmath212 as @xmath213 \hspace*{-2 mm } & = & \hspace*{-2 mm } \int \left(\int k(\mathbf{x},\mathbf{y},f_{\scriptscriptstyle x},f_{\scriptscriptstyle y } ) d\mathbf{x}\right ) + \tilde{k}(\mathbf{y},f_{\scriptscriptstyle y } ) d\mathbf{y},\nonumber\end{aligned}\ ] ] where @xmath214,\nonumber\\ \hspace*{-4mm}&&\hspace*{-4mm}\tilde{k}(\mathbf{y},f_{\scriptscriptstyle y})=f_{\scriptscriptstyle y}(\mathbf{y})[\alpha_1+\hspace*{-1mm}\sum_{i=1}^n \hspace*{-1mm}\eta_{i } y_i+\hspace*{-1mm}\sum_{i=1}^n \sum_{j=1}^n \hspace*{-1mm}\theta_{ij}y_i y_j+\lambda(\mathbf{y } ) ] .\end{aligned}\ ] ] based on the first - order variation condition , we can find the optimal solution @xmath150 and @xmath215 as follows : @xmath216 where @xmath217 , \ ; \boldsymbol{\theta } = \left[\begin{array}{ccc } \theta_{11 } & \cdots & \theta_{1n}\\ \vdots & \ddots & \vdots\\ \theta_{n1 } & \cdots & \theta_{nn } \end{array}\right]\end{aligned}\ ] ] @xmath218^{\scriptscriptstyle t}$ ] and @xmath219^{\scriptscriptstyle t}$ ] .
the following relationships satisfy the necessary conditions ( [ wa_eq21_2 ] ) and ( [ wa_eq21_3 ] ) : @xmath220 and hence : @xmath221 considering the constraints in ( [ wa_eq17_3_1])-([wa_eq17_3_7 ] ) , @xmath122 and @xmath222 in ( [ wa_eq24_1 ] ) are expressed as @xmath223 where @xmath224 , @xmath225 , and @xmath226 is the covariance matrix of @xmath189 .
based on the equations in ( [ wa_eq25_1 ] ) , it turns out that @xmath227 therefore , @xmath150 and @xmath215 are multi - variate gaussian density functions ( without loss of generality , and we can assume that the covariance matrix @xmath153 is invertible due to the arguments mentioned in appendix [ non ] ) .
now , by confirming the second - order variation condition , we will show that the optimal solutions @xmath150 and @xmath215 are necessarily local minima . using theorem [ cal_thm2 ]
, we will show that the following matrix is positive semi - definite : @xmath228 \bigg|_{\fx=\fxstar,\fy=\fystar } \succeq { \bf 0}.\end{aligned}\ ] ] since the elements of the matrix in ( [ wa_eq27_1 ] ) are defined as @xmath229 the matrix is a positive semi - definite matrix , and therefore @xmath230 . because of the convexity of functional @xmath231 wrt variables @xmath52 and @xmath232 , the optimal solutions @xmath150 and @xmath215 actually globally minimize the variational functional in ( [ wa_eq17_1 ] ) . even though these optimal solutions are necessarily optimal
, there exists only one solution , which is the multi - variate gaussian density function and it satisfies euler s equation in ( [ wa_eq21_2 ] ) and ( [ wa_eq21_3 ] ) .
therefore , @xmath150 and @xmath215 are also sufficient in this case .
an alternative more detailed proof of the fact that @xmath150 and @xmath215 represent global optimal solutions is to show that @xmath233 \geq u[\fxstar,\fystar]$ ] , where @xmath234 denote any arbitrary functions satisfying the boundary conditions and the constraints .
first , the following functionals are defined : @xmath235 , \\
f_0(\x,\y,\fx ) & = & \fx(\x)\fw(\y-\x ) , \\
f_1(\x,\y,\fx ) & = & x_i \fx(\x)\fw(\y-\x ) , \\
f_2(\x,\y,\fx ) & = & x_i x_j \fx(\x)\fw(\y-\x),\end{aligned}\ ] ] and thus @xmath236 can be expressed as @xmath237 since the hessian matrix of @xmath236 wrt @xmath89 and @xmath87 is given by @xmath238,\ ] ] which is positive semi - definite , @xmath236 is convex wrt @xmath89 and @xmath87 , and the following inequality holds @xmath239\big|_{\fx=\fxstar,\fy=\fystar},\ ] ] due to the fact that the convex function lies above its tangents .
therefore , it follows that @xmath240 - u[\fxstar,\fystar ] \hspace{-3 mm } & = & \hspace{-3 mm } \iint \hspace{-2 mm } f(\x,\y,\fxhat,\fyhat ) - f(\x,\y,\fxstar,\fystar ) d\x d\y \nonumber \\
\hspace{-3 mm } & = & \hspace{-3 mm } \iint \hspace{-2 mm } f(\x,\y,\fxhat,\fyhat ) - f(\x,\y,\fxstar,\fystar ) d\x d\y + \alpha_0 \iint \hspace{-2 mm } f_0(\x,\y,\fxhat ) - f_0(\x,\y,\fxstar ) d\x d\y \nonumber
\\ \hspace{-3 mm } & + & \hspace{-3 mm } \sum\limits_{i=1}^n \hspace{-1mm}\zeta_i \iint \hspace{-2 mm } f_1(\x,\y,\fxhat ) - f_1(\x,\y,\fxstar ) d\x d\y + \sum\limits_{i=1}^n \sum\limits_{j=1}^n \hspace{-1mm}\gamma_{ij } \iint \hspace{-2 mm } f_2(\x,\y,\fxhat ) - f_2(\x,\y,\fxstar ) d\x d\y \nonumber
\\ \hspace{-3 mm } & + & \hspace{-3 mm } \alpha_1 \int \hspace*{-1 mm } \fyhat(\y ) - \fystar(\y ) d\y \hspace*{-.1 cm } + \hspace*{-.1 cm } \sum\limits_{i=1}^n \eta_i \hspace*{-1 mm } \int \hspace{-2 mm } y_i ( \fyhat(\y ) - \fystar(\y ) ) d\y + \sum\limits_{i=1}^n \sum\limits_{j=1}^n \hspace{-2 mm } \theta_{ij } \int \hspace{-2 mm } y_i y_j ( \fyhat(\y ) - \fystar(\y ) ) d\y \nonumber \\ \hspace{-3 mm } & + & \hspace{-3 mm } \int \hspace{-2 mm } \lambda(\y ) \bigg [ \fyhat(\y ) - \int \hspace{-2 mm } \fxhat(\x)\fw(\y-\x ) d\x \bigg ] d\y - \int \hspace{-2 mm } \lambda(\y ) \bigg [ \fystar(\y ) - \int \fxstar(\x)\fw(\y-\x ) d\x \bigg ] d\y \nonumber \\ \hspace{-3 mm } & = & \hspace{-3 mm } \iint \hspace{-1 mm } k(\x,\y,\fxhat,\fyhat ) - k(\x,\y,\fxstar,\fystar ) d\x d\y + \int \hspace{-1 mm } ( \fyhat - \fystar ) ( \alpha_1 + \bm\eta^{\subt}\y + \y^{\subt}\bm\theta\y + \lambda(\y ) ) d\y \nonumber \\\end{aligned}\ ] ] based on ( [ eq : wanl - tangent ] ) , the righthand side of ( [ eq : wanl - suff ] ) can be lower bounded as follows : @xmath241 - u[\fxstar,\fystar ] & \\ \geq & \hspace*{-.1 cm } \iint \hspace*{-.1 cm } \left [ ( \fxhat \hspace*{-.1 cm } - \hspace*{-.1 cm } \fxstar ) k'_{\fx } \hspace*{-.12 cm } + \hspace*{-.1 cm } ( \fyhat \hspace*{-.1 cm } - \hspace*{-.1 cm } \fystar ) k'_{\fy } \right]\big|_{\fx=\fxstar,\fy=\fystar } \hspace*{-.15 cm } d\x d\y \\ & + \int ( \fyhat - \fystar ) ( \alpha_1 + \bm\eta^{\subt}\y + \y^{\subt}\bm\theta\y + \lambda(\y ) ) d\y \\ \overset{(a)}{= } & \int ( \fxhat-\fxstar ) \left [ \int k'_{\fx } \big|_{\fx=\fxstar,\fy=\fystar } d\y \right ] d\x \\ & + \int ( \fyhat - \fystar ) \left [ \int k'_{\fy } d\x + \tilde{k}'_{\fy } \right ] \big|_{\fx=\fxstar,\fy=\fystar } d\y \\ \overset{(b)}{= } & 0 , \end{split}\ ] ] where ( a ) follows from the fact that @xmath242 and ( b ) is due to ( [ wa_eq21_2 ] ) and ( [ wa_eq21_3 ] ) .
this proves the sufficiency of the gaussian distributions , and therefore , @xmath150 and @xmath215 minimize the variational problem .
the constraints related to the vector means in ( [ wa_eq17_3_2 ] ) and ( [ wa_eq17_3_5 ] ) are unnecessary . without these constraints , the optimal solutions are still multi - variate gaussian density functions but the vector means are equal to zero .
extremal entropy inequality , proposed by liu and viswanath @xcite , was motivated by multi - terminal information theoretic problems such as the vector gaussian broadcast channel and the distributed source coding with a single quadratic distortion constraint .
eei is an entropy power inequality which includes a covariance constraint . because of the covariance constraint
, the extremal entropy inequality could not be proved directly by using the classical entropy power inequality ( epi ) .
therefore , new techniques ( @xcite , @xcite ) were adopted in the proofs reported in @xcite , @xcite . in this section
, the extremal entropy inequality will be proved using a variational approach .
[ ei_thm2 ] assume that @xmath243 is an arbitrary but fixed constant and @xmath244 is a positive semi - definite matrix . a gaussian random vector @xmath189 with positive definite covariance matrix @xmath226
is assumed to be independent of an arbitrary random vector @xmath113 whose covariance matrix @xmath153 satisfies @xmath245 .
then , there exists a gaussian random vector @xmath246 with covariance matrix @xmath247 which satisfies the following inequality : @xmath248 where @xmath249 . by setting @xmath250 , we first consider the following variational problem ( without loss of generality , we assume that @xmath113 , @xmath251 , and @xmath252 have zero mean ) : @xmath253 where @xmath254 is a constant , and @xmath255 stands for the covariance matrix of the optimal solution @xmath252 .
in addition , the term @xmath256 is added to the objective functional ( [ ei_eq20_1 ] ) , and being a constant , it does not affect the optimization problem . without loss of generality , the matrix @xmath244 is assumed to be a positive definite matrix due to the same arguments mentioned in @xcite .
the optimization problem ( [ ei_eq20_1 ] ) can be better addressed by re - casting it as follows : @xmath257 d\mathbf{x } d\mathbf{y}\\ \label{ei_eq21_5_1 } \hspace*{-4mm}&&\hspace*{-4mm}\text{s.t.}\ ; \int \hspace*{-2 mm } \int f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x}d\mathbf{y } = 1,\\ \label{ei_eq21_5_2 } \hspace*{-4mm}&&\hspace*{-1 mm } \int \hspace*{-2 mm } \int \left(y_i y_j -x_i x_j - \left(y - x\right)_i \left(y - x\right)_j \right ) f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x } d\mathbf{y } = 0,\\ \label{ei_eq21_5_3 } \hspace*{-4mm}&&\hspace*{-1mm}\sum_{i=1}^{n}\sum_{j=1}^n\left ( \int\int x_i x_j \xi_i \xi_j f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x}d\mathbf{y } \right ) \leq \sum_{i=1}^{n}\sum_{j=1}^n\sigma^2_{ij } \xi_i \xi_j,\\ \label{ei_eq21_5_4 } \hspace*{-4mm}&&\hspace*{-1 mm } \int \hspace*{-2mm}\int y_iy_j f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x}d\mathbf{y } = { \sigma}^2_{\scriptscriptstyle y^*_{ij } } , \\
\label{ei_eq21_5_5 } \hspace*{-4mm}&&\hspace*{-1 mm } -\int \hspace*{-2mm}\int f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x } ) \log f_{\scriptscriptstyle x}(\mathbf{x } ) d\mathbf{x } d\mathbf{y } \geq p_{\scriptscriptstyle x},\\ \label{ei_eq21_5_6 } \hspace*{-4mm}&&\hspace*{-1 mm } f_{\scriptscriptstyle y}(\mathbf{y } ) = \int f_{\scriptscriptstyle x}(\mathbf{x } ) f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x})d\mathbf{x},\end{aligned}\ ] ] where the arbitrary deterministic non - zero vector @xmath166 is defined as @xmath258^t$ ] , @xmath259 denotes the @xmath260 row and @xmath261 column entry of @xmath255 , @xmath262 , and @xmath263 . using lagrange multipliers , as shown in corollary [ cal_cor1 ]
, the functional problem in ( [ ei_eq21_1 ] ) and the constraints in ( [ ei_eq21_5_1])-([ei_eq21_5_6 ] ) can be expressed in terms of the lagrangian : @xmath264 where @xmath265,\nonumber\\ \tilde{k}(\mathbf{y},f_{\scriptscriptstyle y } ) \hspace*{-2 mm } & = & \hspace*{-2 mm } \lambda(\mathbf{y})f_{\scriptscriptstyle y}(\mathbf{y}).\end{aligned}\ ] ] the lagrange multipliers @xmath205 , @xmath207 , @xmath266 , @xmath267 , @xmath208 , and @xmath211 correspond to the constraints in ( [ ei_eq21_5_1 ] ) , ( [ ei_eq21_5_2 ] ) , ( [ ei_eq21_5_3 ] ) , ( [ ei_eq21_5_4 ] ) , ( [ ei_eq21_5_5 ] ) , and ( [ ei_eq21_5_6 ] ) , respectively . to find the optimal solutions , based on corollary [ cal_cor2 ] , the first - order variation condition is checked as follows : @xmath268 d\y \nonumber \\ \hspace*{-2 mm } & = & \hspace*{-2 mm } 0.\\ \label{ei_eq24_2 } \hspace*{-2mm}&&\hspace*{-2 mm } \int k'_{f_{\scriptscriptstyle y } } d\x + \tilde{k}'_{\fy } \big|_{f_{\scriptscriptstyle x}=f_{\scriptscriptstyle x^ * } , f_{\scriptscriptstyle y}=f_{\scriptscriptstyle y^ * } } = -\frac{\mu \int f_{\scriptscriptstyle x}(\mathbf{x})f_{\scriptscriptstyle w}(\mathbf{y}-\mathbf{x } ) d\mathbf{x}}{f_{\scriptscriptstyle y}(\mathbf{y } ) } + \lambda(\mathbf{y } ) = 0.\end{aligned}\ ] ] the following expressions satisfy the equalities in ( [ ei_eq24_1 ] ) and ( [ ei_eq24_2 ] ) : @xmath269 where @xmath270 , \quad \boldsymbol{\gamma } = \left[\begin{array}{ccc } \gamma_{11 } & \cdots & \gamma_{1n}\\ \vdots & \ddots & \vdots \\ \gamma_{n1 } & \cdots & \gamma_{nn } \end{array}\right ] \nonumber \\ \boldsymbol{\xi } \hspace*{-3 mm } & = & \hspace*{-3 mm } \left[\begin{array}{ccc } \xi_1 \xi_1 & \cdots & \xi_1 \xi_n\\ \vdots & \ddots & \vdots \\ \xi_n \xi_1 & \cdots & \xi_n \xi_n \end{array}\right ] , \nonumber\\ \mathbf{x } \hspace*{-3 mm } & = & \hspace*{-3 mm } \left[x_1 , \cdots , x_n\right]^t,\nonumber\\ \mathbf{y } \hspace*{-3 mm } & = & \hspace*{-3 mm } \left[y_1 , \cdots , y_n\right]^t.\nonumber % \theta \hspace*{-3 mm } & \geq & \hspace*{-3 mm } 0.\end{aligned}\ ] ] now considering the constraints in ( [ ei_eq21_5_1])-([ei_eq21_5_6 ] ) ,
the equations in ( [ ei_eq25_3 ] ) are further processed as follows : @xmath271 where @xmath272 the inequality in ( [ ei_eq28_6 ] ) is due to the second - order variation condition , which will be presented later in this proof .
the inequality ( [ ei_eq28_x ] ) is based on the theory of kkt conditions since the multiplier associated with the inequality constraint is nonnegative .
moreover , the complementary slackness condition in the kkt conditions leads to the following relationship : @xmath273 = 0.\ ] ] based on theorem [ cal_thm2 ] , to make the second variation nonnegative , the positive semi - definiteness of the following matrix is required : @xmath274,\end{aligned}\ ] ] which further reduces to the following condition : @xmath275 \left[\begin{array}{cc } k''_{f_{\scriptscriptstyle x^ * } f_{\scriptscriptstyle x^ * } } & k''_{f_{\scriptscriptstyle x^ * } f_{\scriptscriptstyle y^*}}\\ k''_{f_{\scriptscriptstyle y^ * } f_{\scriptscriptstyle x^ * } } & k''_{f_{\scriptscriptstyle y^ * } f_{\scriptscriptstyle y^ * } } \end{array}\right ] \left[\begin{array}{c } h_{\scriptscriptstyle x } \\
h_{\scriptscriptstyle y}\end{array } \right ] \nonumber\\ \hspace*{-4mm}&=&\hspace*{-4mm}k''_{f_{\scriptscriptstyle x^ * } f_{\scriptscriptstyle x^ * } } h_{\scriptscriptstyle x}^2 + k''_{f_{\scriptscriptstyle y^ * } f_{\scriptscriptstyle y^ * } } h_{\scriptscriptstyle y}^2 + ( k''_{f_{\scriptscriptstyle x^ * } f_{\scriptscriptstyle y^*}}+k''_{f_{\scriptscriptstyle y^ * } f_{\scriptscriptstyle x^ * } } ) h_{\scriptscriptstyle y } h_{\scriptscriptstyle x}\nonumber\\ \hspace*{-4mm}&\geq & \hspace*{-2 mm } 0,\end{aligned}\ ] ] where @xmath276 and @xmath277 are arbitrary admissible functions .
since @xmath278 , @xmath279 , @xmath280 , and @xmath281 are defined as @xmath282 the condition in ( [ ei_eq35_1 ] ) requires @xmath283 which holds true if @xmath284 ( i.e. , @xmath285 ) .
condition @xmath286 is also imposed by the kkt complementary slackness condition corresponding to the constraint ( [ ei_eq21_5_5 ] ) .
therefore , the optimal solutions @xmath150 and @xmath215 minimize the functional problem in ( [ ei_eq21_1 ] ) , and the proof is completed because of convexity of the functional @xmath231 wrt variables @xmath52 and @xmath232 . a more detailed alternative justification of the fact the gaussian distributions @xmath150 and @xmath215 are global minima is next presented .
we will prove the sufficiency of the gaussian distributions by showing @xmath233 \geq u[\fxstar,\fystar]$ ] , where @xmath287 $ ] represents the objective functional in the problem and @xmath288 denote any arbitrary functions satisfying the boundary conditions and the constraints .
first , the following functionals are defined : @xmath289 and thus @xmath290 it can be verified that the hessian matrix of @xmath236 w.r.t @xmath89 and @xmath87 is given by @xmath291,\ ] ] which is positive semi - definite due to ( [ ei_eq28_6 ] ) .
the convexity property of @xmath236 yields that @xmath292 \big|_{\fx=\fxstar,\fy=\fystar},\ ] ] and it follows that @xmath293 \hspace{-1mm}- \hspace{-1mm}u[\fxstar,\fystar ] \hspace*{-4 mm } & = & \hspace*{-4 mm } \iint \hspace*{-1mm}f(\x,\y,\fxhat,\fyhat ) - f(\x,\y,\fxstar,\fystar ) d\x d\y \nonumber \\
\hspace*{-4 mm } & \overset{(a)}{\geq } & \hspace*{-4 mm } \iint \hspace*{-1mm}f(\x,\y,\fxhat,\fyhat ) - f(\x,\y,\fxstar,\fystar ) d\x d\y + \alpha_0 \left [ \iint \hspace*{-1 mm } f_0(\x,\y,\fxhat ) - f_0(\x,\y,\fxstar ) d\x d\y \right ] \nonumber \\ \hspace*{-4 mm } & + & \hspace*{-4 mm } \sum\limits_{i=1}^n \sum\limits_{j=1}^n \gamma_{ij } \left [ \iint \hspace*{-1 mm } f_1(\x,\y,\fxhat ) - f_1(\x,\y,\fxstar ) d\x d\y \right ] + \theta \left [ \iint \hspace*{-1 mm } f_2(\x,\y,\fxhat ) - f_2(\x,\y,\fxstar ) d\x d\y \right ] \nonumber \\ \hspace*{-4 mm } & + & \hspace*{-4 mm } \sum\limits_{i=1}^n \sum\limits_{j=1}^n \phi_{ij } \left [ \iint \hspace*{-1 mm } f_3(\x,\y,\fxhat ) - f_3(\x,\y,\fxstar ) d\x d\y \right ] + \alpha_1 \hspace*{-.1 cm } \left [ \iint \hspace*{-1 mm } f_4(\x,\y,\fxhat ) - f_4(\x,\y,\fxstar ) d\x d\y \right ] \nonumber \\ \hspace*{-4 mm } & + & \hspace*{-4 mm } \int \hspace*{-1 mm } \lambda(\y ) \left [ \fyhat(\y ) - \int \hspace*{-1 mm } \fxhat(\x)\fw(\y-\x ) d\x \right ] d\y - \int \hspace*{-1 mm } \lambda(\y ) \left [ \fystar(\y ) - \int \hspace*{-1 mm } \fxstar(\x)\fw(\y-\x ) d\x \right ] d\y \\ \hspace*{-4 mm } & = & \hspace*{-4 mm } \iint \hspace*{-1 mm } k(\x,\y,\fxhat,\fyhat ) - k(\x,\y,\fxstar,\fystar ) d\x d\y + \int \hspace*{-1 mm } \lambda(\y ) \left ( \fyhat(\y ) - \fystar(\y ) \right ) d\y \nonumber
\\ \hspace*{-4 mm } & \overset{(b)}{\geq } & \hspace*{-4 mm } \iint \hspace*{-1 mm } \left [ ( \fxhat - \fxstar ) k'_{\fx } + ( \fyhat \hspace*{-.1cm}-\hspace*{-.1 cm } \fystar ) k'_{\fy } \right ] \big|_{\fx=\fxstar,\fy=\fystar } d\x d\y + \int \hspace*{-1 mm } \lambda(\y ) \left ( \fyhat(\y ) - \fystar(\y ) \right ) d\y \nonumber \\
\hspace*{-4 mm } & = & \hspace*{-4 mm } \int \hspace*{-1 mm } ( \fxhat-\fxstar ) \left[\int \hspace*{-1 mm } k'_{\fx}\big|_{\fx=\fxstar,\fy=\fystar } d\y \right ] d\x + \int ( \fyhat-\fystar ) \left [ \int \hspace*{-1 mm } k'_{\fy}\big|_{\fx=\fxstar,\fy=\fystar } d\x + \lambda(\y ) \right ] d\y \nonumber \\ \hspace*{-4 mm } & \overset{(c)}{= } & \hspace*{-4 mm } 0,\end{aligned}\ ] ] where the inequality ( a ) follows from the complementary slackness condition in the kkt conditions ( [ eq : eei - vec - comple ] ) .
indeed , since @xmath294 only represents an arbitrary feasible solution and @xmath295 , it follows that @xmath296 = 0,\ ] ] and @xmath297 \leq 0,\ ] ] and therefore , @xmath298
\leq 0.$ ] similarly , the complementary slackness condition associated with ( [ ei_eq21_5_5 ] ) leads to @xmath299\leq 0 $ ] .
in addition , ( b ) is due to ( [ eq : eei - vec - tan ] ) , and ( c ) follows from ( [ ei_eq24_1 ] ) and ( [ ei_eq24_2 ] ) .
this proves the sufficiency of gaussian distributions .
the proposed proof only exploits calculus of variations tools . unlike the previous proofs
, this proof does not adopt neither the channel enhancement technique and epi as in @xcite nor the epi and data processing inequality as in @xcite .
[ ei_thm3 ] assume that @xmath243 is an arbitrary but fixed constant and @xmath244 is a positive semi - definite matrix .
independent gaussian random vectors @xmath189 with covariance matrix @xmath226 and @xmath300 with covariance matrix @xmath301 are assumed to be independent of an arbitrary random vector @xmath113 with covariance matrix @xmath245 .
both covariance matrices @xmath226 and @xmath301 are assumed to be positive definite .
then , there exists a gaussian random vector @xmath246 with covariance matrix @xmath302 which satisfies the following inequality : @xmath303 where @xmath304
. see appendix [ ei_thm3_proof ] .
the proposed proof does not borrow any techniques from @xcite . even though the proposed proof adopts the equality condition for the data processing inequality , a result which was also exploited in @xcite ,
the proposed proof is different from the one in @xcite due to the following features .
first , the proposed proof uses the equality condition of the data processing inequality only once while the proof in @xcite uses it twice . the proof in @xcite exploited the channel enhancement technique twice , which is equivalent to using the equality condition in the data processing inequality .
second , the proposed proof does not use the moment generating function technique unlike the proof proposed in @xcite ; instead the current proof directly exploits a property of the conditional mutual information pertaining to a markov chain .
because of the easiness to incorporate a broad class of constraints , the proposed variational framework finds applicability in a large number of applications
. herein section , we will briefly illustrate some potential applications in this regard and state several open research problems which might be also addressed within the considered functional framework .
the secrecy capacity of gaussian wire - tap channel has been studied by many researchers @xcite , @xcite .
we will approach the gaussian wire - tap problem from the estimation viewpoint , rather than considering the secrecy capacity from an information theoretic perspective .
the following scalar gaussian wire - tap channel is considered : @xmath305 where @xmath158 is an arbitrary but fixed random variable with zero mean and unit variance , @xmath55 is a constant , and @xmath306 and @xmath307 are gaussian random variables with variances @xmath308 and @xmath309 , respectively .
the random variables @xmath306 and @xmath307 are independent of each other , and they have zero mean . in the channel model ( [ wire_eq1_1 ] ) ,
@xmath310 and @xmath311 are considered as a legitimate receiver and as an eavesdropper , respectively .
the goal of this problem is the following .
assume that both receivers use minimum mean square error ( mmse ) estimators .
given the value of the mean square error ( mse ) , which allows to correctly decode the legitimate receiver , what is the optimal distribution which maximizes the difference between the mse in the legitimate receiver and the mse in the eavesdropper ?
the above mentioned problem adopts both practical and reasonable assumptions due to the following reasons .
first , the mmse estimator is an optimal estimator in the sense that it minimizes the mse .
therefore , it is reasonable to use such an optimal estimator .
second , to prevent from eavesdropping , finding the signal distribution that maximizes the difference between the mses corresponding to the legitimate receiver and the eavesdropper , respectively , represents a legitimate design objective . to find the optimal distribution , the following functional problem
is constructed : @xmath312 where @xmath313\right)^2\right]$ ] , @xmath314 $ ] denotes the expectation operator , and @xmath315 is a constant . the optimization problem in ( [ wire_eq2_1 ] ) is expressed as @xmath316|y_2),\\ \label{wire_eq3_2 } & & \text{s.t.}\quad \mathbb{e}\left [ \mathbb{e } \left[x|y_1\right]^2\right ] = 1- r.\end{aligned}\ ] ] the equation in ( [ wire_eq3_1 ] ) is due to the total law of variance and the markov chain @xmath317 . since @xmath318=1 $ ] , the equation ( [ wire_eq3_2 ] ) follows from the constraint in ( [ wire_eq2_1 ] ) .
the objective function in ( [ wire_eq3_1 ] ) is further expressed as @xmath319|y_2\right)=\mathbb{e } \left[\mathbb{e}\left[x|y_1\right]^2\right]-\mathbb{e } \left[\mathbb{e } \left[x|y_2\right]^2\right]\end{aligned}\ ] ] and using the equations ( [ wire_eq3_2 ] ) , ( [ wire_eq4_1 ] ) , the optimization problem in ( [ wire_eq3_1 ] ) is re - formulated in terms of the following variational problem : @xmath320 where @xmath321 and @xmath322 are the probability density functions of @xmath158 and @xmath311 , respectively , and @xmath323 stands for the second - order moment of @xmath311 .
since the first term in ( [ wire_eq4_1 ] ) is given and @xmath324 ^ 2\right ] = \int f_{\scriptscriptstyle y_2}(y)\left(\int x \frac{f_{\scriptscriptstyle y_2|x}(y|x ) f_{\scriptscriptstyle x}(x)}{f_{\scriptscriptstyle y_2}(y ) } dx\right)^2dy,\nonumber\end{aligned}\ ] ] the objective function in ( [ wire_eq5_1 ] ) is derived from the equation ( [ wire_eq3_1 ] ) . also , the additional constraint in ( [ wire_eq5_3 ] ) is required to solve this variational problem .
considering the lagrange multipliers @xmath325 and @xmath99 to account for the constraints in ( [ wire_eq5_3 ] ) and ( [ wire_eq5_3 ] ) , respectively , the following variational problem is constructed : @xmath326 where @xmath327 in accordance with theorem [ cal_thm1 ] , we can determine @xmath328 and @xmath329 to enforce the first - order variation to be zero : @xmath330 taking into account ( [ wire_eq7_1 ] ) , it follows further that @xmath331 & = & \frac{g^*(y)}{f_{\scriptscriptstyle y_2}^*(y ) } = \sqrt{\lambda_1 } y.\end{aligned}\ ] ] since the equation in ( [ wire_eq8_1 ] ) is a linear function of @xmath85 and @xmath332 is a gaussian random variable , therefore , @xmath333 is also a gaussian random variable .
based on theorem [ cal_thm2 ] , it can be verified that the second - order variation is nonnegative . moreover , due to the convexity of @xmath334 wrt @xmath335 and @xmath336 , we can confirm that the solution is optimal , and the proof is completed .
the importance of the variational framework in establishing some fundamental information theoretic inequalities was already illustrated herein paper . at their turn
, these information theoretic inequalities played a fundamental role in establishing other important results and applications .
for example , the minimum fisher information theorem ( cramr - rao inequality ) and maximum entropy theorem were used for developing min - max robust estimation techniques @xcite , results which were recently further extended to the more general framework of noise with arbitrary distribution ( and correlation ) in @xcite and used to explain why the mimo channel estimation scheme proposed in @xcite exhibits a min - max robustness property . along the same line of potential applications , the extensions of the maximum entropy and minimum fisher information results to positive random variables , as stated in theorems [ ef_thm3 ] , [ ef_thm6 ] and [ ef_thm7 ] , play a fundamental role in developing robust clock synchronization algorithms for wireless sensor networks and other wireless networks that rely on message exchanges to acquire the timing information .
a large class of clock synchronization protocols ( see e.g. , tpsn , internet , pbs @xcite ) rely on the two - way message exchange mechanism and for which the timing synchronization approach reduces to estimating a linear regression model for which the distribution of additive noise has positive support but it is otherwise arbitrary @xcite . designing robust timing synchronization algorithms for such protocols is difficult , because of the variability of delay distributions caused by the variable network traffic .
however , this problem can now be resolved at the light of the results brought by theorems [ ef_thm3 ] , [ ef_thm6 ] and [ ef_thm7 ] . by
optimizing the design of timing messages for the scenario of a chi or log - normal distributed delay , then min - max robust time synchronization algorithms could be developed .
the extremal entropy inequality was used in the vector gaussian broadcast channel @xcite , the distributed source coding with a single quadratic distortion constraint problem @xcite , the gaussian wire - tap channel @xcite , and many other problems . even though these applications were traditionally addressed using the information theoretic inequalities
, one can directly approach these applications by means of the proposed variational calculus techniques .
one of the benefits of such a variational approach is the fact that it can cope with many types of constraints as opposed to the eei which is still quite rigid in its formulation . as prof .
max costa suggested the authors of this paper in a private communication , in the context of z gaussian interference channels , such a variational approach might be helpful to develop novel entropy - power - like inequalities , where the limiting variables are gaussian and independent but not anymore identically distributed , and to assess the capacity of the z - gaussian interference channel .
additional important extensions of maximum entropy theorem , minimum fisher information theorem , additive worst noise lemma , and extremal entropy inequality might be envisioned within the proposed variational framework by imposing various restrictions on the range of values assumed by random variables / vectors ( e.g. , random variables whose support is limited to a finite length interval or finite set of values ) or on their second or higher - order moments and correlations .
for example , the problem of finding the worst additive noise under a covariance constraint @xcite as well as establishing multivariate extensions of costa s entropy power inequality @xcite along the lines mentioned by liu et al .
@xcite and palomar @xcite , @xcite might be also addressed within the proposed variational framework .
however , all these challenges together with finding a variational proof of epi remain open research problems for future study .
in this paper , we derived several fundamental information theoretic inequalities using a functional analysis framework .
the main benefit for employing calculus of variations for proving information theoretic inequalities is the fact that the global optimal solution is obtained from the necessary conditions for optimality .
a brief summary of this paper contributions is the following .
first , the entropy maximizing theorem and fisher information minimizing theorem were derived under different assumptions .
second , the worst additive noise lemma was proved from the perspective of a functional problem .
third , the extremal entropy inequality was derived using calculus of variations techniques .
finally , applications and possible extensions that could be addressed within the proposed variational framework were briefly presented .
many open research problems were also formulated .
let @xmath337 and @xmath338 $ ] , where @xmath339 , @xmath109 is a singular matrix , @xmath340 is an orthogonal matrix , and @xmath341 denotes a diagonal matrix .
the correlation matrix of @xmath342 is the zero matrix , and therefore , it is considered as a deterministic vector . without loss of generality , we can assume @xmath343 .
the following matrices are also considered : @xmath344,\nonumber\\ \mathbf{d } & = & \left[\begin{array}{cc } \mathbf{i } & -\mathbf{b}^{\scriptscriptstyle t } \mathbf{c}^{-1 } \\ \mathbf{0 } & \mathbf{i } \end{array}\right],\end{aligned}\ ] ] where the dimensions of @xmath345 , @xmath346 , and @xmath347 are @xmath348 , @xmath349 , and @xmath350 , respectively .
then , @xmath351\left[\begin{array}{c } \bar{\mathbf{x}}_a \\ \mathbf{0 } \end{array}\right ] = \left[\begin{array}{c}\bar{\mathbf{x}}_a \\ \mathbf{0 } \end{array}\right],\nonumber\\ \hspace*{-4 mm } & & \hspace*{-4 mm } \mathbf{d}\mathbf{q}^{\scriptscriptstyle t}_{\scriptscriptstyle \omega } \mathbf{w}_{\scriptscriptstyle g } = \left[\begin{array}{c } \bar{\mathbf{w}}_{\scriptscriptstyle g_a } \\ \bar{\mathbf{w}}_{\scriptscriptstyle g_b } \end{array}\right],\nonumber\\ \label{im_eq0_1 } \hspace*{-4 mm } & & \hspace*{-4 mm } \mathbb{e } \left[\mathbf{d } \mathbf{q}_{\scriptscriptstyle \omega}^{\scriptscriptstyle t } \mathbf{w}_{\scriptscriptstyle g } \mathbf{w}_{\scriptscriptstyle g}^{\scriptscriptstyle t } \mathbf{q}_{\scriptscriptstyle \omega } \mathbf{d}^{\scriptscriptstyle t } \right ] = \left[\begin{array}{cc } \mathbf{a } - \mathbf{b}^{\scriptscriptstyle t } \mathbf{c}^{-1}\mathbf{b } & \mathbf{0 } \\
\mathbf{0 } & \mathbf{c } \end{array}\right].\end{aligned}\ ] ] due to ( [ im_eq0_1 ] ) , the random vectors @xmath352 and
@xmath353 are statistically independent of each other . the left - hand side of the equation in ( [ wa_eq14_1 ] ) can be re - expressed as @xmath354 in ( [ im_eq1_1 ] ) , @xmath137 is considered as a deterministic variable , @xmath353 is given , the term @xmath355 can be ignored in the optimization , and the correlation matrix of @xmath356 is non - singular .
therefore , we can always assume the correlation matrix to be invertible .
first , choose a gaussian random vector @xmath357 whose covariance matrix @xmath358 satisfies @xmath359 and @xmath360 . since the gaussian random vectors @xmath300 and
@xmath189 can be represented as the summation of two independent random vectors @xmath357 and @xmath361 , and the summation of two independent random vectors @xmath357 and @xmath362 , respectively , the left - hand side of the equation in ( [ ei_eq50_1 ] ) is written as follows : @xmath363 since the expression will be minimized over @xmath364 , the last two terms in ( [ ei_eq51_2 ] ) are ignored , and by substituting @xmath252 and @xmath365 for @xmath366 and @xmath367 , respectively , the inequality in ( [ ei_eq50_1 ] ) is equivalently expressed as the following variational problem : @xmath368 where @xmath369 , @xmath370 , @xmath371 , @xmath372 , @xmath373 , @xmath374 , and @xmath375 is the covariance matrix of the optimal solution @xmath376 .
the variational problem in ( [ ei_eq52_1 ] ) is exactly the same as the one in ( [ ei_eq21_1 ] ) .
therefore , using the same method as in the proof of theorem [ ei_thm2 ] , we obtain the following inequality ( see the details in the proof of theorem [ ei_thm2 ] ) : @xmath377 by appropriately choosing @xmath378 and @xmath357 , the right - hand side of the equation in ( [ ei_eq53_1 ] ) is expressed as @xmath379 the equality in ( [ ei_eq54_1 ] ) is due to the equality condition of the data processing inequality in @xcite . for the completeness of the proof , we introduce a technique , which is slightly different from the one in @xcite .
[ ei_lem1 ] when three random vectors @xmath380 , @xmath381 , and @xmath382 represent a markov chain @xmath383 , the following inequality is satisfied : @xmath384 the equality holds if and only if @xmath385 . in lemma [ ei_lem1 ] , @xmath380 , @xmath381 , and @xmath382 are defined as @xmath378 , @xmath386 , and @xmath387 , respectively .
therefore , the equality condition , @xmath385 is expressed as @xmath388 if @xmath389 , the equality in ( [ ei_eq56_5 ] ) is satisfied , the equality condition in lemma [ ei_lem1 ] holds , and therefore , the equality in ( [ ei_eq54_1 ] ) is proved .
the validity of @xmath389 is proved by lemma @xmath390 in @xcite .
g. aubert and p. kornprobst , _ mathematical problems in image processing : partial differential equations and the calculus of variations_. applied mathematical sciences vol . 147 .
springer verlag .
new york , 2006 .
h. weingarten , y. steinberg , and s. shamai , `` the capacity region of the gaussian mutiple - input multiple - output broadcast channel , '' _ ieee trans .
inform . theory _
9 , pp . 3936 - 3964 , sep .
2006 .
m. payaro , m. gregori , and d. palomar , yet another power entropy inequality with an application , " _ 2011 international conference on wireless communications and signal processing ( wcsp ) _ , nanjing , china , nov .
2011 , pp . 1 - 5 .
m. payaro and d. palomar , a multivariate generalization of costa s entropy power inequality , " _ ieee international symposium in information theory 2008 ( isit 2008 ) _ , toronto , canada , jul .
2008 , pp .
1088 - 1092 .
sangwoo park received the b.s .
degree in electrical engineering from chung - ang university ( cau ) , seoul , korea , in 2004 , and the m.s . and ph.d .
degrees in electrical engineering from texas a&m university , college station , in 2008 and 2012 , respectively . from 2004 to 2005 , he worked as a full - time assistant engineer for umts / wcdma projects in samsung electronics . currently , he is a research engineer at kt ( korea telecom ) in korea .
his research interests lie in wireless communications , information theory , and statistical signal processing .
erchin serpedin(f13 ) received the specialization degree in signal processing and transmission of information from ecole superieure delectricite ( supelec ) , paris , france , in 1992 , the m.sc .
degree from the georgia institute of technology , atlanta , in 1992 , and the ph.d .
degree in electrical engineering from the university of virginia , charlottesville , in january 1999 .
he is currently a professor in the department of electrical and computer engineering at texas a&m university , college station .
he is the author of two research monographs , one edited textbook , 100 journal papers and 150 conference papers , and has served as associate editor for about 10 journals such as ieee transactions on information theory , ieee transactions on communications , signal processing ( elsevier ) , ieee transactions on signal processing , ieee transactions on wireless communications , ieee communications letters , ieee signal processing letters , phycom , eurasip journal on advances in signal processing , and eurasip journal on bioinformatics and systems biology .
his research interests include signal processing , wireless communications , computational statistics , bioinformatics and systems biology .
khalid qaraqe(m97-s00 ) received with honors the b.s . degree in ee from the university of technology , baghdad , irak , in 1986 .
he received the m.s .
degree in ee from the university of jordan , jordan , in 1989 , and he earned his ph.d .
degree in ee from texas a&m university , college station , tx , in 1997 . from 1989 to 2004 ,
qaraqe held a variety of positions in many companies .
he has over 15 years of experience in the telecommunications industry .
qaraqe has worked for qualcomm , enad design systems , cadence design systems / tality corporation , stc , sbc and ericsson .
he has worked on numerous gsm , cdma , wcdma projects and has experience in product development , design , deployment , testing and integration .
qaraqe joined texas a&m university at qatar , in july 2004 , where he is now a professor .
qaraqe research interests include communication theory and its application to design and performance analysis of cellular systems and indoor communication systems .
particular interests are in the development of 3 g umts , cognitive radio systems , broadband wireless communications and diversity techniques .
|
this paper proposes a unifying variational approach for proving some fundamental information theoretic inequalities .
fundamental information theory results such as maximization of differential entropy , minimization of fisher information ( cramr - rao inequality ) , worst additive noise lemma , and extremal entropy inequality ( eei ) are interpreted as functional problems and proved within the framework of calculus of variations .
several applications and possible extensions of the proposed results are briefly mentioned .
maximizing entropy , minimizing fisher information , worst additive noise , extremal entropy inequality , calculus of variations
|
the developments of new radioactive ion beam facilities and new detection techniques have largely extended our knowledge on nuclear physics from stable nuclei to unstable nuclei far from the @xmath1-stability line so - called exotic nuclei .
novel and striking features have been found in the nuclear structure of exotic nuclei , such as halo phenomenon @xcite and the disappearance of traditional magic numbers and occurrence of new ones @xcite . in order to describe the exotic nuclei with large space distribution , theoretical approaches
should be developed in coordinate space or coordinate - equivalent space .
density functional theory ( dft ) and its covariant version ( cdft ) have been proved to be effective theories for the description of exotic nuclei @xcite . in comparison with its nonrelativistic counterpart ,
the cdft has many attractive advantages , such as the natural inclusion of nucleon spin freedom , new saturation property of nuclear matter @xcite , large spin - orbit splittings in single particle energies , reproducing the isotopic shifts of pb isotopes @xcite , natural inclusion of time - odd mean field , explaining pseudospin of nucleon and spin symmetries of antinucleon in nuclei @xcite . in the cdft description for exotic nuclei , the dirac equation of the nucleons
should be solved in coordinate space or coordinate - equivalent space . for the spherical system ,
the conventional shooting method works quite well @xcite , which however is rather complicated for the deformed system @xcite .
therefore the dirac woods - saxon basis expansion method was developed @xcite and has been widely used to solve the deformed dirac equation @xcite , which , however , is highly computational time consuming for the heavy system . imaginary time method ( itm ) @xcite is a powerful approach for the self - consistent mean - field calculations in a three - dimensional ( 3d ) coordinate space .
itm has been successfully employed in nonrelativistic self - consistent mean - field calculations @xcite . for a long time
, there exist doubts on the access of itm to dirac equation due to the dirac sea , i.e. , the relativistic ground state within fermi sea is a saddle point rather than a minimum .
this is the so - called variational collapse problem @xcite . to avoid the variational collapse , zhang _ et al _
@xcite applies the itm to the schrdinger - equivalent form of the dirac equation in spherical case .
the same method is used to solve dirac equation with a nonlocal potential in ref .
@xcite . based on the idea of hill and krauthauser @xcite , hagino and
tanimura proposed the inverse hamiltonian method ( ihm ) to avoid variational collapse @xcite .
this method solves the dirac equation directly and the dirac spinor is obtained simultaneously .
meanwhile when the ihm method is applied to lattice space in the numerical calculations , another challenge appears , i.e. , fermion doubling problem @xcite due to the replacement of the derivative by finite difference method @xcite .
this problem appears also in the lattice quantum chromodynamics ( qcd ) @xcite , which has been solved by wilson fermion method @xcite . in ref .
@xcite , tanimura , hagino , and liang followed the same idea and realized the relativistic calculations on 3d lattice by introducing high - order wilson term .
however , the high - order wilson term modified the original dirac hamiltonian and the single particle energies and wave functions need be corrected .
although the corrections can be done with the perturbation theory , numerically it is much more involved .
another problem is that the high - order wilson term introduces artificial symmetry breaking to the system @xcite . in this paper
, we propose a new recipe of imaginary time method to solve dirac equation in 3d lattice space , where the variational collapse problem is avoided by ihm and the fermion doubling problem is avoided by performing the spatial derivatives of dirac equation in momentum space with the help of discrete fourier transform , so called spectral method @xcite .
this method is demonstrated by solving the dirac equation for a given spherical potential in 3d lattice and comparing with the results obtained by shooting method . by extending this method to solve the dirac equations for an axial deformed , non - axial deformed , octupole deformed potential , the corresponding single particle energy levels are obtained .
the corresponding quantum numbers of these energy levels are obtained respectively by projection .
the paper is organized as follows , the variational collapse and the fermion doubling problems will be briefly introduced in sec .
[ numerical_method ] together with the inversion hamiltonian method and the spectral method . in sec .
[ numericaldetails ] the parameters for woods - saxon type potentials and the numerical details are presented .
[ results&discussion ] is devoted to results and discussions .
summary and perspectives are given in sec .
[ summary ] .
itm is a iterative method for mean field problem .
the idea of itm is to replace the time with an imaginary number , and the evolution of the wave function reads @xcite , @xmath2 where @xmath3 is an initial wave function and @xmath4 is the hamiltonian . with the eigenstates @xmath5 of the hamiltonian @xmath4 corresponding to the eigenenergies @xmath6 , the evolution of the wave function @xmath7 can be written as , @xmath8 where @xmath9 . for @xmath10 , @xmath11 will approach the ground state wave function of @xmath4 as long as @xmath12 . in practice ,
the imaginary time @xmath13 is discrete with the interval @xmath14 , i.e. , @xmath15 .
the wave function at @xmath16 is obtained from the wave function at @xmath17 by expanding the exponential evolution operator @xmath18 to the linear order of @xmath14 , @xmath19 since this evolution is not unitary , the wave function should be normalized at every step . in order to find excited states
, one can start with a set of initial wave functions and orthonormalize them during the evolution by the gram - schmidt method .
this method has been successfully employed in the 3d coordinate - space calculations for nonrelativistic systems@xcite .
for the static dirac equation , @xmath20\psi(\bm{r})=\varepsilon\psi(\bm{r}),\ ] ] with @xmath21 and @xmath1 the dirac matrix , @xmath22 the vector potential , @xmath23 the scalar potential , and @xmath24 the dirac spinor , its eigenenergy spectrum extends from the continuum in the dirac sea to the continuum in the fermi sea . because of the existence of dirac sea , the evolution in eq . inevitably dives into the dirac sea ( negative energy states ) as @xmath10 , which is so - called variational collapse problem @xcite . in order to avoid the variational collapse , hagino and tanimura proposed the inverse hamiltonian method @xcite to find the wave function of the dirac hamiltonian @xmath4 by , @xmath25 where @xmath26 is an auxiliary parameter introduced to locate the interested eigenstate . with a given @xmath26 ,
the spectrum of @xmath4 can be labeled as @xmath27 where @xmath28 and @xmath29 are the eigenenergies of the dirac hamiltonian @xmath4 .
accordingly , the spectrum of @xmath30 reads , @xmath31 the evolution of the wave function in eq . will lead to the eigen wave function @xmath32 corresponding to the eigenvalue @xmath33 , @xmath34 as long as @xmath12 . in practice , the imaginary time evolution in eq .
is performed iteratively , @xmath35 the wave function also should be normalized at every step .
the inverse of hamiltonian in eq . , @xmath36 , can be solved by the conjugate residual method iteratively @xcite . in order to find excited states , with a set of initial wave functions
there are two options to choose @xmath26 .
one can take a fixed @xmath26 , then evolve the set of wave functions and orthonormalize them during the evolution by the gram - schmidt method .
alternatively , one can take set of @xmath37 for each eigenstate @xmath38 to evolve the whole set of wave functions
. the details can be found in sec .
[ numericaldetails ] , where an efficient method to choose @xmath37 is suggested to achieve a fast convergence . for a dirac equation on 3d lattice ,
there exist a so - called fermion doubling problem due to the replacement of the first derivatives in dirac equation by the finite difference method @xcite .
taking one dimensional dirac equation as an example , @xmath39 its solution has the following form @xmath40 if one approximates the derivative @xmath41 in eq . with 3-point differential formula with the mesh size @xmath42 , the dirac eq
. becomes , @xmath43\tilde{\psi}(k)=\varepsilon\tilde{\psi}(k).\ ] ] the dispersion relation obtained from eq
. reads , @xmath44 which differs from the exact one , @xmath45 for the dispersion relation eq .
obtained with 3-point differential formula , there are two momentum corresponding to one energy in the momentum interval @xmath46 $ ] .
the lower momentum corresponds to the physical solution , while the higher momentum corresponds to a spurious solution . as illustrated in ref .
@xcite , this problem persists even with more accurate finite differential formula .
similar spurious solution problem in radial dirac equation is also demonstrated in ref .
@xcite . to avoid fermion doubling problem , the derivative in eq .
can be performed in momentum space , @xmath47\tilde{\psi}(k)=\varepsilon\tilde{\psi}(k),\ ] ] which yields the exact dispersion relation , i.e. , fermion doubling problem is avoided naturally .
this is the so - called spectral method , i.e. , to perform spatial derivatives in momentum space . in the following ,
this method is illustrated in 1d case and it is straightforward to generalize this method to 3d case .
we assume that there are even @xmath48 discrete grid points @xmath49 in coordinate space distributing symmetric with the origin point , @xmath50 same number of grid points @xmath51 in momentum space , @xmath52 and the steps in coordinate space @xmath53 and in momentum space @xmath54 are related by , @xmath55 the function in coordinate space @xmath56 and the function in momentum space @xmath57 are connected by the discrete fourier transform , @xmath58 from eq .
, the @xmath59-th order derivative of @xmath56 can be found as , @xmath60 here @xmath61 corresponds to the fourier transform of the @xmath59-th order derivative of @xmath56 , @xmath62 in summary , the procedures to perform derivatives in coordinate space are as follows : ( 1 ) calculate @xmath57 from @xmath56 by the discrete fourier transform in eq . ;
( 2 ) calculate @xmath61 by eq . ;
( 3 ) calculate the @xmath59-th order derivative @xmath63 from @xmath61 by the inverse discrete fourier transform as in eq . .
the spectral method has the advantage to perform the spatial derivatives with a good accuracy .
the information of all grids are used in calculating the spatial derivative of any grid .
different from the finite differential method , all grids are treated on the same foot and the grids near the boundaries do nt need special numerical techniques .
in the following , we will solve dirac equation in 3d lattice in which the variational collapse problem is avoided by inverse hamiltonian method and the fermion doubling problem is avoided by performing spatial derivatives in momentum space with the help of discrete fourier transform , i.e. , spectral method .
the vector potential @xmath22 and the scalar potential @xmath23 in eq .
are woods - saxon type potential satisfying , @xmath64},\\ & v(\bm{r})-s(\bm{r})=\frac{-\lambda v_0}{1+\exp[(r - r_{ls}f(\omega))/a_{ls } ] } , \end{split}\ ] ] where @xmath65 is a function of @xmath66 with potential deformation parameters @xmath67 , @xmath68 and @xmath69 , @xmath70+\beta_{30}y_{30}(\omega).\ ] ] the deformation parameter @xmath67 and @xmath68 in eq .
are related to hill - wheeler coordinates @xmath1 and @xmath71 @xcite by @xmath72 the adopted woods - saxon potential parameters in eq .
are listed in table [ tab_parameter ] , which correspond to the neutron potential in @xmath73ca @xcite .
.[tab_parameter]the parameters in the woods - saxon type potential eq . adopted in the present 3d lattice calculations . [ cols="^,^,^,^,^,^",options="header " , ] in the calculations , the box sizes @xmath74 fm and step sizes @xmath75 fm are respectively chosen along @xmath76 , @xmath77 and @xmath78 axes if not otherwise specified .
the imaginary time step size @xmath79 is taken 100 mev . for the @xmath38-th level ,
the upper component of the initial wave function is generated from a nonrelativistic harmonic oscillator state and the corresponding lower component is taken the same as the upper one .
the energy shift @xmath37 is taken as @xmath80 where @xmath81 is the expectation value of the dirac hamiltonian for the @xmath38-th level .
the choice of @xmath82 is as follows , @xmath83 mev and for @xmath84 , @xmath85 the convergence in the evolution of the wave functions for our interested states is determined by @xmath86 smaller than the required accuracy @xmath87 mev if not otherwise specified . in order to speed up the convergence ,
the dirac hamiltonian is diagonalized within the space of the evolution wave functions every 10 iterations , and the eigenfunctions thus obtained are taken as initial wave functions for future iteration .
the similar technique is also used in ref . @xcite .
in this section , the dirac equation with a given potential is solved in 3d lattic space by the new method ( denoted as _
3d lattice _ ) .
firstly we examine the convergence feature of the present 3d lattice calculation for a spherical potential in eq . .
the results will be compared with those obtained by the shooting method ( denoted as _ shooting _ ) @xcite with a box size @xmath88 fm and a step size @xmath89 fm . with the potential parameters in table [ tab_parameter ] , the evolution of the single particle energies as a function of iteration times is shown in fig .
[ fig_iteration ] .
there are in total 40 bound single particle states in the 3d lattice calculation and some of them are degenerate in energy due to the spherical symmetry . for clarity , only one energy level of the degenerate ones
is shown to illustrate the evolution of single particle energies .
the single particle energies obtained by shooting method are also shown for comparison .
it can be seen that the more deeply levels converge more quickly .
after the 39th iteration , the accuracy of energy for all bound levels are smaller than @xmath0 mev .
a distinct feature is observed at the 10th iteration where the convergence of 1p@xmath90 , 1d@xmath91 , 2s@xmath90 states is speeded up due to the diagonalization of the hamiltonian within the space of the evolution wave functions .
in fact , it will cost tens of thousands of iteration steps to reach the convergence tolerance without this diagonalization procedure .
mev . as a comparison
, the results obtained by the shooting method are also given in figure as well .
, title="fig:",scaledwidth=45.0% ] + in fig .
[ fig_energydifference ] , the absolute deviations of single particle energies between 3d lattice calculation and shooting method are given as a function of single particle energy for different step sizes @xmath42 and box sizes @xmath92 . in fig .
[ fig_energydifference ] ( a ) , for @xmath93 fm and @xmath94 fm , the absolute deviations of single particle energies are smaller than @xmath95 mev , except the weakly bound states 1f@xmath96 , 2p@xmath91 , and 2p@xmath90 . in fig .
[ fig_energydifference ] ( b ) , for @xmath97 fm and @xmath98 fm , the absolute deviations of single particle states are less than @xmath0 mev , except 2p@xmath91 and 2p@xmath90 . in fig .
[ fig_energydifference ] ( c ) , for @xmath97 fm and @xmath99 fm , all absolute deviations including 2p@xmath91 and 2p@xmath90 are smaller than @xmath0mev .
these results indicate that smaller step size can definitely improve the accuracy but not for the weakly bound states with low orbital angular momentum . by choosing suitable step size and box size , accurate descriptions for all the bound states including the weakly bound states 2p@xmath91 and 2p@xmath90 can be achieved in the 3d lattice calculations .
fm and box sizes @xmath94 fm , ( b ) @xmath97 fm and @xmath98 fm , and ( c ) @xmath97 fm and @xmath99 fm .
the spherical quantum numbers are listed in ( b ) .
, title="fig:",scaledwidth=45.0% ] + it is interesting to investigate the spatial distributions of states and examine their agreements with the results obtained by the shooting method . in fig .
[ fig_densityplane ] , as examples , the distributions of the states corresponding to 1d@xmath96 in @xmath100 plane are illustrated .
the states corresponding to 1d@xmath96 are six degenerate single particle states in the 3d lattice calculations .
their spatial distributions are respectively shown in fig .
[ fig_densityplane ] ( a)-(f ) , and fig .
[ fig_densityplane ] ( g ) exhibits their average in @xmath100 plane . as there is no symmetry restriction in the 3d lattice calculations , the six states randomly oriented in space
. however , their average spatial distribution does show the spherical symmetry as shown in fig .
[ fig_densityplane ] ( g ) , which is consistent with the given spherical potential . in @xmath100 plane in the 3d lattice calculation .
figures ( a)-(f ) are the density distributions of the states in 1d@xmath96 , and figure ( g ) is their average spatial distributions.,title="fig:",scaledwidth=45.0% ] + in order to compare with the radial density distribution obtained by the shooting method , one can average the density distributions in the 3d lattice calculation , @xmath101 in fig.[fig_radialdensity ] , the radial density distributions for 1s@xmath102 , 1d@xmath96 , and 2s@xmath90 in the 3d lattice calculation ( open circles ) in comparison with the shooting method ( solid line ) are given , in which a factor @xmath103 has been multiplied in order to amplify the radial density distribution at large distance .
it can be clearly seen that the two distributions are in perfect agreement with each other .
the data points in the 3d lattice calculation are denser for large @xmath104 because the grid points used are uniform in the 3d lattice space . , 1d@xmath96 , and 2s@xmath90 in the 3d lattice calculation ( open circles ) in comparison with the shooting method ( solid line ) .
the radial density distribution in the 3d lattice calculation is extracted by eq .. , title="fig:",scaledwidth=45.0% ] + for the dirac equations with the deformed potentials in eq .
, the single particle energies as functions of deformation parameters @xmath1 , @xmath71 , and @xmath69 are given in fig .
[ fig_deformedlevel ] , which respectively correspond to axial , non - axial and reflection - asymmetric deformed potentials . in fig .
[ fig_deformedlevel](a ) , the potentials have both the space reflection symmetry and axial symmetry with @xmath105 , @xmath106 , and @xmath1 from @xmath107 to @xmath108 . in fig .
[ fig_deformedlevel](b ) , the potentials break the axial symmetry while keep the space reflection symmetry with @xmath109 , @xmath106 , and @xmath71 from @xmath110 to @xmath111 . in fig .
[ fig_deformedlevel](c ) , the potentials break both the space reflection symmetry and axial symmetry with @xmath109 , @xmath112 , and @xmath69 from @xmath107 to @xmath113 .
although there is no symmetry restriction in the 3d lattice calculations , we can search for good quantum numbers from the expectations of physical operators .
for spherical case , total angular momentum @xmath114 and orbital angular momentum @xmath115 can be calculated by the expectation of @xmath116 and @xmath117 with the upper components of the wave functions . for axial case
, the @xmath78-component of the total angular momentum @xmath118 can be calculated by the expectation of @xmath119 . for the space reflection symmetry case
, the parity can be calculated by the expectation of parity operator @xmath120 , where @xmath1 is the dirac matrix and @xmath121 . from fig .
[ fig_deformedlevel ] , it can be seen that the levels in spherical case are split into @xmath122 levels with the potential changing from spherical to deformed .
however , the kramers degeneracy remains as there is no time odd potential .
for axial case , the levels with lower ( higher ) @xmath118 values shift downwards ( upwards ) in consistent with the nilsson model .
comparing fig . [
fig_deformedlevel](a ) and fig .
[ fig_deformedlevel](b ) , it can be seen that the spectrum changes more modestly with @xmath71 than with @xmath1 . in fig .
[ fig_deformedlevel ] ( c ) , all levels trend to shift downwards with @xmath69 , which shows its instability against in fission . , @xmath71 , and @xmath69 .
the red and blue lines in left panel and mid panel represent the levels with positive and negative parity respectively .
the shapes showed in upper panel correspond to the deformed parameters @xmath123 , @xmath124 , @xmath125 , @xmath126 , respectively .
, title="fig:",scaledwidth=45.0% ] + in order to examine the compositions and their evolution of the single particle level with deformation parameters @xmath1 , @xmath71 , and @xmath69 , level a , level b and level c in fig .
[ fig_deformedlevel ] are chosen as examples .
the results are illustrated in fig .
[ fig_components ] . in the left panels ,
the compositions of each level are obtained by overlapping the wave functions with the wave functions obtained with @xmath123 . in the middle panels ,
the compositions of each level are obtained by overlapping the wave functions with the wave functions obtained with @xmath127 . in the right panels ,
the parity compositions of each level are obtained by the expectation of parity operator . in the left panels
, there is only small mixing with other orbits for level a in comparison with level b and c. it s can be understood as follows .
this is due to the special character of level a with @xmath128 and parity @xmath129 .
the possible mixing is from the 1h@xmath130 orbit which lies high in energy .
similar conclusions can be drawn for the level @xmath131 originating from 1p@xmath91 and @xmath132 originating from 1d@xmath96 . in the middle panels , for level a , there is a dramatic change for @xmath128 and @xmath133 components when @xmath71 approaching @xmath111 .
this is due to the interaction between level a and the level originating from @xmath134 and @xmath118=1/2 at @xmath135 , as shown in energy levels in fig .
[ fig_deformedlevel](b ) . in the right panels , for octupole deformed case , the parity composition of level b and level c changes rigorously due to complicated interaction between levels . for level a
, the main composition is negative parity as it mainly interacts with negative parity dominated levels .
all these can be understood from fig .
[ fig_deformedlevel ] ( c ) . as functions of the deformation parameters @xmath1 , @xmath71 , and @xmath69 .
the quantum numbers are given for each compositions .
the total probabilities are shown as black dash lines.,title="fig:",scaledwidth=45.0% ] +
in summary , a new method to solve dirac equation in 3d lattice space is proposed with inverse hamiltonian method to avoid variational collapse and the spectral method to avoid the fermion doubling problem .
this method is demonstrated in solving the dirac equation for a given spherical potential in 3d lattice . in comparison with the results obtained by the shooting method ,
the differences in single particle energies are smaller than @xmath0 mev , and the densities are almost identical , which demonstrate the high accuracy of the present method . applying this method to dirac equations with an axial deformed , non - axial deformed , octupole deformed potential without further modification , the single particle levels as functions of the deformation parameters @xmath1 , @xmath71 , and @xmath69 are shown together with their compositions .
efforts in implanting this method on cdft to investigate nuclei without any geometric restriction are in progress .
possible applications include solving the dirac equation in an external electric potential ( deformation constrained calculation ) to investigate nuclei with an arbitrary shape , and in an external magnetic potential ( coriolis term ) to investigate rotating nuclei with arbitrary shape and an arbitrary rotating axis .
moreover , the 3d time - dependent cdft is also envisioned to be developed to investigate the relativistic effects in heavy ions collision and other nuclear reactions .
we thank p. ring for helpful discussions .
this work was supported in part by the major state 973 program of china ( grant no .
2013cb834400 ) , the national natural science foundation of china ( grants no . 11335002 , no .
11375015 , no . 11461141002 , no .
11621131001 ) .
41ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo
secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/physrevlett.55.2676 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.3963 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.80.460 [ * * , ( ) ] link:\doibase 10.1103/physrevc.82.011301 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.84.5493 [ * * , ( ) ] link:\doibase 10.1103/physrevc.53.2809 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.75.121 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevc.78.064306 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1016/j.ppnp.2005.06.001 [ * * , ( ) ] @noop _ _
( , ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.78.436 [ * * , ( ) ] link:\doibase 10.1103/physrevc.58.r628 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.262501 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.physrep.2014.12.005 [ * * , ( ) ] link:\doibase 10.1103/physrevc.36.354 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevc.85.024312 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) \doibase http://dx.doi.org/10.1016/j.cpc.2014.04.008 [ * * , ( ) ] http://stacks.iop.org/1674-1137/33/i=s1/a=036 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.40.5548 [ * * , ( ) ] @noop _ _ , edited by ( , ) pp
. link:\doibase 10.1103/revmodphys.55.775 [ * * , ( ) ] @noop _ _
( , ) @noop _ _ ( , ) @noop * * ( ) link:\doibase 10.1103/physrev.89.1102 [ * * , ( ) ] @noop _ _ ( , ) http://dx.doi.org/10.1007/bf01282936 [ * * , ( ) ]
|
a new method to solve dirac equation in 3d lattice is proposed , in which the variational collapse problem is avoided by inverse hamiltonian method and the fermion doubling problem is avoided by performing spatial derivatives in momentum space with the help of discrete fourier transform , i.e. , spectral method .
this method is demonstrated in solving the dirac equation for a given spherical potential in 3d lattice . in comparison with the results obtained by the shooting method , the differences in single particle energy are smaller than @xmath0 mev , and the densities are almost identical , which demonstrate the high accuracy of the present method . the results obtained by applying this method without any modification to solve the dirac equations for an axial deformed , non - axial deformed , octupole deformed potential are provided and discussed .
|
the agn / qso luminosity function and its evolution with cosmic time are key observational quantities for understanding the origin of and accretion history onto supermassive black holes , which are now believed to occupy the centers of most galaxies .
since x - ray emission is one of the prominent characters of the agn activity , x - ray surveys are efficient means of sampling agns for luminosity function and evolution studies . an x - ray selected sample of agns is particularly useful because optical surveys often use point - like morphology as a criterion for selecting agns ( qsos ) among numerous other objects , and thus are likely to miss moderate - luminosity intermediate - high redshift agns embedded in their host galaxies .
also , radio surveys sample only a minor population of agns .
the _ rntgen satellite _ ( _ rosat _ ) provided us with soft x - ray surveys with various depths , ranging from the _ rosat _ all - sky survey ( rass , voges et al.@xcite ) to the _ rosat _ deep survey ( rds ) on the lockman hole ( hasinger et al .
various optical identification programs of the survey fields have been conducted and the combination of these now enabled us to construct the soft x - ray luminosity function ( sxlf ) as a function of redshift . in paper i. , we presented a number of global representations of the 0.5 - 2 kev sxlf and investigated the contribution to the soft x - ray background .
we showed that our data are not consistent with the pure - luminosity evolution ( ple ) , contrary to the suggestions of a number of previous analyses ( e.g. boyle et al .
@xcite ; jones et al .
@xcite ) . instead
, we find an excess of intermediate - redshift low - luminosity agns above the ple case , some sign of which was also recognized by page et al .
( @xcite ) . in view of this , we developed two versions of luminosity - dependent density evolution ( ldde1 and ldde2 ) models , which represent the observed data very well .
an extrapolation of these two ldde models below the faintest limit of the survey ( @xmath1 @xmath2 $ ] ) yields significantly different predictions for fainter fluxes , bracketing the range of a possible agn contribution to the soft x - ray background .
chandra ( e.g. mushotzky et al . @xcite ; hornschemeier et al .
@xcite ) and xmm - newton ( @xcite ) are probing much fainter sources and spectroscopic identifications of these will eventually show which of these models is closer to the actual behavior of the agn sxlf .
however , because at least some of the faint x - ray sources are optically too faint for spectroscopic identification even with the largest telescopes , extending the xlf into such faint flux level may be difficult .
in this second paper , we present practical and convenient expressions of the observed sxlf from the _ rosat _ surveys .
we present our results mainly for the investigators who are interested in particular redshift regimes and/or comparing their models with observations . for this purpose ,
we show convenient analytical expressions in several redshift intervals separately .
these give more accurate representations of the data in the redshift ranges of interest than those presented in paper i. we also tabulate the numerical values of the binned sxlf using an improved estimator .
we use a hubble constant @xmath3 @xmath4 $ ] .
the @xmath5 dependences are explicitly shown .
we calculate the results with common sets of cosmological parameters : @xmath6(1.0,0.0 ) , ( 0.3,0.0 ) . and
@xmath6 ( 0.3,0.7 ) .
the symbol `` @xmath7 '' represents the base-10 logarithm .
we have used soft x - ray sources identified with agns with redshift information from a combination of _ rosat _ surveys in various depths / areas from a number of already published and unpublished sources .
detailed description of the definition of the sample , _ rosat _ countrate - to - flux conversion , and survey area are shown in paper i. the summary of the samples , which is a duplicate of table 1 .
of paper i with updated references , is shown in table [ tab : surv ] .
the details of the nature and completeness of each sample were discussed in paper i. the limiting flux versus survey area relation were also shown in paper i. ._rosat _ surveys used in the analysis [ cols="<,^,^,^ " , ] [ tab : surv ] @xmath8 abbreviations rbs : the _ rosat _ bright survey ( fischer et al .
@xcite ; schwope et al .
sa - n : the selected area - north ( zickgraf et al .
@xcite ; appenzeller et al .
@xcite ) , rixos : the _ rosat _ international x - ray optical survey ( mason et al .
@xcite ) , nep : the north ecliptic pole survey ( bower et al .
@xcite ) ; ukd : the uk deep survey ( mchardy et al .
@xcite ) , rds - marano : the _ rosat _ deep survey marano field ( zamorani et al .
@xcite , rds - lh : the _ rosat _ deep survey lockman hole ( hasinger et al .
@xcite ; schmidt et al .
@xcite ; lehmann et al .
@xmath9 excluding agns with @xmath10 .
as in paper i , we present the sxlf in the _ observed _ 0.5 - 2 kev band , i.e. , in the @xmath11 kev range in the object s rest frame .
this is equivalent to assuming an energy index of 1 .
thus no k - correction was applied for our expressions presented here .
the reasons for this choice are explained in detail in paper i. this choice is particularly important for this paper , which is intended to be used as observational constraints for population - synthesis - type models ( e.g. madau et al.@xcite ; comastri et al .
@xcite ; gilli et al .
( @xcite,@xcite ) ; miyaji et al .
@xcite ) with various spectral assumptions . because of that
, it is more useful to provide quantities in a model - independent form rather than applying a particular version of model - dependent k - corrections . by presenting in this manner
, one can avoid the difficulty of reverse k - correcting and re - applying new k - corrections when the new results from _ chandra _ and _ xmm _ provide better knowledge of the x - ray spectra of the population .
ccrcccccc @xmath12-range & @xmath13 & n & a@xmath8 & @xmath14 & @xmath15 & @xmath16 & @xmath17 & ks - prob@xmath9 + + 0.015 - 0.2 & 0.1 & 269 & @xmath18 & 0.60@xmath19 & 0.59@xmath20 & 2.1@xmath21 & 4.22@xmath22 & 0.99,0.68,0.61 + 0.2 - 0.4 & 0.3 & 113 & @xmath23 & 0.89@xmath24 & 0.67@xmath25 & 2.5@xmath21 & 5.25@xmath26 & 0.93,0.56,0.72 + 0.4 - 0.8 & 0.6 & 99 & @xmath27 & 0.54@xmath28 & 0.33@xmath29 & 2.2@xmath30 & 5.95@xmath31 & 0.85,0.85,0.57 + 0.8 - 1.6 & 1.2 & 135 & @xmath32 & 1.48@xmath33 & 0.40@xmath34 & 2.4@xmath35 & 4.07@xmath36 & 0.99,0.96,0.83 + 1.6 - 2.3 & 2.2 & 44 & @xmath37 & 1.2 ( * ) & 0.0 ( * ) & 2.1@xmath38 & 0 ( * ) & 0.27,0.36,0.19 + 2.3 - 4.6 & 3.0 & 25 & @xmath39 & 1.0 ( * ) & & 1.9@xmath35 & 0 ( * ) & 0.72,0.99,0.64 + + 0.015 - 0.2 & 0.1 & 269 & @xmath40 & 0.59@xmath41 & 0.59@xmath42 & 2.1@xmath21 & 4.13@xmath43 & 0.99,0.60,0.54 + 0.2 - 0.4 & 0.3 & 113 & @xmath44 & 0.93@xmath45 & 0.67@xmath46 & 2.4@xmath21 & 5.31@xmath47 & 0.99,0.53,0.79 + 0.4 - 0.8 & 0.6 & 99 & @xmath48 & 0.69@xmath49 & 0.37@xmath50 & 2.3@xmath30 & 5.90@xmath51 & 0.97,0.81,0.57 + 0.8 - 1.6 & 1.2 & 135 & @xmath52 & 2.14@xmath53 & 0.42@xmath54 & 2.4@xmath35 & 4.13@xmath55 & 0.98,0.97,0.72 + 1.6 - 2.3 & 2.2 & 44 & @xmath56 & 1.8 ( * ) & 0.0 ( * ) & 2.1@xmath38 & 0 ( * ) & 0.19,0.41,0.15 + 2.3 - 4.6 & 3.0 & 25 & @xmath57 & 1.0 ( * ) & & 1.9@xmath35 & 0 ( * ) & 0.64,0.96,0.75 + + 0.015 - 0.2 & 0.1 & 269 & @xmath58 & 0.71@xmath59 & 0.62@xmath60 & 2.1@xmath21 & 3.79@xmath61 & 0.99,0.54,0.61 + 0.2 - 0.4 & 0.3 & 113 & @xmath62 & 1.09@xmath63 & 0.67@xmath64 & 2.4@xmath21 & 4.95@xmath47 & 0.97,0.56,0.82 + 0.4 - 0.8 & 0.6 & 99 & @xmath65 & 0.85@xmath66 & 0.36@xmath67 & 2.2@xmath30 & 5.69@xmath68 & 0.96,0.81,0.60 + 0.8 - 1.6 & 1.2 & 135 & @xmath69 & 2.69@xmath70 & 0.43@xmath71 & 2.4@xmath35 & 4.10@xmath36 & 0.98,0.97,0.74 + 1.6 - 2.3 & 2.2 & 44 & @xmath72 & 2.0 ( * ) & 0.0 ( * ) & 2.1@xmath73 & 0 ( * ) & 0.14,0.40,0.16 + 2.3 - 4.6 & 3.0 & 25 & @xmath74 & 1.0 ( * ) & & 1.9@xmath35 & 0 ( * ) & 0.78,0.98,0.76 + [ tab : fit ] parameter values which have been fixed during the fit are labeled by (*).
@xmath8units a : [ @xmath75 , @xmath76 : [ @xmath77 . @xmath9 the three values are probabilities in two 1d - ks test for the distribution , @xmath78 , 1d - ks test for the @xmath12 distribution and the 2d - ks test for the ( @xmath78,@xmath12 ) space respectively .
first , we find a smooth analytical function for each redshift bin using a maximum - likelihood fitting .
the absolute goodness of the resulting expression can then be tested by one- and two - dimensional kolgomorov - smirnov tests ( hereafter , 1d - ks and 2d - ks tests respectively ; press et al.@xcite ; fasano & franceschini @xcite ) .
see paper i for detailed description of these methods .
these fittings and tests can be applied to unbinned data sets thus are free from artifacts and biases from binning . for an analytical expression ,
we use the _ smoothed two- power - law _ formula , as we did in paper i. here , we fit the data in narrow redshift bins and thus evolution in each redshift bin is assumed to be a pure density evolution form : @xmath79^{-1 } \cdot \left(\frac{1+z}{1+z_{\rm c } } \right)^p , \label{eq : ana}\ ] ] where @xmath13 is the central redshift of the bin . for the highest redshift bin where the `` break '' is not apparent , we have used a single power - law form by neglecting the first term in the square bracket in eq.[eq : ana ] .
the luminosity range of the fit is from log @xmath80=41.7 to the maximum available luminosity in the sample .
as shown below and in paper i , the sxlf below the minimum luminosity has a significant excess above the smooth extrapolation .
this excess smoothly connects with the sxlf of the non - agn population ( e.g. hasinger et al .
@xcite ) and the x - ray emission may be significantly contaminated by non - agn activities .
the set of parameters which give the best fit for each redshift bin are shown in table [ tab : fit ] along with the results of the 1d- and 2d- ks tests ( see the notes of the table ) .
the parameter errors correspond to a likelihood change of 2.7 ( 90% confidence errors ) . in any case , eq . [ eq : ana ] gives a statistically satisfactory expression for all redshift bins .
the @xmath0 estimator , which is a generalized version of the original @xmath81 estimator ( schmidt @xcite ) applied to a sample composed of subsamples of different depths ( see paper i ; avni & bahcall @xcite ) , has been widely used for binned luminosity functions ( lf ; we use the acronym lf when the discussion is not limited to the luminosity function in the x - ray band ) in the literature .
as discussed in paper i ( see also wisotzki @xcite ; page & carrera @xcite ) , using it for a binned lf estimator can cause significant biases , especially if the bin covers the flux range where the available solid angle of the survey changes rapidly as a function of flux .
also , the choice of the location in a @xmath82 bin with a non - negligible width at which the data point is plotted significantly changes the impression of the plot . in fig .
3 of paper i , however , we plotted the @xmath0 estimates , because of the lack of a reasonable alternative at the time of writing that paper , with caveats on biases associated with the method .
we note that the estimator can be used in an unbinned manner by considering a set of delta - functions weighted by @xmath83 ( or @xmath84 ) at the positions of sample objects in the luminosity space ( schmidt & green @xcite ) , and this method is free from biases mentioned above . while this unbinned method is a powerful tool to predict , e.g. , the source counts , it does not provide practical means of plotting . in this paper
, we have developed an improved estimator , which is explained in the next subsection .
ccrrc + @xmath12 & log@xmath78@xmath8 & @xmath85 & @xmath86 & @xmath87 + ( 1 ) & ( 2 ) & & & + .015 - 0.2 & 41.30 - 41.70 & 5(3 ) & 1.0 & @xmath88 + .015 - 0.2 & 41.70 - 42.40 & 8(2 ) & 6.3 & @xmath89 + .015 - 0.2 & 42.40 - 43.00 & 23(3 ) & 26.8 & @xmath90 + .015 - 0.2 & 43.00 - 43.50 & 61(7 ) & 58.4 & @xmath91 + .015 - 0.2 & 43.50 - 43.80 & 56(3 ) & 52.6 & @xmath92 + .015 - 0.2 & 43.80 - 44.20 & 55(4 ) & 65.5 & @xmath93 + .015 - 0.2 & 44.20 - 44.50 & 35(1 ) & 31.6 & @xmath94 + .015 - 0.2 & 44.50 - 45.00 & 30 & 25.0 & @xmath95 + .015 - 0.2 & 45.00 - 45.70 & 1 & 3.0 & @xmath96 + + 0.2 - 0.4 & 41.70 - 42.10 & 1 & 0.8 & @xmath97 + 0.2 - 0.4 & 42.10 - 42.70 & 3(2 ) & 4.3 & @xmath98 + 0.2 - 0.4 & 42.70 - 43.30 & 18(5 ) & 15.1 & @xmath99 + 0.2 - 0.4 & 43.30 - 43.80 & 27(3 ) & 26.3 & @xmath100 + 0.2 - 0.4 & 43.80 - 44.30 & 25(3 ) & 29.0 & @xmath101 + 0.2 - 0.4 & 44.30 - 44.90 & 26(1 ) & 23.5 & @xmath102 + 0.2 - 0.4 & 44.90 - 45.40 & 13 & 11.3 & @xmath103 + + 0.4 - 0.8 & 42.30 - 42.70 & 0 & 1.1 & @xmath104 + 0.4 - 0.8 & 42.70 - 43.30 & 13(7 ) & 10.8 & @xmath105 + 0.4 - 0.8 & 43.30 - 43.60 & 14(4 ) & 11.7 & @xmath106 + 0.4 - 0.8 & 43.60 - 44.20 & 38 & 43.2 & @xmath107 + 0.4 - 0.8 & 44.20 - 44.80 & 19(1 ) & 16.8 & @xmath108 + 0.4 - 0.8 & 44.80 - 45.40 & 10 & 8.9 & @xmath109 + 0.4 - 0.8 & 45.40 - 46.50 & 5 & 6.5 & @xmath110 + + 0.8 - 1.6 & 42.70 - 43.30 & 5 & 2.9 & @xmath111 + 0.8 - 1.6 & 43.30 - 43.90 & 23 ( 2 ) & 26.9 & @xmath112 + 0.8 - 1.6 & 43.90 - 44.50 & 55 & 53.0 & @xmath113 + 0.8 - 1.6 & 44.50 - 45.10 & 39 & 36.0 & @xmath114 + 0.8 - 1.6 & 45.10 - 45.70 & 5 & 9.0 & @xmath115 + 0.8 - 1.6 & 45.70 - 46.20 & 4 & 3.1 & @xmath116 + 0.8 - 1.6 & 46.20 - 46.90 & 4 & 2.3 & @xmath117 + + 1.6 - 2.3 & 43.60 - 44.20 & 9 & 13.1 & @xmath118 + 1.6 - 2.3 & 44.20 - 44.80 & 18 & 12.9 & @xmath119 + 1.6 - 2.3 & 44.80 - 45.50 & 14 & 10.5 & @xmath120 + 1.6 - 2.3 & 45.50 - 46.10 & 2 & 2.3 & @xmath121 + 1.6 - 2.3 & 46.10 - 46.80 & 0 & 1.5 & @xmath122 + 1.6 - 2.3 & 46.80 - 47.40 & 1 & 1.0 & @xmath123 + + 2.3 - 4.6 & 43.70 - 44.10 & 2 & 3.4 & @xmath124 + 2.3 - 4.6 & 44.10 - 44.80 & 12 & 10.5 & @xmath125 + 2.3 - 4.6 & 44.80 - 45.40 & 7 & 4.3 & @xmath126 + 2.3 - 4.6 & 45.40 - 46.20 & 2 & 3.2 & @xmath127 + 2.3 - 4.6 & 46.20 - 47.00 & 0 & 1.4 & @xmath128 + 2.3 - 4.6 & 47.00 - 47.50 & 2 & 1.0 & @xmath129 + + [ tab : xlf_f10 ] notes : @xmath130 $ ] in 0.5 - 2 [ kev ] .
@xmath131 $ ] .
@xmath132 the agns in this row are outside of the luminosity range used for the model fit .
ccrrc + @xmath12 & log@xmath78@xmath8 & @xmath85 & @xmath86 & @xmath87 + ( 1 ) & ( 2 ) & & & + .015 - 0.2 & 41.30 - 41.70 & 5(3 ) & 1.0 & @xmath133 + .015 - 0.2 & 41.70 - 42.40 & 7(2 ) & 6.1 & @xmath134 + .015 - 0.2 & 42.40 - 43.00 & 24(3 ) & 26.0 & @xmath135 + .015 - 0.2 & 43.00 - 43.50 & 58(6 ) & 56.4 & @xmath136 + .015 - 0.2 & 43.50 - 43.80 & 55(4 ) & 50.5 & @xmath92 + .015 - 0.2 & 43.80 - 44.20 & 58(4 ) & 64.4 & @xmath137 + .015 - 0.2 & 44.20 - 44.60 & 47(1 ) & 39.8 & @xmath138 + .015 - 0.2 & 44.60 - 45.20 & 19 & 20.6 & @xmath139 + .015 - 0.2 & 45.20 - 45.80 & 1 & 1.3 & @xmath140 + + 0.2 - 0.4 & 41.70 - 42.10 & 1 & 0.6 & @xmath141 + 0.2 - 0.4 & 42.10 - 42.70 & 2(1 ) & 4.2 & @xmath142 + 0.2 - 0.4 & 42.70 - 43.40 & 21(6 ) & 19.4 & @xmath143 + 0.2 - 0.4 & 43.40 - 43.80 & 23(3 ) & 22.0 & @xmath144 + 0.2 - 0.4 & 43.80 - 44.30 & 26(3 ) & 28.7 & @xmath101 + 0.2 - 0.4 & 44.30 - 44.80 & 23(1 ) & 21.3 & @xmath145 + 0.2 - 0.4 & 44.80 - 45.40 & 17 & 14.6 & @xmath146 + + 0.4 - 0.8 & 42.20 - 42.80 & 0 & 1.4 & @xmath147 + 0.4 - 0.8 & 42.80 - 43.25 & 8(5 ) & 7.1 & @xmath148 + 0.4 - 0.8 & 43.25 - 43.60 & 13(6 ) & 10.7 & @xmath149 + 0.4 - 0.8 & 43.60 - 44.10 & 34 & 33.4 & @xmath150 + 0.4 - 0.8 & 44.10 - 44.80 & 25 & 28.4 & @xmath151 + 0.4 - 0.8 & 44.80 - 45.40 & 13(1 ) & 10.0 & @xmath152 + 0.4 - 0.8 & 45.40 - 46.60 & 6 & 7.4 & @xmath153 + + 0.8 - 1.6 & 43.10 - 43.60 & 6 & 6.3 & @xmath154 + 0.8 - 1.6 & 43.60 - 44.10 & 28(2 ) & 26.1 & @xmath155 + 0.8 - 1.6 & 44.10 - 44.60 & 46 & 45.0 & @xmath156 + 0.8 - 1.6 & 44.60 - 45.30 & 42 & 40.6 & @xmath157 + 0.8 - 1.6 & 45.30 - 45.90 & 5 & 9.0 & @xmath158 + 0.8 - 1.6 & 45.90 - 46.50 & 4 & 3.9 & @xmath159 + 0.8 - 1.6 & 46.50 - 47.00 & 4 & 1.6 & @xmath160 + + 1.6 - 2.3 & 43.80 - 44.50 & 13 & 16.7 & @xmath161 + 1.6 - 2.3 & 44.50 - 45.10 & 14 & 12.5 & @xmath162 + 1.6 - 2.3 & 45.10 - 45.70 & 14 & 8.8 & @xmath163 + 1.6 - 2.3 & 45.70 - 46.40 & 2 & 2.9 & @xmath164 + 1.6 - 2.3 & 46.40 - 47.00 & 0 & 1.3 & @xmath165 + 1.6 - 2.3 & 47.00 - 47.60 & 1 & 1.1 & @xmath166 + + 2.3 - 4.6 & 44.10 - 44.80 & 9 & 9.9 & @xmath167 + 2.3 - 4.6 & 44.80 - 45.40 & 6 & 6.1 & @xmath168 + 2.3 - 4.6 & 45.40 - 46.00 & 7 & 4.6 & @xmath169 + 2.3 - 4.6 & 46.00 - 46.60 & 1 & 1.9 & @xmath170 + 2.3 - 4.6 & 46.60 - 47.10 & 0 & 1.2 & @xmath171 + 2.3 - 4.6 & 47.10 - 47.60 & 2 & 1.0 & @xmath172 + + [ tab : xlf_o3 ] notes : @xmath130 $ ] in 0.5 - 2 [ kev ] .
@xmath131 $ ] .
@xmath132 the agns in this row are outside of the luminosity range used for the model fit .
ccrrc + @xmath12 & log@xmath78@xmath8 & @xmath85 & @xmath86 & @xmath87 + ( 1 ) & ( 2 ) & & & + .015 - 0.2 & 41.30 - 41.70 & 5(3 ) & 1.0 & @xmath133 + .015 - 0.2 & 41.70 - 42.60 & 12(3 ) & 10.6 & @xmath173 + .015 - 0.2 & 42.60 - 43.00 & 19(2 ) & 20.5 & @xmath174 + .015 - 0.2 & 43.00 - 43.50 & 58(6 ) & 53.5 & @xmath175 + .015 - 0.2 & 43.50 - 43.80 & 49(3 ) & 49.1 & @xmath176 + .015 - 0.2 & 43.80 - 44.20 & 62(5 ) & 67.5 & @xmath137 + .015 - 0.2 & 44.20 - 44.60 & 41(1 ) & 41.8 & @xmath177 + .015 - 0.2 & 44.60 - 45.10 & 27 & 21.9 &
@xmath178 + .015 - 0.2 & 45.10 - 45.80 & 1 & 2.5 & @xmath179 + + 0.2 - 0.4 & 41.70 - 42.20 & 1 & 0.8 & @xmath180 + 0.2 - 0.4 & 42.20 - 42.80 & 2(1 ) & 4.5 & @xmath181 + 0.2 - 0.4 & 42.80 - 43.40 & 19(6 ) & 14.9 & @xmath182 + 0.2 - 0.4 & 43.40 - 43.80 & 21(2 ) & 21.3 & @xmath183 + 0.2 - 0.4 & 43.80 - 44.30 & 26(3 ) & 28.2 & @xmath184 + 0.2 - 0.4 & 44.30 - 44.70 & 17(2 ) & 19.8 & @xmath185 + 0.2 - 0.4 & 44.70 - 45.30 & 23 & 16.6 & @xmath186 + 0.2 - 0.4 & 45.30 - 46.00 & 4 & 5.7 & @xmath187 + + 0.4 - 0.8 & 42.20 - 42.80 & 0 & 0.8 & @xmath188 + 0.4 - 0.8 & 42.80 - 43.25 & 5(3 ) & 5.5 & @xmath189 + 0.4 - 0.8 & 43.25 - 43.60 & 13(7 ) & 9.8 & @xmath190 + 0.4 - 0.8 & 43.60 - 44.10 & 29(1 ) & 30.3 & @xmath191 + 0.4 - 0.8 & 44.10 - 44.80 & 31 & 34.0 & @xmath192 + 0.4 - 0.8 & 44.80 - 45.40 & 15(1 ) & 11.1 & @xmath193 + 0.4 - 0.8 & 45.40 - 46.00 & 5 & 5.8 & @xmath194 + 0.4 - 0.8 & 46.00 - 46.61 & 1 & 2.5 & @xmath195 + + 0.8 - 1.6 & 43.10 - 43.60 & 6 & 3.9 & @xmath196 + 0.8 - 1.6 & 43.60 - 44.10 & 19(2 ) & 22.5 & @xmath197 + 0.8 - 1.6 & 44.10 - 44.60 & 44 & 41.4 & @xmath198 + 0.8 - 1.6 & 44.60 - 45.30 & 51 & 47.7 & @xmath199 + 0.8 - 1.6 & 45.30 - 46.00 & 7 & 10.7 & @xmath200 + 0.8 - 1.6 & 46.00 - 46.50 & 4 & 3.2 & @xmath201 + 0.8 - 1.6 & 46.50 - 47.00 & 4 & 1.9 & @xmath202 + + 1.6 - 2.3 & 43.80 - 44.50 & 9 & 14.4 & @xmath203 + 1.6 - 2.3 & 44.50 - 45.20 & 18 & 15.0 & @xmath204 + 1.6 - 2.3 & 45.20 - 45.90 & 14 & 8.9 & @xmath205 + 1.6 - 2.3 & 45.90 - 46.60 & 2 & 3.0 & @xmath206 + 1.6 - 2.3 & 46.60 - 47.00 & 0 & 0.6 & @xmath207 + 1.6 - 2.3 & 47.00 - 47.63 & 1 & 1.6 & @xmath208 + + 2.3 - 4.6 & 44.10 - 44.80 & 7 & 9.0 & @xmath209 + 2.3 - 4.6 & 44.80 - 45.40 & 8 & 6.6 & @xmath210 + 2.3 - 4.6 & 45.40 - 46.20 & 7 & 5.8 & @xmath211 + 2.3 - 4.6 & 46.20 - 47.00 & 1 & 2.2 & @xmath212 + 2.3 - 4.6 & 47.00 - 47.63 & 2 & 1.3 & @xmath213 + + [ tab : xlf_f3 ] notes : @xmath130 $ ] in 0.5 - 2 [ kev ] .
@xmath131 $ ] .
@xmath132 the agns in this row are outside of the luminosity range used for the model fit . as an alternative to the @xmath0 method
, we have developed the following estimator for the binned lf , which is free from most biases unavoidable in the @xmath0 method . in sect .
[ sec : ana ] , we have found a smooth analytical function which describes the behavior of the sxlf in a given redshift range .
having the best - fit smooth function , the estimated numerical value for the sxlf in a given bin in the @xmath214-space is : @xmath215 where @xmath216 and @xmath217 are the luminosity and redshift representative of the @xmath218-th bin , @xmath219 is the best - fit analytical expression evaluated at this point , @xmath220 is the actual number of agns observed in the @xmath218-th bin , and @xmath221 is the predicted number of agns in the bin from the best - fit analytical expression .
hereafter , we refer to eq .
[ eq : est ] as the `` @xmath222 estimator '' .
note that the estimator proposed by page & carrera ( @xcite ) ( hereafter pc ) is a special case of eq .
[ eq : est ] where @xmath223 ( or @xmath224 ) .
another advantage of this estimator over @xmath225 is that exact errors at a given significance can be evaluated using poisson statistics .
one disadvantage of this estimator is that it is model - dependent , at least in principle .
since our analytical expressions are satisfactory representations in any case and the estimator is not sensitive to the details of the underlying model , the uncertainties due to the model dependence are practically negligible . in order to compare the goodness of the estimators , we performed simulations . using the actual best - fit model for the @xmath226 bin for @xmath227
, we generated a set of simulated agns .
the number of simulated agns are 10 times those of the actual sample in order to reduce the poisson errors . using the simulated agns and the actual flux - area relation of our combined sample , we estimated binned sxlfs using three different estimators : @xmath0 , @xmath222 ( eq . [ eq : est ] ) and that of pc .
the results are compared with the underlying sxlf , which was used to generate the simulated agns , in fig .
[ fig : compest ] .
for the models to evaluate @xmath86 , we used the re - fitted model using the simulated sample rather than the original model .
the 1@xmath228 errors for the @xmath222 ( eq . [ eq : est ] ) and the pc estimators are poisson errors calculated using eqs .
( 7 ) and ( 12 ) of gehrels ( @xcite ) . on the other hand ,
the errors for the @xmath0 estimator are from eq .
( 3 ) of paper i and are inaccurate for bins with a small number of agns . as shown in fig .
[ fig : compest ] , the @xmath222 estimator best represents the original model and no estimated point deviates from the underlying model by more than 2@xmath228 .
the @xmath0 estimator underestimates the xlf in the lowest luminosity bin as found in pc .
we note that the pc estimator also systematically underestimates the lf in this particular case of the underlying model and the flux - area relation .
this is expected because their estimator implicitly builds in the assumption @xmath223 as the underlying lf shape .
this is much more weighted towards higher luminosities than any part of the realistic agn xlf .
since the amount of this bias depends on the underlying model and the flux - area relation , as well as the points in the bin where the data are plotted , it is not surprising that the bias is not apparent in fig . 2 of pc .
they have also compensated for this bias upon comparing the estimated lf with a model . instead of correcting the estimated binned lf using a good model ( which our @xmath222 estimator does ) , they calculated the `` model - expectated value of the estimator '' to compare with the estimated value from the data . detailed investigation and comparison of these different estimators in various cases are beyond the scope of this paper .
judging from this simulation , the above discussion on biases , and that the exact poisson errors can be used for errors , we choose to use the @xmath222 estimator for our plots and tabulation . using the @xmath222 estimator , we revised the full sxlf plot ( fig . 3 of paper i ) , as shown in fig .
[ fig : exlf ] . instead of connecting the data points , we overplotted the analytical model for each redshift bin .
the resulting binned sxlf are listed in tables [ tab : xlf_f10 ] , [ tab : xlf_o3 ] , and [ tab : xlf_f3 ] for different sets of cosmological parameters respectively .
the columns of these tables are ( 1 ) the redshift range of the bin ; ( 2 ) the luminosity range of the bin ; ( 3 ) the number of agns in the sample for the bin and the number of non-``type 1 '' agns as defined in appendix a of paper i ; ( 4 ) the number of agns expected from the analytical model derived in sect .
[ sec : ana ] .
( 5 ) the binned sxlf estimated using eq .
[ eq : est ] using the model xlf evaluated at the central point of the bin , i.e. @xmath229 ( see table [ tab : fit ] ) and @xmath230 , where the subscripts min and max signify the borders of the bin in columns ( 1 ) and ( 2 ) .
the upper and lower errors correspond to poisson errors estimated by eqs .
( 7 ) and ( 12 ) of gehrels ( @xcite ) respectively using s=1 ( corresponding to the confidence of the gaussian 1@xmath228 .
when there is no object in the bin , the poissonian 90% confidence upper - limit is given ( corresponding to 2.3 objects ) .
we recommend use of the values and errors under this column when , e.g. overplotting observed sxlf values with model predictions .
figure [ fig : exlf ] shows that the lowest - redshift , lowest - luminosity bin has a significant excess over the two power - law analytical expression ( agns belonging to this bin have not been used for the two power - law fit ) , thus the actual underlying sxlf has a much steeper slope than that used for @xmath86 . in order to evaluate the bias caused by this , we have made an @xmath222 estimate of this particular bin using the _ local _ slope of @xmath231 instead of @xmath232 from the two power - law model .
this gave a value about 10% lower , thus , the difference is much smaller than the statistical errors for this bin .
we have shown the tables of observed sxlf values for a number of standard set of cosmological parameters .
these values are intended for direct comparison with models and plotting with realistic error bars .
however , we list a number of caveats and sources of uncertainties , and related issues . * _ countrate - to - flux conversion : _ for the pspc - based observations , where we can limit pulse - height channels , the uncertainty in the countrate - flux ( in 0.5 - 2 [ kev ] ) conversion is small ( @xmath233 3% for photon spectral index of @xmath234 . at the faintest end ( @xmath235 ) , where only the hri data are available ,
the conversion rate varies by @xmath236 for the same spectral index range . *
_ optical classification of agns : _ since different catalogs used in this analysis have different criteria for type i and type ii agns , we did not show separate expressions for these two populations .
see appendix a. of paper i for the approximate difference in behavior of the tentative `` type i '' sample . * _ incompleteness : _ most of the surveys used in the analysis are highly complete or we have selected an appropriate complete subset . in case
there is incompleteness , we have corrected for it by assuming that the redshift distribution / content of the remaining sources are the same as the identified ones in the same flux range .
this assumption is not likely to be the case , considering that they have not been identified not because of a random cause but because of optical faintness and difficulty in obtaining decent optical spectra .
the only place that this could affect significantly is faintest end of rds - lh ( @xmath237 ) , where the identification completeness is @xmath238 .
( at @xmath239 , the identification completeness is @xmath240% in any flux range ) .
this could affect the behavior of , e.g. the apparent break at the low luminosity end in @xmath241 . * _ x - ray spectra / absorption : _ a serious model composer should be aware that the luminosity given here is for @xmath242 [ kev ] in the observer frame .
thus , one should compare the model , with their own spectral assumptions ( spectral index , absorption , fraction of absorbed agns which may depend on luminosity / redshift ) , should compute the apparent luminosity in the @xmath11[kev ] range and compare it with the values listed in this paper . the latest population synthesis model based on absorbed and unabsorbed agns by gilli et al .
( @xcite ) has applied this approach using the tables shown in this paper . however , as discussed in paper i. , the no k - correction case corresponds to a k - correction assuming a power - law photon index of @xmath243 , which is the most representative spectrum for the soft x - ray sources in the sample .
thus for many purposes , considering our tabulated values as k - corrected sxlf would be accurate enough . * _ large - scale structure : _ the lowest redshift bin covers 0.015 @xmath244 0.2 and there is some concern about the effect of the large - scale structure of the universe , which could be confused with the effect of evolution .
zucca et al .
@xcite found an underdensity of galaxies in the local universe out to z@xmath245 0.05 .
however , it might be because of the structure within their survey field of @xmath246 [ deg@xmath247 rather than that of the entire space out to this redshift .
the solid angle surveyed by rbs is @xmath245 50% of the sky and the fields of sa - n are scattered in various directions . in any event
, our lowest redshift bin samples a sufficiently large volume of space to @xmath248 with a uniform redshift coverage , thus it is unlikely that the calculated sxlf is significantly biased by the large - scale structure of the universe .
this work is based on a combination of extensive _ rosat _ surveys from a number of groups .
our work is indebted to the effort of the _ rosat _ team and the optical followup teams in producing data and the catalogs used in the analysis .
in particular , we thank k. mason , a. schwope , g. zamorani , i. appenzeller , and i. mchardy for providing us with and allowing us to use their data prior to publication of the catalogs .
tm was supported by a fellowship from the max - planck - society during his appointment at mpe .
gh acknowledges dlr grant fkz 50 or 9403 5 .
we thank the referee , t. shanks , for useful comments .
|
this is the second paper of our investigation of the 0.5 - 2 kev soft x - ray luminosity function ( sxlf ) of active galactic nuclei ( agn ) using results from _ rosat _ surveys of various depth .
the large dynamic range of the combined sample , from shallow large - area _
rosat _ all - sky survey ( rass)-based samples to the satellite s deepest pointed observation on the lockman hole , enabled us to trace the behavior of the sxlf . while the first paper ( miyaji , hasinger , schmidt @xcite , hereafter paper i ) emphasized the global behavior of the sxlf , cosmological evolution and contribution to the soft x - ray background , this paper presents actual numerical values for practical use of our results . to present the binned sxlf ,
we have used an improved estimator , which is free from biases associated with the conventional @xmath0 estimator .
|
since the mid sixties , the gravitational collapse of stars has been studied based on computer simulations @xcite . investigated topics in the collapse phase included general relativistic dynamics @xcite , magnetic fields @xcite , deleptonization and neutrino trapping @xcite , progenitor rotation @xcite , dissociation energy @xcite , neutrino transport and thermalization @xcite , nuclear electron capture @xcite , and the emission of gravitational waves @xcite , to name only a few of the many possible references .
later , the field has somewhat separated into two rather disjunct lines of research . on the one hand ,
the increasing computing power was focused on the details of the neutrino physics and neutrino transport in spherical symmetry . on the other hand ,
the computational resources were invested in multidimensional dynamics for the investigation of rotating progenitors @xcite , general relativity @xcite , and magnetic fields @xcite .
only few recent multidimensional collapse simulations made the effort to include neutrino physics .
these schemes are either very computationally expensive @xcite or rely on simplifications of the neutrino transport and its microphysics @xcite . among the neutrino physics that is most difficult to capture are accurate composition - dependent rates of electron captures on nuclei @xcite in the early and medium phase of collapse and the competition between neutrino diffusion and neutrino thermalization in the later phase of the collapse @xcite .
electrons at high matter density are degenerate so that they are captured with large energies on bound or free protons .
the produced high energy neutrinos first thermalize by electron scattering until their energy - dependent mean free path is large enough to make diffusion competitive . as the trapped neutrinos
become degenerate themselves in the late stage of collapse , the ability to stream away at low energy actually becomes the bottleneck for further deleptonization .
most recent simulations of this thermalization - diffusion process are based on an individual treatment of different neutrino energy bins with boltzmann neutrino transport @xcite .
the neutrino physics enters the equations of hydrodynamics in the form of three different source terms : as an electron fraction change rate , as an energy or entropy source , and as a source of acceleration by neutrino stress .
this paper aims to bridge the two lines of research by an efficient and very simple prescription of how published and future results of accurate neutrino transport simulations in spherical symmetry could be incorporated in the hydrodynamics of core collapse for a more realistic study of the multidimensional dynamics , the role of magnetic fields , and the emission of gravitational waves than with adiabatic simulations .
a parameterization of the three source terms is individually described and evaluated in @xmath0[sec : electron.fraction]-[sec : velocity ] .
additional tests of the robustness of the parameterization with respect to model changes or differences in the dynamics are conducted in @xmath0[sec : model.dependence ] .
after the conclusion , more code - specific details of the implementation are defined in the appendix .
stellar core collapse proceeds by an imbalance between the self - gravitating forces of the inner core and its fluid pressure .
the baryons contribute the most significant part to the gravitational mass of the stellar core while the degenerate electron gas provides the dominant contribution to the pressure .
the electron fraction , defined as the number of electrons per baryon in the gas , is therefore the most fundamental quantity for the stability of the inner core and the evolution of its size during the dynamical collapse .
the electron fraction evolves by electron captures on nuclei and the emission of the produced electron neutrinos .
see @xcite for a recent review of the nuclear input physics before and during core collapse .
the collapse continues until the matter at the center reaches nuclear density .
strong interactions reduce the compressibility at that point and the inner core bounces .
the outgoing pressure wave turns into a shock wave as soon as it reaches the sonic point at the edge of the homologously collapsing inner core .
the size of the inner core is important because it determines the location of this transition , the initial energy imparted to the shock , and the amount of matter outside of the shock that will be accreted and dissociated in the ongoing evolution .
figure [ fig : ye.trajectories ] shows the electron fraction , @xmath1 , as a function of density , @xmath2 , at different times , @xmath3 .
the data has been taken from a general relativistic core collapse simulation with boltzmann neutrino transport and `` standard '' input physics .
the selected time slices correspond to the instances at which the central density reaches @xmath4 g/@xmath5 , @xmath6 g/@xmath5 , @xmath7 , @xmath8 g/@xmath5 , and finally @xmath9 at bounce .
figure [ fig : ye.trajectories ] demonstrates that the function @xmath10 depends only weakly on time .
hence , it could be interesting to investigate how hydrodynamics simulations behave when the computationally expensive calculation of @xmath10 is replaced by linear interpolation in the logarithmic density of a time - independent tabulated template of @xmath11 . because the electron fraction profile should be as accurate as possible at the time of bounce , when the final size of the inner core is determined , the choice @xmath12 at the time of bounce , @xmath13 , will be investigated .
the data for the bounce electron fraction profile of model g15 are listed in machine - readable tables in @xcite .
alternatively , a fitting formula is provided here to increase the flexibility of the approach .
the fitting of @xmath11 is based on a piecewise linear approximation with a piecewise cubic correction .
the parameters are two points in the @xmath2-@xmath14 space , @xmath15 , and a scale , @xmath16 , of the correction . suggested values of the parameters
are given in table [ tab : fit - formula ] .
.parameters for the fitting - formula.[tab : fit - formula ] [ cols= " < , < , < ,
< , < , < " , ] the fitting formula reads @xmath17 \nonumber \\ y_{e}\left ( x\right ) & = & \frac{1}{2}\left ( y_{2}+y_{1}\right ) + \frac{x}{2}\left ( y_{2}-y_{1}\right ) \nonumber \\ & + & y_{c}\left [ 1-|x|+4|x|\left ( |x|-1/2\right ) \left ( |x|-1\right ) \right ] .\label{eq : fit - formula } \end{aligned}\ ] ] the comparison of the fit with the original data for models n13 and g15 in fig . [ fig : fit - formula ] is very satisfactory .
table [ tab : fit - formula ] shows that the fits for the two models only differ in the density at the base of the silicon - oxygen layer , @xmath18 , and in the central electron fraction , @xmath19 .
the fit can thus easily be modified by variation of these two physical quantities .
the implementation of an electron fraction evolution along @xmath11 in a pure hydrodynamics scheme is achieved by the simple prescription@xmath20 where the variation is taken at the same fluid element .
the minimum function guarantees that the electron fraction monotonically decreases even if transient instances would occur where @xmath21 would be larger than the actual @xmath14 . at the very start of a simulation , for example , fig .
[ fig : ye.trajectories ] indicates that @xmath22 almost everywhere ( @xmath14 in the progenitor data is represented by the profile belonging to the central density @xmath23 g/@xmath5 ; @xmath21 is represented by the profile belonging to @xmath24 ) .
thus , the deleptonization described by eq .
( [ eq : dyedt ] ) sets in smoothly after a short time of adiabatic compression during which the original electron fraction profile moves to the right in the @xmath2-@xmath14 space shown in fig .
[ fig : ye.trajectories ] to join the bounce profile , @xmath21 .
the latter is then followed thereafter .
figure [ fig : yevol.ps ] compares the resulting evolution of the electron fraction to the evolution obtained in the full model g15 with accurate neutrino transport .
the small deviations between the two solutions are readily explained with the profiles shown in fig .
[ fig : ye.trajectories ] : at low density , early time slices ( marked by triangles ) have a lower electron fraction than the bounce profile ( marked by a circle ) . at densities
@xmath25 g/@xmath5 the realistic profile ( marked by a diamond ) assumes a larger electron fraction than the bounce profile .
time slices that reach to even larger densities ( marked by a square and a cross ) show lower electron fractions than the bounce profile . because the bounce profile has been taken as template for the parameterization , the same differences are reflected in fig .
[ fig : yevol.ps ] , most clearly visible in the profile at @xmath26 ms before bounce .
the parameterized deleptonization leads to somewhat higher @xmath14 values in the outer layers , to lower @xmath14 values in intermediate regions , and to higher @xmath14 values near the center .
the deviations are within @xmath27 .
the @xmath14 value in model g15 rises again around nuclear density because of the increasing neutron degeneracy . due to the minimum function , this change is not captured by eq .
( [ eq : dyedt ] ) so that the central @xmath14 value at bounce in the parameterized evolution eventually falls below the corresponding value in the g15 model .
a parameterization of the lepton fraction instead of the electron fraction would improve this , because the @xmath14 would then assume its correct equilibrium value after neutrino trapping .
also the entropy changes would more consistently derive from lepton fraction changes than from electron fraction changes .
the decision to work with the electron fraction was motivated by the dominant role of @xmath14 in the determination of the chandrasekhar mass of the bouncing core , because @xmath14 is a common input variable of realistic equations of state , and because a numerical determination of weak equilibrium would be required to find @xmath14 otherwise .
the @xmath14 profile of model g15 at @xmath26 ms after bounce shows the prominent electron fraction trough that arises when the accretion shock dissociates matter at neutrino - transparent densities around the enclosed mass @xmath28 m@xmath29 .
the huge number of neutrinos emitted in the neutrino burst also leads to neutrino absorption ahead of the shock ( manifest in the @xmath14 peak at @xmath30 m@xmath29 in the time slice at @xmath26 ms after bounce ) .
these @xmath14-changes can of course not be expected to be represented by eq .
( [ eq : dyedt ] ) based on the stationary bounce template .
more sophisticated techniques are necessary to adequately implement the neutrino physics in the long - term postbounce phase .
the electron captures during collapse are not only changing the electron fraction , the matter entropy is affected as well .
the baryons are in nuclear statistical equilibrium and the electrons are in thermal equilibrium .
thus , the increments of the entropy per baryon , @xmath31 , are determined by the values of the chemical potentials @xmath32 , @xmath33 , and @xmath34 of neutrons , protons , and electrons , respectively .
additionally , there is an energy transfer between matter and neutrinos , @xmath35 ( e.g. @xcite and references therein),@xmath36 the temperature of the fluid is denoted by @xmath37 . depending on the density of the material and the energy of the produced neutrinos , the neutrino can either ( i ) directly escape , ( ii )
thermalize and escape , or ( iii ) be trapped for longer than the dynamical time scale . in regime ( i ) , @xmath38 where @xmath39 is
the average energy of the freely escaping neutrinos @xcite , thus @xmath40 in this regime , electron capture on nuclei dominates over electron capture on protons . due to the average q - value of the nuclei @xmath41 mev @xcite the neutrinos
are produced with an average energy that is only marginally larger than @xmath42 .
the corresponding small entropy decrease in this regime shall be neglected in the following parameterization . in regime ( ii )
the produced neutrinos start to be trapped by coherent scattering off heavy nuclei .
the increasing matter density causes the electron chemical potential to rise so that the neutrinos are produced with higher initial energies . as the neutrino mean free path is proportional to @xmath43 , the fastest way of escape proceeds through thermalization at the high end of the neutrino energy spectrum with a transition to diffusion and escape at its low end .
figure 7 in @xcite illustrates this thermalization process , which ends with a final escape energy of order @xmath44 mev . for the following parameterization of the entropy changes
, @xmath45 is used as a constant parameter that defines where regime ( ii ) commences , namely when @xmath46 .
the entropy increase is then evaluated according to@xmath47 where @xmath48 is given by eq .
( [ eq : dyedt ] ) . at even higher densities , in regime ( iii ) , the neutrinos are not able to escape before bounce .
they are in equilibrium with the fluid so that @xmath35 in eq .
( [ eq : entropy.general ] ) is determined by the neutrino chemical potential , @xmath49 . the neutrinos form a normal gas component without transport abilities and the entropy is conserved in accordance with eq .
( [ eq : entropy.general ] ) and the equilibrium condition @xmath50 as second parameter , a density threshold @xmath51 g/@xmath5 is introduced to define the beginning of regime ( iii ) beyond which no further entropy changes are taken into account .
figure [ fig : sevol.ps ] presents the entropy evolution based on eq .
( [ eq : dsdt ] ) in comparison to the entropy profiles of model g15 with comprehensive neutrino transport .
the two parameters used in eq .
( [ eq : dsdt ] ) enable nice agreement in the collapse phase ( profiles at @xmath52 ms and @xmath26 ms before bounce ) .
the density at which the entropy starts to rise is mainly controlled by @xmath53 mev .
the level of the central entropy is mainly determined by the choice of @xmath54 g/@xmath5 .
the values of the two parameters have not been particularly fine - tuned because they should rather represent a generic estimate than a multi - digit optimum for one particular model .
the @xmath55 lower entropy in the parameterized evolution has no noticeable consequences for the dynamics of the cold matter in the inner core .
the entropy profiles at @xmath26 ms after bounce show the expected differences at late time due to the omission of the neutrino burst in the parameterization .
the postshock region in the realistic g15 calculation is significantly cooler because of the energy loss inferred by the neutrino emission .
the adjacent region ahead of the accretion shock is somewhat hotter in model g15 due to neutrino absorptions .
neutrino stress is the third important quantity at the interface between neutrino transport and hydrodynamics .
although the neutrino pressure only contributes a fraction to the gas pressure , it influences the size of the inner core when the gas pressure and the gravitational forces cancel to a large extent . in regime ( iii ) , i.e. the optically thick region where transport processes can be neglected , the neutrino stress is determined by the gradient of the neutrino pressure @xmath56 where @xmath57 is the lagrangean time derivative of the velocity .
the term in the middle is the general expression ; the term on the right hand side is its spherically symmetric limit based on an enclosed mass @xmath58 at radius @xmath59 .
all general relativistic effects shall be neglected in the derivation of this simple parameterization . the neutrino pressure , @xmath60 , is readily evaluated based on the thermodynamic state of the fluid,@xmath61 the neutrino chemical potential is given by @xmath62 .
the constants @xmath63 , @xmath64 , and @xmath65 refer to the planck constant , the speed of light , and the boltzmann constant , respectively .
@xmath66 is the fermi - dirac function of order n @xcite .
for an implementation of neutrino stress in the optically thin regime , an estimate of the neutrino number luminosity is needed .
it can be derived from the deleptonization in eq .
( [ eq : dyedt ] ) and the requirement of lepton conservation .
the simplest possible approximation in the construction of the neutrino flux from distributed sources is that the neutrinos leave isotropically and without time delay from the locations of deleptonization .
even if this assumption is obviously wrong in core collapse , it leads to a useful first approximation of the non - local flux geometry @xcite .
firstly , the deleptonization scheme in eq .
( [ eq : dyedt ] ) already suppresses sources at high opacities .
secondly , lepton conservation requires that an isotropic source contributes with the square of its inverse distance so that closeby sources from this side of the neutrinosphere influence the direction of the neutrino flux more strongly than remote sources from the opposite side . with this assumption
, the neutrino number flux can be expressed by the gradient of a scalar field , @xmath67 , which fulfills the poisson equation@xmath68 in spherical symmetry , the neutrino number luminosity based on eq .
( [ eq : luminosity.poisson ] ) results in a straightforward integration of all enclosed sources,@xmath69 avogadro s number is denoted by @xmath70 .
let us stay in the spherically symmetric limit to derive an extension of the neutrino pressure to the optically thin regions .
the neutrino stress can be expressed as a convolution of the neutrino number luminosity , the neutrino energy , and the energy - dependent reaction cross sections @xcite .
if the reaction cross sections are represented by the inverse mean free path , @xmath71 , one obtains based on the neutrino distribution function , @xmath72,@xmath73 the neutrino momentum phase space element , @xmath74 , is described by the neutrino energy , @xmath75 , and the propagation angle cosine , @xmath76 , with respect to the radial direction .
let s now assume that the spectrum of the number luminosity is well approximated by a fermi - dirac function with degeneracy parameter @xmath77 .
the number luminosity can then be expressed as @xmath78 for the following rough estimate of the decline of the neutrino stress in the outer layers , it is further assumed that the mean free path scales as @xmath79 , so that the energy integration in eq .
( [ eq : acceleration.by.stress ] ) can be evaluated and expressed by the number luminosity in eq .
( [ eq : luminosity]),@xmath80 the neutrino temperature , @xmath81 , and the degeneracy parameter , @xmath77 , can not easily be derived as a function of radius without solving a more detailed transport problem .
though , at high opacity , they limit to the matter temperature , @xmath82 , and the neutrino degeneracy , @xmath83 , respectively . the neutrino stress for the hydrodynamics simulation is now generated by the following procedure : an estimate for the number luminosity is given by eq . ( [ eq : luminosity.estimate ] ) .
then , at all densities larger than a specified neutrino trapping density @xmath84 , the neutrino stress is evaluated according to eqs .
( [ eq : acceleration.by.pressure ] ) and ( [ eq : neutrino.pressure ] ) . with the neutrino stress at density @xmath84
, a constant @xmath85 is defined that attaches the scaling estimate found in eq .
( [ eq : stress.scaling ] ) to the neutrino stress given by eq .
( [ eq : acceleration.by.pressure ] ) , @xmath86 ^{-1}.\ ] ] in eq .
( [ eq : stress.proportionality ] ) , @xmath59 , @xmath87 , @xmath37 , @xmath88 , and @xmath89 are evaluated at the transition density @xmath84 .
now , before eq .
( [ eq : stress.scaling ] ) can be applied to extend the neutrino stress to the region @xmath90 , approximations for @xmath81 and @xmath77 must be found for regime ( i ) .
there , it is assumed that the neutrino spectrum is well represented based on a degeneracy @xmath91 .
it is then consistent with @xmath0[sec : entropy ] and the assumption of spherical symmetry if i express the neutrino temperature by the average neutrino energy parameter , @xmath45 , and set @xmath92 .
the simplest continuous transition of the neutrino stress from regime ( iii ) to regime ( i ) is realized by the adoption of the larger of the two limiting cases in each intermediate point .
hence , at @xmath90 , the neutrino stress is approximated by @xmath93 , \ ] ] where @xmath59 , @xmath37 , @xmath94 , and the estimate of @xmath89 according to eq .
( [ eq : luminosity.estimate ] ) represent the local values at the point where @xmath95 is evaluated .
suggestions with respect to the implementation of this spherically symmetric neutrino stress in a multidimensional hydrodynamics code and convenient approximations for the evaluation of the fermi integrals are collected in the appendix .
this procedure circumvents any explicit reference to cross sections and is therefore simple to apply .
of course , the cross sections do enter the constant @xmath85 implicitly via the relation between the neutrino pressure gradient and the number luminosity at the transition density @xmath84 .
as the value of the number luminosity is derived from eq .
( [ eq : dyedt ] ) , it reflects the deleptonization rate in the reference calculation with full transport which is based on information about the opacities used in the run that produced the electron fraction template at bounce time .
figure ( [ fig : vevol.ps ] ) shows a summary of the evolution of velocity profiles .
it demonstrates that the collapse dynamics is accurately reproduced with the parameterized neutrino physics .
some differences can still be made out : as discussed in @xmath0[sec : electron.fraction ] , eq .
( [ eq : dyedt ] ) implies a transient halt in the deleptonization until the density has increased enough that the initial @xmath14 profile ( labeled by @xmath23 in fig .
[ fig : ye.trajectories ] ) catches up with the template ( labeled by @xmath24 ) .
this leads to a small delay in the infall of the outer layers with respect to the reference g15 model .
this is visible in the velocity profiles at the enclosed mass of @xmath96 - @xmath97 m@xmath29 . the time slice at @xmath26 ms before bounce shows a soft outgoing pressure wave in the solution with the parameterized neutrino physics that is not present in the reference model .
the first suspicion was that it is related to the treatment of the neutrino stress because its appearance coincides to some extent with the moment when the density @xmath84 is reached at the center , i.e. the moment when the applied neutrino stress jumps from zero to the value described by eq .
( [ eq : dvdt.tail ] ) . however , a more careful investigation showed that the dominant reason is the shallower electron fraction profile close to the center in the @xmath98 ms time slice ( see fig .
[ fig : yevol.ps ] ) .
the outgoing wave is caused by a combination of the excess electron pressure at the center with the electron pressure deficit around an enclosed mass of @xmath99 m@xmath29 .
it introduces visible perturbations in the vicinity of the position of shock formation in the time slice at @xmath100 ms before bounce .
nevertheless , the much stronger outgoing pressure wave from bounce and its steepening to the bounce - shock are nicely reproduced in the @xmath101 ms and @xmath102 ms time slices .
the significant differences induced by the neutrino burst in the electron fraction and entropy profiles at @xmath26 ms after bounce seem to have surprisingly little influence on the early dynamics .
the velocity profiles as functions of enclosed mass still agree at @xmath26 ms after bounce .
the differences in the state of the postshock matter rather affect the shock radius than the shock mass .
while the s@xmath103hock in the g15 model is positioned at a radius of @xmath104 km , it already has reached @xmath105 km in the parameterized run due to the omission of the lepton and energy loss by the neutrino burst .
after the demonstration that this simple parameterization works well for the one model g15 , this section aims to explore to what extent the approximations are robust against variations in the dynamics or the initial conditions .
a first test is the application of the same parameterization with the same parameters for @xmath45 and @xmath84 to a different stellar model .
the g15 model discussed above is launched from the @xcite model s15s7b2 .
it has a quite typical iron core @xmath106 m@xmath29 .
the run n13 in @xcite is based on the progenitor of @xcite with an especially small iron core @xmath107 m@xmath108 figure [ fig : model.ps ] compares different runs at the time of core - bounce .
figure [ fig : model.ps]a shows the bounce profile @xmath109 of the n13 run that was used to specify the deleptonization according to eq .
( [ eq : dyedt ] ) for the parameterized run a13 .
the bounce profiles of the n13 and a13 runs are displayed in fig .
[ fig : model.ps]b - d .
the quality of the approximation is very similar to the one discussed above for the g15 model . the same choice of the parameters @xmath45 and @xmath84 seems to fit also other stellar models .
the dynamics in the n13 and g15 models is also different .
the former has been calculated with newtonian hydrodynamics and @xmath110 neutrino transport , the latter with general relativistic dynamics and transport . as a second test
i investigate how the parameterized solution reacts if the bounce template for a newtonian simulation is taken from a general relativistic model of the same progenitor model .
thus , a bounce - template g13 in fig .
[ fig : model.ps]a has been produced from a general relativistic calculation with neutrino transport . a repetition of run a13 with the exchanged template , @xmath111 instead of @xmath109 , leads to the results labeled by x13 g in fig .
[ fig : model.ps]b - d . in spite of the significant dynamical differences in the runs that produced the templates n13 ( solid line ) and g13 ( dashed line ) in fig .
[ fig : model.ps]a , the differences in velocity , electron fraction , and entropy profiles between runs a13 and x13 g are barely visible .
this means that the template for the parameterization of the microphysics can be extracted from a run whose dynamics differs from the run in which the parameterization will be applied . a third test has been performed by the exchange of bounce templates between the progenitor models .
first , a new template has been created with a newtonian simulation of the s15s7b2 progenitor model labeled by n15 in fig .
[ fig : model.ps ] .
the parameterized newtonian run x13p has then been launched from the @xmath112 m@xmath29 progenitor model using this @xmath113 template .
figure [ fig : model.ps]b - d shows a small deviation @xmath114 in the location of the shock formation . although a given template from one progenitor still seems to be useful for many dynamical investigations that are launched from different progenitor models , it might be recommended to use a template that is derived from the same progenitor to achieve the best accuracy .
a closer inspection of the differences reveals the well - known fact that the final deleptonization in the inner core is essential for the location of the shock formation . because the parameterization in eq .
( [ eq : dyedt ] ) does not support electron fraction increases , it is the minimum @xmath14 that counts .
the minimum @xmath14 in the templates of the n13 and g13 models in fig .
[ fig : model.ps]a coincide , while @xmath14 in the n15 run reached slightly lower values .
the described parameterization has been designed to provide a very efficient and reasonably accurate way to lead a hydrodynamics simulation from the onset of collapse of the supernova progenitor star to bounce .
it has also been shown to be accurate for the initial rebound of the stellar core .
however , the simplicity of the described approach entails several limitations .
the signal of gravitational waves is not likely to be limited to the short time interval around bounce that has traditionally been investigated ( @xcite and references therein ) . in
delayed , and especially in neutrino - driven explosions , the signal may decay after the first peak around bounce only to regain strength on a longer time scale of several hundreds of milliseconds , when fluid instabilities in the hot shocked matter between the protoneutron star and the shock front grow to asymmetries on larger scales @xcite .
the dynamics and the input physics for the description of the cold matter in the collapse phase are fundamentally different from the physics in the hydrostatic , but turbulent , hot mantle of the protoneutron star , which eventually determines the supernova explosion after core - bounce . the suggested scheme has not been designed for the postbounce phase and does not describe any of the weak interaction physics relevant after bounce .
firstly , the scheme handles only deleptonization and not neutrino transport .
no attempt has been made to describe the very delicate neutrino _ energy _
deposition behind the stalled accretion shock .
neutrino heating at @xmath115 ms after bounce and beyond is thought to drive important fluid instabilities .
secondly , the scheme is only applicable to situations where the deleptonization is caused by compression of cold matter .
even if the deleptonization in the condensing accreting material continues to be captured after bounce , major contributions from the emission of the neutrino burst at @xmath26 ms after bounce and from neutrino diffusion in the dense core are missed by the simple parameterization .
therefore , if one extends an otherwise adiabatic code with the suggested parameterization for a simulation of the _ postbounce _ phase , one obtains the benefit that simulations pass through a significantly more realistic configuration at bounce and that overly fast prompt explosions for small progenitors are avoided . but no improvement is obtained toward an adequate description of the important neutrino physics in the postbounce phase . another question is the reliability of the scheme in the collapse phase when fast rotation rates are applied .
the density distribution is then significantly less spherically symmetric .
i think one can be somewhat optimistic because the parameterization of the deleptonization during collapse in eqs .
( [ eq : dyedt ] ) and ( [ eq : dsdt ] ) relies more on the local density evolution than on the global geometry .
this is supported by the low sensitivity of the parameterization to global dynamical changes investigated in @xmath116[sec : model.dependence ] .
a global asymmetry changes rather the spatial distribution of local conditions than the quality of the local conditions themselves .
relevant local conditions are : the time scale of compression , the density gradient , or the curvature of the isodensity surface at the point of investigation ( as a measure of the deleptonizing volume per surface area ) .
these quantities as function of the density may change by several percent with respect to a spherically symmetric scenario , but probabely not by orders of magnitude .
if this speculation is true , the parameterization would still be approximately applicable .
it is likely that accuracy is lost with increasing asphericity , but the improvement with respect to adiabatic simulations could still be significant
. it would now be tempting to make also the treatment of the neutrino stress multidimensional .
equation ( [ eq : acceleration.by.pressure ] ) already has a multidimensional form and eq .
( [ eq : dvdt.tail ] ) is readily generalized to @xmath117 , \ ] ] where @xmath67 is determined by eq .
( [ eq : luminosity.poisson ] ) .
unfortunately , it is not clear , how @xmath85 should be chosen in the multidimensional setting .
simple choices may imply discontinuities in the transition of the neutrino stress from eq .
( [ eq : acceleration.by.pressure ] ) to eq .
( [ eq : dvdt.tail.multid ] ) that could introduce undesired artefacts in the velocity field .
the spherically symmetric implementation of the neutrino stress , as layed out in the appendix , is very simple and physically well - defined , but fails to represent any asymmetries in the shape or temperature of the neutrinospheres and in the emitted neutrino luminosities .
this might still be an acceptable limitation because the influence of the neutrino stress ( momentum deposition ) on the collapse dynamics is not as crucial as the much discussed neutrino heating ( energy deposition ) later in the postbounce phase . in the postbounce phase , which is not in the scope of this parameterization , asymmetries in the neutrino field
can have a significant impact on the dynamics of the shock revival @xcite . on the other hand ,
stellar evolution models point to rather slowly rotating inner cores @xcite and corresponding collapse simulations with multidimensional neutrino transport have not shown evidence for sizable asymmetries in the neutrino field @xcite .
the dynamics of core collapse is dominated by electron pressure .
dynamical simulations are only realistic if they can take account of electron captures on bound or free protons during infall . at increasing densities
the electron captures get inhibited by neutrino phase space blocking so that the ability to thermalize and emit the produced neutrinos significantly contributes to the determination of the final electron fraction in the inner core .
the lower the electron fraction , the smaller is the mass of the core that bounces when nuclear densities are reached and the smaller is also the initial energy imparted to the outgoing shock .
these well - known relationships ask for a careful inclusion of neutrino physics in simulations of stellar core collapse .
the most accurate treatment of neutrino transport has been developed in spherical symmetry @xcite where it has been coupled to the most recent evaluation of electron capture rates on heavy nuclei @xcite .
the previous sections present an embarassingly simple and computationally efficient parameterization of the deleptonization in the collapse phase so that the consequences of past and future improvements of the neutrino physics should be straightforward to incorporate in multidimensional simulations that focus on other aspects of core collapse .
the scheme is based on electron fraction profiles as function of density at bounce , @xmath11 , that have been produced by spherically symmetric simulations with full neutrino transport .
a fitting formula is provided to conveniently represent these electron fraction profiles and to allow for adjustments to future developments .
it is found that the time - evolution of the electron fraction in the simulations , @xmath10 , follows in reasonable approximation the profile @xmath11 .
the corresponding parameterization brings the electron fraction at bounce , when it is most important for the shock dynamics , very close to the value in the original simulation .
unfortunately , it is not exactly equal because small differences in the time - dependence of the deleptonization will lead to density differences at bounce so that @xmath11 is not evaluated at exactly the same position as in the original simulation . once the deleptonization is given , the entropy changes and neutrino stress are estimated based on two additional parameters , an average neutrino escape energy , @xmath53 mev , and a neutrino trapping density , @xmath54 g/@xmath5 .
these parameters have not been fine - tuned , but may require adaption if the weak interaction physics is changed . in a comparison of
the parameterized runs with the original ones it is found that the dynamics of core collapse is quite accurately reproduced .
the parameterized neutrino physics presents a significant improvement with respect to adiabatic treatments .
its accuracy may even rival with neutrino transport approximations that neglect neutrino - electron scattering .
however , some clearly visible inaccuracies do develop .
the deleptonization of central zones at high densities proceeds slower than in the reference simulation .
this leads to a weak outgoing pressure wave before bounce - density is reached .
it propagates to the point of shock formation where it causes moderate perturbations .
inaccuracies due to a deviation in the timing of the deleptonization are also found in the hydrodynamic structure ahead of the shock .
these deviations are a direct consequence of the differences between the function @xmath11 and the time evolution of the electron fraction in the reference simulation .
the evolution of the electron fraction and entropy is based on the local thermodynamic state of the fluid .
this should allow an application also in models where the dynamics moderately differs from the spherically symmetric case with which the electron fraction template has been produced .
the less important neutrino stress has been approximated in a spherically averaged manner .
an extension to handle asymmetries in the neutrino stress has been sketched , but its accuracy in models with essentially aspherical density distributions remains to be investigated .
the sensitivity of the parameterized runs to changes in the model has been tested in two experiments that are accessible with comprehensive neutrino transport : firstly , a template produced with general relativistic dynamics has been used in a parameterized run with newtonian gravity and , secondly , a template produced with a 15 @xmath118 progenitor star has been used in a parameterized run launched from a 13 @xmath118 progenitor star with a very different structure of the iron core .
it was found that the parameterization is not sensitive to the dynamical differences between general relativistic and newtonian gravity .
the choice of the progenitor for the template , however , has a small influence on the parameterization .
the extent of the deviation is directly determined by the difference between the central @xmath14 values in the templates . in the investigated case , the @xmath14 difference translates to @xmath114 difference in the point of shock formation .
i thus believe that one and the same template can be used with different progenitors for qualitative studies of the dynamics . but a progenitor - specific template is recommended for more quantitative investigations . the parameterization has been designed for the collapse phase and not for the postbounce evolution .
it does not contain any physics related to neutrino heating . during collapse , it relies on the weak interaction physics used for the production of the bounce template .
the nuclear and weak interaction physics during collapse is an interesting field of research .
progress in its adoption in spherically symmetric simulations has been made @xcite and is likely to continue .
hopefully , this paper enables the efficient transfer of results from state - of - the - art neutrino transport simulations in spherical symmetry to simulations with more spatial degrees of freedom , where the implementation of comprehensive weak interaction physics together with the accurate solution of the energy - dependent multidimensional boltzmann equation is yet computationally prohibitive .
[ sec : appendix ]
the simulations with the parameterized neutrino physics are based on the same spherically symmetric general relativistic hydrodynamics code agile that was used to solve the equations of hydrodynamics in the original production of the electron fraction templates @xcite .
the interaction between neutrino physics and hydrodynamics proceeds through additional source terms in the conservation equations , i.e. @xmath119 , @xmath120 , and @xmath121 in @xcite for electron fraction change rates , total energy change rates , and momentum change rates , respectively .
the simulations use the realistic lattimer - swesty equation of state version 2.7 @xcite . note that this version is likely to crash whenever one tries to enter the nuclear regime ( eosflag=2 ) with a saved guess of the proton fraction obtained from the dissociated regime ( eosflag=3 ) . as
convergence in the dissociated regime is robust , it is advisable to save guesses only when they are returned with eosflag=2 . in
the parameterized runs with agile , eq . ( [ eq : dyedt ] ) is used to set @xmath119 .
the heating rate @xmath120 is found by numerical iteration of the equation of state until the entropy change rate specified by eq .
( [ eq : dsdt ] ) is realized .
the parameter @xmath45 is set to @xmath52 mev .
note that the approximation @xmath122 in the region where @xmath123 will still produce a rate @xmath124 when the electron fraction changes . at densities larger than @xmath54 g/@xmath5 , the spherically symmetric limit of eq .
( [ eq : acceleration.by.pressure ] ) is used to calculate the neutrino stress @xmath121 , while at lower densities , eq .
( [ eq : dvdt.tail ] ) is used . in the general relativistic simulations , the gravitational effect of neutrino
energy and pressure has only been taken into account at densities @xmath125 .
the neutrino pressure is evaluated by eq .
( [ eq : neutrino.pressure ] ) and the specific neutrino energy is set to @xmath126 .
the gravitational effect of the neutrino luminosity was neglected .
all time derivatives , @xmath127 , in eqs .
( [ eq : dyedt],[eq : luminosity.estimate],[eq : dsdt],[eq : acceleration.by.pressure],[eq : acceleration.by.stress],[eq : stress.scaling],[eq : dvdt.tail],[eq : dvdt.tail.multid ] ) are lagrangean , i.e. taken at the same mass element .
most hydrodynamics codes are discretized in space and not in mass ; they rely on eulerian time derivatives , @xmath128 . for the implementation of above equations in these schemes
i suggest to use operator splitting .
for example , the conservation equations of hydrodynamics should be straightforward to extend with a conservation equation for electron number , @xmath129 in this first step , the electron fraction at location @xmath130 is updated from @xmath131 to an intermediate value @xmath132 by the advection of electrons .
the electron fraction update is completed in a second step by application of eq .
( [ eq : dyedt ] ) , @xmath133 the entropy update can similarily be split into an entropy conserving hydrodynamics update ( in whatever form it is realized in the hydrodynamics code ) and a lagrangean change of the specific entropy . for the latter , @xmath48 in eq .
( [ eq : dsdt ] ) is substituted by the result of eq .
( [ eq : dyedt.euler ] ) .
it is numerically stable to apply the neutrino stress in an operator split fashion as well because the neutrino pressure contributes only of order @xmath134 to the total pressure .
the simple spherical limit in eq .
( [ eq : acceleration.by.pressure ] ) is adequate if the deviations from spherical symmetry are small ; the multidimensional form is advised otherwise .
the extension of the neutrino stress to the optically thin regime is less straightforward because its derivation was based on some arguments that apply only in spherical symmetry .
( [ eq : stress.proportionality ] ) , for example , extracts the opacities from the spherically symmetric run that produced the electron fraction templates .
as long as the asphericity in the density distribution does not exceed half a density scale height , i recommend to use spherically averaged conditions of the multidimensional configuration to evaluate the neutrino stress .
the integration of the density over spheres results in an enclosed mass as function of the radius .
the integration of the deleptonization in eq .
( [ eq : dyedt.euler ] ) over spheres leads to a luminosity estimate according to eq .
( [ eq : luminosity.estimate ] ) . based on the spherically integrated density , energy density , and electron density the equation of state delivers the thermodynamic conditions used in eq .
( [ eq : dvdt.tail ] ) to derive the neutrino stress for spheres with densities @xmath135 .
the applicability of the multidimensional eqs .
( [ eq : luminosity.poisson ] ) and ( [ eq : dvdt.tail.multid ] ) in highly asymmetric situations was not numerically investigated because their assessment would require the corresponding reference simulations with multidimensional neutrino transport . in any case , it is important to abandon the application of the neutrino stress in optically thin regions after bounce before the shock reaches the density @xmath136 in order to prevent the growth of the constant @xmath85 in eq .
( [ eq : stress.proportionality ] ) beyond limits .
the fermi integrals in @xmath0[sec : velocity ] are required for degeneracy @xmath137 .
convenient approximations to @xmath138 and @xmath139 have been taken from @xcite , while @xmath140 has been derived along similar lines based on @xcite : @xmath141 with @xmath142 and @xmath143 .
|
a simple and computationally efficient parameterization of the deleptonization , the entropy changes , and the neutrino stress is presented for numerical simulations of stellar core collapse . the parameterization of the neutrino physics is based on a bounce profile of the electron fraction as it results from state - of - the - art collapse simulations with multi - group boltzmann neutrino transport in spherical symmetry .
two additional parameters include an average neutrino escape energy and a neutrino trapping density .
the parameterized simulations reproduce the consequences of the delicate neutrino thermalization / diffusion process during the collapse phase and provide a by far more realistic alternative to the adiabatic approximation , which has often been used in the investigation of the emission of gravitational waves , of multidimensional general relativistic effects , of the evolution of magnetic fields , or even of the nucleosynthesis in simulations of core collapse and bounce . for supernova codes that are specifically designed for the postbounce phase , the parameterization builds a convenient bridge between the point where the applicability of a stellar evolution code ends and the point where the postbounce evolution begins .
|
density functional theory @xcite ( dft ) is by now the most popular method for electronic structure calculations in condensed matter physics and quantum chemistry , because of its unique combination of low computational cost and reasonable accuracy for many molecules and solids . in applying dft to a given electron system , the only quantity that must be approximated in practice
is the functional @xmath0 $ ] for the exchange - correlation energy .
an exact expression for this functional is the coupling - constant integral , e=_0 ^ 1 dw _ [ ] . [ cci ] the integrand
is defined as w_[]= _ [ ] ||_[]-u [ ] .
[ walpha ] here , @xmath1=\frac12\int
d\rv\int d\rv'\rho(\rv)\rho(\rv')/|\rv-\rv'|$ ] is the functional of the hartree energy and the operator @xmath2 describes the coulomb repulsion between the @xmath3 electrons , = _
i=1^n-1_j = i+1^n1|_i-_j| .
[ vee ] ( atomic units are used throughout this work . ) eventually , out of all antisymmetric @xmath3-electron wave functions @xmath4 that are associated with the same given electron density @xmath5 , @xmath6 $ ] denotes the one that yields the minimum expectation of the operator @xmath7 . here
, @xmath8 is the kinetic - energy operator .
notice that the parameter @xmath9 works as an adjustable interaction strength or coupling `` constant '' . at @xmath10 ,
@xmath11 $ ] starts out with the value e=_0[]||_0[]-u [ ] [ ex ] which is the density functional for the exchange energy .
generally , @xmath11 $ ] is a monotonically decreasing function of @xmath9 , since the electrons in the state @xmath6 $ ] increasingly tend to avoid each other in space as the repulsion strength @xmath9 grows .
thus , the expectation of @xmath2 , which is a measure for the average inverse distances between the electrons , must decrease .
however , it can not decrease indefinitely , since the electrons are forced to stay within the fixed density @xmath12 .
schematically , the traditional quantum chemistry approach to electron correlation often consists in trying to _ extrapolate _ the information on the physical system ( @xmath13 ) from the non - interacting limit ( @xmath10 ) by using , e.g. , perturbation theory or more sophisticated methods .
this work , instead , follows the early idea of wigner @xcite , further developed in the dft framework in refs . and , in which the information at @xmath13 is obtained by _ interpolating _ between the two limits of weak interaction , @xmath14 , and infinitely strong interaction , @xmath15 . in this context
, we carry out a detailed study of the limit w_=_w_. [ winf ] although the limit of eq .
was investigated in previous work @xcite , the solution presented here was found only in the special case of two electrons in a spherical density , using physical arguments . for the general case with @xmath3 electrons , the point - charge - plus - continuum ( pc ) model was proposed @xcite , w_=d .
[ wpc ] where @xmath16 and @xmath17 .
this approximation , together with a similar one for the next leading term , @xmath18 $ ] , describing zero - point motion oscillations , was used to construct an interpolation for @xmath19 $ ] between @xmath10 and @xmath20 , called interaction strength interpolation ( isi ) model @xcite .
isi predicts accurate atomization energies ( with a mean absolute error of 3.4 kcal / mole ) @xcite , showing that the general idea of interpolating between the weak- and the strong - interaction limits in dft can work . it should be emphasized that isi uses as ingredients exclusively the two functionals @xmath21 $ ] and @xmath22 $ ] ( of the second - order correlation energy in grling - levy perturbation theory @xcite ) from the relatively simple non - interacting limit @xmath14 ( with single - particle orbitals ) , plus the two functionals @xmath23 $ ] and @xmath18 $ ] from the opposite @xmath15 limit of infinitely strong repulsion , which are , so far , not known exactly ( as said , except for @xmath23 $ ] in the special case of two electrons in a spherical density @xcite ) . in this work
we reformulate the @xmath15 limit of dft in terms of a classical problem with a degenerate minimum .
this allows us to construct the general solution for @xmath23 $ ] , and to clarify many aspects of this limit .
as @xmath24 grows beyond its real - world value @xmath13 , the concept of single - particle orbitals becomes completely meaningless .
when @xmath15 , however , an entirely new type of simplicity with so - called `` co - motion '' functions arises , as we shall see in secs .
[ sec_mr]-[sec_fs ] .
the `` co - motion '' functions , which entirely determine @xmath23 $ ] , can be directly constructed from the one - electron density @xmath12 .
the paper is organized as follows . in sec .
[ sec_mr ] we first give a general overview of the problem , anticipating the solution on intuitive physical grounds , and leaving the mathematical details in secs . [ subsec_as ] and [ sec_fs ] .
we then use our formalism to explicitly evaluate the limit of eq . in the case of
spherical densities , with applications to atoms ( sec .
[ sec_nspherical ] ) . in sec .
[ sec_isi ] we compare our solution with the approximation of eq . , and we discuss the implications for the isi functional . the last sec .
[ sec_conc ] is devoted to conclusions and perspectives . in the appendix
we also consider the simple case of harmonic forces , in order to analyze how the nature of the electron - electron interaction affects the solution .
for a given @xmath3-electron density @xmath5 we generally wish to find the limit or , equivalently , w_+u[]= _ _ [ ] || _ [ ] .
[ winfdef ] if the density @xmath25 is both @xmath3- and @xmath26-representable for every @xmath9 , there exists an @xmath9-dependent external potential @xmath27=\sum_{i=1}^nv\ext^\alpha([\rho],\rv_i)$ ] such that @xmath6 $ ] is the ground state of the hamiltonian ^=++^. [ hv ] as @xmath15 , the binding external potential @xmath28,\rv)$ ] in the hamiltonian must compensate the strong repulsion @xmath29 between the electrons .
therefore , we expect _ = v ( [ ] , ) , [ eq_valphainf ] with @xmath30,\rv)$ ] a smooth function of @xmath31 .
thus , for large @xmath32 , the kinetic energy in the state @xmath6 $ ] is mainly due to zero - point oscillations of strongly repulsive electrons , bound by a strong attractive force that has the order of @xmath29 .
therefore , @xmath33|\hat{t}|\psi_{\alpha}[\rho]\rangle$ ] has the order of @xmath34 and we may write in eq .
@xmath35+u[\rho]&=&\lim_{\alpha\to\infty } \big\langle\psi_{\alpha}[\rho]\big|\frac1{\alpha}\hat{t}+\hat{v}\ee\big| \psi_{\alpha}[\rho]\big\rangle \nonumber\\ & = & \lim_{\alpha\to\infty}\min_{\psi\to\rho } \big\langle\psi\big|\frac1{\alpha}\hat{t}+\hat{v}\ee\big|\psi\big\rangle .
\label{limmin}\end{aligned}\ ] ] in the second step we have applied the definition of the wave function @xmath6 $ ] ; the constraint `` @xmath36 '' addresses all those wave functions @xmath4 that are associated with the same given density @xmath25 .
provided that the limit @xmath15 can be applied directly to the operators in eq .
( there is no rigorous proof for this reasonable conjecture see also the appendix ) , it is simplified to w_+u[]= _ || .
[ minc ] in this case , the expectation of @xmath2 alone is to be minimized , regardless of the kinetic - energy operator @xmath37 .
this apparently purely classical problem corresponds to the quantum - mechanical limit of infinitely large masses . if eq .
( [ eq_valphainf ] ) holds , that is if the density @xmath12 is both @xmath3- and @xmath26-representable also in the very @xmath15 limit , the minimizing @xmath4 in eq .
is the ground state of the pure multiplicative operator @xmath38 .
the local one - body potential @xmath39 is the lagrange multiplier for the constraint `` @xmath36 '' , and can be found via the legendre - transform formulation of eq .
@xcite , w_+u[]=_v\{_|+| -v}. [ maxmin ] to start to address the problem of @xmath3-representability in the @xmath15 limit , we write the minimizing wavefunction @xmath4 in eq . as the product
= ( _ 1, ... ,_n)(_1, ... ,_n ) , [ eq_psi ] where @xmath40 is a spatial wavefunction and @xmath41 is a chosen eigenstate of the total spin @xmath42 and its projection @xmath43 . in sec .
[ subsec_as ] we analyze the @xmath3-representability problem in more detail , showing that , in the special @xmath15 case , we can always construct an antisymmetric @xmath4 of the form , and that the choice of the spin eigenfunction @xmath41 does not affect the square of the spatial wavefunction @xmath44 and thus , as shown by eq . below , the energy .
for this reason , in what follows we only consider the spatial wavefunction @xmath45 . as said , + e([v];_1, ... ,_n ) [ opepot ] is a pure multiplicative operator , so that , after integrating out the spin variables , the unconstrained minimization in the brackets of eq . reads _
d_1 ... d_n|(_1, ... ,_n)|^2 e([v];_1, ...
[ eq_psi2energy ] we see that the minimum in eq .
is reached when the square of the spatial wavefunction @xmath46 is a distribution that is zero everywhere except for values @xmath47 , where m\{(_1, ...
,_n):e([v];_1, ... ,_n)= } [ defm ] is the set of all configurations @xmath48 for which the 3@xmath3-dimensional function @xmath49 of eq
. assumes its absolute ( global ) minimum .
notice that the set @xmath50 is purely determined by the choice of @xmath51 , @xmath52 $ ] .
we shall see in the next sec .
[ subsec_as ] that to such a minimizing distribution @xmath46 we can indeed associate an antisymmetric wavefunction of the form . for a given reasonable ( attractive )
potential @xmath51 , we may expect that @xmath50 will comprise only one single configuration ( plus its permutations ) or a small finite number of configurations : the set @xmath50 corresponds , in fact , to the solution of the classical electrostatic equilibrium problem for @xmath3 identical point charges in the external potential @xmath51 .
the density associated with the corresponding spatial wave function @xmath40 will then be a finite sum of delta functions , ( ) _ k=1^k ( -_k ) , [ rho_delta ] where the @xmath53 are the @xmath54 different position vectors forming the configurations in @xmath50 .
when @xmath51 has some symmetry ( e.g. , spherical ) , @xmath50 can also be a low - dimensional continuum ( e.g. , the surface of a sphere ) .
for a quantum - mechanical system ( atom , molecule , solid ) , the electronic density is typically _ smooth _ , so that eq . ( [ rho_delta ] ) seems in contradiction with the final requirement `` @xmath36 '' [ or the maximization ] . in other words
, it seems like there is a @xmath26-representability problem for smooth densities in the very @xmath15 limit .
the solution to this apparent problem is that the potential @xmath51 must be such that @xmath55 $ ] is a continuum with at least three dimensions .
that is , the absolute minimum of the @xmath56-dimensional function @xmath49 must be degenerate over ( at least ) a 3-dimensional subspace of @xmath57 , which we write as m=\{(,_2(), ... ,_n ( ) ) : p } [ msol ] , where @xmath58 is the region of space where @xmath59 . as we shall see in sec .
[ sec_fs ] , the ansatz can be consistent with the requirement that @xmath49 is minimum over @xmath50 only when the functions @xmath60 satisfy special properties . from the physical point of view , the corresponding spatial wavefunction @xmath40 ( which is zero everywhere except on @xmath50 ) , describes a state in which the position of one of the electrons can be freely chosen in @xmath61 , @xmath62 , but it then fixes the positions of all the other electrons through the functions @xmath60 , @xmath63 .
this is what we call @xcite `` strictly correlated electrons '' ( sce ) .
the requirement `` @xmath36 '' can of course be fulfilled also if the degeneracy of the absolute minimum of @xmath49 is higher , e.g. , if @xmath50 is a 6d subspace described by @xmath64 , which corresponds to the physical state in which the position of two electrons fixes the positions of all the others .
however , we have to keep in mind that we can affect @xmath50 only by varying the one - body potential @xmath51 .
physically , this means that @xmath51 must compensate the repulsive forces of the other @xmath65 electrons acting on the electron in @xmath31 . if the position of the other @xmath65 electrons only depends on @xmath31 [ as in eq .
] , then it may be possible to find such a @xmath51 .
if , instead , we require the degeneracy of @xmath50 to be higher , finding such a @xmath51 seems a daunting task .
we shall go through these physical considerations more in detail in sec .
[ sec_fs ] , where we analyze the solution and write down an explicit equation for @xmath51 . before doing so , we first show that we can consistently construct an antisymmetric wavefunction @xmath4 of the form such that @xmath44 is a distribution that is zero everywhere except on the minimizing set @xmath50 .
we want to construct an antisymmetric wavefunction of the form such that its spatial part @xmath46 is a distribution that is zero everywhere except in the set @xmath50 where @xmath49 has its global minimum . given a configuration @xmath48 of @xmath50
, we first notice that all its electronic positions @xmath66 must be different from each other : without the kinetic energy operator @xmath37 , two or more electrons on - top of each other would make @xmath49 infinite because of the singular nature of coulomb repulsion at short distances .
moreover , for any configuration @xmath67 , also all its @xmath68 permutations are @xmath67 , since @xmath49 is invariant under permutation of its variables @xmath66 . then , given the value of the wavefunction for one configuration @xmath47 , with spins @xmath69 , its value for the permutations @xmath70 of this configuration must be simply defined as @xmath71 where @xmath72 denotes the sign of the permutation @xmath70 .
because the @xmath66 in a given configuration are all different from each other , and @xmath40 is zero everywhere except on @xmath50 , the @xmath3 electrons are always localized in different regions of space that have exactly zero overlap .
this means that when we consider the square of the spatial wavefunction @xmath46 all the cross terms are exactly zero , and thus @xmath73 is completely independent of the choice of the spin eigenstate @xmath41 .
similarly , for repulsive bosons ( with any spin ) , @xmath4 can be _
symmetrized_. both the symmetric and the antisymmetric choice give the same spatial @xmath46 , and thus the same expectation for @xmath2 : when coulomb repulsion between the particles becomes dominant with respect to the kinetic energy , particle - particle overlap is suppressed , so that the particles no longer know wether they are fermions or bosons .
the same construction applies to the sce state in which @xmath50 has the form , since all the considerations made so far are obviously still valid for any configuration in @xmath50 characterized by a given @xmath74 .
we have seen that the potential @xmath51 of eq . for the @xmath15 limit of dft must be such that the minimum of the @xmath56-dimensional function @xmath49 of eq .
is degenerate over the 3-dimensional subspace @xmath50 of eq . .
we shall now analyze how to construct such a @xmath51 , and thus which properties the functions @xmath60 must satisfy , and how they can be determined from the density @xmath12 . since the solution corresponds to a state in which the position of one electron dictates the postions of all the others , we call the functions @xmath60 _ co - motion _ functions . if we assume that the @xmath51 we are looking for is a smooth potential with a continuous gradient @xmath75 , then @xmath76;\rv_1, ... ,\rv_n)$ ] is a continuous and smooth function of its variables @xmath77 , except for configurations with two or more electrons on top of each other , which , as explained in sec .
[ subsec_as ] , can not belong to the minimizing set @xmath50 . in this case
, a minimizing configuration @xmath47 must satisfy the stationarity conditions for the function @xmath76;\rv_1, ... ,\rv_n)$ ] , i.e. , \
{ l v(_1 ) = _
i1^n + v(_2 ) = _
i2^n + . [ gradients ]
if we insert the solution into eqs . we obtain \ { l v ( ) = _
i1^n + v()|_=_2 ( ) = _
i2^n + . [ gradientswithf ] where we have defined @xmath78 .
now , suppose that we have found @xmath65 functions @xmath79 and a potential @xmath51 such that the first of the eqs .
is satisfied for all @xmath80 .
if we now evaluate this first equation for @xmath81 and then put @xmath82 , we see that its left - hand - side coincides with the left - hand - side of the @xmath83 equation .
it is then easy to verify that , if we want also the right - hand - sides to be the same , the functions @xmath79 must satisfy the transformation properties \{_2(_n()), ... ,_n(_n())}= \{_1(), ... ,_n()}\\{_n ( ) } , [ propf ] i.e. , the set of @xmath65 functions @xmath84 with @xmath85 , when evaluated in @xmath86 must yield any permutation of the set of @xmath65 @xmath79 with @xmath87 and @xmath88 ( thus including @xmath89 ) .
this means that applying one co - motion function @xmath90 to one position @xmath77 of a given sce configuration @xmath91 must always yield a position of the same @xmath92 again . if the functions @xmath93 satisfy this property , then the fulfillment of the first of the eqs .
automatically implies the fufillment of the other @xmath65 equations , and thus the stationarity of the solution for any @xmath31 .
one has then to verify that such a stationary solution is the global minimum of the @xmath56-dimensional function @xmath94,\rv_1, ... ,\rv_n)$ ] .
once we have found some functions @xmath79 that satisfy the properties , eqs .
provide @xmath3 equivalent equations for the potential @xmath51 , v ( ) = _ i1^n .
[ eq_vsce ] equation has the clear physical meaning anticipated in sec .
[ sec_mr ] : the potential @xmath51 must compensate the net force acting on the electron in @xmath31 , resulting from the repulsion of the other @xmath65 electrons at positions @xmath79 .
we start now to see that the @xmath15 limit of dft is entirely characterized by the co - motion functions @xmath79 , which also give , as a byproduct via eq .
, the external potential @xmath51 . in the next subsec
. [ sec_ffromrho ] we shall see how to determine these functions from the density @xmath12 .
the sce problem for a given density @xmath12 reduces then to the construction of the appropriate co - motion functions @xmath60 . to do this
, we simply use the quantum mechanical meaning of the electronic density .
since in the sce state the position of the first electron determines the positions of all the others , the probability of finding the first electron in the volume element @xmath95 around the position @xmath31 must be the same of finding the @xmath83 electron in the volume element @xmath96 around the position @xmath60 .
this means that all the co - motion functions @xmath60 must satisfy the differential equation ( _ n ( ) ) d_n()= ( ) d , n=2, ... ,n .
[ detailedbalance ] in order to construct the co - motion functions we thus have to find the initial conditions for the integration of that ( i ) satisfy the properties , ( ii ) yield a smooth potential @xmath51 via eq .
, and ( iii ) give the minimum expectation of @xmath2 . in sec .
[ sec_nspherical ] we solve this problem explicitly for spherically symmetric @xmath3-electron densities .
equations can also be proven by explicitly constructing the square of the spatial wavefunction @xmath46 that is zero everywhere except on @xmath50 of eq . , and by imposing that the expectation value of the operator @xmath97 on @xmath46 is @xmath12 . in this way
, we also see that the value of @xmath98 in a given sce configuration @xmath67 must be set equal to @xmath99 . after all , maximising with respect to the potential @xmath26 in eq
. seems to be equivalent to constructing the co - motion functions @xmath60 . in terms of these functions , the expectation of the operator @xmath2 in the sce state [ see also eqs
. and ] reads w_+u[]=_i=1^n-1_j = i+1^n d , [ winfsce ] where , again , we have used the convention @xmath89 .
equation comes from the fact that , as said in the previous subsec .
[ sec_ffromrho ] , for a given @xmath31 , the configuration @xmath100 in @xmath50 has a weight equal to @xmath101 , and that the only possible electron - electron distances are @xmath102 , with @xmath103 , and @xmath104 .
as an example , we consider here @xmath3 electrons in a spherical density , and we explictly calculate the functions @xmath60 and the value @xmath23 $ ] for a few atoms . given the symmetry of the problem , we choose a spherical - coordinate reference system @xmath105 in which the position of the first electron defines the @xmath106-axis ( @xmath107 ) and the second electron is on the @xmath108 plane ( @xmath109 ) .
the @xmath110 relative angles @xmath111 are globally denoted by @xmath112 .
the total classical energy we want to minimize then reads e_pot([v],r_1, ...
,r_n,)= _ i=1^n v(r_i)+v(r_1, ...
since the external potential does not depend on the relative angles @xmath112 , we can decouple the stationarity equations into an angular part , which only involves @xmath113 , and a radial part that also involves @xmath114 . by solving the angular part for
any given @xmath115 we can define a function @xmath116 for the minimizing angles , valid for any spherically - symmetric @xmath26 .
these angles are the solution of the electrostatic equilibrium problem for @xmath3 neutral sticks of lenghts @xmath115 having the same point - charge @xmath117 glued at one end , and the other end fixed in the origin , in such a way that they are free to rotate in the 3d space .
we now insert the function @xmath116 in the stationarity equations for the radial variables @xmath118 , \ { l v(r_1)=- v(r_1,r_2, ... ,r_n , ) |_(r_1, ... ,r_n ) + v(r_2)=- v(r_1,r_2, ... ,r_n , ) |_(r_1, ... ,r_n ) + .
[ grad_radial ] as explained in the previous sec .
[ sec_fs ] , in order to have a smooth ( e.g. , atomic ) radial density @xmath119 we need to choose a @xmath114 that makes these equations fulfilled by a set of radial distances @xmath120 in which @xmath121 can take any value in the domain in which @xmath122 .
these radial co - motion functions @xmath123 must have the same transformation properties of eq .
, as it can be verified by inspection of eqs . . to construct the radial co - motion functions @xmath123 for a given density @xmath119 we proceed as follows
our starting point is eq . that becomes , in the case of spherical symmetry , 4 f_n(r)^2 ( f_n(r))|f_n(r)|dr=4 r^2 ( r)dr .
[ dbspherical ] we then define the function @xmath124 that gives the expected number of electrons between 0 and @xmath121 , ( r)=_0^r 4 x^2 ( x)dx . @xmath124 is a monotonously increasing function of @xmath121 , and its inverse @xmath125 is well defined for @xmath126 . by integrating both sides of eq .
we see that each function @xmath123 is of the form f_n(r)=\ { ll ^-1((f_n(0))(r ) ) & 0ra_1 + ^-1((f_n(a_1))(a_1 ) ( r ) ) & a_1ra_2 + . [ fgeneral ] where the upper ( lower ) sign corresponds to the case
@xmath127 ( @xmath128 ) , and @xmath129,\;[a_1,a_2], ... ,[a_{k-1},a_k]$ ] are the intervals in which ( i ) @xmath130 does not change sign and ( ii ) @xmath131 $ ] , so that @xmath132 is well defined .
the sce problem for spherical densities thus consists in finding the signs , the intervals @xmath133 $ ] , and the initial conditions @xmath134 in eqs .
that ( i ) make the @xmath123 satisfy the transformation properties , ( ii ) yield , via eq .
, a @xmath114 with a continuous first derivative , and ( iii ) give the minimum expectation value of @xmath2 when inserted in eq . .
we shall now see how to proceed in the case of atomic densities .
we first treat the case of the he atom , and we then generalize the solution to @xmath135 .
two electrons in a smooth spherical density in the @xmath15 limit have been already considered in ref . , where the sce state was proposed on the basis of physical considerations . here ,
we restate the problem in terms of the formalism just derived , and we show that the solution proposed in ref .
was indeed the correct one . when @xmath136 we have only one relative angle , @xmath137 . by solving the stationarity equations for the angular part we immediatly obtain @xmath138 or @xmath70
it is easy to verify that @xmath138 is a maximum and @xmath139 is a minimum for any value @xmath140 and @xmath141 . by inserting the minimizing angle @xmath139 in the stationarity equations for the radial variables , we obtain \ { l v(r_1)=(r_1+r_2)^-2 + v(r_2)=(r_1+r_2)^-2 .
these equations admit a solution of the kind @xmath142 if and only if @xmath143 satisfies the property @xmath144 , in agreement with eq . .
this means that , among all the possible @xmath143 of the form , we can only choose the ones for which @xmath145 .
one can verify that the only possible choices are then @xmath146 we call the choice the `` breathing '' solution : the two electrons are always at the same distance from the center , opposite to each other .
it is a stationary solution valid for any density , since @xmath147 satisfies the differential equation independently of the choice of @xmath119 .
the corresponding external potential is @xmath148 .
the hessian matrix shows that this stationary solution must be ruled out , since it is a maximum for @xmath76,r_1,r_2,\theta_2)$ ] .
the choice corresponds to a discontinuous @xmath149 , except in the case @xmath150 , i.e. , @xmath151 .
we rule out a discontinuous @xmath149 because otherwise our whole construction is inconsitent : if @xmath149 is not continuous it is not necesessary for a minimizing configuration of @xmath152 to satisfy the stationarity equations . as a double check
, we also verified , in the case of the he atom , that the choice @xmath150 in is indeed the one that yields the lowest expectation of @xmath2 , also when compared to the form that , again , corresponds to a discontinous @xmath149 .
the sce solution for @xmath136 electrons in a smooth spherical density is thus the one proposed in ref . ,
i.e. , f(r)=^-1(2-(r ) ) .
[ fsce2 ] it is easy to verify that this stationary solution , with its corresponding potential , v(r)=^r , is the true minimum of @xmath76,r_1,r_2,\theta_2)$ ] .
besides the mathematical arguments , the choice is the most `` physical '' one : it makes the two electrons always be in two different spherical shells , each one containing , on average in the quantum mechanical problem , one electron @xcite .
this is exactly what we expect for two electrons that repel each other infinitely strong , but that have to fulfill the constraint of yielding a given smooth , atomic - like density @xmath119 . when @xmath153 we have 3 angles , @xmath154 , but we immediately obtain @xmath155 : the three electrons must be on the same plane , containing the nucleus , to achieve compensation of the forces ( see sec .
[ sec_vsce ] ) .
we then find numerically , for any given @xmath156 , the minimizing angles @xmath137 and @xmath157 . to construct the radial co - motion functions @xmath158 and @xmath159 we should , in principle , try out all the possible @xmath123 of the form that satisfy the properties , and select the ones that yield a continous @xmath149 and the lowest expectation for @xmath2 .
instead , we use the physical idea that was behind the solution for two electrons of eq .
: we expect that the correct @xmath160 and @xmath161 are the ones that make the three electrons always occupy three different spherical shells , each containing , on average in the quantum mechanical problem , one electron .
such radial co - motion functions read @xmath162 where @xmath163 and @xmath164 .
they are displayed , for the li atom using an accurate fully correlated density @xcite , in fig .
[ fig_f2f3li ] . and @xmath159 for the li atom density .
hartree atomic units are used.,width=317 ] to better grasp the physics behind such solution , we show , in fig .
[ fig_lipositions ] , the positions of the three electrons in the sce state for the same accurate li atom density @xcite .
the two big circles are the radii of the spheres containing , on average in the quantum mechanical problem , one electron ( @xmath165 ) and two electrons ( @xmath166 ) .
the position @xmath121 of the first electron is varied along the vertical @xmath106 axis a ) from 0 to @xmath165 , b ) from @xmath165 to @xmath166 , and c ) from @xmath166 to @xmath167 .
the other two electrons have distances from the center given , respectively , by @xmath158 and @xmath159 , and angular positions given by the minimizing angles @xmath168 and @xmath169 . in each panel ,
the starting position of the three electrons is represented by a full circle ( @xmath170 ) and their final position by an empty circle ( @xmath171 ) .
the dashed curves represent the `` trajectories '' of the three electrons , and the arrows their direction .
thus , for each position @xmath121 on the @xmath106 axis of the first electron , the positions of the other two electrons are completely fixed , apart from the permutation symmetry of electron 2 with electron 3 .
we clearly see from this figure , that in any sce configuration the space is divided in three spherical shells , which never contain more than one electron .
these three spherical shells are the same that , in the quantum mechanical problem , contain _ on average _ one electron : the sce state makes them contain _ always exactly _ one electron , thus suppressing any accidental clustering ( `` correlation suppresses fluctuations '' @xcite ) . because of the properties of the co - motion functions ( which are a consequence of the fact that the electrons are indistinguishable ) , the three panels are equivalent .
our physical problem is in fact invariant for a rigid rotation of any of the configurations corresponding to a given @xmath121 .
this means that in panel b ) we could , for each value of @xmath121 , rigidly rotate the position of the three electrons in such a way that the electron in the inner shell moves on a straight line . in this way
, we would reobtain a rigid rotation of panel a ) . the same
can be done with panel c ) . thus , to calculate the expectation of @xmath2 we only need to consider the case @xmath172 . ) and 2 electrons ( radius @xmath166 ) .
the position @xmath121 of the first electron is varied along the vertical @xmath106 axis a ) from 0 to @xmath165 , b ) from @xmath165 to @xmath166 , and c ) from @xmath166 to @xmath167 .
the other two electrons have distances from the center given , respectively , by @xmath158 and @xmath159 of eqs . - , and angular positions given by the minimizing angles @xmath168 and @xmath169 .
in each panel , the starting position of the 3 electrons is represented by a full circle ( @xmath170 ) and their final position by an empty circle ( @xmath171 ) .
the dashed curves represent the `` trajectories '' of the 3 electrons , and the arrows their direction .
the three panels are actually equivalent ( see text ) .
distances are in atomic units.,title="fig:",width=302 ] ) and 2 electrons ( radius @xmath166 ) .
the position @xmath121 of the first electron is varied along the vertical @xmath106 axis a ) from 0 to @xmath165 , b ) from @xmath165 to @xmath166 , and c ) from @xmath166 to @xmath167 .
the other two electrons have distances from the center given , respectively , by @xmath158 and @xmath159 of eqs . - , and angular positions given by the minimizing angles @xmath168 and @xmath169 .
in each panel , the starting position of the 3 electrons is represented by a full circle ( @xmath170 ) and their final position by an empty circle ( @xmath171 ) .
the dashed curves represent the `` trajectories '' of the 3 electrons , and the arrows their direction .
the three panels are actually equivalent ( see text ) .
distances are in atomic units.,title="fig:",width=302 ] ) and 2 electrons ( radius @xmath166 ) .
the position @xmath121 of the first electron is varied along the vertical @xmath106 axis a ) from 0 to @xmath165 , b ) from @xmath165 to @xmath166 , and c ) from @xmath166 to @xmath167 .
the other two electrons have distances from the center given , respectively , by @xmath158 and @xmath159 of eqs . - , and angular positions given by the minimizing angles @xmath168 and @xmath169 .
in each panel , the starting position of the 3 electrons is represented by a full circle ( @xmath170 ) and their final position by an empty circle ( @xmath171 ) .
the dashed curves represent the `` trajectories '' of the 3 electrons , and the arrows their direction .
the three panels are actually equivalent ( see text ) .
distances are in atomic units.,title="fig:",width=302 ] the solution of eqs . -
is thus very reasonable from the physical point of view .
moreover , it can be easily verified that it satisfies the properties required by stationarity of eq . , and that it corresponds to a smooth external potential @xmath114 , which can be computed via v(r)=- v(r_1,r_2,r_3,_2,_3 ) |_r , f_2(r),f_3(r),_2(r),_3(r ) , [ potfromvee3 ] where @xmath173 is a shortened notation for @xmath174 . of eq .
for the sce state in the case of the li atom . the asymptotic expansion for large @xmath121 , @xmath175 , is also reported .
all quantities are in hartree atomic units.,width=294 ] the potential obtained by integrating eq .
( again using the accurate density for the li atom of ref .
@xcite ) is reported in fig .
[ fig_lipot ] . notice that @xmath114 is finite , with a zero first derivative , at @xmath176 : the electron - nucleus cusp in the density @xmath119 is compensated by the kinetic energy for any value of @xmath9 ( including the very limit @xmath15 ) , as discussed in ref . .
for large @xmath121 , we have @xmath177 , a necessary condition to include as a stationary configuration the one in which one of the @xmath3 electrons is at infinity .
the large-@xmath121 asympotic expansion @xmath178 is also shown in the same fig .
[ fig_lipot ] .
with this potential @xmath114 we have also verified that the stationary solution of eqs . -
is the absolute minimum of @xmath76,r_1,r_2,r_3,\theta_2,\theta_3,\phi_3)$ ] .
all these checks make us believe that eqs . - are the correct sce solution for the li atom : it is difficult to imagine a different solution that yields a lower value for the expectation of @xmath2 , since our solution is the one that makes the 3 electrons be always as far as possible from each other , without violating the constraint of yielding the smooth density @xmath119 . we can thus finally obtain the value @xmath179=-2.6030 $ ] hartree for the li atom , via @xmath180 + u[\rho]=\nonumber \\
\int_0^{a_1 } dr\,4\pi\,r^2\rho(r)\,v\ee(r , f_2(r ) , f_3(r),\theta_2(r),\theta_3(r ) ) , \label{winftyli}\end{aligned}\ ] ] where we have used the fact that , because the three electrons are indistinguishable , integrating from 0 to @xmath167 is the same as integrating three times from 0 to @xmath165 . as a final remark
, we comment briefly on the fact that one never obtains the configuration in which the three electrons are at the same distance from the nucleus , at the vertices of an equilateral triangle .
the reason is that the only @xmath158 and @xmath159 compatible with this configuration ( even for only one single value of @xmath121 , say @xmath181 ) that also satisfy the required properties are the ones corresponding to the `` breathing '' solution , @xmath182 , as it can be easily verified by integrating both sides of eq . with the initial condition @xmath183 .
this gives the two possibilites @xmath184 or @xmath185 : any choice for @xmath158 and @xmath159 which contains the function @xmath186 does not satisfy the properties .
this is a difference with the @xmath136 case in which , instead , the solution is @xmath187 with @xmath188 : two electrons in the sce state in a smooth density have one configuration in which they are both at the same distance from the nucleus , but for three electrons this does not happen .
one can also easiliy check that , again , the `` breathing '' solution , with its potential @xmath189 , is not a minimum for the corresponding @xmath76,r_1,r_2,r_3,\theta_2,\theta_3,\phi_3)$ ] .
we can at this point clearly see the difference between the sce state for a smooth atomic density ( fig .
[ fig_lipositions ] ) and the more familiar wigner - crystal - like state ( for a case with three electrons see , e.g. , ref . ) , in which the three electrons are localized at the vertices of a triangle at a certain distance from the nucleus , say @xmath181 . in this latter case ,
the density becomes very peaked around @xmath181 , loosing any resemblence with an atomic density , and becoming more and more similar to eq . .
we have now 5 relative angles to consider . as in the @xmath153 case
, we compute the minimizing angular function @xmath191 numerically .
we then construct the radial co - motion functions , @xmath158 , @xmath159 , @xmath192 , following the same ideas used for the case @xmath136 and 3 : we divide the space in four spherical shells , each containing , on average in the quantum mechanical problem , one electron . the radial co - motion functions that make the four electrons always be in four distinct shells are then @xmath193 where , again , @xmath194 .
these functions satisfy the transformation properties , and are reported in fig .
[ fig_fbe ] for the case of an accurate correlated density @xcite of the be atom . ,
@xmath159 and @xmath192 for the be atom density .
hartree atomic units are used.,width=317 ] we can then obtain , using the analogue of eq . for @xmath190 , the value @xmath23=-4.0212 $ ] hartree for the be atom . to check our result , we also computed the corresponding external potential @xmath114 , which , as shown in fig .
[ fig_bepot ] , is continuous and smooth .
we also verified that our stationary solution is the absolute minimum for the corresponding @xmath76,r_1,r_2,r_3,r_4,\omega)$ ] . of eq .
for the sce state in the case of the be atom .
the asymptotic expansion for large @xmath121 , @xmath175 , is also reported .
all quantities are in hartree atomic units.,width=294 ] interestingly , we found that , for the case of the be atomic density , the four electrons always lie on the same plane , containing the nucleus .
this effect seems to be due to the shell structure of the atomic density : by repeating the calculation using a simple exponential density , @xmath195 , we found that the four electrons are often non - complanar . instead , by using other test densities with a shell structure , we found again the four electrons always on the same plane .
the shell structure seems to force the four electrons to have distances from the nucleus that are distributed in a highly nonuniform way , so that forces compensation ( see sec .
[ sec_vsce ] ) can be achieved only if they are on the same plane . this way of constructing the solution can be generalized to any number of electrons @xmath3 .
the radial co - motion functions that make the @xmath3 electrons always occupy @xmath3 different shells containing , on average in the quantum mechanical problem , one electron , are as follows .
define an integer index @xmath196 running for odd @xmath3 from 1 to @xmath197 , and for even @xmath3 from 1 to @xmath198 .
then @xmath199 where @xmath194 . for odd @xmath3 ,
these equations give all the needed @xmath65 radial co - motion functions , while for even @xmath3 we have to add the last function , f_n(r)= ^-1(n-(r ) ) . using these radial co - motion functions , which satisfy the properties for any @xmath3 , we calculated the sce value @xmath179 $ ] for accurate sphericalized densities of the b and c atoms @xcite , and for the ne atom density @xcite .
the resulting radial co - motion functions are displayed in figs .
[ fig_fb]-[fig_fne ] . the results for @xmath179 $ ]
are reported in table [ tabwpc ] , and the corresponding potentials @xmath114 are shown in fig .
[ fig_bcnepot ] . in all cases
we find the @xmath3 electrons non - complanar .
of eq . for the sce state in the case of the ne atom and of the sphericalized b and c atoms .
all quantities are in hartree atomic units.,width=294 ]
in the absence of spherical symmetry there seems to be no general strategy for finding the initial conditions to integrate eqs . , although we are still investigating this problem .
so far , such cases can only be treated approximately , using instead of the exact expression the pc model for the functional @xmath23 $ ] .
the pc model has been tested successfully @xcite against the meta - gga functional of ref .
that is expected to be accurate in the limit @xmath15 .
the present exact ( sce ) results , in contrast , provide a rigorous test for the pc model . table [ tabwpc ] documents reasonable values from the pc model for all the atoms considered here .
a severe test is the trivial case of the hydrogen atom h , with the one - electron density @xmath200 , where @xmath201 for any value of @xmath9 and thus @xmath23=-\frac5{16}$ ] hartree . for the other atoms in table [ tabwpc ] ,
the pc error is less than 63 mh @xmath202 39 kcal / mole .
.comparison of the values @xmath23 $ ] in hartree atomic units obtained with the sce construction , and with the pc model @xcite .
the absolute errors of the pc model are also reported . [ cols="<,<,<,^",options="header " , ] for the he , be , and ne atoms , the error introduced by the pc model in @xmath203 $ ] is always less than 1 mh , showing that in the case of neutral atoms the spl - isi correlation functional is not too sensitive to the exact value of @xmath179 $ ] .
however , we have to keep in mind that neutral atoms are systems that resemble much more to the noninteracting ks system ( @xmath10 ) than to the sce state ( @xmath20 ) .
we expect that for more correlated systems ( e.g. , streched bonds ) the isi correlation energy is much more sensitive to the exact value of @xmath179 $ ]
. the investigation of such cases will be the object of future work .
we have reformulated the strong - interaction limit of density functional theory in terms of a classical problem with a degenerate minimum , obtaining a consistent solution for this limit . even if an antisymmetric wavefunction can be explicitly constructed , the strong - interaction limit of dft is entirely characterized by much simpler quantities , the so called `` co - motion '' functions , which are related to the electronic density @xmath12 via the differential equation .
the results of this work can be useful to construct interpolations between the weak- and the strong - interaction limits of dft , and to test other approximate functionals in the @xmath15 limit @xcite .
future work will be mainly devoted to the calculation of this limit for non - spherical densities , and to study the generalization to the next leading term , @xmath18 $ ] .
the calculation of the intracule density ( the probability distribution for the electron - electron distance ) will be also carried out in connection with the correlation energy functional constructed in ref . .
we thank prof . c. bunge for the fully correlated density of the li atom , prof .
c.j umrigar for the vmc densities of the ne and be atoms , and dr .
j. toulouse for the sphericalized densities of the b and c atoms .
one of the authors ( p.g .- g . ) gratefully acknowledges prof .
baerends for the warm hospitality in his group , where part of this work was done .
it is instructive to see how the sce solution changes if we replace the coulomb repulsion with an harmonic interaction .
since in this case we have the exact solution , we can also clarify how the sce construction presented in this paper gives the @xmath15 limit of dft .
we thus consider the exactly solvable 2-electron hamiltonian @xcite ^=-(^2_1+^2_2 ) + ( r_1 ^ 2+r_2 ^ 2)-|_1-_2|^2 , and we analyze the case in which @xmath15 and the density is kept fixed , equal to the one at @xmath13 , by means of a suitable @xmath204
. the ground - state one - electron density of the hamiltonian @xmath205 is given by ( r)=e^-r^2 , = , so that @xmath204 must keep @xmath206 independent of @xmath9 and thus satisfies the equation = , where @xmath207 in order to have a bound system .
we easily find k_=2 + + o(^-1/2 ) .
[ eq_kinf ] the square of the spatial wavefunction is equal to @xmath208^{3/2 } \times \nonumber \\
e^{-\frac{1}{2}(\sqrt{k_\alpha}+\sqrt{k_\alpha-2\alpha } ) ( r_1 ^ 2+r_2 ^ 2)- ( \sqrt{k_\alpha}-\sqrt{k_\alpha-2\alpha})\rv_1 \cdot\rv_2}. \label{eq_psiexact}\end{aligned}\ ] ] now , we want to compare the exact result with the sce construction .
we thus follow the same steps done for the case of the he atom in subsec .
[ sub_he ] .
again , we find that the minimizing angle is @xmath209 and that the radial co - motion function must satisfy the property @xmath144 , so that the only possible choices are the ones of eqs . - . in this case , however , the `` breathing '' solution of eq . , @xmath147 , is the one that yields the minimum expectation of `` @xmath2 '' , and that is the global minimum of the corresponding @xmath49 .
the very different nature of harmonic forces with respect to coulomb repulsion makes the solution completely different also from a qualitative point of view .
the `` breathing '' solution , that was a maximum for coulomb forces , becomes the true minimum in the case of harmonic forces . in this latter case , in fact
, the interaction is not singular at zero electron - electron distance so that the two electrons can get on top of each other ( and they do , at the `` nucleus '' ) .
this means that in this case even in the @xmath15 limit the spin state can still play a role .
the sce one - body potential is obtained by v(r)=r+f(r)=2 r , so that @xmath210 agrees with the leading term ( @xmath29 ) of eq .
( remember that in the sce construction we directly consider all the quantities divided by @xmath9 ) . in order to compare the sce solution with the exact wavefunction of eq . in the @xmath15 limit
, we have first to keep in mind that eq .
is the solution of the hamiltonian @xmath205 , so that it includes also the kinetic energy operator @xmath37 .
equation shows that the external potential @xmath211 that keeps the density fixed has the large-@xmath9 expansion @xmath212 .
the sce construction only gives @xmath213 , but the exact solution of eq .
always includes also the next term , @xmath214 , because it is of the same order of @xmath37 .
the comparison with the sce solution can be done in the following way .
replace in eq .
the large-@xmath9 expansion of @xmath204 up to orders @xmath215 , @xmath216 .
we can rearrange the result , with @xmath217 , @xmath218 when @xmath15 , the first gaussian in eq
. with its factor in front tends to the distribution @xmath219 .
the second gaussian gives @xmath220 .
thus , as @xmath15 we have _ |_(_1,_2)|^2= ( _ 1+_2 ) , which is exactly the sce solution , since in the case of harmonic forces @xmath221 . 99 w. kohn , rev .
phys . * 71 * , 1253 ( 1999 ) .
mattsson , science * 298 * , 759 ( 2002 ) .
c. fiolhais , f. nogueira , and m. marques ( eds . ) , _ a primer in density functional theory _ ( springer - verlag , berlin , 2003 ) .
wigner , phys . rev . * 46 * , 1002 ( 1934 ) ; trans .
faraday soc .
* 34 * , 678 ( 1938 ) .
m. seidl , j. p. perdew , m. levy , phys .
rev . a * 59 * , 51 ( 1999 ) .
m. seidl , j. p. perdew , s. kurth , phys .
lett . * 84 * , 5070 ( 2000 ) .
m. seidl , phys .
a * 60 * , 4387 ( 1999 ) .
m. seidl , j. p. perdew , s. kurth , phys .
a * 62 * , 012502 ( 2000 ) ; * 72 * , 029904 ( 2005 ) ( e ) .
magyar , w. terilla , and k. burke , j. chem .
phys . * 119 * , 696 ( 2003 ) .
t. gimon , a. grling , m. seidl , and j. p. perdew , in preparation .
a. grling and m. levy , phys .
b * 47 * , 13 105 ( 1993 ) ; phys . rev .
a * 52 * , 4493 ( 1995 ) .
e. lieb , int .
j. quantum chem . * 24 * , 243 ( 1983 ) . c. bunge , private communication .
the full ci density of the li atom has been computed with a very large basis set with 8 @xmath222 functions and up to @xmath196 functions .
see , e.g. , p. fulde , _ electron correlation in molecules and solids _ ( springer , berlin , 1995 ) . p. ziesche , j. tao , m. seidl , and j. p. perdew , int . j. quantum chem . * 77 * , 819 ( 2000 ) .
. nagy and zs .
jnosfalvi , philos . mag . * 86 * , 2101 ( 2006 ) .
j. cioslowski and k. pernal , j. chem
. phys . * 125 * , 064106 ( 2006 ) . c. filippi , x. gonze , and c.j .
umrigar , in _ recent developments and applications in modern dft _ , edited by j.m .
seminario , ( elsevier , 1996 ) .
j. toulouse , private communication .
the sphericalized densities of the b and c atoms were obtained from variational monte carlo using accurate optimized wavefunctions ( see j. toulouse and c.j .
umrigar , submitted to j. chem .
c. j. umrigar , j. toulouse , c. filippi , s. sorella , and r.g .
hennig , cond - mat/0611094 ) .
a. i. al - sharif , r. resta , and c. j. umrigar , phys .
a * 57 * , 2466 ( 1998 ) .
j. p. perdew , s. kurth , a. zupan , and p. blaha , phys .
lett . * 82 * , 2544 ( 1999 ) ; * 82 * , 5179 ( 1999 ) ( e ) .
m. ernzerhof , chem .
. lett . * 263 * , 499 ( 1996 ) ; private communication . v.
n. staroverov , g. e. scuseria , j. tao , and j. p. perdew , phys .
b * 69 * , 075102 ( 2004 ) ; j. jung , p. garcia - gonzalez , j. e. alvarellos , and r. w. godby , phys . rev .
a * 69 * , 052501 ( 2004 ) .
gori - giorgi and a. savin , phys .
a * 71 * , 032513 ( 2005 ) ; philos . mag . * 86 * , 2643 ( 2006 ) ; cond - mat/0611324 .
see , e.g. , e.r .
davidson , _ reduced density matrices in quantum chemistry _ ( academic press , new york , 1976 ) .
|
we reformulate the strong - interaction limit of electronic density functional theory in terms of a classical problem with a degenerate minimum .
this allows us to clarify many aspects of this limit , and to write a general solution , which is explicitly calculated for spherical densities .
we then compare our results with previous approximate solutions and discuss the implications for density functional theory .
|
in the standard picture of galactic structure , most galaxies harbour a central super - massive black hole ( smbh ) , which is quiet in normal galaxies and responsible for quasar and seyfert activities in active galaxies .
but the building process of smbhs is still not fully understood .
is the way of creating a smbh with more than @xmath1 a direct collapse of a gas cloud , is it the growth of a small seed black hole in a dense star cluster , or is it the remnant of a hypernova of a super massive star like object ( sms ) ? in the building process of a smbh two basic paths can be distinguished ( rees @xcite ) . at the stellar dynamical path
the gas cloud initially forms a massive central star cluster or bulge .
stellar and dynamical evolution including star - star collisions than form a seed black hole with initial mass up to @xmath2 .
subsequently this black hole grows by accretion of gas and stars . here
the basic problem of the growth is the feeding rate .
only a very small fraction of stars have highly excentric orbits to reach the tidal radius of the black hole and refilling this loss - cone is a complicated matter ( for details see amaro - seoane & spurzem @xcite and references therein ) .
the gas dynamical path starts also with a collapsing gas cloud , but here the cloud collapses directly to a highly condensed gas ball .
the lifetime of this super - massive object is very short ( see section [ sms1 ] ) such that star formation in this condensed phase does not occur .
the dynamical collapse can go on to end up with a smbh , or it can reach a hydrostatic state , which than contracts by radiative cooling .
we call a gas ball with mass @xmath3 in hydrostatic equilibrium a super - massive star ( sms ) .
it is not proofed that a direct collapse is possible , because explosive hydrogen burning would destroy more or less the whole object .
even the final collapse of a sms , which becomes unstable due to the post - newtonian instability , will be stopped by hydrogen burning ( fuller @xcite ) . in a more realistic scenario of galaxy formation , a central bulge or dense star cluster
will form first , and at a later stage there will occur strong gas infall initiated by a galaxy merging process or a bar instability . in this case
the super - massive gaseous object will not be isolated and interaction processes with the star cluster can change the evolution dramatically .
hara ( @xcite ) showed in an analytic approximation , that heating of the sms by drag forces of the stars crossing the sms can stop the contraction for some time . in a numerical simulation langbein ( @xcite ) showed that the energy input due to star - star collisions can dominate over heating by drag forces and leads to a rebounce of the sms . in the next section
we give an overview over the properties of a sms and in section [ sms2 ] we discuss the interaction processes of the sms with the surrounding star cluster in a more complete way .
in a sms radiation pressure @xmath4 dominates strongly over thermal pressure @xmath5 , the fraction of which is @xmath6 where @xmath7 is the mass in units of @xmath8 .
@xmath9 is independent of radius ( @xmath10 given in @xmath11 ) and does not change during contraction of the sms . for @xmath12
the structure of the sms is very near to a n=3 polytrope , which is the stability limit with zero total energy .
the sms is determined by mass @xmath13 and radius @xmath14 .
the central temperature @xmath15 , central density @xmath16 , gravitational and total energy ( @xmath17 and @xmath18 ) are given by @xmath19 respectively .
the contraction rate of the sms is determined by the eddington luminosity @xmath20 leading to a contraction time @xmath21 in parameter space ( the @xmath22-plane ) , the regime of hydrostatic , stable equilibrium ( where a sms can exist ) is bounded at the high radius end ( at @xmath23 ) where the kelvin - helmholtz time @xmath24 ( contraction time ) equals the free fall time . at the low radius end
it is bounded by the post - newtonian instability ( @xmath25 ) or for lower masses when nuclear burning at the centre sets in ( @xmath26 ) . in figure [ fig1 ] the relevant regime for a sms
the schwarzschild radius @xmath27 is far below the living region of a sms .
the evolutionary tracks are horizontal lines from right to left .
the contraction time is also plotted showing that without re - heating the lifetime of the sms is at most @xmath28 before nuclear burning or the post - newtonian collapse sets in .
in the central region of young galaxies with sms both , stellar and gas densities , are very high exceeding @xmath29 . therefore the interaction rate of stars is very high and drag forces of loss - cone stars ( this are the stars crossing the sms ) can drive the evolution considerably . earlier investigations
( hara @xcite , langbein @xcite ) show that the timescale for dynamical coupling of the stellar system with the gas and for heating by stellar collisions is comparable or even shorter than the lifetime of a sms .
here we concentrate on the star - gas interaction and discuss the effects on the structure and evolution of the sms .
stars moving through the gas produce density and velocity perturbations by gravitational forces and the geometrical cross section .
the consequences of these local perturbations on the global behaviour of the sms can be consistently estimated using a quasi - linear perturbation theory . in this method
, all physical quantities are decomposed into a local mean value and a perturbation ( @xmath30 for the gas density e.g. ) .
then the time dependent perturbations are computed from the set of linearised equations and finally the quasistationary fluctuation field of the nonlinear quantities as functions of the mean quantities are derived by taking the statistical average over all stars .
dynamical friction between stars and gas is given by @xmath31 with gravitational potential @xmath32 ( see just @xcite ) .
the level of approximation is similar to that of using chandrasekhars dynamical friction formula .
the advantage of the fluctuation theory is that the turbulent velocity field , corrections to the equation of state and to radiation transfer can all be derived consistently .
it is also straightforward to include all relevant physical processes in the gas like radiation pressure , heating and cooling processes , viscosity or even magnetic fields .
the structure of the sms is determined by hydrostatic equilibrium and a stationary radiation field in the diffusion limit .
these assumptions are not automatically valid for the perturbations , because here the length and timescales are much shorter . the perturbations are time dependent and it depends on the scale , whether the radiation field will be compressed also .
the character of the perturbations will be between the two extremes : isothermal : : in small scale fluctuations the inhomogeneities of the radiation field diffuse quickly away . since the temperature is determined by the radiation ,
the compression of the gas is isothermal and the sound speed relevant for determining the mach number of the stellar motion is the usual isothermal sound speed ( @xmath33 for a million solar mass sms with radius @xmath34 ) .
most stars are highly supersonic in this case .
adiabatic : : larger fluctuations , where the radiation field is also compressed , are adiabatic .
the mach number is a factor of @xmath35 smaller than in the isothermal case .
it is determined by the total pressure leading to typical mach numbers of order unity for the stars .
the resultant density and velocity fluctuations of the gas and the variations in the radiation field change the mean local properties of the sms systematically . in the sms basically three different effects of the fluctuations can be important : dynamical heating : : the energy loss of the stars by dynamical friction heats the sms in the inner region .
the heating rate depends on the relative masses and densities of the star cluster and the sms , and of the filling factor of the loss cone , which is a complicated matter to estimate .
if the heating rate exceeds the cooling of the sms , then the contraction can be stalled for a while .
mass segregation : : dynamical friction leads also to mass segregation of the stars .
loss cone stars become confined to the sms .
this happens , if the friction timescale @xmath36 becomes smaller than the contraction time @xmath24 . in figure [ fig1 ]
an estimation of @xmath36 is shown and also the coupling radius @xmath37 of the sms for stars with an initial velocity of 5000 km / s . in more compact sms loss - cone stars
are effectively stopped by dynamical friction . since dynamical friction
has its maximum just above the sound speed , confined stars will be decelerated quickly below the sound speed building a new highly condensed stellar core
. then the question is , whether this new core can decouple from the surrounding to build a new smaller sms - star cluster system with higher density .
equation of state : : the turbulent velocity field and non - isothermal density fluctuations change the relative contribution of radiation pressure .
this results in a different mean @xmath38 of the form @xmath39 which stabilises the sms
. additionally the direct influence of the inhomogeneities on radiation transfer may generally destroy the @xmath0 behaviour of the sms .
then the post - newtonian instability can not occur . as a bottom line we find , that the interaction of the stellar component and the sms can give rise to a secular evolution away from the unstable zero - energy state to a more stable structure
. additionally mass segregation may lead to a decoupling of the core region resulting in a sequence of sms , the smaller and more compact ones nested in the larger ones . in order to show , whether this kind of evolution can happen , we will do detailed semi - analytical model calculations .
99 amaro - seoane , p. & spurzem , r. 2001 , mnras , 327 , 995 fuller , g.m . , woosley , s.e . , weaver , t.a . ,
apj , 307 , 675 hara , t. 1978 , progress of th .
physics , 60 , 711 just , a. , kegel , w.h . & deiss , b.m .
1986 , a&a , 164 , 337 langbein , t. , spurzem , r. , fricke , k.j . ,
yorke , h.w .
1990 , a&a , 227 , 333 rees , m.j . 1984 , ara&a , 22 , 471
|
highly condensed gaseous objects with masses larger than are called _ super - massive stars_. they are thought to be possible precursors of super - massive black holes in the centres of galaxies . in the quasistationary contraction phase , the hydrostatic equilibrium is determined by radiation pressure and gravitation .
the global structure is that of an @xmath0 polytrope at the stability limit .
small relativistic corrections for example can initiate a free fall collapse due to the post newtonian instability . since the outcome of the final collapse a super - massive black hole or hypernova depends sensitively on the structure and the size of the object ,
when the instability sets in , it is important to investigate in more detail the contraction phase of the sms .
if the gaseous object is embedded in a dense stellar system , the central star cluster , the interaction and coupling of both components due to dynamical friction changes the energy balance and evolution of the sms dramatically . dynamical friction between stars and gas , which can be estimated semi - analytically ( see just @xcite ) , has three different effects on the two - component system : 1 ) the gas is heated by decelerating the stars .
this may stall the contraction process for a while until the stars in the loss cone , these which cross the sms , lost their kinetic energy ( for the total heating rate see amaro - seoane & spurzem @xcite ) .
2 ) this cooling of the loss cone stars lead to a mass segregation in the stellar component resulting in a much more condensed central stellar core .
3 ) the inhomogeneities due to the gravitational wakes in the gas changes the effective absorption coefficient of the gas . this affects the condition for hydrostatic equilibrium and may give essential deviations from the @xmath0 polytrope .
we discuss in which evolutionary stages and parameter range these interaction processes are relevant and how they can influence the stability and evolution of the sms .
|
this paper is motivated by our endeavour to better understand isotropy conditions in higher gradient elasticity theories .
therefore , we make use of the old idea expressed by truesdell ( * ? ? ?
* lecture 6 ) albeit including higher gradient continua .
rotations can not be detected by any experiment .
the response of the material with respect to the reference configuration @xmath1 is the same as that with respect to any other obtained from it by rotation . '' ] to this end we reconsider frame - indifference and isotropy first in nonlinear elasticity and then in classical linear elasticity theories .
these investigations are extended to linear higher gradient elasticity with a focus on the indeterminate couple stress theory .
we do not discuss general material symmetry conditions a la noll s group theoretic framework , as e.g. presented in @xcite . in nonlinear elasticity ,
isotropy requirements are fundamentally different from frame - indifference requirements .
indeed , isotropy rotates the referential coordinate system , while frame - indifference rotates the spatial coordinate system . both spaces , referential and spatial , are clearly independent of each other and the corresponding rotations of frames are not connected to each other . in this paper , however
, we will later apply the * same rotation * to the referential and spatial frame simultaneously .
it is clear that this is not equivalent to frame - indifference or isotropy as we show with simple explicit examples .
what , then , is the use of such a specific transformation ( which we will call the @xmath2-transformation in section [ kaprotinvariance ] ) .
indeed , once we consider reduced energy expressions which already encode frame - indifference , then applying the @xmath2-transformation and requiring form - invariance of the energy under this transformation is equivalent to isotropy .
surprisingly , the @xmath2-transformation remains operative in exactly the same way when applying it to the geometrically linear context .
it is this insight that we follow when discussing isotropy conditions for higher gradient continua .
we do not believe that our development leads to new results for the representation of isotropic formulations , but we believe that there is a conceptual gain of understanding when working directly with transformation connected to finite rotations . in this respect
we remind the reader that for linearized frame - indifference it would be sufficient to work with the lie - algebra of skew - symmetric rotations instead of the orthogonal group .
this is , however , not possible with respect to isotropy considerations .
+ further , we give a contribution to the discussion in @xcite , where a fundamental aspect of strain gradient elasticity as proposed by mindlin @xcite is doubted .
there we read : `` however , the approach described by equation ( 1 ) includes a serious flaw : .
here , @xmath3 is the second - order couple stress tensor and @xmath4 represents the skew - symmetric part of the total force stress tensor @xmath5 .
the equation is given for the planar situation , see also eq .. ] couple stresses are components of a tensor which has a rank higher than two . in conclusion
, the torque arising from an asymmetric stress tensor can not consequently be compensated by the torque associated with a tensor of higher rank .
in fact , there are different transformation rules for tensors of different rank during rotation of a coordinate system .
therefore , couple stress theory does not consequently fulfill the requirement of invariance with respect to rotation of the coordinate system . in spite of this deficiency
, strain gradient theory still contains important aspects which are corrected . ''
motivated by this statement we show to the contrary that all proper invariance requirements are satisfied .
+ the paper is now structured as follows .
after notational agreements we recall the gradient of continuum rotation , which is the curvature measure in linear couple stress - elasticity .
identities from the scalar triple product built a basis to study transformation rules for this curvature measure later in the text . in section [ kapobjnonlinearelas ]
we start our investigation in the context of nonlinear hyperelasticity by discussing frame - indifference and isotropy for first and second gradient continua .
then we consider the simultaneous transformation of the deformation to new spatial and referential coordinates .
the consequence of this transformation in linearized elasticity follows in section [ kapobjlinearelas ] .
further , we prepare for higher gradient elasticity by giving details on transformation rules and apply them to the linear momentum balance equations of several models in section [ kaprotinvariance ] .
we then define form - invariance in higher gradient elasticity in section [ kapobjectiso ] and specify our result for several couple stress theories in section [ kapindeterminate ] .
finally , we conclude and give an outlook .
by @xmath6 we denote the set of real @xmath7 second order tensors , written with capital letters .
vectors in @xmath8 are denoted by small letters .
additionally , tensors in @xmath9 are necessary for our discussion , where significant symbols like @xmath10 , @xmath11 , and @xmath12 will be used .
the components of vectors and tensors are given according to orthogonal unit vectors @xmath13 , which may be rotated by @xmath14 . throughout this paper latin subscripts
specify the direction of components in index notation and take the values @xmath15 . for repeating subscripts einstein
s summation convention applies .
we adopt the usual abbreviations of lie - algebra theory , i.e. , @xmath16 is the lie - algebra of skew symmetric tensors and @xmath17 is the lie - algebra of traceless tensors . for all
@xmath18 we set @xmath19 , @xmath20 and the deviatoric part @xmath21 and we have the _ orthogonal cartan - decomposition of the lie - algebra _
@xmath22 @xmath23 simply allowing to split every second order tensor uniquely into its trace - free symmetric part , skew - symmetric part and spherical part , respectively . for @xmath24
\ , , \qquad \overline{{{{a } } } } = \operatorname{anti}[{{{a } } } ] \ , , \end{aligned}\ ] ] the functions @xmath25 and @xmath26 are given by @xmath27 \colonequals -\frac{1}{2 } \
, \overline{a}_{ij } \ ,
\epsilon_{ijk } \ , { { { e}}}_k \ , , \quad \quad \operatorname{anti}[{{{a } } } ] : = - \epsilon_{ijk } \ , a_k \ , { { { e}}}_i \otimes { { { e}}}_j = \overline{a}_{ij } \ , { { { e}}}_i \otimes { { { e}}}_j \ , , \end{aligned}\ ] ] with @xmath28 for even permutation , @xmath29 for odd permutation , and @xmath30 else .
note that the skew - symmetric part of a tensor @xmath31 can be written by the combination of permutations via @xmath32 from eq .
and eq.@xmath33 one obtains @xmath34 = - \12 \ , \underbrace{\epsilon_{ijk } \ , \12 \ , \epsilon_{nij}}_{\displaystyle \delta_{nk }
} \ , \epsilon_{nab } \ , x_{ab } \ , { { { e}}}_k = - \12 \ , x_{ab } \ , \epsilon_{abk } \ , { { { e}}}_k \,.\end{aligned}\ ] ] in index notation the typical conventions for differential operations are implied such as comma followed by a subscript to denote the partial derivative with respect to the corresponding coordinate .
the gradient of a scalar field @xmath35 and the gradient of a vector field @xmath36 are given by @xmath37 } \colonequals \frac{\partial \
, \phi}{\partial \ , { { { x } } } }
= \frac{\partial \ ,
\phi}{\partial \ ,
x_i } \ , { { { e}}}_i = \phi_{,i } \ , { { { e}}}_i \quad \in \operatorname{\mathbb{r}^3}\,,\end{aligned}\ ] ] @xmath38 } \colonequals \frac{\partial \ , { { { b}}}}{\partial \ , { { { x } } } } = \frac{\partial \ , b_i}{\partial \
, x_j } \ , { { { e}}}_i \otimes { { { e}}}_j = b_{i , j } \ , { { { e}}}_i \otimes { { { e}}}_j \quad \in \operatorname{\mathbb{r}}^{3 \times 3 } \,.\end{aligned}\ ] ] similarly , we also define the gradient of a tensor field @xmath18 : @xmath39 } \colonequals \frac{\partial \ , { { { x}}}}{\partial \ , { { { x } } } } = \frac{\partial \ , x_{ij}}{\partial \ , x_k } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k = x_{ij , k } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \quad \in \operatorname{\mathbb{r}}^{3 \times 3 \times 3 } \,.\end{aligned}\ ] ] for vectors @xmath40 we let @xmath41 denote the inner and outer product on @xmath42 with associated vector norm @xmath43 .
the standard euclidean inner product on @xmath6 is given by @xmath44 and thus the frobenius tensor norm is @xmath45 .
similarly , on @xmath46 we consider the inner product @xmath47 and the tensor norm @xmath48 .
the identity tensor on @xmath49 will be denoted by @xmath50 , so that @xmath51 and @xmath52 .
further , we will make repeated use of the identity @xmath53 = { \langle{{{x } } } \ , { { { y } } } \ , { { { z } } } , { { { \mathbbm{1}}}}\rangle } = { \langle{{{y } } } , { { { x}}}^t \ , { { { z}}}\rangle}$ ] . the divergence of a vector field @xmath54 reads : @xmath55 } ) = \langle { { \rm grad}_{{{{x}}}}[{{{{v } } } } ] } , { { { \mathbbm{1}}}}\rangle = v_{i , j } \ , \delta_{ij } = v_{i , i } \quad \in \operatorname{\mathbb{r}}\ , .\end{aligned}\ ] ] note that we do not introduce symbolic notation for the double contraction or more complicated contractions for higher order tensors , since this may be confusing and also limited for certain cases . instead , we use generally index notation for contractions being not defined by the scalar product .
thus , the divergence of a second order tensor is given by @xmath56})_{ijk } \
, \delta_{jk } \ , { { { e}}}_i = y_{ij , j } \ , { { { e}}}_i \quad \in \operatorname{\mathbb{r}^3}\ , .\end{aligned}\ ] ] similarly , the divergence of a third order tensor reads @xmath57 the curl of a vector @xmath54 is given by @xmath58 similarly , the curl of a second order tensor field @xmath59 is defined by @xmath60 using eq .
the curl of a vector can be written in terms of a gradient via @xmath61 } ) & = - \12 \ , ( { { \rm grad}_{{{{x}}}}[{{{{v}}}}]})_{ab } \ , \epsilon_{abk } \ , { { { e}}}_k = \12 \,{{\rm curl}_{{{{x}}}}}{{{v } } } \,.\end{aligned}\ ] ] in this work we consider a body which occupies a bounded open set @xmath62 of the three - dimensional euclidian space @xmath42 and assume that its boundary @xmath63 is a piecewise smooth surface .
an elastic material fills the domain @xmath64 and
we refer the motion of the body to the displacement field @xmath65 shifting any point @xmath66 of the reference configuration to the actual configuration @xmath67 .
+ since the gradient of a scalar @xmath68 is curl - free @xmath69 } & = { { \rm curl}_{{{{x}}}}}[\phi_{,i } \ , { { { e}}}_i ] = - \phi_{,ij } \ , \epsilon_{ijk } \ , { { { e}}}_k \notag \\ & = -(\phi_{,12 } - \phi_{,21 } ) \ , { { { e}}}_3 - ( \phi_{,23 } - \phi_{,32 } ) \ , { { { e}}}_1 - ( \phi_{,31 } - \phi_{,13 } ) \ , { { { e}}}_2 = 0 \,,\end{aligned}\ ] ] the definition in eq .
is such that the gradient of a vector field @xmath70 is also curl - free : @xmath71 } & = { { \rm curl}_{{{{x } } } } } [ u_{i , j } \ , { { { e}}}_i \otimes { { { e}}}_j ] = - u_{i , jb } \ , \epsilon_{jbk } \ , { { { e}}}_i \otimes { { { e}}}_k \notag \\ & = - ( u_{i,12 } - u_{i,21 } ) \ , { { { e}}}_i \otimes { { { e}}}_3 - ( u_{i,23 } - u_{i,32 } ) \ , { { { e}}}_i \otimes { { { e}}}_1 - ( u_{i,31 } - u_{i,13 } ) \ , { { { e}}}_i \otimes { { { e}}}_2 = 0 \ , .\end{aligned}\ ] ] however , the skew - symmetric part of the gradient of a vector field @xmath70 is generally * not * curl - free : @xmath72 } & = { { \rm curl}_{{{{x}}}}}(\12 \ , \epsilon_{nij } \ , \epsilon_{nab } \ , u_{a , b } \ , { { { e}}}_i \otimes { { { e}}}_j ) = -\12 \ , \epsilon_{nij } \ , \epsilon_{nab } \ , u_{a , bk } \ , \epsilon_{jkp } \ , { { { e}}}_i \otimes { { { e}}}_p \notag \\ & = - \12 \ , ( \delta_{kn } \ , \delta_{ip } - \delta_{ki } \ , \delta_{pn } ) \ , \epsilon_{nab } \ , u_{a , bk } \ , { { { e}}}_i \otimes { { { e}}}_p = - \12 ( u_{a , bn } \ , \epsilon_{abn } \ , \delta_{ip } - u_{a , bi } \ ,
\epsilon_{abp } ) \ , { { { e}}}_i \otimes { { { e}}}_p \notag \\ & = - \12 \underbrace{(u_{1,23 } - u_{1,32 } + u_{2,31 } - u_{2,13 } + u_{3,12 } - u_{3,21})}_{\displaystyle \operatorname{tr}{{\rm curl}_{{{{x}}}}}{{\rm grad}_{{{{x}}}}[{{{{u } } } } ] } = 0 } \ , { { { \mathbbm{1}}}}+ \frac{\partial}{\partial \ , x_i } ( \underbrace{\12 \ , u_{a , b } \ , \epsilon_{abp}}_{\displaystyle ( - \operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{u}}}}]})_p } ) \ , { { { e}}}_i \otimes { { { e}}}_p \notag \\ & = - ( { { \rm grad}_{{{{x}}}}[{\operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{u}}}}]}}]})_{pi } \ , { { { e}}}_i \otimes { { { e}}}_p \notag \\ & = - ( { { \rm grad}_{{{{x}}}}[{\operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{u}}}}]}}]})^t = - \12 \ , ( { { \rm grad}_{{{{x}}}}[{{{\rm curl}_{{{{x}}}}}{{{u}}}}]})^t \ , .\end{aligned}\ ] ] in the indeterminate couple stress model the gradient of the continuum rotation @xmath73 defines the curvature measure @xmath74 } = - \12 \ , u_{a , bj } \ ,
\epsilon_{abi } \ , { { { e}}}_i \otimes { { { e}}}_j \ , .\end{aligned}\ ] ] from eq . and
it follows that @xmath75 } } ] } = - ( { { \rm curl}_{{{{x}}}}}\skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]})^t \ , .\end{aligned}\ ] ] further , it has been shown in @xcite ( see also @xcite ) that @xmath76 can be also written in terms of @xmath77})^t \ , .\end{aligned}\ ] ] the representations of @xmath76 with the skew - symmetric gradient in eq . on the one hand , and the symmetric gradient in eq . on
the other hand seems amazing . for the proof of relation we use eq .
, , and @xmath78 to obtain the above statement : @xmath79}}_{\displaystyle = \ , 0 } & = { { \rm curl}_{{{{x}}}}}\ { \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } + \skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } \notag \\ \leftrightarrow \ , 0 & = { { \rm curl}_{{{{x}}}}}\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } + \underbrace{{{\rm curl}_{{{{x}}}}}\skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}}_{\displaystyle = - ( { \widetilde{{{{k}}}}({{{x}}})})^t } \notag \\ \leftrightarrow \quad ( { \widetilde{{{{k}}}}({{{x}}})})^t & = { { \rm curl}_{{{{x}}}}}\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \,.\end{aligned}\ ] ] note that the trace of @xmath76 is zero : @xmath80 } ) = { { \rm div}_{{{{x}}}}}{{\rm curl}_{{{{x}}}}}{{{{u } } } ( { { { x}}})}= { \langle{{\rm grad}_{{{{x } } } } [ { { { \rm curl}_{{{{x}}}}}{{{{u } } } ( { { { x } } } ) } } ] } , { { { \mathbbm{1}}}}\rangle } = - u_{a , bj } \ , \epsilon_{abi } \ , \delta_{ij } \notag \\ & = - u_{a , bi } \ ,
\epsilon_{abi } = - u_{1,23 } + u_{1,32 } - u_{2,31 } + u_{2,13 } - u_{3,12 } + u_{3,21 } = 0 \ , .\end{aligned}\ ] ] in section [ kapnotations ] we have used the baseless permutation @xmath81 to define some functions . therefore , the base of the functions argument yields the base of the function automatically . on the one hand , this is convenient but on the other hand not sufficiently precise to investigate some transformation rules to be considered in section [ kaprotinvariance ] .
thus , we make use of the third order levi - civita tensor @xmath82 the components of the levi - civita tensor are the scalar triple product of an orthogonal unit base , which may be rotated
. therefore , let @xmath83 be a constant rotation tensor with @xmath84 and @xmath85 , mapping the orthogonal referential system of euclidean vectors @xmath86 to a rotated system @xmath87 via @xmath88 note that the components @xmath89 are defined by the commutative inner product .
however , the rotation tensor @xmath83 is generally not symmetric : @xmath90 .
since the referential basis vectors @xmath86 are considered to be orthogonal and of unit length the rotated basis vectors @xmath87 are orthogonal and of unit length as well : @xmath91 thus , the levi - civita tensor @xmath92 can be written in terms of the rotated base with components @xmath93 since the scalar triple product does not depend on the direction of the orthogonal unit base , the components of the levi - civita tensor are independent concerning the direction of the orthogonal unit base .
its components are independent concerning the direction of the orthogonal unit base . ] in @xmath94 .
thus , the levi - civita tensor is an isotropic tensor of order 3 with respect to proper orthogonal transformations , which is reading @xmath95 in eq . , and
yielding the identity @xmath96 since the symbolic notation of tensor products is only established in @xmath97 we prescind from defining a new kind of simple contraction .
this would be necessary to switch from index to symbolic notation in eq .
, where each rotation tensor @xmath83 has one simple contraction with the corresponding index of the permutation . in eq . is not possible via standard symbolic notation . ] since the rotation @xmath83 in eq .
is arbitrary , the identity must also hold for the transposed of @xmath83 , which is also in @xmath98 .
thus , a similar identity to eq .
is easily deduced : @xmath99 another identity can be found via simple contraction of @xmath83 and the levi - civita tensor , which is considered in the basis @xmath86 on the left hand side and by the basis @xmath87 on the right hand side : @xmath100 note that eq .
links a linear to a quadratic term of @xmath83 .
in geometrically nonlinear hyperelasticity we know that frame - indifference is left - invariance of the energy under so(3)-action and isotropy is right - invariance under so(3 ) , i.e. for the energy density @xmath101 we have @xmath102 therefore , by specifying @xmath103 we obtain also the necessary invariance condition @xmath104 note that condition does not imply objectivity , as can be seen from considering the energy expressions @xmath105 satisfying but not being frame - indifferent @xcite . with the same example one sees that condition does not imply isotropy .
therefore , applying the transformation @xmath106
( i.e. the @xmath2-transformation defined in eq . ) has * no * intrinsic meaning in nonlinear elasticity theory as such .
+ it is clear that every frame - indifferent elastic energy @xmath101 can be expressed in the right cauchy - green tensor @xmath107 in the sense that there is a function @xmath108 such that @xmath109 @xmath110 of course , any @xmath101 of the form is automatically frame - indifferent . applying the isotropy condition to the representation in eq .
we must have @xmath111 therefore , for reduced energy expressions the isotropy requirement can be equivalently stated as @xmath112 in eq .
we see that isotropy is invariance of the function @xmath113 under simultaneous spatial and referential rotation of the coordinate system with the same rotation @xmath83 by interpreting the cauchy - green tensor @xmath114 as a linear mapping @xmath115 , which transforms under such a change of coordinate system as @xmath116 we may now define a transformation of the deformation @xmath117 to new spatial and referential coordinates via to be a finite - sized ball @xmath118 which is invariant under rotations . then @xmath119 . ]
@xmath120 then in his discussion of isotropy for frame - indifferent formulations , however , he remains on the level of stresses . ] @xmath121 } = { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { { q } } } \ , { { \mbox{$\varphi$ } } } ( { { { q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } = \frac{\partial \ , [ { { { q } } } \ , { { \mbox{$\varphi$ } } } ( { { { q}}}^t \ , { { \mbox{$\xi$}}})]}{\partial \ , { { \mbox{$\xi$}}}}= { { { q } } } \ , \frac{\partial \ , { { \mbox{$\varphi$ } } } ( { { { x}}})}{\partial \ , { { { x } } } } \ , \frac{\partial \ , { { { x}}}}{\partial \ , { { \mbox{$\xi$}}}}= { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$ } } } ( { { { x } } } ) } ] } \ , { { { q}}}^t\end{aligned}\ ] ] and @xmath122})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{\mbox{$\varphi$}}}^{\sharp}({{\mbox{$\xi$}}})}}]}=({{\rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } \ , { { { q}}}^t)^t \ , { { \rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } { { { q}}}^t = { { { q } } } \ , { { { { c}}}({{{x}}})}\ , { { { q}}}^t \,.\end{aligned}\ ] ] gathering our findings so far we can state : @xmath123[2cm][c]{14 cm } { the nonlinear hyperelastic formulation is frame - indifferent and isotropic { \bf if and only if } there exists $ \psi : { { \rm{sym } } } ^+(3 ) \rightarrow \operatorname{\mathbb{r}}$ such that $ w({{{f } } } ) = \psi({{{f}}}^t \ , { { { f}}})$ and $ w({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{\mbox{$\varphi$}}}^{\sharp}({{\mbox{$\xi$ } } } ) } } ] } ) \overbrace{=}^{\rm form - invariance } w({{\rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } ) \quad \left[{\rm i.e. } \quad \psi({{{q } } } \ , { { { c } } } \ , { { { q}}}^t ) = \psi({{{c } } } ) \quad \forall \ , { { { q } } } \in { { \rm so}}(3 ) \right ] \ , . $ } } \end{aligned}\ ] ] in nonlinear elasticity , invariance of the formulation under eq . is a consequence of objectivity ( left @xmath98-invariance of the energy ) . therefore , the transformation in eq . may be used to probe objectivity and isotropy of the formulation , while it is not equivalent to both . if we assume already the reduced representation in @xmath113 , then rotational invariance of the formulation under eq .
is equivalent to isotropy , see also @xcite .
+ the well - known representation theorems imply that any @xmath101 satisfying the isotropy condition and the classical format of material frame - indifference , @xmath124 must be expressible in terms of the principal invariants of @xmath125 , i.e. , @xmath126 where @xmath127 incidentally , @xmath128 is not only invariant under compatible changes of the reference configuration with rigid rotations @xmath129 ( constant rotations ) , but also under inhomogeneous rotation fields @xmath130 , that is , @xmath131 since @xmath132 therefore , the principal invariants @xmath133 are unaffected by inhomogeneous rotations .
however , we must stress that isotropy per se is not defined as invariance under right multiplication with @xmath134 .
the difference between form - invariance under compatible transformations with rigid rotations @xmath129 ( isotropy ) and right - invariance under inhomogeneous rotation fields @xmath134 will only become visible in higher gradient elasticity treated in the next section . in any first gradient theory both requirements coincide .
note that likewise , is invariant under left multiplication with inhomogeneous rotation fields .
we define for further use @xmath135[1.0cm][c]{14 cm } { $ \text{\bf left - local so(3)-invariance } \,\,\,\,\ , : \quad w(q(x ) \ , f(x ) ) = w(f(x ) ) \quad \forall q(x ) \in { { \rm so}}(3 ) \\ \\ \text{\bf right - local so(3)-invariance } \ , : \quad w(f(x ) \ , q(x ) ) = w(f(x ) ) \quad \forall q(x ) \in { { \rm so}}(3 ) \ , . $ } } \end{aligned}\ ] ] similarly , we define for * constant * rotations @xmath136 @xmath137[1.0cm][c]{14 cm } { $ \text{\bf left - global so(3)-invariance } \,\,\,\,\ , : \quad w(\overline{q } \ , f(x ) ) = w(f(x ) ) \quad \forall \overline{q } \in { { \rm so}}(3 ) \quad \text{\bf objectivity } \\
\\ \text{\bf right - global so(3)-invariance } \ , : \quad w(f(x ) \ , \overline{q } ) = w(f(x ) ) \quad \forall \overline{q } \in { { \rm so}}(3 ) \quad \text{\bf isotropy } \ , .
$ } } \end{aligned}\ ] ] with these definitions we have shown that for @xmath138 both , local and global , left and right so(3 ) invariance are satisfied .
left - global so(3)-invariance is identical to the cosserat s invariance under `` action euclidienne '' @xcite . in this subsection
we would like to extend the `` local '' picture of the previous subsection to second gradient materials . using the noether theorem and lie - point symmetries ,
the explicit expressions of the isotropy condition in linear gradient elasticity of grade-2 and in linear gradient elasticity of grade-3 have been derived by lazar and anastassiadis ( * ? ? ?
* eq.(4.76 ) ) and agiasofitou and lazar ( * ? ? ?
* eqs.(57 ) and ( 58 ) ) , respectively , as consequence of global rotational invariance . + for the sake of clarity we leave first objectivity aside and discuss only isotropy .
moreover , we make the following simplifying assumptions . we consider a * homogeneous material * given as a finite - sized ball @xmath139 and we want to formalize the statement that first rotating the ball @xmath140 and then to apply the loads leaves invariant the response in the sense that this is indistinguishable from not rotating the ball , see fig .
[ objectivitypic ] .
this is in accordance with the statement in truesdell & noll @xcite : `` for an isotropic material in an undistorted state , a physical test can not detect whether or not the material has been rotated arbitrarily before the test ist made . ''
( 130,110 ) ( 0,0 ) let the elastic energy of the body @xmath139 depend also on second gradients of the deformation , i.e. we consider therefore , @xmath141 } \right ) \ , { \,{\rm d}}x = \int \limits_{\displaystyle x \in b } \!\!\!\
! w \left ( { { \rm grad}_{{{{x}}}}[{\varphi(x ) } ] } , \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\varphi(x ) } ] } } ] } \right ) \ , { \,{\rm d}}x \,.\end{aligned}\ ] ] we first transform eq . to new coordinates
@xmath142 via introducing an orientation preserving diffeomorphism @xmath143 , @xmath144 see also @xcite . for further reference ,
we compute some relations connected to the transformation . taking the first and second derivative with respect to @xmath145 in @xmath146 and using @xmath147 we obtain @xmath148 } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] }
\quad \leftrightarrow \quad { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , , \end{aligned}\ ] ] and @xmath149 } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \notag \\ & 0= { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\zeta^{-1}(x ) ) } ] } } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } + { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\zeta^{-1}(x ) ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \notag \\ \leftrightarrow \,\ , & 0= { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\zeta^{-1}(x ) ) } ] } } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } + { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\zeta^{-1}(x ) ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \notag \\ \leftrightarrow \,\ , & { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { \zeta(\zeta^{-1}(x ) ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } = - { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \,.\end{aligned}\ ] ] with help of eq .
the latter implies @xmath150 } } ] } = -({{\rm grad}_{{{\mbox{$\xi$ } } } } [ { \zeta(\xi)}]})^{-1 } \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \,.\end{aligned}\ ] ] connected to the coordinate transformation we consider the deformation expressed in these new coordinates via setting @xmath151 the standard chain rule induces for the first and second derivative of @xmath152 with respect to @xmath145 the following relations @xmath153 } = { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \ , , \end{aligned}\ ] ] and @xmath154 } } ] } \notag \\ & \qquad = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } + { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \,.\end{aligned}\ ] ] the elastic energy , expressed in new coordinates via the transformation of integrals formula reads then , inserting and @xmath155 } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \ , { { \rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } \notag \\ & \qquad \qquad \qquad \qquad + { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \bigg ) \ , \det({{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } ) \ , { \,{\rm d}}\xi \,.\end{aligned}\ ] ] we rewrite as follows , using again @xmath156 } = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1}$ ] from @xmath157 } \ , & ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \notag \\ & \qquad \qquad \qquad + { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x ) } ] } } ] } \bigg ) \ , \det({{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } ) \ , { \,{\rm d}}\xi \,.\end{aligned}\ ] ] further , we use relation to substitute @xmath158}}]}$ ] yielding @xmath159 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \notag \\
\quad & - { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { \zeta(\xi)}]})^{-1}\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \bigg ) \ , \det({{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } ) \ , { \,{\rm d}}\xi \,.\end{aligned}\ ] ] we say that the elastic energy is * form - invariant * with respect to the ( referential ) coordinate transformation @xmath160 if and only if @xmath161 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \notag \\
\quad & - { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { \zeta(\xi)}]})^{-1}\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } } ] } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \bigg ) \ , \det({{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } ) \ , { \,{\rm d}}\xi \notag \\ & \qquad \qquad \qquad \qquad \qquad \qquad = \int \limits_{\displaystyle \xi \in \zeta^{-1}(b ) } \!\!\!\!\!\!\!\ ! w \bigg ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \bigg ) \ , { \,{\rm d}}\xi \,,\end{aligned}\ ] ] and we call then @xmath160 a * material symmetry transformation * , cf .
* eq.(11 ) ) , @xcite and ( * ? ? ?
it is simple to see that for equality to be satisfied at all , we must have @xmath162 } ) = 1 \,.\end{aligned}\ ] ] it can be shown as well that we may restrict attention to the situation @xmath163 } \in { \mathcal g } \subseteq { { \rm so}}(3 ) \qquad \forall \
, \xi \in \zeta^{-1}(b ) \,,\end{aligned}\ ] ] which is consistent with the local statement in the previous subsection [ kapenerconsider ] . here
, @xmath164 is a subgroup of the proper rotation group @xmath98 . and
@xmath165 then the material is called centro - symmetric .
we do not discuss this possibility . ]
the group @xmath166 can be discrete has been given in hessel @xcite . ]
( for e.g. orthotropy ) or continuous ( continuous for transverse isotropy and isotropy )
. we may therefore specify to the case of @xmath167 and writing @xmath168})^{-1 } = q(\xi ) \,,\end{aligned}\ ] ] we obtain as first concise form - invariance statement for material symmetry @xmath169 } \ ,
q(\xi ) , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , q(\xi ) \ , q(\xi ) \notag \\ & \qquad - { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , q(\xi ) \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{q^t(\xi ) } ] } \ , q(\xi ) \ , q(\xi ) \bigg ) \ , 1 \ , { \,{\rm d}}\xi \notag \\ & \qquad
\qquad \qquad \qquad = \int \limits_{\displaystyle \xi \in \zeta^{-1}(b ) } \!\!\!\!\!\!\!\ ! w \bigg ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \bigg ) \ , { \,{\rm d}}\xi \quad \forall q(\xi)\in { \mathcal g } \,.\end{aligned}\ ] ] the format can be found e.g. in @xcite .
the homogeneous material is usually said to be * isotropic * following the local reasoning , in the first entry and the zero in the second , @xmath170 , these transformations can be interpreted as rigid rotations , and we call the respective reference placement an undistorted state . if a material allows for such undistorted states , it is a solid .
if it contains all orthogonal dyadics in the first entry , then the material is called isotropic '' .
this statement leaves open that for isotropy , we may choose inhomogeneous rotations @xmath130 .
in our setting we understand by isotropy using necessarily rigid rotations and therefore only @xmath171 .
] whenever @xmath172 and holds for all non - constant rotation fields @xmath173 . in the following ,
consistent with previous definitions , we will call .
+ let us proceed by showing that right - local so(3)-invariance in is , however , * misconceived * as a general condition for isotropy .
the misconception in consists in suggesting a generality of transformation behaviour followed for the rotation field @xmath174 that is , in reality , not present . to see this we refer to figure [ objectivitypic ] . requiring in that @xmath175})^{-1 }
= q^t(\xi ) \in { { \rm so}}(3 ) \quad
\leftrightarrow \quad { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } = q(\xi ) \ , \in { { \rm so}}(3)\end{aligned}\ ] ] means , by a * classical geometric rigidity * result ( see e.g. @xcite ) that indeed @xmath176[1.0cm][c]{14 cm } { $ { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } = q(\xi ) \ , \in { { \rm so}}(3 ) \quad
\rightarrow \quad q(\xi ) = { \overline{q}}={\rm const . }
\quad { \rm and } \quad \zeta(\xi)= { \overline{q}}\ , \xi + \overline{b } \\ \\ \text{and therefore } \quad { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi ) } ] } } ] } = 0 \qquad \text{\bf compatibility of rotations } \ , .
$ } } \end{aligned}\ ] ] thus , the correct statement for isotropy , in our view is , ) = w({\overline{q}}\ , ( f^t f - { { { \mathbbm{1 } } } } ) \ , { \overline{q}}^t , \ , \operatorname{curl\,}[{\overline{q}}\ , ( f^t f - { { { \mathbbm{1 } } } } ) \ , { \overline{q}}^t ] ) \notag \\ & = w({\overline{q}}\ , ( f^t \ , f - { { { \mathbbm{1 } } } } ) \ , { \overline{q}}^t , \ , { \overline{q}}\ , \operatorname{curl\,}[f^t \ , f - { { { \mathbbm{1 } } } } ] \ , { \overline{q}}^t ) = w(f^t \ , f - { { { \mathbbm{1 } } } } , \ , \operatorname{curl\,}[f^t \ , f - { { { \mathbbm{1 } } } } ] ) \ , , \notag\end{aligned}\ ] ] where @xmath101 is the free - energy of the nonlinear couple stress model of toupin @xcite
. additionally , in @xcite also the anisotropic case is given , c.f . @xcite .
] noting that for a global rotation of the coordinates we have @xmath177 @xmath178 } & \ , \overline{q}^t , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \overline{q}^t \ , \overline{q}^t \bigg ) \ , { \,{\rm d}}\xi \notag \\ & = \int \limits_{\displaystyle \xi \in b } \!\!\!\ ! w \bigg
( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \bigg ) \ , { \,{\rm d}}\xi \quad \forall \overline{q } \in { { \rm so}}(3 ) \,.\end{aligned}\ ] ] we will refer to this condition as * right - global so(3)-invariance * , which , for us , is * isotropy*.$ ] . '' ] we appreciate that the right - local so(3)-invariance condition is much to restrictive in that arbitrary , inhomogeneous rotation fields are allowed instead of only constant rotations @xmath129 .
the reader should carefully note that we started by using a coordinate transformation @xmath179 and therefore we require in the end that @xmath180 , in line with our understanding to really rotate the finite sized ball .
there is no other coordinate transformation @xmath160 such that @xmath181}=q(\xi ) \ , \in { { \rm so}}(3)$ ] everywhere , provided a minimum level of smoothness is assumed .
we note that @xmath182 . } ) = w_{\rm local}(f ) + w_{\rm curv}({{\rm grad}[{f}]})$ ] .
there , @xmath183 must be isotropic and the condition on @xmath184 can be localized , yielding in the end @xmath185 } } ] } \ , q^t\!\!(\xi ) \ , q^t\!\!(\xi ) ) \ , { \,{\rm d}}\xi = \int \limits_{\displaystyle \xi \in b } \!\!\!\ ! w_{\rm curv } ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } ) \ , { \,{\rm d}}x \quad \forall q(\xi ) \in { { \rm so}}(3 ) \ , .\end{aligned}\ ] ] thus , can be expressed with inhomogeneous rotation fields since the space derivatives do not act on the rotations in this representation . ]
@xmath186[1.0cm][c]{14 cm } { $ \qquad \text { \bf right - local so(3)-invariance } \hspace{0.2 cm } \rightarrow \hspace{0.2 cm } \text { \bf right - global so(3)-invariance } \leftrightarrow \text{\bf isotropy}\\ \\
\text { } \hspace{2.0 cm } \eqref{intwgradgrad5 } \hspace{6 cm } \eqref{intwgradgrad6 } $ } } \end{aligned}\ ] ] the condition is consistent with ( * ? ? ?
* eq.16 ) in that the term with additional first @xmath187-derivatives on @xmath188 in is absent .
we remark again , that in the local theory the condition @xmath189 } \ , \overline{q}^t \right ) { \,{\rm d}}\xi = \int \limits_{\displaystyle \xi \in b } \!\!\!\ ! w \left ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \right ) \ , { \,{\rm d}}\xi \quad \forall \overline{q } \in { { \rm so}}(3)\end{aligned}\ ] ] can not distinguish between constant or non - constant rotations , as seen in .
this means for the local theory that implies and is therefore equivalent to @xmath190 } \ , \overline{q}^t\!\!(\xi )
\right ) { \,{\rm d}}\xi = \int \limits_{\displaystyle \xi \in b } \!\!\!\ ! w \left
( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \right ) \ , { \,{\rm d}}\xi \ , .\end{aligned}\ ] ] the latter might explain why one may be inclined to allow non - constant rotation fields in , which is forbidden for higher gradient materials , as we saw . in higher gradient
elasticity , however , not every ( left - global so(3)-invariant ) objective energy is also left - local so(3)-invariant . } ) = \psi(c , { { \rm grad}[{c}]})$ ] is automatically already left - local so(3)-invariant . ]
this can be seen by looking at @xmath191 } ) = \psi(c ) + \sum_{i=1}^3 { \|\ , f^t \ , \frac{\partial f}{\partial x_i } \ , \|}^2 \ , .\end{aligned}\ ] ] the curvature expression @xmath192 is objective , but not left - local so(3)-invariant , i.e. @xmath193 implying that @xmath194 in order to obtain directly a left - local so(3)-invariant expression , we could choose @xmath195 } ) = \psi(c ) + \sum_{i=1}^3 { \|\ , \operatorname{sym}\left(f^t \ , \frac{\partial f}{\partial x_i}\right )
\ , \|}^2 \ , , \end{aligned}\ ] ] since @xmath196})$ ] is depending on the full gradient of @xmath197 via @xmath198 } & = \left [ \frac{\partial ( f^t f)}{\partial x_1 } \ , , \
, \frac{\partial ( f^t f)}{\partial x_2 } \ , , \ , \frac{\partial ( f^t f)}{\partial x_3 } \ , \right ] \notag \\ & = \left [ \frac{\partial ( f^t)}{\partial x_1 } \ , f + f^t \ , \frac{\partial ( f)}{\partial x_1 } \ , , \ , \frac{\partial ( f^t)}{\partial x_2 } \ , f+ f^t \ , \frac{\partial ( f)}{\partial x_2 } \ , , \ ,
\frac{\partial ( f^t)}{\partial x_3 } \ , f+ f^t \ , \frac{\partial ( f)}{\partial x_3 } \ , \right ] \ , .\end{aligned}\ ] ] thus , using @xmath199 column - wise we get that @xmath200 } ) & = { \|\ , \12 \ , \left [ \left ( f^t \ , \frac{\partial ( f)}{\partial x_1}\right)^t \!\!\ ! + f^t \ ,
\frac{\partial ( f)}{\partial x_1 } \ , , \
, \left ( f^t \ , \frac{\partial ( f)}{\partial x_2}\right)^t \!\!\ ! + f^t \ , \frac{\partial ( f)}{\partial x_2 } \ , , \ , \left ( f^t \ , \frac{\partial ( f)}{\partial x_3}\right)^t \!\!\ ! + f^t \ , \frac{\partial ( f)}{\partial x_3 } \right ] \ , \|}^2 \notag \\ & = { \|\ , \12 \ , \left [ 2 \ , \operatorname{sym}\left ( f^t \ , \frac{\partial ( f)}{\partial x_1}\right ) \ , , \ , 2 \ , \operatorname{sym}\left ( f^t \ , \frac{\partial ( f)}{\partial x_2}\right ) \ , , \ , 2 \ , \operatorname{sym}\left ( f^t \ , \frac{\partial ( f)}{\partial x_3}\right ) \right ] \ , \|}^2 \notag\\ & = \sum_{i=1}^3 { \|\ , \operatorname{sym}\left ( f^t \ , \frac{\partial ( f)}{\partial x_i}\right ) \ , \|}^2 = \widehat{\psi}_{\rm curv}(f^t \ , { { \rm grad}_{{{{x}}}}[{f } ] } ) \ , .\end{aligned}\ ] ] note that is right - global so(3)-invariant , but not right - local so(3)-invariant , which can be seen similarly to the above discussion . + next ,
we assume for simplicity a free energy density of the form @xmath201})$ ] .
moreover , we already consider a further reduced objective format @xmath202})$ ] , which is left - local and left - global so(3 ) invariant as seen above . now , for investigation of isotropy
, it suffices to concentrate on the induced conditions on @xmath203})$ ] .
we repeat the ( referential ) coordinate transformation and we already observed in that we can always write @xmath204 } ) = \widehat{\psi}_{\rm curv}(f^t \ , { { \rm grad}_{{{{x}}}}[{f } ] } ) \ , .\end{aligned}\ ] ] moreover , from the results of the previous section right - global so(3)-invariance under * constant rotations * @xmath136 is equivalent to @xmath205 } \ , & \overbrace{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x)}]}}^{\overline{q}})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \overline{q } \ , \ , \overline{q } \ , \big ) \ , { \,{\rm d}}\xi \notag \\ & \qquad \qquad= \int \limits_{\displaystyle \xi \in b } \!\!\!\ ! \widehat{\psi}_{\rm curv } \big ( ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \big ) \ , { \,{\rm d}}\xi \,.\end{aligned}\ ] ] and this is equivalent to @xmath206})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \overline{q } \ , \ , \overline{q } \ , \big ) \ , { \,{\rm d}}\xi \notag \\ & \qquad = \int \limits_{\displaystyle \xi \in b } \!\!\!\ !
\widehat{\psi}_{\rm curv } \big ( ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \big ) \ , { \,{\rm d}}\xi \quad \forall \overline{q } \in { { \rm so}}(3 ) \,.\end{aligned}\ ] ] as an example , let us quickly check that the objective ( left - global so(3)-invariant ) curvature expression @xmath207 } } ] } \ , \|}^2 \,,\end{aligned}\ ] ] is isotropic in the sense of right - global so(3)-invariance . indeed , we have to check only that @xmath208 } } ] } { \overline{q}}\ , \ , { \overline{q}}\ , \|}^2 = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \|}^2 \,.\end{aligned}\ ] ] the latter is verified since in indices we get @xmath209 } } ] } { \overline{q}}\ , { \overline{q}}\ , \|}^2 = { \|\ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } { \overline{q}}_{an } \ , { \overline{q}}_{jk } \ , e_i \otimes e_k \otimes e_n \ , \|}^2 \notag\\ & = { \langle({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } { \overline{q}}_{an } \ , { \overline{q}}_{jk } \ , e_i \otimes e_k \otimes e_n \ , , \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{opq } { \overline{q}}_{qs } \ , { \overline{q}}_{pt } \ , e_o \otimes e_t \otimes e_s\rangle } \notag \\ & = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{opq } \ , \delta_{io } \ , { \overline{q}}_{an } \ ,
\delta_{ns } \ , { \overline{q}}_{jk } \ , \delta_{kt } \ , { \overline{q}}_{qs } \ , { \overline{q}}_{pt } \notag \\ & = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ipq } \ , { \overline{q}}_{as } \ , { \overline{q}}^t_{sq } \ , { \overline{q}}_{jt } \ , { \overline{q}}^t_{tp } \notag \\ & = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ipq } \ , \delta_{aq } \ , \delta_{jp } \notag \\ & = ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi)}]}}]})_{ija } = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \|}^2 \,.\end{aligned}\ ] ] note , however , that is a prototype expression which does not satisfy the right - local so(3)-invariance condition .
this is the case , since for inhomogeneous rotation fields @xmath173 we have @xmath210 } } ] } q(\xi ) \ , \ ,
q(\xi ) - { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } \ , q(\xi ) \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{q(\xi ) } ] } & \ , q(\xi ) \ , \ , q(\xi ) \ , \|}^2 \notag\\ & \neq { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\xi ) } ] } } ] } \ , \|}^2 \,.\end{aligned}\ ] ] it is instructive to see that the right - local so(3)-invariance can be satisfied nevertheless within a second gradient model by taking @xmath211 } ) = \psi(c , \ , { { \rm grad}_{{{{x}}}}[{c } ] } ) = \widehat{\psi}(i_1(c ) , \ ,
i_2(c ) , \ , i_3(c ) ) + \sum_{k=1}^3 { \|\ , { { \rm grad}_{{{{x}}}}[{i_k(c ) } ] } \ , \|}^2 \,,\end{aligned}\ ] ] with @xmath212 arbitrary . in this formulation , objectivity is clear , as the functional dependence is expressed only in @xmath125 and isotropy will be secured by working only with the principal invariants @xmath133 and their derivatives . in order to see that right - local so(3)-invariance is satisfied we look exemplarily at @xmath213 } \
, \|}^2_{\operatorname{\mathbb{r}}^3 } = { \|\ , { { \rm grad}_{{{{x}}}}[{{\|\ , { { \rm grad}_{{{{x}}}}[{\varphi(x ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^{3 \times 3 } } } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \,.\end{aligned}\ ] ] we compute directly , following the steps leading to @xmath214 } \ , \|}^2_{\operatorname{\mathbb{r}}^{3 \times 3 } } & = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , ( \hspace{-0.8cm}\underbrace{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x)}]}}_{\displaystyle ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \in { { \rm so}}(3 ) } \hspace{-0.8 cm } ) \ , \|}^2_{\operatorname{\mathbb{r}}^{3 \times 3 } } = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , q(\xi ) \
, \|}^2_{\operatorname{\mathbb{r}}^{3 \times 3 } } \notag \\ & = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^{3 \times 3 } } \,,\end{aligned}\ ] ] where we have allowed that @xmath173 may be inhomogeneous .
further , using indices @xmath215 } \ , \|}^2 } ] } \notag \\ & \qquad \qquad \qquad = 2 \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]})_{ij } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]}}]})_{ijb } \ , ( \hspace{-0.8cm}\underbrace{{{\rm grad}_{{{{x}}}}[{\zeta^{-1}(x)}]}}_{\displaystyle ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]})^{-1 } \in { { \rm so}}(3 ) } \hspace{-0.8 cm } ) _ { bk } \ , e_k \,.\end{aligned}\ ] ] this implies finally @xmath216 } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } & = { \|\ , 2 \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]})_{ij } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]}}]})_{ijb } ( \hspace{-0.6 cm } \underbrace{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\zeta(\xi)}]}}_{\displaystyle \in { { \rm so}}(3 ) \ , \forall \xi \in \zeta^{-1}(b ) } \hspace{-0.6 cm } ) ^{-1}_{bk } \ , e_k \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \notag \\ & = { \|\ , ( 2 \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]})_{ij } \ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]}}]})_{ijk } \ , e_k \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \notag \\ & = { \|\ , 2 \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \notag \\ & = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{\|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ , \|}^2 } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \notag \\ & = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{tr}\{({{\rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x))}]})^t \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\varphi^{\flat}(\zeta^{-1}(x ) ) } ] } \ } } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \notag \\ & = { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{tr}c^{\flat } } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3}={\|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{i_1(c^{\flat } ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}}^3 } \,,\end{aligned}\ ] ] which is form - invariance in the sense of condition and implies therefore also condition . note again that we did not use constant rotations . therefore , @xmath217 } \ , \|}^2_{\operatorname{\mathbb{r}}^3}$ ] is already right - local so(3)-invariant . in the lack of having an encompassing representation theorem for right - local @xmath98-invariant functions let us describe a large class of such expressions . for this
we assume that @xmath218 } ) = \widehat{\psi}(i_1(c ) , \ , i_2(c ) , \ , i_3(c ) ) + \widehat{\psi}_{\rm curv}({{\rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } ) \,,\end{aligned}\ ] ] where we defined the vector of principal invariants @xmath219 and we specify , furthermore , that @xmath220 } ) = \widehat{\widehat{\psi}}_{\rm curv}(({{\rm grad}_{{{{x}}}}[{\widehat{i}(c)}]})^t { { \rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } ) \,.\end{aligned}\ ] ] then right - local @xmath98-invariance is entirely dependent upon properties of @xmath221 , namely @xmath222}^t { { \rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } \ , q(\xi)^t ) = \widehat{\widehat{\psi}}_{\rm curv}({{\rm grad}_{{{{x}}}}[{\widehat{i}(c)}]}^t { { \rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } ) \quad \forall \ ,
q(\xi ) \in { { \rm so}}(3 ) \,.\end{aligned}\ ] ] thus , similar to the local case , @xmath223 can be further reduced to a * symmetric function * of the * singular values * of @xmath224}$ ] .
+ in fact , another representation , also reminiscent of the local theory suggests itself .
any @xmath225})$ ] satisfies right - local so(3)-invariance if an only if @xmath226 } \ , q(\xi ) ) = \widehat{\psi}_{\rm curv}({{\rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } ) \ , .\end{aligned}\ ] ] the latter may be used to obtain the reduced format @xmath227 } ) = \widetilde{\psi}_{\rm curv}(\underbrace{{{\rm grad}_{{{{x}}}}[{\widehat{i}(c ) } ] } \ , ( { { \rm grad}_{{{{x}}}}[{\widehat{i}(c)}]})^t}_{\displaystyle \colonequals \widetilde{b}_{\nabla } } ) \,,\end{aligned}\ ] ] where @xmath228 } \ , ( { { \rm grad}_{{{{x}}}}[{\widehat{i}(c)}]})^t = { { \rm grad}_{{{{x}}}}[{\widehat{i}(b ) } ] } \ , ( { { \rm grad}_{{{{x}}}}[{\widehat{i}(b)}]})^t \ , , \qquad b = f \ , f^t \,,\end{aligned}\ ] ] is a finger - type stretch tensor based on @xmath229}$ ] .
is an isotropic tensor function and the strain energy is an isotropic function of @xmath140 .
however , this further reduction possibility is not viable here since we would need an additional objectivity requirement .
this is absent in . ]
let us approach the problem with an easy example .
we have seen in sect.[kapoinonlin - non - local - case ] that @xmath230 is frame - indifferent ( left - global so(3)-invariant ) but not left - local so(3)-invariant . however , using an involved compatibility argument @xcite one can show that @xmath231 } ) = \widehat{\widehat{\psi}}(c , { { \rm grad}[{c } ] } ) \ , , \end{aligned}\ ] ] and the right - hand side is clearly left - local so(3)-invariant . in fact , it can be shown that every objective second gradient energy can be represented as } = c^{-1 } ( f^t { { \rm grad}[{f}]})$ ] .
this is done in bertram @xcite . ]
@xmath232 } ) = \psi(f^t f , f^t { { \rm grad}[{f } ] } ) = \widehat{\widehat{\psi}}(c , { { \rm grad}[{c } ] } ) \ , , \end{aligned}\ ] ] see toupin ( * ? ? ?
* eq.(10.24)+(10.26 ) ) and ( * ? ? ?
* eq.(5.12 ) ) .
this may lead one to assume that working solely with @xmath233})$ ] exhausts all possible fomulations for frame - indifferent second gradient models .
surprisingly , the situation is more complicated . returning to we observe @xmath234 } \\
\displaystyle \text{compatible}}}\quad & \hspace{2.3 cm } \not{\ !
\big\| } \,\ , { \substack{\displaystyle q(x ) \ , f \neq { { \rm grad}[{\vartheta } ] } \\ \displaystyle \text{incompatible } } } \notag \\ \psi \left(\sqrt{f^t f } , { \rm grad}\!\left[\!\sqrt{f^t f}\right]\right ) \quad & = \qquad
\widehat{\psi}\left(\sqrt{(q(x ) \ , f)^t ( q(x ) \ , f ) } , { \rm grad}\!\left[\!\sqrt{(q(x ) \ , f)^t ( q(x ) \ , f)}\right]\right ) \,.\end{aligned}\ ] ] the situation is completely analogous to the linearized case in which e.g. @xmath235 } \ , \|}^2 = { \|\ , { { \rm grad}[{\operatorname{axl}(\skew { { \rm grad}[{u } ] } ) } ] } \ , \|}^2 = { \|\ , \operatorname{curl\,}\operatorname{sym}{{\rm grad}[{u } ] } \ , \|}^2 \ , , \end{aligned}\ ] ] see eq . and , and both expressions are ( left ) global @xmath236-invariant , i.e. invariant w.r.t .
the transformation @xmath237 } \mapsto \overline{a } + { { \rm grad}[{u } ] } \ , , \ , \overline{a } \in \operatorname{\mathfrak{so}}(3)$ ] .
equality in holds only for gradients @xmath237}$ ] and uses compatibility . as equal energy expressions ,
both terms in give rise to equivalent euler - lagrange equations , while the formulation based on @xmath238})}]}$ ] delivers * non - symmetric force - stresses * and the formulation based on @xmath239}$ ] gives * symmetric * force stresses , see .
the difference resides in an extra stress contribution which has zero divergence , see @xcite .
therefore , both formulas , although based on the same energy , may yield essentially different boundary value problems .
the difference occurs for certain boundary conditions and load cases @xcite .
thus it makes a fundamental difference starting with the left or right term in , similarly with . in summary , @xmath240[2cm][c]{14.5 cm } {
the complete model formulation based on left - global so(3)-invariance is not necessarily captured by starting with a left - local so(3)-invariant expression .
hence , not all frame - indifferent variational formulations can be obtained by assuming $ w(f , { { \rm grad}[{f } ] } ) = \psi(c , { { \rm grad}[{c}]})$ and working with the latter expression $ \psi(c , { { \rm grad}[{c}]})$.\footnote{the reader should be aware that the energy in a higher - gradient model does not completely determine the model .
decisive is the way in which partial integration steps are performed and this may change boundary conditions . now , using $ { \|\ , { { \rm grad}[{\operatorname{axl}(\skew { { \rm grad}[{u } ] } ) } ] } \ , \|}^2 $ suggests some partial integration steps which leads in a natural way to nonsymmetric total force stresses . } } } \end{aligned}\ ] ] moreover , while representation theorems for isotropic energies in the reduced format @xmath241})$ ] are well - known , cf .
sect.[kaprayleighproduct ] and @xcite , the more complete format is given by @xmath242})$ ] and representation theorems for isotropic models in this extended setting are not standard .
the form of the stress - strain law for linear elastic solids @xcite reads @xmath243 in order to discuss isotropy , the classical procedure is to consider @xmath244 and @xmath245 as linear mappings and refer these linear mappings to new axes of coordinates .
we note that any linear mapping has a meaning independent of the used coordinate system .
then isotropy means that the formula connecting stress components with strain components is independent of the direction of the new axis . in this case
the quadratic energy function @xmath246 is invariant for all transformations from one set of orthogonal axes to another : @xmath247 hence , isotropy is equivalent to @xmath248 in terms of the elasticity tensor , eq .
implies @xmath249 in an isotropic material , when we do not want to directly work with the elastic energy , the principal axes of stress @xmath244 and strain @xmath245 must coincide : @xmath250 up to this point , the isotropy condition in eq . is not directly related to our @xmath2-transformation since @xmath251 are taken merely as linear mappings .
next we show that by properly introducing the @xmath2-transformation , isotropy in linear elasticity can be completely characterized . due to linear isotropic
elasticity the simultaneous transformation of spatial and referential coordinates yields the cauchy stress @xmath252 } + \lambda \
, \operatorname{tr}({{\rm grad}[{{{{u}}}}]})\ , { { { \mathbbm{1}}}}$ ] in the transformed representation @xmath253 } ) & = 2 \ , \mu \ , \operatorname{sym}({{{q } } } \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } { { { q}}}^t ) + \lambda \operatorname{tr}({{{q } } } \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } { { { q}}}^t ) { { { \mathbbm{1}}}}\notag \\ & = 2 \ , \mu \ , \operatorname{sym}({{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } { { { q}}}^t ) + \lambda \operatorname{tr}({{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } { { { q}}}^t ) { { { \mathbbm{1}}}}\notag \\ & = { { { q } } } \ , \ { 2\ , \mu \ , \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } + \lambda \ , \operatorname{tr}({{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) { { { \mathbbm{1}}}}\ } { { { q}}}^t \notag \\ & = { { { q } } } \ , { { \mbox{$\sigma$ } } } ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) { { { q}}}^t = { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t \,.\end{aligned}\ ] ] but in any ( rotated ) coordinate system the elasticity tensor has the same format from assuming isotropy ] is form - invariant with respect to a full orthogonal group of transformations of the natural state ... '' ] @xmath254 where @xmath255 thus , the requirement for isotropy is simply @xmath256 in terms of our @xmath2-transformation
. this yields @xmath257 eq . and
imply @xmath258 . since this is true for any @xmath259 we have in components : @xmath260 the general solution , for symmetric tensors
@xmath261 is the well - known relation @xmath262 where @xmath263 is the shear modulus and @xmath264 is the second lam parameter . in linear
elasticity the quadratic energy @xmath265 is defined as a function of the displacement gradient @xmath266 } = { { \rm grad}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } - { { { \mathbbm{1}}}}\qquad { \rm for } \quad { { \mbox{$\varphi$}}}({{{x } } } ) = { { { x } } } + { { { u}}}({{{x}}})\,.\end{aligned}\ ] ] linearized frame - indifference requires that @xmath267 } + \overline{{{{w}}}})= w_{\rm lin}({{\rm grad}[{{{{u } } } } ] } ) \qquad \forall \
, \overline{{{{w } } } } \in \operatorname{\mathfrak{so}}(3 ) \,.\end{aligned}\ ] ] this implies the reduced representation @xmath268})= \psi_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ) = \psi_{\rm lin}({{\mbox{$\varepsilon$ } } } ) \ , , \qquad { { \mbox{$\varepsilon$ } } } \colonequals \operatorname{sym}{{\rm grad}[{{{{u } } } } ] } \,.\end{aligned}\ ] ] describing the deformation @xmath269 by the displacement field @xmath270 one also gets @xmath271 = { { \mbox{$\xi$ } } } + { { { q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) = { { \mbox{$\xi$ } } } + { { { u}}}^\sharp({{\mbox{$\xi$ } } } ) \ , , \end{aligned}\ ] ] which shows that the tranformation in eq . is compatible with respect to displacements and deformations .
considering @xmath272})^t \ , { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { \mbox{$\varphi$}}}^{\sharp } } ] } & = { { { q } } } \ , ( { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$}}}}]})^t \ , { { { q}}}^t \ , { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$ } } } } ] } \ , { { { q}}}^t= { { { q } } } \ , ( { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$}}}}]})^t \ , { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$ } } } } ] } \ , { { { q}}}^t \notag \\ & = { { { q } } } \ , [ { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } + { \rm h.o.t.}]\ , { { { q}}}^t = { { { \mathbbm{1}}}}+ 2 \ , { { { q } } } \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t + { \rm h.o.t . }
\notag \\ & = { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym}{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t + { \rm h.o.t . }
= { { { \mathbbm{1}}}}+ 2 \operatorname{sym}{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{u}}}^\sharp({{\mbox{$\xi$ } } } ) } ] } + { \rm h.o.t . } \
, , \end{aligned}\ ] ] we may use transformation for a geometrically linear displacement formulation and probe objectivity and isotropy in problems with linearized kinematics , which is investigated in the next section is not compatible with the linear displacement formulation , since , @xmath273 } = ( { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$}}}({{{x } } } ) } ] } ) \ , { { { q}}}$ ] and @xmath274})^t \ , { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { \mbox{$\varphi$}}}^{\flat } } ] } & = { { { q } } } \ , ( { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$}}}}]})^t \ , { { \rm grad}_{{{{x}}}}[{{{\mbox{$\varphi$ } } } } ] } \ , { { { q}}}^t = { { { q } } } \ , [ { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym } ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } + { \rm h.o.t.})\ , { { { q}}}^t \notag \\ & = { { { \mathbbm{1}}}}+ 2 \ , { { { q } } } \operatorname{sym } ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , { { { q}}}^t + { \rm h.o.t . } = { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym } ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t ) + { \rm h.o.t . } \notag \\ & \neq { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym}{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{u}}}^\flat({{\mbox{$\xi$ } } } ) } ] } + { \rm h.o.t . } = { { { \mathbbm{1}}}}+ 2 \ , \operatorname{sym}({{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q } } } ) + { \rm h.o.t.}\end{aligned}\ ] ] note that @xmath275 is the correct transformation in nonlinear elasticity when we only want to discuss isotropy there as seen in section [ kapoinonlin - non - local - case ] . ] . as a preliminary , for the isotropic cauchy stress tensor @xmath276 in linear elasticity @xmath277 } + \lambda \operatorname{tr}(\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , { { { \mathbbm{1}}}}\,,\end{aligned}\ ] ] we obtain @xmath278 } + \lambda \operatorname{tr}(\operatorname{sym}{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{u}}}^\sharp({{\mbox{$\xi$ } } } ) } ] } ) \ , { { { \mathbbm{1}}}}\,,\end{aligned}\ ] ] as expected . + it is illuminating to see that , although in a linearized context , we may still use the full @xmath2-transformation @xmath279 } = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t\end{aligned}\ ] ] for the displacement @xmath70 , since it is fully consistent with the corresponding transformation for the deformation : @xmath280 } = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } \ , { { { q}}}^t\end{aligned}\ ] ] and @xmath281 } - { { { \mathbbm{1}}}}&= { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } \ , { { { q}}}^t - { { { \mathbbm{1}}}}= { { { q } } } \ , ( { { \rm grad}_{{{{x}}}}[{{{{\mbox{$\varphi$}}}({{{x } } } ) } } ] } - { { { \mathbbm{1 } } } } ) \ , { { { q}}}^t = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t \notag \\ & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } \,.\end{aligned}\ ] ] with the help of the @xmath2-transformation we may alternatively characterize frame - indifference and isotropy .
@xmath282[3cm][c]{14.5 cm } { the linear hyperelastic formulation is ( linearized ) frame - indifferent and isotropic { \bf if and only if } there exists $ \psi_{\rm lin } : { { \rm{sym } } } ( 3 ) \rightarrow \operatorname{\mathbb{r}}$ such that\\ \\ $ w_{\rm lin}({{\rm grad}[{{{{u } } } } ] } ) \overbrace{=}^{{\rm linearized } \atop { \rm frame - indifferent } } \psi_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ) \ , , \quad { \rm and } $ \\ \\ $ w_{\rm lin}({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } ) \overbrace{=}^{\rm form - invariance } w_{\rm lin}({{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \quad \left [ { \rm i.e. } \quad \psi_{\rm lin}({{{q } } } \ , { { \mbox{$\varepsilon$ } } } \ , { { { q}}}^t ) = \psi_{\rm lin}({{\mbox{$\varepsilon$ } } } ) \ , \forall \ , { { { q } } } \in { { \rm so}}(3 ) \ , . \right ] $ } } \end{aligned}\ ] ] the latter result is in complete correspondence to the finite strain setting in eq .. in conclusion : if we work already within the reduced format of @xmath283 } ) = \psi_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ) \ , \rightarrow \ , ( { \rm linearized \ , frame - indifference } ) \,,\end{aligned}\ ] ] then , isotropy is nothing else but form - invariance of the energy @xmath265 under the @xmath2-transformation .
it is this simple observation that will be subsequently made operable for linear higher - gradient elasticity .
the reader should carefully note that linearized elasticity , as we treat it here , is not frame - indifferent in the precise sense that it is not left - invariant under rotations in @xmath98 . by this
we mean that for @xmath284 one can write , formally , the linear elastic energy ( wether isotropic or anisotropic ) as @xmath285 } ) = w_{\rm lin}\left(\operatorname{sym } ( { { \rm grad}[{{{\mbox{$\varphi$ } } } } ] } - { { { \mathbbm{1 } } } } ) \right ) \,,\end{aligned}\ ] ] and in this representation , @xmath265 is not left - invariant under the @xmath98-action .
let us consider the lagrangian description of the displacement field @xmath286 with components @xmath287 in the basis @xmath86 at the referential position @xmath66 .
@xmath288[2cm][c]{12 cm } { we investigate transformation rules for the { \bf simultaneous } transformation of spatial and referential coordinates by the { \bf same } rigid rotation $ { { { q } } } \in { { \rm so}}(3)$ \begin{center } $ { { { u}}}^{\sharp}({{{q } } } \ , { { { x } } } ) \colonequals { { { q } } } \ , { { { u}}}({{{x } } } ) \ , , \quad { { \mbox{$\xi$ } } } : = { { { q } } } \ , { { { x } } } \ , , \quad { { { x } } } : = { { { q}}}^t \ , { { \mbox{$\xi$ } } } \ , $ \end{center } \begin{center } $ { { { { u}}}^\sharp ( { { \mbox{$\xi$}}})}= { { { q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \quad \leftrightarrow \quad { { { u } } } ( { { { x } } } ) = { { { q}}}^t\,{{{{u}}}^\sharp ( { { \mbox{$\xi$}}})}\ , . $ \end{center } } } \end{aligned}\ ] ] then the jacobian matrix is given by the transposition of the rotation tensor : @xmath289 the simultaneous rotation of spatial and referential coordinates leaves the energy content stored due to deformation of a finite sized ball invariant .
it depends only on the shape of the ellipse , not on its ( irrelevant ) inner structure nor its rotation in space , see fig .
[ objectivitypic2 ] .
( 130,110 ) ( 0,0 ) in this section the transformation of standard vectorial and tensorial functions is under investigation . as already seen in eq .
the gradient of the displacement field @xmath70 transforms via @xmath290 } = { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { { q } } } \ , { { { u } } } ( { { { q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } = \frac{\partial \ , [ { { { q } } } \ , { { { u } } } ( { { { q}}}^t \ , { { \mbox{$\xi$}}})]}{\partial \ , { { \mbox{$\xi$}}}}= { { { q } } } \ , \frac{\partial \ , { { { u } } } ( { { { x}}})}{\partial \ , { { { x } } } } \ , \frac{\partial \ , { { { x}}}}{\partial \ , { { \mbox{$\xi$}}}}= { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u } } } ( { { { x } } } ) } ] } \ , { { { q}}}^t \,.\end{aligned}\ ] ] the divergence of the displacement field yields @xmath291 & = \operatorname{tr}({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } ) = \operatorname{tr}({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } ) = \operatorname{tr}({{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t ) \notag \\ & = { \langle{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t , { { { \mathbbm{1}}}}\rangle } = { \langle { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } , { { { q}}}^t \ , { { { q}}}\rangle } = { \langle { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } , { { { \mathbbm{1}}}}\rangle}= \operatorname{tr}({{\rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } ) \notag \\ & = { { \rm div}_{{{{x}}}}}[{{{u}}}({{{x } } } ) ] \,,\end{aligned}\ ] ] which is independent of @xmath83 . however , the divergence of the displacement gradient reads @xmath292 } \ } & = ( { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$}}})}]}}]})_{ijk } \ , \delta_{jk } \ , { { { e}}}_i = ( { { \rm grad}_{{{\mbox{$\xi$ } } } } [ { { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t } ] } ) _ { ijk } \ , \delta_{jk } \ , { { { e}}}_i \notag \\ & = \left ( \frac{\partial \ , ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t)_{ij}}{\partial \ , x_a } \ , \underbrace{\frac{\partial \ , x_a}{\partial \ , \xi_k}}_{\displaystyle { { { q}}}^t } \right)_{ijk } \!\!\!\!\!\ ! \delta_{jk } \ , { { { e}}}_i = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t}]})_{ija } \
, \underbrace{({{{q}}}^t)_{ak } \ , \delta_{jk}}_{\displaystyle q^t_{aj } } \ , { { { e}}}_i \notag \\ & = ( q_{im } ( { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}})_{mb } \ , q^t_{bj})_{,a } \ , q^t_{aj } \ , { { { e}}}_i = q_{im } ( { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}}]})_{mba } \ , \underbrace{q^t_{bj } \ , q_{ja}}_{\displaystyle \delta_{ba } } \ , { { { e}}}_i \notag \\ & = q_{im } ( { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}}]})_{mbb } \ , { { { e}}}_i = { { { q } } } \ , { { \rm div}_{{{{x}}}}}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ }
\,.\end{aligned}\ ] ] the curl of the displacement gradient is given by definition for a second order tensor @xmath293 . from eq .
it follows @xmath294 } \ } & = { { \rm curl}_{{{\mbox{$\xi$}}}}}\ { { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } \ }
= 0 \ , .\end{aligned}\ ] ] of course , eq . is not a transformation rule since the curl of the gradient of the displacement field is generally zero .
however , the transformation rule for an arbitrary argument which is not the gradient of a vector field will be discussed in section [ kaptrafoenhanced ] . before , let us see that the transformation is compatible with the application of a sym or skew operator to the gradient of the displacement : @xmath295 } \ } & = \operatorname{sym}\{{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } \ } = \operatorname{sym}\{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t \ } \notag \\ & = \12 \{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t + { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x}}})}]}^t \ , { { { q}}}^t \ } = { { { q } } } \ , \operatorname{sym}\ { { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ } \ , { { { q}}}^t \,,\end{aligned}\ ] ] @xmath296 } \ } & = \skew \{{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } \ } = \skew \{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t \ } \notag \\ & = \12 \{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t - { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x}}})}]}^t \ , { { { q}}}^t \ } = { { { q } } } \ , \skew \ { { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ } \ , { { { q}}}^t \,.\end{aligned}\ ] ] from eq .
it follows that the trace of @xmath297}$ ] is invariant under the @xmath2-transformation , yielding @xmath298 } ) \ } & = \operatorname{tr}\ { { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } \ } = \operatorname{tr}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } \,,\\ \operatorname{tr}\ { \skew ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } ) \ } & = \operatorname{tr}\ { \skew ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ } = 0 \,.\end{aligned}\ ] ] thus , the deviator of the symmetric gradient of @xmath70 transforms accordingly as @xmath299 } ) & = { { { q } } } \ , \operatorname{sym}\ { { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ } \ , { { { q}}}^t - \frac{1}{3 } \ , \operatorname{tr}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } \ , { { { \mathbbm{1}}}}\notag \\ & = { { { q } } } \ , \ { \operatorname{sym } ( { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } ) - \frac{1}{3 } \ , \operatorname{tr}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } { { { \mathbbm{1}}}}\ } \ , { { { q}}}^t \notag \\ & = { { { q } } } \ , \ { \operatorname{dev}\operatorname{sym } ( { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } ) \ } \ , { { { q}}}^t \,.\end{aligned}\ ] ] the divergence of the symmetric and skew - symmetric part of @xmath297}$ ] then transforms according to @xmath300 } ) \ } & = { { \rm div}_{{{\mbox{$\xi$}}}}}\ { \operatorname{sym } ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } ) \ } = { { \rm div}_{{{\mbox{$\xi$}}}}}\ { \operatorname{sym } ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t ) \ }
\notag \\ & = { { \rm div}_{{{\mbox{$\xi$}}}}}\ { { { { q } } } \ , \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t \ } = { { { q } } } \ , { { \rm div}_{{{{x}}}}}\ { \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } \ , , \end{aligned}\ ] ] @xmath301 } ) \ } & = { { \rm div}_{{{\mbox{$\xi$}}}}}\ { \skew ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { u}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } ) \}= { { \rm div}_{{{\mbox{$\xi$}}}}}\ { \skew ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t ) \ } \notag \\ & = { { \rm div}_{{{\mbox{$\xi$}}}}}\ { { { { q } } } \ , \skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t ) \ } = { { { q } } } \ , { { \rm div}_{{{{x}}}}}\ { \skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } \,.\end{aligned}\ ] ] to investigate the axial vector @xmath302 of a skew symmetric tensor @xmath303 we use eq . and
consider the components of the tensor @xmath304 in the basis @xmath305 , reading @xmath306 since the functions axl and anti are free of a basis their results depend on the basis of the argument , which may be rotated : @xmath307 = q_{ak}(\operatorname{axl}[{{{a}}}])_k \ , { { { e}}}_a = -\12 \ , q_{ak } \ , a_{ij } \ , \epsilon_{ijk } \ , { { { e}}}_a = -\12 \ , a_{ij } \ , \underbrace{q_{ak } \ , \epsilon_{kij}}_{\mathclap{\displaystyle q_{bi } \ , q_{cj } \ , \epsilon_{bca } } } \ , { { { e}}}_a = -\12 \ , q_{bi } \ , a_{ij } \ , q^t_{jc } \ , \epsilon_{bca } \ , { { { e}}}_a \notag \\ & = \operatorname{axl}[\underbrace{{{{q } } } \ , { { { a } } } \ , { { { q}}}^t}_{\displaystyle { { { a}}}^\sharp } ] \qquad \leftrightarrow \qquad \operatorname{axl}[{{{a}}}^\sharp ] = { { { q } } } \ , \operatorname{axl}[{{{a } } } ] \ , .\end{aligned}\ ] ] two transformation rules can be found in eq .
: the transformation for the axial - operator and how the simultaneous transformation of spatial and referential coordinates transform a tensor , reading @xmath308 additionally , the transformation of @xmath309 is given by eq . , reading @xmath310 = \operatorname{anti}[{{{q } } } \ , { { { a } } } ] = \operatorname{anti}\operatorname{axl}[{{{a}}}^\sharp ] = \operatorname{anti}\operatorname{axl}[{{{q } } } \ , { { { a } } } \ , { { { q}}}^t ] = { { { q } } } \ , { { { a } } } \ , { { { q}}}^t \ , \qquad \leftrightarrow \qquad \operatorname{anti}[{{{a}}}^\sharp ] = { { { q } } } \ , \operatorname{anti}[{{{a } } } ] { { { q}}}^t \ , .\end{aligned}\ ] ] with help of eq . and
we obtain for the curl of the displacement field @xmath70 : @xmath311 & = 2 \operatorname{axl}\skew [ { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } ] = 2 \operatorname{axl}\skew [ { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x}}})}]}\ , { { { q}}}^t ] = 2 \operatorname{axl}[{{{q } } } \ , \skew { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \ , { { { q}}}^t ] \notag \\ & = { { { q } } } \ , [ 2 \operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } ] = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{{u}}}({{{x } } } ) ] \ , .\end{aligned}\ ] ] the gradient of the curl of the displacement field transforms via @xmath312 } ] } & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , \underbrace{{{\rm curl}_{{{{x}}}}}{{{u}}}({{{q}}}^t \ , { { \mbox{$\xi$}}})}_{\displaystyle \colonequals { { { v}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } } ] } = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { v}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{v}}}({{{x } } } ) } ] } \ , { { { q}}}^t \notag \\ & = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\rm curl}_{{{{x}}}}}{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t \ , .\end{aligned}\ ] ] thus , the transformation of spatial and referential coordinates by the same rigid rotation yields the curvature @xmath313}$ ] from eq . to transform via @xmath314 } = { { { q } } } \ , \underbrace{\12 \ , { { \rm grad}_{{{{x}}}}[{{{\rm curl}_{{{{x}}}}}{{{u}}}({{{x}}})}]}}_{\displaystyle = \ , { \widetilde{{{{k}}}}}({{{x } } } ) } \ , { { { q}}}^t= { { { q } } } \ , { \widetilde{{{{k}}}}}({{{x } } } ) \ , { { { q}}}^t \ , .\end{aligned}\ ] ] the symmetric and skew symmetric part of @xmath315 transform as @xmath316 = \12 \ , [ { \widetilde{{{{k}}}}}^\sharp({{\mbox{$\xi$ } } } ) + ( { \widetilde{{{{k}}}}}^\sharp({{\mbox{$\xi$}}}))^t ] = { { { q } } } \ , \12 \ , [ { \widetilde{{{{k}}}}}({{{x } } } ) + ( { \widetilde{{{{k}}}}}({{{x}}}))^t ] \ , { { { q}}}^t = { { { q } } } \ , \operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] \ , { { { q } } } ^t \,,\end{aligned}\ ] ] @xmath317 = \12 \ , [ { \widetilde{{{{k}}}}}^\sharp({{\mbox{$\xi$ } } } ) - ( { \widetilde{{{{k}}}}}^\sharp({{\mbox{$\xi$}}}))^t ] = { { { q } } } \ , \12 \ , [ { \widetilde{{{{k}}}}}({{{x } } } ) -({\widetilde{{{{k}}}}}({{{x}}}))^t ] \ , { { { q}}}^t = { { { q } } } \ , \skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] \ , { { { q } } } ^t \,.\end{aligned}\ ] ] since the trace of @xmath318 is zero , it is independent of @xmath83 with @xmath319 . for the transformations in this section we often use the cauchy stress tensor @xmath244 as the argument , which is a second order tensor .
the divergence lowers a tensor of second order , transforming according to @xmath320 & = { { \rm div}_{{{\mbox{$\xi$ } } } } } [ { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t ] = \left ( \frac{\partial \ , ( { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t)_{ij}}{\partial \ , x_a } \ , \frac{\partial \ , x_a}{\partial \ , \xi_k}\right)_{ijk } \!\!\!\!\ !
\delta_{jk } \ , { { { e}}}_i = q_{ib } \ , \sigma_{bc , a } \ , q^t_{cj } \ , q^t_{ak } \ , \delta_{jk } \ , { { { e}}}_i \notag \\ & = q_{ib } \ , \sigma_{bc , a } \ , \underbrace{q^t_{cj } \ , q_{ja}}_{\displaystyle \delta_{ca } } \ , { { { e}}}_i = q_{ib } \ , \sigma_{ba , a } \ , { { { e}}}_i = { { { q } } } \ , { { \rm div}_{{{{x } } } } } [ { { { \mbox{$\sigma$}}}({{{x } } } ) } ] \,.\end{aligned}\ ] ] in contrast , the gradient lifts a second order tensor to a tensor of third order transforming via @xmath321 } & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t } ] } = \frac{\partial \ , ( { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t)_{ij}}{\partial \ , x_a } \ , \frac{\partial \ , x_a}{\partial \ , \xi_k } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t}]})_{ija } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = ( q_{ib } \ , \sigma_{bc } \ , q^t_{cj})_{,a } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = q_{ib } \ , \sigma_{bc , a } \ , q^t_{cj } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}{{{q}}}^t } ] } \ , { { { q}}}^t \,.\end{aligned}\ ] ] since the divergence lowers the order of @xmath322}$ ] to second order , one obtains @xmath323 } ] & = { { \rm div}_{{{\mbox{$\xi$ } } } } } [ { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \mbox{${{{\mbox{$\sigma$}}}({{{x}}})}$ } } } \ , { { { q}}}^t } ] } \ , { { { q}}}^t ] \notag \\ & = \left ( \frac{\partial \ , ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t } ] } \ , { { { q}}}^t)_{ijk}}{\partial \ , x_n } \ , \frac{\partial \ , x_n}{\partial \ , \xi_m } \right)_{ijkm } \!\!\ ! \delta_{km } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t } ] } \ , { { { q}}}^t}]})_{ijkn } \ , q^t_{nm } \ , \delta_{km } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = [ ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t}]})_{ija } \ , q^t_{ak}]_{,n } \ , q_{kn } \ , { { { e}}}_i \otimes { { { e}}}_j \notag\\ & = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t}]})_{ija , n } \ , \underbrace{{{{q}}}^t_{ak } \ , q_{kn}}_{\displaystyle \delta_{an } } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t}]})_{ijn , n } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = ( { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t)_{ij , nn } \ , { { { e}}}_i \otimes { { { e}}}_j = q_{ia } \ , ( { { { \mbox{$\sigma$}}}({{{x}}})})_{ab , nn } \ , q^t_{bj } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = { { \rm div}_{{{{x}}}}}[{{\rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \mbox{${{{\mbox{$\sigma$}}}({{{x}}})}$ } } } { { { q}}}^t } ] } ] = { { { q } } } \ , ( { { \rm div}_{{{{x}}}}}{{\rm grad}_{{{{x}}}}[{{{{\mbox{$\sigma$}}}({{{x } } } ) } } ] } ) \ , { { { q}}}^t \,.\end{aligned}\ ] ] similarly , the gradient of the divergence of @xmath244 yields @xmath324 } & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { \rm div}_{{{{x}}}}}{{{\mbox{$\sigma$}}}({{{x } } } ) } } ] } = \frac{\partial \ , ( { { { q } } } \ , { { \rm div}_{{{{x}}}}}{{{\mbox{$\sigma$}}}({{{x}}})})}{\partial \ , { { { x } } } } \ , \frac{\partial \ , { { { x}}}}{\partial \ , { { \mbox{$\xi$ } } } } = { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm div}_{{{{x}}}}}{{{\mbox{$\sigma$}}}({{{x } } } ) } } ] } \ , { { { q}}}^t \notag \\ & = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\rm div}_{{{{x}}}}}{{{\mbox{$\sigma$}}}({{{x } } } ) } } ] } \ , { { { q}}}^t \,.\end{aligned}\ ] ] next , let us discuss the transformation of the curl operator for arbitrary second order tensors in @xmath325 . therefore , we consider a generally non - symmetric field @xmath326 . the transformation to rotated coordinates
is given , similarly to the gradient @xmath297}$ ] , see eq . , by @xmath327 .
then @xmath328 = { { \rm curl}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { p}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t ] = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{{{p}}}({{{x } } } ) } ] \ , { { { q}}}^t \,.\end{aligned}\ ] ] this last formula can be seen in several ways .
first , with the help of eq .
we obtain @xmath329 & = { { \rm curl}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { p}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t ] = - \left ( \frac{\partial \ , ( { { { q } } } \ , { { { { p}}}({{{x}}})}\ , { { { q}}}^t)_{ia}}{\partial \ , x_c } \ , \frac{\partial \ , x_c}{\partial \ , \xi_b } \right)_{iab } \ , \epsilon_{abj } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = - ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { { { p}}}({{{x}}})}\ , { { { q}}}^t}]})_{iac } \ , q^t_{cb } \ , \epsilon_{abj } \ , { { { e}}}_i \otimes { { { e}}}_j = - ( { { { q } } } \ , { { { { p}}}({{{x}}})}\ , { { { q}}}^t)_{ia , c } \ , q^t_{cb } \ , \epsilon_{abj } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = - q_{im } \ , p_{mn , c } \ , q^t_{na } \ , q^t_{cb } \ , \epsilon_{abj } \ , { { { e}}}_i \otimes { { { e}}}_j = - q_{im } \ , p_{mn , c } \ , \underbrace{q_{an } \ , q_{bc } \ , \epsilon_{abj}}_{\displaystyle q_{jp } \ , \epsilon_{pnc } } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = q_{im } \ , \underbrace{(- p_{mn , c } \ , \epsilon_{ncp})}_{\displaystyle ( { { \rm curl}_{{{{x}}}}}[{{{{p}}}({{{x}}})}])_{mp } } \ , q^t_{pj } \ , { { { e}}}_i \otimes { { { e}}}_j = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{{p}}}({{{x } } } ) ] \ , { { { q}}}^t \,.\end{aligned}\ ] ] alternatively , from @xcite , p. 318
, we have the transformation law @xmath330 = ( { { \rm curl}_{{{{x}}}}}[{{{p}}}({{{x } } } ) ] ) \ , { { { q}}}^t \,.\end{aligned}\ ] ] combining eq .
with @xmath331 = \operatorname{lin}_{{{{p}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } \underbrace{({{{\mbox{$\nabla$}}}_{\ ! { { \mbox{$\xi$}}}}}\ , { { { q}}})}_{\displaystyle = \ , 0 } + { { { q } } } \ , { { \rm curl}_{{{\mbox{$\xi$}}}}}[{{{p}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) ] \,,\end{aligned}\ ] ] where @xmath332 is a linear operator , we obtain formally @xmath333 = { { { q } } } \ , { { \rm curl}_{{{\mbox{$\xi$}}}}}[{{{p}}}({{{x } } } ) \ , { { { q}}}^t ] = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{{p}}}({{{x } } } ) ] \ , { { { q}}}^t \,.\end{aligned}\ ] ] obviously , repeating curl operators yield the same structure of transformation as in eq . , e.g. @xmath334 & = { { \rm curl}_{{{\mbox{$\xi$ } } } } } [ { { { q } } } \ , \underbrace{{{\rm curl}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x}}})}]}_{\displaystyle \colonequals { { { { p}}}({{{x } } } ) } } \ , { { { q}}}^t ] = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{{{p}}}({{{x } } } ) } ] \ , { { { q}}}^t = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}[{{\rm curl}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x } } } ) } ] ] \ , { { { q}}}^t \,.\end{aligned}\ ] ] note that the curl of a second order tensor field has generally no divergence @xmath335 ) & = \div ( - \sigma_{ia , b } \ , \epsilon_{abj } \ , { { { e}}}_i \otimes { { { e}}}_j ) = - \sigma_{ia , bk } \ , \epsilon_{abj } \ , \delta_{jk } \ , { { { e}}}_i = - \sigma_{ia , bk } \ , \epsilon_{abk } \ , { { { e}}}_i \notag \\ & = - ( \sigma_{i1,23}-\sigma_{i1,32}+\sigma_{i2,31}-\sigma_{i2,13}+\sigma_{i3,12}-\sigma_{i3,21 } ) \ , { { { e}}}_i = 0 \,,\end{aligned}\ ] ] which is independent of @xmath83 , reading @xmath336 ) = { { \rm div}_{{{{x}}}}}({{\rm curl}_{{{{x}}}}}[{{\mbox{${{{\mbox{$\sigma$}}}({{{x}}})}$ } } } ] ) = 0 $ ] .
+ let us reformulate the result in eq . by using @xmath297}$ ] instead of
@xmath244 as second order argument : @xmath337 } } ] } & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t } ] } \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t}]})_{ija } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = ( q_{ib } \ , u_{b , c } \ , q^t_{cj})_{,a } \ ,
q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = q_{ib } \ , u_{b , ca } \ , q^t_{cj } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t } ] } ) \ , { { { q}}}^t \,.\end{aligned}\ ] ] then , lowering the tensorial order of @xmath338 } } ] } $ ] by the divergence yields @xmath339}}]}\ } & = { { \rm div}_{{{\mbox{$\xi$}}}}}\{{{\rm grad}_{{{{x}}}}[{\underbrace{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}\ , { { { q}}}^t}_{\displaystyle \colonequals { { { { p}}}({{{x}}})}}}]}\,{{{q}}}^t \ } \notag \\ & = \left ( \frac{\partial \ , ( { { \rm grad}_{{{{x}}}}[{{{{{p}}}({{{x } } } ) } } ] } \ , { { { q}}}^t)_{ijk}}{\partial \ , x_a } \ , \frac{\partial \ , x_a}{\partial \ , \xi_l } \right)_{ijkl } \ , \delta_{kl } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = ( { { \rm grad}_{{{{x}}}}[{{{{{p}}}({{{x}}})}}]})_{ijb , a } \ , \underbrace{q^t_{bk } \ , q^t_{al } \ , \delta_{kl}}_{\displaystyle \delta_{ba } } \ , { { { e}}}_i \otimes { { { e}}}_j = ( { { \rm grad}_{{{{x}}}}[{{{{{p}}}({{{x}}})}}]})_{ija , a } \ , { { { e}}}_i \otimes { { { e}}}_j \notag \\ & = [ { { { { p}}}({{{x}}})}]_{ij , aa } \ , { { { e}}}_i \otimes { { { e}}}_j = { { \rm div}_{{{{x}}}}}\{{{\rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}\ , { { { q}}}^t } ] } \ } \,.\end{aligned}\ ] ] since @xmath83 is a constant rotation tensor it is possible to reformulate the result of eq . into @xmath340}}]}\ } & = { { \rm div}_{{{{x}}}}}\{{{\rm grad}_{{{{x}}}}[{{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}\ , { { { q}}}^t } ] } \ } = [ { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}\ , { { { q}}}^t]_{ij , aa } \notag \\ & = ( q_{ib } \ , ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]})_{bc } \ , q_{cj}^t ) _ { , aa } \ , { { { e}}}_i \otimes { { { e}}}_j = q_{ib } \ , ( { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]})_{bc , aa } \ , q_{cj}^t \notag \\ & = { { { q } } } \ , { { \rm div}_{{{{x}}}}}\{{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } \ } { { { q}}}^t = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\rm div}_{{{{x}}}}}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } } ] } { { { q}}}^t \,.\end{aligned}\ ] ] finally , with help of eq .
let us lower the tensorial order of @xmath341 } } ] } $ ] by the divergence again to find @xmath342}}]}\ } ) & = { { \rm div}_{{{\mbox{$\xi$ } } } } } ( { { { q } } } \ , \underbrace{{{\rm grad}_{{{{x}}}}[{{{\rm div}_{{{{x}}}}}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } } ] } } _ { \displaystyle \colonequals { { { { p}}}({{{x } } } ) } } { { { q}}}^t ) \notag \\ & = \left ( \frac{\partial \ , ( { { { q } } } \ , { { { { p}}}({{{x}}})}\ , { { { q}}}^t)_{ij}}{\partial \ , x_a } \ , \frac{\partial \ , x_a}{\partial \ , \xi_k } \right)_{ijk } \ , \delta_{jk } \ , { { { e}}}_i = ( { { { q } } } \ , { { { { p}}}({{{x}}})}\ , { { { q}}}^t)_{ij , a } \ , q^t_{ak } \ , \delta_{jk } \ , { { { e}}}_i \notag \\ & = q_{ib } \ , ( { { { { p}}}({{{x}}})})_{bc , a } \ , q^t_{cj } \ , q_{ja } \ , { { { e}}}_i = q_{ib } \ , ( { { { { p}}}({{{x}}})})_{bc , a } \ , \delta_{ca } \ , { { { e}}}_i = q_{ib } \ , ( { { { { p}}}({{{x}}})})_{ba , a } \ , { { { e}}}_i\notag \\ & = { { { q } } } \ , { { \rm div}_{{{{x}}}}}({{{{p}}}({{{x } } } ) } ) = { { { q } } } \ , { { \rm div}_{{{{x}}}}}({{\rm grad}_{{{{x}}}}[{{{\rm div}_{{{{x}}}}}\ { { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ } } ] } ) \,.\end{aligned}\ ] ] in this subsection we consider the question of form - invariance of the balance equations . to start these considerations ,
we consider the laplace equation .
note that the laplacian @xmath343 of a scalar field @xmath344 is * form - invariant * concerning our simultaneous transformation of spatial and referential coordinates by the same rigid rotation : @xmath345 } = { { \rm div}_{{{\mbox{$\xi$}}}}}\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{h({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } = { { \rm div}_{{{\mbox{$\xi$}}}}}\left [ \frac{\partial \ , h({{{q}}}^t \ , { { \mbox{$\xi$}}})}{\partial { { \mbox{$\xi$ } } } } \right ] = { { \rm div}_{{{\mbox{$\xi$}}}}}\left [ \frac{\partial \ , h({{{x}}})}{\partial x_i } \ , \frac{\partial \ , x_i}{\partial \ , \xi_j } \ , { { { e}}}_j \right ] \notag \\ & = { { \rm div}_{{{\mbox{$\xi$}}}}}[h_{,i } \ , q^t_{ij } \ , { { { e}}}_j ] = { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } ] = { \langle\frac{\partial \ , ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{h({{{x}}})}]})}{\partial \ , { { { x } } } } \ , \frac{\partial \ , { { { x}}}}{\partial \ , { { \mbox{$\xi$ } } } } , { { { \mathbbm{1}}}}\rangle } \notag \\ & = { \langle{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } } ] } \ , { { { q}}}^t , { { { \mathbbm{1}}}}\rangle } = { \langle{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } } ] } , { { { q}}}^t \ , { { { q}}}\rangle } = { \langle{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } } ] } , { { { \mathbbm{1}}}}\rangle } \notag \\ & = \operatorname{tr}{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } } ] } = { { \rm div}_{{{{x}}}}}\ , { { \rm grad}_{{{{x}}}}[{h({{{x } } } ) } ] } = { \mathop{}\!\mathbin\bigtriangleup}_{{{{x}}}}h({{{x } } } ) \ , .\end{aligned}\ ] ] therefore , the laplace - equation is form - invariant with respect to the @xmath2-transformation for scalars } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \ , { \,{\rm d}}x \ , , \quad h : \ , \operatorname{\mathbb{r}^3}\rightarrow \operatorname{\mathbb{r}}\ , .\end{aligned}\ ] ] consider @xmath346 .
then @xmath347}$ ] is a row - vector and it holds @xmath348 } = { { \rm grad}_{{{{x}}}}[{h(q^t \ , \xi ) } ] } \ , q^t \ , .\end{aligned}\ ] ] moreover , @xmath349 } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } & = \frac{1}{2 } \ , { \|\ , ( { { \rm grad}_{{{\mbox{$\xi$}}}}[{h^{\flat}(\xi)}]})^t \ , \|}^2_{\operatorname{\mathbb{r}^3 } } = \frac{1}{2 } \ , { \|\ , ( { { \rm grad}_{{{{x}}}}[{h(q^t \ , \xi ) } ] } \ , q^t)^t \ ,
\|}^2_{\operatorname{\mathbb{r}^3 } } = \frac{1}{2 } \ , { \|\ , q \ , ( { { \rm grad}_{{{{x}}}}[{h(q^t \ , \xi)}]})^t \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \notag \\ & = \frac{1}{2 } \ , { \|\ , ( { { \rm grad}_{{{{x}}}}[{h(x)}]})^t \ , \|}^2_{\operatorname{\mathbb{r}^3 } } = \frac{1}{2 } \ , { \|\ , { { \rm grad}_{{{{x}}}}[{h(x ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \ , .\end{aligned}\ ] ] therefore @xmath350 } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \ , { \,{\rm d}}x = \int_{\xi \in b } \frac{1}{2 } \ , { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{h^{\flat}(\xi ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \ , \det q { \,{\rm d}}\xi = \int_{\xi \in b } \frac{1}{2 } \ , { \|\ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{h^{\flat}(\xi ) } ] } \ , \|}^2_{\operatorname{\mathbb{r}^3 } } \ , { \,{\rm d}}\xi \,,\end{aligned}\ ] ] which is form - invariance and in terms of our previous terminology right - global so(3)-invariance . ] , as is well - known . + the classical balance of linear momentum in linear
elasticity includes the divergence of the cauchy stress @xmath244 and the net force @xmath351 which both transform via @xmath352 + { { { f}}}^{\sharp}({{\mbox{$\xi$ } } } ) & = & 0 \ , , \qquad \qquad { { { f}}}^{\sharp}(\ , { { \mbox{$\xi$ } } } ) \colonequals { { { q } } } \ , { { { f}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { \mbox{$\sigma$}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t ] + { { { q } } } \ , { { { f}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { { q } } } \ , { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x } } } ) } ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { { q } } } \ , \ { { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x } } } ) } ] + { { { f}}}({{{x } } } ) \ } & = & 0 \vspace{0.2 cm } \\
\leftrightarrow \quad & { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x } } } ) } ] + { { { f}}}({{{x } } } ) & = & 0 \ , . \end{array}}\ ] ] therefore , the linear elasticity equations are form - invariant with respect to the @xmath2-transformation .
the transformation with @xmath83 from the left also holds for the divergence of second order couple stress @xmath353 concerning its symmetric part @xmath354 ) = { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , \operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] \ , { { { q}}}^t ] = { { { q } } } \ , { { \rm div}_{{{{x}}}}}(\operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \,,\end{aligned}\ ] ] and also concerning its skew - symmetric part @xmath355 ) = { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , \skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] \ , { { { q}}}^t ] = { { { q } } } \ , { { \rm div}_{{{{x}}}}}(\skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \,.\end{aligned}\ ] ] in the isotropic indeterminate couple stress theory the nonlocal force stress is given by @xmath356 and adds to the local force stress yielding the total stress @xmath357 . since @xmath358
we find with help of eq .
that its transformation reads for the symmetric part @xmath359)\ } = \operatorname{anti}\{{{{q } } } \,{{\rm div}_{{{{x}}}}}\ , ( \operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } = { { { q } } } \ , \operatorname{anti}\ { { { \rm div}_{{{{x}}}}}(\operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } \ , { { { q}}}^t \,,\end{aligned}\ ] ] and similarly for its skew - symmetric part @xmath360)\ } = \operatorname{anti}\{{{{q } } } \,{{\rm div}_{{{{x}}}}}\ , ( \skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } = { { { q } } } \ , \operatorname{anti}\ { { { \rm div}_{{{{x}}}}}(\skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } \ , { { { q}}}^t \,.\end{aligned}\ ] ] thus , the divergence of the nonlocal stress contribution transforms like the local stress via @xmath361)\ } ] & = { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \,\operatorname{anti}\{{{\rm div}_{{{{x}}}}}\ , ( \operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \}\,{{{q}}}^t ] \notag \\ & = { { { q } } } \,{{\rm div}_{{{{x}}}}}[\operatorname{anti}\ { { { \rm div}_{{{{x}}}}}(\operatorname{sym}[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } ] \,,\end{aligned}\ ] ] and @xmath362)\ } ] & = { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \,\operatorname{anti}\{{{\rm div}_{{{{x}}}}}\ , ( \skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \}\,{{{q}}}^t ] \notag \\ & = { { { q } } } \,{{\rm div}_{{{{x}}}}}[\operatorname{anti}\ { { { \rm div}_{{{{x}}}}}(\skew[{\widetilde{{{{k}}}}}({{{x } } } ) ] ) \ } ] \,.\end{aligned}\ ] ] this can also be seen if we directly write the ( extended ) balance of linear momentum of the indeterminate couple stress model via @xmath363 \ } + { { { f}}}^{\sharp}({{\mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\
\leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}\ { { { { q } } } \ , { { \mbox{$\sigma$}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t - \12 \ , \operatorname{anti}{{\rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { m}}}({{{x } } } ) \ , { { { q}}}^t ] \ } + { { { q } } } \ , { { { f}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}\ { { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t - \12 \ , \operatorname{anti}({{{q } } } \ , { { \rm div}_{{{{x } } } } } [ { { { m}}}({{{x } } } ) ] ) \ } + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}\ { { { { q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t - \12 \ , { { { q } } } \ , ( \operatorname{anti}{{\rm div}_{{{{x}}}}}[{{{m}}}({{{x } } } ) ] ) \ , { { { q}}}^t \ } + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { { q } } } \ , { { \rm div}_{{{{x}}}}}\ { { { { \mbox{$\sigma$}}}({{{x}}})}- \12 \ , \operatorname{anti}{{\rm div}_{{{{x}}}}}[{{{m}}}({{{x } } } ) ] \ } + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{{x}}}}}\ { { { { \mbox{$\sigma$}}}({{{x}}})}- \12 \ , \operatorname{anti}{{\rm div}_{{{{x}}}}}[{{{m}}}({{{x } } } ) ] \ } + { { { f}}}({{{x } } } ) & = & 0 \ , . \end{array}}\ ] ] therefore , the balance equations of the indeterminate couple stress model are form - invariant with respect to the @xmath2-transformation . + alternatively , the hyperstress tensor @xmath11 of * third order * may appear in higher gradient theories via @xmath364 } } ] } = \l.{\rm d}_{{{\mbox{$\xi$}}}}^2 { { { { u}}}^\sharp ( { { \mbox{$\xi$}}})}}$ ] , where @xmath365 is a 6th order tensor
. we only consider @xmath365 to map like the identity @xmath366 for any @xmath367 for simplicity .
thus , we obtain from eq . the transformation @xmath368 & = { { \rm div}_{{{\mbox{$\xi$}}}}}{{\rm div}_{{{\mbox{$\xi$}}}}}\{{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$}}})}}]}}]}\ } = { { { q } } } \ , { { \rm div}_{{{{x}}}}}({{\rm grad}_{{{{x}}}}[{{{\rm div}_{{{{x}}}}}\{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}\ } } ] } ) \notag \\ & = { { { q } } } \ , { { \rm div}_{{{{x}}}}}{{\rm div}_{{{{x}}}}}\underbrace{\{{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}}]}\}}_{\displaystyle = { \mathfrak{m}}({{{x } } } ) } \notag \\ & = { { { q } } } \ , { { \rm div}_{{{{x}}}}}\{{{\rm div}_{{{{x}}}}}[{\mathfrak{m}}({{{x}}})]\ } \,.\end{aligned}\ ] ] hence the balance of linear momentum in a full higher gradient theory transforms according to @xmath369 ] + { { { f}}}^{\sharp}({{\mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\
\leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{\mbox{$\sigma$}}}^{\sharp}({{\mbox{$\xi$ } } } ) } ] + { { { f}}}^{\sharp}({{\mbox{$\xi$ } } } ) + { { \rm div}_{{{\mbox{$\xi$}}}}}\ , { { \rm div}_{{{\mbox{$\xi$}}}}}[{\mathfrak{m}}^\sharp({{\mbox{$\xi$ } } } ) ] & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { \mbox{$\sigma$}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t ] + { { { q } } } \ , { { { f}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) + { { \rm div}_{{{\mbox{$\xi$}}}}}\ , { { \rm div}_{{{\mbox{$\xi$}}}}}[{\mathfrak{m}}^\sharp({{\mbox{$\xi$ } } } ) ] & = & 0 \vspace{0.2 cm } \\
\leftrightarrow \quad & { { { q } } } \ , { { \rm div}_{{{{x}}}}}[{{\mbox{$\sigma$}}}({{{x } } } ) ] + { { { q } } } \ , { { { f}}}({{{x } } } ) + { { { q } } } \ , \ { { { \rm div}_{{{{x}}}}}\ , { { \rm div}_{{{{x}}}}}[{\mathfrak{m}}({{{x } } } ) ] \ } & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{{x}}}}}[{{\mbox{$\sigma$}}}({{{x } } } ) + { { \rm div}_{{{{x}}}}}{\mathfrak{m}}({{{x } } } ) ] + { { { f}}}({{{x } } } ) & = & 0 \ , . \end{array}}\ ] ] therefore , the balance equations of the higher gradient model are form - invariant with respect to the @xmath2-transformation .
+ finally , we show that the form invariance condition for the system of eq . in a variant of the indeterminate couple stress theory @xcite always holds true : @xmath370 + { { { f}}}^{\sharp}({{\mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\
\leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { \mbox{$\sigma$}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t + \operatorname{sym}{{\rm curl}_{{{\mbox{$\xi$}}}}}({{{q } } } \ , { { { m}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) \ , { { { q}}}^t ) ] + { { { q } } } \ , { { { f}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t + \operatorname{sym}\ { { { { q } } } \ , { { \rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) \ , { { { q}}}^t \ } ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , { { { \mbox{$\sigma$}}}({{{x}}})}\ , { { { q}}}^t + { { { q } } } \ , \operatorname{sym}{{\rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) \ , { { { q}}}^t ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{\mbox{$\xi$}}}}}[{{{q } } } \ , \ { { { { \mbox{$\sigma$}}}({{{x}}})}+ \operatorname{sym}{{\rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) \ } \ , { { { q}}}^t ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0\vspace{0.2 cm } \\ \leftrightarrow \quad & { { { q } } } \ , { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x}}})}+ \operatorname{sym}{{\rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) ] + { { { q } } } \ , { { { f}}}({{{x } } } ) & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { { q } } } \ , \ { { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x}}})}+ \operatorname{sym}{{\rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) ] + { { { f}}}({{{x } } } ) \ } & = & 0 \vspace{0.2 cm } \\ \leftrightarrow \quad & { { \rm div}_{{{{x}}}}}[{{{\mbox{$\sigma$}}}({{{x}}})}+ \operatorname{sym}{{\rm curl}_{{{{x } } } } } ( { { { m}}}({{{x } } } ) ) ] + { { { f}}}({{{x } } } ) & = & 0 \ , . \end{array}}\ ] ]
now we turn back to the invariance considerations in sections [ kapobjnonlinearelas ] and [ kapobjlinearelas ] .
motivated by the insights obtained there , we require * form - invariance * of the * energy expression * also in higher gradient or couple - stress models . as a primer ,
consider the higher order curvature expression @xmath371 } ) \|^2 \ , .\end{aligned}\ ] ] linearized objectivity in eq
. is clear , since the curvature is defined in terms of linear combinations of partial derivatives of @xmath0}$ ] .
moreover , since @xmath372}\ } ) & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{axl}\skew { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } } ] } = { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{axl}\skew ( { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t ) } ] } \notag \\ & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{axl}({{{q } } } \ , ( \skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , { { { q}}}^t ) } ] } = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , \underbrace { ( \operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]})}_{\displaystyle \colonequals { { { v}}}({{{x } } } ) } } ] } \notag \\ & = { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{{q } } } \ , { { { v}}}({{{q}}}^t \ , { { \mbox{$\xi$ } } } ) } ] } = { { { q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{v}}}({{{x } } } ) } ] } \ , { { { q}}}^t = { { { q } } } \ , ( { { \rm grad}_{{{{x}}}}[{\operatorname{axl}\skew { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) \notag \\ & = { { { q } } } \ , { { \rm curl}_{{{{x}}}}}(\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , { { { q}}}^t \,,\end{aligned}\ ] ] eq . yields @xmath373}\ } ) = ( { \widetilde{{{{k}}}}}^{\sharp } ( { { \mbox{$\xi$}}}))^t = { { { q } } } \ , ( { \widetilde{{{{k}}}}}({{{x}}}))^t \ , { { { q}}}^t\,.\end{aligned}\ ] ] thus , the curvature energy in eq . is form - invariant since @xmath374}\ } ) \|^2 & = \| \operatorname{sym}({\widetilde{{{{k}}}}}^{\sharp } ( { { \mbox{$\xi$ } } } ) \|^2 = \| \operatorname{sym}{{{q } } } \ , { \widetilde{{{{k}}}}}({{{x } } } ) \ , { { { q}}}^t \|^2 = \| \operatorname{sym}{\widetilde{{{{k}}}}}({{{x } } } ) \|^2 \notag \\ & = \| \operatorname{sym}{{\rm curl}_{{{{x}}}}}\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{u}}}({{{x } } } ) } ] } \|^2 \,.\end{aligned}\ ] ] here , we consider ( linearized ) frame - indifferent second gradient elasticity written in the reduced format } , \ , { { \rm grad}[{{{\rm grad}[{u}]}}]})$ ] is also ( linearized ) frame - indifferent . the reduction to @xmath375 } , \ , { { \rm grad}[{\operatorname{sym}{{\rm grad}[{u}]}}]})$ ] is possible since any second derivative @xmath376}}]}$ ] is a linear combination of gradients of strain @xmath377}$ ] .
however , there remains a subtle difference between @xmath375 } , \ , { { \rm grad}[{{{\rm grad}[{u}]}}]})$ ] and @xmath375 } , \ , { { \rm grad}[{\operatorname{sym}{{\rm grad}[{u}]}}]})$ ] in that formulations based on either expression may differ in the higher order traction boundary conditions while the energies coincide , see subsection [ kapproblemcompinv ] . ]
@xmath378 } , \ , { { \rm grad}[{{{\rm grad}[{u } ] } } ] } ) = w(\operatorname{sym}{{\rm grad}[{u } ] } , \ , { { \rm grad}[{\operatorname{sym}{{\rm grad}[{u } ] } } ] } ) \,,\end{aligned}\ ] ] where @xmath101 is a quadratic form . assuming the natural additive split @xmath379 } ,
\ , { { \rm grad}[{\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } } ] } ) = w_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ) + w_{\rm curv}({{\rm grad}[{\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } } ] } ) \,,\end{aligned}\ ] ] we are interested in further reduced forms for isotropy .
the reduced format for @xmath265 is nothing but isotropic linear elasticity , in which isotropy is in effect the requirement of form - invariance of @xmath380 } ) \overbrace{=}^{\rm form - invariant } w_{\rm lin}(\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \,,\end{aligned}\ ] ] under the transformation . from eq .
we obtain the standard condition @xmath381 } \ , { { { q}}}^t ) = w_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ) \qquad \forall \ , { { { q } } } \in { { \rm so}}(3 ) \,,\end{aligned}\ ] ] which implies , as is well - known , the representation @xmath382 } ) & = \mu \ , \| \operatorname{sym}{{\rm grad}[{{{{u } } } } ] } \|^2 + \frac{\lambda}{2 } \ , [ \operatorname{tr}\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ] ^2 \notag\\ & = \mu \ , \|\operatorname{dev}\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } \|^2 + \frac{\kappa}{2 } \ , [ \operatorname{tr}\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } ] ^2 \,,\end{aligned}\ ] ] where @xmath263 , @xmath264 are the classical lam - constants , and @xmath1 is the bulk modulus .
+ for the general curvature term in eq .
we require for isotropy the same formal form - invariance as in eq .
, namely @xmath383 } } ] } ) \overbrace{=}^{\rm form - invariant } w_{\rm curv}({{\rm grad}_{{{{x}}}}[{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) \,.\end{aligned}\ ] ] from eq . and
we obtain @xmath384 } } ] } = { { \rm grad}_{{{{x}}}}[{\operatorname{sym}{{{q } } } \ , { { \rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , { { { q}}}^t } ] } \ , { { { q}}}^t \ , , \end{aligned}\ ] ] yielding @xmath385 } \ , { { { q}}}^t } ] } \ , { { { q}}}^t ) = w_{\rm curv}({{\rm grad}_{{{{x}}}}[{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}}]})\end{aligned}\ ] ] in this general case of a strain gradient model . +
in the case of the indeterminate couple stress model with symmetric total stress tensor we have that @xmath386 } } ] } ) = \widehat w_{\rm curv}\left ( ( { { \rm curl}_{{{{x}}}}}{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}})^t \right ) \,,\end{aligned}\ ] ] and we obtain similarly the form - invariance condition @xmath387 } } ) ^t ) \overbrace{=}^{\rm form - invariant } \widehat w_{\rm curv}(({{\rm curl}_{{{{x}}}}}{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}})^t ) \,,\end{aligned}\ ] ] which implies @xmath388 } } ) ^t \ , { { { q}}}^t ) = \widehat w_{\rm curv}(({{\rm curl}_{{{{x}}}}}{\ , \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x}}})}}]}})^t ) \,.\end{aligned}\ ] ] if this is true for all @xmath14 as in it can be concluded that @xmath389 } } ) ^t}_{\displaystyle { \widetilde{{{{k}}}}}({{{x } } } ) } ) = \mu \ , l_c^2 \ , \ { \alpha_1 { \|\ , \operatorname{dev}\operatorname{sym}{\widetilde{{{{k}}}}}({{{x } } } ) \ , \|}^2 + \alpha_2 \ , { \|\ , \skew { \widetilde{{{{k}}}}}({{{x } } } ) \ , \|}^2 + \alpha_3 \ , [ \operatorname{tr}{\widetilde{{{{k}}}}}({{{x}}})]^2 \}\end{aligned}\ ] ] with nondimensional constants @xmath390 , @xmath391 , @xmath392 and we note that @xmath393 .
@xmath394[5cm][c]{14.5 cm } { the linear second - gradient hyperelastic formulation is ( linearized ) frame - indifferent and isotropic { \bf if and only if } there exists a function $ \widehat{\psi}_{\rm lin } : { { \rm{sym } } } ( 3 ) \times \operatorname{\mathbb{r}}^{18 } \rightarrow \operatorname{\mathbb{r}}$ such that\footnotemark\\ \\ $ w_{\rm lin}({{\rm grad}[{u } ] } , \ , { { \rm grad}[{{{\rm grad}[{u } ] } } ] } ) \!\!\!\!\!\!\!\
! \overbrace{=}^{\text{linearized frame - indifferent } \atop \text { reduced format } } \!\!\!\!\!\!\!\ ! \widehat{\psi}_{\rm lin}(\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } \ , , \ , { { \rm grad}[{\operatorname{sym}{{\rm grad}[{{{{u } } } } ] } } ] } ) \ , , \quad { \rm and } $ \\ \\ $ w_{\rm lin}({{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } \ , , \ , { { \rm grad}_{{{\mbox{$\xi$}}}}[{{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } } ] } ) \overbrace{=}^\text { form - invariance } w_{\rm lin}({{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , , \ , { { \rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) $ \\ \\ $ [ { \rm i.e. } \quad \widehat{\psi}_{\rm lin}(\operatorname{sym}{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } \ , , \
, { { \rm grad}_{{{\mbox{$\xi$}}}}[{\operatorname{sym}{{\rm grad}_{{{\mbox{$\xi$}}}}[{{{{{u}}}^\sharp ( { { \mbox{$\xi$ } } } ) } } ] } } ] } ) $ \center { $ = \widehat{\psi}_{\rm lin}(\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } \ , , \ , { { \rm grad}_{{{{x}}}}[{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) \quad \forall \ , { { { q } } } \in { { \rm so}}(3 ) ] \ , .
$ } } } \end{aligned}\ ] ] in terms of @xmath395 a second gradient material is said to be isotropic if the stored energy satisfies for all @xmath396 : @xmath397 } \ , , \ , { { \rm grad}_{{{{x}}}}[{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) \notag \\ & = \widehat{\psi}_{\rm lin } \bigg ( q \ , ( \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , q^t \ , , \ , \underbrace{{{\rm grad}_{{{{x } } } } [ { q \ , ( \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , q^t } ] } \ , q^t } _ { \displaystyle ( q_{ib } \ , \12 [ u_{b ,
c}+u_{c , b } ] \ , q^t_{cj})_{,a } \ , q^t_{ak } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k\footnotemark}\bigg)\end{aligned}\]]and this is an immediate consequence of .
considering condition in the case of the theory without higher gradients we have @xmath398 } ) = { \psi}_{\rm lin } \left ( q \ , ( \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \
, q^t \right ) \quad \forall { { { q } } } \in { { \rm so}}(3)\end{aligned}\ ] ] and we may probe also with non - constant rotation fields @xmath399 without altering the imposed invariance condition . however
, it is important to realize that the adopted invariance requirement for isotropy in higher - gradient materials via the @xmath2-transformation does not allow for space dependent rotation fields and the invariance condition for isotropy * does not * read @xmath400 } \ , , \ , { { \rm grad}_{{{{x}}}}[{\operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } } ] } ) \notag \\ & = \widehat{\psi}_{\rm lin } \left ( q({{{x } } } ) \ , ( \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \
, q^t({{{x } } } ) \ , , \ , { { \rm grad}_{{{{x } } } } [ { q({{{x } } } ) \ , ( \operatorname{sym}{{\rm grad}_{{{{x}}}}[{{{{{u } } } ( { { { x } } } ) } } ] } ) \ , q^t({{{x } } } ) } ] } \ , q^t({{{x } } } ) \right ) \,.\end{aligned}\ ] ]
this section does not contain new results but is included to connect our investigation of isotropy conditions to specific models proposed in the literature .
the linear indeterminate couple stress model is a specific second gradient elastic model , in which the higher order interaction is restricted to the continuum rotation @xmath401 , where @xmath402 is the displacement of the body .
it is therefore interpreted to be sensitive to rotations of material points and it is possible to prescribe boundary conditions of rotational type .
superficially , this is the simplest possible generalization of linear elasticity in order to include the gradient of the local continuum rotation as a source of additional strains and stress with an associated energy . in this paper , therefore , we limit our analysis to isotropic materials and only to the second gradient of the displacement : @xmath403 } } ] } \sim { \rm d}^2_{{{{x } } } } { { { u}}}=\underbrace{\frac{\partial^2 u_i}{\partial x_j \ , \partial x_k}}_{\displaystyle u_{i , jk } } \ , { { { e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k = ( \varepsilon_{ji , k}+\varepsilon_{ki , j}-\varepsilon_{jk , i } ) \,{{{e}}}_i \otimes { { { e}}}_j \otimes { { { e}}}_k \ , , \end{aligned}\ ] ] where @xmath404 + ( \text{grad } [ { { { u}}}])^t ) = \operatorname{sym}\text{grad } [ { { { u } } } ] \end{aligned}\ ] ] is the symmetric linear strain tensor .
thus , from eq .
all second derivatives of the displacement field @xmath70 can be obtained from linear combinations of derivatives of @xmath245 . in general
, strain gradient models do not introduce additional independent degrees of freedoms aside the displacement field @xmath70 .
thus , the higher derivatives introduce a latent - microstructure " ( constraint microstructure @xcite ) . however , this apparent simplicity has to be payed with more complicated and intransparent boundary conditions , as treated in a series of papers @xcite . + we assume the elastic energy to be given by @xmath405 ^ 2 + \mu\,l_c^2\,(\alpha_1\ , \|\operatorname{dev}\operatorname{sym}[\widetilde{{{{k}}}}]\|^2 + \alpha_2 \ , [ \operatorname{tr}(\widetilde{{{{k}}}})]^2 + \alpha_3\ , \|\skew[\widetilde{{{{k}}}}]\|^2 ) \ , { \,{\rm d}}v \ , , \end{aligned}\ ] ] where @xmath263 and @xmath264 are the constitutive lam constants and the curvature energy is based on the second order curvature tensor @xmath318 @xmath406})}]}=\12 \ , { { \rm grad } [ { \operatorname{curl}\ , { { { u } } } } ] } \ , , \end{aligned}\ ] ] with additional dimensionless constitutive parameters @xmath390 , @xmath391 , @xmath392 , and the characteristic length @xmath407 . taking free variations
@xmath408 of the elastic energy @xmath409 yields the virtual work principle @xmath410 } \rangle+\lambda \operatorname{tr}({{\mbox{$\varepsilon$}}})\,\operatorname{tr } ( { { \rm grad } [ { \delta { { { u } } } } ] } ) \notag\\ & + 2 \ , \mu \ , l_c^2 \ , \alpha_1\ , \langle \operatorname{dev}\operatorname{sym}({\widetilde{{{{k}}}}}),\operatorname{dev}\operatorname{sym}{{\rm grad}[{\operatorname{axl}\ , \skew { { \rm grad } [ { \delta { { { u } } } } ] } } ] } \rangle \notag \\ & + 2 \ , \mu \ , l_c^2 \ , \alpha_2\ , \operatorname{tr}({\widetilde{{{{k } } } } } ) \ , \operatorname{tr}({{\rm grad}[{\operatorname{axl}\ , \skew { { \rm grad } [ { \delta { { { u } } } } ] } } ] } ) \notag \\ & + 2 \ , \mu \ , l_c^2 \ , \alpha_3\ , \langle \skew({\widetilde{{{{k } } } } } ) , \skew({{\rm grad } [ { \operatorname{axl}\,\skew \ , { { \rm grad } [ { \delta { { { u}}}}]}}]})\rangle + \langle { { { f } } } , \delta { { { u}}}\rangle { \,{\rm d}}v = 0 \ , .\end{aligned}\ ] ] using the classical divergence theorem for the curvature term in eq .
it follows after some simple algebra that @xmath411 } \rangle { \,{\rm d}}a=0,\end{aligned}\ ] ] where @xmath244 is the symmetric local force - stress tensor @xmath412 and @xmath413 represents the nonlocal force - stress tensor @xmath414 \quad \in \operatorname{\mathfrak{so}}(3 ) \,,\end{aligned}\ ] ] which here is automatically skew - symmetric - transformation : since @xmath413 and @xmath3 are tensors of second order it holds @xmath415 and @xmath416 , respectively .
further , eq . yields @xmath417 = { { { q } } } \ , { { \rm div}_{{{{x}}}}}[{{{m}}}({{{x}}})]$ ] and with help of eq
. we obtain @xmath418 ) = \operatorname{anti}({{{q } } } \ , { { \rm div}_{{{{x}}}}}[{{{m}}}({{{x } } } ) ] ) = { { { q } } } \ , \operatorname{anti}({{\rm div}_{{{{x}}}}}[{{{m}}}({{{x } } } ) ] ) \ , { { { q}}}^t$ ] . ] .
+ the second order couple stress tensor @xmath3 in eq . reads @xmath419 } ) + \alpha_2 \ , \underbrace{\operatorname{tr}({{\rm grad } [ { \operatorname{curl}\ , { { { u}}}}]})}_{\displaystyle = \,0 } \ , { { { \mathbbm{1}}}}+ { \alpha_3}\,\skew({{\rm grad } [ { \operatorname{curl}\ , { { { u } } } } ] } ) ] \notag \\ & = 2 \ , \mu \ , l_c^2\,[\alpha_1 \ , \operatorname{dev}\operatorname{sym}({\widetilde{{{{k } } } } } ) + \alpha_3 \ , \skew({\widetilde{{{{k } } } } } ) ] \,,\end{aligned}\ ] ] which may or may not be symmetric , depending on the material parameters @xmath390 , @xmath392 .
however , @xmath3 in eq . is automatically trace free since both the deviator and the skew operator yield trace free tensors .
however , the asymmetry of the nonlocal force stress @xmath413 appears as a constitutive assumption .
thus , if the test function @xmath420 also satisfies @xmath421})=0 $ ] on @xmath422 ( equivalently @xmath423 ) , then we obtain the balance of momentum @xmath424 since the balance of angular momentum is given by eq . , we can combine them to see compact equilibrium equation @xmath425[2cm][c]{12cm}{$ \div \widetilde{{{\mbox{$\sigma$ } } } } + { { { f } } } = 0 \ , , \quad \quad \widetilde{{{\mbox{$\sigma$ } } } } = { { \mbox{$\sigma$ } } } + \widetilde{{{\mbox{$\tau$ } } } } \ , \notin { { \rm{sym } } } ( 3 ) \
, , \quad \text{total force stress } \\ \\ { { \mbox{$\sigma$ } } } = 2 \ , \mu \ , \operatorname{sym}{{\rm grad } [ { { { { u } } } } ] } + \lambda \operatorname{tr}({{\rm grad } [ { { { { u } } } } ] } ) { { { \mathbbm{1}}}}\ , , \quad \quad \widetilde{{{\mbox{$\tau$ } } } } = - \frac{1}{2 } \operatorname{anti}\div { { { m } } } \in \operatorname{\mathfrak{so}}(3 ) \\ \\ \div{{{{m } } } } + 2 \operatorname{axl}(\widetilde{{{\mbox{$\tau$ } } } } ) = 0 \quad \leftrightarrow \quad \div{{{{m } } } } + 2 \operatorname{axl}(\skew \widetilde{{{\mbox{$\sigma$ } } } } ) = 0 \ , . $ } } \end{aligned}\ ] ] let us consider the strain and curvature energy as minimization problem @xmath426}\|^2+\frac{\lambda}{2}\ , [ \operatorname{tr}({\rm sym } \ , { { \rm grad } [ { { { { u}}}}]})]^2+w_{\rm curv}({\rm d}^2_{{{{x } } } } { { { u}}})\right ] { \,{\rm d}}v \quad \mapsto \quad \text{min
. w.r.t . } \ , { { { u}}},\end{aligned}\ ] ] admitting unique minimizers under some appropriate boundary condition . here @xmath427 are the lam constitutive coefficients of isotropic linear elasticity , which is fundamental to small deformation gradient elasticity . if the curvature energy has the form @xmath428})$ ] , the model is called * a strain gradient model*. we define the hyperstress tensor of third order as @xmath429 .
note that @xmath430 is in general not symmetric . in the following we recall some curvature energies proposed in different isotropic second gradient elasticity models : * * the indeterminate couple stress model * ( grioli - koiter - mindlin - toupin model ) @xcite in which the higher derivatives ( apparently ) appear only through derivatives of the infinitesimal continuum rotation @xmath401 .
hence , the curvature energy has the equivalent forms @xmath431}\|^2+\alpha_3\,\| \skew { { \rm grad } [ { \operatorname{curl}\ , { { { u}}}}]}\|^2\notag\\ & = \mu\,l_c^2(\alpha_1\ , \|\operatorname{sym}\underbrace{{{\rm grad}[{\operatorname{axl}(\skew { { \rm grad } [ { { { { u}}}}]})}]}}_{\displaystyle { \widetilde{{{{k}}}}}}\|^2+\,\alpha_3\,\| \skew \underbrace { { { \rm grad}[{\operatorname{axl}(\skew { { \rm grad } [ { { { { u}}}}]})}]}}_{\displaystyle { \widetilde{{{{k}}}}}}\|^2 \ , , \\ { { { m } } } & = 2\,\mu\,l_c^2(\alpha_1\,\operatorname{sym}{\widetilde{{{{k}}}}}+\alpha_3\,\skew{\widetilde{{{{k } } } } } ) \ , .\end{aligned}\ ] ] although this energy admits the equivalent forms @xmath33 and @xmath147 , the equations and the boundary value problem of the indeterminate couple stress model is usually formulated only using the form of the energy .
we remark that the spherical part of the couple stress tensor is zero since @xmath432 as seen before . in order to ensure the pointwise uniform positive definiteness it is assumed that @xmath433 .
note that pointwise uniform positivity is often assumed @xcite when deriving analytical solutions for simple boundary value problems because it allows to invert the couple stress - curvature relation .
* * the modified symmetric couple stress model - the conformal model*. on the other hand , in the conformal case @xcite one may consider @xmath434 , which makes the couple stress tensor @xmath435 symmetric and trace free @xcite .
this conformal curvature case has been derived by neff in @xcite , the curvature energy having the form @xmath436 indeed , there are two major reasons uncovered in @xcite for using the modified couple stress model .
first , in order to avoid non physical singular stiffening behaviour for smaller and smaller samples in bending @xcite one has to take @xmath434 .
second , based on a homogenization procedure invoking a natural micro - randomness " assumption ( a strong statement of microstructural isotropy ) implies conformal invariance , which is again @xmath434 .
such a modell is still well - posed @xcite leading to existence and uniqueness results with only one additional material length scale parameter , while it is * not * pointwise uniformly positive definite .
* * the skew - symmetric couple stress model*. hadjesfandiari and dargush strongly advocate @xcite the opposite extreme case , @xmath437 and @xmath438 , i.e. they used the curvature energy @xmath439}\|^2=\mu\,l_c^2\,\alpha_3\ , \|\skew { \widetilde{{{{k}}}}}\|^2\ , , \quad \quad { { { m}}}=2\,\mu\,l_c^2\,\alpha_3\,\skew\,{\widetilde{{{{k}}}}}\,.\end{aligned}\ ] ] in that model the nonlocal force stress tesnor @xmath413 is skew - symmetric and the couple stress tensor @xmath3 is assumed to be completely skew - symmetric .
their reasoning , based on boundary conditions , is critically discussed in neff et al.@xcite . in ghiba
et al.@xcite the isotropic , linear indeterminate couple stress model has been modified so as to have symmetric total force stress @xmath440 , while retaining the same weak form of the euler - lagrange equations .
this is possible since the force stresses appearing in the balance of forces is only determined up to a self - equilibrated stress - field @xmath441 , i.e. @xmath442 the curvature energy expression of this new model is @xmath443 } ) \ ,
\|}^2 + \alpha_2 \ , { \|\ , \skew \operatorname{curl\,}(\operatorname{sym}{{\rm grad } [ { { { { u } } } } ] } ) \ , \|}^2 ) \ , .\end{aligned}\ ] ] the strong form of the new model reads @xmath444[2cm][c]{12cm}{$ \div \widehat{{{\mbox{$\sigma$ } } } } + { { { f } } } = 0 \ , , \quad \quad \widehat{{{\mbox{$\sigma$ } } } } = { { \mbox{$\sigma$ } } } + \widehat{{{\mbox{$\tau$ } } } } \ , \in { { \rm{sym } } } ( 3 ) \
, , \quad \text{total force stress } \\ \\ { { \mbox{$\sigma$ } } } = 2 \ , \mu \ ,
\operatorname{sym}{{\rm grad } [ { { { { u } } } } ] } + \lambda \operatorname{tr}({{\rm grad } [ { { { { u } } } } ] } ) { { { \mathbbm{1}}}}\ , , \quad \quad \widehat{{{\mbox{$\tau$ } } } } = \operatorname{sym}\operatorname{curl\,}(\widehat{{{{m } } } } ) \ , \in { { \rm{sym } } } ( 3 )
\\ \\ \widehat{{{{m } } } } = 2 \ , \mu \ , l_c^2 ( \alpha_1 \ , \operatorname{dev}\operatorname{sym}\operatorname{curl\,}(\operatorname{sym}{{\rm grad } [ { { { { u}}}}]})+ \alpha_2 \ , \skew \operatorname{curl\,}(\operatorname{sym}{{\rm grad } [ { { { { u } } } } ] } ) ) \ , .
$ } } \end{aligned}\ ] ] + the total force stress tensor is now @xmath445 and the second order couple stress tensor is @xmath446 .
note that similarly as in the indeterminate couple stress theory we have @xmath447 .
compared to the indeterminate couple stress theory on can show that @xmath448 , as claimed .
thus , eq . is a _ couple stress model with symmetric total force stress @xmath440 and trace free couple stress tensor @xmath446 .
_ moreover , the couple stress tensor @xmath3 is symmetric itself for the possible choice @xmath449 .
the strong form of the general strain gradient model reads @xmath450[3cm][c]{14cm}{$ \div({{\mbox{$\sigma$ } } } + \div { \mathfrak{m } } ) + { { { f } } } = 0 \ , , \qquad { \mathfrak{m}}\ , \ , \text{third order hyperstress tensor } \\ \\ { { \mbox{$\sigma$ } } } = 2 \ , \mu \ , \operatorname{sym}{{\rm grad } [ { { { { u } } } } ] } + \lambda \operatorname{tr}({{\rm grad } [ { { { { u } } } } ] } ) { { { \mathbbm{1}}}}\ , , \quad \breve{\tau}=\div { \mathfrak{m}}\ , \notin { { \rm{sym } } } ( 3)\\ \\ \breve{\sigma } = \sigma + \breve{\tau } \ , \notin { { \rm{sym } } } ( 3 ) \ , , \qquad \div \breve{\sigma } + { { { f } } } = 0\\ \\ { \mathfrak{m}}= \l.{{\rm grad}_{{{{x}}}}[{{{\rm grad}_{{{{x}}}}[{u(x ) } ] } } ] } \ , \in
\operatorname{\mathbb{r}}^{3 \times 3 \times 3 } , \quad \l : \operatorname{\mathbb{r}}^{3 \times 3 \times 3 } \rightarrow \operatorname{\mathbb{r}}^{3 \times 3 \times 3 } , \quad \text{linear 6th order tensor . } $ } } \end{aligned}\ ] ] for isotropy , it has been shown in @xcite that @xmath365 has 5 independent coefficients .
we have investigated the isotropy requirements in elasticity theory by going back to the basic invariance requirements under transformation of coordinates . in nonlinear elasticity
it is important to observe the difference between right - local @xmath98-invariance and right - global @xmath98-invariance .
we have given examples that both notions do not coincide in general and have shown another example where even right - local @xmath98-invariance is satisfied . for us , isotropy is right - global @xmath98-invariance while right - local @xmath98-invariance is a stronger condition imposing more restrictive invariance requirements .
next , in linear theories ( whether first or second gradient or indeed higher order gradient theories ) defined in the linearized frame - indifferent object @xmath451}$ ] and gradients of it , isotropy can be probed by simultaneous rigid rotation of spatial and referential coordinates .
we have called this transformation the @xmath2-mapping .
isotropy then is equivalent to * form - invariance * of all energy terms with respect to the @xmath2-transformation .
similarly , isotropy is equivalent to form - invariance of the balance equations with respect to the @xmath2-transformation .
we have investigated , in this respect , several variants of higher - gradient models .
+ our approach can be immediately extended to discuss anisotropy for higher gradient theories simply by requiring form - invariance of the energy in the @xmath2-transformation not for all @xmath14 ( isotropy ) but only for @xmath452 ( material symmetry group ) .
this will lead to very succinct statements of anisotropy for higher gradient models .
|
for homogeneous higher gradient elasticity models we discuss frame - indifference and isotropy requirements . to this end , we introduce the notions of local versus global so(3)-invariance and identify frame - indifference ( traditionally ) with global left so(3)-invariance and isotropy with global right so(3)-invariance
. for specific restricted representations , the energy may also be local left so(3)-invariant as well as local right so(3)-invariant .
then we turn to linear models and consider a consequence of frame - indifference together with isotropy in nonlinear elasticity and apply this joint invariance condition to some specific linear models .
the interesting point is the appearance of finite rotations in transformations of a geometrically linear model .
it is shown that when starting with a linear model defined already in the infinitesimal symmetric strain @xmath0}$ ] , the new invariance condition is equivalent to isotropy of the linear formulation . therefore , it may be used also in higher gradient elasticity models for a simple check of isotropy and for extensions to anisotropy . in this respect
we consider in more detail variational formulations of the linear indeterminate couple stress model , a new variant of it with symmetric force stresses and general linear gradient elasticity .
+ + * key words : * invariance conditions , frame - indifference , covariance , isotropy , orthogonal group , strain gradient elasticity , couple stress , polar continua , symmetric stress , hyperstresses , modified couple stress model , rotational invariance , form - invariance , rayleigh product + ams math 74a10 ( stress ) , 74a35 ( polar materials ) , 74b05 ( classical linear elasticity ) , 74a30 ( nonsimple materials ) , 74a20 ( theory of constitutive functions ) , 74b20 ( nonlinear elasticity )
|
let @xmath0 be a connected oriented surface of finite type , of genus @xmath1 with @xmath2 punctures .
we assume that the euler characteristic of @xmath3 is negative .
we shall consider hyperbolic structures on @xmath3 and any such structure will be complete and of finite area .
we say that two complex structures ( respectively hyperbolic structures ) @xmath4 and @xmath5 on @xmath3 are equivalent if there is a conformal map ( respectively an isometry ) from @xmath4 to @xmath5 which is homotopic to the identity map of @xmath3 .
the teichmller space @xmath6 of @xmath3 is the space of complex structures ( or , equivalently , the space of hyperbolic structures ) on @xmath3 up to equivalence .
an _ asymmetric metric _ on a set @xmath7 is a nonnegative function @xmath8 on @xmath9 which satisfies the axioms of a metric except the symmetry axiom , that is , we do not require that @xmath10 for all @xmath11 .
thurston defined in @xcite an asymmetric metric @xmath12 on @xmath6 by setting for each @xmath13 @xmath14 where @xmath4 and @xmath5 are considered as ( equivalence classes of ) complete finite area hyperbolic structures on @xmath3 , where the infimum is take over all homeomorphsims @xmath15 which are homotopic to the identity map of @xmath3 and where @xmath16 is the lipschitz constant of @xmath17 , that is , @xmath18 for @xmath19 and @xmath20 in @xmath4 .
we shall call this asymmetric metric the _ thurston asymmetric metric_. thurston @xcite proved that there is a ( not necessarily unique ) extremal lipschitz homeomorphsim that realizes the infimum in ( [ equ : lip ] ) , and that @xmath21 where @xmath22 denotes the hyperbolic length of @xmath23 in @xmath4 and @xmath23 ranges over all essential ( that is , neither homotopic to a point nor a puncture ) simple closed curves on @xmath3 .
we denote the lipschitz constant of such an extremal lipschitz map by @xmath24 .
thurston showed that this function @xmath12 is indeed an asymmetric metric on teichmller space , that it is a finsler metric and that any two points in teichmller space can be joined by a ( not necessarily unique ) @xmath12-geodesic ( that is , a shortest path with respect to @xmath12 ) .
thurston also made a relation between extremal lipschitz maps and a class of @xmath12-geodesics called _
stretch lines_. ( we shall recall the definition below . )
the teichmller metric is defined by @xmath25 where @xmath26 is the quasiconformal dilatation of a homeomorphsim @xmath15 homotopic to the identity map of @xmath3 . using a lemma of wolpert , we have @xcite , @xmath27 . by definition ,
the topology induced by thurston s asymmetric metric on @xmath6 is the one induced by its symmetrization , @xmath28 which is a genuine metic .
this topology coincides with the one induced by the teichmller metric , see li @xcite and papadopoulos - thret @xcite for information about the symmetrization . in an asymmetric metric space @xmath29 ,
the _ left _ and _ right _ open balls centered at a point @xmath19 and of radius @xmath30 , defined respectively by @xmath31 and @xmath32 may have completely different behaviors .
endowed with thurston s asymmetric metric , the space @xmath6 is complete , and the left and right closed balls are compact .
see @xcite , @xcite and @xcite for the topology induced by this asymmetric metric .
the _ mapping class group _ @xmath33 of @xmath3 is the group of homotopy classes of orientation - preserving homeomorphism of @xmath3 .
this group acts properly discontinuously and isometrically on @xmath6 endowed with teichmller s metric or with thurston s asymmetric metric .
the quotient space is the moduli space @xmath34 thurston used his compactification of @xmath6 by the space of projective measured foliations to prove the following theorem ( see thurston @xcite , theorem 4 or @xcite for details ) .
each @xmath35 is either periodic , reducible or pseudo - anosov .
pseudo - anosov mapping classes are neither periodic nor reducible . here
, @xmath36 is said to be _ periodic _ if it is of finite order ; @xmath36 is said to be _
reducible _ if there is a multicurve ( i. e. a union of disjoint essential curves ) on @xmath3 which is globally fixed by @xmath36 ; @xmath36 is said to be _ pseudo - anosov _ if it preserves a pair of projective classes of transverse measured foliations having no singular closed leaves , multiplying the transverse measure on one by some constant @xmath37 and on the other by the constant @xmath38 .
we shall recall this classification more precisely in the following sections .
let @xmath36 be an element of @xmath33 .
we set @xmath39 we call @xmath40 the _ translation distance _ of @xmath36 with respect to thurston s asymmetric metric .
we say that @xmath40 is _ realized _ if there exists some @xmath41 such that @xmath42 .
it is interesting to compare @xmath40 with the translation distance defined by the teichmller metric @xmath43 : @xmath44 since @xmath27 , we have @xmath45 . since thurston s asymmetric
metric is not symmetric , we may also consider @xmath46 to motivate the work done in this paper , we recall that bers @xcite studied the translation distance @xmath47 and proved the following classification result .
[ bers @xcite ] let @xmath35
. then 1 .
[ p1 ] @xmath48 and realized @xmath49 @xmath36 is periodic .
@xmath47 is not realized @xmath49 @xmath36 is reducible .
3 . @xmath50 and realized @xmath49 @xmath36 is pseudo - anosov . in case ( [ p1 ] ) , saying that @xmath36 is _ periodic _ means that it is represented by a periodic homeomorphism of the surface .
bers interpreted the realization of @xmath47 as an extremal problem . in the case where @xmath47 is not realized
, bers further showed that there is a generalized solution of the extremal problem which is given by a noded riemann surface , and that the restriction of @xmath36 on each nonsingular component of the noded riemann surface is either periodic or pseudo - anosov .
daskalopoulos and wentworth @xcite obtained an analogous classification of elements of @xmath33 in terms of the weil - petersson translation distance . in analogy with bers classification of elements of @xmath33 with respect to the teichmller metric , we classify @xmath35 into four types : [ cols="^,^,^,>",options="header " , ] the aim of our investigation in this paper is to analyze each of these cases .
it is easy to see that @xmath36 is elliptic if and only if @xmath36 has a fixed point in @xmath6 .
indeed , if @xmath36 has a fixed point in @xmath6 , then there is a complex structure @xmath4 on @xmath3 such that @xmath36 induces a biholomorphic automorphism of @xmath4 . using a classical result of hurwitz
, we deduce that @xmath36 is of finite order , therefore periodic .
conversely , if @xmath36 is periodic , nielsen showed that @xmath36 has a fixed point in @xmath6 .
in fact , kerckhoff proved a stronger result , namely , that the action of every finite subgroup of @xmath33 has a fixed point in @xmath6 .
( this was his solution to the so - called the _
nielsen realization problem_. ) a stretch line is a @xmath12-geodesic of a special kind , defined by stretching " along a complete geodesic lamination .
we shall recall the definition in section [ sec : pa ] . in this paper
, we prove the following : [ thm : pa ] if @xmath36 is pseudo - anosov , then it leaves a @xmath12-geodesic invariant .
furthermore , for some @xmath51 , @xmath52 leaves a stretch line invariant .
we also prove the following : [ thm : ab ] if @xmath36 is pseudo - anosov , then @xmath53 .
the projection on moduli space of the @xmath36-invariant @xmath12-geodesic provided by theorem [ thm : pa ] is a closed @xmath12-geodesic , and it is homotopic to a teichmller closed geodesic of the same length . in analogy with the case of the teichmller metric , we prove the following : [ thm : irreducible ] if @xmath36 is irreducible , then @xmath36 is either elliptic or hyperbolic .
[ thm : hyperbolic ] if @xmath36 is of infinite order and leaves a @xmath12-geodesic invariant , then @xmath36 is hyperbolic .
the translation distance @xmath40 is attained at any point on the geodesic . [ thm : parabolic ] the mapping class @xmath36 is of infinite order , reducible and periodic on all reduced components on @xmath3 if and only if @xmath36 is parabolic . to show that a reducible map @xmath35 can not attain @xmath47 in teichmller space ,
bers @xcite used the notion of nielsen extension of riemann surfaces to show that for any @xmath54 , one can shrink " the reduced curves to get a new riemann surface @xmath5 satisfying @xmath55 .
bers construction does not apply to the setting of thurston s asymmetric metric .
we shall see in
[ s : red - hyp ] of this paper that there exists a reducible map @xmath36 which is hyperbolic with respect to the asymmetric @xmath12 , and therefore for which @xmath40 is realized .
these results give quite a good picture of the behavior of the action of a mapping class with respect to thurston s asymmetric metric , in terms of thurston s classification .
it remains to study the case where @xmath36 is reducible and has at least one pseudo - anosov reduced component . in the setting of the teichmller metric , to show that a hyperbolic map @xmath35 ( @xmath50 and realized ) leaves a teichmller geodesic invariant , bers @xcite used the fact that any two points in teichmller space are connected by a unique teichmller geodesic .
since thurston s asymmetric metric is not uniquely geodesic , bers argument does not work here . in the paper @xcite by daskalopoulos and wentworth , the strict convexity property of hyperbolic length functions along weil - petersson geodesics
is used to show that a reducible map can not leave a weil - petersson geodesic invariant .
we do nt know whether geodesic length functions are convex along @xmath12-geodesics . 1 .
is theorem [ thm : ab ] true in general ? 2 .
if @xmath36 is reducible and has at least one pseudo - anosov reduced component , is it pseudo - hyperbolic ?
can it be hyperbolic ?
if @xmath36 is hyperbolic , does it leave a @xmath12-geodesic invariant ?
is it pseudo - anosov ?
if @xmath36 is of infinite order and leaves a @xmath12-geodesic ( or even a stretch line ) invariant , is it pseudo - anosov ? in [ s : red - hyp ] , we give an example , on any surface of genus @xmath56 , of a reducible mapping class which is of hyperbolic type .
it is known that closed teichmller geodesics in moduli space correspond to conjugacy classes of pseudo - anosov maps .
the above questions are related to the question of whether closed @xmath12-geodesics in moduli space correspond to conjugacy classes of pseudo - anosov maps .
in this section , we prove theorems [ thm : irreducible ] and [ thm : hyperbolic ] .
let @xmath57 be the set of homotopy classes of essential simple closed curves on @xmath3 .
a finite non - empty subset @xmath58 is called _ admissible _ if @xmath59 for all @xmath60 and if @xmath61 can be realized simultaneously as mutually disjoint curves .
we call @xmath35 _ reducible _ if there is an admissible set @xmath62 such that @xmath63 an element @xmath35 is called _ irreducible _ if there is no admissible set reduced by @xmath36
. we shall use the following hyperbolic geometry estimate : given a simple closed geodesic @xmath64 of length @xmath65 on a hyperbolic surface , there exists a collar neighborhood about @xmath64 of width @xmath66 , with @xmath67 see buser @xcite for a reference .
it follows that for small @xmath65 , the length of any simple closed curve that intersects @xmath64 is large .
the following consequence is well - known : [ lem : collar ] there is a universal constant @xmath68 such that any two distinct simple closed geodesics on a hyperbolic surface with hyperbolic length less than @xmath69 are disjoint .
we also need the following theorem .
see bers @xcite for a proof . for each @xmath70 ,
the subset @xmath71 is compact .
here @xmath72 denotes the length of the shortest essential simple closed curve of a hyperbolic surface @xmath4 .
note that this quantity is well - defined for elements @xmath4 in @xmath6 or in @xmath73 . for each @xmath70 , the subset @xmath74
is called the _ @xmath30-thick part _ of @xmath75 .
mumford s compactness theorem is equivalent to saying that for any sequence @xmath76 in an @xmath30-thick part of @xmath75 , there is a subsequence of @xmath76 , still denoted by @xmath76 , and a sequence of @xmath77 in @xmath78 , such that @xmath79 converges in @xmath75 .
[ lem : systol ] let @xmath35 be irreducible and @xmath80 .
then , with @xmath69 being the constant given in lemma [ lem : collar ] , we have : @xmath81 suppose that @xmath82 let @xmath83 such that @xmath84 .
consider the curves @xmath85 , @xmath86 , @xmath87 , @xmath88 .
since @xmath89 each @xmath90 satisfies @xmath91 by @xmath92 , we have @xmath93 by lemma [ lem : collar ] , the curves @xmath94 are mutually disjoint .
since there are at most @xmath95 isotopy class of pairwise disjoint simple closed curves in @xmath3 , there is a least number @xmath96 such that @xmath97 .
note that @xmath98 is admissible and @xmath99 as a result , @xmath36 is reducible , which contradicts the assumption . by assumption
, @xmath36 is irreducible .
consider a sequence @xmath76 in @xmath75 with @xmath100 there is a constant @xmath101 such that for any @xmath102 , @xmath103 . by lemma [ lem : systol ]
, @xmath104 therefore , the sequence @xmath105 lies in a thick part of @xmath75 .
now we use a compactness argument of bers @xcite to show that @xmath40 attains its infimum in @xmath75 . by mumford s compactness theorem , there is a subsequence of @xmath76 , still denoted by @xmath76 , and a sequence @xmath77 in @xmath33 , such that @xmath79 converges to some point @xmath106 .
let @xmath107 .
since @xmath33 acts isometrically with respect to thurston s asymmetric metric , we have @xmath108 by @xmath109 , we have @xmath110 since @xmath111 for some sufficiently large constant @xmath112 and since @xmath6 is locally compact , up to a subsequence , we can assume that @xmath113 converges to some @xmath114 .
since @xmath115 converges to @xmath5 , we conclude that the sequence @xmath116 converges to @xmath117 . for any @xmath70
, there is a number @xmath118 , such that for any @xmath119 , @xmath120 since @xmath33 acts properly discontinuously on @xmath75 , we have , for sufficiently large @xmath121 , @xmath122 it follows from @xmath123 that @xmath124 [ coro : reducible2 ] suppose that @xmath35 is parabolic or pseudo - hyperbolic and suppose that @xmath76 is a sequence in @xmath75 with @xmath125 then for any @xmath70 , @xmath126 leaves @xmath127 for any sufficiently large @xmath128 . otherwise , using bers compactness argument as we did in the proof of theorem [ thm : irreducible ] , we can show that @xmath36 is elliptic or hyperbolic .
the proof follows bers argument in @xcite , theorem 5 .
we reproduce it here for the convenience of the reader .
suppose that a @xmath12-geodesic through @xmath80 is invariant under @xmath36 .
let @xmath129 be any point in @xmath6 . since @xmath36 is an isometry , we have @xmath130 since @xmath2 is arbitrary , @xmath131 for any @xmath106 .
as a result , @xmath132 .
bers also proved that @xmath133 is realized if and only if @xmath36 leaves a teichmller geodesic invariant , cf .
bers @xcite , theorem 5 . by an analysis of the teichmller geodesic left invariant by @xmath36
, he showed that the initial and terminal quadratic differentials of the teichmller map in the homotopy class of @xmath36 coincide .
it follows that @xmath36 preserves a pair of transverse measured foliations , multiplying the transverse measure of one by @xmath134 ( the dilation of the teichmller map ) and of the other one by @xmath38 .
this means that @xmath36 is pseudo - anosov .
we will study pseudo - anosov maps in section [ sec : pa ] .
in this section , we recall some necessary background on laminations and on thurston s construction of strectch lines . fix a hyperbolic structure on the surface @xmath3 .
a _ geodesic lamination _
@xmath135 on @xmath3 is a closed subset of @xmath3 which is the union of disjoint simple geodesics ( called the _ leaves _ of @xmath135 ) .
a _ measured geodesic lamination _ is a geodesic lamination @xmath135 with a transverse invariant measure of full support @xmath135 .
if a geodesic lamination @xmath135 has not all of its leaves going at both ends to cusps , then @xmath135 contains a compactly - supported sublamination admitting a transverse measure .
the _ stump _ of a geodesic lamination @xmath135 is the maximal compact sublamination of @xmath136 admitting a transverse measure . a geodesic lamination @xmath135 on a hyperbolic surface @xmath4 is _ complete _ if its complementary regions are all isometric to ideal triangles .
associated with such a pair @xmath137 is a partial ( that is , supported on a subsurface ) measured foliation @xmath138 , satisfying the following : a. the foliation @xmath138 intersects @xmath135 transversely and in each ideal triangle of the complement of @xmath135 , the leaves of @xmath138 are composed of horocycles perpendicular to the boundary of the ideal triangle ; b. the non - foliated region in each ideal triangle is bounded by three pairwise tangent leaves , as in figure [ horo ] ; c. the transverse measure for @xmath138 agrees with arc length on @xmath135 . by collapsing each non - foliated region of @xmath138 onto a tripod
, we get a measured foliation on @xmath3 of full support , still denoted by @xmath138 , which is well defined up to isotopy and which is called the _
horocyclic foliation _ corresponding to @xmath137 .
note that since the hyperbolic structures we consider are complete , each puncture of @xmath4 has a neighborhood isometric to a cusp .
if the set of punctures of @xmath3 is nonempty , each complete geodesic lamination on @xmath3 has some leaves going to cusps .
the corresponding horocylic foliations are standard near the cusps , meaning that in a neighborhood of each cusp , the leaves are circles homotopic to the puncture , and the total transverse measure of an arc converging to the cusp is infinite .
we denote by @xmath139 the space of equivalence classes of measured foliations that are transverse to @xmath135 and standard near the cusps .
thurston @xcite proved the following fundamental result .
[ thm : thurston1 ] the map @xmath140 is a homeomorphism .
we also recall the following definition from thurston @xcite .
a geodesic lamination @xmath141 is _ totally transverse _
to @xmath135 if each leaf of @xmath141 intersects @xmath135 transversely infinitely often and if each leaf of @xmath135 which does not go to a cups intersects @xmath141 transversely infinitely often .
( in counting intersections , the leaves are parametrized by the reals . in particular , with this convention , a simple closed geodesic meets any transverse arc infinitely often . ) by a result of thuston @xcite , @xmath139 can be identified with the space of ( equivalence classes ) of measured laminations with compact support and totally transverse to @xmath135 . from now on
, we always assume that a measured geodesic lamination has compact support .
we shall say that a measured geodesic lamination @xmath141 is _ supported by _
a geodesic lamination @xmath135 if @xmath141 is a sublamination of @xmath135 .
we fix a complete geodesic lamination @xmath135 on @xmath3 .
the _ stretch line _ directed by @xmath136 and passing through @xmath142 is the ( image of a ) path @xmath143 we shall call a segment of a stretch line a _ stretch path_. stretch lines are geodesics for the thurston asymmetric metric .
more precisely , suppose that @xmath135 supports a measured geodesic lamination @xmath141 .
thus , for any two points @xmath144 on the stretch line , the lipschitz distance @xmath145 is equal to @xmath146 and realized by @xmath147 in fact , laminations that realize the lipschitz distance @xmath148 ( or , equivalently , that maximize the ratio of length ) have been characterized by thurston .
a geodesic lamination @xmath135 is _ ratio - maximizing _ for @xmath4 and @xmath149 if there is an @xmath24-lipschitz map ( homotopic to the identity map ) from a neighborhood of @xmath135 in @xmath4 to a neighborhood of @xmath135 in @xmath5 .
a geodesic lamination @xmath135 is called _ chain - recurrent _
if for any @xmath70 and for any @xmath150 there is a closed @xmath30-trajectory of @xmath135 through @xmath151 , that is , a closed unit speed path in the surface such that for any interval of length @xmath152 on the path there is an interval of length @xmath152 on some leaf of @xmath135 such that the two paths remain within @xmath30 of each other in the @xmath153 sense . by a result of thurston @xcite , for any two distinct points @xmath154 , there is a unique maximal ( in the sense of inclusion ) ratio - maximizing chain - recurrent lamination @xmath155 which contains all other ratio - maximizing chain - recurrent laminations . with the above definitions , we can state the following theorem of thurston .
[ thm : thurston2 ] any two points @xmath156 can be connected by a @xmath12-geodesic which consists of a finite concatenation of stretch paths .
moreover , each of these stretch paths stretches along some complete lamination containing the unique ratio - maximizing chain - recurrent lamination @xmath155 .
since each element @xmath36 of @xmath33 acts isometrically on @xmath157 , it maps @xmath12-geodesics to @xmath12-geodesics . we have the following : any @xmath35 maps a stretch line @xmath158 through @xmath4 to a stretch line through @xmath159 , given by @xmath160 it is easy to see that the image of @xmath158 under @xmath36 is a stretch line which passes through @xmath159 , stretched along the complete lamination @xmath161 and with corresponding horocyclic measured foliation @xmath162 . since the horocyclic measured foliation is uniquely determined by @xmath159 and @xmath161 , and
since @xmath163 is also a stretch line through @xmath159 directed by @xmath161 , it follows that @xmath164 .
we denote by @xmath165 $ ] the projective class of a measured lamination ( or measured foliation ) @xmath141 .
such a @xmath165 $ ] determines a element in thurston s boundary of teichmller space .
we refer to thurston @xcite and @xcite for definitions and properties of thurston s compactification and thurston s boundary . here
we just recall that thurston s compactification is homeomorphic to a closed @xmath166-dimensional ball whose boundary is identified with the space of projective measured foliations and that the mapping class group action on teichmller space extends continuously to thurston s compactification .
the following two theorems are important for our study .
[ thm : papa ] the stretch line @xmath158 converges in the positive direction to @xmath167 $ ] in thurston s boundary .
recall that a measured geodesic lamination is said to be _ uniquely ergodic _ if its transverse measure is unique up to scalar multiples .
[ thm : gt ] suppose that the stump @xmath141 of the geodesic lamination @xmath135 is non - empty and is uniquely ergodic .
then @xmath158 converges in the negative direction to @xmath165 $ ] in thurston s boundary .
now consider a pair of totally transverse measured geodesic laminations @xmath168 and @xmath169 .
suppose that @xmath168 is uniquely ergodic .
we can choose a complete geodesic lamination @xmath135 whose stump is @xmath168 .
by theorem [ thm : thurston1 ] , there exists a unique @xmath170 such that @xmath138 is equivalent to @xmath169 .
( as we noted above , we identify @xmath138 with @xmath169 . )
it follows from theorem [ thm : papa ] and theorem [ thm : gt ] that the stretch line @xmath158 converges in the positive direction to @xmath171 $ ] and converges in the negative direction to @xmath172 $ ] .
we have the following consequence .
[ thm : converge ] let @xmath168 and @xmath169 be a pair of totally transverse measured geodesic laminations and suppose that @xmath168 is uniquely ergodic .
then there exists a stretch line which converges in the positive direction to @xmath171 $ ] and in the negative direction to @xmath172 $ ] . in general , the stretch line in corollary [ thm : converge ] is not unique .
if @xmath168 is not maximal , there are different completions of @xmath168 which give different stretch lines in teichmller space .
in this section , we prove theorem [ thm : pa ] and [ thm : ab ] . recall that an element @xmath35 is said to be _ pseudo - anosov _ if it has a representative , still denoted by @xmath36 , which is a homeomorphism of @xmath3 that preserves the projective classes of a pair of transverse measured foliations @xmath173 and @xmath174 and if there exists a constant @xmath37 such that @xmath175 @xmath176 the measured foliations @xmath168 and @xmath169 are called the _ stable _ and _ unstable _ foliations of @xmath36 respectively . the quantity @xmath177 is the _ topological entropy _ of @xmath36 @xcite @xcite .
we recall some more terminology : a measured foliation @xmath141 is _ minimal _ if no closed curve in @xmath3 can be realized as a leaf of @xmath141 .
equivalently , after whitehead moves , the foliation has only dense leaves on @xmath3 .
two measured foliations are _ topologically equivalent _ if after isotopy and whitehead moves , the leaf structure of the two foliations is the same .
a theorem of masur @xcite states that if a measured foliation @xmath141 is minimal and uniquely ergodic , then for any other measured foliation @xmath178 satisfying @xmath179 we have @xmath180 for some constant @xmath181 .
[ lemma : ergodic ] suppose that @xmath35 is pseudo - anosov .
then each of the invariant foliations of @xmath36 is minimal and uniquely ergodic . in what follows
, we shall often pass from the measured foliations @xmath168 and @xmath169 to the measured laminations that represent them and vice versa .
we endow @xmath3 with a hyperbolic structure and we identify @xmath168 and @xmath169 with the corresponding measured geodesic laminations . since @xmath168 is minimal and uniquely ergodic , each connected component of @xmath182 is isometric to an ideal polygon or an ideal polygon with one puncture .
it follows that there are finitely many different complete geodesic laminations with stump @xmath168 .
let @xmath135 be a complete geodesic lamination with stump @xmath168
. then @xmath135 and @xmath169 are totally transverse . by proposition [ lemma : ergodic ] ,
both @xmath168 and @xmath169 are minimal and uniquely ergodic .
we replace @xmath168 and @xmath169 by their respective measured foliations . up to whitehead moves , we may assume that for both @xmath168 and @xmath169 , each half - leaf is dense in the surface .
then it is easy to see that each leaf of @xmath169 intersects @xmath135 transversely infinitely often , and each leaf of @xmath135 which does not go to a cusp intersects @xmath169 transversely infinitely often .
it follows from the definition that @xmath169 is totally transverse to @xmath135 .
[ thm : pa - stretch ] suppose that @xmath35 is pseudo - anosov .
then there exists an integer @xmath183 such that @xmath52 leaves a stretch line invariant .
let @xmath168 and @xmath169 be the stable and unstable measured laminations of @xmath36 .
choose a complete geodesic lamination @xmath135 with stump @xmath168 . by the discussion before corollary [ thm :
converge ] , there is a unique stretch line , denoted by @xmath184 , which converges in the positive direction to @xmath171 $ ] and converges in the negative direction to @xmath172 $ ] .
since @xmath36 fixes @xmath172 $ ] and @xmath171 $ ] on thurston s boundary , each @xmath185 is a stretch line which also converges in the positive direction to @xmath171 $ ] and converges in the negative direction to @xmath172 $ ] .
a stretch line satisfying the above conditions is uniquely determined by @xmath169 and the completion of @xmath168 . since there are finitely many different completions of @xmath168 , there is a integer @xmath2 such that @xmath52 leaves @xmath184 invariant . by theorem [ thm : hyperbolic ]
, @xmath52 is hyperbolic .
the translation distance @xmath186 attains its infimum at any point on @xmath184 .
[ pro : stretch ] with the above notations , @xmath187 . since @xmath188 , for any @xmath142 , @xmath189 thus @xmath190 .
take @xmath4 on the stretch line @xmath184 .
since the measure of @xmath169 is stretched " by @xmath191 , by definition of stretch line , the length of the measured lamination @xmath168 is also stretched " by @xmath191 .
we have @xmath192 .
the proof of theorem [ thm : pa - stretch ] gives the following : [ thm : pa - stretch2 ] suppose that @xmath35 is a pseudo - anosov mapping class with stable and unstable measured laminations @xmath168 and @xmath169 and suppose that @xmath168 is complete
. then @xmath36 leaves a unique stretch line invariant . in the general case
, we do nt know whether a pseudo - anosov mapping class leaves a stretch line invariant .
we will show that such a mapping class leaves a @xmath12-geodesic invariant .
first notice the following property : [ pro : stretch2 ] suppose that @xmath35 is pseudo - anosov with topological entropy @xmath177 .
then @xmath36 is hyperbolic and @xmath193 . by the above results ,
there is a power @xmath52 which fixes a stretch line and with translation length @xmath187 . by the triangle inequality ,
@xmath194 it follows that @xmath195 . to show the reverse inequality , we use the following lemma . [
lem : papa ] there is a constant @xmath196 such that for any simple closed curve @xmath64 and for any integer @xmath197 , @xmath198 by this lemma , we have @xmath199 by taking @xmath200 , we have @xmath201 since @xmath64 is arbitrary , we have @xmath202 moreover , the infimum of @xmath40 is attained at any point on the stretch line invariant under @xmath52 .
now we show the following : [ thm : pa - stretch4 ] a pseudo - anosov map @xmath35 leaves a @xmath12-geodesic invariant . by theorem [ thm : pa - stretch ] , there exists an integer @xmath2 such that @xmath52 leaves a stretch line @xmath203 invariant .
note that @xmath204 are also invariant by @xmath52 .
choose any point @xmath4 on @xmath203 and choose a geodesic segment between @xmath4 and @xmath159 .
denote this geodesic segment by @xmath205 .
we construct a path @xmath206 in teichmller space by setting @xmath207 where @xmath208 .
we also extend the path to the reverse direction by iteration of @xmath209 . in this way
, we get an infinite path @xmath206 in teichmller space which is invariant by the action of @xmath36 .
we claim that @xmath206 is a geodesic . to see this , consider the segment on @xmath206 connecting @xmath210 and @xmath211 .
the length of this segment is equal to @xmath212 times @xmath213 .
on the other hand , both @xmath210 and @xmath211 lie on the stretch line @xmath203 , and the lipschitz distance between @xmath210 and @xmath211 is equal to @xmath212 times @xmath177 ( the entropy of @xmath36 ) .
we have shown in proposition [ pro : stretch2 ] that @xmath214 . as a result ,
the length of the above segment on @xmath206 connecting @xmath210 and @xmath211 is equal to the lipschitz distance between @xmath210 and @xmath211 . since @xmath128 is arbitrary
, @xmath206 is a @xmath12-geodesic .
combining theorem [ thm : pa - stretch ] and theorem [ thm : pa - stretch4 ] , we obtain theorem [ thm : pa ] . note that if @xmath36 is pseudo - anosov with entropy @xmath177 , it was proved by bers @xcite that @xmath215 .
it follows that @xmath216 . from this
, we deduce theorem [ thm : ab ] .
the projection of a @xmath36-invariant @xmath12-geodesic on the moduli space is a closed @xmath12-geodesic , which is homotopic to a closed teichmller geodesic with the same length .
note that by results of thret in @xcite , any two @xmath12-geodesics left invariant by a pseudo - anosov map are asymptotic , and two @xmath12-geodesics left invariant by two distinct pseudo - anosov maps are divergent .
by bers classification , we have [ pro : reducible ] an element @xmath35 is of infinite order and reducible if and only if @xmath217 is not realized .
moreover , such an element @xmath36 has no pseudo - anosov reduced component if and only if @xmath218 .
suppose that @xmath35 is of infinite order , reducible and has no pseudo - anosov reduced component . since @xmath45 and @xmath218 , @xmath36 is parabolic , i. e. , @xmath219 .
on the other hand , suppose that @xmath36 is of infinite order , reducible and has at least one pseudo - anosov reduced component .
let @xmath5 be a pseudo - anosov reduced component of @xmath36 .
up to taking a power of @xmath36 , we may assume that @xmath36 maps @xmath5 to @xmath5 .
there is a pair of transverse measured foliations @xmath168 and @xmath169 on the surface @xmath5 such that @xmath220 for some constant @xmath37 .
it follows that for any @xmath80 , @xmath221 as a result , @xmath222 .
we deduce the following : [ thm : parabolic ] @xmath36 is parabolic if and only if @xmath36 is infinite order , reducible and has no pseudo - anosov reduced component .
now suppose that @xmath36 is pseudo - hyperbolic . by theorem [ thm : parabolic ] , @xmath36 has at least one pseudo - anosov reduced component .
the maximum of the topological entropy on all of the pseudo - anosov components gives a lower bound for @xmath40 .
in fact , from bers results , it follows that @xmath223 . by theorem [ thm : irreducible ]
, @xmath36 is reducible .
assume that @xmath36 is reduced by a maximal admissible set @xmath62 .
up to a power of @xmath36 , we may assume that @xmath36 fixes each reduced component .
note that any power of @xmath36 is also pseudo - hyperbolic . by definition
, there exists a sequence @xmath76 in teichmller space such that @xmath224 by corollary [ coro : reducible2 ] , for any @xmath70 , @xmath126 leaves the @xmath30-thick part of @xmath6 for @xmath128 sufficiently large .
this means that @xmath225 .
let @xmath5 be any pseudo - anosov reduced component of @xmath5 .
but assumption , @xmath36 fixes @xmath5 .
p - h ] let @xmath226 be a sequence of simple closed curves such that @xmath227 . for sufficiently large @xmath128 ,
each @xmath228 is not contained in the interior of @xmath5 .
suppose not and take a subsequence of @xmath226 , still denoted by @xmath229 , contained in the interior @xmath5 . since @xmath227 and @xmath230 is uniformly bounded above , using the proof of lemma [ lem : systol ]
, @xmath36 is reducible restricted on @xmath5 .
this contradicts the assumption that @xmath5 is a pseudo - anosov component .
in this section , we prove the following consider a pseudo - anosov map @xmath36 on a hyperbolic surface of genus @xmath1 with @xmath231 boundary components , which we denote by @xmath232 in order to distinguish it from the surface @xmath233 with punctures .
we assume that @xmath36 does not permute the complementary components of its stable laminations .
we take two copies of this map @xmath36 , which we denote by @xmath234 and @xmath235 , and we denote the two underlying surfaces by @xmath236 and @xmath237 .
we glue @xmath236 and @xmath237 in a symmetric manner along their boundary components and we obtain a homeomorphism of a closed hyperbolic surface @xmath238 of genus @xmath239 . we denote the homeomorphism @xmath240 by @xmath36 .
the surface @xmath3 has an order - two symmetry @xmath241 which commutes with @xmath36 ( i. e. @xmath242 ) and which fixes the multicurve @xmath243 .
@xmath248 @xmath249 .
the homeomorphism @xmath36 is reducible and with no dehn twist on the reducing multicurve , and it fixes the measured geodesic laminations @xmath250 and @xmath251 ; cf . figure [ phi ] . 1 .
[ c1 ] an _ ideal polygon _ ( that is , a hyperbolic polygon with cusps , or , equivalently , a surface isometric to the convex hull in @xmath254 of a finite number of points at infinity ) ; 2 . [ c2 ] a _ crown _ ( that is , a hyperbolic annulus with cusps on one of its boundary components ) .
we consider a component @xmath255 of @xmath256 and we lift it to the universal cover @xmath257 .
the map @xmath36 also lifts to @xmath257 and we can choose a lift @xmath258 of @xmath255 such that it is fixed ( setwise ) by the lift @xmath259 of @xmath36 .
the ideal points of @xmath258 cut the circle at infinity into segments ( figure [ lift ] ) .
each segment between two ideal vertices of @xmath258 contains in its interior exactly one fixed point ; these fixed points of @xmath259 on @xmath260 are alternatively ideal vertices of lifts @xmath261 of components @xmath262 of @xmath253 and ideal vertices of lifts @xmath263 of components @xmath264 of @xmath265 and they are alternatively sinks and sources ; cf .
thurston and handel @xcite and @xcite .
consider now a completion @xmath266 of @xmath135 such that all the extra leaves of @xmath266 are isolated .
since @xmath36 has no dehn twists along its reducing multicurve , @xmath266 is fixed by @xmath36 .
the measured geodesic lamination @xmath141 is totally transverse to @xmath266 , that is , @xmath267 .
let @xmath268 be a measured foliation corresponding to @xmath141 . by a result of thurston ( cf .
theorem [ thm : thurston1 ] above ) , there exists a unique hyperbolic structure @xmath269 on @xmath3 such that @xmath270 , where @xmath271 is the horocyclic foliation .
since @xmath273 and since @xmath269 and @xmath274 are isometric hyperbolic structures on @xmath3 , the leaves of the foliations @xmath275 and those of @xmath276 have the same endpoints on the circle at infinity , once we lift the situation to the universal coverings . but these endpoints are those of @xmath141 and they are fixed by @xmath259 setwise . hence @xmath277 ( here , equality forgets about the transverse measures ) , or , equivalently , @xmath278 setwise .
let @xmath279 be the multiplicative factor : @xmath280 then , l. bers , an extremal problem for quasiconformal mappings and a theorem by thurston , acta math . vol .
141 , no . 1 ( 1978 ) , 7398 .
l. bers , a remark on mumford s compactness theorem , israel journal of math .
12 , no . 4 ( 1972 ) , 400407 .
p. buser , geometry and spectra of compact riemann surfaces , birkhuser , 1992 .
g. daskalopoulos and r. wentworth , classification of weil - petersson isometries , amer .
125 , no . 4 ( 2003 ) , 941975 .
a. fathi , f. laudenbach and v. poenaru , travaux de thurston sur les surfaces ( second edition ) , astrisque , vol
. 6667 .
a. papadopoulos , on thurston s boundary of teichmller space and the extension of earthquakes , topology and its application 41 ( 1991 ) , 147177 .
a. papadopoulos and g. thret , on the topology defined by thurston s asymmetric metric , math .
soc . , 142 ( 2007 ) , 487496 . a. papadopoulos and g. thret , on teichmller s metric and thurston s asymmetric metric on teichmller space . a. papadopoulos ( ed . ) , handbook of teichmller theory .
volume i. zrich , european mathematical society , irma lectures in mathematics and theoretical physics 11 , 111 - 204 ( 2007 ) .
w. p. thurston , on the geometry and dynamics of diffeomorphisms of surfaces , bull .
( n.s.)19(2 ) ( 1988 ) , 417431 .
w. p. thurston , minimal stretch maps between hyperbolic surfaces , arxiv : math.gt/9801039 .
|
we study the action of the elements of the mapping class group of a surface of finite type on the teichmller space of that surface equipped with thurston s asymmetric metric .
we classify such actions as elliptic , parabolic , hyperbolic and pseudo - hyperbolic , depending on whether the translation distance of such an element is zero or positive and whether the value of this translation distance is attained or not , and we relate these four types to thurston s classification of mapping classes .
the study is parallel to the one made by bers in the setting of teichmller space equipped with teichmller s metric , and to the one made by daskalopoulos and wentworth in the setting of teichmller space equipped with the weil - petersson metric .
ams mathematics subject classification : 32g15 ; 30f60 ; 57m50 ; 57n05 .
keywords : teichmller space ; thurston s asymmetric metric ; mapping class group .
|
hard x - ray emission in the brightest low - mass x - ray binaries ( hereafter lmxbs ) , the so - called z - sources , was occasionally detected in the past ( see e.g. peterson & jacobson 1966 ) .
these results received relatively little attention , mostly because the lack of a broad - band spectral coverage did not permit to establish whether an extra component was indeed required to fit the hard spectrum of these sources .
renewed interest in the hard x - ray emission properties of bright lmxbs was motivated by recent broad - band studies mainly performed with rxte ( 2 200 kev ) and bepposax ( 0.1 200 kev ) .
these have shown that most z - sources display variable , hard power - law shaped components , dominating their spectra above @xmath2 kev ( see di salvo & stella 2000 for a review ) .
the hard component detected in bright ( otherwise soft ) lmxbs can be fitted by a power law , with photon index in the range 1.93.3 , contributing from 1% to 10% of the observed ( @xmath3 kev ) source luminosity .
the presence of these components in z sources seems sometimes to be related to the source state or its position in the x - ray color - color diagram ( hereafter cd ) .
the clearest example to date is in the bepposax observation of gx 17 + 2 , where the hard component ( a power - law with photon index of @xmath4 ) showed the strongest intensity in the horizontal branch ( hb ) of its cd ( di salvo et al .
a factor of 20 decrease was observed when the source moved from the hb to the normal branch ( nb ) , i.e. from low to high ( inferred ) mass accretion rate .
a hard tail was also detected in almost all the currently known z sources ( e.g. di salvo et al .
2001 ; iaria et al . 2001
; di salvo et al .
2002 ; asai et al .
the fact that a similar hard component has been observed in several z sources indicates that this is probably a common feature of these sources .
however , the origin of this hard component is still poorly understood .
while in most cases the hard component becomes weaker at higher accretion rates , hexte observations of sco x1 showed a hard power - law tail in 5 out of 16 observations , without any clear correlation with the position in the cd ( damico et al .
the thermal vs. non - thermal nature of this component remains to be addressed , yielding important information on the production mechanism .
sco x1 , the brightest persistent x - ray source in the sky , is also the brightest radio source among neutron star lmxbs , with a mean radio flux about 10 times higher than that of the other z sources ( e.g. fender & hendry 2000 ) .
a hard x - ray power - law component has been observed in rxte / hexte ( 20200 kev ) data of this source ( damico et al .
as already mentioned , contrary to the case of gx 17 + 2 , in sco x1 the flux of this component was observed to vary without any clear correlation with the position in the cd .
interestingly , strickman & barret ( 2000 ) report that the hard x - ray emission present in osse data of sco x1 may be correlated with periods of radio flaring . to study the hard x - ray emission in , the brightest of these sources , as well as its correlation with other source properties ( such as radio emission and fast timing variability )
, we have performed a campaign of observations of sco x1 with integral and rxte .
part of these observations were also done simultaneously with radio vlbi observations ( which will be discussed elsewhere ) .
the integral spectrum of shows with high statistical significance the presence of a hard ( power - law ) component , without any clear exponential cutoff up to @xmath5 kev .
we also find clear evidence that the intensity of this component is correlated with the position of the source in the x - ray cd .
sco x-1 was observed during two complete integral revolutions on 2003 july 30 august 1 and 2003 august 11 13 .
the integral payload consists of two main @xmath6-ray instruments , a spectrometer , spi ( vedrenne et al .
2003 ) and an imager , ibis , and of two monitor instruments , the x - ray monitor jem - x ( 3 35 kev , lund et al .
2003 ) and the optical monitoring camera ( v band , 500 - 600 nm , mas - hesse et al .
the ibis instrument ( ubertini et al . 2003 ) covers the energy range between 20 kev and 8 mev with two detectors , isgri ( lebrun et al .
2003 ) and picsit ( di cocco et al . 2003 ) and has a field of view of @xmath7 at half sensitivity ( @xmath8 fully coded ) with a point spread function of @xmath9 fwhm .
the spi also covers the energy range between 20 kev and 8 mev with a fov of @xmath10 diameter ( @xmath11 fully coded ) and an angular resolution of @xmath12 .
observations were performed in the usual integral format , i.e. the observation was split in separate exposures ( `` science windows '' , or scws ) , each lasting @xmath13 sec , followed by a 5 minutes slew .
the exposures were arranged following a @xmath14 dither pattern .
the ibis and spi instruments were operated in standard mode during the whole observation , while jem - x was in a non - standard mode ( spec ) which , unfortunately , is not yet calibrated .
we therefore discarded the jem - x data from our analysis .
the effective exposure time was 372 ks for ibis and 358 ks for spi .
the ibis data were analyzed using the standard analysis procedures of osa version 5 and the latest response matrices ( october 2004 ) rebinned to 26 channels between 15 and 800 kev and the latest spectral extraction routines ( goldwurm et al .
a systematic error of @xmath15 was applied to all the integral / isgri spectra . during all the observations ,
sco x-1 was the only source detected in the wide fov of the instruments .
the source coordinates as derived from the isgri mosaic image in the 20 35 kev energy band are : ra = 16 19 54.9 , dec = 15 38 34.4 ( uncertainty @xmath16 , 1 sigma confidence level ) at about @xmath17 from the simbad position ( mcnamara et al .
2003 ) . to study the source spectral state in the standard x - ray band , we also analyzed data from the proportional counter array ( pca ; zhang et al .
1993 ) on board rxte , which consists of five co - aligned proportional counter units ( pcus ) , with a total collecting area of 6250 @xmath18 and a field of view , limited by collimators , of @xmath19 fwhm , sensitive in the energy range @xmath20 kev .
the rxte observation , performed simultaneously with the integral observation , is divided into two parts of 175 ks each , approximately two weeks apart from each other .
we selected intervals for which the elevation angle of the source above the earth limb was greater than 10 degrees . on a few occasions ,
some of the five pcus were off ; we therefore used only data from pcus 2 and 3 which where on for most of the observation .
the x - ray cd of sco x-1 during the integral observations , obtained from the pca data , is shown in figure 1 . during each of the two observations the source described a fairly complete z - track in the cd . in order to check for the presence of the hard tail in our data , we first analysed the integral ( ibis / isgri , energy band 20 kev 200 kev ) spectrum integrated over the whole observation ( see figure 2 , left panel ) .
similar to what has been done for other lmxbs of the z - class , we fitted the integral spectrum with the comptonization model comptt ( whose description is given in titarchuk 1994 ) .
comptt is an analytical model describing the thermal comptonization of soft photons ( for which a wien spectrum is assumed ) inverse - compton scattered in a hot electron cloud with optical depth @xmath21 and whose temperature can range from a few kev to 500 kev .
this model includes relativistic effects and works for optically thick and optically thin regimes .
the geometry of the comptonizing cloud can be either spherical or disk - like . in this paper
we assume a spherical geometry ( this assumption only affects the value of the optical depth derived from the fit ) .
this model gives a good fit of the soft part of the spectrum up to @xmath22 kev . above this energy ,
a hard excess is clearly visible in the residuals , independently of the particular comptonization model used to fit the soft part of the spectrum .
the fit is significantly improved by adding to the comptt model a power - law with photon index @xmath23 ( this gives a reduction of the @xmath24/dof from 892/13 to 14.4/11 ) .
we tested the presence of a thermal cutoff in the hard power - law ; substituting the power law with a cutoff power - law ( that is a power law multiplied by an exponential cutoff ) does not improve the fit significantly ( the latter model gives a @xmath24/dof = 15.6/10 ) , and the temperature of the exponential cutoff is @xmath25 kev ( @xmath26 confidence level ) .
the best - fit parameters for the isgri ( 20 200 kev ) spectrum are reported in table 1 ; data and residuals with respect to the best fit model are shown in figure 2 ( left panel ) . to look for variability in the hard component with the spectral state of the source , as measured by its position in the x - ray cd
, we divided the z - track in the cd of into four parts corresponding to the hb / upper - nb , the nb , the flaring branch ( fb ) , and the nb / fb vertex , respectively .
we therefore extracted integral / isgri spectra for each of the time intervals mentioned above , resulting in four cd resolved spectra .
unfortunately there was no superposition between rxte and integral data in the fb during the first part of the observation .
the integral / isgri exposure times for the four intervals were 52.4 ks , 45.5 ks , 60.6 ks and 6 ks .
we fitted each of this spectra with comptt and a cutoff power law .
this model gave a good fit of the first three spectra , with little variability in the spectral parameters ( see table 1 ) .
for three of these spectra , the hard power law component was required in order to fit the data .
remarkably the fb spectrum did not require a power - law component and could be fitted with a simple comptonization model ( see fig . 2 , right panel ) .
including the power law in the spectral fit of the fb spectrum , with the photon index fixed at 3.1 , a good fit requires a decrease of the normalization of the power law component by a factor at least 5 with respect to the average spectrum . in table 1
we show the total x - ray flux of calculated in the @xmath27 kev and in the @xmath28 kev energy range , flux ( 2040 ) and flux ( 40200 ) , respectively . while the flux ( 2040 ) decreases from the hb to the nb / fb vertex and
then increases when the source goes to the fb , the flux ( 40200 ) always decreases , varying by about one order of magnitude when the source moves from the hb to the fb .
although from table 1 there seems to be a clear trend of the @xmath28 kev flux to decrease from the hb to the nb and fb vertex , the uncertainties on the hard x - ray flux are still quite large to draw a firm conclusion .
in fact , in order to calculate the flux in a given energy range and the associated uncertainty , xspec needs to use the total best - fit model , which includes both the comptt and the cutoff power - law component ; each of these components has several parameters and the uncertainties on all the parameters are taken into account in the calculation of the source flux and its uncertainty .
if we are interested in knowing the flux of the hard power - law component alone in a given energy range ( to see how this component evolves when the source moves in the x - ray cd ) , it is more convenient to use the model named pegpwrlw in xspec .
this model allows to directly calculate the power - law flux in a given energy range ( @xmath29 kev is our choice ) and the associated uncertainty .
we therefore substituted the cutoff power - law model with the pegpwrlw model to fit the integral spectra of .
the results of the fits with the comptt plus pegpwrlw model are also reported in table 1 ( since the parameters of the comptt did not change significantly these are shown only once in the table ) .
the decrease of the power - law @xmath29 kev flux along the cd is evident .
for clarity , we have also plotted these power - law fluxes in figure 1 ( inset ) .
we report on a spectral analysis of a simultaneous @xmath30 ks - long observation of with integral and rxte .
we show that the addition of a hard power - law component dominating the x - ray spectrum above @xmath2 kev proves necessary for a good fit of the integral ( @xmath29 kev ) spectrum .
the power law is quite steep , with a photon index of about 3.1 , contributing up to 12% of the observed @xmath31 kev luminosity , and does not show any evidence of a high energy cutoff up to @xmath32 kev . similarly to what was observed in the bepposax observation of the z source gx 17 + 2 , the presence of the hard component in seems to be related to the position of the source in the z track ; in fact we observe a clear trend of the @xmath29 kev power - law flux to decrease when the inferred mass accretion rate increases ( i.e. from the hb to the nb / fb vertex ) . at
the highest inferred mass accretion rate ( i.e. in the fb of the x - ray cd ) , the hard x - ray emission seems to disappear completely ( the hard x - ray flux decreases by about one order of magnitude with respect to the flux measured in the hb / upper - nb ) .
note , however , that the behavior of the hard component in the fb may be variable ( c.f .
damico et .
al . 2001 ) .
one of the most interesting results of this analysis is that the hard power law detected in does not show any evidence of a high energy cutoff , with lower limits on the electron temperature that are in most cases above 100 kev ( the lower limit on a thermal cutoff in the averaged integral spectrum , where the statistics is the highest , is about 200 kev ) .
such high temperatures are not expected in a system like , because of the strong compton cooling due to the primary soft spectrum , where most of the energy is emitted .
these results therefore indicate that , in analogy with the steep hard power - law emission detected in some soft states ( e.g. intermediate and very high ) of systems containing a black hole candidates , which do not show any energy cutoff up to a few hundreds kev , the hard power law observed in z sources may be of non - thermal origin .
the most probable origin of these components are therefore comptonization in a hybrid thermal / non - thermal corona ( where the maxwellian velocity distribution of the electrons have a non - thermal high - velocity tail , e.g. poutanen & coppi 1998 ; see farinelli et al .
2005 and da et al .
2006 for a successful application of this model to the case of gx 17 + 2 and , respectively ) .
for instance , it is possible that a hard population of electrons is accelerated in internal shocks at the base of a jet .
otherwise the hard power law may originate in a mildly relativistic ( @xmath33 ) bulk motion of matter close to the compact object .
therefore non - thermal comptonization in outflows ( or jets ) may be the origin of these components , with flatter power laws corresponding to higher optical depths of the scattering medium and/or higher bulk electron velocities ( e.g. psaltis 2001 ) .
it has been proposed that non - thermal , high energy electrons , responsible for the hard tails observed in z sources , might be accelerated in a jet ( di salvo et al .
this is also confirmed by the observed correlation between the strength of the hard x - ray emission and the position of the source in the x - ray cd .
in fact all z sources are detected as variable radio sources with the highest radio flux associated with the hb and the lowest with the fb ( migliari & fender 2006 , and references therein ) .
in other words , both the radio emission from these objects , which is probably due to jets ( fender & hendry 2000 ) , and the hard x - ray emission are anticorrelated with the inferred mass accretion rate . in this respect , it will be very important to study the direct correlation between the presence and strength of the hard x - ray emission and the radio activity of the source .
the radio ( vla ) observations of , simultaneous to part of the integral observations , is under investigation and will be discussed elsewhere .
asai , k. , dotani , t. , mitsuda , k. , et al .
1994 , pasj , 46 , 479 da , a. , et al .
2006 , apj , submitted damico , f. , heindl , w.a . ,
rothschild , r.e . , gruber , d.e .
2001 , apj 547 , l147 di cocco , g. , caroli , e. , celesti , e. , et al .
2003 , a&a , 411 , l189 di salvo , t. , stella , l. , robba , n.r . , et al .
2000 , apj 544 , l119 di salvo , t. , & stella , l. 2000 , proceedings of the xxxviith rencontres de moriond , eds .
a. goldwurm , d. neumann , j. tran thanh van , p. 67
di salvo , t. , robba , n.r . ,
iaria , r. , et al .
2001 , apj , 554 , 49 di salvo , t. , farinelli , r. , burderi , l. , et al .
2002 , a&a , 386 , 535 farinelli , r. , et al .
2005 , a&a , 434 , 25 fender , r. p. , & hendry m. a. 2000 , mnras , 317 , 1 fomalont , e. b. , geldzahler , b. j. , & bradshaw , c. f. 2001a apj , 553 , l27 fomalont , e. b. , geldzahler , b. j. , & bradshaw , c. f. 2001b apj , 558 , 283 goldwurm , a. , david , p. , foschini , l. , et al .
2003 , a&a , 411 , l223 iaria , r. , burderi , l. , di salvo , t. , et al .
2001 , apj , 547 , 412 lebrun , f. , leray , j. p. , lavocat , p. , et al .
2003 , a&a , 411 , l141 lund , n. , budtz - joergensen , c. , westergaard , n. l. , et al .
2003 , a&a , 411 , l231 mcnamara , b. j. , et al .
2003 , aj , 125 , 1437 mas - hesse , j. m. , gimenez , a. , culhane , j. l. , et al .
2003 , a&a , 411 , l261 migliari , s. , & fender , r. p. , 2006 ,
mnras , in press peterson , l. e. , & jacobson , a. s. , 1966 , apj , 145 , 962 poutanen , j. , & coppi , p. s. , 1998 , phys .
scripta , t77 , 57 psaltis , d. , 2001 , apj , 555 , 786 strickman , m. , & barret , d. 2000 , proc . of the fifth compton symposium , eds . m. l. mcconnell and j.m .
ryan , aip conference proceedings , vol .
510 , p. 222
titarchuk , l. 1994 , apj , 434 , 570 ubertini , p. , lebrun , f. , di cocco , g. , et al .
2003 , a&a , 411 , l131 vedrenne , g. , roques , j .-
p . , schnfelder , v. , et al .
2003 , a&a , 411 , l63 zhang , w. , giles , a.b .
, jahoda , k. , soong , y. , swank , j.h . ,
morgan , e.h .
1993 , proc .
spie 2006 , 324 the model consists of a comptonized spectrum modeled by comptt , and a cutoff power law or the pegpwrlw model to fit the hard component .
@xmath34 is the temperature of the seed photon ( wien ) spectrum , @xmath35 the electron temperature and @xmath21 the optical depth in a spherical geometry .
fluxes ( 2040 ) and ( 40200 ) are the total flux from the source in units of @xmath36 erg @xmath37 s@xmath38 in the 20 40 kev energy range and in units of @xmath39 erg @xmath37 s@xmath38 in the 40 200 kev energy range , respectively . on the other hand , flux ( 20200 )
is the power - law flux in the 20200 kev range in units of @xmath39 erg @xmath37 s@xmath38 .
@xmath40 is the variation of @xmath24 for the addition of a power law to the model .
note that the addition of a power law does not change the @xmath24 for the fb spectrum .
@xmath41 is the final reduced @xmath24 for the best - fit model .
all the uncertainties are calculated at 90% confidence level , upper limits at 95% confidence level .
kev]/[@xmath42 kev ] and hard color is the ratio [ @xmath43 kev]/[@xmath44 kev ] , respectively . to take into account possible small gain changes during the rxte observations , the hard and soft colors of are normalized with respect to the colors ( calculated in the same energy range ) obtained from rxte observations of crab acquired close to the dates of our observations .
only data form pcus 2 and 3 , which were on for most of the observation , have been used .
the four colors indicate the four regions ( also indicated with numbers from 1 to 4 ) in which the cd has been divided , from which integral spectra were extracted . * inset : * the measured power - law flux in the @xmath29 kev range plotted for each of the four cd - resolved spectra .
[ fig1],width=264,height=302 ]
|
we present here simultaneous integral / rxte observations of sco x-1 , and in particular a study of the hard x - ray emission of the source and its correlation with the position in the z - track of the x - ray color - color diagram . we find that the hard x - ray ( above about 30 kev ) emission of sco x-1 is dominated by a power - law component with a photon index of @xmath0 .
the flux in the power - law component slightly decreases when the source moves in the color - color diagram in the sense of increasing inferred mass accretion rate from the horizontal branch to the normal branch / flaring branch vertex .
it becomes not significantly detectable in the flaring branch , where its flux has decreased by about an order of magnitude .
these results present close analogies to the behavior of gx 17 + 2 , one of so - called sco - like z sources .
finally , the hard power law in the spectrum of sco x-1 does not show any evidence of a high energy cutoff up to @xmath1 kev , strongly suggesting a non - thermal origin of this component .
|
models possessing global and/or local discrete @xmath0 symmetry play an important role in many branches of physics ranging from the solid state physics to the description of the universality features of the deconfining transition in @xmath25 gauge theories . in this paper
we are interested in the @xmath0 lattice gauge theory ( lgt ) at zero temperature . assigning the gauge fields @xmath26 to the links of a simple hypercubic lattice ,
the most general action of the isotropic @xmath0 lgt can be written as @xmath27 where @xmath28 , @xmath29 , denotes the unit vector in the @xmath30-th direction .
similarly , the most general action of the @xmath0 spin model is given by @xmath31 in both cases we used the convention @xmath32 the standard potts model corresponds to the choice when all @xmath33 are equal .
then , the sum over @xmath34 reduces to a delta - function on the @xmath0 group .
the conventional vector model corresponds to @xmath35 for all @xmath36 . for @xmath37
the potts and vector models are equivalent . two - dimensional ( 2d )
standard and vector @xmath0 lgts are exactly solvable both in the finite volume and in the thermodynamic limit .
they exhibit no phase transition at any finite value of the coupling constant @xmath38 .
in particular , the rectangular @xmath39 wilson loop in the representation @xmath34 obeys the area law @xmath40 thus implying a permanent confinement of static charges .
for example , the string tension of the vector model in the thermodynamic limit reads @xmath41 \
> 0 \ \ , \ \
c_k(n,\beta ) \ = \
\sum_{r=-\infty}^{\infty } \
i_{nr+k}(\beta ) \ . % \label{2d_sigma}\end{aligned}\ ] ] here , @xmath42 is the modified bessel function
. no exact solution has been found for any @xmath0 model in @xmath9 , where the phase structure becomes highly non - trivial .
while the phase structure of the general model defined by ( [ action_gauge ] ) remains unknown , it is well established that potts models and vector models with only @xmath43 non - vanishing have one phase transition from a confining phase to a phase with vanishing string tension @xcite . via duality ,
@xmath0 gauge models can be exactly related to @xmath9 @xmath0 spin models . in particular
, a potts gauge theory is mapped to a potts spin model , and such a relation allows to establish the order of the phase transition .
hence , potts lgts with @xmath44 have second order phase transition , while for @xmath45 one finds a first order phase transition .
since @xmath46 lgt is equivalent to the ising model , its critical behavior is well known ( see refs .
@xcite and references therein ) .
generally , the @xmath0 global symmetry of the finite - temperature @xmath47 @xmath25 gauge theory motivated thorough investigations , both analytical and numerical , of the @xmath9 spin models , especially for @xmath37 @xcite ( for more recent studies , see @xcite and references therein ) .
the @xmath9 potts models for @xmath48 have been simulated in @xcite and studied by means of the high - temperature expansion in @xcite .
surprisingly , much less is known about the critical behavior of @xmath0 vector lgts when @xmath5 .
they have been studied numerically in @xcite up to @xmath49 on symmetric lattices with size @xmath50 $ ] .
it was confirmed that zero - temperature models possess a single phase transition which disappears in the limit @xmath51 .
a scaling formula proposed in @xcite shows that the critical coupling diverges like @xmath52 for large @xmath1 .
thus , the @xmath53 lgt has a single confined phase in agreement with theoretical results @xcite .
we are not aware , however , of any detailed study of the critical behavior of the vector models with @xmath54 in the vicinity of this single phase transition .
slightly more is known about the critical properties of @xmath0 vector spin models . in particular
, it has been suggested that all vector spin models exhibit a single second order phase transition @xcite .
an especially detailed study was performed on the @xmath55 model , because the @xmath55 global symmetry appears as an effective symmetry of the @xmath56 antiferromagnetic potts model @xcite .
the computed critical indices suggest that the @xmath55 vector model belongs to the universality class of the @xmath9 @xmath6 model .
an interesting feature of the @xmath55 model and , possibly , of all vector models with @xmath5 , is the appearance of an intermediate rotationally symmetric region below the critical temperature of the second order phase transition .
the mass gap was however found to be rather small , but non - vanishing in this region @xcite .
combined with a renormalization group ( rg ) study , the analysis concluded that this intermediate region presents a crossover to a low - temperature massive phase , where the discreteness of @xmath55 plays an essential role @xcite . the main goal of the present work is to fill the gap in our knowledge about the critical behavior of the @xmath9 @xmath0 lgts .
another motivation comes from our recent studies of the deconfinement transition in the @xmath0 vector lgt for @xmath5 at finite temperatures @xcite .
the major findings of these papers was the demonstration of two phase transitions of the berezinskii - kosterlitz - thouless type and the existence of an intermediate massless phase .
the critical indices at these transitions have been found to coincide with the indices of the @xmath57 vector spin models .
an interesting question then arises regarding the construction of the continuum limit of the finite - temperature models in the vicinity of the critical points . for this to accomplish it might be useful , and even necessary , to know the scaling of quantities such as string tension , correlation length , _ etc .
_ near the critical points of the corresponding zero temperature models . in this work we are going to
: * perform an analytical study of the general @xmath9 @xmath0 lgt for various @xmath1 using a `` phenomenological '' rg ; * locate critical points of the vector models via monte carlo simulations of the dual of the @xmath0 lgt and determine their scaling with @xmath1 ; * compute some critical indices and establish the universality class of the models ; * illustrate the existence of a rotationally symmetric region within the ordered phase for all @xmath8 at least in the finite volume .
this paper is organized as follows . in section
2 we formulate our model and recall the exact duality relation with a generalized @xmath9 @xmath0 spin model .
section 3 is devoted to a rg study of the models . within a modified version of the phenomenological rg
, we explore the space of coupling constants , find fixed points and calculate the critical index @xmath58 . in section 4
we present the setup of monte carlo simulations , define the observables used in this work and present the numerical results .
in particular , we locate the position of critical points and compute various critical indices at these points . as a cross - check , we also simulated @xmath59 models for which high - precision results .
the same section deals also with the computation of the average action and the heat capacity in the vicinity of critical points and the derivation of the index @xmath7 from a finite size scaling ( fss ) analysis of the heat capacity . in section 5
we discuss some findings regarding the symmetric region below the critical point .
all results are finally summarized in section 6 .
we work on a @xmath9 lattice @xmath60 with linear extension @xmath61 ; @xmath62 , where @xmath63 $ ] denote the sites of the lattice and @xmath28 , @xmath29 , denotes the unit vector in the @xmath30-th direction .
periodic boundary conditions ( bc ) on gauge fields are imposed in all directions .
we introduce conventional plaquette angles @xmath64 as @xmath65 the @xmath9 @xmath0 lgt on an isotropic lattice can generally be defined as @xmath66 the most general @xmath0-invariant boltzmann weight is @xmath67 \ . \label{qpgen}\ ] ] the @xmath53 lgt is defined as the limit @xmath51 of the above expressions . to study the phase structure of @xmath9 @xmath0 lgts
one can map the gauge model to a generalized @xmath9 @xmath0 spin model with the action given by eq .
( [ action_spin ] ) .
the relation between spin @xmath68 and gauge @xmath69 couplings can be computed exactly and reads @xmath70 \
\cos \left(\frac{2 \pi p k}{n } \right ) \ , \label{couplings}\ ] ] where the dual boltzmann weight is defined as @xmath71 in what follows we are going to simulate the @xmath0 vector lgt .
in this case we use @xmath72 and the dual weight becomes @xmath73 an example of the explicit relations between couplings in this case together with some further important comments on the relations can be found in @xcite . with the goal of performing rg transformations
, it is somewhat more convenient to use a different but equivalent representation for the boltzmann weight , namely @xmath74 \ = \ \sum_{k=0}^{n-1 } \ t_k \
\exp\left [ \frac{2\pi i}{n } \ k \ s \right ] \ .
\label{bweight1}\ ] ] the set of coupling constants @xmath75 can be chosen to satisfy @xmath76 the coupling constants @xmath77 and @xmath33 can be connected to each other via the fourier transform on the @xmath0 group .
the dual of the partition function ( [ ptdef ] ) can then be presented as @xmath78 here , the summations over @xmath79 , enforce the global bianchi constraints due to the periodic bc and @xmath80 on a set of links dual to any fixed closed surface wrapping the original lattice in the directions perpendicular to @xmath30 , otherwise @xmath81 .
as a first step we consider the general @xmath9 @xmath0 lgt with the boltzmann weight ( [ bweight1 ] ) and study its phase structure with the help of the phenomenological renormalization group ( ph rg ) @xcite . for the @xmath9 ising model
this rg was used in @xcite .
as is well known , the ph rg gives in many cases not only the qualitatively correct phase diagram of a model , but also a good quantitative approximation for the observables near the critical points , and this approximation can be systematically improved .
we follow the general strategy described in the original papers @xcite and only briefly outline our modifications ( for a detailed account of these modifications , together with application to other models , see @xcite ) . as is common for this type of rg , we preserve the mass gap during the rg steps .
we proceed as follows . 1 .
as a starting point we always use the dual formulations , _
i.e. _ eq . ( [ pt_dual ] ) in the present case .
2 . as a simplest @xmath9 strip
, we take a lattice @xmath82 with periodic bc in both transverse directions and periodic or free bc in the longitudinal direction .
correlation functions are computed on @xmath83 and on a one - dimensional lattice for all representations .
the requirement of mass gap preservation leads to equations for the fixed points @xmath84 ^ 2 \ , \label{rrm2}\ ] ] where @xmath85 is the eigenvalue of the transfer matrix corresponding to the correlation function in the representation @xmath86 .
this set of equations is treated as the recursion relations for the renormalized coupling constants .
the summation over the spins lying in one plane perpendicular to the longitudinal direction introduces all possible interactions between the remaining spins .
the transfer matrix is constructed for the evolution of all independent couplings of this general interaction .
this substantially reduces the size of the matrix .
5 . combining the ph rg in the form described above with the cluster decimation approximation of ref .
@xcite , we can construct new partition and correlation functions on a decimated lattice with double lattice spacing .
the next iteration is performed with newly computed constants @xmath87 . in the framework of this approach
we have explored the phase structure of the generalized @xmath0 model . only in the case of the standard potts model
we can restrict ourselves to one iteration , since the rg steps do not generate new interactions .
for the general case one should perform many iterations to locate precisely the critical points .
the critical indices are calculated in the fixed point of the iterations where the critical points of the models of a given universality class are flowing to .
thorough discussion of the general case will be given in @xcite .
here we report on the results for the vector lgts relevant for this paper . .@xmath9
@xmath0 standard potts models : @xmath88 from the ph rg ( column two ) and from the monte carlo simulations of ref .
@xcite ( column three ) .
@xmath9 @xmath0 vector models : @xmath88 from the ph rg ( column 4 ) and from the monte carlo simulations of this work ( column five ) ; critical index @xmath58 from the ph rg . [ cols="^,^,^,^,^,^ " , ] [ indices_nu_zero_hc ]
@xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] @xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] @xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] @xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] @xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] @xmath55 vector model on a @xmath89 lattice for @xmath90 ( left ) and @xmath91 ( right).,title="fig:",scaledwidth=44.0% ] another interesting phenomenon we have encountered during our study is the appearance of a symmetric phase just below @xmath88 for all @xmath8 . that such phase exists in the vector @xmath55 spin model
has been known for a long time @xcite . here
we confirm its existence for all vector @xmath92 lgts .
the phase exhibits itself , _
e.g. _ , in the behavior of magnetization . as an example
, we give in fig [ scatter_plots ] the scatter plots of magnetization and rotated magnetization , together with the histogram of the magnetization angle for @xmath55 on a @xmath89 lattice .
one sees that below the critical coupling at @xmath91 the symmetry is not broken on this lattice . only starting from approximately @xmath90 one
can observe the appearance of a symmetry - broken phase .
in addition , we have studied the behavior of the population susceptibility below @xmath93 .
fig [ chipplot ] shows that it has a second broad maximum , which slowly moves to @xmath93 with increasing lattice size .
a similar picture is observed for all @xmath8 .
while , however , for @xmath94 the peak of the population susceptibility moves rather fast and practically collapses with the peak at the critical coupling on the largest available lattice @xmath4 , for larger @xmath1 the peak stays rather far from the corresponding critical coupling , even for @xmath4 ( with our data we can not even exclude a situation when the convergence of the second maximum is logarithmic ) .
we can imagine two scenarios to explain such behavior : 1 .
this symmetric phase exists only in finite volume .
when the lattice size increases , the second maximum approaches the critical coupling and , eventually , the symmetric region shrinks and disappears .
the explanation proposed in @xcite might work in this case , too .
namely , the symmetric phase on the finite lattice is a phase with a very small mass gap and describes a crossover region to the symmetry - broken phase .
2 . for @xmath95
the second maximum of the population susceptibility stays away from the critical couplings even in the infinite volume limit . in this case
it might correspond to some higher order phase transition and the symmetric phase with tiny or even vanishing mass gap exists also in the thermodynamic limit . in both cases
it is tempting to speculate that this symmetric region is reminiscent of the massless phase which appears in these models at finite temperature @xcite .
whichever scenario of the above two is realized , one needs to study the models on much larger lattices to uncover it .
in this paper we have studied the @xmath9 @xmath0 lgt at zero temperature aiming at shedding light on the nature of phase transitions in these models for @xmath54 .
this study was based on the exact duality transformations of the gauge models to generalized @xmath9 @xmath0 spin models . in section 2
we presented an overview of the exact relation between couplings of these two models . in section 3
we have studied the models analytically using a version of the phenomenological rg based on the preservation of the mass gap combined with a cluster decimation approximation .
this study provided us with an approximate location of the critical couplings as well as with the value of the index @xmath58 .
these calculations show that the index @xmath58 is approximately the same for @xmath96 , @xmath97 , indicating that these two models might belong to the same universality class .
we find @xmath98 and approximately equal for @xmath5 .
hence , @xmath99 vector lgts belong to a different universality class .
the numerical part of the work has been devoted to the localization of the critical couplings , computation of the various critical indices and check of the hyperscaling relation .
the main results can be shortly summarized as follows : * we have determined numerically the position of the critical couplings for various @xmath0 models . for @xmath37
we find a reasonable agreement with the values quoted in the literature . for larger @xmath1
, we have significantly improved the values given in @xcite .
this allowed us to improve the scaling formula for the critical couplings with @xmath1 . *
the critical indices @xmath58 and @xmath100 derived here for @xmath96 suggest that these models are in the universality class of the @xmath9 ising model , while our results for all @xmath5 hint all vector @xmath92 lgts belong to the universality class of the @xmath9 @xmath6 model .
this is especially evident from the value of the index @xmath58 , which stays very close to the @xmath6 value , @xmath101 , given in @xcite .
the index @xmath7 in this case takes a small negative value .
it thus follows that a third order phase transition takes place for @xmath8 .
* a careful investigation of the specific heat and of the index @xmath7 extracted from it suggests however a more complicated picture of the critical behavior . in this case
we find a value which roughly agrees with the value of @xmath9 ising model for all @xmath1 studied .
the fact that we observe two different values of the index @xmath7 dependently on whether we approach the critical point from below or from above leads to the conclusion that the first derivative of the free energy could exhibit a cusp in the thermodynamic limit if @xmath5 .
* our data also revealed the existence of a symmetric phase for all @xmath0 vector lgts if @xmath5 .
however , substantially larger lattices are required to see if this phase survives the transition to the thermodynamic limit .
the work of ukrainian co - authors was supported by the ukrainian state fund for fundamental researches under the grant f58/384 - 2013 .
numerical simulations have been partly carried out on ukrainian national grid facilities .
thanks for warm hospitality the dipartimento di fisica delluniversit della calabria and the infn gruppo collegato di cosenza during the final stages of this investigation .
|
three - dimensional @xmath0 lattice gauge theories at zero temperature are studied for various values of @xmath1 . using a modified phenomenological renormalization group
, we explore the critical behavior of the generalized @xmath0 model for @xmath2 .
numerical computations are used to simulate vector models for @xmath3 for lattices with linear extension up to @xmath4 .
we locate the critical points of phase transitions and establish their scaling with @xmath1 .
the values of the critical indices indicate that the models with @xmath5 belong to the universality class of the three - dimensional @xmath6 model .
however , the exponent @xmath7 derived from the heat capacity is consistent with the ising universality class .
we discuss a possible resolution of this puzzle .
we also demonstrate the existence of a rotationally symmetric region within the ordered phase for all @xmath8 at least in the finite volume .
* critical behavior of @xmath9 @xmath0 lattice gauge theories at zero temperature * 0.3 cm o. borisenko@xmath10 , v. chelnokov@xmath11 , g. cortese@xmath12 , m. gravina@xmath13 , a. papa@xmath14 , i. surzhikov@xmath15 0.6 cm @xmath16 _ bogolyubov institute for theoretical physics , _ _ national academy of sciences of ukraine , _ _ 03680 kiev , ukraine _
0.2 cm @xmath17 _ instituto de fsica terica uam / csic , _
_ cantoblanco , e-28049 madrid , spain _ _ and departamento de fsica terica , _ _ universidad de zaragoza , e-50009 zaragoza , spain _ 0.2 cm @xmath18 _ dipartimento di fisica , universit della calabria , _ _ and istituto nazionale di fisica nucleare , gruppo collegato di cosenza _
_ i-87036 arcavacata di rende , cosenza , italy _ 0.6 cm ' '' '' : @[email protected] , @[email protected] , @[email protected] , @[email protected] , @[email protected] , @[email protected]
|
in his foundational work on buildings , j.tits @xcite proved that all thick irreducible spherical and affine buildings of rank @xmath0 ( including nondiscrete ones ) are associated with algebraic groups and , hence , have weyl groups coming from root systems .
he then showed @xcite how to use the _ free construction _ to prove existence of thick spherical buildings of rank 2 modeled on arbitrary finite coxeter groups ( see definition [ modeled ] ) .
such buildings are necessarily infinite except for dihedral groups of the order @xmath1 .
m.ronan @xcite constructed a variety of irreducible rank 2 discrete affine buildings not coming from algebraic groups .
the corresponding finite coxeter groups are necessarily of crystallographic type , i.e. , they are the dihedral groups @xmath2 .
however , examples of nondiscrete affine buildings of rank 2 corresponding to the rest of finite dihedral groups remained elusive .
the main result of this paper is a construction of such buildings .
consider the dihedral group @xmath3 of order @xmath4 .
take a countable dense subgroup @xmath5 of translations of the apartment @xmath6 , invariant under the action of @xmath7 by conjugations , so that for the group @xmath8 ( which is , in fact , a coxeter group ) , every vertex of the coxeter complex @xmath9 is special .
such coxeter complexes will be called _
special_. for instance , when @xmath10 is a prime number , the complex @xmath9 is special no matter what @xmath5 is .
the main goal of this paper is to prove [ main ] for every special coxeter complex @xmath9 with nondiscrete countable group @xmath11 , there exists a thick euclidean building @xmath12 of rank 2 modeled on @xmath9 . after proving this theorem we found out that for @xmath13 , existence of thick euclidean buildings ( for some countable @xmath11 ) was also proven in a recent work of p.hitzelberger , l.kramer and r.weiss @xcite by a totally different method : their buildings are embedded in buildings of the type @xmath14 .
it is quite likely that theorem [ main ] holds for semidirect products @xmath11 of @xmath7 with arbitrary ( nondiscrete ) groups of translations @xmath15 , or , at least countable ones .
the strategy for proving theorem [ main ] is the same as in tits free construction of rank 2 spherical buildings , see @xcite , i.e. , we run a _ `` ponzi - scheme '' _ : we inductively construct a sequence of cat(0 ) spaces @xmath16 where each @xmath17 is _ finite _
, i.e. , is a finite union of apartments .
( the embeddings @xmath18 are _ not _ isometric . )
each @xmath17 is modeled on @xmath19 but fails , rather badly , the rest of axioms of a thick building , as `` most '' pairs of points are not contained in a common apartment .
given points ( say , vertices ) @xmath20 , we obtain @xmath21 by attaching an apartment to @xmath17 along a closed convex subcomplex @xmath22 containing @xmath23 . then @xmath24 belong to a common apartment .
this , of course , produces more points in @xmath21 which do not belong to a common apartment , as well as more walls which are not _
thick_. however , these will be taken care of at some later stages of the construction . the hardest part of the proof is to establish the following _ combinatorial convexity theorem _ , proven in sections [ convexity ] and [ generalized ] : [ conv ]
let @xmath12 be a cat(0 ) metric space , which is a finite space modeled on @xmath19 .
then , given any pair of points @xmath25 and germs of weyl chambers @xmath26 with the tips at @xmath27 , there exists a _ planar _
convex subcomplex @xmath28 containing @xmath27 and @xmath29 .
the same applies to each pair of _ super - antipodal _
chambers @xmath30 at infinity , i.e. , pair of chambers which admit regular points connected by a geodesic in @xmath12 .
planarity of @xmath31 means existence of a _ weakly isometric embedding
_ @xmath32 .
( we introduce @xmath33-isometry and weak isometry in sections [ metric ] and [ spaces ] as certain generalizations of the ordinary isometry between metric spaces . )
the embedding @xmath32 allows one to attach @xmath34 to @xmath12 along @xmath31 so that @xmath35 is again a cat(0 ) space which is finite , modeled on @xmath19 .
the convex subcomplexes @xmath31 used in our paper are ( extended ) _ corridors _ obtained by taking a finite union of _ tiles _ in @xmath12 .
the appearance of corridors in this context is natural since , in the case when @xmath12 is a euclidean building , every pair of vertices @xmath25 admits a combinatorial convex hull , called a _ weyl hull _ , which is a tile with the vertices @xmath23 . since each @xmath17 will have only countably many corridors , we enumerate them and form the sequence @xmath36 so that for every convex corridor @xmath31 in @xmath37 there exists @xmath38 such that @xmath21 is formed by attaching an apartment along a convex corridor @xmath39 containing @xmath31 .
organization of the paper . in sections [ complexes ] ,
[ metric ] , [ spaces ] we review standard concepts related to buildings : coxeter complexes , cat(0 ) and cat(1 ) metric spaces and spaces modeled on coxeter complexes .
we review the building axioms in section [ spaces ] .
the only nonstandard material in these sections is the one of _ @xmath33-isometries _ between metric spaces and the corresponding concept of _ weak isometries _ in the context of spaces modeled on @xmath19 . in section [ corridors ]
we define the main technical tool of this paper _ corridors _ , which are finite unions of _ tiles _ , parallelograms in @xmath34 with the angles @xmath40 and @xmath41 . in section [ orders ]
we define two orders on corridors and prove existence of maximal elements with respect to these orders .
the _ maximal corridors _ are proven to be convex in section [ convexity ] . in the same section
we prove a weaker version of theorem [ conv ] , i.e. , we prove that every two points in @xmath12 belong to a corridor ( and , therefore , to a convex corridor ) . in section
[ generalized ] we introduce the notion of _ extended corridors _ , which are noncompact analogues of corridors and finish the proof of theorem [ conv ] . in section [ category ]
we define _ enlargement _ of a space @xmath12 modeled on @xmath19 , which is the operation of attaching to @xmath12 an apartment along an extended convex corridor in @xmath12 .
we then show that the image of a convex corridor under enlargement is again a corridor ( possibly nonconvex ) .
in section [ mainproof ] we put it all together and prove our main result , theorem [ main ] , by constructing an increasing sequence @xmath36 of spaces modeled on @xmath19 , whose union is an affine building . in section [ conclusion ]
we make some speculative remarks on possible generalizations of our results , in particular , constructing rank 2 buildings with highly transitive automorphism groups .
our collaboration on this project started at the aim workshop `` buildings and combinatorial representation theory '' in 2007 and we are grateful to aim for this opportunity .
the first author was supported by the nsf grant dms-08 - 00247 .
the second author was supported by the nsf grant dms-05 - 54349 .
he is grateful to linus kramer and petra hitzelberger for the reference to their work .
let @xmath34 , the _ apartment _ , be either the euclidean space @xmath42 or the unit @xmath43-sphere @xmath44 ( we will be primarily interested in the case of euclidean plane and the circle ) . if @xmath45 , a _ coxeter group _ acting on @xmath34 is a finite group @xmath7 generated by isometric reflections . if @xmath46 , a _ coxeter group _ acting on @xmath34 is a group @xmath11 generated by isometric reflections in hyperplanes in @xmath34 , so that the linear part of @xmath11 is a coxeter group acting on @xmath47 .
thus , @xmath8 , where @xmath5 is a certain ( countable or uncountable ) group of translations in @xmath42 .
a spherical or euclidean _ coxeter complex _ is a pair @xmath48 , of the form @xmath49 or @xmath50 .
a _ wall _ in the coxeter complex @xmath48 is the fixed - point set of a reflection in @xmath51 .
a _ regular point _ in a coxeter complex is a point which does not belong to any wall .
singular point _ is a point which is not regular .
half - apartment _ in @xmath34 is a closed half - space bounded by a wall .
cell _ in is a convex polyhedron which is the intersection of finitely - many half - apartments .
note that in the spherical case , there is a natural cell complex in @xmath52 associated with @xmath7 , where the _ cells _ are those cells ( in the above sense ) which can not be separated by a wall .
however , the affine case , when @xmath11 is nondiscrete , there will be no cell natural complex attached to @xmath11 . _
chambers _ in @xmath49 are the fundamental domains for the action @xmath53 , i.e. , the closures of the connected components of the complement to the union of walls .
a _ dilation _ of an affine coxeter complex is a map of the form @xmath54 , where @xmath55 is a vector and @xmath56 .
thus , we regard translations as a ( limiting form of ) dilations . a _ vertex _ in @xmath48 is a ( component of , in the spherical case ) the 0-dimensional intersection of walls .
we will consider almost exclusively only those coxeter complexes which have at least one vertex ; such complexes are called _
essential_. equivalently , these are spherical complexes where the group @xmath51 does not have a global fixed point and those euclidean coxeter complexes where @xmath7 does not have a fixed point in @xmath47 .
a _ special point _ in @xmath50 is a point whose stabilizer in @xmath11 is isomorphic to @xmath7 .
we will always choose coordinates in @xmath46 so that the origin is special .
a coxeter complex where every vertex is special is called a _ special coxeter complex_. in general , the stabilizer of a point @xmath57 is naturally isomorphic a coxeter subgroup of @xmath7 .
consider now the case of @xmath58 , when @xmath3 , the dihedral group of order @xmath59 . under the identification of the vertices of the regular @xmath59-gon with the @xmath59-th roots of unity
, the group @xmath5 always contains ( or equals to ) the ring @xmath60 $ ] , where @xmath61 is a primitive @xmath59-th root of unity if @xmath10 is odd and a primitive @xmath10-th root of unity if @xmath10 is even .
it is well known that @xmath62 is discrete in @xmath63 if and only if @xmath64 . if @xmath10 is prime , then every proper coxeter subgroup of @xmath7 is isomorphic to @xmath65 and , hence , every vertex of @xmath66 is special .
if @xmath10 is not prime and @xmath5 is the smallest nontrivial group of translations normalized by @xmath7 , the complex @xmath19 is non - special
. nevertheless , for every ( countable ) @xmath67 , there exists a ( countable ) group @xmath68 containing @xmath5 and normalized by @xmath7 , so that the coxeter complex @xmath69 is special . set @xmath70 and @xmath71 and @xmath72 .
we continue inductively : given a coxeter complex @xmath73 with ( countable ) coxeter group , we define @xmath74 as the group of translations generated by all vertices of @xmath75 ( since we have chosen the origin in @xmath34 , this makes sense ) . then , every vertex of @xmath75 is special in @xmath76 , @xmath77 . by taking @xmath78
we obtain the required special coxeter complex @xmath79 . given a point @xmath80 , we then let @xmath81 denote the stabilizer of @xmath82 in @xmath11 . then
the _ type _ , @xmath83 , of the point @xmath82 is the @xmath7-conjugacy class of the _ linear part _ of @xmath81 ( i.e. , the image under the canonical projection @xmath84 ) . in particular
, for special coxeter complexes , all vertices have the same type . note that this concept of type is weaker is transitive on @xmath34 . ] than the standard ( in the discrete case ) notion of type given by the coset @xmath85 of @xmath82 in @xmath86 ( equivalently , the intersection of the orbit @xmath85 with the fundamental alcove ) .
however , in the case of nondiscrete coxeter groups , the latter concept does not seem to be of much use since @xmath86 can not be identified with an alcove or any reasonable ( commutative ) geometric object . in the spherical case ,
the notion of _ type _ is given by the projection @xmath87 where the quotient is the spherical weyl chamber , a fundamental domain for @xmath53 .
we now assume that @xmath58 , @xmath88 and @xmath3 , the dihedral group of order @xmath59 .
then there will be at most @xmath89 types of points in @xmath19 , where @xmath90 is the number of divisors of @xmath10 .
we choose the coordinates in @xmath34 so that the positive weyl chamber @xmath91 of @xmath7 is bisected by the @xmath82-axis .
a _ sub - wall _ in @xmath34 is a codimension 1 cell in @xmath34 , i.e. , a nondegenerate geodesic segment contained in a wall and bounded by two vertices .
for a subset @xmath92 in a metric space @xmath12 and @xmath93 , we let @xmath94 denote the _ open @xmath95-neighborhood _ of @xmath92 in @xmath12 , i.e. , @xmath96 for instance , if @xmath97 is a single point , then @xmath98 is the open @xmath95-ball centered at @xmath99 .
a metric space @xmath12 is called _ geodesic _ if every two points in @xmath12 are connected by a ( globally distance - minimizing ) geodesic .
most metric spaces considered in this paper will be geodesic ; the only occasional exceptions will be links of vertices in certain metric cell complexes . for a pair of points @xmath100 in a metric space @xmath12 we let @xmath101 denote a closed geodesic segment ( if it exists ) in @xmath12 connecting @xmath82 and @xmath99 . as , most of the time
, we will deal with spaces where every pair of points is connected by the unique geodesic , this is a reasonable notation .
we let @xmath102 denote the _ open geodesic segment _ , the _ interior of _ @xmath101 : @xmath103 geodesics will be always parameterized by their arc - length .
* @xmath33-isometries .
* let @xmath104 be metric spaces and @xmath105 be a real number .
we say that a map @xmath106 is a _
@xmath33-isometry _ if it satisfies the following : \1 .
@xmath107 .
\2 . @xmath108 .
the motivation for this definition comes from considering maps where the diameter @xmath33 of the target is less than the diameter of the domain , but we still would like to have maps @xmath109 which are as close to being isometries as possible .
we will use this concept , however , even if @xmath110 .
the following properties of @xmath33-isometries are immediate : * each @xmath33-isometry is injective . * for each @xmath33-isometry @xmath111 , its inverse @xmath112 is also a @xmath33-isometry . *
composition of @xmath33-isometries is again a @xmath33-isometry .
* cat(0 ) and cat(1 ) spaces .
* we refer the reader for instance to @xcite for the detailed treatment of this material .
let @xmath12 be a geodesic metric space .
the space @xmath12 is called cat(0 ) , if the geodesic triangles in @xmath12 are _ thinner _ than triangles in the euclidean plane @xmath113 .
one can state this in terms of quadruples of points in @xmath12 .
let @xmath114 ( vertices of a triangle @xmath115 ) .
let @xmath116 be such that @xmath117 ( i.e. , @xmath10 will be a point on the side @xmath101 of the triangle ) .
find points @xmath118 in @xmath113 so that @xmath119 these are vertices of a _ comparison triangle _ @xmath120 in @xmath113 .
next , find a point @xmath121 so that @xmath122 i.e. , @xmath123 belongs to the side of @xmath124 which corresponds to the side @xmath101 of @xmath115
. then we require the cat(0 ) inequality : @xmath125 similarly , one defines cat(1 ) spaces @xmath92 , except : \1 .
triangles of the perimeter @xmath126 in @xmath92 are compared to triangles on the unit 2-sphere .
\2 . @xmath92
need not be geodesic , however , if @xmath127 can not be connected by a geodesic , we require the distance between these points be @xmath128
. we will think of the distances in cat(1 ) spaces as _
angles _ and , in many cases , denote these distances @xmath129 .
the following characterization of 1-dimensional cat(1 ) spaces will be important : a 1-dimensional metric space ( a metric graph ) is a cat(1 ) space if and only if the length of the shortest embedded circle in @xmath12 is @xmath130 .
if @xmath131 is a metric graph , where each edge is given the length @xmath132 , then the cat(1 ) condition is equivalent to the assumption that girth of @xmath131 is @xmath133 . in a cat(0 ) space , any pair of points is connected by a unique geodesic , while in cat(1 ) spaces this is true for points within distance @xmath134 .
a subset @xmath135 in a cat(0 ) space @xmath12 is called _ convex _ if for every two points @xmath136 the geodesic @xmath137 is contained in @xmath135 .
a path - connected subset @xmath135 in a cat(0 ) space @xmath12 is convex if and only if it is locally convex , i.e. , every point @xmath138 possesses a neighborhood in @xmath135 which is convex in @xmath12 .
let @xmath12 be a cat(0 ) space .
the _ space of directions _ @xmath139 at a point @xmath140 is defined as the set of germs of geodesics emanating from @xmath82 .
the distances in @xmath139 are given by the _ angles _ between the geodesics .
if @xmath12 is cat(0 ) , then @xmath139 is cat(1 ) for every @xmath140 .
conversely , if @xmath12 is simply - connected , and @xmath139 is cat(1 ) for every @xmath140 , then @xmath12 is cat(0 ) .
suppose that @xmath12 is a 2-dimensional euclidean metric cell complex , i.e. , @xmath12 is obtained by gluing isometrically convex polygons in @xmath113 .
the space of directions @xmath139 is just the link @xmath131 of @xmath82 in @xmath12 . to describe the metric on @xmath139 note that a small neighborhood of @xmath82 in @xmath12 is covered by flat euclidean sectors @xmath141 with the tip at @xmath82 .
the edges of @xmath131 correspond to maximal flat sectors @xmath141 ( not containing any edges or vertices in its interior ) .
the angles of these maximal sectors at @xmath82 are the lengths of the corresponding edges of @xmath131 .
we refer to @xcite for the detailed discussion . * spaces modeled on coxeter complexes .
* [ modeled ] a space _ modeled on the coxeter complex @xmath143 _ is a metric space @xmath12 together with an atlas where charts are isometric embeddings @xmath144 and the transition maps are restrictions of the elements of @xmath11 .
the maps @xmath144 and their images are called _ apartments _ in @xmath12 .
note that ( unlike in the definition of an atlas in a manifold ) we do not require the apartments to be open in @xmath12 .
the same definition applies to spaces modeled on the spherical coxeter complex @xmath49 , only now the charts are maps of @xmath47 .
this definition generalizes to the case where we replace @xmath11 with an arbitrary group @xmath51 of affine transformations of @xmath42 and assume that @xmath145 is a @xmath51invariant affine subspace . in this generality , one has to remove , of course , the assumption that @xmath12 is a metric space and the charts are isometric embeddings .
instead , one has to assume that for every two charts @xmath146 , @xmath147 is a closed convex subset of @xmath34 .
interesting examples of this setup are provided by _ tropical geometry _
, where @xmath148 .
therefore , all @xmath11-invariant ( resp .
@xmath7-invariant ) notions defined in @xmath34 , extend to @xmath12 .
in particular , we will talk about vertices , walls , sub - walls , cells , etc .
note that the intersection of finitely many cells in @xmath12 is again a cell .
we also will have a well - defined notion of _ type _ of a point @xmath140 : it is given by computing the type of @xmath82 in an apartment @xmath149 containing @xmath82 .
we retain the notation @xmath150 in the affine , resp .
spherical , case .
if @xmath3 and @xmath92 is modeled on @xmath151 , then @xmath92 has the natural structure of a bipartite metric graph , where every edge has the length @xmath132 .
accordingly , if @xmath12 is a cat(0 ) space modeled on @xmath152 , then for every @xmath140 , @xmath153 is a cat(1 ) metric space modeled on @xmath154 , where @xmath155 , a subgroup of @xmath7 .
a space @xmath12 modeled on a coxeter complex is called _ locally finite _ if every point of @xmath12 has a neighborhood covered by finitely many apartments .
a space @xmath12 modeled on @xmath19 is said to be _ finite _ if it is covered by finitely many apartments .
( this , of course , does not mean finiteness of @xmath12 as a set or finiteness of its diameter . )
[ complex ] we let @xmath156 denote the collection of all finite spaces @xmath12 modeled on @xmath19 . when @xmath19 is fixed , we will use the notation @xmath157 for this collection .
similarly , we define @xmath158 the collection of all locally finite spaces @xmath12 modeled on @xmath19 . a _ finite subcomplex _
@xmath92 in @xmath159 is a finite union of cells in @xmath12 .
note that such @xmath92 need not be modeled on @xmath142 .
in what follows , all spaces modeled on an affine coxeter complexes are assumed to be cat(0 ) and all metric spaces modeled on spherical coxeter complexes are assumed to be cat(1 ) . for a space @xmath92 modeled on a spherical coxeter complex @xmath160 , two elements @xmath161 are called _ antipodal _ if @xmath162 .
elements of @xmath92 are _ super - antipodal _ if @xmath163 .
accordingly , chambers @xmath164 are antipodal ( rep .
super - antipodal ) if they contain antipodal ( resp .
super - antipodal ) regular elements . for instance , if @xmath92 consists of a single apartment , then chambers @xmath165 are antipodal if and only if the element @xmath166 sending @xmath167 to @xmath168 is the longest element of @xmath7 , i.e. , @xmath169 let @xmath12 be modeled on a 2-dimensional complex @xmath19 .
a _ thick sub - wall _ in @xmath12 is a sub - wall so that for every @xmath170 , @xmath139 is not a single apartment .
in other words , @xmath82 has no neighborhood which is locally isometric to @xmath34 .
of course , not every sub - wall is thick .
similarly , we say that @xmath140 is a _
thick vertex _ of @xmath12 if either @xmath82 locally separates @xmath12 or @xmath82 belongs to at least three thick sub - walls @xmath171 , so that @xmath172 equivalently , the graph @xmath139 is either disconnected or contains at least 3 vertices of valence @xmath0 . [ examples ] 1 .
take two apartments @xmath173 and glue them at a vertex @xmath55 .
then @xmath82 is a thick vertex of the resulting space .
take apartments @xmath174 and glue them along walls @xmath175 , @xmath176 as follows : @xmath177 now , take an isomorphism @xmath178 , so that @xmath179 crosses ( transversally ) @xmath176 .
next , glue @xmath180 via @xmath181 to form a space @xmath12 .
then the vertex @xmath182 corresponding to @xmath183 is a thick vertex in @xmath12 .
a _ euclidean ( or , affine ) building _ modeled on @xmath19 is a cat(0 ) space @xmath12 modeled on @xmath19 which satisfies the following two extra conditions : axiom 1 .
( `` connectedness '' ) every two points @xmath25 are contained in a common apartment .
( `` angle rigidity '' ) for every @xmath140 , the space of directions @xmath153 satisfies the following : @xmath184 here @xmath185 is the _
type projection_. for instance , one can see that every ( non - flat ) nonpositively curved symmetric space satisfies axiom 1 but not axiom 2 .
a _ spherical building _ is defined similarly , except that it is required to be modeled on a spherical coxeter complex , be cat(1 ) and the axiom 2 is not needed in this case ( it follows from _ discreteness _ of the coxeter complex ) .
it was proven in @xcite ( and in an unpublished paper by b. kleiner ) that a piecewise - euclidean cat(0 ) cell complex is a euclidean building if and only if all its links are spherical buildings .
a building @xmath12 is called _ thick _ if every wall in @xmath12 is the intersection of ( at least ) three half - apartments .
below is an alternative set of axioms for euclidean buildings due to ann parreau @xcite .
replace axioms 1 and 2 with : let @xmath12 be a cat(0 ) space modeled on @xmath19 so that : axiom 1@xmath186 .
let @xmath25 and @xmath187 be spherical chambers , where we endow the spherical buildings @xmath188 with the @xmath189 structure .
then there exists an apartment @xmath149 so that @xmath190 , @xmath191 .
axiom 2@xmath186 . for every @xmath140 the following holds .
let @xmath192 be antipodal _ weyl sectors _ with the tip @xmath82 ( i.e. , @xmath193 are antipodal in @xmath139 ) .
then there exists an apartment @xmath149 containing @xmath194 . here
a _ weyl sector _ in @xmath12 is a subset @xmath195 of an apartment @xmath149 , so that @xmath195 is a parallel translate of a weyl chamber in @xmath34 .
note that if the tip @xmath82 of @xmath195 is a special vertex , then @xmath195 is a weyl chamber in @xmath12 .
parreau s numbering of axioms is different from ours .
she also does not require the cat(0 ) condition , but has other axioms requiring intersections of pairs of apartments to be closed and convex .
these conditions follow from the cat(0 ) property .
we will be using axioms 1@xmath186 and 2@xmath186 since verifying angle rigidity would require proving a slightly weaker form of axiom 1@xmath186 anyhow .
we refer the reader to tits original paper @xcite and to appendix 3 of ronan s book @xcite for alternative axiomatization of nondiscrete euclidean buildings .
a _ morphism _ between two spaces @xmath104 modeled on @xmath196 , where @xmath197 or @xmath198 , is defined as a map @xmath106 which , being written in `` local coordinates given by apartments '' is the restriction of elements of @xmath51 .
more generally , given two spaces @xmath104 modeled on @xmath196 , and subsets @xmath199 , a _
@xmath200 is a map @xmath201 , which , in local coordinates , appears as the restriction of an element of @xmath51 . in other words , for every point @xmath202 , and for every pair of apartments @xmath203 ( with images containing @xmath204 respectively ) the ( partially defined ) map @xmath205 is the restriction of an element of @xmath51 .
it follows that , if a geodesic @xmath206 is contained in @xmath92 , then its image under a morphism @xmath200 is a piecewise - geodesic path .
therefore , every morphism @xmath200 induces a morphism ( again denoted @xmath111 to simplify the notation ) @xmath207 let @xmath135 be a subcomplex in a metric space @xmath208 modeled on a spherical coxeter complex @xmath49 .
we always endow @xmath135 with the restriction of the path - metric on @xmath135 ( not the induced path - metric ! ) .
let @xmath209 denote the vertex set of the complex @xmath49 .
set @xmath210 suppose now that @xmath58 and @xmath3 .
we set @xmath211 .
then , clearly , @xmath212 more generally , for arbitrary finite coxeter groups @xmath7 we have : @xmath213 with strict inequality if and only if the coxeter number of @xmath7 is odd .
this is the case when @xmath7 is a product of symmetric groups @xmath214 with odd @xmath10 and dihedral groups @xmath215 with odd @xmath10 .
for instance , if @xmath216 with odd @xmath10 , then @xmath217 .
we then say that a _ weak isometry _ between subcomplexes @xmath218 is a surjective morphism @xmath219 which is a @xmath33-isometry .
it is easy to see that in the case @xmath3 and @xmath58 , the latter condition is equivalent to being a @xmath220-isometry .
we now define weak isometries between subsets of spaces modeled on affine coxeter complexes .
let @xmath104 be modeled on the affine coxeter complex @xmath19 .
we first consider the easier case when @xmath221 are subcomplexes .
then a _ weak isometry _
@xmath200 is a morphism @xmath200 which sends cells isometrically to cells and induces a weak isometry @xmath222 for every @xmath223 .
this is the concept akin to the notion of an isometric embedding of riemannian manifolds ( rather than isometry between metric spaces ) .
now , suppose that @xmath199 are _ locally conical subsets _ , i.e. , each point of @xmath224 admits a neighborhood in @xmath224 which is a subcomplex in @xmath104 .
therefore , for every @xmath225 , @xmath226 are subcomplexes .
then a _ weak isometry _
@xmath200 is a surjective morphism @xmath227 inducing a weak isometry @xmath228 for every @xmath229 .
we will use the name _ weakly isometric embedding _ for the map @xmath230 , when we do not want to specify the image of @xmath111 .
( conical sets , cf .
let @xmath231 be a subset and @xmath232 .
then we say that @xmath135 is _ conical _ with respect to @xmath233 if : \1 . for every @xmath234 , @xmath235 ( i.e.
, @xmath135 is star - like with respect to @xmath233 ) .
geodesics starting at @xmath233 do not branch in @xmath135 : if @xmath236 , @xmath237 are geodesics of the same length in @xmath31 , and the germs of @xmath238 agree near @xmath233 , then @xmath239 .
let @xmath92 be a metric graph .
we define the _ cone over @xmath92 _
@xmath240 as follows .
for each edge @xmath241 of @xmath92 of length @xmath242 , let @xmath243 be the ( infinite ) sector in @xmath244 with the tip @xmath233 ( at the origin ) and the angle @xmath242 .
if @xmath55 is a vertex of @xmath241 , the cone @xmath245 the ray bounding @xmath243 .
we now glue all the sectors @xmath243 via isometries @xmath246 whenever the edges @xmath247 share the vertex @xmath55 .
the result is @xmath240 .
there is a more general definition of a metric cone over a cat(1 ) space @xmath92 , obtained by taking an appropriate metric on @xmath248 , see @xcite , but we will not need it . now , if @xmath135 is a conical subset in @xmath12 ( modeled on @xmath66 ) , then at every point @xmath232 , the set @xmath135 is isometric to a star - like subset of the euclidean cone @xmath249 since @xmath12 is locally finite , each point @xmath140 has a convex neighborhood which is conical with respect to @xmath82 . in particular , every subcomplex in @xmath12 is locally conical .
the following lemma describing the geometry of @xmath12 near sub - walls generalizes the above observation .
we first need let @xmath250 be a building which contains at most one thick wall @xmath251 , i.e. , @xmath250 is obtained by attaching several half - apartments along @xmath251 .
then , for each closed subsegment @xmath252 and @xmath253 we consider @xmath254 .
we will refer to @xmath255 as an _
open book with the binding @xmath256_. let @xmath12 be locally finite , modeled on @xmath19 .
let @xmath206 be a sub - wall which does not cross ( transversally ) any thick sub - walls .
( we allow @xmath256 to be a sub - wall . )
then there exists @xmath253 so that @xmath255 is conical with respect to every @xmath257 .
moreover , @xmath255 is isomorphic to an _ open book with the binding @xmath256_. \1 .
since @xmath12 is cat(0 ) , the neighborhood @xmath258 is convex . in particular
, it is star - like with respect to every point @xmath257 .
if two geodesics in @xmath31 `` branch '' , then branching occurs at a point of @xmath256 , since elsewhere , @xmath31 is locally isometric to @xmath6 .
if a geodesic @xmath259 intersects @xmath256 at a point different from @xmath233 , then it is entirely contained in @xmath256 .
the non - branching of geodesics follows .
pick a point @xmath257 which is not an end - point .
we identify @xmath250 with @xmath260 .
we have the map @xmath261 which sends @xmath262 to the pair @xmath263 , where @xmath264 is the direction of @xmath265 in @xmath266 and @xmath267 .
we leave it to the reader to verify that @xmath268 is an isometry of @xmath31 to an open book in @xmath250 .
we observe that every @xmath12 modeled on @xmath19 has extendible geodesics : if @xmath269\to x$ ] is a geodesic , it extends to a geodesic ray @xmath270 .
this follows since every @xmath271 has an antipodal direction ( coming from an apartment in @xmath139 containing @xmath264 ) . for a subcomplex @xmath272 ( or ,
more generally , a locally conical subset ) , we define the `` unit tangent bundle '' @xmath273 as the space of pair @xmath274 .
then @xmath275 .
we topologize @xmath276 as follows .
let @xmath277 denote the space of geodesic rays in @xmath12 with the topology of uniform convergence on compacts .
then @xmath276 is the quotient of @xmath277 , where we identify rays emanating from @xmath82 having the same germs at @xmath82 .
we give @xmath276 the quotient topology and @xmath273 the subspace topology .
the following are elementary properties of @xmath276 : * @xmath276 is hausdorff . * the projection @xmath278 is continuous . *
if @xmath272 is closed then @xmath275 is closed .
( note that in general @xmath279 . ) * if @xmath12 is locally finite then the map @xmath280 is proper ( arcela - ascoli theorem ) . in particular
, @xmath276 is locally compact in this case .
let @xmath281 ( which is modeled on 2-dimensional @xmath19 ) be a connected subset which is homeomorphic to an open 2-dimensional manifold .
we also assume that each point @xmath234 has a neighborhood which is contained in a single apartment .
we thus obtain an _ open _
atlas @xmath282 on @xmath208 , whose charts @xmath283 are isomorphisms . thus , the transition maps for this atlas are restrictions of the elements of @xmath11 .
this gives us a geometric structure on @xmath208 modeled on @xmath19 and , therefore , we obtain a multi - valued developing map @xmath284 which is , locally , a morphism ( see e.g. @xcite ) .
suppose that @xmath208 is convex in @xmath12 .
then @xmath111 is single - valued and injective . since @xmath208 is convex , it is simply - connected .
hence , @xmath111 is single - valued . in order to check injectivity , note that
, being a local isomorphism , @xmath111 is a local isometry .
let @xmath285 be distinct points . since @xmath208 is convex , the ( unique ) geodesic @xmath286 is contained in @xmath208 .
since @xmath111 is a local isomorphism , it sends local geodesics to local geodesics . since
a local geodesic in @xmath34 is a global geodesic , @xmath287 is a straight - line segment with the end - points @xmath288 and the length equal to that one of @xmath256
. therefore , @xmath289 .
a subset @xmath272 is called _ planar _ if there exists a weakly isometric embedding @xmath290 .
in this section we will assume that the coxeter complex @xmath142 is _ special _ , the group of translations of @xmath11 is dense in @xmath244 , and spaces @xmath12 modeled on this complex are locally finite .
below we define _ tiles _ and _ corridors _ , which are the main tools for proving a combinatorial convexity theorem [ convex ] , which , in turn , is the key for constructing the building @xmath12 as in theorem [ main ] .
a _ model tile _ in a coxeter complex @xmath9 is a parallelogram @xmath33 whose edges are sub - walls parallel to the walls bounding @xmath91 . in particular , the angles of @xmath33 are of the form @xmath291
. we will regard a single sub - wall , with the slope @xmath292 , as a ( degenerate ) tile .
we let @xmath293 denote the left and right vertices of @xmath33 .
then @xmath33 is uniquely determined by @xmath294 .
we now define _ model corridors _ in @xmath34 .
we will frequently omit the adjective `` model '' to simplify the terminology .
every model corridor @xmath31 is determined a pair of piecewise - linear functions @xmath295\subset \r \to \r$ ] satisfying the following : \1 .
the points @xmath297 are ( distinct ) vertices of @xmath34 : they will be called the _ extreme vertices _ ( or _ extremes _ ) of the corridor @xmath31 .
@xmath298 for all @xmath299 $ ] .
the slopes of the graphs of @xmath300 are of the form @xmath301 .
the function @xmath302 is called _ upper roof function _ and @xmath303 is called _ lower roof function .
_ the slope of @xmath302 at @xmath304 is @xmath305 and at @xmath306 it is @xmath307 , for @xmath303 it is the other way around . in section
[ generalized ] , we will generalize the above concepts to allow tiles of infinite diameter and roof functions defined on infinite intervals .
( model ) corridor @xmath31 _ is the closed region between the graphs of the functions @xmath308 . a corridor @xmath31 is called _ simple _ if either @xmath309 on the open interval @xmath310 or @xmath311 on @xmath312 $ ] . in other words ,
simple corridors are the ones which are homeomorphic to a closed disk or isometric to a segment .
a corridor is _ degenerate _ if it is a sub - wall , i.e. , is a single segment .
thus , every degenerate corridor is simple .
figure [ f11 ] provides examples of corridors .
we note that each corridor is a union of finitely many simple corridors , every two of which intersect in at most one point , an extreme vertex of both corridors .
these intersection points will be called _ cut - vertices _ of @xmath31 , as they separate the corridor @xmath31 . away from these points ,
a corridor @xmath31 is a manifold with boundary ( of dimension 1 or 2 ) .
this decomposition of @xmath31 into maximal simple sub - corridors @xmath313 is unique ; the sub - corridors @xmath313 are called the _ components _ of @xmath31 .
we let @xmath314 denote the graphs of @xmath308
. then @xmath315 .
we note that each corridor with the extremes @xmath316 is contained in the tile with the vertices @xmath316 .
it follows immediately from the definition that each edge of the polygon @xmath31 is a sub - wall and each vertex is a vertex of @xmath9 . despite the fact that a corridor is not convex ( unless it is a single tile ) , it satisfies the following weak convexity property : let @xmath251 be a line in @xmath34 with the slope @xmath292 . then @xmath317 is either empty or is a segment .
although we will not need this , one can characterize simple nondegenerate corridors in @xmath34 geometrically ( without using the roof functions ) as follows : 1 .
@xmath31 is a subcomplex in @xmath19 .
@xmath31 has only finitely many edges and is homeomorphic to the closed 2-disk .
3 . all edges of @xmath31 have slopes @xmath292 .
@xmath31 does not have any two consecutive ( interior ) angles each @xmath318 .
the following lemma is elementary and is left to the reader . [ approx ] suppose we are given @xmath253 and a continuous function @xmath319\to \r$ ] , so that @xmath320 are vertices of the coxeter complex @xmath19
. then there are upper and lower roof functions @xmath321\to \r$ ] so that : \1 . @xmath322 .
@xmath323 .
[ limit ] let @xmath324 ( @xmath325 ) be an increasing family of corridors with common extreme points @xmath316 : @xmath326 . assume also that the numbers of edges in the corridors @xmath324 are uniformly bounded from above .
then @xmath327 is a corridor in @xmath328 with the extreme points @xmath329 .
let @xmath330 be the upper / lower roof function for the corridor @xmath324 .
since the corridors @xmath324 are increasing , we obtain : @xmath331 moreover , all the roof functions in question are uniformly bounded ( from above and below ) by the roof functions of the tile with the vertices @xmath316 .
it is then immediate that there exist limits @xmath332 and these limits are roof functions with respect to @xmath333 .
it is clear from the definition that every point in a corridor @xmath31 belongs to the interior ( relative to @xmath31 ) of a tile @xmath334 . by compactness
, we can therefore cover the corridor @xmath31 by finitely many tiles .
a _ tiled corridor _ in @xmath34 is a finite set of tiles @xmath335 whose union @xmath336 is a corridor @xmath31 in @xmath34 , so that @xmath337 is minimal with this property , i.e. , any proper subset @xmath338 will have strictly smaller union .
when convenient , we will abuse the notation and conflate @xmath337 and @xmath339 .
we set @xmath340 .
we now extend our discussion to corridors in @xmath12 , a space modeled on @xmath9 .
a _ tile _ in @xmath12 is a subcomplex isomorphic to a ( model ) tile in @xmath34 .
we now define corridors in @xmath12 .
they will satisfy two axioms c1 and c2 , the first of which is easy to state , the second axiom will take more work . every corridor in @xmath12 is a subcomplex @xmath28 weakly isometric to a model corridor in @xmath34
. in particular , if @xmath55 is a cut - vertex of @xmath31 , it disconnects @xmath31 into two sub - corridors @xmath341 , so that @xmath342 is a chamber or a vertex in @xmath343 , and the distance between these is at least @xmath344 . using the embedding @xmath290
, we will be identifying corridors @xmath31 in @xmath12 with corridors in @xmath34 .
suppose @xmath28 is a corridor with extreme vertices @xmath316 .
since @xmath345 is contained in a tile with the extreme vertices @xmath346 , we obtain [ diam ] @xmath347 , where @xmath348 .
we now introduce the second corridor axiom . to motivate the definition , consider a model corridor @xmath349 as in figure [ f8 ] , where @xmath350 is a degenerate tile .
let @xmath351 be the reflection in the wall containing @xmath350 .
then the union @xmath352 is not a model corridor in @xmath34 , but satisfies axiom c1 : the isomorphism @xmath353 is given by the map which fixes @xmath354 pointwise and sends @xmath355 to @xmath356 via @xmath357 .
for a variety of reasons , we would like to exclude @xmath358 ( and other similar `` twisted corridors '' ) from being corridors in spaces @xmath12 modeled on @xmath19 . the most obvious reason (
although not the critical one ) is that we would like every corridor in @xmath34 to be the image of a model corridor under some @xmath359 .
the key reason for excluding such @xmath358 will become clear in the proof of lemma [ wmax ] . one can tell apart a `` twisted corridor '' @xmath360 from a model corridor by observing that there exists a _ trapezoid _
@xmath361 _ connecting _ the tiles @xmath362 as in figure [ f8 ] , while a model corridor can not admit such trapezoids since its existence would mean that boundary slopes of a model corridor would take at least 3 different values .
equivalently , there is a nontrivial continuous family of segments parallel to @xmath350 with end - points contained in @xmath358 .
( these segments foliate the trapezoid @xmath361 . )
a _ trapezoid _ in @xmath34 is a nondegenerate quadrilateral bounded by sub - walls and having the angles @xmath363 .
we will say that the set of polygons @xmath364 is a _ bridge _ in @xmath34 . by abusing the notation
, we will refer to the union of polygons in a bridge as a bridge as well .
we define a _ bridge _ in a space @xmath12 ( modeled on @xmath19 ) as a set of polygons whose union is isometric to a bridge in @xmath34 .
we are now ready to state the second corridor axiom : * axiom c2*. a corridor in @xmath31 can not admit a `` bridge '' , i.e. , @xmath31 can not contain a degenerate sub - corridor @xmath350 which , together with tiles @xmath365 , and a trapezoid @xmath366 form a bridge .
equivalently , if @xmath350 is a degenerate irreducible component of @xmath31 , there is no nontrivial family of segments in @xmath12 parallel to @xmath350 whose end - points are contained in @xmath31 .
given a corridor @xmath28 and its weak isometry @xmath367 , we declare the _ boundary _ @xmath368 of @xmath31 , to be the preimage @xmath369 .
it is clear that the boundary is independent of the embedding @xmath111 .
we then similarly extend to corridors in @xmath12 all the concepts defined for model corridors .
[ isomembed ] suppose that @xmath28 is a corridor and a morphism @xmath370 is an isometric embedding .
then @xmath371 is again a corridor in @xmath372 .
let @xmath373 be a weakly isometric embedding .
then @xmath374 is again a weakly isometric embedding .
this verifies axiom c1 .
suppose that @xmath371 admits a bridge @xmath375
. then @xmath361 is foliated by parallel geodesic segments whose endpoints belong to @xmath376 .
since @xmath377 is convex in @xmath372 , it follows that each of these geodesics is contained in @xmath377 . in particular , @xmath28 admits a bridge .
contradiction .
[ nonconvex ] it important to note that , unless @xmath12 is a building , if @xmath378 is a convex corridor , its image @xmath379 is not necessarily convex .
this follows from the fact that a weakly isometric embedding is not in general an isometric embedding .
the reader can construct examples of convex corridors with nonconvex images @xmath345 using the example of a space @xmath12 described in example [ examples ] , part 2 .
in this section , @xmath159 ( as in notation [ complex ] ) is a space modeled on @xmath19 .
let @xmath380 and @xmath381 be tiled corridors in @xmath12 with @xmath382 .
we say that @xmath383 if : \1 . @xmath31 and
@xmath358 have the same extreme vertices .
\2 . for every @xmath384 there exists
@xmath385 so that @xmath386 .
@xmath387 , where @xmath388 are weak isometries to model corridors in @xmath34 .
we say that @xmath389 if @xmath383 and the cardinality of @xmath390 is at most the cardinality of @xmath391 , i.e. , @xmath392 .
it is easy to check that both @xmath393 and @xmath394 are ( partial ) orders . on
can replace the order @xmath394 with the inclusion relation between the corridors @xmath31 , but the proofs will become somewhat more complicated .
a tiled corridor @xmath337 in @xmath12 is _ weakly maximal _ if it is maximal with respect to the order @xmath393 .
a tiled corridor @xmath337 in @xmath12 will be called _
maximal _ if it is maximal with respect to the order @xmath394 .
let @xmath12 be modeled on @xmath19 .
we will use the notation @xmath395 for the same metric space @xmath12 , but regarded as a space modeled on @xmath328 , where @xmath396 is the closure of @xmath11 .
then [ wmax ] for every tiled corridor @xmath397 in @xmath12 there exists a tiled corridor @xmath337 such that @xmath337 is weakly maximal in @xmath12 and @xmath398 .
let @xmath399 .
. we will first prove existence of a weakly maximal corridor in @xmath395 ; in the second part of the proof we will verify that this corridor is actually a corridor in @xmath12 . by zorn s lemma , it suffices to consider a maximal totally ordered ( with respect to the order @xmath393 ) family of tiled corridors @xmath400 and to show that this family contains a maximal element . without loss of generality
we can assume all tiled corridors have the same cardinality .
then each tiled corridor @xmath401 is of the form @xmath402 where @xmath403 .
set @xmath404 and let @xmath405 be the isomorphism onto a model corridor @xmath406 , so that @xmath407 then we obtain an increasing family of ( model ) tiled corridors @xmath408 . by lemma [ limit ] , the closure @xmath358 of the union @xmath409 is again a corridor in @xmath34 . defining a tiled corridor
@xmath410 we get @xmath411 .
similarly , set @xmath412 we may have to omit some of the tiles in @xmath337 and @xmath413 in order to obtain minimal tiled corridors @xmath414 with @xmath415 . in order to simplify the notation
we will ignore this issue .
then the maps @xmath416 yield a map @xmath417 .
we claim that @xmath31 is a corridor in @xmath395 .
we first verify that @xmath111 is a weak isometry .
it is clear that the restriction of @xmath111 to every simple sub - corridor is an isometry to its image , as the limit of isometries , is an isometry . in particular
, the restriction of @xmath111 to each simple sub - corridor is a weak isometry .
therefore , we have to consider the behavior of @xmath111 at the cut - vertices .
if @xmath55 is a cut - vertex of @xmath31 then it is a cut - point of every @xmath418 .
if a cut - vertex @xmath55 of @xmath31 was separating nondegenerate sub - corridors for some @xmath419 , the same is true for all @xmath420 .
moreover , in this case @xmath421 and @xmath422 is independent of @xmath423 for @xmath420 .
thus , @xmath111 is a weak isometry at @xmath55 in this case .
suppose now that @xmath55 belongs to a degenerate component , say , @xmath424 , of @xmath324 for all @xmath425 .
note that the point @xmath55 need not be an end - point of this geodesic segment .
this is the part of the proof where the corridor axiom c2 will be used .
we will consider only the case when @xmath426 is _ not _ of the form @xmath427 for sufficiently large @xmath425 .
the remaining case will be similar and is left to the reader . if @xmath55 is not a vertex of @xmath424 we let @xmath428 ( @xmath191 ) denote the antipodal directions tangent to @xmath424 .
otherwise , we define @xmath428 so that one of them ( say , @xmath429 ) is tangent to @xmath424 and the other is a fixed direction which belongs to @xmath430 ( for all @xmath425 ) and is super - antipodal to @xmath429 : @xmath431 by the construction of @xmath31 , the germ @xmath432 at @xmath55 consists of two components @xmath433 containing @xmath434 ( @xmath191 ) .
each @xmath435 is either a chamber or equals @xmath434 .
therefore , the distance between @xmath30 in @xmath436 is at least @xmath437 .
if the distance is @xmath438 then @xmath111 is indeed a weak isometry at @xmath55 and we are done .
otherwise , the distance equals @xmath439 and both @xmath435 are chambers @xmath440 : @xmath441 .
moreover , we necessarily have @xmath442 .
in particular , @xmath55 separates nondegenerate sub - corridors in @xmath31 .
since the distance between @xmath30 is @xmath439 , there exists a flat triangle @xmath366 with the tip @xmath55 and the angle @xmath439 at @xmath55 , so that @xmath443 combining the germ of this flat triangle with the germs ( at @xmath55 ) of @xmath444 , we obtain a bridge in @xmath12 containing @xmath445 for sufficiently large @xmath423 .
this contradicts axiom c2 for the corridors @xmath446
. see figure [ f9 ] .
the same argument excludes existence of a bridge for @xmath31 : one can regard the triangle @xmath361 in the above argument as a degenerate trapezoid .
since the family of tiled corridors @xmath447 was chosen to be maximal , it shows that @xmath448 for all sufficiently large @xmath425 and @xmath337 is a weakly maximal tiled corridor in @xmath395 .
note that even if all the corridors @xmath324 were corridors in @xmath12 , then @xmath31 might be a corridor only in @xmath395 , as the walls of @xmath19 are dense among the walls of @xmath328 .
suppose that one of the edges @xmath241 of @xmath31 is not contained in a wall of @xmath12 .
since its image under @xmath111 has the slope @xmath301 , it follows that @xmath241 can not contain any vertices of @xmath12 .
this is the only place of the paper where we _ truly need _ the fact that @xmath19 is special : all other arguments could be modified to handle the general case .
in particular , @xmath241 contains no extreme points of @xmath31 and no thick vertices of @xmath12 .
therefore , all the tiles of @xmath337 which overlap with the edge @xmath241 can be slightly expanded in the direction of this edge to form a new corridor @xmath449 which is strictly larger than @xmath31 ( it will have an edge parallel to @xmath241 ) .
thus , @xmath450 contradicting maximality of @xmath337 .
our next goal is to prove existence of maximal tiled corridors with respect to the order @xmath394 . in view of the previous proposition
, it suffices to restrict our discussion to weakly maximal tiled corridors .
let @xmath451 be model tiles ; we denote the ( closed ) edges of these tiles @xmath452 and @xmath453 respectively , so that the edges @xmath454 correspond to each other under the dilation mapping @xmath33 to @xmath455 . given a subset @xmath456 , we define @xmath457 which is _ the number of sides of the tiles @xmath458 `` sharing '' a point from @xmath459 . _
the following lemma is elementary : [ trivial ] let @xmath458 be tiles in @xmath34 and @xmath456 .
\1 . if @xmath460 , then @xmath461 .
if @xmath462 , then either @xmath463 or @xmath464 . in view of this lemma , we will say that a tile @xmath465 that has a thick vertex of @xmath12 ( or an extreme point of @xmath336 ) on every edge is _ rigid _ and a tile that has such a point on all but one edges is _ semi - rigid_. in particular , as @xmath12 is locally finite , a compact subset of @xmath12 contains only finitely many rigid tiles .
[ edge ] let @xmath33 be a tile in a weakly maximal corridor @xmath337 , @xmath466 , and @xmath241 is an edge of @xmath33 which contains neither thick vertices , nor extreme points of @xmath31
. then @xmath467 . if not
, we can expand @xmath33 slightly to the direction of @xmath241 to a new tile @xmath455 ( which shares with @xmath33 three other sides ) , see figure [ f3 ] .
then @xmath468 is still a tiled corridor in @xmath12 ( which has the same extreme points as @xmath337 ) contradicting weak maximality of @xmath337 .
= 4.5 in suppose that @xmath337 is a weakly maximal tiled corridor in @xmath12 , @xmath339 .
let @xmath459 denote the union of the set of extreme points @xmath469 of @xmath337 with the set of thick vertices contained in @xmath470 , where @xmath348 .
let @xmath423 denote the ( finite ) cardinality of @xmath459 .
set @xmath471 .
\1 . if @xmath472 are semi - rigid tiles so that @xmath473 , then @xmath461 .
suppose that @xmath472 are not semi - rigid tiles . if @xmath474 , then @xmath461 .
suppose @xmath475 .
by lemma [ trivial ] , @xmath463 or @xmath476 .
therefore , we can eliminate one of the tiles @xmath477 from the tiled corridor @xmath337 ( without changing @xmath336 ) , contradicting the definition of a tiled corridor .
suppose that , say , @xmath458 are not semi - rigid and share two elements @xmath478 of @xmath459 , both of which belong to @xmath479 . then , by lemma [ edge ] , @xmath480 are both contained in @xmath481 .
suppose that @xmath475 .
definition of a tiled corridor implies that @xmath33 can not contain @xmath455 and vice versa .
but this means that , @xmath482 has nonempty intersection with the interior of @xmath455 , see figure [ f2 ] . contradiction .
the case when one of the shared points is on @xmath479 and the other on @xmath481 is similar and is left to the reader .
= 4.5 in the following three corollaries are then immediate : each tile in @xmath337 is uniquely determined by 2 or by 3 elements of @xmath459 . [ bound ] let @xmath337 be a weakly maximal tiled corridor with the extreme points @xmath316
. then number of tiles in @xmath337 is at most @xmath483 .
[ finiteness ] given a tiled corridor @xmath337 in @xmath12 , there exists only finitely many weakly maximal tiled corridors @xmath413 in @xmath12 so that @xmath383 .
we then obtain : [ maxexists ] for every tiled corridor @xmath397 there exists a maximal corridor @xmath337 so that @xmath484 . by lemma [ wmax ] ,
the set of weakly maximal tiled corridors @xmath485 so that @xmath486 is nonempty .
this set is finite according to corollary [ finiteness ] .
therefore , it contains a maximal element .
we continue with the notation of the previous section .
our next objective is to build convex subcomplexes in @xmath12 containing the given pair of points : [ convex ] let @xmath12 be locally finite and modeled on a special coxeter complex @xmath19 , whose group of translations is countable and dense in @xmath244 . then : \1 .
for any two points @xmath25 , there exists a corridor @xmath31 in @xmath12 so that @xmath487 .
moreover , if @xmath27 is not a thick vertex and @xmath101 is not a sub - wall near @xmath27 , then we can choose @xmath31 so that @xmath488 . \2 . for every corridor @xmath31 in @xmath12
, there exists a convex corridor @xmath358 so that @xmath489 .
the proof of this theorem will occupy the rest of the section . \1 .
since @xmath12 contains only countably many walls , it contains only countably many ( convex ) subcomplexes .
the above theorem is clear in the case when @xmath12 is a building , as any two points belong to an apartment . in this case
, we can use @xmath490 , a single diamond .
\3 . if @xmath12 is not a building , the convex corridor @xmath358 is , in general , not a diamond .
accordingly , the weakly isometric embedding @xmath491 will not send @xmath358 to a convex subset of @xmath34 .
see remark [ nonconvex ] .
the theorem is probably also true for arbitrary cat(0 ) locally euclidean complexes ( not necessarily 2-dimensional ones ) provided that we allow cells in @xmath31 to be arbitrary convex euclidean polyhedra .
however , requiring @xmath31 in theorem [ convex ] to be planar , imposes serious restrictions on the allowed subcomplexes .
\5 . in the main application of the above theorem
, @xmath12 will be finite , i.e. , covered by finitely many apartments .
we prove this theorem by : \i . constructing a corridor @xmath31 containing @xmath492 and @xmath493 .
we will subdivide the geodesic @xmath494 into subsegments @xmath495 not passing through thick vertices .
we then construct for each @xmath495 a simple ( possibly degenerate ) corridor @xmath313 , so that @xmath496 and the end - points of @xmath495 are the extreme vertices of @xmath313 .
the corridor @xmath31 will be the union of the sub - corridors @xmath313 .
the most difficult part of the proof is constructing a _ planar neighborhood _ @xmath497 of @xmath495 .
the corridor @xmath313 will be contained in this neighborhood . \ii . given a tiled corridor @xmath337 with @xmath466 , we will take a maximal tiled corridor @xmath413 so that @xmath383 .
in particular , @xmath498 .
we then verify convexity of @xmath358 .
we first assume that @xmath494 is a geodesic with distinct end - points , which contains no thick vertices in its interior and is not contained in a wall .
if @xmath256 is contained in a wall @xmath499 , then @xmath500 is a planar set and we obtain an isometric embedding @xmath501 any sub - wall @xmath502 containing @xmath256 , is therefore a degenerate convex corridor in @xmath12 containing @xmath256 .
* planar neighborhoods .
* at first , we assume , in addition , that neither @xmath492 , nor @xmath493 is a thick vertex and that @xmath256 does not cross a thick wall .
then a _ planar neighborhood _ of @xmath206 is a connected closed planar subset @xmath503 of @xmath12 whose interior @xmath208 ( in @xmath12 ) contains @xmath256 .
[ openplanar ] under the above assumptions , @xmath256 has a convex planar neighborhood @xmath503 .
in particular , an embedding @xmath504 will be distance - preserving .
let @xmath505 be the minimal distance from @xmath256 to the union of thick sub - walls of @xmath12 .
our assumptions imply that @xmath253 .
we then let @xmath208 be the open @xmath505-neighborhood @xmath255 and @xmath503 be its closure in @xmath12 .
then both @xmath208 and @xmath503 are convex in @xmath12 , since @xmath12 is a cat(0 ) space . moreover , by the choice of @xmath505 , @xmath506 where each @xmath507 is planar .
therefore , by considering the developing map of @xmath508 , and using convexity of this set we obtain a morphism @xmath509 .
convexity of @xmath208 implies that @xmath111 is an isometric embedding .
therefore , it extends to an isometric embedding @xmath510 .
we will have to generalize the concept of a planar neighborhood to take into account the cases when one of the @xmath27 s is a thick vertex , or when @xmath256 crosses a thick sub - wall : in both cases a planar neighborhood of @xmath256 does not exist , so we will have to modify the definition . a _ relative _ planar neighborhood of @xmath256 is a closed connected subset @xmath511 ( the closure of some @xmath281 ) so that the following holds : \1 . @xmath512 .
if @xmath27 is not a thick vertex , then @xmath208 contains @xmath27 .
\3 . if @xmath27 is a thick vertex , then ( for sufficiently small @xmath253 ) @xmath513 is conical with respect to @xmath27 and @xmath514 is a chamber in @xmath188 .
there exists an injective isometric morphism @xmath284 onto an open subset of @xmath34 .
( note that we do not require @xmath208 itself to be open in @xmath12 . )
figure [ f7 ] gives an example of relative neighborhood .
we observe that if @xmath256 is a geodesic satisfying the assumptions of lemma [ openplanar ] , then the notions of a planar neighborhood and of a neighborhood coincide .
every geodesic @xmath515 ( which has no thick vertices in its interior ) has a convex relative planar neighborhood .
first of all , without loss of generality we may assume that for both @xmath191 , either @xmath27 is a thick vertex or it is not contained in a wall ( otherwise , we can extend @xmath256 slightly in the direction of @xmath27 ) .
let @xmath253 be such that @xmath255 intersects only those thick sub - walls which cross @xmath256 , possibly at its endpoint(s ) .
as before , @xmath255 is convex , however , it is no longer planar ( unless we are in the situation of lemma [ openplanar ] ) .
our goal is to decrease this neighborhood of @xmath256 to make it planar and to satisfy the rest of the properties of a relative neighborhood .
suppose that @xmath27 is a thick vertex .
consider the germ of @xmath256 at @xmath27 .
since @xmath256 is not contained in a wall , it follows that this germ determines a regular point @xmath516 of @xmath188 .
therefore , @xmath516 is contained in a unique chamber @xmath517 .
hence , there exists a closed conical ( with respect to @xmath27 ) subset @xmath518 so that @xmath519 thus @xmath520 is isomorphic to an appropriate subset of @xmath521 in particular , @xmath520 is planar .
we then remove from @xmath522 the subset @xmath523 if @xmath27 is not a thick vertex , we do no not modify @xmath255 near @xmath27 .
this ensures that the resulting convex subset @xmath524 satisfies parts 1 , 2 , 3 of the definition .
we also observe that @xmath525 were planar if the interior of @xmath256 did not cross any thick sub - walls .
we now deal with the thick sub - walls which @xmath256 might cross .
let @xmath526 denote the thick sub - walls of @xmath12 crossed by the interior of @xmath256 at the points @xmath527 .
then @xmath525 is locally planar except at the points of the geodesics @xmath528 : near those points , @xmath525 is not even a manifold , but looks like an `` open book '' with the binding @xmath529 .
our goal is to `` tear off '' all but two pages of this open book ( for each @xmath530 ) , i.e. , the `` pages '' which are not tangent to @xmath256 .
consider the link @xmath531 . since @xmath532 is not a thick vertex ,
this link is a building : this graph has exactly 2 thick vertices , which are antipodal points representing the tangent directions to @xmath529 at @xmath532 .
the tangent directions to @xmath256 at @xmath532 are represented by two _
regular _ antipodal points in @xmath531 .
therefore , there exists a unique apartment @xmath533 containing these regular antipodal points .
we now remove from @xmath525 the union of those open geodesic segments @xmath534 which are _ not _ tangent to @xmath357 .
( the closure of the removed set is conical with respect to @xmath532 . )
we do this for each @xmath535 .
see figure [ f7 ] .
the resulting set @xmath208 is still convex in @xmath12 .
now , however , it is a topological 2-manifold .
we construct the injective morphism @xmath536 as in the proof of lemma [ openplanar ] .
the fact that the image of @xmath111 is open follows from the invariance of domain theorem .
[ simplecorridor ] for any pair of points @xmath25 so that the open geodesic segment @xmath537 is not a sub - wall and contains no thick vertices , there exists a simple corridor @xmath28 containing @xmath23 .
if @xmath27 is a vertex then we can choose @xmath31 that has @xmath27 as one of its two extremes .
if @xmath27 is not a thick vertex , we can also choose @xmath31 so that @xmath488 .
let @xmath503 be a convex ( relative ) planar neighborhood of the geodesic @xmath494 , given by the above lemma .
we begin by considering the case when both @xmath27 are thick vertices ; thus , the germ of @xmath503 near @xmath27 is a sector @xmath520 ( with the angle @xmath132 at @xmath27 ) .
since @xmath503 is planar , it suffices to consider the case when it is contained in @xmath34 .
moreover , we can assume that it is embedded in @xmath34 so that the slopes of the boundary of @xmath538 have the form @xmath292 .
then , the same will be true for the slopes of @xmath539 .
the geodesic segment @xmath540 is the graph of a linear function @xmath541\to \r$ ] , with @xmath542 now , approximate @xmath543 ( from above and from below ) by upper and lower roof functions @xmath300 using lemma [ approx ] .
note that the graphs of the germs of @xmath544 near @xmath545 are the boundary segments of @xmath520 , @xmath237 . by taking a sufficiently fine approximation
, we obtain that the graphs of @xmath300 are contained in @xmath503 .
therefore , the region bounded by these graphs is the required simple corridor .
suppose that neither @xmath492 nor @xmath493 is a thick vertex and , hence , @xmath546 .
if the slope of @xmath540 represents a vertex in @xmath151 , we can as well assume that this slope equals @xmath292 .
in this case , using convexity of @xmath208 , it is easy to construct a single ( `` nearly degenerate '' ) tile @xmath547 containing @xmath23 .
we therefore assume that the slope of @xmath256 is in the open interval @xmath548 .
pick disjoint congruent tiles @xmath549 containing @xmath550 .
let @xmath551 denote the extreme vertices of the smallest tile @xmath552 containing @xmath553 .
then @xmath554 .
if @xmath27 is a vertex , we can , if necessary , choose @xmath555 so that @xmath556 .
we assume that @xmath557 are the projections of @xmath558 to the horizontal axis .
the convex hull @xmath559 is a hexagon , with four edges @xmath560 incident to @xmath551 , @xmath237 and the remaining two parallel edges @xmath561 , so that the boundary of @xmath562 is the union of two arcs @xmath563 these arcs are the graphs of functions @xmath564\to \r$ ] .
approximating @xmath565 by the respectively upper and lower roof functions @xmath300 we obtain a simple model corridor @xmath566 containing @xmath23 . since @xmath567 is
a simple model corridor , as a subcomplex in @xmath12 is automatically satisfies axiom c2 .
the construction in the case when one of the points @xmath23 is a vertex and the other is not , is a combination of the two arguments in case 1 and 2 , and is left to the reader .
we now drop the assumption that the open geodesic @xmath568 does not contain any thick vertices .
[ ecorridor ] for every pair of points @xmath25 , there is a corridor @xmath31 containing @xmath23 .
if @xmath27 is not a thick vertex and the germ of the geodesic @xmath494 at @xmath27 is a regular direction in @xmath188 , we can assume that @xmath27 does not belong to @xmath569 .
we subdivide the geodesic @xmath256 into maximal subsegments @xmath570 , so that the interior of each @xmath571 does not contain any thick vertices .
according to corollary [ simplecorridor ] , there exist simple ( possibly degenerate ) corridors @xmath572 so that each @xmath313 contains @xmath573 .
moreover , for @xmath574 , the vertices @xmath575 are the extremes of @xmath313 . if @xmath576 or @xmath577 is non - degenerate and the corresponding point @xmath27 in this corridor is not a thick vertex , by corollary [ simplecorridor ] , we can assume that @xmath27 does not belong to the boundary of this corridor .
let @xmath31 denote the union @xmath578 .
we claim that @xmath31 is the required corridor . for each @xmath579
we have an isomorphic ( and isometric ) embedding @xmath580 , whose image is a simple model corridor @xmath581 .
we combine these maps to obtain a weakly isometric embedding @xmath582 as follows : we choose @xmath583 arbitrarily .
we then normalize each @xmath584 ( by a translation in @xmath11 ) , so that @xmath585 and @xmath586 is the extreme rightmost point of @xmath587 and extreme leftmost point of @xmath588 .
since each @xmath587 is a model corridor , this ensures that the germs of @xmath589 at @xmath586 are antipodal in @xmath590 , provided that both are chambers . in any case , @xmath591 with equality to @xmath220 , except for the case when both @xmath589 are degenerate and their union is a geodesic segment in @xmath34 . in the latter case the angle in @xmath34 equals @xmath280 and the angle in @xmath592 is @xmath128 , since @xmath256 is geodesic .
we thus obtain a morphism @xmath290 whose restriction to each @xmath313 is @xmath584 . for each @xmath579 , the convex hull of @xmath587 is a tile @xmath593 with the extreme vertices @xmath594 . by the construction , for @xmath595
, @xmath596 is at most one point ( the common extreme vertex ) .
therefore , @xmath111 is injective .
we next need to verify the weak isometry at the vertices @xmath551 .
suppose that both @xmath597 are nondegenerate corridors .
then their germs at @xmath551 are chambers @xmath598 .
since @xmath256 is a geodesic , its angle at @xmath551 is @xmath128 .
therefore , the chambers @xmath599 are _ super - antipodal _ : the angular distance between these chambers in @xmath592 is @xmath600 . on the other hand ,
the germs of @xmath587 at @xmath601 are antipodal in @xmath602 .
therefore , the map @xmath603 is a @xmath220-isometry .
thus , @xmath290 is a weakly isometric embedding .
suppose that @xmath313 is degenerate , but @xmath604 is not .
the germ @xmath435 represents the germ of the geodesic @xmath495 at @xmath551 and @xmath605 is a chamber containing the germ of the geodesic @xmath606 at @xmath551 .
therefore , since @xmath605 contains a point antipodal to @xmath435 , the angular distance in @xmath592 between the vertex @xmath435 and the chamber @xmath605 is @xmath607 .
the angular distance between their images @xmath608 in @xmath609 is exactly @xmath220 .
therefore , @xmath111 is a weak isometry at @xmath610 .
the argument when both @xmath597 are degenerate follows from our discussion of the angles between @xmath597 and @xmath589 .
thus , @xmath31 satisfies axiom c1 of corridors .
we need to check axiom c2 .
suppose that @xmath31 admits a bridge @xmath611 where @xmath612 is degenerate .
then , by the construction of the corridor @xmath31 , one of the @xmath613 ( say , @xmath614 ) is a vertex @xmath551 and @xmath615 ( one can actually assume that @xmath616 ) .
therefore , the germ of the geodesic @xmath256 at @xmath614 is contained in @xmath617 .
since @xmath618 is a planar region , this means that the angle of @xmath256 at @xmath619 is less than @xmath280 , contradicting the fact that @xmath256 is a geodesic .
[ maxconvex ] if @xmath337 is a maximal tiled corridor ( with respect to the order @xmath394 ) in @xmath12 , then @xmath339 is convex in @xmath12 .
suppose that @xmath31 is not convex in @xmath12 .
then , since @xmath12 is cat(0 ) , there exists a vertex @xmath55 of @xmath31 so that @xmath31 fails to be locally convex at @xmath55 .
since all the angles of @xmath31 are of the form @xmath620 , it follows that @xmath31 has the angle @xmath621 at @xmath55 .
this means that @xmath622 is not convex in @xmath343 . in other words ,
if @xmath623 denote the tangent directions to @xmath569 at @xmath55 , then @xmath624 .
recall that @xmath343 is a bipartite graph with the edges of the length @xmath132 . since @xmath625 have different colors
, it follows that @xmath626 .
since @xmath343 has girth @xmath627 , it follows that @xmath628 .
therefore , there exists a small flat sector @xmath629 with the vertex @xmath55 and @xmath630 .
thus , by density of the walls of @xmath34 , we can find a small tile @xmath631 with the vertex @xmath55 and the angle @xmath344 at @xmath55 .
the union @xmath632 is , therefore , again a corridor and @xmath633 is a tiled corridor so that @xmath383 . since @xmath634 is strictly larger than @xmath31 , it follows that @xmath337 is not maximal .
contradiction .
in general , weakly maximal corridors ( i.e. , maximal with respect to the order @xmath393 ) need not be convex .
[ convexexist ] every corridor @xmath28 is contained in a convex corridor @xmath635 .
let @xmath339 .
let @xmath413 be a maximal corridor so that @xmath636 ( see proposition [ maxexists ] )
. then @xmath498 .
the corridor @xmath358 is convex by the above lemma . although we do not need this , given @xmath337 so that @xmath339 , we can find a canonical convex corridor @xmath358 containing @xmath31 .
namely , the set of weakly maximal tiled corridors @xmath485 so that @xmath637 is finite according to corollary [ finiteness ] . the number of maximal corridors among the corridors @xmath485 is finite as well .
let @xmath358 denote the intersection @xmath638 taken over all maximal corridors @xmath485 .
one can verify that @xmath358 is indeed a corridor .
convexity of @xmath358 follows from the fact that it is intersection of convex sets .
we can now finish the proof of theorem [ convex ] .
the first assertion of theorem follows from proposition [ ecorridor ] .
the second assertion is corollary [ convexexist ] .
in this section we prove two generalizations of theorem [ convex ] . our first goal is to provide a generalization which allows corridors to have infinite diameter .
we first have to generalize the notion of a _
tile_. we define _ extended ( infinite ) tiles _ , which are infinite sectors bounded by walls with the slopes @xmath292 and having the vertex angle @xmath132 or @xmath639 .
we will also allow the whole apartment as a single tile .
such infinite tiles appear as limits ( in the chabauty topology on the space of closed subsets of @xmath34 ) of sequences of the ordinary tiles .
accordingly , we generalize the upper and lower roof functions to allow functions defined on @xmath640 .
we still require such piecewise - linear functions to have only finitely many break - points : every roof function admits a partition of @xmath640 in finitely many subintervals , where the function is linear .
then , a _ extended model corridor _ is the closed region bounded from above and below by graphs of upper and lower roof functions ; we also declare @xmath34 to be a corridor ( consisting of a single infinite tile ) .
it is clear that each corridor @xmath567 is a union of finitely many tiles , two of which are infinite , with the vertex angles @xmath132 .
therefore , for every corridor @xmath31 , there exists a extended tiled corridor @xmath337 so that @xmath466 : the corridor @xmath337 still has only finitely many tiles , but some of them are allowed to be infinite .
thus the `` extremes '' of the extended corridors @xmath31 are points at infinity of @xmath31 .
namely , if @xmath540 is any complete geodesic contained in @xmath31 , we declare the points at infinity of @xmath256 to be the extremes of @xmath31 .
therefore , @xmath31 now has ( typically ) more than 2 extreme points .
let @xmath641 , a space modeled on @xmath19 and covered by finitely many apartments .
( it is no longer enough to assume local finiteness . )
we define ( infinite ) corridors in @xmath12 by repeating the definition for the ordinary corridors . every complete geodesic @xmath206 is contained in a corridor .
the arguments remain pretty much the same as in the proof of proposition [ ecorridor ] , except that we start by observing that , for sufficiently large @xmath423 , the subrays @xmath642 in @xmath256 do not cross any walls in @xmath12 . therefore , there are infinite tiles @xmath643 containing these subrays . then @xmath644 is a finite geodesic segment and we can cover it with a tiled corridor @xmath31 using proposition [ ecorridor ] .
one then shrinks the tiles @xmath645 , to @xmath646 so that @xmath647 is a corridor covering @xmath256 , see figure [ f6 ] .
the definitions of the orders @xmath393 and @xmath394 on corridors generalize almost verbatim to the corridors having infinite tiles .
the only modification one needs is that instead of dealing with finite extreme vertices , we require an extended corridors to contain certain points at infinity .
equivalently , one can ask for corridors covering a fixed complete geodesic in @xmath12 .
one again sees that for every extended tiled corridor @xmath337 , there exists an extended weakly maximal tiled corridor @xmath413 so that @xmath389 : while considering an infinite increasing sequence of tiled corridors we no longer can argue that the tiles have uniformly bounded diameter , but such unbounded sequences of tiles still converge to infinite tiles . ( the limit could be the whole apartment @xmath34 . ) given a weakly maximal extended corridor @xmath337 we construct a maximal extended corridor @xmath413 so that @xmath383 by appealing to the fact that @xmath12 has only finitely many thick vertices .
the proof that every maximal extended corridor is convex is exactly the same as that one of lemma [ maxconvex ] . by combining these observations
, we obtain [ convex1 ] 1 .
every complete geodesic @xmath648 is contained in an extended corridor @xmath31 .
every extended corridor @xmath31 is contained in a maximal extended corridor @xmath358 .
the corridor @xmath635 is convex .
let @xmath649 be weyl sectors which share the tip @xmath650 and which are super - antipodal at @xmath82 .
then there exists an extended corridor @xmath28 so that @xmath651 moreover , the corridor @xmath31 contains a geodesic with the ideal points in @xmath652 . it remains to prove part 3 .
we start with the case when @xmath82 is a vertex and , hence , @xmath653 are bounded by sub - walls .
then , @xmath649 , are extended tiles in @xmath12 and , by our assumptions , @xmath654 is an extended corridor .
the geodesic contained in this corridor is the union of rays @xmath655 which are super - antipodal at @xmath82 and asymptotic to points in @xmath649 .
if @xmath82 is not a vertex , @xmath194 have to be antipodal at @xmath82
. moreover , the boundary rays of @xmath656 are not sub - walls .
in particular , they do not contain thick vertices .
therefore , there exist ( special ) vertices @xmath657 close to @xmath82 and weyl sectors @xmath658 with the tips @xmath27 , so that @xmath659 and @xmath660 is a tile .
therefore , we obtain an extended corridor @xmath661 containing @xmath662 .
we now return to the discussion of ordinary corridors . in the following theorem it suffices to assume that @xmath12 is locally finite , modeled on @xmath19 .
[ convex2 ] for @xmath25 let @xmath435 be chambers in the buildings @xmath188 , which are given structure of thick buildings , @xmath191 . then there exists a convex corridor @xmath28 containing @xmath23 , so that @xmath663 as before , set @xmath494 .
we start with a convex corridor @xmath31 containing @xmath23 , given by theorem [ convex ] .
we modify @xmath31 twice : first to ensure @xmath664 and next , to ensure @xmath665 we describe only the first modification , since the second is obtained by applying the same procedure to the modified convex corridor and switching the roles of @xmath23 . to simplify the notation , set @xmath666 , @xmath667 .
we let @xmath668 be the tangent direction to @xmath256 at @xmath82 .
case 1 ( the generic case ) : @xmath82 is not a thick vertex , is not contained in a thick sub - wall and the germ of @xmath256 at @xmath82 is not contained in a sub - wall . in particular , @xmath669 is the apartment .
then , according to theorem [ convex ] , we can assume that @xmath670 .
therefore @xmath671 .
case 2 : we now allow @xmath82 to be contained in a thick sub - wall , but retain the rest of the restrictions . since @xmath82 is not a thick vertex , @xmath139 is a building ( but not a single apartment ) . the chamber @xmath357 is a half - apartment in @xmath139 and @xmath668 is a regular direction .
\(a ) if @xmath672 , then @xmath673 and we are done .
\(b ) if @xmath674 , then , there exists a ( necessarily regular ) direction @xmath675 antipodal to @xmath668 and we can replace @xmath256 with a slightly longer geodesic @xmath676 tangent to @xmath264 .
the convex corridor @xmath677 containing @xmath678 will then contain @xmath82 in its interior and , therefore @xmath679 hence , we are done in this case as well .
case 3 : @xmath82 is a thick vertex , but @xmath668 is a regular direction , i.e. , the germ of @xmath256 at @xmath82 is not contained in a sub - wall .
\(a ) suppose that @xmath680 this means that @xmath357 is contained in an arc @xmath681 of the length @xmath682 .
since @xmath139 has extendible geodesics , we can assume that the length of @xmath683 is @xmath280 and , hence , @xmath61 is antipodal to @xmath668 .
we now use the same arguments as in case 2a : extend @xmath256 to a longer geodesic @xmath678 tangent to @xmath61 .
let @xmath677 be the convex corridor containing @xmath684 .
by convexity , @xmath685 will contain a geodesic in @xmath139 connecting @xmath686 .
since @xmath668 is regular , this geodesic is unique and equals @xmath683 .
in particular , @xmath687 .
\(b ) if ( a ) fails then @xmath357 contains a regular direction @xmath264 so that @xmath688 we again repeat the arguments in the case 2(b ) : take a longer geodesic @xmath678 tangent to @xmath264 .
then the corresponding convex corridor @xmath677 will contain a germ of @xmath357 at @xmath82 .
case 4 : @xmath668 is a singular direction and the germ of @xmath256 at @xmath82 is contained in a sub - wall .
then @xmath31 contains a degenerate irreducible component @xmath576 so that @xmath689 .
we have the same dichotomy as in case 3 : \(a ) suppose that @xmath357 is contained in an arc @xmath681 of the length @xmath280 .
we construct @xmath684 as before .
there exists a flat sector @xmath141 ( germ of a half - apartment ) with the tip @xmath82 so that @xmath690 .
we then find a tile @xmath691 so that @xmath692 , i.e. , one of the edges of @xmath33 is contained in @xmath684 , so that @xmath693 is a corridor .
see figure [ f10 ] . set @xmath694 by the construction , @xmath695 .
we claim that @xmath677 is a corridor .
we obtain a weakly isometric embedding @xmath696 by combining the weakly isometric embeddings of @xmath697 and @xmath31 .
if @xmath677 were to admit a bridge , there would have to be at least two vertices where @xmath677 fails to be locally convex .
one of them is in @xmath697 .
since @xmath31 is convex , it does not have any nonconvex vertices .
thus , @xmath677 is a corridor .
\(b ) if @xmath357 contains a regular direction @xmath264 so that @xmath698 , we repeat the arguments in case 3(b ) .
let @xmath699 be ( convex compact ) cells in @xmath12 .
then there exists a finite set of corridors @xmath572 in @xmath12
so that for every @xmath700 and chambers @xmath701 there exists a corridor @xmath702 containing @xmath23 and so that @xmath703 the product @xmath704 is compact . for every subcomplex @xmath31 ( we are interested only in corridors , of course ) , @xmath705 is an open subset , since @xmath706 is contained in an apartment in @xmath12 .
therefore , we obtain an open covering of @xmath704 by the sets of the form @xmath707 therefore , by compactness , there exists a finite number of corridors @xmath572 so that for every @xmath708 and every chamber @xmath709 , @xmath710
for a space @xmath641 we define @xmath711 , the set of all _ convex extended corridors _ in @xmath12 . in this section
we will frequently omit the adjective _ extended _ since almost all the discussion of corridors will be local .
given @xmath641 , we define an _ enlargement of @xmath12 along a ( extended ) convex corridor @xmath28 _ to be the space @xmath712 obtained by attaching to @xmath12 the apartment @xmath34 along @xmath28 .
the attachment is given by a weakly isometric embedding @xmath713 .
we will also use the term _ enlargement _ for the tautological embedding @xmath714 .
it is important to note that every wall in @xmath12 is an extended corridor , therefore , enlargements include attaching apartments along walls . if @xmath372 is obtained from @xmath12 by enlargement , then @xmath372 is cat(0 ) .
since @xmath31 is contractible , @xmath372 is simply - connected .
since @xmath372 is a locally - euclidean cell complex , it suffices to check that the links of vertices @xmath55 of @xmath12 are cat(1 ) spaces .
let @xmath55 be a vertex of @xmath31 .
if @xmath345 is locally convex at @xmath715 then @xmath716 is again cat(1 ) , since @xmath717 is convex in both @xmath718 and @xmath719 .
consider therefore , a vertex @xmath715 where @xmath720 is not locally convex
. then @xmath721 is obtained from @xmath718 by attaching either one arc @xmath242 ( if @xmath55 is not a cut - vertex of @xmath358 ) or two arcs @xmath722 ( if @xmath55 is a cut - vertex ) ; in either case , the length of @xmath242 ( and @xmath722 ) is greater than or equal to @xmath723 .
we will consider the case of one arc since the case of two arcs is similar . since @xmath28 is convex , the arc @xmath242 is attached to vertices @xmath625 of @xmath92 , so that @xmath724 ( which is the inner angle of @xmath31 at @xmath55 ) .
let @xmath725 be the shortest embedded circle . if @xmath357 does not contain @xmath242 , then @xmath726 and @xmath727 .
if @xmath357 contains @xmath242 , then @xmath728 , where @xmath729 is an arc in @xmath92 connecting @xmath625
. then , @xmath730 and , hence @xmath731 thus , @xmath732 is cat(1 ) and @xmath372 is cat(0 ) .
we therefore obtain a category @xmath733 , whose objects are element of @xmath157 ( i.e. , finite spaces modeled on @xmath9 ) and whose morphisms , @xmath734 , are compositions of enlargements .
[ conserve ] every enlargement @xmath735 is a weakly isometric embedding .
the map @xmath736 is 1-lipschitz , in particular , its restriction to every apartment in @xmath12 is an isometry . verifying that a map is a weakly isometric embedding is a local problem .
let @xmath182 be a vertex
. then @xmath721 is obtained from @xmath718 by attaching at most two arcs , each having length @xmath737 .
therefore , the inclusion @xmath738 is a @xmath220-isometric embedding .
it follows that the map @xmath739 is a weak isometry .
the fact that @xmath736 does not increase distances is clear .
suppose that @xmath740 is an embedded circle of length @xmath741 ( say , an apartment ) .
set @xmath742 .
suppose that there exist @xmath743 so that @xmath744 then , combining the geodesic @xmath745 with the geodesic @xmath746 , we obtain an embedded circle of length @xmath747 in @xmath92 .
this contradicts the fact that @xmath372 is cat(0 ) and @xmath732 is cat(1 ) .
therefore , for every apartment @xmath748 , its image @xmath749 is convex in @xmath372 and , hence , the restriction of @xmath736 to @xmath750 is an isometry .
note that @xmath714 is very seldom an isometric embedding , namely , if and only if @xmath720 is convex , which means that @xmath31 is a single ( extended ) tile .
[ corridorconservation ] let @xmath751 be obtained from @xmath12 by enlargement along a ( extended ) convex corridor @xmath635 and @xmath714 be the tautological embedding .
let @xmath752 .
then the image @xmath753 is again a corridor ( not necessarily convex ) .
proof of axiom c1 is the same as in lemma [ isomembed ] . we need to verify axiom c2
. we will identify @xmath31 and @xmath371 .
suppose that @xmath754 does admit a bridge @xmath755 .
let @xmath756 be the edges adjacent to the trapezoid . since @xmath31 admits a bridge , it is not convex in @xmath372 .
since @xmath31 was convex in @xmath12 , this means that @xmath736 is not a local isometry at the vertices @xmath613 of the bridge ( where @xmath757 ) .
it follows that @xmath613 are vertices of @xmath758 . by convexity , @xmath759 as well .
moreover , the germs of @xmath760 at @xmath761 have to be contained in @xmath758 ( @xmath237 ) , otherwise , they would be convex in @xmath372 .
thus , @xmath762 and , therefore , @xmath763 admits a bridge with trapezoid @xmath361 .
since the convex hull of @xmath763 is contained in the apartment @xmath34 ( which was attached to @xmath372 along @xmath358 ) , we conclude that @xmath764 .
this , however , contradicts the fact that the image of @xmath358 in @xmath34 is a model corridor .
[ largercorridor ] let @xmath714 be a morphism in @xmath733 .
then for every @xmath765 , there exists a ( extended ) convex corridor @xmath763 containing @xmath371 .
since @xmath736 is a composition of enlargements , we construct @xmath358 inductively starting with @xmath766 . if @xmath767 is a convex corridor and @xmath768 is an enlargement , then @xmath769 is a corridor according to lemma [ corridorconservation ] .
we then replace the corridor @xmath770 with a convex corridor in @xmath771 containing @xmath770 ( see part 1 of theorem [ convex1 ] ) .
we are now ready to prove theorem [ main ] .
we proceed by constructing inductively an increasing sequence @xmath772 , @xmath773 , by using enlargements along ( extended ) convex corridors is in @xmath774 . ]
then the building @xmath12 will be the direct limit of the sequence @xmath36 .
set @xmath776 since @xmath774 is countable for every @xmath772 , we get a bijection @xmath777 .
define @xmath778 by @xmath779 let @xmath780 denote the standard enumeration , so that both @xmath781 are 1-lipschitz functions @xmath782 .
in particular , @xmath783 lastly , set @xmath784 .
we start with @xmath785 , the model apartment .
however , we could as well start with any @xmath786 .
then , for @xmath787 we compute the ( extended ) convex corridor @xmath788 , @xmath789 , where @xmath790 .
we have a morphism @xmath791 which is a composition of enlargements . according to corollary [ largercorridor ] , there exists a convex ( extended ) corridor @xmath792 containing @xmath793 .
the corridor @xmath358 is not unique .
this does not matter with one exception : if @xmath31 is a geodesic in an apartment in @xmath37 , @xmath371 is still a geodesic in @xmath17 ; in this case we take @xmath794 .
we then set @xmath795 be the enlargement of @xmath17 along @xmath358 . to make sure that @xmath21 is well - defined
, we need that @xmath796 is a corridor in @xmath37 , for some @xmath797 .
this immediately follows from the inequality ( [ lip ] ) . our next task is to give @xmath12 structure of a building .
observe that each apartment in @xmath17 is also an apartment in @xmath21 ( see lemma [ conserve ] ) .
therefore , we define the set of apartments in @xmath12 as the union of the sets of apartments in @xmath17 , @xmath787 .
since each @xmath17 is modeled on @xmath19 , we conclude that @xmath12 is also modeled on @xmath19 : the transition maps are in @xmath11 and for every pair of apartments @xmath798 @xmath799 is a closed and convex subset in @xmath34 bounded by sub - walls .
let @xmath800 and @xmath801 be germs of weyl chambers .
then there exists an apartment in @xmath12 containing @xmath23 and the germs @xmath30 .
in particular , every two points of @xmath12 belong to a common apartment .
take the smallest @xmath802 so that @xmath803 and @xmath804 , @xmath191 . according to theorem [ convex2 ]
, there exists a convex corridor @xmath805 containing @xmath27 and @xmath29
. then there exists @xmath38 so that @xmath789 .
therefore , @xmath21 is obtained by attaching an apartment @xmath34 along a convex corridor @xmath39 containing @xmath371 , @xmath791 .
it follows that the images of @xmath806 in @xmath21 will belong to an apartment @xmath34 .
hence , the image of @xmath34 in @xmath12 will contain @xmath806 , @xmath191 .
we can now give @xmath12 structure of a cat(0 ) space , so that every apartment is isometrically embedded .
even though the morphisms @xmath791 are not isometric , nevertheless , for every @xmath807 there exists @xmath808 so that @xmath23 belong to an apartment in @xmath17 ; therefore , for every @xmath809 we have : @xmath810 we , hence , obtain a metric on @xmath12 .
since every morphism in @xmath733 is 1-lipschitz , it follows that for every @xmath808 , @xmath811 is also 1-lipschitz .
the metric on @xmath12 is a direct limit of the metrics on the cat(0 ) spaces @xmath17 ; therefore @xmath12 is also cat(0 ) ( as the cat(0 ) condition is a condition on quadruples of points ) .
let @xmath812 be antipodal weyl sectors with the common tip @xmath82 .
then there exists an apartment @xmath149 containing @xmath662 .
we first find @xmath802 so that @xmath813 .
since @xmath194 are antipodal in @xmath12 , and @xmath814 is 1-lipschitz , it follows that @xmath194 are super - antipodal in @xmath37 .
therefore , according to theorem [ convex1 ] , there exists an extended convex corridor @xmath805 containing @xmath815 .
the corridor @xmath31 equals @xmath796 for some @xmath38 . recall that @xmath21 is obtained from @xmath17 by enlargement along a convex corridor containing the image @xmath816 .
it follows that the images of @xmath194 in @xmath21 ( and , hence , in @xmath12 ) are contained in an apartment .
we thus verified axioms a1@xmath186 and a2@xmath186 of euclidean buildings for @xmath12 .
@xmath12 is thick .
let @xmath251 be a wall in @xmath12 . then @xmath251 is contained in an apartment @xmath149 , which means that @xmath251 is a wall in an apartment @xmath817 for some @xmath802 .
since @xmath251 is also a ( degenerate ) corridor it follows that there exists @xmath38 so that @xmath21 is obtained from @xmath17 by attaching an apartment @xmath750 along @xmath251 .
therefore , the wall @xmath499 will be the intersection of three ( actually , four ) half - apartments in @xmath12 and , hence , thick .
this concludes the proof of theorem [ main ] . one can modify the construction of the building @xmath12 as follows . besides
_ enlargements _
, one allows isometric embeddings @xmath818 , where @xmath21 is obtained from @xmath17 by attaching an apartment ( or another @xmath819 ) along a convex subcomplex _
isometric _ to a convex subcomplex in @xmath34 ( resp . in @xmath820 ) . as a corollary of theorem [ main ] , we prove let @xmath821 , the closure of @xmath11 in the full group of isometries of @xmath34
. then there exists a thick euclidean building @xmath372 of rank 2 modeled on @xmath822
. let @xmath12 be a building as in theorem [ main ] .
pick a nonprincipal ultrafilter @xmath823 and consider the ultralimit @xmath824 of the constant sequence @xmath825 with fixed base - point ( see e.g. @xcite for the definitions ) . then , according to @xcite , @xmath826 is the required building : it is a thick euclidean building modeled on @xmath822 .
one would like to generalize theorem [ main ] in several directions : at the moment , we do not know how to deal with 1 .
however , 2 and 3 can be approached along the same lines .
namely , instead of constructing a building @xmath12 as a direct limit of a totally - ordered ( by inclusion ) set of spaces @xmath772 , one would like to allow direct limits using partially ordered sets .
this hinges on the following problem , to which we expect a positive solution : fix a coxeter complex @xmath19 on which all spaces @xmath641 are modeled .
recall that in section [ category ] , we defined a category @xmath733 of finite complexes modeled on @xmath19 , where morphisms are given by compositions of enlargements . defining such
@xmath828 will solve both 2 and 3 : see @xcite for the explanation of how to use the amalgamation property to form buildings with `` large '' automorphism groups .
constructing @xmath828 in the context of spherical buildings modeled on @xmath151 is the main technical result of the paper by k. tent @xcite and is nontrivial already in this case .
doing this for affine coxeter groups appears to be much harder .
existence of the category @xmath828 would imply that the affine buildings constructed in our paper embed in rank 2 affine buildings with highly transitive automorphism group .
a partial confirmation to this comes from the following observation .
let @xmath12 be a building constructed as in section [ mainproof ] ; such a building is unlikely to have any nontrivial automorphisms .
let @xmath372 be a rank 2 building of countably infinite thickness which is modeled on @xmath19 ( i.e. , every wall in @xmath372 is the intersection of countably infinitely many distinct half - apartments ) .
then there exists an embedding @xmath109 .
in particular , if @xmath7 is of crystallographic type , there exists a building @xmath372 corresponding to a reductive algebraic group and , hence , having highly transitive automorphism group .
w. goldman , _ geometric structures on manifolds and varieties of representations _ , in `` geometry of group representations ( boulder , co , 1987 ) , '' p. 169198
, contemp .
74 , amer . math .
soc . , 1988 .
a. parreau , _ immeubles affines : construction par les normes et tude des isomtries _ , in : `` crystallographic groups and their generalizations '' ( kortrijk , 1999 ) , p. 263302
, contemp .
262 , amer .
soc . , providence , ri , 2000 .
m. ronan , _ a construction of buildings with no rank 3 residues of spherical type _ , in : `` buildings and the geometry of diagrams , ( como , 1984 ) '' , p. 242248 .
lecture notes in math .
1181 , springer , berlin , 1986 .
|
we construct rank 2 thick nondiscrete affine buildings associated with an arbitrary finite dihedral group .
|
pulsar timing arrays ( ptas ) are collections of millisecond pulsars @xcite whose joint timings will show correlations that are a specific signature of gravitational waves ( gws ) passing between the earth and the pulsars @xcite . currently , there are three ptas working collaboratively to detect gws in this way @xcite
. these arrays will soon have the sensitivity to detect single extragalactic sources of gws , @xcite which are resolvable above a gw background @xcite . to increase the efficiency of pta observations and
potentially decrease the time until a detection is made , various groups have considered the optimization of pta observations @xcite , but all have assumed an equal probability of detection across the sky . while all directions are equally likely to contain a gw source , there should exist more probable locations for the brightest gw source . and
while the probability is quite small for there to exist a source that stands out above the background , the analysis in this paper identifies the potential locations of that source using what is currently known about the distribution of galaxies in the local universe , and allows others to enhance discussions on the optimization of a pta @xcite .
supermassive black hole binary ( smbhb ) systems with periods of months to years are thought to be the most important source of gravitational waves @xcite .
binaries like these form when galaxies containing nuclear black holes merge @xcite .
corresponsingly , the number of such binaries should be greatest where galaxy mergers are more frequent ; i.e. , in galaxy clusters .
all things being equal , the nearest clusters will play host to the brightest sources .
this sugests that , as pta sensitivities increase , we look toward the largest or richest nearby clusters ( virgo , fornax , norma , perseus and coma ) as the most likely location of the earliest detectable smbhb source . here
we make this expectation quantitative .
below in 2 , we use a compilation of several galaxy surveys to identify the mass , distance , and location of all galaxies within @xmath0mpc . from this data
we estimate the smbhb mass and the lifetime of gw emission in a detectable pta band for each galaxy in 3 , and together with the distance , we identify the probability of the existence of a detectable source in a given direction .
4 contains a full explanation of the results with sky position maps .
we investigate the probability of detection given an increasing pta sensitivity in 5 and a summary of our findings is found in 6 .
we require knowledge of the mass and distance of each galaxy in order to estimate the amplitude of the gws emitted by a smbhb that may exist at the center of that galaxy .
we start by searching the extragalactic distance database ( edd ) , created by @xcite .
this database is a compilation of many extragalactic surveys with the intention of compiling all visible galaxies within @xmath1 mpc ( z=0.03 ) , where @xmath2 km / s / mpc .
the eddhttp://edd.ifa.hawaii.edu ] recently updated its records with the 2m++ galaxy reshift catalog which reaches @xmath3 complete out to 200@xmath4 mpc and is conservatively complete to 60@xmath4 mpc @xcite .
this gives our sample the same completeness .
additionally , the edd was updated with an extensive survey of all galaxies within 11 mpc @xcite , later in the paper local sources will be highlighted and this recent addition to the edd gives us great confidence in our ability to talk about neighboring galaxies .
the lyon - meudon extragalactic database ( leda ) @xcite is the largest database that the edd draws on , but while leda compiles over three million objects @xcite the edd only gathers information from leda for galaxies that are found in other surveys @xcite which at the time of this paper was just under a hundred thousand galaxies@xmath5 .
both the edd and leda take advantage of the labeling mechanism started in the principal catalogue of galaxies ( pgc ) @xcite .
the pgc number of every known galaxy is used to access the same galaxy across several surveys and to prevent any single galaxy from being counted more than once . to obtain an accurate distance ,
the databases use various methods based on the known parameters of each galaxy .
there are two main distance moduli calculated in leda , mod0 " , which is calculated from a distance catalog using the tully - fisher relation or the faber - jackson relation , and modz " , which is calculated using redshift @xcite .
the edd primarily uses the tully - fisher relation to get distances , but compares the results with other distances to assure a common scale @xcite .
distance measurement using the tully - fisher relation is only available for about @xmath6 of galaxies , with the rest being derived from redshift measurements .
solar masses .
the above log - log plot shows a clear trend that is best represented by the line @xmath7 .
this plot was made using all galaxies with a known @xmath8 and a well defined b - band luminosity found in the extragalactic databases.,scaledwidth=45.0% ] we calculate the total central black hole mass of a galaxy using the m-@xmath8 relation when an accurate central velocity dispersion ( @xmath8 ) is found .
this is available for @xmath9 of the galaxies . in an attempt to expand the pool of usable galaxies , b - band luminosity is used as a surrogate for mass . in leda ,
most known galaxies have a calculated total b - band luminosity , which we converted to a total luminosity using a bolometric correction @xcite .
the number of galaxies for this work increased from 4,258 to 83,816 .
fig [ fig : bhvl ] compares the luminosity and black hole mass for galaxies with known @xmath8 .
while the relationship is most likely non - linear in lower mass regions , since ptas are only sensitive to a certain range of black hole masses , @xmath10 solar masses @xcite , we can ignore the low mass trend and only use the clear linear trend in the pta sensitive region .
we advise the reader that the line looks like a poor fit to the data because of the invisible density gradient in the gray region of the plot . for consistency , all gathered luminosities were converted to black hole mass using the above trend line which gives the power law , @xmath11 where l is the corrected luminosity of each galaxy . in fig
[ fig : totalgalaxies ] , we plot all found galaxies with a recorded b - band luminosity to look for survey biases .
this plot was created with equal - area pixels generated using mealpixprogram developed by gwastro research group and available from http://gwastro.org ] .
the empty areas in the plot are caused by the milky way galaxy , whose plane renders these areas of the sky unobservable , also called the zone of avoidance .
the largest value pixels in this plot , appearing as black and dark gray , contain galaxy clusters .
significant clusters are labeled on the plots throughout this paper .
the virgo cluster contains approximately 1300 galaxies @xcite ; the fornax cluster contains around 60 galaxies @xcite ; the norma cluser contains around 600 galaxies @xcite ; the perseus cluster contains about 500 galaxies @xcite ; and the coma cluster contains more than a thousand galaxies @xcite .
since all pixels not obstructed by the plane of the milky way contain a number of galaxies that are within the same order of magnitude , we deem the distribution across the sky to be reasonably non - bias , particularly with respect to nearby sources , which are most important to us . fig [ fig : disthist ] contains a plot of all galaxies with a recorded b - band luminosity in the combined databases as a histogram over distance .
we label the mean distance of some galaxy clusters , and in general the larger spikes in galaxies are due to clusters .
the number of galaxies grow out to approximately 150 mpc , and then fall off inversely with distance . from the completeness of the edd , we feel confident that in this paper we achieve a reasonably accurate representation of our local universe , i.e. the galaxies within 150 mpc .
now that we have a significant sample size of galaxies , we cut down to only the galaxies with luminosities great enough to potentially harbor a detectable source ; galaxies with central black holes larger than @xmath12 . in an attempt to remove bias to our closest neighbors
, we remove the andromeda galaxy and our galaxy from the dataset as well .
the effect of this mass cut can be seen on the gray histogram overlayed on fig [ fig : disthist ] , which leaves us with 75,486 galaxies .
the region of sky within 50 mpc , is the most effected by this cut , which makes sense given that less luminous galaxies will only be observable at smaller distances .
the virgo cluster is the only galaxy cluster strongly effected by this cut for similar reasons .
after the mass cut was made , we cut down to only the galaxies within 150 mpc of the milky way , which leaves us with 40,560 galaxies , just under half of the starting number .
the effect of this distance cut verses sky position can be seen in fig [ fig : massdistcutgalaxies ] .
solar masses , and have distances less than 150 mpc out to which the surveys are roughly complete.,scaledwidth=45.0% ] solar masses and are within 150 mpc . in comparison
to fig [ fig : totalgalaxies ] , this plot clearly shows the distribution of number density of galaxies in the local universe to be dominated by galaxy clusters .
this distribution shows no clear bias and is the distribution used in the rest of the paper.,scaledwidth=45.0% ]
to estimate the detectibility of gws with ptas , we use the smbhb mass and the distance of each galaxy to calculate the gw signal strength from each potential source .
we also calculate the lifetime of emission in a detectable pta band , and estimate the number of detectable sources .
this allows us to identify the probability of the existence of a source at a given amplitude in a given direction .
we start with the simplified assumption that all galaxies contain a binary with equal mass black holes . since we are only trying to estimate the hotness " of sky positions relative to each other , we work in proportionalities .
the overall factors are irrelevant at this stage of our understanding .
a more mathematical description of these proportionalities can be found in the appendix .
we use mealpix@xmath13 to divide the sky into equal - area pixels , and associate each galaxy with a particular pixel based on its sky location .
if we assume a given galaxy contains or contained a smbhb then the probability that the smbhb exists now is the ratio of the binary s lifetime , @xmath14 , to the age of the universe . if we further assume that the source is detectable as long as the smbhb exists then the expectation value of the number density of sources in a certain pixel , @xmath15 , is directly proportional to the lifetime , @xmath14 , of all sources in that pixel ; @xmath16 where n is the total number of galaxies in a given pixel and @xmath17 represents a particular galaxy in that pixel .
@xmath15 serves as one of the metrics we use to characterize the `` hotness '' of the gw sky .
however , @xmath15 says nothing about the relative strength of sources , just the number of them .
we therefore also use the metric @xmath18 that is proportional to the expectation value of the power in gws emitted from a particular pixel . consider the gw power contained in the residual response from a single source , p ; the response in pulsar timing to a gravitational wave of strain @xmath19 is proportional to @xmath19 times the period of the binary @xcite .
however , we assume that all the smbhbs have the same period , and drop the length of the period from our calculations . therefore , the residual response is proportional to @xmath19 .
the power in the pulsar timing residuals is the square of this response : @xmath20 the total energy emitted over the lifetime of the source is proportional to the power times the lifetime of the source , @xmath14 .
therefore @xmath21 where the integral is over all time .
when we add up all the sources in a particular pixel we get @xmath22 where n is the total number of galaxies in a given pixel and @xmath17 represents a particular galaxy in that pixel .
in essence , this is the total amount of energy a pta can expect to receive from this pixel over all time , and after dividing by the age of the universe would be the average power in the pixel .
therefore @xmath23 is also proportional to the expectation value of the power from this pixel and we have an expression for our second metric , @xmath18 , @xmath24 the above equations only depend on the gw strain , @xmath19 , and the lifetime of each smbhb , @xmath14 .
we use the standard dipole approximation from @xcite to estimate the strain @xmath19 ; @xmath25 where @xmath26 is the chirp mass of the smbhb .
the lifetime of the source , @xmath14 , is given by : @xmath27 the chirp mass of a binary system , @xmath26 , is proportional to the total mass of the binary , @xmath28 , since we have already assumed that the masses in the binary are roughly equal , @xmath29 we can now rewrite both the number of detectable sources and the gw signal strength in terms of @xmath28 and @xmath30 , which are the observed quantities gathered in 2 .
@xmath31 maps of the above quantities can be found in the next section .
there are many factors that contribute to the detection of a gw signal by a pta , and so measuring only one quantity is insufficient to assess the likelihood of detecting a gw source in a particular region of the sky .
for example , when calculating the number of detectable sources we rank the galaxies in terms of their likelihood of being ` on ' during an observation , while when calculating gw signal strength we rank the sources in terms of the overall power they are expected to contribute to the pulsar timing band over the course of their lifetimes .
these plots give different pictures of our local universe and are both needed to accurately understand a ptas probability of making a detection .
one way of finding probable locations for pta detection is to look at the expectation value of the number of sources in any given direction @xmath15 , which we estimate using eqn [ eqn : dect ] .
this value is proportional to the probability of a detectable source currently being in the pta band in a given pixel . using the equations in the appendix ,
the amount of time each potential source spends in the pta band is calculated , which is converted to a probability by dividing by the age of the universe .
we find the total probability of a single source currently being detectable to be @xmath32 , with the ` brightest ' location on the sky having a @xmath33 probability of currently containing a single source that stands out about the background in the pta band .
the probability of a detectable source currently being in the pta band is plotted verses sky position in fig [ fig : f1 ] , and as a function of distance in fig [ fig : f1hist ] .
these plots are dominated by galaxy clusters in the distance range of 50 to 100 mpc .
specifically the coma , norma and perseus clusters , which all contain numerous massive galaxies , while also being at a close enough distance for those galaxies to be resolvable by a pta .
these plots highlight a region of space between 50 and 75 mpc where a large portion of galaxies are a part of the centaurus supercluster .
this region of space which stretches down towards the norma cluster , is partially in the zone of avoidance and contains the great attractor @xcite .
while optical observations will likely not reveal the great attractor , ptas have the potential to discover the source of this attraction from gw observations . , from eqn [ eqn : power ] is plotted verses sky position .
this plot is dominated by a handful of local sources , including the virgo cluster.,scaledwidth=45.0% ] calculated using eqn [ eqn : power ] verses distance gives a sense of spatial depth to the sky plot in fig [ fig : f2 ] .
this plot shows that a local source has the potential to dominate a ptas search of single gw sources .
the black peak is the brightest source of @xmath18 , and we removed it to make the gray overlay showing the importance of local powerful sources.,scaledwidth=45.0% ] @xmath18 is estimated using the total gw power for each source in a particular pixel and at a particular distance , integrated over its lifetime , which effectively weights the power from each source with the probability of whether or not it will be caught ` on ' during an observation .
estimated using eqn [ eqn : power ] , @xmath18 is plotted verses sky position in fig [ fig : f2 ] and as a function of distance in fig [ fig : f2hist ] .
these plots highlight a handful of local sources that have the potential to dominate a ptas detection of a single gw source . in an attempt to show just how dominating a single source
can be , we removed the brightest source , ( ra @xmath34h , dec @xmath35 , distance @xmath36 mpc ) and overlayed the total power as a function of distance in fig [ fig : f2hist ] . with the largest source of gw power removed from the plots ,
a small number of galaxies , specifically those around the virgo cluster , continue to dominate the picture . unlike the @xmath15 plots which highlight the region of space between 50 and 100 mpc , the galaxies with the largest @xmath18 values are within a distance of 20 mpc .
for a pta to detect a source , that source must not only be emitting gw radiation in a detectable band , but also at a sufficient amplitude . while the above plots provide an accurate picture of the local landscape for pta detectable sources , we have not yet taken into account the threshold of a pta to detect these sources .
any pta will have a minimum detectability threshold defined by its sensitivity .
this threshold is directly related to the strain amplitude , @xmath19 of a potential gw source . over time
, the ptas will continually lower this threshold as improvements to timing , data analysis , and amount of data continue .
eventually , virtually all potential sources in section 4 will be detectable , and the maps presented are the best guides of where to focus efforts .
however , the first sources detected will be the sources with the largest @xmath19 . in this section , we make an educated guess as to where those first sources might be located based on our collected data .
we start by considering the ten brightest potential sources , which we label case a. to mimic improvements to the sensitivity , we also consider two more cases , b @xmath37 c , that reduce the initial cutoff value by subsequent factors of @xmath38 .
our goal is to predict which of these bright sources will be detected first .
therefore , once all the sources above a given threshold have been identified , a probability is assigned to each source by considering its lifetime over the total lifetime of the sample , or @xmath15 . in case
a we consider the 10 sources with the highest amplitude @xmath19 according to eqn [ eqn : strain ] .
the likelihood of one of those sources being detected by a pta is plotted in fig [ fig : ha ] .
the pixel with the most likelihood contains two sources .
each of the two sources in that pixel has a lifetime of about one quarter of the total lifetime of all the detectable sources . in fig
[ fig : hahist ] , we see that half of these sources are within 20 mpc .
when the sensitivity is increased by a factor of @xmath38 in case b , there are thirty three potential sources .
fig [ fig : hb ] plots the likelihood that one of these sources is detected . as in case a , the plot is dominated by one very likely pixel , however in this case there is only one source in that pixel and it contains 20@xmath39 of the total lifetime of all detectable sources .
this source has the largest @xmath18 value , discussed in 4.2 , and while it has a smaller strain than the initial ten , its lifetime is six times longer than any of the potential sources in case a. this source is at a distance of 2.2 mpc , and the distance distribution of the other sources is plotted in fig [ fig : hbhist ] , where a third of the potential sources are within 20 mpc . in case c , the sensitivity
is increased one order of magnitude from case a , and there are 237 potential sources .
the likelihood of detecting a certain source is plotted in fig [ fig : hc ] .
this plot is unlike the case a and case b plots since it has sources that are more distributed in clusters .
the pixel with the largest total lifetime contains three sources , and contains five percent of the total lifetime of all detectable sources at this cutoff .
fig [ fig : hchist ] shows that these sources are evenly distributed throughout distance with a small hotspot " appearing around the distance of the virgo cluster .
overall , the first two cases reveal very specific sources that dominate the likelihood of detection .
it is important to note that as sensitivity is lowered the new sources that will become detectable have the potential for significantly longer lifetimes and these sources are more likely to be found in galaxy clusters , specifically the virgo cluster .
in this paper , data is gathered from a compilation of at least @xmath3 complete galactic surveys out to 200 mpc .
these galaxies are all assumed to contain a smbhb with equal mass black holes , and the total central mass is calculated using b - band luminosity as a surrogate for mass .
this data set is then cut down to only detectable sources ( i. e. sources with a total central mass larger than @xmath12 and with in a distance of 150 mpc ) . from this data ,
two metrics are used to estimate the hotness " of the gw sky : @xmath15 , which is proportional to the number density of sources in a given direction ; and @xmath18 , which is proportional to the expectation value of the power in gws emitted from a particular direction .
using @xmath15 , we are able to calculate the specific probability that each pixel contains a currently radiating gw source in the pta band . while this number is very small for any individual pixel , we find a total probability of @xmath32 that one of the galaxies we considered is detectable .
the ` brightest ' location on the sky has a @xmath33 probability of currently containing a single source that stands out about the background in the pta band .
overall the distribution of single sources potentially detectable by a pta has a larger number density around local galaxy clusters . while the gw signal strength is dominated by a handful of sources , with the region of sky around the virgo cluster
having a larger number density of these ` bright ' sources .
work by @xcite has been done to show that if a new powerful timing pulsar was discovered in the direction of the virgo cluster , a pta would have twice the sensitivity to a region about @xmath40 around the discovered pulsar .
given that the sensitivity of a single pulsar to a gw source falls off as @xmath41 , where @xmath42 is the angle between a pulsar and a gw source @xcite , we recommend focusing the search for new pulsars in the vicinity of the virgo cluster . as stated earlier in this paper , this is only a broad estimate that assumed an equal probability for all galaxies to contain a binary with equal mass black holes .
future work is planned to incorporate work done by @xcite , which uses the sloan digital sky survey and the millenium simulation data @xcite to search for smbhbs in the redshift range of @xmath43 .
combining these two data sets is the next step towards creating realistic population distributions for single source gws detectable by ptas .
while the distribution of detectable single sources will most likely scale with any new estimate , @xmath18 is more affected by these specific probabilities , and a future paper will address these factors . in this section , we will derive two statistics which can be calculated from existing survey data and are proportional to two measures of gravitational wave strength in a given direction .
the two measures are the number of detectable binary systems and the strength of the stochastic gw signal coming from a particular region of the sky .
we will denote these quantities as @xmath44 and @xmath45 , respectively .
our calculation starts with the differential rate of smbh coalescence given by @xmath46 .
this quantity represents the number of smbh binary systems coalescing per unit observer time , per unit chirp mass , @xmath26 , of the system , per unit red shift z , occurring within a solid angle @xmath47 as seen by an observer at earth .
given the amount of time , @xmath48 , that a binary system spends emitting a gw with frequency between @xmath49 and @xmath50 , the number of detectable binary systems per unit solid angle may be expressed as : @xmath51 in the above , the integration should be performed over that region of @xmath52 , @xmath26 , and @xmath49 where gws would be detectable by a given pta configuration .
given the amplitude of the gravitational wave strain , @xmath53 , emitted by a particular smbh system , the strength of the stochastic gw signal may be written as : @xmath54 we want to make estimates of the relative strength of both @xmath44 and @xmath55 using observational data from galaxy surveys . in order to do this
, we will make the following assumptions : 1 ) @xmath56 is proportional to the number of observable galaxies per unit solid angle in a given direction , 2 ) the evolution of the smbh binaries are dominated by the effects of gw emission , 3 ) the probability of a galaxy harboring a smbh is the same for all galaxies , 4 ) the chirp mass of the binary system is proportional to the total luminosity of the galaxy , @xmath57 , 5 ) all galaxies of interest have @xmath58 . using the fact that @xmath56 is independent of frequency , equation [ dndomega ]
may be written as : @xmath59 where @xmath60 is the total time a smbh with chirp mass @xmath26 is detectable in a given pta and is given by : @xmath61 the frequencies @xmath62 and @xmath63 are the lowest and highest detectable frequencies given the sensitivity of a pta and a particular smbh binary system . note that both @xmath62 and @xmath63 depend on @xmath26 and @xmath52 .
these frequencies are calculated using the following expression for the gw strain amplitude : @xmath64 together with an expression for the minimum detectable strain of a pta .
let @xmath65 be the total number of galaxies per unit total luminosity , per unit red shift , per unit solid angle .
assumptions one and three allow us to write @xmath66 where @xmath67 is a constant .
the total number of galaxies per unit solid angle in a given direction is given by @xmath68 with assumption 4 , we can write @xmath69 where @xmath70 is a constant . putting this all into equation [ dndomega2 ] , we have @xmath71 given that @xmath72 is the fractional number of galaxies in a given direction with total luminosity between @xmath57 and @xmath73 located at a red shift between @xmath52 and @xmath74 , we can write the detectable number of systems as @xmath75 where the angle brackets represent an average over all galaxies in a particular direction . since @xmath76 , we find that @xmath77 where @xmath78 , defined in the above equation , is a quantity that may be estimated from galaxy catalogue data and is directly proportional to the number of detectable sources per unit solid angle in a given direction .
we can also define an idealized statistic which is valid for the case where the pta can detect all frequencies down to some cuttoff frequency . in this case
, we can ignore the frequency dependent terms and use @xmath79 like @xmath78 , @xmath80 is proportional to @xmath81 but it only depends on the properties of the galaxy distribution and not the details of a particular pta .
an estimate for @xmath80 is simply the sum of @xmath82 over all galaxies in a particular small region of the sky . using the same assumptions as with equation [ dh2domega ]
, one can show that : @xmath83 here , the range of integration is over all @xmath84 at a fixed frequency @xmath49 , unlike the expression for @xmath44 . using assumption 2 ) , we know that @xmath85 this , together with equation [ hs ] gives us the following : @xmath86 where @xmath87 as defined above may be estimated from existing data and is proportional to the strength of the stochastic gw emission in a particular direction .
an estimate for @xmath87 is obtained by summing @xmath88 over all galaxies in a particular direction .
|
the steadily improving sensitivity of pulsar timing arrays ( ptas ) suggests that gravitational waves ( gws ) from supermassive black hole binary ( smbhb ) systems in the nearby universe will be detectable sometime during the next decade .
currently , ptas assume an equal probability of detection from every sky position , but as evidence grows for a non - isotropic distribution of sources , is there a most likely sky position for a detectable single source of gws ? in this paper , a collection of galactic catalogs is used to calculate various metrics related to the detectability of a single gw source resolvable above a gw background , assuming that every galaxy has the same probability of containing a smbhb .
our analyses of these data reveal small probabilities that one of these sources is currently in the pta band , but as sensitivity is improved regions of consistent probability density are found in predictable locations , specifically around local galaxy clusters .
|
since the original theoretical proposals@xcite of the existence of topological insulators ( tis ) and their eventual realization in hgte / cdte and inas / gasb quantum wells ( qws),@xcite extensive studies of their properties have been performed .
because of the topological non - triviality of topological insulators , it was expected that the dissipationless current carried by topologically protected edge states could be utilized to construct electronic devices with low power consumption .
experimentally , however , the edge states that were expected to be robust were found to become unstable when the sample of hgte / cdte quantum wells became somewhat larger and reached sizes of a few microns .
this experimental finding stirred up a lot of interest .
how disorder , magnetic impurity , electron - electron interaction , electron - phonon interaction , etc , and the various combinations of these effects , could affect transport in edge states as well as the topological properties of tis themselves have been intensively studied.@xcite the notion that topological superconductors ( tss ) , similar to tis , possess topologically protected gapless edge or surface states traversing through the superconductor gap , was also put forward .
it is anticipated@xcite that superconducting vortices in two dimensional chiral @xmath0-wave tss can bind majorana fermions ( mfs ) , which persist at zero energy and obey non - abelian braiding statistics .
this implies that adiabatically exchanging two majorana fermions noncommutatively transforms the system from one ground state to another .
due to the protection of the non - abelian braiding properties a quantum bit encoded in mfs is immune to external environmental noise .
this gives mfs considerable promise in topological quantum computing .
motivated by this idea , a great deal of effort has been expended to examine many materials and structures to find ways to produce mfs . because mfs can appear when @xmath1 wave superconductors are put in close proximity to tis , great attention has recently been focused on this kind of superconducting hybrid system.@xcite one phenomenon that also appears in superconducting hybrid systems is andreev reflection ( ar ) , which is an important transport process found about fifty years ago.@xcite ar occurs near the interface of a conductor and a superconductor : an electron incident from the metallic side is reflected as a hole while a cooper pair is created in the superconductor .
recently , it has been reported theoretically that ar in ti - superconductor ( ti - s ) hybrid junctions should manifest quantized characteristics due to topologically protected helical edge states@xcite and thus may be used as a powerful method to probe helical edge modes in ti system.@xcite more intriguingly , it is anticipated that the promising existence of mfs in josephson junctions with a superconductor - ti - superconductor ( s - ti - s ) structure will make the josephson supercurrent oscillate with a period of @xmath2-twice that of conventional junctions.@xcite theoretical proposals focus on correlated phase measurements to prove the existence of mfs in these hybrid systems .
this means that the first crucial thing to do is to study the behavior of the helical edge or surface states in close proximity to a superconductor . to date
, many experimental groups have observed proximity - induced superconductivity in the surface or edge states of tis.@xcite in particular , proximity - induced superconductivity in the quantum spin hall edge of hgte / cdte qws has been reported , and clear superconducting quantum interference patterns due to the superconducting helical edge states of tis were obtained independently by two experimental groups.@xcite following the dynes - fulton analysis process , the supercurrent as a function of transverse position in a s - ti - s junction was obtained and found to be localized on two edges of the junction . however , one large puzzle remained in the discussion in ref .
[ ] ; namely , that the supercurrent was localized on the two edges of the junction even though the system still had measurable resistance due to residual bulk states , which are supposed to spread widely in whole sample .
the dynes - fulton approach was discussed in very recent theoretical work,@xcite where it was pointed out the space - distributed supercurrent obtained through the frourier transformation of the dynes - fulton process is not quantitatively accurate and sometimes deviates significantly from the correct value .
therefore , it is very important to perform a direct numerical simulation on this s - ti - s hybrid junction and clearly verify the existence of proximity - induced superconductivity in the helical edge states of quantum spin hall systems , and also present reasonable explanations of the experimental results .
that is the purpose of the present work .
the rest of this paper is organized as follows : in sec .
[ sec : models ] , we introduce the hamiltonian of superconducting leads and hgte / cdte qws , present the formula for the current through the superconducting leads , as well as for the local current through the cross section of hgte / cdte qws .
the numerical results are discussed in sec .
[ sec : discussions ] .
finally , a brief summary is given in sec .
[ sec : conclusions ] .
[ sec : hamiltonians ] the two - dimensional ti phase has been experimentally realized in the hgte / cdte@xcite and inas / gasb@xcite qws . in order to easily compare our numerical simulation to the experimental results in refs .
[ ] and [ ] , here we shall adopt the parameters appropriate for hgte / cdte qws in the following calculations .
the results should be also qualitatively correct for other 2d ti systems , e.g. inas / gasb qws .
we consider a superconductor - hgte / cdte qw - superconductor josephson junction as shown in fig .
[ figure1 ] .
this hybrid device is described by the hamiltonian @xmath3 , where @xmath4 , @xmath5 , and @xmath6 are the hamiltonians of the hgte / cdte qws , superconducting leads , and the coupling between them , respectively . in the nambu representation , the tight binding hamiltonian @xmath4 of the hgte / cdte qws
is represented as : @xmath7 where @xmath8 is the site index , @xmath9 or @xmath10 represents the unit vector along @xmath11 or @xmath12 direction , @xmath13 , and @xmath14 are annihilation operators of electrons at the site @xmath15 in the states @xmath16 , @xmath17 , @xmath18 , and @xmath19 . in eq.([m2 ] ) , @xmath20{ll } h_{\mathbf{i},\mathbf{i}/\mathbf{i},\mathbf{i}+\hat{\mathbf{r } } } & \mathbf{0}\\ \mathbf{0 } & -h_{\mathbf{i},\mathbf{i}/\mathbf{i},\mathbf{i}+\hat{\mathbf{r } } } \end{array } \right ) $ ] and @xmath21{ll } e^{i\phi_{\mathbf{i},\mathbf{i}+\hat{\mathbf{r } } } } & \mathbf{0}\\ \mathbf{0 } & e^{-i\phi_{\mathbf{i},\mathbf{i}+\hat{\mathbf{r } } } } \end{array } \right ) $ ] are @xmath22 matrices , where @xmath23{ll } e_{s } & 0\\ 0 & e_{p } \end{array } \right ) $ ] , @xmath24{ll } t_{ss } & -it_{sp}\\ -it_{sp } & t_{pp}% \end{array } \right ) $ ] , and @xmath25{ll } t_{ss } & -t_{sp}\\ t_{sp } & t_{pp } \end{array } \right ) $ ] , with @xmath26 , @xmath27 and @xmath28 .
adopting a definite gauge for the magnetic vector potential ( @xmath29 with @xmath30 ) , the additional hopping phase arising from the external magnetic field is @xmath31 and @xmath32 . here , @xmath33 is the fermi energy ( pinned by the superconductor condensate ) , @xmath34 is the lattice constant , and @xmath35 , @xmath36 , @xmath37 , @xmath38 , and @xmath39 are system parameters which can be experimentally controlled .
the hamiltonian @xmath5 of the two - dimensional superconducting leads is : @xmath40 where the sum over @xmath41 refers to the left and right superconducting leads and @xmath42 ( @xmath43 ) is the creation ( annihilation ) operator in the superconducting leads for electrons with the momentum @xmath44 . here we consider a general @xmath1-wave superconductor ; @xmath45 is the superconductor gap , and @xmath46 represents the superconducting phase of the left or right superconducting lead . noted that , it is straightforward to compute all momentum - space green s functions , as well as the self - energy functions of the superconducting leads , using the superconducting hamiltonian given above . to simplify use of the landauer - btiker formula in real space
, it necessary to transform all functions into the form appropriate to the tight - binding approximation .
following the same method used in previous work,@xcite we show the result in appendix a. then , the coupling hamiltonian @xmath6 becomes : @xmath47 , where the operator @xmath48 and @xmath49{llll}% t_{{\scriptscriptstyle{\mathbf{s}}}s } & t_{{\scriptscriptstyle{\mathbf{s}}}p } & 0 & 0\\ 0 & 0 & -t_{{\scriptscriptstyle{\mathbf{s}}}s } & -t_{{\scriptscriptstyle{\mathbf{s}}}p}% \end{array } \right ) .$ ] here the parameters @xmath50 and @xmath51 are the coupling strengths between the superconductors and the hgte / cdte qws , which depends on the interface contact potential and , experimentally , the quality of the coupling .
the supercurrent through superconducting leads is carried by cooper pairs .
however , one does not have to introduce the creation operator for the cooper pair in order to calculate the current through the superconducting lead .
generally , the supercurrent from the left or right superconducting lead to the central device can be calculated from the evolution of the number operator of the electrons in the superconducting lead : @xmath52\bigg\rangle\nonumber\\ & & = \frac{ie}{\hbar}\sum_{\alpha,\mathbf{i}_y,\sigma } [ t_{{\scriptscriptstyle{\mathbf{s}}}\alpha}\langle a^{\dagger}_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \sigma}c_{\alpha\mathbf{i}_y\sigma}\rangle - t^*_{{\scriptscriptstyle{\mathbf{s}}}\alpha}\langle c^{\dagger}_{\alpha\mathbf{i}_y\sigma}a_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \sigma}\rangle]\nonumber\\ & & = \frac{e}{\hbar}\sum_{\alpha,\mathbf{i}_y } tr[t_{{\scriptscriptstyle{\mathbf{s}}}\alpha}g^{<}_{\alpha { \scriptscriptstyle{\mathbf{s}}}}(t , t;\mathbf{i}_y,\mathbf{i}_y)-t^*_{{\scriptscriptstyle{\mathbf{s}}}\alpha}g^ { < } _ { { \scriptscriptstyle{\mathbf{s}}}\alpha } ( t , t;\mathbf{i}_y,\mathbf{i}_y)]\nonumber\\ & & = \frac{e}{\hbar } tr[\mathbf{\gamma_z}\mathbf{g}^{<}_{{\scriptscriptstyle{c\mathbf{s}}}}(t , t){\bf t}_{{\scriptscriptstyle{\mathbf{s}c}}}-\mathbf{\gamma_z } { \bf t}^\dagger_{{\scriptscriptstyle{\mathbf{s}c } } } \mathbf{g}^ { < } _ { { \scriptscriptstyle{\mathbf{s}c}}}(t , t)],\end{aligned}\ ] ] where particle number operator @xmath53 and @xmath54 . in order to obtain the green s function in the above formula , the contour - ordering green s functions @xmath55 is defined in the nambu representation : @xmath56,\\ & & \mathbf{g}_{{\scriptscriptstyle{\mathbf{s}c}}}(\tau,\tau';\mathbf{i}_y,\mathbf{i}'_y)=\nonumber\\ & & -i\left [ \begin{array}{ll } \big\langle \mathcal{t } a_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \uparrow}(\tau)\psi^{\dagger}_{\mathbf{i}'_y\uparrow}(\tau')\big\rangle & \big\langle \mathcal{t } a_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \downarrow}(\tau)\psi_{\mathbf{i}'_y\uparrow}(\tau')\big\rangle \\ \big\langle \mathcal{t } a^{\dagger}_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \uparrow}(\tau)\psi^{\dagger}_{\mathbf{i}'_y\downarrow}(\tau')\big\rangle & \big\langle \mathcal{t } a^{\dagger}_{{\scriptscriptstyle{\mathbf{s } } } \mathbf{i}_y \downarrow}(\tau)\psi_{\mathbf{i}'_y\downarrow}(\tau')\big\rangle \end{array}\right],\end{aligned}\ ] ] where the contour - ordering operator , @xmath57 , orders all operators following it along the contour time @xmath58 loop , and @xmath59 . in this case
, a general relation for the contour - ordered green s function @xmath60 can be derived rather easily , either using the equation of motion technique or with a direct expansion of the scattering matrix , as the follows : @xmath61 here @xmath62 and @xmath63 represent the contour - ordered green s functions of the central region and the decoupled superconducting lead in the nambu representation , respectively . using the standard analytic continuation
formula@xcite @xmath64 $ ] , the lesser green s function in the nambu representation is found to be @xmath65 similarly , @xmath66 can be also obtained .
then , substituting the above expression for @xmath67 into eq .
( [ current1 ] ) , the current is @xmath68 + h.c.\nonumber\\ & & = \frac{e}{\hbar}\int dt_1 tr[\mathbf{\gamma_z}\mathbf{g}^r_{\scriptscriptstyle{cc}}(t , t_1 ) \mathbf{\sigma}^<_{{\scriptscriptstyle{\mathbf{ss}}}}(t_1,t)\nonumber\\ & \ & + \mathbf{\gamma_z}\mathbf{g}^<_{\scriptscriptstyle{cc}}(t , t_1 ) \mathbf{\sigma}^a_{{\scriptscriptstyle{\mathbf{ss}}}}(t_1,t ) ] + h.c.,\end{aligned}\ ] ] where @xmath69 represents the self - energy function of the supercondcuting lead . in the case of zero bias voltage , because both the green s function @xmath70 and the self - energy function @xmath71 become functions of @xmath72,@xcite eq .
( [ current2 ] ) can be simplified as : @xmath73.\end{aligned}\ ] ] the self - energy functions @xmath74 , @xmath75 and @xmath76 for the superconducting leads can be calculated as in refs .
[ ] . for the sake of completeness , surface green s functions and retarded or lesser self - energy function of superconducting leads are given in appendix a. the green s function of the central region can be calculated from @xmath77 where @xmath78 is the hamiltonian of the central ti region as shown in fig.[figure1](a ) ( light blue region ) and @xmath79 represents the linewidth function of states , which is always chosen to be an infinitesimal quantity for clean samples . consequently , the current through the left or right superconducting lead can be obtained straightforwardly using eq .
( [ current3 ] ) . in the following numerical calculations ,
we choose parameters appropriate for the realistic materials:@xcite ( 1 ) the hgte / cdte qw parameters are @xmath80 , @xmath81 , @xmath82 , and @xmath83 ; ( 2 ) the superconductor parameters are the gap energy @xmath84 , the superconducting phases @xmath85 and @xmath86 . the lattice constant @xmath34 is set to @xmath87 and the ti - s coupling strengths are taken to be @xmath88 . besides , it is worth pointing out that actually we do not need the full green s function of the whole central region , and a recursive algorithm about how to obtain parts of retarded green s function , used in eqs .
( [ current3 ] ) , ( [ lcurrent5 ] ) and ( [ distributionofcooperpair1 ] ) , is briefly introduced in appendix b. in order to check the real - space distribution of the supercurrent , it is necessary to derive a local current formula perpendicular to the cross section of the hgte / cdte qws . using the non - equilibrium green s function method , the current flowing through site @xmath15 of the central region
can be written as@xcite @xmath89\bigg\rangle\nonumber\\ & & = \frac{ie}{\hbar}\sum_{\alpha,\beta,\hat{\mathbf{r } } } [ h_{\mathbf{i}\bf i+\hat{\mathbf{r}},\alpha\beta } \langle c^{\dagger}_{\alpha \mathbf{i } \uparrow}c_{\beta{\bf i}+\hat{\mathbf{r}}\uparrow}\rangle - h^*_{\mathbf{i}\mathbf{i}+\hat{\mathbf{r}},\alpha\beta } \langle c^\dagger_{\beta{\bf i}+\hat{\mathbf{r}}\uparrow}c_{\alpha \mathbf{i } \uparrow}\rangle]\nonumber\\ & & = \frac{e}{\hbar}\sum_{\hat{\mathbf{r } } } tr[h_{\mathbf{i}\mathbf{i}+\hat{\mathbf{r}}}\mathbf{g}^{<}_{\scriptscriptstyle{cc},\uparrow\uparrow}(t , t;{\bf
i+\hat{\mathbf{r}}},{\bf i } ) + h.c.]\nonumber\\ & & = \frac{e}{\hbar}\sum_{\hat{\mathbf{r } } } tr[h_{\mathbf{i}\mathbf{i}+\hat{\mathbf{r}}}\mathbf{g}^{<}_{\scriptscriptstyle{cc}}(t ,
t;{\bf i+\hat{\mathbf{r}}},{\bf i } ) + h.c.]_{\uparrow\uparrow}\nonumber\\ & & = \frac{2e}{\hbar } re\sum_{\hat{\mathbf{r } } } tr[h_{\mathbf{i}\mathbf{i}+\hat{\mathbf{r}}}\mathbf{g}^{<}_{\scriptscriptstyle{cc}}(t ,
t;{\bf i+\hat{\mathbf{r}}},{\bf i})]_{\uparrow\uparrow}\end{aligned}\ ] ] and similarly in the nambu representation , @xmath90_{\downarrow\downarrow}.\end{aligned}\ ] ] here , the hamiltonian matrices are the same to those defined in the previous section and the keldysh lesser green s function of the central region is defined as : @xmath91 . the ribbon width is @xmath92 and the effective mass is @xmath93 . ]
if we set the voltage of the superconducting leads to zero ( @xmath94 ) , it can be shown that all green s functions of the central region depend only on the time difference ; namely @xmath95.@xcite after taking the fourier transform of the lesser green s function , the local current between neighboring sites @xmath15 and @xmath96 can be calculated from the formula @xmath97\big\}\nonumber\\ & = & \frac{2e^2}{h}re\int^{\infty}_{-\infty}de\ tr[\mathbf{\gamma}_zh_{\mathbf{i}\mathbf{j}}\mathbf{g}_{\scriptscriptstyle{cc}}^{<}(e;\mathbf{j},\mathbf{i})].\end{aligned}\ ] ] by applying the keldysh equation @xmath98,@xcite eq .
( [ lcurrent4 ] ) can be simplified to @xmath99_{\mathbf{j}\mathbf{i}}\big\},\end{aligned}\ ] ] where the linewidth functions of superconducting leads are @xmath100 , and the fermi energies of the left and right superconducting leads are set to zero , and thus @xmath101 below / above the fermi energy .
thus , given that the retarded and advanced green s functions are known , the local current can be readily calculated using the above formula .
in addition , the local cooper pair distribution can be approximately obtained by the computing superconductor order parameter , @xmath102_{\uparrow,\downarrow}\big\ } \nonumber\\ & & = re\int^{\infty}_{-\infty}\frac{de}{2\pi } tr\big\{[\mathbf{g}_{\scriptscriptstyle{cc}}^r(\mathbf{\gamma}_lf_l+\mathbf{\gamma}_rf_r ) \mathbf{g}_{\scriptscriptstyle{cc}}^a]_{\mathbf{i}\uparrow,\mathbf{i}\downarrow}\big\},\nonumber\\ & & = re\int^{0}_{-\infty}\frac{de}{2\pi } tr\big\{[\mathbf{g}_{\scriptscriptstyle{cc}}^r ( \mathbf{\gamma}_l+\mathbf{\gamma}_r ) \mathbf{g}_{\scriptscriptstyle{cc}}^a]_{\mathbf{i}\uparrow,\mathbf{i}\downarrow}\big\},\end{aligned}\ ] ] where @xmath103 has been adopted .
[ sec : discussions ] [ sec : discussions ] in order to facilitate comparison of our results with previous experimental measurements , we have performed calculations for essentially the same device as used in the experiments . as shown in fig .
[ figure1 ] , the josephson junction consists of the central region of hgte / cdte qws and two superconducting leads . applying a perpendicular magnetic field ,
a supercurrent interference pattern is expected to be observed as the magnetic strength is varied . in order to discuss the following calculation results explicitly ,
the energy dispersion in the absence and presence of external magnetic field are given in fig .
1(b ) and 1(c ) , respectively .
note that the ribbon width is set to @xmath92 , and the effective mass to @xmath104 which corresponds to the hgte / cdte qw being a ti .
in addition , the hopping phase arising from the external magnetic field in fig .
1(c ) is set to @xmath105 .
one of the most intriguing observations reported in one recent experiment@xcite is that although there is still residual resistance from bulk states , it seems that only topologically protected edge states contribute to the supercurrent interference pattern while the bulk states appear to be ` frozen ' in the transport measurement .
one of the underlying reasons may be the special characteristics of tis , which result in the edge states predominating in the proximity effect to a superconductor , while at the same time suppressing the contribution from the bulk states .
if this supposition is true , ti - superconductor hybrid devices could open a fascinating new chapter in this field ; if not , a reasonable explanation of the experimental results is still necessary . an effective and feasible approach to verify this assumption , that the system topology has endowed topologically protected edge states with some special characteristics but
does not do the same for bulk states , is to examine the real - space distribution of cooper pairs along the cross section through the central ti sample .
when @xmath93 and @xmath106 as shown in fig .
[ figure2](a ) , cooper pairs localize mainly near the two edges of the josephson junction .
this distribution origins with the topologically protected edge states being confined on the sample edges .
as the fermi energy is shifted to the conduction band edge , @xmath107 , cooper pairs spread over the whole cross section of the qws , which indicates that the bulk states become dominant . when qws are tuned to be non - topological , @xmath108 , the occupation number of cooper pairs decreases to zero for @xmath106 because there is no state within the band gap .
similarly , cooper pairs spread over all the whole cross section of qws when the fermi energy is again tuned to the edge of conductance band , @xmath107 in fig .
[ figure2](a ) .
there is no change in essence when an external magnetic field is applied , as shown in fig . [ figure2](b ) . ) , respectively .
noted that @xmath109 represents the @xmath12 index of the site located at the middle cross section of hgte / cdte qws .
the ribbon width is @xmath110 and the length of the central region is @xmath111 . ] from fig .
[ figure2 ] , it can be concluded that there is no essential difference between topologically - protected edge states and bulk states in proximity to a superconductor .
even though topologically - protected edge states , because of their high mobility , could have some advantages in proximity to a superconductor , the bulk states are not forbidden to contribute and can experience the proximity effect .
this conclusion is still correct when an external magnetic field is turned on .
this does not resolve the problem of why , in the experiment@xcite mentioned above , only topologically - protected edge states appear to contribute to the supercurrent interference pattern even though there are residual bulk states present . to understand the connection between our results and the experimental results,@xcite we present interference patterns for hgte / cdte hybrid system from the perspective of numerical simulation .
this will illuminate the underlying physics .
when the fermi energy is set to @xmath106 in fig .
[ figure3](a ) , which means there are only two pairs of topologically - protected edges states confined to the top and bottom boundaries of the central sample , an interference pattern similar to a two - slit interference pattern is observed . in particular , when counting period of the fluctuations in fig.[figure3](a ) , it is found that they do correspond to a change of @xmath112 in the magnetic flux threading the whole area of the central hgte / cdte qws .
when the fermi energy is shifted up to @xmath113 , the perfect two - slit interference pattern is damaged because of the addition of some bulk states , as shown in fig.[figure3](b ) .
note first that the heights of the interference peaks decrease gradually as the threading magnetic flux increases .
secondly , the modulation period of the interference pattern is not exactly equal to the change of the magnetic flux @xmath112 threading the whole area of the central hgte / cdte qws , and in particular the width of the central lobe clearly increases in size . as the fermi energy is further moved up to @xmath114 , an interference pattern similar to a single - slit frauhofer interference pattern
is obtained , as seen in fig .
in this case , many bulk states mask the role of the two pairs of edge states in the interference . besides
, it was reported in ref .
[ ] that both the temperature and the length of josephson junction can influence the shape of interference pattern obviously . , @xmath113 and @xmath114 , respectively .
the ribbon width is @xmath115 , the length of the central region is @xmath116 , and the effective mass is @xmath93 . note that @xmath117 represents the additional hopping phase due to the external magnetic field , which is defined in sec .
[ sec : hamiltonians ] . ] comparing our theoretical results to the experimental ones reported in ref .
[ ] , it is easy to see that there are always residual bulk states even for the perfect ` double - slit interference pattern ' , namely fig .
2(c ) in ref .
[ ] , while the authors thought there were only edge states .
the question mentioned in the introduction and elsewhere therefore arises as to why no bulk states and only edge states are found when transforming the interference pattern into a real - space current distribution along the cross section of the hete / cdte qws using the dynes - fulton method.@xcite one possible reason , which has been discussed in ref .
[ ] , is that the lack of bulk states may arise from the highly simplifying assumptions used in the dynes - fulton analysis of ref .
[ ] . however , we want to present an alternate explanation of the experimental results . in fig .
[ figure4 ] , we plot the distribution of cooper pairs and supercurrent along the cross section as a function of the sample width . in order to compare our theoretical results conveniently with those of ref .
[ ] , here we maintain a system with a definite resistance ( about @xmath118)@xcite by adjusting the fermi energy appropriately no matter what the width of sample is in fig .
[ figure4 ] .
it can be clearly seen from fig .
[ figure4 ] that both the distribution of cooper pairs and the magnitude of the local supercurrent at the sample center decrease gradually as the sample width increases .
when the sample width is increased to @xmath119 , the contribution of the bulk states to cooper pair or supercurrent is so small as to be negligible .
however , the contribution of the edge states undergoes almost no change as the sample width changes from @xmath120 to @xmath119 , because the topological edge states are located in a finite range with characteristic decay length @xmath121.@xcite moreover , the interference pattern is determined by the square of the local supercurrent density .
therefore , one clear result is that only cooper pairs or supercurrent confined near the top and bottom boundaries of the ti are detectable experimentally .
in fact , this was just the case in ref .
[ ] , where the sample width was @xmath122 and the residual resistance from bulk states was approximately @xmath123 . because the wave functions of these residual bulks states spread over the whole cross - section of the sample and thus are much weaker than those of two pairs of edge states which are always confined at the sample boundaries , to detect the contribution of finite residual bulk states to the supercurrent becomes very difficult experimentally .
that is why only induced superconductivity in the quantum spin hall edge states is observed in ref .
[ ] . through changing the ti sample width and tuning the fermi energy by gate voltage , the results in fig .
[ figure4 ] can be easily checked for experiments .
finally , it should be noted that in order to obtain interference patterns similar to those seen in the experiment,@xcite a relative large linewidth , @xmath124 , had to be used for every lattice site of the central ti sample .
it is found that the interference pattern deviates significantly from that reported in the experiment@xcite if the linewidth @xmath79 is set to be small .
the reason is twofold : ( 1 ) the finite width of the central sample adopted in our simulation leads to discrete energy levels , therefore a relative large linewidth @xmath125 is needed to smear discrete energy levels and makes them look like continuum energy spectrum ; ( 2 ) the experimental samples are somewhat disordered , which leads to broadening of states , however this supposition needs to be further confirmed experimentally . since residual bulk states and dephasing are two significant factors influencing the robustness of the edge states of tis , production of a clean , high quality ti sample is a long sought goal of this field . at present ,
residual bulk states are still always observed , no matter whether in 2d ti samples or in 3d ones . how to tune the fermi energy completely away from the conduction / valence band edges into the band gap
is still a big challenge before us . , and @xmath109 represents the @xmath12 index of the site locating at the middle cross section of the hgte / cdte qws
. noted that a definite resistance ( about @xmath126 ) are preserved by choosing a suitable fermi energy no matter what the width of sample is .
the length of the central region is taken as @xmath116,and the effective mass is @xmath93 . ]
[ sec : conclusions ] we have studied the transport properties of the superconductor - hete / cdte qws - superconductor josephson junction in the presence of an applied magnetic field . it was found that the proximity effect can induce superconductivity for both edge and bulk states . compared to bulk states , topologically - protected edge states do not possess any obvious advantages arising from being in close proximity to a superconductor .
the fact that experiments have only observed induced superconductivity only in quantum spin hall edge states is just due to the effect of size : the magnitude of the wave functions from bulk states become very small when these residual bulk states spread over the relatively large cross section of the sample ( @xmath122 ) . from the results of this work
, one meaningful thing is that even though there is limited residual bulk resistance for hgte / cdte qws , it does not affect the superconducting quantum interference pattern for s - ti - s josephson junction .
however , on the other hand the experimental results tell us that residual bulk states are still present even after many years attempts in experiment , and the fermi energy for hete / cdte qws does not lie inside the expected bulk gap . in this sense
, there is still a long way to go to realize a perfect quantum spin hall system , which is a prerequisite for obtaining useful topological devices for technological purposes .
we gratefully acknowledge the financial support from nsf - china under grant nos . 11204065(j.t.s . ) , 11474085(j.t.s . ) , 11504008(h.w.l . ) , 11574245(j.l . ) , 11274364(q.f.s . ) , 11574007(q.f.s . ) , and 11504008(x.c.x . ) , and nbrp of china under grand nos .
2012cb921303 and 2015cb921102 .
is also supported by nsf - hebei province under grant no .
in this appendix , we intend to obtain the surface green s function @xmath127 of the two - dimensional superconducting lead and the self - energies @xmath128 associated with the superconducting lead in real space .
note that a detailed derivation can be found in a published paper.@xcite using the hamiltonian of the superconducting lead introduced in sec .
[ sec : hamiltonians ] , the bulk green s function @xmath127 of the superconductor can be obtained readily in momentum space : @xmath129 where @xmath130 and @xmath131 .
here , the space of the above matrix refers to the wavevectors @xmath132 and @xmath133 . using the fourier transformation ,
we can write the surface green s function of two - dimensional superconducting leads in real space@xcite @xmath134 where @xmath135 is the bessel function of the first kind , @xmath136 for @xmath137 and @xmath138 for @xmath139 .
in addition , the density of electron states @xmath140 is assumed to be independent of the superconducting spectrum @xmath141 , which makes the detailed expression for @xmath141 unnecessary.@xcite straightforwardly , using the surface green s function @xmath127 for the superconducting lead , the self - energies @xmath142 take the form : @xmath143(\mathbf{\sigma}^r_{{\scriptscriptstyle{s}},nm}-\mathbf{\sigma}^a_{{\scriptscriptstyle{s}},nm}),\nonumber\end{aligned}\ ] ] where @xmath144 represent the coupling strength between the supperconducting leads and the central sample @xmath50 or @xmath51 , and @xmath145 is the fermi function .
if we suppose now that the couplings of the superconducting leads to the @xmath1 and @xmath0 orbitals in the hgte / cdte qws are identical , the final self energy takes the form @xmath146 .
in this appendix , we present a numerical method about how to compute the retarded green s function of the central region in eq .
( [ current3 ] ) or ( [ lcurrent5 ] ) .
a straightforward method to obtain the retarded green s function of the central region is via a direct inversion of large matrix , as shown in eq .
( [ tigreenfunction ] ) . however , the direct inversion of large matrix is greatly time - consuming for computers . in order to obtain a current by landauer - bttiker formula , in fact we only need parts of the retarded green s function of the central region , e.g. that of the left - most layer contacting with the left lead , that of the right - most layer contacting with the right lead , that between the left - most layer and the right - most layer , and that of one specific layer in the central region , which can be represented as @xmath147 , @xmath148 , @xmath149 or @xmath150 , and @xmath151 , respectively .
using dyson equation , it can be easily proved that hopping green s function between the left - most layer and the right - most layer as well as that of the right - most layer can be obtained by the following recursive algorithm : @xmath152 where @xmath153 , @xmath154 , @xmath155 represents the hamiltonian of an arbitrary layer , the hopping hamiltonians to the right neighboring layer and the left neighboring layer , respectively . @xmath156 and @xmath157 are self energies from left lead and right lead .
if there are many sites for one layer , all quantities above will become a matrix with the same dimension to the site number in one layers .
it should be pointed out explicitly that @xmath149 and @xmath148 are , namely , green s function from the left - most layer to the right - most one and that of the right - most layer contacting with the left lead , respectively .
similarly , we can obtain @xmath150 and @xmath147 if @xmath156 and @xmath157 commute each other , as well as @xmath154 and @xmath158 . from the recursive algorithm above
, the green s functions of one specific layer in the central region can be obtained as follows : @xmath159 here , @xmath151 , @xmath160 , and @xmath161 are the green s function of the @xmath162-th layer , the hopping green s function from the left - most layer to the @xmath162-th layer , and the hopping green s function from the right - most layer to the @xmath162-th layer , respectively .
similarly , the hopping green s functions from the @xmath162-th layer to the left - most layer and to the right - most layer can be also obtained through this recursive method .
m. knig , s. wiedmann , c. brne , a. roth , hartmut buhmann , laurens w. molenkamp , x .-
qi and s .- c .
zhang , science * 318 * , 766 ( 2007 ) ; m. knig , h. buhmann , l. w. molenkamp , t. l. hughes , c .- x .
qi , and s .- c .
zhang , j. phys .
. jpn . * 77 * , 031007 ( 2008 ) . because the maximum normal device resistance is about @xmath126 in ref .
, it could be simply regarded as to consist of two pairs of helical edge states and six residual bulk states .
in fact , the discussion about fig . [ figure4 ] refers to when the contribution of the topological edge states outweighs that of the residual bulk states in the superconducting quantum interference as the sample width increases .
|
using non - equilibrium green s functions , we studied numerically the transport properties of a josephson junction , superconductor - topological insulator - superconductor hybrid system .
our numerical calculation shows first that proximity - induced superconductivity is indeed observed in the edge states of a topological insulator adjoining two superconducting leads and second that the special characteristics of topological insulators endow the edge states with an enhanced proximity effect with a superconductor but do not forbid the bulk states to do the same . in a size - dependent analysis of the local current
, it was found that a few residual bulk states can lead to measurable resistance , whereas because these bulk states spread over the whole sample , their contribution to the interference pattern is insignificant when the sample size is in the micrometer range .
based on these numerical results , it is concluded that the apparent disappearance of residual bulk states in the superconducting interference process as described in ref .
[ ] is just due to the effects of size : the contribution of the topological edge states outweighs that of the residual bulk states .
|
the nature of anomalous velocity discovered during @xmath221 cm surveys of the galaxy has been the subject of much debate in the past three decades .
there appear to be at least four reasonably distinct classes of this gas , namely : ( 1 ) localized outflows and subsequent inflows associated with massive star formation which are commonly referred to as a `` galactic fountain '' ( shapiro & field , 1976 ; bregman , 1980 ) .
( 2 ) the stream of tidal debris from the interaction of the galaxy with the magellanic clouds is another major source of anomalous velocity .
recent imaging ( putman & gibson , 1999 ) has begun to reveal the true extent of this debris system , and allow it to be distinguished from other components .
( 3 ) the extended hvc complexes ( named a , c , h , m , ) which span ten s of degrees on the sky and have recently been determined to have both nearby distances of about 10 kpc ( van woerden et al .
1999 ) and rather low ( @xmath30.1 solar ) metal abundance ( wakker et al . 1999 ) .
these diffuse structures appear to be currently merging with the galaxy and each account for some 10@xmath4 m@xmath5 of fresh gas . and finally ( 4 ) the system of compact high velocity clouds ( chvcs ) cataloged by braun & burton ( 1999 ) .
the chvcs form a distinct class of compact , high - contrast emission features at anomalous velocity , which can be readily distinguished from the more diffuse components in the recent high quality surveys ; the leiden / dwingeloo survey in the north ( hartmann & burton 1997 ) and the parkes multibeam survey in the south ( see putman & gibson 1999 ) .
their average angular size is only 1 degree fwhm and the total velocity width 30 fwhm .
the 65 confirmed members of the chvc class share a well - defined kinematic pattern .
a global search for the reference frame that minimizes the line - of - sight velocity dispersion of the system , returns the local group standard of rest with high confidence . in this system
the velocity dispersion of the population is only 70 , although it is in - falling toward the local group barycenter at about 100 .
these properties suggest that the objects are : ( 1 ) associated with the local group rather than the galaxy as such , ( 2 ) that they are likely to reside at quite substantial distances and ( 3 ) have as yet had little tidal interaction with the more massive local group members .
a program of high resolution imaging of the chvcs has begun to provide further insights into the nature of these sources .
an initial sub - sample of six chvcs was imaged with the upgraded westerbork array ( wsrt ) in december 1998 .
a complete description of those results can be found in braun & burton ( 2000 ) .
the sub - sample was chosen to span a wide range of positions on the sky and in line - of - sight velocity , as well as gas content and linewidth .
an additional eight sources have been imaged with the wsrt during the summer of 1999 , specifically targeting objects suspected of having very bright and narrow emission features .
a sample of ten chvcs have also been imaged with the upgraded arecibo telescope in november of 1999 .
in addition to fully - sampled images of the central square degree , a very deep cross - cut of two degree length was obtained for each source reaching an rms sensitivity of about @xmath6 = 10@xmath7 @xmath8 at 10 resolution .
full accounts of these recent results will be given elsewhere .
all of the imaged sources share a number of common properties .
each source contains between one and ten compact cores ( of 1 to 10 arcmin extent ) embedded in a diffuse halo of about 1 degree diameter .
an average of 40% of the line flux arises in the compact cores ( although the range varies between 1 and 60% ) .
the peak column density in these cores varies between 10@xmath9 and 10@xmath10 @xmath8 .
the fractional surface area of each source which exceeds a column density of 5@xmath11 is only 15% on average , although this also varies greatly from source to source .
the remainder of the emission arises in a diffuse halo which reaches peak column densities of @xmath12@xmath13 @xmath8 but declines smoothly with an exponential profile to values of @xmath14 @xmath8 or less ( as illustrated in fig.2 ) .
the resolved linewidths toward the core components ( with 1 arcmin resolution ) are typically only 5 fwhm ( but range from as little as 2 to as much as 15 ) .
all of the core linewidths are sufficiently narrow that it is possible to unambiguously identify these with the cool neutral medium phase of , with typical kinetic temperature of 100 k. the diffuse halos , on the other hand , all have resolved linewidths ( with 3 arcmin resolution ) of 2125 fwhm , as seen in fig.2d .
the one - to - one correspondence of halo emission with linewidths equal to or exceeding the thermal linewidth of 8000 k gas ( 21 fwhm ) provides an unambiguous identification of these halos with the warm neutral medium phase of .
a shielding column of wnm is one prerequisite for the condensation of cnm cores in calculations of thermodynamics ( eg .
wolfire et al .
the second prerequisite for cnm condensation is a sufficiently high thermal pressure , in excess of about p / k = 100 @xmath15k .
both of these aspects are illustrated in the phase diagram shown in fig.1d .
an adequate shielding column of wnm is clearly detected in the chvcs , as just discussed above .
we will return to the question of an adequate thermal pressure below .
another general result is that the cnm core components are dynamically decoupled from their wnm halos .
this is illustrated in fig.2 , where a substantial velocity gradient of some 30 is seen within the major core component of chvc158@xmath1639@xmath16285 .
such large gradients are not observed in the halo gas , which instead is always centered near the systemic velocity of the system .
this dichotomy suggests that while the cores may represent flattened rotating disks , the halos are more nearly spherical distributions .
one of the objects imaged with the wsrt , chvc125 + 41@xmath16207 , has proven particularly interesting for several reasons .
this object has a number of opaque core components of only @xmath17 = 90angular size , a brightness temperature in emission of 75 k , and linewidths so narrow ( less than 2 fwhm ) that the kinetic temperature ( t@xmath18 = 85 k ) , opacity ( @xmath19 = 2 ) and column density ( = 10@xmath10@xmath8 ) , can be accurately derived .
a good estimate of the volume density ( n@xmath20 = 2@xmath11 @xmath15 ) for this object has been made by modeling the thermodynamics ( wolfire et al .
comm . ) . with these quantities in hand and only the assumption of crude spherical symmetry for the cores ,
it is possible to estimate the source distance from d = /(n@xmath20@xmath17 ) = 600@xmath1300 kpc .
the same object has also allowed the first measurement of chvc metallicity , since the bright uv source mrk 205 is located behind its diffuse halo .
bowen , blades and pettini ( 1995 ) detect unsaturated mgii absorption , with 0.15 ew at the chvc velocity , while we detect a neutral column = 5@xmath2110@xmath22 @xmath8 at the same position .
the implied metal abundance is about 0.05 solar .
although many of the chvcs have only one or two core components , a handful of these objects might best be called chvc clusters .
as many as ten cores are seen in close proximity ( within 3050 arcmin ) , each with its own systemic velocity and internal kinematics ( which we address below ) but sharing the same diffuse halo of shielding gas .
the most extreme case is that of chvc115 + 13@xmath16275 , where the cores have relative velocities as high as 70 and angular separations as large as 30 arcmin . if these are gravitationally bound systems ( and their cool gas content and isolation on the sky make this seem plausible ) then they have a dynamical mass , @xmath23 , and gas mass @xmath24 . if for example the distance were 0.7 mpc , then @xmath25 m@xmath26 and @xmath27 m@xmath26
. the dark to gas mass ratio in this case has the rather substantial value of @xmath28 = 51 , which scales with @xmath29 for other distance estimates .
as already noted above , many of the chvc cores give an indication for organized internal kinematics .
many of the cores have a systematic velocity gradient along the long axis of their elliptical distribution which is very suggestive of rotation in a flattened disk system .
several examples are shown in figs.1 and 2 .
the best - resolved examples of this pattern have been subjected to the standard tilted ring fitting algorithms used in deriving galactic rotation curves .
good solutions for rotation are found ( fig.1c ) which slowly rise over some 6 arcmin to constant values of 15 to 20 .
the shapes of these rotation curves are in very good agreement with those predicted by navarro , frenk & white ( 1997 ) for a cdm halo at 0.7 mpc distance ( as shown in fig.1c ) .
the dark to visible mass ratios in these two cases are @xmath28 = 36 and 29 .
the presence of a significant dark matter component is important not only for understanding the kinematics of the individual cores and the chvc clusters , but also for understanding the thermodynamics . as noted earlier , a significant thermal pressure in excess of about p / k = 100 @xmath15k is required to allow condensation of the cool cores observed in the chvcs ( see fig.1d , upper ) .
the mass density provided within the central few kpc of an nfw halo in this mass range ( fig.1d , lower ) is sufficient to provide the required hydrostatic pressure in combination with the observed warm halo column densities of about 10@xmath9@xmath8 ( see also braun & burton 2000 ) .
we are grateful to m.g .
wolfire , a. sternberg , d. hollenbach , and c.f .
mckee for providing the equilibrium temperature curves shown in fig.1d .
the westerbork synthesis radio telescope is operated by the netherlands foundation for research in astronomy , under contract with the netherlands organization for scientific research .
the arecibo observatory is part of the national astronomy and ionosphere center , which is operated by cornell university under a cooperative agreement with the national science foundation .
bowen , d.v .
, blades , j.c .
, & pettini , m. 1995 , , 448 , 662 braun r. , & burton w.b . 1999 , , 341 , 437 braun r. , & burton w.b . 2000 ,
, in press , astro - ph/9912417 bregman j. 1980 , , 236 , 577 hartmann d. , & burton w.b .
1997 , `` atlas of galactic neutral hydrogen '' , cambridge : cambridge university press navarro j.f .
, frenk c.s .
, & white s.d.m . , 1997 , , 490 , 493 putman m.e . , & gibson b.k .
1999 , pasa , 16 , 70 shapiro p.r . , &
field g. 1976 , , 205 , 762 wakker b.p .
, howk j.c . ,
savage b.d . , van woerden h. , tufte s.l . ,
schwarz u.j .
, benjamin r. , reynolds r.j .
, peletier r.f . &
kalberla p.m.w . 1999 ,
nature , 402 , 388 van woerden h. , schwarz u.j . , peletier r.f .
, wakker b.p . , &
kalberla p.m.w .
1999 , nature , 400 , 138 wolfire m.g .
, hollenbach d. , mckee c.f .
, tielens a.g.g.m . , & bakes e.l.o .
1995 , , 443 , 152
|
a distinct sub - class of anomalous velocity emission features has emerged from recent high quality surveys of the local group environment , namely the compact high velocity clouds ( chvcs ) .
a program of high - resolution imaging with the westerbork array and the arecibo telescope has begun to provide many insights into the nature of these objects .
elongated core components with a velocity gradient consistent with rotation ( v@xmath0 15 ) are seen in many objects .
comparison of volume and column densities has allowed the first distance estimates to be made ( 600@xmath1300 kpc ) .
the objects appear to be strongly dark - matter dominated with dark - to - gas mass ratios of 3050 implied if the typical distance is 700 kpc .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
|
the first radiation zeros were discovered as a consequence of a general investigation @xcite of the production of electroweak pairs in hadronic collisions @xmath0 and neutrino reactions @xmath1 addressed to very - high - energy cosmic - ray physics .
the investigation probed the sensitivity of these reactions to the trilinear gauge boson couplings and , at the present time , useful limits on their deviations from gauge theory predictions have been obtained in @xmath2 and hadron collider experiments @xcite .
pronounced dips were found @xcite in the angular distributions for the production of @xmath3 and @xmath4 in the two - body parton and lepton reactions .
subsequent work @xcite by mikaelian , samuel , and sahdev proved that the @xmath3 dips were in fact exact zeros at particular angles , which would be ruined by non - gauge couplings .
the wisconsin brain trust @xcite followed with a general demonstration of the implied amplitude factorization .
since it is so easy to do so , we show how a radiation zero arises in lowest - order radiative charged - scalar fusion @xmath5 whose feynman diagrams lead to the born amplitude @xmath6 with charges @xmath7 , four - momenta @xmath8 , and photon momentum @xmath9 and polarization @xmath10 .
( the trilinear scalar coupling is taken to be unity . ) using momentum conservation , we observe that @xmath11 if all @xmath12 are the same .
this is exactly the same condition found for the @xmath3 amplitude , independent of the spins .
it leads to a zero when @xmath13 for the center - of - mass angle @xmath14 of particle 3 relative to the direction of particle 1 ( or between the photon and particle 2 ) .
this reduces to @xmath15 for @xmath16 ( @xmath17 ) .
faced with the vanishing of the above amplitudes , brodsky asserted that the way to look at these zeros was as the complete destructive interference of classical radiation patterns . following this lead , and making a long story short ( see the longer story with better referencing in @xcite ) , an arbitrary number @xmath18 of external charged particles was considered and a general set of radiation interference theorems were found @xcite . again considering the emission of a photon with momentum @xmath9 , the tree amplitude approximation vanishes , independent of any particle s spin @xmath19 , for common charge - to - light - cone - energy ratios ( the radiation null zone ) , @xmath20 where the @xmath21 particle has electric charge @xmath7 and four - momentum @xmath8 .
all couplings must be prescribed by local gauge theory .
we see why it took so long to discover radiation zeros since the first null zone requirement is that all charges must have the same sign .
fractionally charged quarks and weak bosons are needed in order to get three things : same - sign charges , a process well - approximated by a born amplitude , and a four - particle reaction so the null zone was simple .
while there are zeros associated with any gauge group when the corresponding massless gauge bosons are emitted , in qcd , color charges are averaged or summed over in hadronic reactions . in thinking of the weak bosons themselves ,
electroweak symmetry is broken and nonzero weak - boson masses ruin radiation interference .
what about other photonic zeros ? the zero in electron - electron bremsstrahlung is less interesting .
zeros in electron - quark and quark - antiquark bremsstrahlung require jet identification along with the more complicated phase space @xcite . while we shall say more about other tests , we find ourselves returning again and again to the original @xmath3 zero .
there are various corrections that turn the @xmath3 zero into a dip . theoretically , higher - order ( closed - loop )
corrections will not vanish in the null zone , since the internal loop momenta can not be fixed .
structure function effects , higher order qcd corrections , finite @xmath23 width effects , and photon radiation from the final state lepton line all tend to fill in the dip .
the main complication in the extraction of the @xmath24 distribution in @xmath3 production , however , originates from the finite resolution of the detector and ambiguities in reconstructing the parton center of mass frame .
the ambiguities are associated with the nonobservation of the neutrino arising from @xmath23 decay . identifying the missing transverse momentum , @xmath25 , with the transverse momentum of the neutrino of a given @xmath3 event ,
the unobservable longitudinal neutrino momentum , @xmath26 , and thus the parton center of mass frame , can be reconstructed by imposing the constraint that the neutrino and charged lepton four momenta combine to form the @xmath23 rest mass .
the resulting quadratic equation , in general , has two solutions . in the approximation of a zero @xmath23-decay width ,
one of the two solutions coincides with the true @xmath26 . on an event by event basis , however , it is impossible to tell which of the two solutions is the correct one .
this ambiguity considerably smears out the dip caused by the amplitude zero .
problems associated with the reconstruction of the parton center of mass frame could be avoided by considering hadronic @xmath23 decays . the horrendous qcd background
, however , renders this channel useless .
at present there is only a preliminary study by the cdf collaboration of the @xmath24 distribution @xcite from a partial data sample of the 1992 - 95 tevatron run .
the event rate is still insufficient to make a statistically significant statement about the existence of the radiation zero .
one can at best say that `` there is a hint of gauge zero '' in fig .
[ fig : fourteen2 ] . [ cols="^ " , ] + a)7.cm b )
the radiation zeros are the generalization of the well - known vanishing of classical non - relativistic electric and magnetic dipole radiation occurring for equal charge / mass ratios ( indeed , the low - energy limit of the null zone conditions ) and equal gyro - magnetic g - factors .
the null zone is exactly the same as that for the completely destructive interference of radiation by charge lines ( a classical convection current calculation@xcite ) and is preserved by the fully relativistic quantum born approximation for gauge theories .
stan brodsky has long emphasized the magic of the gyro - magnetic ratio value @xmath27 predicted by gauge theory for spinor and vector particles . only for this value
will born amplitudes have the same null zone as the classical radiation patterns for soft photons . and only for this value
will born amplitudes have good high - energy behavior . in this way we have a connection between the large and small distance scales , with the value @xmath27 as a bridge .
it is great to have this chance to acknowledge stan brodsky s unique leadership and collaboration in these radiation matters .
ub would like to thank s. errede , t. han , n. kauer , g. landsberg , j. ohnemus , r. sobey and d. zeppenfeld for pleasant collaboration and many fruitful discussions .
rb is grateful to k. kowalski , sh .
shvartsman and c. taylor for advice and collaboration through the years .
this research was supported in part by nsf grant phy-9600770 .
r. w. brown and k. o. mikaelian , phys .
* d19 * , 922 ( 1979 ) .
r. w. brown , d. sahdev , and k. o. mikaelian , phys . rev . * d20 * , 1164 ( 1979 ) .
j. ellison and j. wudka , annu .
nucl . part .
sci . * 48 * , 33 ( 1998 ) ; the lep collaborations , cern - ep/99 - 15 ( report , february 1999 ) .
k. o. mikaelian , m. a. samuel , and d. sahdev , phys .
lett . * 43 * , 746 ( 1979 ) .
d. zhu , phys . rev . *
d22 * , 2266 ( 1980 ) ; c. j. goebel , f. halzen , and j. p. leveille , phys
. rev . * d23 * , 2682 ( 1981 ) . r. w. brown,_understanding something about nothing : radiation zeros _ and t. han , _ exact and approximate radiation amplitude zeros - phenomenological aspects _ , february 1995 , international symposium on vector boson self - interactions , aip conference proceedings 350 , eds .
u. baur , s. errede , and t. mller , pp .
261 and 224 .
s. j. brodsky and r. w. brown , phys .
lett . * 49 * , 966 ( 1982 ) ; r. w. brown , k. l. kowalski , and s. j. brodsky , phys . rev . *
d28 * , 624 ( 1983 ) ; r. w. brown and k. l. kowalski , phys . rev . *
d29 * , 2100 ( 1984 ) .
bilchak , j. phys .
phys . * 11 * , 1117 ( 1985 ) ; m. a. doncheski and f. halzen , z. phys .
* c52 * , 673 ( 1992 ) .
d. benjamin , proceedings of the 10th topical workshop on proton - antiproton collider physics , fermilab , may 1995 , aip conference proceedings 357 , eds .
r. raja and j. yoh , p. 370 .
u. baur , s. errede , and g. landsberg , phys .
* d50 * , 1917 ( 1994 ) . c. l. bilchak
, r. w. brown , and j. d. stroughair , phys . rev . *
d29 * , 375 ( 1984 ) .
u. baur , t. han , and j. ohnemus , phys .
rev . * d48 * , 5140 ( 1993 ) .
u. baur , s. errede , and j. ohnemus , phys .
* d48 * , 4103 ( 1993 ) . r. w. brown , m. e. convery and m. a. samuel , phys . rev . *
d49 * , 2290 ( 1994 ) .
u. baur , t. han , n. kauer , r. sobey and d. zeppenfeld , phys .
* d56 * , 140 ( 1997 ) .
u. baur , t. han and j. ohnemus , phys . rev . * d51 * , 3381 ( 1995 ) .
u. baur , t. han and j. ohnemus , phys .
lett . * 72 * , 3941 ( 1994 ) .
w. j. stirling and a. werthenbach , preprint hep - ph/9905341 ; m. heyssler and w. j. stirling , eur .
j. * c4 * , 289 ( 1998 ) .
|
we celebrate two birthdays connected with the radiation zero phenomenon .
first , a striking dip in the theoretical angular distributions of radiative weak - boson production was discovered twenty years ago .
the key experimental interest is that this will not occur in any deviation from the standard model .
second , the classical training of stanley brodsky began sixty years ago , which was instrumental in understanding why theoretical spin - independent radiation zeros appear in almost all born amplitudes for the radiation of photons and gluons and other massless gauge bosons ( but rarely in physical kinematic regions ) . and there are approximate zeros for massive bosons and `` type ii '' zeros that can also be studied .
we discuss how the difficulties in observing the original mikaelian - samuel - sahdev zero finally may be surmounted next year .
|
in many cell types , cell membranes are composed @xcite of a diverse array of lipids , organized as a lipid bilayer , and membrane proteins , which play a central role in most cellular processes .
membrane proteins are rigid compared to the surrounding lipid bilayer @xcite .
thus , the lipid bilayer typically deforms to accommodate membrane proteins and , in particular , the bilayer hydrophobic thickness is compressed or expanded compared to the preferred bilayer thickness in the absence of membrane proteins @xcite .
distinct conformations of a membrane protein generally yield distinct energy costs of protein - induced lipid bilayer deformations . as a result ,
the lipid bilayer can serve as a `` splint '' stabilizing certain protein conformations @xcite and thereby regulate protein function @xcite . in agreement with this general picture ,
experiments have revealed @xcite that , across the kingdoms of life , central biological functions of integral membrane proteins such as ion exchange and signaling are regulated by the mechanical properties of the surrounding lipid bilayer , with the hydrophobic regions of membrane proteins coupling to the hydrophobic regions of lipid bilayers @xcite . in particular , elastic bilayer thickness deformations have been found @xcite to regulate the functions of a diverse range of integral membrane proteins .
cell membranes are crowded with membrane proteins @xcite , with a typical mean center - to - center distance @xmath0 nm between neighboring proteins @xcite . as a result ,
the elastic decay length of protein - induced lipid bilayer thickness deformations @xcite is comparable to the typical edge - to - edge spacing of proteins in cell membranes @xcite , yielding thickness - mediated interactions between membrane proteins @xcite .
for the small protein separations relevant for cell membranes , thickness - mediated interactions between integral membrane proteins can be @xmath1 @xmath2 in magnitude @xcite and , depending on the hydrophobic thickness of neighboring membrane proteins , be energetically favorable or unfavorable .
the lipid bilayer elasticity theory @xcite underlying the description of protein - induced bilayer deformations and bilayer - mediated protein interactions has a rich and distinguished history , dating back to the classic work of w. helfrich @xcite , p. b. canham @xcite , e. a. evans @xcite , and h. w. huang @xcite .
according to this classic theory , membrane proteins may , in addition to thickness - mediated interactions @xcite , also interact @xcite via bilayer curvature deformations @xcite and bilayer fluctuations @xcite . while the competition between thickness-,curvature- , and fluctuation - mediated protein interactions depends on the properties of the specific lipids and membrane proteins under consideration , one generally expects @xcite that thickness - mediated protein interactions are strong and short - ranged , and that curvature- and fluctuation - mediated protein interactions are weak and long - ranged .
the classic elasticity theory of protein - induced lipid bilayer deformations can be extended in various ways @xcite to account for detailed molecular properties of lipids such as lipid tilt and lipid intrinsic curvature @xcite , yielding additional modes of bilayer - mediated protein interactions .
theoretical studies of bilayer - mediated protein interactions have largely focused on idealized ( often cylindrical or conical ) protein shapes which do not correspond to any particular membrane protein , and proteins at large @xmath3 .
in contrast , bilayer - mediated protein interactions at small @xmath3 are most relevant for the crowded environment provided by cell membranes @xcite , while modern structural biology suggests a rich picture of membrane protein shape with experimental surveys of the protein content in various cell membranes @xcite indicating great diversity in the oligomeric states and symmetries of membrane proteins .
the central goal of this article is to provide a detailed discussion of a combined analytic and numerical framework @xcite which allows prediction of lipid bilayer - mediated elastic interactions between integral membrane proteins at arbitrary @xmath3 for the protein shapes suggested by structural studies .
we focus here on protein - induced lipid bilayer thickness deformations , which have been found @xcite to play central roles in regulation of protein function and bilayer - mediated protein interactions in a wide range of experimental systems @xcite . using this mathematical framework
we have shown previously that the shape of integral membrane proteins , and resulting structure of lipid bilayer thickness deformations , can play a crucial role in the regulation of protein function by lipid bilayers @xcite , and that bilayer thickness - mediated interactions between integral membrane proteins can be strongly directional and dependent on protein shape @xcite .
thus , in addition to the magnitude of the bilayer - protein hydrophobic mismatch @xcite , protein shape may be a crucial determinant of membrane protein regulation by lipid bilayers and bilayer - mediated protein interactions .
we develop , illustrate , and test our analytic and numerical framework for calculating bilayer - mediated protein interactions using the protein shapes shown in fig .
[ figillust ] , which embody key mechanisms by which protein crowding and protein shape may affect protein - induced lipid bilayer deformations . the most straightforward model of protein - induced lipid bilayer thickness deformations assumes a circular protein cross section with constant boundary conditions along the bilayer - protein interface [ see fig .
[ figillust](a ) ] .
the resulting `` cylinder model '' of membrane proteins allows investigation of thickness - mediated protein interactions in crowded membranes without the further complications introduced by a complicated protein shape .
the cylinder model has been used before in a number of different settings @xcite to describe protein - induced bilayer thickness deformations .
due to interacting membrane proteins obtained using ( a ) the cylinder model , ( b ) the crown model , and ( c ) the clover - leaf model of integral membrane proteins ( see sec . [ secelasticmodel ] for further details ) .
the thickness deformations in panels ( a ) and ( b ) were obtained through exact analytic minimization of the thickness deformation energy ( see sec . [ secanalyticsol ] ) , and the thickness deformations in panel ( c ) were calculated using finite elements ( see sec . [ secfe ] ) .
the clover - leaf shape in panel ( c ) provides a simple coarse - grained model @xcite of the observed closed - state structure of the pentameric mechanosensitive channel of large conductance @xcite ( protein structural data shown as ribbon diagrams ; protein data bank accession number 2oar ) .
the calculated lipid bilayer thickness deformations depend on lipid and protein properties , the protein center - to - center distance @xmath3 , and , for the crown and clover - leaf models , the protein orientations @xmath4 .
the color scale ranges from @xmath5 nm to @xmath6 nm . ] aside from interactions with neighboring membrane proteins , angular variations in the bilayer - protein boundary conditions @xcite or a non - circular protein cross section @xcite may also break rotational symmetry of bilayer thickness deformations about the protein center .
simple representations of these two features of protein shape are provided @xcite by the `` crown model '' [ see fig .
[ figillust](b ) ] , in which we allow for angular variations of the protein hydrophobic thickness while assuming a circular protein cross section , and by the `` clover - leaf model '' [ see fig .
[ figillust](c ) ] , in which we assume constant boundary conditions along the bilayer - protein interface but allow for a non - circular shape of the protein cross section . in general , membrane proteins will have both a non - circular cross section and variable hydrophobic thickness .
the crown and clover - leaf models allow us to isolate these two possible origins of angular anisotropy of protein - induced lipid bilayer deformations .
an important difference between the cylinder model , and the crown and clover - leaf models , is that for the latter models bilayer - mediated protein interactions are inherently directional and depend not only on the protein separation @xmath3 but also on the protein orientations @xmath7 , where the index @xmath8 denotes different membrane proteins [ figs .
[ figillust](b , c ) ] .
the organization of this article is as follows .
section [ secelasticmodel ] provides a detailed discussion of the elastic energy of lipid bilayer thickness deformations , and the cylinder , crown , and clover - leaf models of integral membrane proteins . in sec .
[ secanalyticsol ] we obtain , based on refs .
@xcite , analytic solutions of the thickness deformation fields and thickness deformation energies due to cylinder , crown , and clover - leaf shapes at arbitrary @xmath3 and protein orientations .
these analytic solutions are exact for cylinder and crown shapes , and perturbative for clover - leaf shapes . as described in sec .
[ secnumericalsol ] , we complement our analytic solutions , and assess their validity , by developing numerical solution schemes based on finite element ( fe ) and finite difference ( fd ) solution procedures
. in particular , the fe approach described here offers a straightforward way of representing the complicated protein shapes suggested by membrane structural biology , and is efficient and accurate enough to enable prediction of the directional thickness - mediated interactions of hundreds of integral membrane proteins at arbitrary @xmath3 and protein orientations @xcite with reasonably coarse computational grids . in sec .
[ seccylinder2 ] we provide a detailed comparison between analytic and numerical results for the thickness deformation fields and thickness deformation energies implied by the cylinder model in the non - interacting and interacting regimes of @xmath3 .
sections [ seccrownresults ] and [ seccloverresults ] provide similar comparisons between analytic and numerical results for the crown and clover - leaf models .
a summary of our results and conclusions can be found in sec .
[ secsummary ] .
in the standard elasticity theory of bilayer - protein interactions @xcite , integral membrane proteins are assumed to be rigid membrane inclusions which deform the surrounding lipid bilayer @xcite . in the simplest formulation , bilayer deformations
can then be captured by two coupled scalar fields @xmath9 and @xmath10 which define the positions of the hydrophilic - hydrophobic interface in the outer and inner lipid bilayer leaflets , respectively .
it is mathematically convenient to express @xmath9 and @xmath10 in terms of the midplane deformation field @xmath11 and the thickness deformation field @xmath12 in which @xmath13 is one - half the unperturbed hydrophobic thickness of the lipid bilayer . to leading order , the elastic energies governing @xmath14 and @xmath15 decouple from each other @xcite . in the most straightforward model of bilayer - protein interactions
@xcite , the energy cost of midplane deformations is then captured by the helfrich - canham - evans energy @xcite and , within the monge representation , the energy cost of thickness deformations is of the form @xcite @xmath16 \right\}}\,,\ ] ] where @xmath17 is the bending rigidity of the lipid bilayer , @xmath18 is the stiffness associated with thickness deformations , and @xmath19 is the membrane tension .
the effective parameters @xmath17 , @xmath18 , and @xmath13 in eq .
( [ energy ] ) encapsulate bilayer material properties relevant for protein - induced bilayer thickness deformations and depend on the bilayer composition @xcite .
typical values measured in experiments are @xmath20 @xmath2 , @xmath21 @xmath22nm@xmath23 , and @xmath24 nm @xcite , which we use for all the numerical calculations described here .
the classic model of protein - induced lipid bilayer thickness deformations in eq . ( [ energy ] ) employs the monge representation of surfaces , @xmath25 , with cartesian coordinates @xmath26 , and only considers leading - order terms in @xmath15 and its derivatives .
the former assumption can be justified by noting that thickness deformations generally decay rapidly compared to midplane deformations , with typical thickness and midplane decay lengths @xmath27 nm and @xmath28500 nm @xcite , respectively .
the validity of the latter assumption depends on the specific properties of the lipid bilayer and protein under consideration , but for experimental model systems @xcite one typically finds @xmath29 and @xmath30 .
hence , bilayer overhangs and higher - order corrections to eq .
( [ energy ] ) can often be neglected when describing protein - induced lipid bilayer thickness deformations , with the bilayer midplane being approximately parallel to the reference plane invoked in the monge representation of surfaces .
the terms @xmath31 and @xmath32 in eq .
( [ energy ] ) provide lowest - order descriptions of the energy cost of bilayer bending , and compression and expansion , of the bilayer hydrophobic core , respectively . for generality
we allow for the two tension terms @xmath33 and @xmath34 in eq .
( [ energy ] ) , which account @xcite for stretching deformations tangential to the leaflet surfaces and changes in the projection of the bilayer area onto the reference plane , respectively . the minimal model in eq .
( [ energy ] ) can be extended in a variety of ways to account for more detailed properties of lipid bilayers including lipid tilt @xcite , lipid intrinsic curvature @xcite , inhomogeneous deformation of lipid volume and effects of gaussian curvature on protein - induced bilayer deformations @xcite , asymmetric bilayer thickness deformations @xcite , and protein - induced local modulation of bilayer elastic properties @xcite . the lipid bilayer thickness deformation energy in eq . ( [ energy ] )
provides a simple model of thickness - mediated protein interactions , as well as the coupling between protein function and bilayer thickness deformations , and has the appealing property that all the material parameters entering eq . ( [ energy ] ) can be measured directly in experiments . for given bilayer - protein boundary conditions ( see sec . [ secshape ] ) , minimization of eq .
( [ energy ] ) completely specifies the lowest - energy bilayer thickness configuration and its associated energy cost .
models based on eq .
( [ energy ] ) have been found to capture the basic experimental phenomenology of bilayer - protein interactions for gramicidin channels @xcite , the mechanosensitive channel of large conductance ( mscl ) @xcite , g - protein coupled receptors @xcite , the bacterial leucine transporter @xcite , and chemoreceptor lattices @xcite , as well as a variety of other integral membrane proteins @xcite .
while we focus here on elastic thickness deformations , midplane deformations may generally also contribute @xcite to bilayer - mediated protein interactions , and the regulation of protein function by bilayer mechanical properties .
minimization of eq .
( [ energy ] ) can be performed by solving the appropriate euler - lagrange equation , which is given by @xmath35 the analytic solution of eq .
( [ genbihpreu ] ) is facilitated @xcite by introducing the function @xmath36 in terms of which eq .
( [ genbihpreu ] ) can be expressed as @xmath37 where @xmath38\,.\ ] ] to analytically calculate the thickness deformation energy associated with the solutions of eq .
( [ genbihpreu ] ) we use eqs .
( [ genbihpreu ] ) and ( [ transfu ] ) to rearrange the thickness deformation energy in eq .
( [ energy ] ) as @xmath39\ , , \nonumber\\&&\label{energy2}\end{aligned}\ ] ] where the term @xmath40 is independent of @xmath41 and arises , for @xmath42 , due to relaxation of the `` loading device '' producing membrane tension @xcite via uniform compression of the bilayer hydrophobic core . since @xmath43 does not contribute to the energy cost of protein - induced bilayer deformations we subtract @xmath43 from @xmath44 , @xmath45 .
using gauss s theorem , we then find @xmath46}\,,\ ] ] where the line integrals @xmath47 are to be taken along all bilayer - protein interfaces , with the bilayer unit normal vectors @xmath48 perpendicular to the bilayer - protein interfaces and pointing towards the proteins , and we assume @xcite that @xmath41 , as well as its derivatives , go to zero far from the proteins .
following previous work on lipid bilayer - protein interactions @xcite we model integral membrane proteins as rigid membrane inclusions of fixed shape and hydrophobic thickness .
the specific properties of a given membrane protein enter our description of bilayer - protein interactions through the shape of the protein cross section , and through the boundary conditions on @xmath15 along the bilayer - protein interface .
as described in sec .
[ secintro ] , we consider here three distinct models of protein shape : the cylinder model , the crown model , and the clover - leaf model
. these minimal models do not provide detailed descriptions of protein shape but , rather , aim to encapsulate the features of a given protein hydrophobic surface most crucial for protein - induced lipid bilayer thickness deformations . as noted above
, midplane deformations decouple to leading order from thickness deformations .
hence , the models of protein - induced bilayer thickness deformations considered here can be easily complemented by corresponding models of protein - induced bilayer midplane deformations @xcite , which capture separate aspects of bilayer - protein interactions . a straightforward description of the effect of protein shape on protein - induced lipid bilayer thickness deformations is provided by the cylinder model of integral membrane proteins [ fig .
[ figillust](a ) ] @xcite . introducing the polar coordinates @xmath49 with the center of membrane protein @xmath50 as the origin , the boundary conditions for protein @xmath50 can be written as @xmath51 where the boundary curve @xmath52 for a membrane protein with a circular cross section of radius @xmath53 , and the constants @xcite @xmath54 where @xmath55 is the hydrophobic thickness of protein @xmath50 and the @xmath56 correspond to the normal derivatives of @xmath57 evaluated along the bilayer - protein boundary .
equation ( [ bc1 ] ) assumes perfect hydrophobic matching between the membrane protein and the lipid bilayer @xcite .
this assumption is expected to break down for a large enough hydrophobic mismatch between the protein and the ( undeformed ) lipid bilayer @xcite , in which case @xmath55 corresponds to the effective hydrophobic thickness of the membrane protein . unless indicated otherwise
, we use @xmath58 nm for the numerical calculations described here , which approximately corresponds to @xcite the hydrophobic thickness of the observed structure of closed pentameric mscl @xcite . following previous work on mscl - induced lipid bilayer thickness deformations employing the cylinder model of integral membrane proteins @xcite , we use @xmath59 nm , which yields an area of the transmembrane protein cross section consistent with the observed structure of closed pentameric mscl @xcite . a number of different choices for the boundary condition in eq .
( [ bc2f ] ) have been investigated @xcite .
in particular , @xmath60 may be chosen based on experimental observations or molecular dynamics simulations , or may be regarded as a free parameter to be fixed as part of the energy minimization procedure .
we follow here previous theoretical work on the experimental phenomenology of gramicidin channels @xcite and mscl @xcite which suggests that , to a first approximation , @xmath61 . the hydrophobic thickness of integral membrane proteins is generally expected to vary along the bilayer - protein interface @xcite , yielding anisotropic bilayer thickness deformations and , in the case of two or more membrane proteins in sufficiently close proximity , directional interactions @xcite . to study the generic effects of a variable protein hydrophobic thickness on bilayer - mediated protein interactions we replace the constant @xmath62 in eq .
( [ bc1f ] ) by @xcite @xmath63 where @xmath64 is the average hydrophobic mismatch , @xmath65 is the magnitude of mismatch modulations , @xmath66 is the protein symmetry , and @xmath7 parametrizes the orientation of protein @xmath50 . for each bilayer leaflet , eq .
( [ varu ] ) yields a periodic modulation of the protein hydrophobic surface which resembles the shape of a crown [ fig .
[ figillust](b ) ] , and we therefore refer to eq .
( [ varu ] ) as the crown model of integral membrane proteins . for our numerical calculations
we use the values @xmath67 nm , @xmath68 nm , and @xmath69 in eq .
( [ varu ] ) , and vary @xmath7 to explore bilayer thickness - mediated interactions for a range of relative protein orientations .
we choose all other parameter values as described for the cylinder model of membrane proteins . even for non - interacting membrane proteins , this parametrization of the crown model in eq .
( [ varu ] ) yields a maximum magnitude of the gradient of bilayer thickness deformations @xmath27 , and therefore produces thickness deformations which lie at the limit of applicability of the leading - order energy in eq .
( [ energy ] ) .
furthermore , for interacting membrane proteins we generally find with this parametrization of the crown model that the maximum magnitude of the gradient of bilayer thickness deformations @xmath70 . as a result , the numerical estimates of the thickness deformation energy obtained for this parametrization of eq .
( [ varu ] ) are of limited physical significance .
we allow here for such large magnitudes of the gradient of bilayer thickness deformations in order to explore the mathematical limits of applicability of our analytic and numerical solution procedures ( see sec .
[ seccrownresults ] ) .
membrane structural biology has produced a rich and diverse picture of membrane protein shape , which suggests @xcite that integral membrane proteins can occur in a variety of different oligomeric states and transmembrane shapes .
distinct oligomeric states of membrane proteins generally yield distinct symmetries of the protein cross section which , in turn , induce distinct symmetries of lipid bilayer thickness deformations @xcite .
the resulting non - trivial structure of bilayer thickness deformations can yield substantial deviations from the energy cost of protein - induced bilayer thickness deformations implied by the cylinder model of integral membrane proteins . in particular , for mscl it has been found @xcite that the elastic energy of protein - induced bilayer thickness deformations provides a signature of the protein oligomeric state , with distinct mscl oligomeric states yielding distinct mscl gating characteristics and directional bilayer - mediated protein interactions . a simple coarse - grained model of the cross sections of a diverse range of membrane proteins is provided by the clover - leaf model @xcite @xmath71\,,\ ] ] where @xmath72 parametrizes the magnitude of the deviation of the protein cross section from the circle [ fig .
[ figillust](c ) ] .
in particular , the structure of pentameric mscl observed in _
mycobacterium tubercolosis _
@xcite suggests @xcite @xmath69 , @xmath73 nm , and @xmath74 , which we use for all the numerical calculations involving clover - leaf shapes described here . in general , the hydrophobic thickness of integral membrane proteins
is expected to vary along the boundaries of clover - leaf shapes . however , in order to isolate the effect of anisotropy in protein shape on bilayer thickness - mediated protein interactions we focus here on the simpler scenario of a constant hydrophobic thickness , and use for the clover - leaf model of integral membrane proteins the same boundary conditions on @xmath15 along the bilayer - protein interface as for the cylinder model [ see eqs .
( [ bc1f ] ) and ( [ bc2f ] ) ] .
building on earlier work on the lipid bilayer thickness deformations induced by cylindrical membrane proteins @xcite and bilayer curvature - mediated interactions between conical membrane proteins in the far - field limit @xcite , we develop in this section analytic solutions @xcite of the bilayer thickness - mediated interactions between integral membrane proteins at arbitrary protein separations and relative orientations . to solve for the thickness - mediated interactions implied by eq .
( [ energy ] ) for two membrane proteins , we employ a two - center bipolar coordinate system ( see fig .
[ figbipol ] ) . for the sake of simplicity ,
we assume in fig .
[ figbipol ] that the two proteins have circular cross sections with radii @xmath75 .
we relax this assumption below to capture interactions between clover - leaf shapes . to mathematically relate the polar
coordinates @xmath76 centered about proteins 1 and 2 , we note from fig .
[ figbipol ] the bipolar coordinate transformations @xmath77 @xmath78 , and @xmath79 .
the corresponding transformations for @xmath80 and @xmath81 are symmetric in the protein indices , but @xmath82
. separated by a center - to - center distance @xmath3 along the @xmath83-axis .
the @xmath83-@xmath84 plane corresponds to the reference plane used in the monge representation of surfaces .
expressions for @xmath85 in terms of @xmath86 can be obtained by considering the coordinates of the red ( left ) point , and expressions for @xmath86 in terms of @xmath85 can be obtained by considering the coordinates of the blue ( right ) point . ]
we solve the euler - lagrange equation ( [ genbih ] ) by making the ansatz @xcite @xmath87 where the @xmath88 are the solutions of eq .
( [ genbih ] ) for a single protein @xmath89 , which are of the form @xcite @xmath90 where the @xmath91 are solutions of the helmholtz equations @xmath92 for the exterior of a circle of radius @xmath53 , the above helmholtz equations are readily solved by separation of variables @xcite .
thus , the general single - protein solution of eq .
( [ genbih ] ) can be constructed from @xmath93 where @xmath94 and @xmath95 are constants , the fourier - bessel terms @xmath96 the @xmath97 and @xmath98 with @xmath99 are the modified bessel functions of the first and second kind , and @xmath100 , @xmath101 , @xmath102 , @xmath103 with @xmath104 are constants .
assuming that @xmath105 as @xmath106 @xcite , we have @xmath107 for @xmath104 , and eq .
( [ gensolsingle ] ) reduces to @xcite @xmath108 where @xmath109 corresponds to the full single - protein solution .
if the boundary conditions on @xmath15 along the bilayer - protein interfaces are such that @xmath110 in fig . [ figbipol ] we have , by symmetry , that @xmath111 for @xmath104 .
substitution of eq .
( [ constructgensol ] ) with eq .
( [ gensolsingle2 ] ) into eq .
( [ gensol ] ) yields the solution of the thickness deformations induced by two membrane proteins . to use eq .
( [ evalenergyline ] ) to evaluate the elastic energy associated with these thickness deformations , and to impose suitable boundary conditions along the bilayer - protein interfaces , we recast along the bilayer - protein boundary associated with protein 2@xmath112 in terms of @xmath113 , @xmath114 , and @xmath115 , and vice versa . for protein 2 , this is achieved @xcite by first expanding the bipolar coordinate transformations for @xmath80 , @xmath81 , and @xmath116 , and then the expression for @xmath112 in eq .
( [ constructgensol ] ) , in terms of @xmath117 up to some order @xmath118 in @xmath119 .
steric constraints mandate @xmath120 and , hence , @xmath121 along the bilayer - protein boundary associated with protein 2 . following a similar procedure for protein 1 , eq .
( [ gensol ] ) yields explicit expressions for @xmath41 in terms of @xmath86 in the vicinity of protein 2 and in terms of @xmath85 in the vicinity of protein 1 .
we note that expansion of @xmath122 around protein 2 up to order @xmath118 in @xmath119 produces angular variations in @xmath123 up to @xmath124 and @xmath125 .
we set @xmath126 to ensure that these `` secondary '' angular variations , which are introduced into the general solution in eq .
( [ gensol ] ) via expansion of the bipolar coordinate transformations , are of the same maximum order as the angular variations captured directly by the fourier - bessel series in eq .
( [ gensolsingle2 ] ) @xcite .
the expansions described above yield explicit expressions for @xmath41 in terms of @xmath85 and @xmath86 in the vicinity of proteins 1 and 2 , respectively .
thus , the expression for the thickness deformation energy in eq .
( [ evalenergyline ] ) can be written as @xmath127 where the overall minus signs arise because the bilayer normal vectors point towards decreasing @xmath128 along the bilayer - protein interfaces , and the boundary energy densities @xmath129_{r_1=r_1}\,,\nonumber\\&&\\ \label{eanaytic2 } g_2(\theta_2)&= & \left[k_b \frac{\partial \bar u}{\partial r_2 } \nabla_2 ^ 2 \bar u - k_b \bar u \frac{\partial}{\partial r_2}\nabla_2 ^ 2 \bar u+\tau \bar u \frac{\partial \bar u}{\partial r_2 } \right]_{r_2=r_2}\nonumber\\&&\end{aligned}\ ] ] are evaluated using @xmath130 and @xmath131 , respectively , in which we have noted that @xmath132 along the circumference of a circle , and the laplace operators @xmath133 in polar coordinates , where @xmath89 . as a result of our expansions of the bipolar coordinate transformations , @xmath41 around proteins 1 and 2 only depends on @xmath134 through linear sums over @xmath135 and @xmath136 with @xmath99 , and it is therefore straightforward @xcite to analytically evaluate the angular integrals in eq .
( [ intgder ] ) , resulting in an algebraic expression for @xmath44 . in the case of two interacting membrane proteins , the general solution in eq .
( [ gensol ] ) contains the @xmath137 coefficients @xmath94 , @xmath100 , and @xmath101 with @xmath89 and @xmath138 .
these coefficients are determined by the boundary conditions through the linear system of equations @xmath139 where the vector @xmath140 is of length @xmath137 and contains each coefficient appearing in eq .
( [ gensol ] ) as a separate element .
the vector @xmath141 contains the boundary conditions on @xmath41 and its radial derivatives , at proteins 1 and 2 , at each order in @xmath142 and @xmath143 with @xmath144 as separate elements . at @xmath145
, we have two boundary conditions for each protein , yielding four elements in @xmath141 .
similarly , for @xmath146 we have four boundary conditions at each order in @xmath147 for each protein yielding , in total , the @xmath137 independent boundary conditions required to fix the @xmath137 independent coefficients appearing in eq .
( [ gensol ] ) .
finally , the rows of the matrix @xmath148 are constructed from the coefficients of @xmath142 and @xmath143 with @xmath99 in eq .
( [ gensol ] ) and their radial derivatives , at proteins 1 and 2 , with each column corresponding to a particular coefficient . for two proteins one therefore obtains four rows at @xmath145 and eight rows at each @xmath149 yielding , as required , a square matrix of order @xmath137 .
solution of eq .
( [ matrixeq ] ) for the coefficients @xmath140 @xcite yields @xcite , as @xmath109 , the exact thickness deformations @xmath15 in eqs .
( [ transfu ] ) and ( [ gensol ] ) induced by arbitrary protein configurations and , via eq .
( [ intgder ] ) , the associated thickness deformation energy @xmath44 .
however , to make the above solution procedure analytically tractable it is , in practice , necessary to truncate the respective fourier - bessel series , and expansions of the bipolar coordinate transformations , at some finite value of @xmath150 .
such a truncation relies on the assumption that , beyond @xmath150 , angular variations in eq .
( [ gensol ] ) can be neglected .
the validity of this assumption for a given value of @xmath150 can be confirmed @xcite by systematically including higher - order terms . for large @xmath150
, it can be convenient to substitute numerical values for all model parameters , and to numerically solve for @xmath140 in eq .
( [ matrixeq ] ) . the above analytic solution procedure can be implemented directly for the boundary conditions associated with the cylinder and crown models of membrane proteins discussed in secs .
[ seccylinder ] and [ seccrown ] . for the clover - leaf model discussed in sec .
[ secclover ] , suitable ( approximate ) boundary conditions can be obtained perturbatively from eq .
( [ boundcclover ] ) @xcite by expanding the left - hand sides of eqs .
( [ bc1f ] ) and ( [ bc2f ] ) in terms of the small parameter @xmath72 @xcite , and noting that @xmath151 to leading order in @xmath72 , eqs .
( [ bc1f ] ) and ( [ bc2f ] ) are then given by @xmath152 where @xmath153 note that , for protein configurations which satisfy @xmath154 , only cosine modes must be considered in the above expressions because , for such configurations , the arrangement in fig .
[ figbipol ] is symmetric under @xmath155 . for the sake of simplicity , we approximate @xmath41 in eqs .
( [ genbc1trimerfin2 ] ) and ( [ genbc2trimerfin2 ] ) by only including the rotationally symmetric `` background '' fields about proteins 1 and 2 in eq .
( [ gensol ] ) .
bilayer thickness - mediated interactions in the clover - leaf model of integral membrane proteins can then be analyzed following the same steps as for the cylinder and crown models , but with the additional approximations inherent in eqs .
( [ genbc1trimerfin ] ) and ( [ genbc2trimerfin ] ) .
the fe framework provides a versatile numerical approach for handling protein - induced lipid bilayer deformations in crowded membranes .
we have developed a general fe scheme @xcite for the numerical study of bilayer - protein interactions which allows reliable and efficient minimization of eq .
( [ energy ] ) for hundreds of interacting integral membrane proteins with complicated shapes and boundary conditions . to complement our fe approach ,
we have also developed a fd scheme for the minimization of eq .
( [ energy ] ) . in this section
we provide a detailed description of our fe and fd solution procedures , which permit minimization of eq .
( [ energy ] ) for effectively arbitrary protein shapes , separations , and orientations . while the analytic solution described in sec .
[ secanalyticsol ] allows for infinitely large system sizes , numerical solutions are necessarily restricted to finite solution domains . with the exception of special cases such as provided by periodic systems @xcite , numerical minimization of eq .
( [ energy ] ) therefore relies on the assumption that the energy cost of thickness deformations decays sufficiently rapidly at large distances from the proteins so that finite size effects can be neglected .
this assumption is violated for finite membrane tensions in eq .
( [ energy ] ) , in which case the magnitude of @xmath43 in eq .
( [ energyg1 ] ) increases with bilayer area and , indeed , becomes infinite in the limit of infinitely large lipid bilayers .
however , @xmath43 does not contribute to the energy cost of protein - induced lipid bilayer thickness deformations .
thus , direct comparisons between analytic and numerical solutions of the energy cost of protein - induced lipid bilayer thickness deformations in the non - interacting as well as interacting regimes can be made , even for @xmath156 , by subtracting the ( finite ) value of @xmath43 associated with a specific choice for the size of the solution domain from the numerical solution , and ( formally ) subtracting the corresponding ( infinite ) value of @xmath43 from the analytic solution as in eq .
( [ evalenergyline ] ) . furthermore , as we discuss below , comparisons between analytic and numerical results for the thickness deformation field , rather than the thickness deformation energy , over identical ( finite ) bilayer areas for the analytic and numerical solutions provide an alternative approach for testing our numerical solution procedures
. this approach does not rely on subtracting @xmath43 in eq .
( [ energy2 ] ) .
the fe framework @xcite was developed to permit reliable and computationally efficient numerical solutions of boundary value problems involving large computational domains with complicated boundary shapes and boundary conditions .
this makes the fe approach well suited for the calculation of lipid bilayer - mediated protein interactions in crowded membranes with many interacting proteins . using mscl as a model system
, we have shown previously @xcite that the fe approach makes it feasible to predict directional bilayer - mediated protein interactions in systems composed of hundreds of integral membrane proteins . while standard fe methods based on lagrange interpolation functions @xcite are sufficient to compute the thickness stretch and tension terms in eq .
( [ energy ] ) , the curvature term , being second order in derivatives , requires @xmath157 continuity @xcite and therefore can not be handled through lagrange interpolation functions . the discrete kirchhoff triangle ( dkt )
method offers an elegant and efficient way to circumvent this limitation @xcite .
in particular , to bypass @xmath157 continuity , the dkt approach employs a plate theory allowing for transverse shear deformations , in which case @xmath158 continuity is sufficient .
the kirchhoff hypothesis of zero transverse shear is then enforced discretely along the edges of the triangular elements , thus ensuring the conformity of curvatures at element interfaces . in our fe framework for calculating bilayer thickness - mediated interactions between integral membrane proteins @xcite
we adopt a hybrid fe approach , in which we combine the dkt formulation for the bending terms @xcite with standard lagrange interpolation for the thickness stretch and gradient terms @xcite . to derive the stiffness matrix associated with this hybrid fe approach , we rewrite eq .
( [ energy ] ) in cartesian coordinates , @xmath159 + \frac{2\tau}{a } u \right\ } \ , .
\label{feenergy}\end{aligned}\ ] ] the variation of eq .
( [ feenergy ] ) is given by @xmath160 with the generalized strain vector @xmath161\,,\ ] ] and the constitutive matrix @xmath162 \ , .\ ] ] while the displacements @xmath163 , @xmath164 , and @xmath165 of the corner nodes of each fe triangle are sufficient to define lagrange interpolation functions , the dkt approach requires nine degrees of freedom per triangle , @xmath166\,,\ ] ] where the partial derivatives @xmath167 and @xmath168 with @xmath169 or 3 correspond to rotations at the corner nodes of each fe triangle .
we use the strain - displacement transformation matrix @xmath170\ ] ] to construct the strain vector @xmath171 , where the linear triangular shape functions @xmath172 are given in ref .
@xcite and the dkt shape functions @xmath173 are given in ref .
@xcite .
finally , the fe thickness deformation energy is obtained by summing over all finite elements , @xmath174 in which the element stiffness matrix @xmath175 and `` internal tension '' @xmath176 are given by @xmath177 the above integrals are performed over the local coordinates @xmath178 using second - order ( three points per element ) gaussian quadrature , and scaled by the area @xmath179 of the element . minimizing eq .
( [ energy ] ) obtained using our fe approach for two clover - leaf pentamers . the triangular mesh ( black overlay )
indicates the mesh used in the fe calculation , and is generated using the frontal algorithm of the gmsh package @xcite .
all model parameters were chosen as described in sec .
[ secelasticmodel ] . ] to enforce the general boundary conditions along the bilayer - protein interfaces in eqs .
( [ bc1f ] ) and ( [ bc2f ] ) we fix @xmath180 for the nodes defining the protein boundaries , where @xmath181 and @xmath182 are @xmath83 and @xmath84 projections of the thickness gradient @xmath183 along the contour of the bilayer - protein interfaces . for the cylinder and clover - leaf models of protein shape
the hydrophobic thickness is constant along the bilayer - protein interface , and we therefore impose @xmath184 . for the crown model we have @xmath185 from the hydrophobic thickness variation in eq .
( [ varu ] ) .
for the nodes defining the outer boundary of the simulation domain we do not constrain @xmath15 and its derivatives .
we assemble the thickness deformation energy in eq .
( [ eq : enfe ] ) in c++ using the variational mechanics library voom and minimize eq .
( [ eq : enfe ] ) by employing the l - bfgs - b solver @xcite .
figure [ fe1 ] shows a representative thickness deformation profile obtained for two clover - leaf pentamers together with the corresponding triangular mesh used in the fe calculation , which we generate using the frontal algorithm of the gmsh package @xcite .
we check the accuracy of our fe procedure by comparing the total thickness deformation energies predicted by analytic and fe approaches for the cases in which exact analytic results on protein - induced lipid bilayer thickness deformations are available @xcite ( see sec . [ secanalyticsol ] ) . as discussed above
, such comparisons necessitate subtracting @xmath43 in eq .
( [ energyg1 ] ) . to complement these tests , we also compare our analytic and fe approaches using the analytic and fe solutions for the thickness deformation field .
this approach for quantifying the level of agreement between analytic and fe approaches does not rely on subtracting @xmath43 in eq .
( [ energyg1 ] ) .
in particular , following ref .
@xcite we monitor the percentage error in the thickness deformations obtained from the fe approximation , @xmath186 , relative to the analytic solution , @xmath15 , @xmath187 and the corresponding percentage error in curvature deformations , @xmath188 where the @xmath189 norm @xmath190 and the sobolev semi - norm @xmath191 are evaluated by numerical quadrature over identical ( finite ) bilayer areas for the analytic and numerical solutions , using an approximately circular integration domain with radius @xmath192 nm .
fd methods have been employed previously to study the bilayer thickness deformations induced by gramicidin channels @xcite , mscl @xcite , g - protein coupled receptors @xcite , and the bacterial leucine transporter @xcite . in our fd scheme , we discretize the lipid bilayer domain of interest using a hexagonal grid with @xmath193 nodes and lattice spacing @xmath14 [ see fig . [ fd1](a ) ] .
we denote the nodal values of the thickness deformations by @xmath194 .
taylor series expansion then yields the discretized laplace operator @xmath195\,.\end{aligned}\ ] ] minimization of the thickness deformation energy in eq .
( [ energy ] ) via solution of the euler - lagrange equation ( [ genbihpreu ] ) using fd requires an expression for the discretized biharmonic term @xmath196 .
we obtain @xmath196 by applying the discretized laplace operator in eq .
( [ eqfdlap ] ) two times , resulting in a 19-point stencil with the coefficients shown in fig .
[ fd1](a ) . are indexed by @xmath50 along the horizontal direction and by @xmath147 along the oblique direction .
the 19-point stencil discretizing the biharmonic operator at position @xmath197 is found by multiplying @xmath194 and nearby nodal values of the thickness deformation field by the indicated coefficients , adding up these contributions , and dividing by @xmath198 .
( b ) we choose the protein boundary points ( blue circles ) to correspond to the nodes closest to the exact protein boundary curve ( grey clover - leaf shape ; see sec . [ secclover ] ) , and impose the general boundary condition in eq . ( [ bc1f ] ) at these nodes .
we use a layer of interior points ( red squares ) , together with corresponding exterior points ( see main text ) , to impose the general boundary condition in eq .
( [ bc2f ] ) . ] for our fd calculations we focus on the case of zero membrane tension , for which eq .
( [ energy ] ) yields the fd thickness deformation energy @xmath199\,,\ ] ] in which nodal contributions are scaled with the unit area associated with each node .
collecting nodal values of the thickness mismatch @xmath194 into a vector @xmath200 of length @xmath201 , we recast the energy in eq .
( [ energyfd ] ) into the matrix form @xmath202 \nonumber \\ & = & \frac{\sqrt{3 } h^2}{2}\left [ \frac{k_b}{2}\left(\vec{u}^t \vec{n } \vec{u}\right ) + \frac{k_t}{2a^2 } \vec{u}^t \vec{u } \right ] \nonumber\\ & \equiv & \vec{u}^t \vec{q } \vec{u}\ , , \label{calculategfd}\end{aligned}\ ] ] where @xmath203 and @xmath204 , @xmath205 , and @xmath206 are @xmath207 laplacian , biharmonic , and identity matrices , respectively . in each row ,
the matrices @xmath204 and @xmath208 have the coefficients of the discrete laplace and biharmonic operators associated with a single node as their elements , respectively .
the matrices @xmath204 and @xmath208 are therefore highly sparse , with non - zero elements organized according to the node ordering of the vector @xmath200 . to enforce the general bilayer - protein boundary conditions in eqs .
( [ bc1f ] ) and ( [ bc2f ] ) in our fd scheme we find all grid points at a distance less than @xmath209 from the protein boundary , as illustrated in fig .
[ fd1](b ) for clover - leaf shapes .
we take these nodes to define the protein boundary in our fd scheme , and impose the boundary condition in eq .
( [ bc1f ] ) by setting @xmath210 along the discretized protein boundary . to enforce the boundary condition on @xmath211 in eq .
( [ bc2f ] ) we find , for each protein , the nodes in the interior of the protein boundary curve within a distance @xmath212 from the protein boundary so that @xmath213 . for each of these interior points
we then find a mirror symmetric point exterior to the protein boundary curve such that interior and exterior points are connected by a line of length @xmath214 which is normal to the exact protein boundary curve . in order to satisfy the boundary condition @xmath61 used here
, we impose the constraint that the values of @xmath194 at the interior and exterior points are equal to each other ( other choices for the value of @xmath60 could be implemented following analogous steps ) . here
a complication arises in that the exterior points are typically not grid points .
we address this issue by interpolating , for each exterior point , the values of the thickness deformation field at the three nearest grid points . due to the substantial computational cost associated with our fd scheme
we only employ here the fd approach to study mirror - symmetric protein configurations . for two proteins located at @xmath215 the size of the solution domain
can then be reduced by a factor of one - half by imposing mirror - symmetric boundary conditions along the boundary line @xmath216 ( see fig .
[ figbipol ] ) . for the remaining boundaries of the solution domain we set @xmath217 . for all our fd calculations we use a rectangular solution domain of side lengths ( @xmath218 , @xmath219 ) nm with the protein center placed on the longer midline of the rectangle .
we check that our results are robust with respect to increases in the size of the solution domain .
to minimize the thickness deformation energy we construct the discretized version of the euler - lagrange equation ( [ genbihpreu ] ) for @xmath220 , @xmath221 defined for the nodal values @xmath200 of all fd grid points . to enforce the boundary conditions in eqs .
( [ bc1f ] ) and ( [ bc2f ] ) we adjust the rows of the matrix @xmath222 in eq .
( [ eqfdeuler ] ) corresponding to the fd boundary nodes so as to fix the value of @xmath194 at these nodes , as described above .
accordingly , the vector @xmath223 contains non - zero elements at rows corresponding to the boundary nodes .
we solve the linear system in eq .
( [ eqfdeuler ] ) using the sparse matrix structures and solvers provided by the numerical computing environment matlab @xcite .
we calculate the corresponding thickness deformation energy using eq .
( [ calculategfd ] ) for the nodal values of all fd grid points lying outside the protein domains .
in this section we focus on the most straightforward scenario of lipid bilayer thickness deformations induced by cylindrical membrane proteins [ fig . [
figillust](a ) ] , and compare numerical results obtained using our fe and fd schemes to the corresponding exact analytic solutions . as discussed above
, we subtract @xmath43 in eq .
( [ energyg1 ] ) to compare thickness deformation energies obtained using numerical and analytic solution procedures .
we first consider the bilayer thickness deformations induced by cylindrical membrane proteins in the non - interacting regime of large protein separations , and then discuss the bilayer thickness deformations induced by two interacting cylindrical membrane proteins . versus radial coordinate @xmath224 obtained from the exact analytic solution in eq .
( [ constructgensol ] ) for @xmath225 @xcite , the fe approach , and the fd approach .
the grey vertical line indicates the protein boundary .
( b ) bilayer thickness deformation energy versus thickness mismatch @xmath226 obtained using analytic , fe , and fd approaches .
the insets show ( a ) the difference in thickness deformation profile , @xmath227 , and ( b ) the percentage difference in thickness deformation energy , @xmath228 in eq .
( [ eq : defetag ] ) , between the analytic solution and the corresponding results of fe ( blue solid curves ) and fd ( green dashed curves ) calculations , respectively . for the fe solution we used an average edge size of the fe mesh @xmath229 nm , and for the fd solution we used a lattice spacing @xmath230 nm . we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] + versus radial coordinate @xmath224 obtained from the exact analytic solution in eq .
( [ constructgensol ] ) for @xmath225 @xcite , the fe approach , and the fd approach .
the grey vertical line indicates the protein boundary .
( b ) bilayer thickness deformation energy versus thickness mismatch @xmath226 obtained using analytic , fe , and fd approaches .
the insets show ( a ) the difference in thickness deformation profile , @xmath227 , and ( b ) the percentage difference in thickness deformation energy , @xmath228 in eq .
( [ eq : defetag ] ) , between the analytic solution and the corresponding results of fe ( blue solid curves ) and fd ( green dashed curves ) calculations , respectively . for the fe solution we used an average edge size of the fe mesh @xmath229 nm , and for the fd solution we used a lattice spacing @xmath230 nm .
we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] for a single cylindrical membrane protein , the protein - induced lipid bilayer thickness deformations are rotationally symmetric about the protein , with the exact analytic solution @xcite corresponding to the zeroth - order terms in the general solution in eq .
( [ constructgensol ] ) .
the radial profile of this exact analytic solution about the membrane protein is governed by zeroth - order modified bessel functions of the second kind , yielding an approximately exponential decay of thickness deformations with a periodic modulation @xcite and a characteristic length scale of thickness deformations @xmath231 nm @xcite [ see fig .
[ fig : single_cyl](a ) ] . the thickness deformation profiles obtained using our fe and fd solution procedures are in excellent quantitative agreement with the corresponding analytic solution [ fig .
[ fig : single_cyl](a ) ] .
computing the percentage difference between numerical and analytic results for the thickness deformation energy , @xmath232 we find that the thickness deformation energies obtained using analytic , fe , and fd approaches are also in excellent quantitative agreement [ fig .
[ fig : single_cyl](b ) ] .
as expected from scaling arguments @xcite , all three solution procedures yield an approximately quadratic dependence of the thickness deformation energy on hydrophobic mismatch . )
( @xmath233 ; orange dashed curve ) and percentage error in curvature deformations in eq .
( [ percentageerror2 ] ) ( @xmath234 ; magenta solid curve ) versus @xmath235 for a single cylindrical membrane protein .
errors in the thickness mismatch and curvature deformations decay as @xmath235 and @xmath236 , respectively .
( b ) percentage difference between analytic and fe results for the thickness deformation energy per cylindrical membrane protein , @xmath228 in eq .
( [ eq : defetag ] ) , versus @xmath235 for a system consisting of two identical cylindrical membrane proteins in the non - interacting regime ( @xmath237 nm ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] + ) ( @xmath233 ; orange dashed curve ) and percentage error in curvature deformations in eq .
( [ percentageerror2 ] ) ( @xmath234 ; magenta solid curve ) versus @xmath235 for a single cylindrical membrane protein .
errors in the thickness mismatch and curvature deformations decay as @xmath235 and @xmath236 , respectively .
( b ) percentage difference between analytic and fe results for the thickness deformation energy per cylindrical membrane protein , @xmath228 in eq .
( [ eq : defetag ] ) , versus @xmath235 for a system consisting of two identical cylindrical membrane proteins in the non - interacting regime ( @xmath237 nm ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] the convergence of the fe and fd solutions towards the exact analytic solution can be quantified by systematically increasing the spatial resolution of the numerical solution schemes .
we find that for the fe solution the percentage errors in thickness and curvature deformations in eqs .
( [ percentageerror ] ) and ( [ percentageerror2 ] ) , which are summed over the entire solution domain , monotonically decrease with decreasing average edge size of the fe mesh , @xmath235 [ see fig . [
fig : cyl_convergence](a ) ] .
in particular , the thickness deformation error decreases approximately quadratically with decreasing @xmath235 , while the curvature deformation error decreases approximately linearly with decreasing @xmath235 . as shown in fig .
[ fig : cyl_convergence](b ) ( red solid curve ) , the error in the thickness deformation energy decreases quadratically with decreasing @xmath235 .
while the results in fig . [ fig : cyl_convergence ] were obtained at zero membrane tension , we find that membranes at finite tension yield similar scaling of the errors in the fe thickness deformations , the fe curvature deformations , and the fe thickness deformation energy with @xmath235 . for the fd solution ( red solid curve in fig .
[ fig : cyl_convergence2 ] ) the error in the thickness deformation energy decreases approximately linearly with decreasing lattice spacing @xmath14 .
the central - difference laplacian fd stencil we used here is second - order accurate .
the linear convergence is therefore most likely an indication that the fd error is dominated by the enforcement of the slope boundary conditions .
furthermore , we find that , for the parameter values used for fig .
[ fig : cyl_convergence2 ] , the percentage error in the thickness deformation energy is smaller than @xmath240 for resolutions @xmath241 nm .
thus , we find that the fe and fd solutions both yield good agreement with the exact analytic solution for high enough spatial resolutions of the numerical solution schemes .
however , figs .
[ fig : cyl_convergence](b ) and [ fig : cyl_convergence2 ] also show that , compared to the fd solution procedure , the fe solution procedure yields accurate results even at relatively low average spatial resolutions , and converges more rapidly towards the exact analytic result for the thickness deformation energy with increasing average spatial resolution .
this suggests that the fe solution procedure is more efficient and , for a given average spatial resolution , more accurate than the fd scheme used here . in eq .
( [ eq : defetag ] ) , versus lattice spacing , @xmath14 , for a system consisting of two identical cylindrical membrane proteins in the non - interacting regime ( @xmath237 nm for the fd solution ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath220 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] . ] for cylindrical membrane proteins of the same hydrophobic thickness , our analytic and numerical solution schemes imply that , for the bilayer and protein parameter values used here , bilayer thickness - mediated interactions are strongly favorable for protein center - to - center distances smaller than @xmath242 nm ( see fig .
[ fig : cylinders ] ) . for intermediate protein separations
@xmath243@xmath244 nm , thickness - mediated interactions are weakly unfavorable .
we find that thickness - mediated interactions are practically negligible for protein separations greater than @xmath245 nm or minimum protein edge - to - edge separations greater than @xmath246 .
, for two cylindrical membrane proteins versus center - to - center protein distance , @xmath3 , calculated analytically at @xmath247 , @xmath248 , @xmath249 , and @xmath239 in eq .
( [ gensol ] ) , and numerically using our fe and fd solution procedures .
the inset shows the percentage difference in thickness deformation energy , @xmath228 in eq .
( [ eq : defetag ] ) , between the analytic solution at @xmath239 and the corresponding results of fe ( blue dashed curve ) and fd ( orange solid curve ) calculations . for the fe solution we used @xmath250 nm and for the fd solution we used @xmath230 nm . we set @xmath220 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] . ]
the non - monotonic dependence of bilayer thickness - mediated interactions on protein separation ( fig .
[ fig : cylinders ] ) can be understood @xcite by considering how the bilayer thickness deformations due to single isolated membrane proteins would interfere . as noted in sec .
[ secsinglecylind ] , the thickness deformation profile induced by a single cylindrical membrane protein relaxes rapidly away from the protein , but overshoots with respect to the unperturbed lipid bilayer thickness @xcite [ fig .
[ fig : single_cyl](a ) ] .
indeed , the zeroth - order modified bessel functions of the second kind appearing in the general solution in eq .
( [ gensol ] ) imply successive expansion and compression zones of the lipid bilayer thickness around each protein .
when two identical cylindrical membrane proteins are in close proximity to each other , bilayer thickness deformation zones of the same sign overlap and the overall bilayer deformation footprint of the interacting proteins is strongly reduced compared to the large-@xmath3 limit , resulting in strongly favorable thickness - mediated protein interactions .
in contrast , for intermediate protein separations , the protein - induced bilayer thickness deformation profiles are out of phase , yielding substantial overlap of compressed and expanded bilayer regions .
this results in frustration of bilayer thickness deformations , and produces weakly unfavorable protein interactions . as discussed in sec .
[ secanalyticsol ] , the analytic solution of bilayer thickness - mediated protein interactions in eq .
( [ gensol ] ) is , in practice , only obtained up to some finite order @xmath251 . while modes with @xmath252 are irrelevant at large protein separations , modes with non - zero @xmath253 are essential to correctly account for the thickness deformations induced by interacting proteins . to confirm the validity of the truncated expansion in eq .
( [ gensol ] ) , we compare the analytic solution of thickness - mediated protein interactions to high - resolution numerical solutions obtained by fe and fd methods ( fig . [
fig : cylinders ] ) . at low orders of the analytic solution , @xmath247 and @xmath248 , our analytic estimates of the thickness - mediated interaction energy reproduce the qualitative features of the numerical interaction potentials , but exceed the numerical results by several @xmath2 . increasing the order of the analytic solution to @xmath249 and @xmath239
, we obtain convergence of the analytic solution even for small values of @xmath3 .
we find that these high - order analytic solutions are in good quantitative agreement with the corresponding fe solution .
the agreement between fe and high - order analytic solutions is approximately independent of protein separation [ fig .
[ fig : cylinders](inset ) ] .
in particular , fe and analytic approaches agree similarly well in the non - interacting regime , for which only zeroth - order modes of the analytic solution must be considered , and in the strongly interacting regime for which , in principle , and infinite number of higher - order modes should be considered in the analytic solution .
this suggests that the fe and high - order analytic solutions correctly account for thickness - mediated protein interactions even at very small @xmath3 .
furthermore , we find that , in the strongly interacting regime , the discrepancy between fe and high - order analytic results for the thickness deformation energy decreases approximately quadratically with decreasing @xmath235 [ blue dashed curve in fig .
[ fig : cyl_convergence](b ) ] , in agreement with the corresponding result obtained for non - interacting membrane proteins [ red solid curve in fig .
[ fig : cyl_convergence](b ) ] .
while the results in figs . [ fig : cyl_convergence](b ) and [ fig : cylinders ] were obtained with @xmath220 , we find similar agreement between fe and high - order analytic solutions for finite membrane tensions .
in contrast to the fe solution procedure , the discrepancy between fd and high - order analytic solutions tends to increase with decreasing @xmath3 [ fig .
[ fig : cylinders](inset ) ] . this can be understood by noting that , in the fd scheme , very small lattice spacings are required to resolve protein - induced lipid bilayer deformations in the strongly interacting regime . as in the case of non - interacting proteins ( red solid curve in fig .
[ fig : cyl_convergence2 ] ) , the discrepancy between fd and high - order analytic results for the thickness deformation energy decreases approximately linearly with decreasing lattice spacing ( blue dashed curve in fig .
[ fig : cyl_convergence2 ] ) . as in sec .
[ secsinglecylind ] this points to approximate enforcement of slope boundary conditions as the likely dominant source of error in the fd solutions
. moreover , fig .
[ fig : cyl_convergence2 ] shows that , consistent with fig .
[ fig : cylinders ] , the fd scheme produces larger discrepancies with fe and high - order analytic solutions at smaller @xmath3 than larger @xmath3 , independent of the lattice spacing considered . for instance , for small @xmath3 a too coarse lattice spacing @xmath254 nm can yield errors in the thickness deformation energy @xmath255 @xmath2 , while the same lattice spacing only produces errors @xmath256 @xmath2 at large protein separations .
in contrast , the convergence of fe and high - order analytic solutions with decreasing @xmath235 is not diminished in the interacting regime compared to the non - interacting regime [ fig .
[ fig : cyl_convergence](b ) ] .
the cylinder model of integral membrane proteins @xcite provides a beautiful `` zeroth - order '' description of thickness - mediated protein interactions , but is not able to capture the discrete symmetries and distinct hydrophobic shapes of membrane proteins suggested by membrane structural biology . as discussed in sec .
[ seccrown ] , a straightforward way to account for rotational asymmetry of the hydrophobic surface of membrane proteins is to allow for angular variations in protein hydrophobic thickness while maintaining a circular protein cross section @xcite , resulting in the crown model of membrane proteins [ fig .
[ figillust](b ) ] .
angular variations in protein hydrophobic thickness yield rotationally asymmetric distributions of compression and expansion zones about the protein . for two or more such proteins in close enough proximity ,
the anisotropy of overlapping deformation fields produces directionality of thickness - mediated protein interactions @xcite , which is expected to affect protein organization and function . , versus center - to - center protein distance , @xmath3 , calculated analytically at @xmath239 , and numerically using fe and fd schemes , for the minus - minus configuration [ @xmath257 and @xmath258 in eq .
( [ varu ] ) ] , the plus - minus configuration [ @xmath257 and @xmath259 in eq .
( [ varu ] ) ] , and a configuration intermediate between the minus - minus and plus - minus configurations [ @xmath257 and @xmath260 in eq .
( [ varu ] ) ] .
we only consider here fd solutions for the mirror - symmetric minus - minus protein configuration .
the inset shows the percentage difference in thickness deformation energy , @xmath228 in eq .
( [ eq : defetag ] ) , between the analytic result at @xmath239 and the corresponding results of the fe [ blue curves ; minus - minus ( solid ) , intermediate ( short dashed ) , and plus - minus ( long dashed ) configurations ] and fd ( orange triangles with @xmath261 at @xmath238 nm ) calculations .
we used @xmath250 nm for the fe and @xmath230 nm for the fd solutions , and set @xmath220 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] . ]
we used our analytic and numerical solution procedures to determine the thickness deformation energy associated with two crown shapes in the `` minus - minus configuration '' [ @xmath257 and @xmath258 in eq .
( [ varu ] ) ] , in which protein boundary regions with minimal hydrophobic thickness face each other at the point of closest protein edge - to - edge separation , in the `` plus - minus configuration '' [ @xmath257 and @xmath259 in eq .
( [ varu ] ) ; see fig .
[ figillust](b ) ] , in which protein boundary regions with maximal and minimal hydrophobic thickness face each other at the point of closest protein edge - to - edge separation , and in a protein configuration corresponding to @xmath257 and @xmath260 in eq .
( [ varu ] ) , which is intermediate between minus - minus and plus - minus configurations ( see fig . [
fig : crowns ] ) .
we find that , in the non - interacting regime , fe and exact analytic solutions are in good quantitative agreement , with the discrepancy in thickness deformation energy @xmath262 for the parameter values used in fig .
[ fig : crowns ] [ see fig . [
fig : crowns](inset ) ] .
in contrast , our fd solution yields more substantial discrepancies with the exact analytic solution . we attribute this to the difficulty of accurately and unambiguously imposing complicated boundary conditions in the fd scheme . in the interacting regime
, we find that analytic , fe , and fd solutions predict the same basic qualitative properties of bilayer thickness - mediated interactions between crown shapes ( fig .
[ fig : crowns ] ) .
depending on relative protein orientation , thickness - mediated interactions can switch from being strongly favorable to being strongly unfavorable at small protein separations .
in particular , when the periodic undulations of the thickness deformations induced by the two proteins are in phase in the membrane region separating the two proteins , as in the case of the minus - minus configuration in fig .
[ fig : crowns ] , similar patterns of protein - induced compression and expansion zones of the lipid bilayer overlap for small @xmath3 , yielding strongly favorable interactions .
in contrast , when the thickness undulations induced by the two proteins are out of phase , as in the case of the plus - minus configuration in fig .
[ fig : crowns ] , there is substantial overlap of out - of - phase compression and expansion zones for small @xmath3 , resulting in frustration of bilayer thickness deformations and strongly unfavorable bilayer thickness - mediated interactions . as the relative protein orientation is changed continuously from in - phase to out - of - phase configurations , the interaction potentials change smoothly @xcite from being favorable to being unfavorable at small @xmath3 which , as in the case of the intermediate configuration in fig .
[ fig : crowns ] , can produce a minimum in the thickness - mediated interaction energy at a characteristic protein separation . on a quantitative level
, we find that our fe results on bilayer thickness - mediated interactions between crown shapes are generally in good quantitative agreement with our analytic solution ( fig .
[ fig : crowns ] ) .
in contrast , the fd solution procedure yields more substantial discrepancies with our analytic results .
in particular , analytic and fe solutions are in good quantitative agreement for all configurations in fig .
[ fig : crowns ] with @xmath263 nm , with a discrepancy in thickness deformation energy @xmath264 .
we find a similar level of quantitative agreement between analytic and fe solutions even for the smallest protein separations considered for the minus - minus configuration in fig .
[ fig : crowns ] . )
( @xmath233 ; orange dashed curve ) and percentage error in curvature deformations in eq .
( [ percentageerror2 ] ) ( @xmath234 ; magenta solid curve ) versus @xmath235 for two crown shapes in the minus - minus configuration in the interacting regime at @xmath265 nm .
the error is evaluated using the corresponding analytic result at @xmath239 .
errors in the thickness mismatch and curvature deformations decay as @xmath235 and @xmath236 , respectively .
( b ) percentage difference between analytic and fe results for the thickness deformation energy per protein , @xmath228 in eq .
( [ eq : defetag ] ) , versus @xmath235 for two crown shapes in the minus - minus configuration in the non - interacting regime ( @xmath237 nm for the fe solution ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath266 @xmath267 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
[ fig : crown_convergence],title="fig : " ] + ) ( @xmath233 ; orange dashed curve ) and percentage error in curvature deformations in eq .
( [ percentageerror2 ] ) ( @xmath234 ; magenta solid curve ) versus @xmath235 for two crown shapes in the minus - minus configuration in the interacting regime at @xmath265 nm . the error is evaluated using the corresponding analytic result at @xmath239 .
errors in the thickness mismatch and curvature deformations decay as @xmath235 and @xmath236 , respectively .
( b ) percentage difference between analytic and fe results for the thickness deformation energy per protein , @xmath228 in eq .
( [ eq : defetag ] ) , versus @xmath235 for two crown shapes in the minus - minus configuration in the non - interacting regime ( @xmath237 nm for the fe solution ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath266 @xmath267 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
[ fig : crown_convergence],title="fig : " ] the discrepancy between analytic and fe results is most pronounced at very small @xmath3 in the strongly unfavorable regime of bilayer thickness - mediated interactions , @xmath268 nm , for the plus - minus and intermediate configurations in fig . [ fig : crowns ] , where the interaction energy diverges .
however , even in this regime the percentage difference between analytic and fe results is @xmath269 for the plus - minus and intermediate configurations in fig .
[ fig : crowns ] , for which the thickness deformation energy can exceed @xmath270 @xmath2 for the smallest value @xmath271 nm we allow here .
we find that in the strongly unfavorable regime @xmath268 nm the magnitude of the gradient of @xmath15 can exceed @xmath272 .
the maximum value of @xmath273 in the strongly interacting regime is therefore substantially greater than the maximum magnitude of the gradient of bilayer thickness deformations @xmath274 in the non - interacting regime of the crown shapes considered here , which we take to induce large gradients of the thickness deformation field so as to test the mathematical limits of applicability of our analytic and numerical solution procedures ( see sec .
[ seccrown ] ) .
the more pronounced discrepancy between fe and analytic results in the strongly unfavorable regime of bilayer thickness - mediated interactions between crown shapes arises because , in this regime , bilayer thickness deformations show a strong variation over the small membrane region separating the two proteins .
high accuracy in the strongly unfavorable regime of thickness - mediated interactions therefore requires highly refined meshes in the fe scheme and , to capture pronounced angular variations in the thickness deformation field , a large value of @xmath150 in the analytic approach . in eq .
( [ eq : defetag ] ) , versus lattice spacing , @xmath14 , for two crown shapes in the plus - plus configuration in the non - interacting regime ( @xmath237 nm for the fd solution ; red solid curve ) and in the strongly interacting regime ( @xmath238 nm ; blue dashed curve ; @xmath239 for the analytic solution ) .
we set @xmath220 .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] . ] following sec .
[ seccylinder2 ] we quantify the discrepancy between numerical and analytic solutions by systematically increasing the spatial resolution of the numerical solutions . as in the case of cylindrical membrane proteins
we find that , for the fe solution procedure , the errors in thickness deformations and thickness deformation energy decrease approximately quadratically with decreasing average edge size of the fe mesh , while the error in curvature deformations decreases approximately linearly with @xmath235 ( see fig . [
fig : crown_convergence ] ) .
we find similar scaling of the discrepancy between fe and analytic solutions with @xmath235 in the interacting and non - interacting regimes , as well as for zero and finite membrane tensions . as for cylindrical membrane proteins , the error in the thickness deformation energy obtained from the fd solution procedure decreases approximately linearly with decreasing lattice spacing in the interacting and non - interacting regimes ( see fig .
[ fig : crown_convergence2 ] ) , which again points to approximate enforcement of slope boundary conditions as the likely dominant source of error in the fd solutions . furthermore , fig .
[ fig : crown_convergence2 ] suggests that , similar to the case of cylindrical membrane proteins , the fd scheme generally produces larger discrepancies with high - order analytic solutions at smaller @xmath3 than larger @xmath3 .
in contrast , the convergence of fe and high - order analytic solutions is not diminished in the interacting regime compared to the non - interacting regime [ fig .
[ fig : crown_convergence](b ) ]
. finally , comparison of figs .
[ fig : cyl_convergence ] and [ fig : crown_convergence ] , and figs .
[ fig : cyl_convergence2 ] and [ fig : crown_convergence2 ] , shows that , for a given spatial resolution , the discrepancy between fe and fd solutions and the analytic solution is more pronounced for crown shapes than for cylindrical membrane proteins .
the clover - leaf model of integral membrane proteins discussed in sec .
[ secclover ] can be used to capture non - circular bilayer - protein boundary curves [ fig .
[ figillust](c ) ] , and provides a generalization of the cylinder model of membrane proteins complementary to the crown model . in particular , to model the discrete symmetries of integral membrane proteins suggested by membrane structural biology , the clover - leaf model allows for periodic modulations in the shape of the hydrophobic cross section of membrane proteins . as a result ,
the bilayer thickness deformations induced by clover - leaf proteins show a characteristic pattern of compression and expansion zones about the protein @xcite , which provides a simple description of the effect of protein shape and oligomeric state on lipid bilayer thickness deformations .
furthermore , for two clover - leaf proteins in close enough proximity , bilayer thickness - mediated interactions are directional @xcite and bear characteristic signatures of protein shape , symmetry , and orientation . , versus center - to - center protein distance , @xmath3 , calculated analytically at @xmath239 , and numerically using fe and fd schemes , for the face - on configuration [ @xmath257 and @xmath258 in eq .
( [ boundcclover ] ) ] and the tip - on configuration [ @xmath275 and @xmath259 in eq .
( [ boundcclover ] ) ] .
( b ) thickness - mediated interaction energy @xmath276 versus @xmath3 obtained by subtracting the protein - induced thickness deformation energies in the non - interacting regime from the respective perturbative analytic , fe , and fd solutions in panel ( a ) .
the inset shows the difference in the thickness - mediated interaction energies obtained from the fe solution procedure , and the analytic and fd solution procedures .
we use the same labeling conventions for panel ( b ) as for panel ( a ) . for our numerical calculations we used @xmath277 nm for the fe and @xmath230 nm for the fd solutions .
we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] + , versus center - to - center protein distance , @xmath3 , calculated analytically at @xmath239 , and numerically using fe and fd schemes , for the face - on configuration [ @xmath257 and @xmath258 in eq .
( [ boundcclover ] ) ] and the tip - on configuration [ @xmath275 and @xmath259 in eq .
( [ boundcclover ] ) ] .
( b ) thickness - mediated interaction energy @xmath276 versus @xmath3 obtained by subtracting the protein - induced thickness deformation energies in the non - interacting regime from the respective perturbative analytic , fe , and fd solutions in panel ( a ) .
the inset shows the difference in the thickness - mediated interaction energies obtained from the fe solution procedure , and the analytic and fd solution procedures .
we use the same labeling conventions for panel ( b ) as for panel ( a ) . for our numerical calculations we used @xmath277 nm for the fe and @xmath230 nm for the fd solutions .
we set @xmath220 for both panels .
all model parameters were chosen as described in sec .
[ secelasticmodel ] , and analytic and numerical solutions were obtained as discussed in secs .
[ secanalyticsol ] and [ secnumericalsol ] .
, title="fig : " ] to compare our analytic and numerical solutions in the case of clover - leaf shapes , we consider the bilayer thickness deformation energies associated with two clover - leaf proteins in the `` face - on configuration '' [ @xmath257 and @xmath258 in eq .
( [ boundcclover ] ) ] and the `` tip - on configuration '' [ @xmath275 and @xmath259 in eq .
( [ boundcclover ] ) ] using analytic , fe , and fd solution procedures [ see fig . [
fig : clover5](a ) ] .
we find that the thickness deformation energies obtained using our fe and fd solution procedures are in agreement within the numerical accuracy of the numerical solution schemes .
in contrast , the thickness deformation energies obtained via the perturbative analytic solution procedure differ substantially from the fe and fd solutions , in the non - interacting as well as interacting regimes in fig . [
fig : clover5](a ) .
some discrepancy between analytic and numerical results is to be expected , given that somewhat different boundary value problems are solved in the perturbative analytic and numerical approaches , with the analytic solution only being first order in the perturbation parameter @xmath72 in eq .
( [ boundcclover ] ) .
this produces a systematic error in the thickness deformation energy obtained through the perturbative analytic approach , which yields disagreement between perturbative analytic and numerical solution procedures even in the non - interacting regime . as far as bilayer thickness - mediated interactions between clover - leaf shapes
are concerned [ see fig . [
fig : clover5](b ) ] , the discrepancy between analytic and numerical results is most pronounced at small @xmath3 , where the interaction potentials obtained from perturbative analytic and numerical approaches can differ by @xmath70 @xmath2 .
this can be understood intuitively by noting that the analytic calculation of thickness - mediated interactions between clover - leaf shapes relies on a perturbative mapping of clover - leaf boundary curves with constant hydrophobic thickness onto circular boundary curves with varying hydrophobic thickness [ see eqs .
( [ genbc1trimerfin])([genbc2trimerfin2 ] ) ] .
we use here a first - order perturbative mapping , which is expected to become increasingly inaccurate at small protein separations since , as @xmath3 is being decreased , the structure of thickness deformations in close proximity to the clover - leaf proteins comes to dominate the interaction energy .
we therefore attribute the disagreement between analytic and numerical approaches in fig . [
fig : clover5](b ) to shortcomings of the analytic approach .
note , however , that the first - order perturbative analytic solution and the corresponding numerical solutions yield the same basic scenario for the competition between the different orientations of pentameric clover - leaf shapes considered in fig .
[ fig : clover5 ] .
indeed , similar agreement between first - order perturbative and numerical approaches is obtained for other clover - leaf shapes and orientations @xcite , which suggests that the first - order perturbative approach can accurately capture the directionality of thickness - mediated interactions between integral membrane proteins with clover - leaf shapes .
a wide range of experiments indicate @xcite that protein - induced lipid bilayer thickness deformations can play a crucial role in the regulation of protein function through bilayer material properties and bilayer - mediated protein interactions .
cell membranes are generally crowded with membrane proteins @xcite , suggesting that the protein center - to - center distance @xmath3 is typically small _ in vivo _ , while modern structural biology suggests a rich picture of membrane protein shape with great diversity in the oligomeric states and symmetries of membrane proteins .
motivated by these experimental observations , we have developed a combined analytic and numerical framework @xcite which allows prediction of the protein - induced lipid bilayer thickness deformations implied by the classic model in eq .
( [ energy ] ) for arbitrary @xmath3 and the protein shapes suggested by structural studies .
our analytic solution procedure , which is based on refs .
@xcite , allows exact solutions of the protein - induced lipid bilayer thickness deformation field and elastic thickness deformation energy for proteins with constant or varying hydrophobic thickness in the ( strongly ) interacting as well as non - interacting regimes , provided that the proteins have circular transmembrane cross sections . through a perturbative approach ,
our analytic solution scheme can also account for non - circular protein cross sections . following similar steps as in refs .
@xcite , our analytic solution procedure is readily applied @xcite to calculate curvature - mediated protein interactions at arbitrary protein separations .
the exact analytic solutions described here are in excellent quantitative agreement with our numerical solutions for arbitrary protein orientations and arbitrary @xmath3 with the exception of the strongly unfavorable regime of bilayer thickness - mediated interactions , for which the interaction energy diverges as the edge - to - edge protein separation approaches zero .
we regard this regime as being of limited physical significance because , for proteins of distinct hydrophobic thickness , the leading - order model in eq .
( [ energy ] ) is expected to break down at small @xmath3 due to the large gradients of protein - induced lipid bilayer deformations in the bilayer region separating the two proteins . in principle , higher - order analytic solutions than those considered here could be developed to access thickness - mediated interactions in this regime . for proteins of non - circular cross section ,
our comparisons between analytic and numerical solution procedures show that the first - order perturbative solution can accurately capture the dependence of the thickness deformation energy on protein oligomeric state @xcite , as well as the directionality of bilayer thickness - mediated protein interactions @xcite .
however , the first - order perturbative approach does not yield the exact value of the thickness deformation energy .
the discrepancy between our perturbative analytic and numerical solutions of bilayer thickness - mediated protein interactions is most pronounced at small @xmath3 , where the perturbative analytic approach is expected to break down .
we have developed both fd and fe solution schemes to numerically calculate the lipid bilayer thickness deformations induced by membrane proteins .
the fd scheme has the advantage of being conceptually simple and relatively straightforward to implement .
we find that our fd scheme accurately accounts for the lipid bilayer thickness deformations induced by cylindrical membrane proteins in the non - interacting regime and can also capture , albeit with less accuracy , bilayer thickness - mediated interactions between cylindrical membrane proteins .
however , the convergence of the numerical solutions to the exact analytic solutions is slower for the fd scheme than for the fe scheme , most likely due to errors in the fd slope boundary conditions at the protein surface . furthermore , we find that the fd solution procedure can introduce substantial numerical errors for non - cylindrical membrane proteins .
these errors are particularly pronounced in the interacting regime .
in contrast , we find that the fe scheme described here yields rapid numerical convergence for all available exact analytic solutions of the minima of eq .
( [ energy ] ) @xcite .
the convergence properties of the fe scheme do not seem to be diminished substantially in the interacting regime compared to the non - interacting regime of bilayer thickness - mediated protein interactions , or by complicated boundary shapes and boundary conditions .
the combined presence of both first and second derivatives in the energy in eq .
( [ energy ] ) places special demands on the fe formulation .
in particular , while standard lagrange interpolation functions @xcite are adequate to compute the thickness stretch and gradient terms in eq .
( [ energy ] ) , they fail to produce conforming curvatures at element interfaces @xcite .
our fe solution procedure overcomes this challenge by combining @xcite lagrange shape functions for the thickness stretch and gradient terms with the dkt method @xcite for curvature deformations .
the resulting fe approach is computationally efficient and allows accurate solutions of the complicated boundary value problems posed by many interacting membrane proteins .
indeed , we have shown previously @xcite that our fe approach permits calculation of directional thickness - mediated protein interactions in systems composed of hundreds of integral membrane proteins for arbitrary protein separations and orientations using protein shapes suggested by membrane structural biology .
the combined analytic and numerical framework described here shows that the shape of integral membrane proteins , and resulting structure of lipid bilayer thickness deformations , can play a crucial role in the regulation of protein function by lipid bilayers @xcite , and that bilayer thickness - mediated interactions between integral membrane proteins are strongly directional and dependent on protein shape @xcite .
taken together , our results suggest that , in addition to bilayer - protein hydrophobic mismatch @xcite , protein shape may be a crucial determinant of membrane protein regulation by lipid bilayers and bilayer - mediated protein interactions . the classic model of protein - induced lipid bilayer thickness deformations in eq . ( [ energy ] ) and modifications thereof have been found to capture the basic experimental phenomenology of bilayer - protein interactions in a wide range of experimental systems @xcite , only involve parameters which can be measured directly in experiments , and are simple enough to allow analytic solutions .
analogous models have been formulated @xcite to describe protein - induced curvature deformations @xcite and fluctuation - mediated interactions @xcite . in general , thickness- , curvature- , and fluctuation - mediated interactions all contribute to bilayer - mediated interactions between integral membrane proteins , but the relative strengths of these interactions depend on the specific experimental system under consideration .
in addition to bilayer - mediated interactions , membrane proteins may , in principle , also interact via electrostatic forces . however , electrostatic interactions in aqueous environments are generally screened @xcite and charged protein residues are typically excluded from the transmembrane regions of membrane proteins @xcite . the model of protein - induced lipid bilayer thickness deformations in eq .
( [ energy ] ) and the corresponding `` zeroth - order '' models @xcite capturing curvature- and fluctuation - mediated protein interactions absorb the molecular details of lipids and membrane proteins into effective material parameters . to provide a more detailed description of bilayer - protein interactions , a number of extensions and refinements of these models
have been developed @xcite .
for instance , the effect of bilayer - protein interactions on the elastic properties of lipid bilayers can be captured @xcite by allowing for spatial variations in the values of the elastic bilayer parameters .
furthermore , the microscopic roughness of lipid bilayers due to area fluctuations can affect @xcite protein - induced lipid bilayer deformations .
these model refinements yield bilayer thickness deformation profiles which are quantitatively but not qualitatively different from those implied by eq .
( [ energy ] ) [ see , for instance , fig . [
fig : single_cyl](a ) ] .
however , additional structural properties of the lipid bilayer , such as lipid tilt @xcite , may have a more substantial effect on bilayer - mediated protein interactions .
moreover , integral membrane proteins may tilt to reduce hydrophobic mismatch , as suggested by monte carlo and molecular dynamics simulations @xcite .
tilting of membrane proteins is expected to be most pronounced for small membrane proteins with only a single transmembrane @xmath278-helix . while protein tilting generally competes with protein - induced lipid bilayer thickness deformations as a mechanism for alleviating bilayer - protein hydrophobic mismatch , experiments have suggested @xcite that protein tilting is in general too weak to fully offset a hydrophobic mismatch between membrane proteins and the surrounding lipid bilayer . in this article
we have focused on bilayer thickness - mediated interactions between two integral membrane proteins . in general , more than two membrane proteins are expected to interact in the crowded membrane environments provided by living cells . we have shown previously @xcite that , in contrast to curvature- and fluctuation - mediated protein interactions @xcite , bilayer thickness - mediated protein interactions are approximately pairwise additive , at least for large enough protein separations . for small protein separations ,
non - pairwise contributions to bilayer thickness - mediated interactions between integral membrane proteins can modify the interaction strength by @xmath279 @xcite .
however , except in special cases @xcite , non - pairwise contributions to bilayer thickness - mediated protein interactions do not alter how bilayer thickness - mediated interactions vary with the shape and arrangement of proteins .
the approximate pairwise additivity of bilayer thickness - mediated protein interactions presents a considerable simplification @xcite for the mathematical analysis of systems composed of many interacting integral membrane proteins .
recent breakthroughs in superresolution light microscopy and electron cryo - tomography have revealed that integral membrane proteins can form large clusters with intricate translational and orientational protein ordering , which provides @xcite a general mechanism for cells to modulate protein function through cooperative interactions and local modification of bilayer mechanical properties .
but , to date , the physical mechanisms giving rise to the observed lattice architectures and collective functions of membrane protein clusters remain largely unknown .
the directionality of bilayer thickness - mediated protein interactions implied by the observed protein structures presents one possible physical mechanism for membrane protein organization and collective function .
such directional interactions can yield @xcite ordering of integral membrane proteins , which is also consistent with molecular dynamics simulations @xcite .
the combined analytic and numerical framework we have discussed here allows calculation of the lipid bilayer - mediated protein interactions implied by bilayer elasticity theory @xcite for the protein shapes suggested by structural studies at arbitrary protein separations and orientations .
our framework thus presents a step towards a general physical theory of how directional bilayer - mediated protein interactions affect the molecular structure , organization , and biological function of proteins in the crowded membrane environments provided by living cells .
|
hydrophobic thickness mismatch between integral membrane proteins and the surrounding lipid bilayer can produce lipid bilayer thickness deformations .
experiment and theory have shown that protein - induced lipid bilayer thickness deformations can yield energetically favorable bilayer - mediated interactions between integral membrane proteins , and large - scale organization of integral membrane proteins into protein clusters in cell membranes . within the continuum elasticity theory of membranes , the energy cost of protein - induced bilayer
thickness deformations can be captured by considering compression and expansion of the bilayer hydrophobic core , membrane tension , and bilayer bending , resulting in biharmonic equilibrium equations describing the shape of lipid bilayers for a given set of bilayer - protein boundary conditions . here
we develop a combined analytic and numerical methodology for the solution of the equilibrium elastic equations associated with protein - induced lipid bilayer deformations .
our methodology allows accurate prediction of thickness - mediated protein interactions for arbitrary protein symmetries at arbitrary protein separations and relative orientations .
we provide exact analytic solutions for cylindrical integral membrane proteins with constant and varying hydrophobic thickness , and develop perturbative analytic solutions for non - cylindrical protein shapes .
we complement these analytic solutions , and assess their accuracy , by developing both finite element and finite difference numerical solution schemes .
we provide error estimates of our numerical solution schemes and systematically assess their convergence properties . taken together
, the work presented here puts into place an analytic and numerical framework which allows calculation of bilayer - mediated elastic interactions between integral membrane proteins for the complicated protein shapes suggested by structural biology and at the small protein separations most relevant for the crowded membrane environments provided by living cells .
|
our era has witnessed the massive explosion of data and a dramatic improvement of technology in collecting and processing large data sets .
we often encounter huge data sets that the number of features greatly surpasses the number of observations .
it makes many traditional statistical analysis tools infeasible and poses great challenge on developing new tools .
regularization methods have been widely used for the analysis of high - dimensional data .
these methods penalize the least squares or the likelihood function with the @xmath0-penalty on the unknown parameters ( lasso , @xcite ) , or a folded concave penalty function such as the scad @xcite and the mcp@xcite .
however , these penalized least - squares methods are sensitive to the tails of the error distributions , particularly for ultrahigh dimensional covariates , as the maximum spurious correlation between the covariates and the realized noises can be large in those cases . as a result
, theoretical properties are often obtained under light - tailed error distributions @xcite . to tackle the problem of heavy - tailed errors , robust regularization methods have been extensively studied .
@xcite , @xcite and @xcite developed robust regularized estimators based on quantile regression for the case of fixed dimensionality .
@xcite studied the @xmath0-penalized quantile regression in high dimensional sparse models .
@xcite further considered an adaptively weighted @xmath0 penalty to alleviate the bias problem and showed the oracle property and asymptotic normality of the corresponding estimator .
other robust estimators were developed based on least absolute deviation ( lad ) regression .
@xcite studied the @xmath0-penalized lad regression and showed that the estimator achieves near oracle risk performance under the high dimensional setting .
the above methods essentially estimate the conditional _ median ( or quantile ) _ regression , instead of the conditional _ mean _ regression function . in the applications where the mean regression is of interest , these methods are not feasible unless a strong assumption is made that the distribution of errors is symmetric around zero .
a simple example is the heteroscedastic linear model with asymmetric noise distribution .
another example is to estimate the conditional variance function such as arch model @xcite . in these cases ,
the conditional mean and conditional median are very different .
another important example is to estimate large covariance matrix without assuming light - tails .
we will explain this more in details in section 5 .
in addition , lad - based methods tend to penalize strongly on small errors .
if only a small proportion of samples are outliers , they are expected to be less efficient than the least squares based method .
a natural question is then how to conduct ultrahigh dimensional mean regression when the tails of errors are not light ? how to estimate the sample mean with very fast concentration when the distribution has only bounded second moment ?
these simple questions have not been carefully studied .
lad - based methods do not intend to answer these questions as they alter the problems of the study .
this leads us to consider huber loss as another way of robustification .
the huber loss @xcite is a hybrid of squared loss for relatively small errors and absolute loss for relatively large errors , where the degree of hybridization is controlled by one tuning parameter . unlike the traditional huber loss , we allow the regularization parameter to diverge ( or converge if its reciprocal is used ) in order to reduce the bias induced by the huber loss for estimating conditional mean regression function . in this paper ,
we consider the regularized approximate quadratic ( ra - lasso ) estimator with an @xmath0 penalty and show that it admits the same @xmath1 error rate as the optimal error rate in the light - tail situation .
in particular , if the distribution of errors is indeed symmetric around 0 ( where the median and mean agree ) , this rate is the same as the regularized lad estimator obtained in @xcite .
therefore , the ra - lasso estimator does not lose efficiency in this special case . in practice , since the distribution of errors is unknown , ra - lasso is more flexible than the existing methods in terms of estimating the conditional mean regression function .
a by - product of our method is that the ra - lasso estimator of the population mean has the exponential type of concentration even in presence of the finite second moment .
@xcite studied this type of problem and proposed a class of losses to result in a robust @xmath2-estimator of mean with exponential type of concentration .
we further extend his idea to the sparse linear regression setting . as done in many other papers , estimators with nice sampling properties
are typically defined through the optimization of a target function such as the penalized least - squares .
the properties that are established are not the same as the ones that are computed . following the framework of @xcite , we propose the composite gradient descent algorithm for solving the ra - lasso estimator and develop the sampling properties by taking computational error into consideration .
we show that the algorithm indeed produces a solution that admits the same optimal @xmath1 error rate as the theoretical estimator after sufficient number of iterations .
this paper is organized as follows .
first , in section [ sec2 ] , we introduce the ra - lasso estimator and show that it has the same @xmath1 error rate as the optimal rate under light - tails . in section [ sec3 ] , we study the property of the composite gradient descent algorithm for solving our problem and show that the algorithm produces a solution that performs as well as the solution in theory . in section [ sec4 ]
, we show the connection between huber loss and catoni loss and establish an concentration inequality for robust estimation of mean .
the estimation of the error s variance is investigated in section [ sec5 ] .
numerical studies are given in section [ sec6 ] and [ sec7 ] to compare our method with two competitors .
all technical proofs are presented in section [ sec : proofs ] .
we consider the linear regression model @xmath3 where @xmath4 are independent and identically distributed ( i.i.d ) @xmath5-dimensional covariate vectors , @xmath6 are i.i.d errors , and @xmath7 is a @xmath5-dimensional regression coefficient vector .
we consider the high - dimensional setting , where @xmath8 for some constant @xmath9 .
we assume the distributions of @xmath10 and @xmath11 are independent and both have mean 0 . under this assumption , @xmath12 is the mean effect of @xmath13 conditioning on @xmath10 , which is assumed to be of interest . to adapt for different magnitude of errors and robustify the estimation
, we propose to use the huber loss @xcite : @xmath14 the huber loss is quadratic for small values of @xmath15 and linear for large values of @xmath15 .
the parameter @xmath16 controls the blending of quadratic and linear penalization .
the least squares and the lad can be regarded as two extremes of the huber loss for @xmath17 and @xmath18 , respectively . deviated from the traditional huber s estimator
, the parameter @xmath16 converges to zero in order to reduce the biases of estimating the mean regression function when the conditional distribution of @xmath19 is not symmetric . on the other hand
, @xmath16 can not shrink too fast in order to maintain the robustness . in this paper , we regard @xmath16 as a tuning parameter , whose optimal value will be discussed later in this section . in practice
, @xmath16 needs to be tuned by some data - driven method . by letting @xmath16 vary , we call @xmath20 the robust approximate quadratic ( ra - quadratic ) loss . to estimate @xmath12
, we propose to solve the following convex optimization problem : @xmath21 to assess the performance of @xmath22 , we study the property of @xmath23 , where @xmath24 is the euclidean norm of a vector .
when @xmath25 converges to zero sufficiently fast , @xmath22 is a natural @xmath2-estimator of @xmath26 , which is the population minimizer under the ra - quadratic loss and varies by @xmath16 . in general
, @xmath27 differs from @xmath12 .
but , since the ra - quadratic loss approximates the quadratic loss as @xmath16 tends to 0 , @xmath27 is expected to converge to @xmath12 .
this property will be established in theorem [ thm1 ] .
therefore , we decompose the statistical error @xmath28 into the approximation error @xmath29 and the estimation error @xmath30 .
the statistical error @xmath23 is then bounded by @xmath31 in the following , we give the rate of the approximation and estimation error , respectively .
we show that @xmath23 admits the same rate as the optimal rate under light tails , as long as the two tuning parameters @xmath16 and @xmath25 are properly chosen .
we first give the rate of the approximation error under some moment conditions on @xmath10 and @xmath11 .
we assume both @xmath12 and @xmath27 are interior points of an @xmath1 ball with sufficiently large radius .
[ thm1 ] it holds that @xmath32 , under the following conditions : * @xmath33 , for some @xmath34 .
* @xmath35)\leq\lambda_{\max}(\operatorname{e}[\bx\bx^t])\leq \ku<\infty$ ] , * for any @xmath36 , @xmath37 is sub - gaussian with parameter at most @xmath38 , i.e. @xmath39 , for any @xmath40 .
theorem [ thm1 ] reveals that the approximation error vanishes faster if higher moments of error distribution exists .
we next give the rate of the estimation error @xmath41 .
this part differs from the existing work regarding the estimation error of high dimensional regularized @xmath2-estimator @xcite as the population minimizer @xmath27 now varies with @xmath16 .
however , we will show that the estimation error rate does not depend on @xmath16 , given a uniform sparsity condition . in order to be solvable in the high - dimensional setting ,
@xmath12 is usually assumed to be sparse or weakly sparse , i.e. many elements of @xmath12 are zero or small . by theorem [ thm1 ]
, @xmath27 converges to @xmath12 as @xmath16 goes to 0 . in view of this fact ,
we assume that @xmath27 is uniformly weakly sparse when @xmath16 is sufficiently small .
in particular , we assume that there exists a small constant @xmath42 , such that @xmath27 belongs to an @xmath43-ball with a uniform radius @xmath44 that @xmath45 for all @xmath46 $ ] , and some @xmath47 $ ] . when the conditional distribution of @xmath19 is symmetric , @xmath48 for all @xmath16 and @xmath49 .
therefore the condition reduces to that @xmath12 is in the @xmath43 ball . in a special case where @xmath50
, it follows from theorem [ thm1 ] that if @xmath12 belongs to the @xmath0-ball with radius @xmath51 and @xmath52^{\frac{1}{k-1}}$ ] , where @xmath53 is a generic constant , then @xmath27 belongs to the @xmath0-ball with radius @xmath54 for all @xmath46 $ ] . for a general @xmath55
, we assume a uniform upper bound @xmath44 as in ( [ eq2.4 ] ) , which is allowed to diverge to infinity . since the ra - quadratic loss is convex
, we show that with high probability the estimation error @xmath56 belongs to a star - shaped set , which depends on @xmath16 and the threshold level @xmath57 of signals .
[ lem1 ] under conditions ( c1 ) and ( c3 ) , by choosing @xmath58 and @xmath59 , where @xmath60 , @xmath61 and @xmath62 are some constants , with probability greater than @xmath63 , @xmath64 where @xmath65 , @xmath57 is a positive constant , @xmath66 and @xmath67 denotes the subvector of @xmath68 with indices in set @xmath69 .
we further verify a restricted strong convexity ( rsc ) condition , which has been shown to be critical in the study of high dimensional regularized @xmath2-estimator @xcite .
let @xmath70^t\bdelta,\ ] ] where @xmath71 , @xmath68 is a @xmath5-dimensional vector and @xmath72 is the gradient of @xmath73 at the point of @xmath74 .
[ def : rsc ] the loss function @xmath73 satisfies rsc condition on a set @xmath75 with curvature @xmath76 and tolerance @xmath77 if @xmath78 next , we show that with high probability the ra - quadratic loss ( [ eq2.2 ] ) satisfies rsc for all @xmath79 with uniform constants @xmath80 and @xmath77 that do not depend on @xmath16 .
lemma [ lem2 ] is a preliminary result , based on which rsc is checked in lemma [ lem3 ] .
[ lem2 ] under conditions ( c1)-(c3 ) , for all @xmath81 , there exist uniform positive constants @xmath82 and @xmath83 , such that @xmath84 with probability at least @xmath85 for some positive constants @xmath86 and @xmath87 .
[ lem3 ] suppose conditions ( c1)-(c3 ) hold and assume that @xmath88 with the choice @xmath89 , with probability at least @xmath85 , the rsc condition holds for @xmath90 and @xmath91 with @xmath92 and @xmath93 .
lemma [ lem3 ] shows that , even though @xmath27 is unknown and the set @xmath94 depends on @xmath16 , rsc holds with uniform constants that do not depend on @xmath16 .
this further gives the following upper bound of the estimation error @xmath41 , which also does not depend on @xmath16 .
[ thm2 ] under conditions of lemma [ lem1 ] and , with probability at least @xmath95 , @xmath96^{1/2-q/4}).\ ] ] finally , theorem [ thm1 ] and [ thm2 ] together lead to the following main result , which gives the rate of the statistical error @xmath23 .
[ thm3 ] under conditions of lemma [ lem1 ] and , with probability at least @xmath95 , @xmath97^{1/2-q/4}).\ ] ] next , we compare our result with the existing results regarding the robust estimation of high dimensional linear regression model . 1 .
when the distribution of @xmath11 is symmetric around 0 , then @xmath98 for any @xmath16 , which has no approximation error .
if @xmath11 has heavy tails in addition to being symmetric , we would like to choose @xmath16 sufficiently large to robustify the estimation .
it then follows from theorem [ thm2 ] that @xmath99^{1/2-q/4})$ ] , where @xmath100 . the rate is the same as the minimax rate @xcite for weakly sparse model under the light tails . in a special case that @xmath101 , it gives @xmath102 , where @xmath103 is the number of nonzero elements in @xmath12
this is the same rate as the regularized lad estimator in @xcite and the regularized quantile estimator in @xcite .
it suggests that our method does not lose efficiency for symmetric heavy - tailed errors . 2 .
if the distribution of @xmath11 is asymmetric around 0 , the quantile and lad based methods are inconsistent , since they estimate the median instead of the mean .
theorem [ thm3 ] shows that our estimator still achieves the optimal rate given that @xmath104^{1-\frac{q}{2}}\}^{\frac{1}{2(k-1)}})$ ] even though the @xmath105-th moment of @xmath106 is assumed . recall from conditions in lemma [ lem1 ] that we also need to choose @xmath107 for some constant @xmath108 .
given the sparsity condition ( [ eq2.7 ] ) , @xmath16 can be chosen to meet the above two requirements . in terms of estimating the conditional mean effect , errors with heavy but asymmetric tails
give the case where the ra - lasso has the biggest advantage over the other estimators .
in practice , the distribution of errors is unknown .
however , we proved that our method is no worse than the existing methods for any type of errors , as long as the tuning parameters are chosen properly .
hence , our method is more flexible .
the gradient descent algorithm @xcite is usually applied to solve the convex problem ( [ eq2.3 ] ) . for example , we can replace the ra - quadratic loss with its local quadratic approximation ( lqa ) and iteratively solve the following optimization problem : @xmath109^t ( \bbeta-{\widehat}{\bbeta}^t ) + \frac{\gamma_{u}}{2 } { \lvert\bbeta-{\widehat}{\bbeta}^t\rvert_2}^2+\lambda_n { \lvert\bbeta\rvert_1}\right\},\ ] ] where @xmath110 is a fixed constant at each iteration , and the side constraint `` @xmath111 '' is introduced to guarantee good performance in the first few iterations and @xmath112 is allowed to be sufficiently large . to solve ( [ eq3.1 ] ) , the update can be computed by a two - step procedure .
we first solve ( [ eq3.1 ] ) without the norm constraint by soft - thresholding the vector @xmath113 at level @xmath25 and call the solution @xmath114 .
if @xmath115 , set @xmath116 . otherwise , @xmath117 is obtained by further project @xmath114 onto the @xmath0-ball @xmath118
. the projection can be done @xcite by soft - thresholding @xmath114 at level @xmath119 , where @xmath119 is given by the following procedure : ( 1 ) sort @xmath120 into @xmath121 ; ( 2 ) find @xmath122 and let @xmath123 .
@xcite considered the computational error of such first - order gradient descent method .
they showed that , for a convex and differentiable loss functions @xmath124 and decomposable penalty function @xmath125 , the error @xmath126 has the same rate as @xmath23 for all sufficiently large @xmath127 , where @xmath128 , and @xmath129 . different from their setup , our population minimizer @xmath27 varies by @xmath16 . nevertheless ,
as @xmath27 converges to the true effect @xmath12 , by a careful control of @xmath16 , we can still show that @xmath126 has the same rate as @xmath23 , where @xmath22 is the theoretical solution of ( [ eq2.3 ] ) and @xmath130 is as defined in ( [ eq3.1 ] ) .
the key is that the ra - quadratic loss function @xmath73 satisfies the restricted strong convexity ( rsc ) condition and the restricted smoothness condition ( rsm ) with some uniform constants , namely @xmath131 as defined in ( [ eq2.5 ] ) satisfies the following conditions : @xmath132 for all @xmath74 and @xmath68 in some set of interest , with parameters @xmath133 , @xmath134 , @xmath135 and @xmath136 that do not depend on @xmath16 .
we show that such conditions hold with high probability .
[ lem4 ] under conditions ( c1)-(c3 ) , for all @xmath137 and @xmath138 , with probability greater than @xmath85 , ( [ eq3.2 ] ) and ( [ eq3.3 ] ) hold with @xmath139 , @xmath140 , @xmath141 , @xmath142 .
we further show in theorem [ thm4 ] that , whenever @xmath143 , which is required for consistency of _ any method _ over the weak sparse @xmath43 ball by the known minimax results @xcite , it holds that @xmath144 for sufficiently many iterations with lower bound specified in theorem [ thm4 ] .
hence , @xmath145^{1/2-q/4}),\end{aligned}\ ] ] which has the same rate as @xmath23 .
hence , from a statistical point of view , there is no need to iterate beyond @xmath127 steps .
[ thm4 ] under conditions of theorem [ thm3 ] , suppose we choose @xmath25 as in lemma [ lem1 ] and also satisfying @xmath146\frac{\log p}{n},\ ] ] where @xmath147 denotes the cardinality of set @xmath148 and @xmath149 , then with probability at least @xmath150 , we have @xmath151 \right),\ ] ] for all iterations @xmath152 where @xmath153 and @xmath154 is the initial value , @xmath155 is the tolerance level , @xmath156 and @xmath106 are some constants as will be defined in ( [ eq7.21 ] ) and ( [ eq7.22 ] ) , respectively .
@xcite considered the estimation of the mean of heavy - tailed distribution with fast concentration .
he proposed an @xmath2-estimator by solving @xmath157=0,\ ] ] where the influence function @xmath158 is chosen such that @xmath159 .
he showed that this @xmath2-estimator has the exponential type of concentration by only requiring the existence of the variance .
it performed as well as the sample mean under the light - tail case . in section [ sec2 ]
, we essentially showed the same type of concentration for the ra - quadratic loss under the linear regression setting .
the estimation of mean can be regarded as a univariate linear regression where the covariate equals to 1 . in that special case , we have a more explicit concentration result for the ra - mean estimator , which is the estimator that minimizes the ra - quadratic loss .
let @xmath160 be an i.i.d sample from some unknown distribution with @xmath161 and @xmath162 .
the ra - mean estimator @xmath163 of @xmath164 is the solution of @xmath165=0,\ ] ] for parameter @xmath166 , where the influence function @xmath167 if @xmath168 , @xmath169 , if @xmath170 and @xmath171 if @xmath172 .
the following theorem gives the exponential type of concentration of @xmath163 around @xmath164 .
[ thm5 ] assume @xmath173 and let @xmath174 where @xmath175 .
then , @xmath176 the above result provides fast concentration of the mean estimation with only two moments assumption .
this is very useful for last scale hypothesis testing @xcite and covariance matrix estimation @xcite , where uniform convergence is required . taking the estimation of large covariance matrix as an example , in order for the elements of the sample covariance matrix to converge uniformly , the aforementioned authors require the underlying multivariate distribution be sub - gaussian .
this restrictive assumptions can be removed if we apply the robust estimation with concentration bound . regarding @xmath177 as the expected value of the random variable @xmath178 ( it is typically not the same as the median of @xmath179 ) , it can be estimated with accuracy @xmath180 where @xmath181 and @xmath182 is ra - mean estimator using data @xmath183 .
since there are only @xmath184 elements , by taking @xmath185 and the union bound , we have @xmath186 when @xmath187 .
this robustified covariance estimator requires much weaker condition than the sample covariance and has far wide applicability than the sample covariance .
it can be regularized further in the same way as the sample covariance matrix .
on the other hand , catoni s idea could also be extended to the linear regression setting .
suppose we replace the ra - quadratic loss @xmath20 in ( [ eq2.3 ] ) with catoni loss @xmath188 where the influence function @xmath189 is given by @xmath190 let @xmath191 be the corresponding solution .
then , @xmath191 has the same convergence rate as the ra - lasso , when the second or the third moment of errors exists . [ thm6 ] suppose condition ( c1 ) holds for @xmath192 or 3 , ( c2 ) , ( c3 ) and ( [ eq2.7 ] ) hold , then with probability at least @xmath95 , @xmath193^{1/2-q/4}).\ ] ] unlike the ra - lasso , the order of bias of @xmath191 can not be further improved , even when higher moments of errors exist beyond the third order .
the reason is that the catoni loss is not exactly the quadratic loss over any finite intervals .
similar results regarding the computational error of @xmath191 could also be established as in theorem [ thm4 ] , since the rsc / rsm conditions also hold for catoni loss with uniform constants .
we estimate @xmath194 based on the ra - lasso estimator and a cross - validation scheme . to ease the presentation
, we assume the data set can be evenly divided into @xmath195 folds with @xmath196 observations in each fold .
then , we estimate @xmath197 by @xmath198 where @xmath199 is the ra - lasso estimator obtained by using data points outside the @xmath105-th fold .
we show that @xmath200 is asymptotically efficient .
[ thm7 ] under conditions of theorem [ thm3 ] , if @xmath201 for @xmath202 , and @xmath203^{1-\frac{q}{2}}\}^{\frac{1}{2(k-1)}}\right)$ ] , then @xmath204
in this section , we assess the finite sample performance of the ra - lasso and compare it with other methods through various models .
we simulated data from the following high dimensional model @xmath205 where we generated @xmath206 observations and the number of parameters was chosen to be @xmath207 .
we chose the true regression coefficient vector as @xmath208 where the first 20 elements are all equal to 3 and the rest are all equal to 0 . to involve various shapes of error distributions , we considered the following five scenarios : 1 .
normal with mean 0 and variance 4 ( n(0,4 ) ) ; 2 . two times
the t - distribution with degrees of freedom 3 ( @xmath209 ) ; 3 .
mixture of normal distribution(mixn ) : @xmath210 ; 4 .
log - normal distribution ( lognormal ) : @xmath211 , where @xmath212 is standard normal .
weibull distribution with shape parameter = 0.3 and scale parameter = 0.5 ( weibull ) . in order to meet the model assumption ,
the errors were standardized to have mean 0 .
table [ tab1 ] categorizes the five scenarios according to the shapes and tails of the error distributions .
.summary of the shapes and tails of five error distributions [ tab1 ] [ cols="^,^,^",options="header " , ] ]
let @xmath213 . since @xmath12 minimizes @xmath214 , it follows from condition(c2 ) that @xmath215=(\betaa-\betas)^t\operatorname{e}(\bx\bx^t)(\betaa-\betas)\geq \kappa_l { \lvert\betaa-\betas\rvert_2}^2.\ ] ] let @xmath216 . then , since @xmath217 is the minimizer of @xmath218 , we have @xmath219\\ & = & \operatorname{e}[\ell(y-\bx^t\betaa)-\ell_{\alpha}(y-\bx^t\betaa ) ] + \operatorname{e}[\ell_{\alpha}(y-\bx^t\betaa)-\ell_{\alpha}(y-\bx^t\betas)]\\ & & \hspace{3ex}+\operatorname{e}[\ell_{\alpha}(y-\bx^t\betas)-\ell(y-\bx^t\betas)]\\ & \leq & \operatorname{e}[g_{\alpha}(y-\bx^t\betaa)]-\operatorname{e}[g_{\alpha}(y-\bx^t\betas ) ] . \end{aligned}\ ] ] by taylor s expansion , we have @xmath220 \leq 2\operatorname{e } [ ( z-\alpha^{-1 } ) i(z>\alpha^{-1 } ) |\bx^t(\betaa-\betas)|],\ ] ] where @xmath221 is a vector lying between @xmath12 and @xmath27 and @xmath222 . with @xmath223 denoting the conditional expectation with respect to @xmath11 given @xmath10 , we have @xmath224 \leq \operatorname{e}_{\epsilon } [ z i(z > \alpha^{-1 } ) ] \leq \alpha^{k-1}\operatorname{e}_{\epsilon } z^k .
\end{aligned}\ ] ] therefore , @xmath225 $ ] is further bounded by @xmath226,\ ] ] where the constant @xmath227 is defined in condition ( c1 ) .
next , we show that @xmath228)=o(1)$ ] .
let @xmath229 be a @xmath5-dimensional vector with @xmath230 . by the cauchy - schwartz inequality ,
@xmath231\leq [ \operatorname{e}(m_k+|\bx^t(\tilde{\bbeta}-\betas)|^k)^4]^{1/2 } [ \operatorname{e}(\bx^t\bnu)^4]^{1/2}.\ ] ] by ( c3 ) , @xmath37 is sub - gaussian with parameter @xmath232 . under the assumption that @xmath12 and @xmath27 are interior points of an @xmath1-ball with sufficiently large radius , @xmath233 is sub - gaussian with parameter @xmath234 , which is no larger than @xmath235 . using the moment results of sub - gaussian random variables @xcite , @xmath236 .
similarly , @xmath237 .
therefore , @xmath238=o(1)$ ] .
hence , by definition , @xmath228)=o(1)$ ] . using this result and ( [ eq7.3 ] ) , @xmath239&\leq 2(2\alpha)^{k-1}[\lambda_{\max}(\operatorname{e}[(m_k+|\bx^t(\tilde{\bbeta}-\betas)|^k)^2\bx\bx^t])]^{1/2 } { \lvert\betaa-\betas\rvert_2}\\ & = o(\alpha^{k-1 } { \lvert\betaa-\betas\rvert_2 } ) .
\end{aligned}\ ] ] this together with ( [ eq7.1 ] ) completes the proof .
first of all , it follows from lemma 1 of @xcite that @xmath240 on the event @xmath241 .
hence , we need to show that the event @xmath241 holds with high probability .
the latter will be established by using bernstein s inequality along with the union bound .
the gradient of @xmath73 , @xmath242\bx_i,\ ] ] where @xmath167 , for @xmath168 ; @xmath169 , for @xmath170 ; and @xmath171 , for @xmath172 . using @xmath243 and assumption ( c3 ) , we have @xmath244x_{ij } \}^2 & \leq 4\operatorname{e}\{(y_i-\bx_i^t\betaa)^2x_{ij}^2 \}\\ & \leq 8\operatorname{e}\{(\epsilon_i^2+|\bx_i^t(\betaa-\betas)|^2)x_{ij}^2\}\\ & \leq v , \end{aligned}\ ] ] where @xmath61 is a constant depending on @xmath245 and @xmath246 and
the last inequality follows from a similar argument as in the proof of theorem [ thm1 ] . by ( c3 ) and that @xmath247 , @xmath248x_{ij}$ ] is also sub - gaussian . for any @xmath249 , using the relation between the @xmath105th moment and the second moment of sub - gaussian random variables @xcite , @xmath250x_{ij}|^k\leq \frac{k!}{2}l^{k-2}\operatorname{e}|\psi[\alpha(y_i-\bx_i^t\betaa)]x_{ij}|^2,\ ] ] where @xmath62 is a constant depending on @xmath245 only .
hence , @xmath251x_{ij}|^k
\leq \frac{k!}{2}(2l/\alpha)^{k-2 } v.\ ] ] by bernstein inequality ( proposition 2.9 of @xcite ) and note that @xmath252\bx_i)={\ensuremath{\boldsymbol{0}}}$ ] , we have @xmath253x_{ij}\right| \geq \sqrt{\frac{2vt}{n}}+\frac{lt}{\alpha
n } \right)\leq 2\exp(-t).\ ] ] let @xmath254 and observe that @xmath255 by the choice of @xmath25 and @xmath16 .
we have @xmath253x_{ij}\right| \geq \frac{\lambda_n}{2}\right ) \leq
2\exp \left(-\frac{n\lambda_n^2}{32 v } \right).\ ] ] it then follows from union inequality that @xmath256\bx_i \right \rvert_{\infty}}>\frac{\lambda_n}{2 } \right ) & \leq 2 \exp \left(-\frac{n\lambda_n^2}{32 v}+\log p \right)=2p^{-c_0 } , \end{aligned}\ ] ] where @xmath257 .
this completes the proof .
denote @xmath258 . applying a second - order taylor expansion to @xmath259 between @xmath27 and @xmath260 ,
we conclude that for some @xmath261 $ ] , @xmath262(\bx_i^t\bdelta)^2 , \label{eq7.5 } \end{aligned}\ ] ] where @xmath263 for @xmath168 , and @xmath264 otherwise .
note that each term in is nonnegative .
however , the quadratic component in is not lipschitz continuous with a bounded lipschitz coefficient . in order to apply the contraction theorem of @xcite
, we introduce a truncation function that is lipschitz and bound from below .
let @xmath265 where @xmath266 is the indicator function .
clearly , @xmath267 and satisfies the lipschitz condition with lipschitz coefficient bounded by @xmath268 .
we first show @xmath269 for @xmath270 , where the thresholds @xmath271 and @xmath272 will be chosen as in ( [ eq7.10 ] ) .
let @xmath273(\bx_i^t\bdelta)^2 $ ] .
we need only to show that @xmath274 when @xmath275 or @xmath276 , the right hand side is zero and the inequality holds trivially .
thus , we need only to consider the case @xmath277 and @xmath278 . in this case ,
@xmath279 and hence @xmath280=1 $ ] . using this , @xmath281 using ( [ eq7.7 ] ) , to prove the lemma ,
we need to show that , for any fixed @xmath282 $ ] , with high probability @xmath283 where constants @xmath82 and @xmath83 do not depend on @xmath16 and @xmath284 this is equivalent to proving ( [ eq7.8 ] ) for @xmath285 . indeed , from the definition ( [ eq7.6 ] ) , for any @xmath286 and @xmath287
, we have @xmath288 .
thus , the event @xmath289 is the same as the event @xmath290 which equals to the event with @xmath291 . to establish ,
let us consider its complementary event .
define @xmath292 let @xmath293 be the unit sphere with @xmath1-radius one , and @xmath294 be the sphere of @xmath0-radius @xmath127 , which is to be chosen later .
the complementary event of is given by @xmath295<\kappa_1\{1-\kappa_2 \sqlogpn { \lvert\bdelta\rvert_1}\ } , \text { for some } \bdelta\in \bbs_2(1)\big\}.\ ] ] our goal is to show that the probability of this event is very small , which is demonstrated through the following three steps . * first
, we show that with the following choice of truncation @xmath296 for any fixed @xmath297 , we have @xmath298\geq \kl/2.\ ] ] * second , with @xmath299-\operatorname{e}[\gd(\bx)]|$ ] , we prove the tail probability bound for @xmath300 is bounded by @xmath301 for each given @xmath127 . *
finally , we use a standard peeling argument @xcite to establish @xmath302 the result ( c ) together with ( [ eq7.11 ] ) show that the probability of the complementary event of ( [ eq7.9 ] ) with @xmath303 and @xmath304 is bounded by @xmath305 , which completes the proof .
we first prove ( [ eq7.11 ] ) .
in fact , by condition ( c2 ) , for any @xmath306 , @xmath307\geq \kl { \lvert\bdelta\rvert_2}^2=\kl$ ] .
so , it suffices to show that @xmath308\leq \kl/2 $ ] .
note that , @xmath309 for all @xmath10 such that @xmath310 and @xmath311 .
therefore , we have @xmath312\leq \operatorname{e}[(\bx^t\bdelta)^2i(|y-\bx^t\betaa|\geq t)]+\operatorname{e}[(\bx^t\bdelta)^2i(|\bx^t\bdelta|\geq \tau/2)].\ ] ] to bound the first term on the right hand side of ( [ eq7.13 ] ) , it follows from the cauchy - schwartz inequality that @xmath313 \leq [ \operatorname{e}(\bx^t\bdelta)^4]^{1/2 } [ p(|y-\bx^t\betaa|\geq t)]^{1/2}.\ ] ] since @xmath314 is sub - gaussian with parameter at most @xmath315 by assumption ( c3 ) , we have @xmath316 .
meanwhile , for any @xmath317 , it follows from the chebyshev inequality and theorem [ thm1 ] that @xmath318\\ & \leq 2\operatorname{e}\epsilon^2 + 2\operatorname{e}[\bx^t(\betas-\betaa)]^2\\ & \leq 2 { m_k}^{2/k}+o(\alpha^{2k-2})\\ & \leq 4 { m_k}^{2/k}. \end{aligned}\ ] ] to bound the second term on the right hand side of ( [ eq7.13 ] ) , by the concentration inequality of sub - gaussian variables , we have @xmath319 then , by the choice of @xmath271 and @xmath272 in ( [ eq7.10 ] ) , @xmath313\leq \frac{\kl}{4 } \qquad \text { and }
\qquad \operatorname{e}[(\bx^t\bdelta)^2i(|\bx^t\bdelta|\geq \tau/2 ) ] \leq \frac{\kl}{4}.\ ] ] hence , ( [ eq7.11 ] ) follows .
next , we give the tail bound as in ( b ) . indeed , for any @xmath297 , we have @xmath320 .
therefore , by massart concentration inequality ( theorem 14.2 of @xcite ) , for any @xmath321 , we have @xmath322 . by choosing @xmath323 , we have @xmath324 next , we bound @xmath325 .
let @xmath326 be an i.i.d .
sequence of rademacher variables .
a symmetrization theorem ( theorem 14.3 of @xcite ) yields @xmath327\leq 2\operatorname{e}\left[\sidx |\frac{1}{n } \sum_{i=1}^n \omega_i \gd(\bx_i)|\right]=2\operatorname{e}\left[\sidx |\frac{1}{n } \sum_{i=1}^n \omega_i\varphi_{\tau}(\fd(\bx_i))|\right].\ ] ] by definition , the function @xmath328 is lipschitz with parameter at most @xmath329 and @xmath330 .
therefore , by the ledoux - talagrand contraction theorem ( @xcite , p.112 ) , we have @xmath331&\leq 8\tau^2\operatorname{e}\left[\sidx |\frac{1}{n } \sum_{i=1}^n \omega_i \fd(\bx_i)|\right ] \\ & = 8\tau^2\operatorname{e}\left[\sidx |\frac{1}{n } \sum_{i=1}^n \omega_i\bx_i^t\bdelta i(|y_i-\bx_i^t\betaa|\leq t)|\right]\\ & \leq 8\tau^2 t \operatorname{e}{\left \lvert\frac{1}{n } \sum_{i=1}^n \omega_i\bx_ii(|y_i-\bx_i^t\betaa|\leq t ) \right \rvert_{\infty}}. \end{aligned}\ ] ] since the variables @xmath332 are zero - mean i.i.d .
sub - gaussian with parameter at most @xmath315 , so are @xmath333 . since @xmath334 is the maxima of @xmath5 such terms , known bounds on the expectation of sub - gaussian maxima ( e.g. see @xcite , p.79 ) yield @xmath335 hence , @xmath336\leq 24\tau^2\ko t \sqlogpn.\ ] ] combining ( [ eq7.14 ] ) and ( [ eq7.15 ] ) , we have @xmath337 where constants @xmath338 and @xmath339 depends on @xmath340 and @xmath341 only .
this result holds for each given @xmath127 .
next , we furnish the peeling argument in ( c ) .
let @xmath342 and @xmath343 .
since @xmath344 , the set can be covered by partition @xmath345 with @xmath346 .
thus , by union bound , @xmath347 since @xmath348 for @xmath349 . by letting @xmath350 as in ( [ eq7.12 ] ) and solving for @xmath127 , by , we obtain @xmath351 where the last inequality follows from sum of geometric series .
note that , @xmath352 therefore , @xmath353 . let @xmath354 , we have @xmath355 hence , for any @xmath356 , we have @xmath357 by the cauchy - schwartz inequality and , we can bound further that @xmath358 it then follows from lemma [ lem2 ] that @xmath359 \}\\ & = \big(\kappa_1 - 4\kappa_1\kappa_2\rq^{1/2}\eta^{-q/2}\sqlogpn\big ) { \lvert\bdelta\rvert_2}^2 - 4\kappa_2\rq\eta^{1-q}\sqlogpn .
\end{aligned}\ ] ] with @xmath360 and @xmath89 , it holds that @xmath361 which is no larger than @xmath362 under assumption ( [ eq2.7 ] ) . on the other hand
, @xmath363 therefore , rsc holds with @xmath364 and @xmath365 .
let @xmath366 and @xmath367 denote the events that lemma [ lem1 ] and lemma [ lem3 ] hold , respectively . by theorem 1 of @xcite , within @xmath368
, it holds that @xmath369^{1-(q/2)}\right ) , \end{aligned}\ ] ] where ( i ) follows from the choice of @xmath89 . on the other hand , by lemma [ lem1 ] and [ lem3 ] , @xmath370 . from the proof of lemma [ lem2 ]
, we can see that indeed holds for all @xmath74 and @xmath371 that @xmath372 using the fact that @xmath373 , we conclude that @xmath374 therefore , ( [ eq3.2 ] ) holds with @xmath139 and @xmath140 . meanwhile , since @xmath375 , it follows from ( [ eq7.5 ] ) that @xmath376 under the sub - gaussianity assumption ( c3 ) , it follows from some existing work ( e.g. page 18 of @xcite ) that , with probability great than @xmath85 , it holds that @xmath377 where @xmath86 and @xmath87 are some generic constants .
hence , ( [ eq3.3 ] ) holds with @xmath378 and @xmath379 .
\(a ) we first show that , for any @xmath380 , @xmath381 , for all @xmath127 greater than the right hand side of ( [ eq7.20 ] ) , where @xmath382 is a contraction constant and @xmath106 is a tolerance parameter , which will be given in ( [ eq7.21 ] ) and ( [ eq7.22 ] ) , respectively . _ [ theorem 2 of @xcite ] suppose for any data set @xmath385 , the loss function @xmath386 is convex and differentiable and the regularizer @xmath387 is a norm . consider the optimization problem of @xmath388 for a radius @xmath112 such that @xmath389 is feasible , where @xmath390 , and a regularization parameter @xmath25 satisfying bound @xmath391 where @xmath392 is the dual norm of the regularizer .
in addition , suppose that the loss function @xmath73 satisfies the rsc / rsm condition with parameters ( @xmath133 , @xmath134 ) and ( @xmath135 , @xmath136 ) , respectively .
let @xmath393 be any @xmath387-decomposable pair of subspaces such that @xmath394 where @xmath395 , @xmath396 , @xmath397 , and @xmath398 .
denote @xmath399 , where @xmath400 is the projection of @xmath389 onto @xmath401 .
then for any @xmath380 , we have @xmath402 for all @xmath403 where @xmath404 , @xmath405 is the solution by the gradient descent algorithm after @xmath406 iteration , and @xmath407 is the initial value of @xmath408 . _ in fact ,
theorem 2 of @xcite is a deterministic statement for all choices of pairs @xmath393 . from lemma [ lem1 ] and lemma [ lem4 ]
, we have shown that with our choice of @xmath25 , the ra - quadratic loss function satisfy ( [ eq7.18 ] ) and rsc / rsm with probability at least @xmath63 and @xmath85 , respectively .
hence , theorem 2 of @xcite applies to our problem with high probability .
we further choose the pair @xmath409 and give the explicit expression of constants for our problem as the follows : @xmath410 where @xmath411 , @xmath412 , and @xmath413 .
it remains to check ( [ eq7.19 ] ) . by ( [ eq7.21 ] )
, @xmath382 is equivalent to requiring @xmath414 with @xmath89 , it follows from ( [ eq7.16 ] ) that @xmath415 hence , ( [ eq7.23 ] ) holds when @xmath416 is sufficiently large .
moreover , from we need @xmath146\frac{\log p}{n},\ ] ] which is satisfied under the stated assumption .
it then follows from theorem 2 of @xcite that , for any @xmath417 , @xmath381 , for all iterations @xmath127 greater than the right hand side of ( [ eq7.20 ] ) . for step ( b )
, it follows from rsc condition that @xmath418^t ( { \widehat}{\bbeta}^t-{\widehat}{\bbeta } ) \geq \frac{\gamma_l}{2 } { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_2}^2-\tau_l { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_1}^2.\ ] ] then we have @xmath419^t({\widehat}{\bbeta}^t-{\widehat}{\bbeta})+ \lambda_n({\lvert{\widehat}{\bbeta}^t\rvert_1}-{\lvert{\widehat}{\bbeta}\rvert_1})+\frac{\gamma_l}{2 } { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_2}^2-\tau_l { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_1}^2 .
\end{aligned}\ ] ] since @xmath22 is the minimizer of @xmath420 , by the first - order condition , @xmath421^t({\widehat}{\bbeta}^t-{\widehat}{\bbeta})\geq 0 $ ] .
therefore , @xmath422^t({\widehat}{\bbeta}^t-{\widehat}{\bbeta})+
\lambda_n({\lvert{\widehat}{\bbeta}^t\rvert_1}-{\lvert{\widehat}{\bbeta}\rvert_1})+\frac{\gamma_l}{2 } { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_2}^2-\tau_l { \lvert{\widehat}{\bbeta}^t-{\widehat}{\bbeta}\rvert_1}^2.\ ] ] by the convexity of the @xmath0-norm , @xmath423^t({\widehat}{\bbeta}^t-{\widehat}{\bbeta } ) \geq 0 $ ] .
hence , @xmath424 next , we bound @xmath425 .
it follows from lemma 3 of @xcite that @xmath426 where @xmath427 is defined as in ( a ) .
then , by the cauchy - schwartz inequality , @xmath428 equations ( [ eq7.24 ] ) and ( [ eq7.25 ] ) together with results in ( a ) imply that , @xmath429 letting @xmath430 , we have @xmath431 we now bound the second term in ( [ eq7.26 ] ) . by ( [ eq7.16 ] ) and ( [ eq7.17 ] ) , we have @xmath432 . \end{split}\ ] ] meanwhile , from ( a ) we have @xmath433 since @xmath434 , @xmath435 , @xmath436 , @xmath437 , and @xmath438 , it follows from ( [ eq7.26 ] ) , ( [ eq7.27 ] ) and ( [ eq7.28 ] ) that @xmath439 \right ) .\ ] ] the proof follows the same spirit of the proof of proposition 2.4 of @xcite .
the influence function @xmath440 satisfies @xmath441 using this and independence , with @xmath442 $ ] , we have @xmath443 \right\ } & \leq \bigl ( \operatorname{e}\{\exp \ { \psi[\alpha(y_i-\theta ) ] \ } \big ) ^n \\ & \leq \{1+\alpha(\mu-\theta)+\alpha^2[\sigma^2+(\mu-\theta)^2 ] \}^n\\ & \leq \exp \left\{n\alpha(\mu-\theta)+n\alpha^2 [ v^2+(\mu-\theta)^2 ] \right\}. \end{aligned}\ ] ] similarly , @xmath444
\right\}\leq \exp \left\{-n\alpha(\mu-\theta)+n\alpha^2 [ v^2+(\mu-\theta)^2 ] \right\}$ ]
. define @xmath445+\frac{\log ( 1/\delta)}{n\alpha}\\ b_-(\theta)&=\mu-\theta-\alpha [ v^2+(\mu-\theta)^2]-\frac{\log ( 1/\delta)}{n\alpha } \end{aligned}\ ] ] by chebyshev inequality , @xmath446 \right\ } } { \exp\{\alpha n(\mu-\theta)+n\alpha^2 [ v^2+(\mu-\theta)^2]+\log(1/\delta ) \}}\leq \delta\ ] ] similarly , @xmath447 .
let @xmath448 be the smallest solution of the quadratic equation @xmath449 and @xmath450 be the largest solution of the equation @xmath451 . under the assumption that @xmath452 and the choice of @xmath453 , we have @xmath454 .
therefore , @xmath455 similarly , @xmath456 with @xmath457 , @xmath458 , @xmath459 .
since the map @xmath460 is non - increasing , under event @xmath461 @xmath462 i.e. @xmath463 . meanwhile , @xmath464 . first , we prove that the approximation error rate @xmath465 , where @xmath466 is the population minimizer under the catoni loss .
let @xmath467 \text{d}t$ ] .
it follows from ( [ eq7.2 ] ) that @xmath468\leq \operatorname{e}[|g_{\alpha}'(y-\bx^t\tilde{\bbeta})\bx^t(\betacs-\betas)|],\ ] ] where @xmath221 is a vector lying between @xmath12 and @xmath469 . since @xmath470 , by the second - order taylor expansion with an integral remainder , @xmath471 hence , by the cauchy - schwartz inequality , @xmath472&\leq \alpha^2 \operatorname{e}[|y-\bx^t\tilde{\bbeta}|^3|\bx^t(\bbeta_{\alpha}^{c\ast}-\betas)| ] \\ & \leq 4\alpha^2\operatorname{e}[(m_3+|\bx^t(\tilde{\bbeta}-\betas)|^3)|\bx^t(\bbeta_{\alpha}^{c\ast}-\betas)|]\\ & \leq 4\alpha^2[\lambda_{\max}(\operatorname{e}[(m_3+|\bx^t(\tilde{\bbeta}-\betas)|^3)^2 \bx\bx^t])]^{1/2 } { \lvert\bbeta_{\alpha}^{c\ast}-\betas\rvert_2 } , \end{aligned}\ ] ] which is of order @xmath473 , as @xmath474)=o(1)$ ] by an analogous argument as in the proof of theorem [ thm1 ] .
similarly as in ( [ eq7.1 ] ) , @xmath475\geq \kappa_l { \lvert\betaa-\betas\rvert_2}^2.\ ] ] hence , @xmath476 .
if @xmath11 only has the second moment exist , by a first - order taylor expansion of @xmath477 similarly as in ( [ eq7.29 ] ) , we have @xmath478 .
next , since @xmath479 , by the same argument as in the proof of lemma [ lem3 ] , rsc holds for catoni s loss with probability no less than @xmath85 .
hence , similarly as in theorem [ thm2 ] , @xmath480^{1/2-q/4})$ ] .
this together with @xmath465 completes the proof .
first of all , observe that @xmath481 given that @xmath482 exists , by central limit theorem , @xmath483 .
let @xmath484 .
we now need to prove that the last two terms are negligible .
conditioning on data outside the @xmath105th fold , @xmath485 hence , @xmath486 , where the last equality follows from theorem [ thm3 ] . by an analogous argument
, we have @xmath487^{1-q/2 } \})=o(1/\sqrt{n } ) .
\end{aligned}\ ] ] this completes the proof of the theorem .
duchi , j. , shalev - shwartz , s. , singer , y. , and chandra , t. ( 2008 ) .
efficient projections onto the @xmath0-ball for learning in high dimensions .
_ proceedings of the 25th international conference on machine learning _ , 272279 .
huang , c. c. , liu , k. , pope , r. m. , du , p. , lin , s. , rajamannan , n. m. , et al .
activated tlr signaling in atherosclerosis among women with lower framingham risk score : the multi - ethnic study of atherosclerosis . _
plos one _ , * 6 * , e21067 .
wainwright , m. j. ( 2013 ) . regularized @xmath2-estimators with nonconvexity : statistical and algorithmic theory for local optima .
_ advances in neural information processing systems _ , 476484 .
negahban , s. n. , ravikumar , p. , wainwright , m. j. , and yu , b. ( 2012 ) . a unified framework for high - dimensional analysis of m - estimators with decomposable regularizers .
_ statistical science _ * 27 * , 538557 .
|
data subject to heavy - tailed errors are commonly encountered in various scientific fields , especially in the modern era with explosion of massive data . to address this problem ,
procedures based on quantile regression and least absolute deviation ( lad ) regression have been developed in recent years .
these methods essentially estimate the conditional median ( or quantile ) function
. they can be very different from the conditional mean functions when distributions are asymmetric and heteroscedastic .
how can we efficiently estimate the mean regression functions in ultra - high dimensional setting with existence of only the second moment ? to solve this problem , we propose a penalized huber loss with diverging parameter to reduce biases created by the traditional huber loss .
such a penalized robust approximate quadratic ( ra - quadratic ) loss will be called ra - lasso . in the ultra - high dimensional
setting , where the dimensionality can grow exponentially with the sample size , our results reveal that the ra - lasso estimator produces a consistent estimator at the same rate as the optimal rate under the light - tail situation .
we further study the computational convergence of ra - lasso and show that the composite gradient descent algorithm indeed produces a solution that admits the same optimal rate after sufficient iterations . as a byproduct
, we also establish the concentration inequality for estimating population mean when there exists only the second moment .
we compare ra - lasso with other regularized robust estimators based on quantile regression and lad regression .
extensive simulation studies demonstrate the satisfactory finite - sample performance of ra - lasso . _
key words _ : high dimension , huber loss , m - estimator , optimal rate , robust regularization .
|
the problem we address in this paper is the numerical integration of the second order time - dependent linear equation @xmath0 where @xmath1 , @xmath2 and primes denote time - derivatives .
it has many practical applications , for example , in periodically variable systems such as quadrupole mass filter and quadrupole devices , @xcite , microelectromechanical systems @xcite , in bose
einstein condensates , spatially linear electric fields , dynamic buckling of structures , electrons in crystal lattices , waves in periodic media , etc .
( see @xcite and references therein ) . in these cases
the matrix @xmath3 is usually time periodic with period @xmath4 ( ( [ hill ] ) is then an example of a matrix hill equation ) and of moderately large size @xmath5 .
parametric resonances can occur so that it is very important to know the stability regions in terms of the parameters of the system .
another important example of this type is the linear time - dependent wave equation @xmath6 equipped with initial conditions @xmath7 , and @xmath8 .
once discretized in space in a bounded region , [ eq.se ] also leads to [ hill ] . when the dimension @xmath5 of [ hill ] is relatively small , the numerical computation of its fundamental matrix solution for one period ( usually repeatedly many times for different values of the parameters of the system ) is feasible and allows one to analyse the stability of the configuration . in this case
numerical schemes that involve matrix matrix products can be used . however , when [ hill ] results from a semidiscretized pde like [ eq.se ] then @xmath9 , and so only numerical schemes involving matrix
vector products are suitable . taking these considerations into account , in this work we present two classes of numerical schemes : one that involves matrix matrix products showing a high performance for the computation of the fundamental matrix solution for one period , and another class involving matrix vector products , addressed for the numerical treatment of pdes .
equation ( [ hill ] ) can be written as a first order system by introducing new variables @xmath10 as @xmath11 and @xmath12 , @xmath13 .
the solution of [ firstordersystem ] evolves through a linear transformation ( evolution operator or fundamental matrix solution ) given by @xmath14 . in the usual case
in which @xmath15 is a real and symmetric @xmath16 matrix valued function ( @xmath17 ) then @xmath18 is a symplectic transformation .
the eigenvalues of @xmath18 occur in reciprocal pairs , say @xmath19 , where @xmath20 denotes the complex conjugate of @xmath21 . as a result , for stable systems ,
all of the eigenvalues must lie on the unit circle ( see for example @xcite ) .
this is a very important property that is not preserved in general by standard numerical integrators .
thus , one can be forced to use very small time steps to avoid the undesirable numerical instabilities or asymptotic stabilities arising from this fact , with the resulting degradation in the efficiency of the algorithms .
the solution of the non - autonomous [ firstordersystem ] can not be written in general in closed form . nevertheless , we can always formally write it as a single exponential by using the magnus expansion @xcite , @xmath22 where @xmath23 satisfies a ( highly nonlinear ) differential equation . although different approximations can be found in the literature ( see e.g. @xcite and references therein ) , they involve the computation of nested commutators of @xmath24 , evaluated at different times , and so for many problems may be computationally expensive for this reason , here we consider two different strategies starting from the formal solution provided by the magnus expansion .
the first leads us to optimised methods specifically addressed to problems defined by ( [ hill ] ) when matrix matrix products are feasible ; to these we call _ magnus - decomposition methods_. the second strategy allows us to build optimised methods when only matrix vector products are suitable .
these are referred to as _ magnus
splitting methods_. in _ magnus - decomposition methods _ , in particular , we approximate @xmath25 by a composition of simpler symplectic maps that for the problem at hand are considerably faster to compute .
as an illustration , a 4th - order method within this family contains only one costly matrix exponential and this exponential is cheaply approximated by a symplectic approximation of order @xmath26 leading to an scheme given by the composition : @xmath27 } = \begin{pmatrix } i & 0 \\ d_{3}^{[4,q ] } & i \end{pmatrix } \begin{pmatrix } 0 & d_{2}^{[4,q ] } \\ i & 0 \end{pmatrix } \begin{pmatrix } i & 0 \\ d_{1}^{[4,q ] } & i \end{pmatrix},\ ] ] where @xmath28 } , \
i=1,2,3 $ ] , are certain matrices to be defined later on . on the other hand , in _ magnus splitting methods _ , we consider different decompositions that are suitable for low - oscillatory and high - dimensional problems , where only matrix vector products are reasonable in the numerical scheme
. methods of order @xmath29 within this family are of the form @xmath30 } = \begin{pmatrix } i & h a_{m+1 } i \\ 0 & i \end{pmatrix } \begin{pmatrix } i & 0 \\ h c_m & i \end{pmatrix } \begin{pmatrix } i & h a_m i \\ 0 & i \end{pmatrix } \
\cdots \begin{pmatrix } i & 0 \\ h c_1 & i \end{pmatrix } \begin{pmatrix } i & a_1 i \\ 0 & i \end{pmatrix } , \ ] ] where the coefficients @xmath31 are obtained numerically by solving a set of non - linear polynomial equations and the matrices @xmath32 are linear combinations of @xmath3 evaluated at certain quadrature points . the product @xmath33 } z_0 $ ] is done with only @xmath34 products of matrices @xmath32 on a vector of dimension @xmath5 .
since all the schemes presented in this work are based on the magnus expansion , they become explicit symplectic integrators when @xmath35 is real and symmetric and can be considered as geometric integrators @xcite , showing a favourable behaviour on long - time simulations .
as stated before , the magnus expansion @xcite expresses the solution of [ firstordersystem ] on the time interval [ @xmath36,@xmath37 in the form of a single exponential of an infinite series @xmath38 whose first terms are given by @xmath39 \ , d\tau_{2 } \ , d\tau_{1},\ ] ] where @xmath40 = \alpha\beta - \beta\alpha$ ] is the matrix commutator .
when matrix @xmath24 is given by , the exponent @xmath41 and as well as any truncation of the series at order @xmath42 , @xmath43}$ ] , belong to the symplectic lie algebra , and thus symplecticity is automatically preserved , approximations of @xmath41 in terms of @xmath24 evaluated at the nodes of some quadrature rule can be obtained as follows .
first we consider the polynomial @xmath44 of degree @xmath45 in @xmath46 that interpolates @xmath24 on @xmath47 $ ] at the points @xmath48 , where @xmath49 are the nodes of the gauss legendre quadrature rule of order @xmath50 .
the perturbed problem reads @xmath51,\ ] ] where @xmath52 is the exact solution of ( [ firstordersystem ] ) at @xmath36 . from a direct application of the alekseev
grbner lemma @xcite ( see also @xcite ) , we have that @xmath53 letting @xmath54 , we write the interpolation polynomial as @xmath55,\ ] ] with @xmath56 and the usual lagrange polynomials @xmath57 .
notice that for our problem ( [ firstordersystem ] ) we have @xmath58 where @xmath59 and @xmath60 in particular , for sixth - order methods ( @xmath61 ) we have @xmath62 where @xmath63 .
one can readily check that @xmath64 .
the integrals in the magnus expansion for @xmath44 can be computed from analytically .
this results in an approximation of @xmath65 expressed in terms of @xmath66 up to order @xmath50 .
specifically , a sixth - order approximation @xmath67 } = \omega + \mathcal{o } ( h^{7})$ ] results from @xmath68 } = \alpha_{1 } + \frac{1}{12 } \alpha_{3 } - \frac{1}{12 } [ 12 ] + \frac{1}{240 } [ 23 ] + \frac{1}{360 } [ 113 ] - \frac{1}{240 } [ 212 ] + \frac{1}{720 } [ 1112],\ ] ] where @xmath69 $ ] represents the nested commutator @xmath70\ldots]]]$ ] . at this point
, we can proceed in different ways and we present two different strategies to build new methods . in both cases , the respective methods ( [ eq : phi4q1 ] ) and ( [ eq : rkn2 ] ) will be compositions of exponentials of elements of the lie algebra generated by @xmath66 whose
coefficients will be determined in such a way that the compositions coincide with ( [ sixth ] ) up to the desired order ( six in this case ) .
firstly , we examine the structure of the lie algebra generated by the @xmath71 .
we immediately notice that , @xmath72=0 $ ] for @xmath73 .
furthermore , @xmath74=\begin{pmatrix } 0 & 0 \\ 2 \mu_2 ^ 2 & 0 \end{pmatrix}.\ ] ] we distinguish the following types of exponentials that may appear as elements of the lie algebra generated by @xmath75 : @xmath76 clearly , we want to avoid the computation of the full matrix exponential @xmath77 and instead focus on types @xmath78 .
the latter is nilpotent and its exponential comes virtually for free . in @xcite , it was obtained methods of order 4 , 6 and 8 for the hill equation , denoted by @xmath79}$ ] ( a @xmath80-order method containing @xmath81 exponentials of type @xmath82 above ) .
in particular @xmath83 } = \begin{pmatrix } i & 0 \\ hc_{2}^{[4 ] } & i \end{pmatrix } \,\ , \exp \left(h \begin{pmatrix } 0 & i \\ d_{1}^{[4 ] } & 0 \end{pmatrix}\right ) \,\ , \begin{pmatrix } i & 0 \\ hc_{1}^{[4 ] } & i \end{pmatrix}\ ] ] @xmath84 } = \begin{pmatrix } i & 0 \\ hc_{2}^{[6 ] } & i \end{pmatrix } \,\ , \exp \left(\frac{h}{2 } \begin{pmatrix } 0 & i \\ d_{2}^{[6 ] } & 0 \end{pmatrix}\right ) \,\ , \exp \left(\frac{h}{2 } \begin{pmatrix } 0 & i \\ d_{1}^{[6 ] } & 0 \end{pmatrix}\right ) \,\ , \begin{pmatrix } i & 0 \\ hc_{1}^{[6 ] } & i \end{pmatrix},\ ] ] where @xmath85}$ ] are linear combinations of @xmath3 evaluated at a set of quadrature points of order @xmath29 or higher and @xmath86}$ ] are linear combinations of @xmath3 that additionally contain one product of such linear combinations . due to the exponentials ,
these schemes are specially appropriate when the solution is oscillatory or stiff . since the averaged matrices @xmath87}$ ]
can be considered as constant matrices in each time subinterval , the following result is useful when computing the corresponding exponentials for a real - valued matrix @xmath88 @xcite : @xmath89 although there exist efficient methods to compute matrix trigonometric functions @xcite as well as matrix exponentials , such as pad approximants , krylov / lanczos methods , chebyshev method and others , for this particular case a more efficient procedure is obtained by decomposing the exponential into a product of simple matrices . when @xmath90 is symmetric , the resulting approximations preserve the symplectic structure by construction
specifically , if @xmath91 , where @xmath92 is the spectral radius of @xmath93 , then @xcite @xmath94 where @xmath95 the truncated series expansions of @xmath96 and @xmath97 up to order @xmath98 and @xmath99 in @xmath100 , denoted by @xmath101}$ ] and @xmath102}$ ] , respectively , can be simultaneously computed with only @xmath103 products ( e.g. , @xmath104},\ : p^{[4]}$ ] can be computed with only one product ) .
we denote by @xmath105}$ ] a @xmath106-order approximation to @xmath107 obtained by replacing @xmath108 and @xmath109 by @xmath101}$ ] and @xmath102}$ ] .
notice that if @xmath90 is a symmetric matrix then @xmath101},\:r^{[q]}$ ] are also symmetric matrices and , by construction , @xmath105}$ ] is a symplectic matrix @xmath110 .
taking into account all these considerations , we substitute [ eq : factoring , eq : factoring : sintan ] into @xmath79}$ ] given by ( [ eq : phi41 ] ) and ( [ eq : phi62 ] ) and combine commuting matrices to get finally @xmath111 } , \upsilon_{2}^{[6,q]}$ ] given in table [ tab : methods41_62 ] .
the algorithm proceeds by multiplying @xmath112}$ ] by the result from the previous computational step .
as the last matrix in a step commutes with the first one in the following step , some products can be saved , hence for an even @xmath113 the computational cost of @xmath114}_{1 } $ ] and @xmath115}_{2 } $ ] are @xmath116 and @xmath117 , respectively .
the schemes and the relevant parameters are collected in .
@xmath118 @xmath119 @xmath120 @xmath121}$ ] and @xmath122 } $ ] are obtained by applying the expansions [ eq : factoring : sintan ] to @xmath123 } \!.$ ] @xmath124}=\frac{\sqrt{15}}{36}k + \frac{5}{36}l \qquad & d_1^{[4]}=-m_2 \\
c_2^{[4]}=-\frac{\sqrt{15}}{36}k + \frac{5}{36}l & \end{array}\ ] ] @xmath125}= \begin{pmatrix } i & 0 \\ hc_2^{[4]}+r_1^{[4,q ] } & i \end{pmatrix } \begin{pmatrix } i & q_1^{[4,q+2 ] } \\ 0 & i \end{pmatrix } \begin{pmatrix } i & 0 \\ hc_1^{[4]}+r_1^{[4,q ] } & i \end{pmatrix}.\ ] ] @xmath126}=-\frac{\sqrt{15}}{180}k + \frac1{18}l+\frac1{12960}f \qquad & d_1^{[6]}=-m_2 -\frac{4}{3\sqrt{15}}k + \frac1{6}l \\ c_2^{[6]}=+\frac{\sqrt{15}}{180}k + \frac1{18}l+\frac1{12960}f & d_2^{[6]}=-m_2 + \frac{4}{3\sqrt{15}}k + \frac1{6}l \end{array}\ ] ] @xmath127}= \begin{pmatrix } i & 0 \\ hc_2^{[6]}+r_2^{[6,q ] } & i \end{pmatrix } \begin{pmatrix } i & q_2^{[6,q+2 ] } \\
0 & i \end{pmatrix } \begin{pmatrix } i & 0 \\ r_2^{[6,q]}+r_1^{[6,q ] } & i \end{pmatrix } \begin{pmatrix } i & q_1^{[6,q+2 ] } \\ 0 & i \end{pmatrix } \begin{pmatrix } i & 0 \\ hc_1^{[6]}+r_1^{[6,q ] } & i \end{pmatrix}.\ ] ] for deriving the second class of schemes considered in this work , we first split the matrix @xmath24 of [ firstordersystem ] as @xmath128 and denote @xmath129 where @xmath130 is the interpolating polynomial of @xmath131 in the interval @xmath132 $ ] , and @xmath133 @xmath134 .
it is easy to check that @xmath135=0 $ ] and @xmath136=[\beta_i,\beta_j,\beta_k,\delta_1]=0 $ ] for any value of @xmath137 . as a consequence ,
the formal solution [ sixth ] simplifies to @xmath138 } & = & \delta_1 + \beta_1 + \frac{1}{12 } \beta_3 + \frac{1}{12 } [ \beta_2,\delta_1]+ \frac{1}{360 } \big ( -[\delta_1,\beta_3,\delta_1 ] + [ \beta_1,\delta_1,\beta_3 ] \big ) \nonumber\\ & & - \frac{1}{240 } [ \beta_2,\delta_1,\beta_2 ] + \frac{1}{720 } \big([\delta_1,\beta_1,\delta_1,\beta_2 ] - [ \beta_1,\delta_1,\beta_2,\delta_1 ] \nonumber \big ) .
\end{aligned}\ ] ] it is then clear that a composition of type ( [ eq : rkn2 ] ) can be recovered by considering the following product of exponentials : @xmath139 } = \prod_{i=1}^{m+1 } \exp\left(\sum_{j=1}^3 y_{i , j}\beta_j \right ) \ , \exp\left(a_i\delta_1\right)\ ] ] @xmath140 .
specifically , taking to account ( [ eq.ab ] ) , we have @xmath141 } = \prod_{i=1}^{m+1 } \begin{pmatrix } i & 0 \\ \displaystyle \sum_{j=1}^3b_{i , j}h m_j & i \end{pmatrix } \begin{pmatrix } i & a_ihi \\ 0 & i \end{pmatrix}\ ] ] where @xmath142 and @xmath143 , so that , in practice , corresponds to a @xmath34-stage composition .
all methods are symplectic when applied to hamiltonian systems and , moreover , the coefficients are chosen so that time symmetric is preserved . to obtain particular methods we extend the analysis carried out in @xcite , where several 6th - order schemes were derived for the more general problem @xmath144 although all of them can be applied to the present problem simply by letting @xmath145 ,
new methods have also been obtained by taking into consideration the simpler structure that this system possesses .
specifically , we have taken @xmath146 in as the coefficients of an optimised 11-stage 6th - order method designed in @xcite for when @xmath35 and @xmath147 are constant . in this way the commutators involving only @xmath148 and @xmath149 ( e.g. @xmath150 $ ] , @xmath151 $ ] , etc . )
vanish up to order six , whereas the higher - order contributions are minimised by considering more stages than strictly necessary to solve the order conditions ; this extra cost is compensated by a much improved accuracy and stability in an autonomous case .
next , we look for new coefficients @xmath152 , which now have to satisfy a much reduced set of order conditions .
the 11-stage composition has five coefficients @xmath153 and six coefficients @xmath154 to solve seven equations .
taking into account the structure of the equations , this leaves one of the @xmath153 and three of the @xmath154 as free parameters , that are chosen in order to minimize the objective function @xmath155 .
once the coefficients @xmath156 are chosen , the matrices @xmath157 are replaced by the corresponding linear combinations of @xmath158 evaluated at the quadrature rule of order 6 , so that one ends up with a composition of the form ( [ comp.2 ] ) . specifically , the following 11-stage 6th - order method is obtained : @xmath159 } = \begin{pmatrix } i & h a_{12 } i \\ 0
& i \end{pmatrix } \begin{pmatrix } i & 0 \\ hc_{11 } & i \end{pmatrix } \begin{pmatrix } i & ha_{11}i \\ 0 & i \end{pmatrix }
\ \cdots \begin{pmatrix } i & 0 \\ hc_{1 } & i \end{pmatrix } \begin{pmatrix } i & ha_1i \\ 0 & i \end{pmatrix } , \ ] ] where @xmath160 @xmath161 and @xmath49 are the nodes of the 6th - order gauss
legendre quadrature rule , i.e. , @xmath162 the corresponding coefficients @xmath163 are : @xmath164 @xmath165 verifying the time - symmetry condition @xmath166 , and @xmath167
in the following section we study and demonstrate the performance of the new methods with respect to well established explicit and implicit standard runge kutta ( rk ) and explicit symplectic runge kutta
nystrm ( rkn ) methods from the literature .
six types of methods are considered : * @xmath168 } $ ] from @xcite and @xcite : symplectic @xmath80-order methods requiring computing of @xmath81 matrix exponentials .
* @xmath169 } $ ] : new sets of methods obtained from @xmath170}_k $ ] by decomposing matrix exponentials and taking taylor series expansion up to @xmath171 .
* @xmath172 } $ ] : the new 11-stage 6th - order magnus splitting method ( [ eq : psi2_11,11stage_bs ] ) .
* @xmath173 } $ ] : an 11-stage 6th - order magnus splitting method for non - autonomous systems from @xcite . * @xmath174}$ ] : an explicit @xmath81-stage @xmath80-order method that uses the radau quadrature rule of order six for the time - dependent part . * @xmath175}$ ] : implicit @xmath81-stage @xmath80-order symplectic gauss legendre methods .
* @xmath176}$ ] : @xmath81-stage @xmath80-order explicit symplectic methods from @xcite .
@xmath177 } $ ] has been selected instead of the more effective @xmath178 } $ ] ( ibid . ) because the former has the same number of stages as @xmath172 } $ ] .
the cost of each method is estimated for two different problem types .
the first one is when a numerical method acts on the fundamental matrix @xmath179 .
let @xmath180 .
then , the cost is expressed in the number of matrix matrix products @xmath181 , required to propagate for one time step @xmath182 .
evaluations of @xmath183 , scalar matrix multiplications and linear combinations of matrices are not included to the cost .
the second case is when the same method is used to integrate a system whose state is represented by a vector @xmath184 .
similarly , the cost is expressed in the number of matrix vector products @xmath185 .
the methods costs are summarized in table [ tab : cost ] .
the parameter @xmath186 } $ ] in the implicit methods refers to the number of iterations per step for a @xmath187-order method .
typically , @xmath186}=4 ... 7 $ ] to attain convergence and preservation of symplecticity to high accuracy . for the numerical experiments in this paper ,
we assume them to be @xmath188}=4 $ ] for the 4th - order method and @xmath189}=6 $ ] for the 6th - order one .
.cost of the methods in terms of matrix matrix products , @xmath181 , and matrix vector products , @xmath185 . [ cols="<,^,^ " , ] however , some of these methods are not optimised for problems where matrix matrix are exceedingly costly , and they are not used in these problems .
moreover , the schemes @xmath190}$ ] acting on a vector could be carried out using propagators like krylov methods to compute the action of the exponential of a matrix on a vector , but this is not considered in this work .
the reference solutions are obtained numerically using sufficiently small time steps .
the first performance test is executed employing the mathieu equation @xmath191 written as a first - order system ( [ firstordersystem ] ) . at first
, we family - wise compare the performance of the new methods @xmath192}_{k } $ ] .
then , we integrate for @xmath193 $ ] with the identity matrix as the initial condition and then measure the @xmath194-norm of the error of the fundamental matrix at the final time .
this procedure is repeated for different time steps and different choices of @xmath195 . in table
[ tab : best_q_of_omega ] we show the values of @xmath26 that provided the best performances for the choices @xmath196 and @xmath197 .
we observe that the choices @xmath198 are generally better in the 4th - order family , and @xmath199 are better among the 6th - order methods . these selected schemes will be considered in the following numerical examples .
cccccccc @xmath200 & 1/125 & 1/25 & 1/5 & 1 & 5 & 25 & 125 + ' '' '' & + ' '' '' @xmath29 of @xmath201}$ ] & 6 , 10 & 6 , 10 & 6 , 10 & 6 , 10 & 12 , 10 & 8 , 12 & 8 , 12 + ' '' '' @xmath26 of @xmath202}$ ] & 10 , 8 & 10 , 8 & 10 , 8 & 10 , 8 & 12 , 8 & 12 , 8 & 12 , 8 + ' '' '' & + ' '' '' @xmath29 of @xmath201}$ ] & 10 , 6 & 10 , 6 & 10 , 6 & 10 , 12 & 12 , 8 & 8 , 12 & 8 , 12 + ' '' '' @xmath26 of @xmath202}$ ] & 8 , 10 & 8 , 10 & 8 , 10 & 8 , 10 & 12 , 8 & 12 , 8 & 12 , 8 + we select now @xmath203 and @xmath204 for plotting the @xmath194-norm of the error of the fundamental matrix at the final time versus the computation cost in units of @xmath181 for all methods previously considered . show methods performance .
we observe that the standard explicit and implicit rk methods perform considerably worse while the new methods show the best performances for oscillatory cases . to illustrate how the accuracy depends on the frequency @xmath200 , we take @xmath205 , the time step @xmath206 and measure the @xmath194-norm of the error in the fundamental matrix solution for @xmath207 } $ ] .
the results are shown in .
we observe that the new methods show smaller error growth as frequency of the problem increases ( oscillatory problems ) . 0.5 in the mathieu [ ex : mathieu ] on a logarithmic scale.,title="fig : " ] 0.5 in the mathieu [ ex : mathieu ] on a logarithmic scale.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] the second benchmark to consider is the matrix hill equation : @xmath210 where @xmath211 .
we assume @xmath212 , where @xmath213 is a pascal matrix : @xmath214 we set @xmath215 , @xmath216 and @xmath217 and compute solutions for @xmath218 and @xmath219 on the interval @xmath220 $ ] , and then we measure the @xmath194-norm of the error in the fundamental matrix solution at the final time . in @xcite , it was determined that for matrix hill - type problems @xmath190 } $ ] perform no worse than @xmath177 } $ ] .
demonstrate that , thanks to lower computational cost , the new @xmath111 } , \upsilon_{2}^{[6,q ] } $ ] and @xmath172 } $ ] methods consistently produce better results in both oscillatory ( larger @xmath221 ) and nearly autonomous ( small @xmath222 ) cases . to analyse the performance of the methods that only involve matrix
vector products we consider the following trapped wave equation @xmath223 equipped with initial conditions @xmath224 , and @xmath225 . the solution for @xmath226 can be easily be obtained by separation of variables and it is given by @xmath227 .
when an external interaction appears , @xmath228 , the equation has no analytical solution in general and one has to consider , for example , a numerical scheme .
we assume the solution is confined to a region @xmath229 $ ] and hence the solution and all spatial derivatives vanish at these boundaries .
this allows us to treat the problem as periodic and spectral methods can be used .
we divide the spatial region into @xmath147 intervals of length @xmath230 and , after spatial discretization , we obtain an equation similar to [ hill ] that we write as the first order system [ firstordersystem ] where @xmath231 and @xmath232 . for the numerical experiments we take @xmath233 and the external interaction
@xmath234 we take @xmath235 , @xmath236 , and integrate for the period @xmath237 $ ] .
the reference solution is obtained numerically with a sufficiently small time step and we measure the @xmath238-norm of the solution versus the number of matrix vector products for each method .
the results are shown in the where @xmath172}$ ] , with only three evaluations of the time - dependent functions per step , shows the best performance for smooth time - dependencies and it is nearly the best one when such time - dependency increases . 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ] 0.5 on a @xmath208@xmath208 scale ; @xmath209.,title="fig : " ]
starting from the well know magnus expansion for linear time - dependent differential equations , in this work we have presented two families of time - integrators especially designed for the second order linear system ( [ hill ] ) . both share with the exact solution relevant qualitative properties ( in particular , its symplectic character ) and thus provide an accurate description of the system over long time intervals , due to their favourable error propagation , but they differ in significant aspects .
whereas magnus - decomposition methods are addressed to problems with small to moderate dimensions where the numerical computation of matrix matrix products ( and thus the fundamental matrix ) is feasible , magnus splitting methods are advantageous when the dimension of the system is exceedingly high , as is the case when ( [ hill ] ) results from a linear time - dependent wave equation previously discretized in space . the numerical experiments reported here clearly illustrate this difference : whereas the first family performs more efficiently on the scalar mathieu equation and a low - dimensional and oscillatory matrix hill equation , it is the second class of integrators which shows a better behaviour on a discretised trapped wave equation , where algorithms based only in matrix - vector products are advisable .
magnus - decomposition methods can be considered as an adaptation of the schemes presented in @xcite involving appropriate truncated expansions of the otherwise computationally costly matrix exponentials initially present in them and useful combinations of matrices .
splitting methods , on the other hand , belong to the class of integrators analysed in @xcite : taking as a starting point an efficient splitting method for the autonomous case , the scheme is formed by averaging the matrix @xmath3 at each step ( with different weights ) .
the new coefficients are obtained by solving the additional order conditions arising from the time - dependency .
compared with the general case , the number of order conditions is considerably reduced , which allows to get new schemes within this family in a rather straightforward way .
bader , blanes , casas and kopylov acknowledge the ministerio de economa y competitividad ( spain ) for financial support through the coordinated project mtm2013 - 46553-c3 - 3-p . additionally , kopylov has been partly supported by fellowship grisolia/2015/a/137 from the generalitat valenciana .
al - mohy , n.j .
higham and s.d .
relton , new algorithms for computing the matrix sine and cosine separately or simultaneously , siam j. sci . comput .
* 37 * ( 1 ) ( 2015 ) , a456-a487 .
p. alonso , j. ibnez , j. sastre , j. peinado , and e. defez , efficient and accurate algorithms for computing matrix trigonometric functions , _ j. comp . appl .
_ * 309 * , pp . 325332 , 2017 .
p. bader , s. blanes , solving the perturbed quantum harmonic oscillator in imaginary time using splitting methods with complex coefficients , in f. casas , v. martinez ( eds . ) , advances in differential equations and applications , sema simai springer series 4 , ( 2014 ) , pp .
217227 .
p. bader , s. blanes , f. casas , and e. ponsoda , efficient numerical integration of nth - order non - autonomous linear differential equations .
_ j. comp .
appl . math .
_ * 291 * , pp . 380390 , 2016 .
p. bader , s. blanes , e. ponsoda and m. seydaolu , symplectic integrators for the matrix hill equation .
_ j. comp .
_ in press .
p. bader , a. iserles , k. kropielnicka , and p. singh , efficient methods for linear schrdinger equation in the semiclassical regime with time - dependent potential , proc .
a * 472 * ( 2016 ) 20150733 .
d. s. bernstein .
matrix mathematics : theory , facts , and formulas , vol . 1 , 2009 . s. blanes , f. casas , a concise introduction to geometric numerical integration .
crc press , boca raton , ( 2016 ) .
s. blanes , f. casas , and a. murua , splitting methods for non - autonomous linear systems , _ int . j. comput
_ , * 84 * ( 2007 ) , pp . 713 - 727 .
s. blanes , f. casas , and a. murua , splitting methods in the numerical integration of non - autonomous dynamical systems , _ racsam _ , * 106 * ( 2012 ) , pp .
s. blanes , f. casas , j. a. oteo , j. ros . the magnus expansion and some of its applications . , * 470 * , ( 2009 ) , pp .
. s. blanes and p.c .
moan , practical symplectic partitioned runge - kutta and runge - kutta - nystrm methods , j. comput .
, * 142 * ( 2002 ) , pp . 313 - 330 .
a. j. dragt , lie methods for nonlinear dynamics with applications to accelerator physics , university of maryland , 2015 . m. drewsen and a. brner , harmonic linear paul trap : stability diagram and effective potentials , phys . rev . a 62 ( 2000 ) 045401 . ,
geometric numerical integration , 2nd ed . , springer , berlin , 2006 .
a. iserles and s.p .
nrsett , on the solution of linear differential equations in lie groups , philos .
royal soc .
london ser .
a 357 ( 1999 ) , pp .
9831019 . , simulating hamiltonian dynamics , cambridge university press , cambridge , 2005 .
w. magnus . on the exponential solution of differential equations for a linear operator . ,
* vii * , ( 1954 ) pp .
. w. magnus and s.winkler , hill equation .
wiley , new york , 1966 .
f. g. major , v. n. gheorghe , and g. werth , charged particle traps .
physics and techniques of charged particle field confinement , springer , 2005 .
moan , efficient approximation of sturm liouville problems using lie - group methods .
damtp , tech .
report 1998/na11 , university of cambridge , united kingdom ( 1998 ) .
mclachlan , theory and application of mathieu functions , dover , new york , 1964 .
h. munthe - kaas and b. owren , computations in a free lie algebra , philos .
royal soc .
london ser .
a , 357 ( 1999 ) , pp .
w. paul , electromagnetic traps for charged and neutral particles , rev .
modern phys .
62 ( 1990 ) 531540 . , numerical hamiltonian problems , chapman and hall , london , 1994 .
k. l. turner , s. a. miller , p. g. hartwell , n. c. macdonald , s. h. strogatz , and s. g. adams , five parametric resonances in a microelectromechanical system , nature 396 ( 1998 ) 149152 .
a. zanna , collocation and relaxed collocation for the fer and the magnus expansions , siam j. numer .
, 36 ( 1999 ) , pp .
|
two families of symplectic methods specially designed for second - order time - dependent linear systems are presented .
both are obtained from the magnus expansion of the corresponding first - order equation , but otherwise they differ in significant aspects .
the first family is addressed to problems with low to moderate dimension , whereas the second is more appropriate when the dimension is large , in particular when the system corresponds to a linear wave equation previously discretised in space .
several numerical experiments illustrate the main features of the new schemes .
second - order linear differential equations ; non - autonomous ; symplectic integrators ; magnus expansion ; matrix hill s equation 65l07 ; 65l05 ; 65z05
|
the super - kamiokande experiments @xcite in the past several years , joined by sno @xcite more recently , have given strong indications of neutrino oscillation that are corroborated and constrained by a variety of other experiments .
these experiments started a new era in the study of neutrino physics and offered the best indication to date of physics beyond the standard model . to further probe neutrino physics , there are a number of ongoing and planned neutrino oscillation experiments .
these experiments promise to give a full description of the phenomenology of neutrino mixing .
the most attractive experiments among the new generation of neutrino oscillation experiments are the long baseline ( lbl ) experiments .
they are performed in the controlled environment of traditional experimental high energy physics and expected to allow precision measurements of the oscillation parameters , including the leptonic @xmath0 phase .
notably , the recently approved superbeam facility , the high intensity proton accelerator ( hipa ) @xcite , which can provide intensive high energy neutrino beams from its 50 gev proton synchrotron , offers the possibility of even more desirable lbl experiments .
so far , the possibility of two lbl experiments using the hipa superbeam have been discussed .
one is hipa to kamiokande at a baseline length of about 300 km @xcite known as j2k , and the other is hipa to a detector located 2100 km away near beijing @xcite called h2b .
it is well - known that there are parameter ambiguities that are generally associated with oscillation measurements at a single baseline@xcite .
measurements at more than one baseline can be beneficial @xcite ; our previous studies @xcite showed that the joint analysis of the j2k and h2b experiments can offer extra leverages to resolve some of these ambiguities .
our results , however , also showed that @xmath0 violation effects can not be determined at 3@xmath7 level even with the joint analysis considered in the study , in which no antineutrino beams were used . since the leptonic @xmath0 phase and mass squared difference sign are pertinent information in the physics of neutrino mixings , which seems to be very different from that of the quark sector , it is necessary to find out how to pin down these neutrino mixing parameters accurately in new experiments .
it has been widely recognized that the neutrino factory @xcite is an ideal facility for the study of neutrino mixings . however , because of the technical and budgetary challenges faced with building a workable neutrino factory in the near future , and because of the availability of a conventional superbeam from hipa in about five years , it is obviously advisable that we explore the full potential of the hipa superbeams . in this work ,
we examine in further detail the measurement of the @xmath0 phase and the mass squared difference sign from the joint analysis at the two long baselines with specific hipa superbeams and more suitable detectors .
the neutrino beams are the 2-degree off - axis superbeams for the lbl at 300 km and a narrow band superbeam for vlbl at 2100 km .
we evaluate the event rates and investigate their sensitivities to the @xmath0 phase and the sign of @xmath8 .
taking into account all possible experimental errors , we find that a fairly precise measurement of the @xmath0 phase , the sign of the mass squared difference and the mixing angle @xmath9 is possible but requires : ( 1 ) the joint analysis at the two baselines ; ( 2 ) that both a @xmath10 beam and a @xmath11 beam are needed at 300 km ; ( 3 ) that a @xmath11 beam is needed at 2100 km if @xmath8 is negative ; and ( 4 ) a significant increase in the statistics at both 300 km and 2100 km .
we find that @xmath12 can be probed to very small values , depending on the value of the @xmath0 phase . in sec .
ii we describe how the simulations are performed . in sec .
iii our results are presented .
a brief discussion and conclusion can be found in sec .
our oscillation analyses will be restricted to 3 flavors of active neutrinos .
the parameters of the system consists of 2 mass squared differences ( msd ) , 3 mixing angles and 1 measurable @xmath0 phase .
the unitary mixing matrix in the vacuum is parameterized as usual @xmath13 where @xmath14 , @xmath15 , and @xmath16 , @xmath17 defined for @xmath18 are the mixing angles of mass eigenstates @xmath19 and @xmath20 , and @xmath1 is the @xmath0 phase angle .
the three mass eigenvalues are denoted as @xmath21 , @xmath22 , and @xmath23 .
the two independent msd are @xmath24 and @xmath25 .
the inputs of the mixing angles and msd s are obtained from solar , atmospheric and reactor experiments : @xmath26 note that the sign of @xmath8 is unknown .
the currently favored large mixing angle solar solution requires @xmath27 . in lbl experiments the neutrino beam interacts with electrons contained in earth matter @xcite . in the present study we use the preliminary reference earth model @xcite , generally known as prem , for earth density profiles @xcite and numerically integrate the schrdinger equation that descibes the propagation of the neutrino in matter for the treatment of distance dependent matter density
however , we note that there exist more sophisticated approaches to earth matter effect , including both an updated average density profile known as the ak135 @xcite and treatments of uncertainties of the density profile @xcite which can affected the determination of the @xmath0 phase angle
. the hipa 50 gev proton synchrotron beam calls for a power of 0.77 mw in phase i , to be upgraded to 4 mw in phase ii . the superbeam provided by hipa can be a wide band beam ( wbb ) , a pulsed narrow band beam ( nbb ) , or an off - axis beam ( oab ) .
the wbb contains neutrinos with widely distributed energy . in a nbb
the neutrino flux is concentrated in a narrow range of energies , with maximum energy @xmath28 where the intensity is peaked , and the intensity decreases rapidly below @xmath28 .
an oab also peaks at a certain energy , but has a longer high - energy tail than that of a nbb .
more details of the various beam profiles can be found in ref .
@xcite . for the 2100 km baseline
we use a nbb @xcite with peak energy of 4 gev .
we have also investigated the beams of peak energies of 5 , 6 and 8 gev , but found that the 4 gev beam gives the best results . for 300 km
we use the 2-degree oab ( 2@xmath29oab ) which has a peak energy at 0.8 gev @xcite . the detector at 2100 km is assumed to be a water cerenkov calorimeter with resistive plate chambers @xcite located in beijing , tentatively called the beijing astrophysics and neutrino detector ( band ) .
the size of the detector will be 100 kt at the beginning and can be upgraded to a much larger one depending on the physics requirements .
the detector at 300 km is initially the kamiokande detector of the present size of 22.5 kt and upgraded later to 450 kt . for the experimental errors we use 3@xmath7 throughout this work .
all independent errors , statistical and systematic , are added in quadrature . for the statistical error we used @xmath30 poisson errors as described in the appendix of ref .
@xcite , including a background at the 1% level of the rate of the survival channel . for the systematic error we assumed that the background is known at the 2% level as given in @xcite . to estimate the error due to the uncertainty in the measurement of the mixing angle @xmath31
, we assumed that @xmath32 is measured via the survival channel at @xmath33 km , with the event rate given by @xmath34 , where @xmath35 and @xmath36 is the number of events in the absence of oscillations .
then the statistical uncertainty on @xmath32 is @xmath37 we then find the variation in the rate in question for a 3@xmath7 deviation in @xmath31 , and added this in quadrature to the other 3@xmath7 errors described above to obtain the total 3@xmath7 error .
we consider three scenarios in the present investigation .
the scenarios are summarized in table 1 .
for scenario i the first stage involves a 5-year experiment with a water cerenkov detector of 22.5 kt at 300 km with the 2@xmath29oab @xmath10 beam @xcite from hipa at 0.77 mw .
this stage is contained in the plan of j2k @xcite .
the second stage of this scenario has hipa upgraded to 4 mw , as discussed in ref .
@xcite , to deliver a nbb @xmath10 beam to the water cerenkov detector of 100 kt at l=2100 km to run for 5 years .
scenario ii has an upgraded 4 mw hipa and calls for both @xmath10 and @xmath11 beams .
the experiment at 2100 km is the same as in scenario i. for 300 km , however , we assume a much larger water cerenkov detector of 450 kt .
it will run for 2 years with 2@xmath29oab @xmath10 beam , and then 6 years with the 2@xmath29oab @xmath38 beam .
scenario iii is similar to scenario ii , but calls for a much larger water cerenkov detector at 2100 km , to run either the @xmath39 or @xmath40 beam , for example , for 5000 kt - yr .
whether a @xmath41 or @xmath11 beam is delivered to the 2100 km site depends on the sign of @xmath3 , as will be explained below .
let us first describe briefly the strategy of our calculation .
there are 6 measurable oscillation parameters in the 3-flavor neutrino scheme .
there are two mass scales ; one is the atmospheric scale and the other the solar scale .
existing oscillation experiments have determined that the two mass scales are widely separated and therefore sensitive to different @xmath42 regions .
the lbl and vlbl we are considering are affected by the solar scale only at next to leading order , but can be strongly affected by the unknown parameters @xmath43 and @xmath9 . hence we will take @xmath44 and @xmath45 to be the values determined in solar experiments , as given in eq .
( [ input12 ] ) .
this leaves 4 parameters to be determined , but @xmath46 and @xmath31 are known already to a fair degree of accuracy from atmospheric neutrino experiments . therefore obtaining @xmath9 , @xmath1 , and
the sign of @xmath3 is the main goal of our calculation .
the current and soon to be online lbl experiments will determine @xmath46 and @xmath32 to a better accuracy ; @xmath47 and @xmath48 will be measured more precisely by kamland @xcite .
the first task of the scenarios we propose is to determine @xmath49 and @xmath32 more accurately , using the @xmath39 survival probability .
this will allow us to have as small an uncertainty as possible in the determination of the parameters @xmath9 and @xmath43 .
let us define @xmath50 as the number of charge
current events involving the @xmath51 charged lepton at the baseline @xmath52 .
we assume that @xmath53 , which depends on the @xmath39 survival probability , is used to determine @xmath54 and @xmath31 with as small an error as possible .
then in the various scenarios @xmath55 can not be used to determine @xmath9 and @xmath43 .
= 300 km with better statistics are made , the improved @xmath55 will be used to update the values of @xmath54 and @xmath31 . ] in this scenario we assume that only the @xmath10 beam is employed to run at @xmath52=300 km and 2100 km . since @xmath55 has already been used to determine @xmath54 and @xmath32 , we are left with three types of independent measurements for the determination of @xmath9 , @xmath1 and the sign of @xmath8 : @xmath56 , @xmath57 and @xmath58 .
the measurements of these three types of events form a surface in a three - dimensional space when @xmath9 and @xmath43 are varied in their allowed ranges .
the angle @xmath9 is constrained by the chooz reactor experiment @xcite and the @xmath0 phase is completely unconstrained .
therefore , we take their ranges to be : @xmath59 and @xmath60 ) .
such three dimensional surfaces , which are tube like , are displayed in fig .
the upper and lower surfaces are for negative and positive @xmath3 , respectively .
the closed curves around the axes of the tubes are traced out by varying @xmath1 from @xmath61 to @xmath62 , while the lines running parallel to the axes of the tubes are determined by varying @xmath12 from @xmath63 to @xmath64 . for fixed @xmath12 values ,
we then obtain the ellipses in fig .
[ fig2 ] .
when @xmath56 , the number of the @xmath65 appearance events at 300 km , is measured , it determines a closed curve which is obtained from the three - dimensional surface by a cut at a given value on the @xmath56 axis .
the value of @xmath56 does not determine @xmath9 directly since @xmath66 is unknown ; for each of the closed curves we obtain a definite relation between @xmath66 and @xmath12 when the sign of @xmath8 is given . we show in fig .
[ fig3 ] two sets of such relations for each @xmath3 sign .
as shown , we choose two extreme values of @xmath56 , each of which leads to a range of values for @xmath12 , depending on the sign of @xmath3 . for positive @xmath3 , the larger @xmath56 curve limits @xmath12 to the range ( @xmath67 ) , while the smaller one corresponds to @xmath12 lying in the range ( @xmath68 ) .
similarly , the ranges of the values of @xmath12 for negative @xmath3 can be read off from fig .
[ fig3 ] .
we plot in fig .
[ fig4 ] the two - dimensional curves with fixed @xmath56 .
note that the scale of the horizontal axis is logarithmic .
if the scale was linear , the curves would be ellipses .
an open square indicates the point on a curve with @xmath1=0@xmath29 , solid square 90@xmath29 , open circle 180@xmath29 , and solid circle 270@xmath29 .
we also show a representative 3@xmath7 error bar for each curve .
the assignment of statistical and systematic errors has been discussed in the preceding section .
the total error is dominated mostly by the statistical error .
one sees that although the sign of @xmath8 can be determined at the 3@xmath7 level , there is no sensitivity to the value of the @xmath0 phase .
in particular , the error in the @xmath69 channel is very large in comparison with the range of variation in the number of events when @xmath1 varies ; we will encounter similar situation in the next scenario . by including the @xmath11 beam aimed at the detector at 300 km , we have two more types of events , i.e. , @xmath71 and @xmath72 .
so , in addition to the three - dimensional surface in the @xmath56-@xmath57-@xmath58 space shown earlier in fig .
[ fig1 ] , we also have surfaces in the spaces @xmath56-@xmath71-@xmath72 and @xmath56-@xmath71-@xmath57 , as shown in figs .
[ fig5 ] and [ fig8 ] , respectively . with several fixed values of @xmath12
we obtain the curves shown in figs .
[ fig6 ] and [ fig9 ] .
we plot the two - dimensional projections of fixed @xmath56=1000 and 10000 in figs .
[ fig7 ] and [ fig10 ] .
we found in scenario i that the 2100 km data with just a @xmath39 source can determine the sign of @xmath8 at 3@xmath7 , but can not measure the @xmath0 phase ( see fig .
[ fig4 ] ) .
in contrast , as shown in fig .
[ fig7 ] , the 300 km data using both a @xmath41 and @xmath70 source can determine the @xmath0 phase in the ranges ( @xmath73 ) or ( @xmath74 ) , but can not distinguish between the two ranges since the measurement is only sensitive to @xmath75 . furthermore , unless @xmath43 is close to @xmath76 or @xmath77 the sign of @xmath8 can not be determined once all of the experimental errors , including the error in the determination of @xmath31 , are taken into account , leaving a four - fold ambiguity . the problem lies in the fact that @xmath78 is used ; as already noted in scenario i , survival data provide poor resolution to the @xmath0 phase , and the matter effect is small at the relatively short distance of 300 km .
six more three - dimensional plots which will contain either @xmath72 , or @xmath69 , or both , can be made , but they are not very useful in the present analysis because they involve the survival data . in order to obtain good resolution in the sign of the msd and to distinguish the two ranges of the @xmath0 phase as discussed in the preceding section , we have to use data of the electron flavor only .
hence we need two experiments with different @xmath42 ratios and one of them should be a vlbl for a good sensitivity to the matter effect .
this brings us to the combined analysis of @xmath56 , @xmath71 and @xmath57 , as shown in fig .
[ fig10 ] .
the sign of @xmath8 can be easily determined if @xmath12 is not too small and the @xmath0 phase can be measured with again the ambiguity between the two ranges ( @xmath73 ) and ( @xmath74 ) , as in the case of fig .
the problem lies in the fact that the resolution in @xmath57 is poor due to the low number of events , while the resolution of the 300 km @xmath79 is excellent .
so we have to increase the statistics at 2100 km .
this takes us to scenario iii below .
the situation of scenario ii can be improved if the statistics in @xmath57 are significantly increased .
this can be achieved by using a larger detector and/or running for a longer period of time for the @xmath57 measurement . for scenario
iii we set the detector size times the running time at @xmath52=2100 km to be 10 times larger than that of scenario ii , assuming that the number of events can be straightforwardly scaled up with the detector size .
the running at 300 km is the same as in scenario ii .
the resultant two - dimensional plot is shown fig .
[ fig11 ] . in this scenario ,
the sign of @xmath8 can be clearly determined at the 3@xmath7 level , even for @xmath56=1000 which corresponds to a very small @xmath12 lying in the range ( @xmath68 ) , as indicated by the dotted curves in fig .
[ fig11 ] .
if @xmath8 is positive , a reasonably accurate determination of @xmath43 can be made with no sgn(@xmath8 ) or @xmath9 ambiguity , and the @xmath11 beam is not needed at 2100 km even for very small @xmath12 in the range ( @xmath68 ) .
this is consistent with the results of ref .
@xcite , where it was found that a @xmath80 and @xmath81 measurement at short distance and a @xmath80 measurement at a long distance could resolve parameter ambiguities for @xmath82 . to see the sensitivity more clearly for positive @xmath8 , we replot the results in figs .
[ fig12 ] and [ fig13 ] respectively for @xmath56=10000 and 1000 .
we see that for @xmath83 , which corresponds to larger @xmath12 ( @xmath84 ) as shown in fig .
[ fig3 ] , the @xmath0 phase can be determined better than 10@xmath29 at 3@xmath7 for @xmath1 small or around 180@xmath29 .
the sensitivity deteriorates slowly when @xmath43 moves away from @xmath85 or @xmath86 , and the uncertainty becomes of the order of 25@xmath29 when @xmath1 is close to 90@xmath29 or 270@xmath29 . even for @xmath56=1000 , which corresponds to very small @xmath12 in the range of ( @xmath68 ) ,
the measurement of the @xmath0 phase is still reasonably good .
it is interesting to note that the sensitivity of the @xmath0 measurement near @xmath87=0@xmath29 and 180@xmath29 for @xmath88 is comparable to that of the much higher number of events of @xmath56=10000 .
hence , in this scenario , either case can establish whether or not @xmath0 in the lepton sector is violated if @xmath1 deviates by than 10@xmath29 from the @xmath0 conserving points of @xmath87=0@xmath29 or 180@xmath29 .
if @xmath8 is negative , the @xmath89 oscillation is the favorable channel to investigate . hence once it is clear that @xmath3 is negative ( see the next section for a detailed discussion ) , the @xmath11 beam should be delivered to 2100 km to run for 5000 kt - yr .
the results , which are the counterparts to fig .
[ fig11 ] , are shown in fig .
[ fig14 ] . with positron events at 2100 km the @xmath0 phase can be well measured . to see the sensitivity more clearly , we replot the results in figs . [ fig15 ] and [ fig16 ] for @xmath56=10000 and 1000 respectively .
the accuracy of the @xmath43 measurement for @xmath90 using the @xmath40 beam is about the same as that of the @xmath39 beam for @xmath91 , although the distinction between @xmath43 in the range ( @xmath74 ) and @xmath43 in the range ( @xmath73 ) is not as good for a @xmath40 beam with @xmath90 .
we have also done the analysis assuming a @xmath41 or @xmath70 nbb of peak energy 5 , 6 , or 8 gev is delivered to the detector at 2100 km .
we found that for these cases the ellipses in figs .
[ fig11 ] to [ fig15 ] are much flatter than for the 4 gev nbb , so that they do not do as well in resolving the degeneracy in @xmath43 .
we conclude that with a superbeam , such as that delivered by hipa , the joint analysis at two baselines , of which one is an lbl at 300 km and the other a vlbl at 2100 km , can determine the @xmath8 sign and give a reasonably precise measurement of the @xmath0 phase and @xmath9 . to achieve this ,
both @xmath39 and @xmath40 beams are needed for the lbl experiment .
the survival events @xmath92 and @xmath93 are generally insensitive to the matter and @xmath0 effects . the initial hipa @xmath39 beam with power 0.77 mw will run with exposure 22.5@xmath945 kt - yr at a detector at @xmath52=300 km to obtain both survival events @xmath53 and appearance events @xmath56 .
the former is used to improve the determination of the mixing angle @xmath31 and mass squared difference @xmath95 , so as to reduce the uncertainty of these crucial input parameters .
the latter can show the existence of an appearance signal for @xmath96 and find a crude relation between @xmath43 and @xmath9 as shown in fig .
[ fig3 ] .
a detailed determination of the oscillation parameters will require an upgrade of the hipa beam power to 4 mw .
using our studies in this paper as a guide , we suggest as one possibility the following experimental steps using the upgraded hipa beam : 1ex * stage 1 * : deliver a 4 mw @xmath41 2@xmath29oab to a 450 kt detector at a distance of @xmath52=300 km for 2 years .
the survival events @xmath53 are used to determine more precisely the parameters @xmath31 and @xmath97 .
the appearance events @xmath56 are used to refine the relation between @xmath43 and @xmath9 , as shown in fig .
[ fig3 ] .
* stage 2 * : a 4 mw @xmath41 nbb with peak energy around 4 gev is delivered to a detector at @xmath52=2100 km , to run for 100@xmath945 kt - yr .
the survival and appearance events @xmath69 and @xmath57 are used to determine the sign of @xmath98 , with the most sensitivity coming from @xmath57 ( see fig . [ fig4 ] ) . *
stage 3 * : a 4 mw @xmath99 2@xmath29oab is delivered to the 300 km baseline detector for 450@xmath946 kt - yr and @xmath78 and @xmath71 are obtained .
the data can only determine @xmath43 and @xmath100 up to a 2-fold degeneracy because of the poor separation between @xmath66 and @xmath101 in the @xmath78 measurement , as demonstrated in fig .
[ fig7 ] .
* stage 4 * : a 4 mw @xmath39 ( @xmath40 ) nbb with peak energy around 4 gev is delivered to the 2100 km baseline detector for 1000@xmath945 kt - yr if @xmath102 ( @xmath103 ) .
then at 3@xmath7 , the value of @xmath43 can be determined to about @xmath104 for values close to @xmath85 or @xmath86 , or to about @xmath105 for values close to @xmath106 or @xmath107 .
the distinction between @xmath43 in the ranges @xmath108 and @xmath109 is better for @xmath110 than for @xmath90 . the @xmath111 case is shown in figs . [ fig12 ] and [ fig13 ] and that of @xmath112 in figs .
[ fig15 ] and [ fig16 ] . 1ex it is apparent from our calculation that in order to obtain enough statistics to provide a reasonably precise measurement of @xmath9 and @xmath43 the total detector size and running time have to be sufficiently large .
we have not attempted a detailed optimization ; rather , we offer our calculation as an example for illustration .
a search is still required to determine the optimal conditions for the measurement .
eventually uncertainties in the earth matter density along a given baseline as well as uncertainties in the solar neutrino oscillation parameters @xmath45 and @xmath113 must also be taken into account .
we thank t. kobayashi for providing neutrino beam profiles and k. hagiwara for discussions .
we also thank l .- y .
shan and our colleagues of the h2b collaboration @xcite for discussions .
this work is supported in part by doe grant no . de - fg02-g4er40817 .
hipa : a multipurpose high intensity proton synchrotron at both 50 gev and 3 gev to be constructed at the jaeri tokai campus , japan has been approved in december , 2000 by the japanese funding agency .
the long baseline neutrino oscillation experiment is one of projects of the particle physics program of the facility .
more about hipa can be found at the website : `` http://jkj.tokai.jaeri.go.jp '' .
j2k : y. ito , et .
letter of intent : a long baseline neutrino oscillation experiment the jhf 50 gev proton - synchrotron and the super - kamiokande detector _
, jhf neutrino working group , feb . 3 , 2000
; see also arxiv : hep - ex/0106019 .
j. burguet - castell , m. b. gavela , j. j. gomez - cadenas , p. hernandez and o. mena , nucl .
b * 608 * , 301 ( 2001 ) [ arxiv : hep - ph/0103258 ] ; m. freund , p. huber and m. lindner , nucl .
b * 615 * , 331 ( 2001 ) [ arxiv : hep - ph/0105071 ] ; j. pinney and o. yasuda , phys .
d * 64 * , 093008 ( 2001 ) [ arxiv : hep - ph/0105087 ] ; h. minakata and h. nunokawa , jhep * 0110 * , 001 ( 2001 ) [ arxiv : hep - ph/0108085 ] ; v. barger , d. marfatia and k. whisnant , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed
. n. graf , arxiv : hep - ph/0108090 ; phys
. rev . * d 65 * , 073023 ( 2002 ) [ arxiv : hep - ph/0112119 ; t. kajita , h. minakata , and h. nunokawa , arxiv : hep - ph/0112345 ; g. barenboim , a. de gouvea , m. szleper , and m. velasco , arxiv : hep - ph/0204208 ; a. donini , d. meloni and p. magliozzi , arxiv : hep - ph/0206034 .
l. wolfenstein , phys . rev . *
d17 * , 2367 ( 1978 ) ; * d20 * , 2634 ( 1979 ) ; v. barger , k. whisnant , s. pakvasa and r.j.n .
phillips , phys . rev . *
d22 * , 2718 ( 1980 ) ; p. langacker , j.p .
leveille and j. sheiman , phys . rev . * d27 * , 1228 ( 1983 ) .
lian - you shan , bing - lin young and xinmin zhang , phys . rev .
* d66 * , 053012 ( 2002 ) [ arxiv : hep - ph/0110414 ] ; lian - you shan and xinmin zhang , phys .
* d65 * , 113011 ( 2002 ) ; lian - you shan , futian liu , yi - fang wang , changgeng yang , bing - lin young and xinmin zhang , in preparation .
b. jacobsson , t. ohlsson , h. snellman and w. winter , phys . lett . * 532 * , 259 ( 2002 ) [ arxiv : hep - ph/0112138 ] ; for a summary of different approaches to approximate the earth s density , see b. jacobsson , t. ohlsson , h. snellman and w. winter , talk given at nufact 02 , london , 2002 , arxiv : hep - ph/0209147 .
|
we propose to measure the @xmath0 phase @xmath1 , the magnitude of the neutrino mixing matrix element @xmath2 and the sign of the atmopheric scale mass squared difference @xmath3 with a superbeam by the joint analysis of two different long baseline neutrino oscillation experiments .
one is a long baseline experiment ( lbl ) at 300 km and the other is a very long baseline ( vlbl ) experiment at 2100 km .
we take the neutrino source to be the approved high intensity proton synchrotron , hipa .
the neutrino beam for the lbl is the 2-degree off - axis superbeam and for the vlbl , a narrow band superbeam .
taking into account all possible errors , we evaluate the event rates required and the sensitivities that can be attained for the determination of @xmath1 and the sign of @xmath4 .
we arrive at a representative scenario for a reasonably precise probe of this part of the neutrino physics .
= # 1,nucl . phys .
b * # 1 * , # 1,phys .
lett .
b * # 1 * , # 1,phys .
rev .
d * # 1 * , # 1,phys .
rev .
lett . *
# 1 * , # 1,z .
phys .
c * # 1 * , * measuring @xmath0 violation and mass ordering in + joint long baseline experiments with superbeams * + k. whisnant@xmath5 , jin min yang@xmath6 , bing - lin young@xmath5 + _ @xmath5 department of physics and astronomy , iowa state university , ames , iowa 50011 , usa + @xmath6 institute of theoretical physics , academia sinica , beijing 100080 , china _
|
we discuss models of directed polymer which have unbounded jumps introduced in @xcite .
see @xcite for the background and related works .
the previous paper shows the following three results : ( i ) the continuity of the free energy with respect to inverse temperature and the appearance probability of obstacles ; ( ii ) the asymptotic of the free energy as the appearance probability of obstacles goes to 1 ; ( iii ) the continuity of the time constant of first passage percolation ( fpp ) related to the model . in this paper , we show concentration bounds for the ground state energy , which is nothing but the passage time , and the free energy . as applications ,
we derive the so - called `` rate of convergence '' results and , based on them , we are able to remove the restriction left in @xcite on a parameter in ( i ) and also give an alternative proof of ( iii ) .
in fact , a concentration bound for the lower tail for fpp is shown in @xcite , proposition 3.1 .
however the upper tail is significantly more difficult as we need to control the maximum jump of the optimal path , see section [ key ] below . in this paper , we succeed in controlling jumps not only of the optimal path but also of low energy paths in the directed polymer model . as a result
, we can show the concentration for the passage time and the free energy .
+ let @xmath0 be the random walk on @xmath1 starting from @xmath2 and with the transition probability @xmath3 where @xmath4 for @xmath5 and @xmath6 is a function of the form @xmath7 where @xmath8 and @xmath9 is a positive constant determined as to be @xmath10 ( for the choice of @xmath11 , see remark [ choice ] below ) .
the random environment is modelled by independent and identically distributed bernoulli random variables @xmath12 with parameter @xmath13 ; @xmath14 we introduce the hamiltonian @xmath15 and define the partition functions by @xmath16 \textrm { for $ \beta\in{\mathbb{r}}$ and } z_n^{\eta,-\infty}=p(h_n^\eta=0),\ ] ] where @xmath17 $ ] denotes the expectation with respect to @xmath18 .
note that @xmath19 is positive for @xmath20-almost every @xmath21 , since the random walk has unbounded jumps .
an important quantity in this model is the so - called free energy defined by @xmath22 , \end{split}\end{aligned}\ ] ] whose existence can be shown by using subadditive ergodic theorem . in @xcite ,
the continuity property and some asymptotic beahvior of the free energy were studied .
next , we introduce first passage percolation models related to this directed polymer
. denote the ( scaled ) points where @xmath23 by @xmath24 with the scaling factor @xmath25 ( @xmath26 ) .
this scaling is natural since @xmath27 converges as @xmath26 to the poisson point process @xmath28 on @xmath29 whose intensity is the product of the counting measure and lebesgue measure . with some abuse of notation we will frequently identify @xmath27 , and more generally any point measure , with its support . given a realization of @xmath27 , we define the minimum passage time from @xmath2 to @xmath30 by @xmath31 this is the directed version of the howard - newman s euclidean fpp model in @xcite .
now , a direct application of the subadditive ergodic theorem shows that the limit @xmath32 exists @xmath20-almost surely and equals to @xmath33 $ ] .
the limit @xmath34 , so - called time constant , is non random .
observe also that definition ( [ eq : passtime ] ) makes perfect sense when @xmath35 , yielding a limit @xmath36 in ( [ mu_p ] ) .
it is again shown in @xcite that @xmath34 is continuous as @xmath37 .
this fpp is related to the _ ground state at @xmath38 _ of the directed polymer introduced above .
namely , @xmath39 furthermore , it is shown in @xcite that as @xmath40 , the ground state gives dominant contribution to the free energy : @xmath41 as one sees above , the subadditive ergodic theorm is useful to show the existence of the free energy and the time constant .
however , it prevent us from getting more information about them , such as the continuity . in this paper
, we will use a method of concentration of measures to get the rate of convergence in and and derive the continuity results as a corollary .
+ first , we shall present the results for fpp .
the first result is the concentration of the passage time around the mean .
[ conc;fpp ] for any @xmath42 there exist positive constants @xmath43,@xmath44 , and @xmath45 $ ] which are independent of @xmath13 and @xmath30 such that for any @xmath46 , @xmath47|>n^{\frac{1}{2}+\delta})<{}c_1\exp\{-c_2n^{\lambda}\}.\label{con1;fpp}\ ] ] the next result is about the rate of convergence .
it implies that the fluctuation exponent ( see @xcite ) is @xmath48 or less .
[ 21 ] for any @xmath49 there exist positive constants @xmath50 which are independent of @xmath13 such that for any @xmath46 , latexmath:[\ ] ] and letting @xmath358 , @xmath359 \le n\mu_p+cn^{\chi}$ ] follows . on the other hand ,
since it is easy to check the usual subadditivity , we have . from and , we have .
+ + the proof is almost the same as that for fpp , so we only mention the outline here . to prove theorem [ 21 ] , we again use the argument of zhang .
let @xmath360 let @xmath361 and we define an event @xmath350 by @xmath362^d}{z(n,2n , x,\omega)},z_{2n}(\omega)\leq{}2\tilde{z}_{2n}(\omega)\}.\ ] ] one can show , as in the case of fpp , that for sufficiently large @xmath30 , @xmath363 on this event , @xmath364^d}\sum_{\underset{\gamma(0)=0}{\gamma:\gamma(n)=x}}e^{-f_n(\gamma;\omega)}z(n,2n , x,\omega)\right\ } } , \end{split}\ ] ] where we have used the argument below in the second inequality and divided paths at its @xmath30-th point in the third ineqality .
thanks to the choice of @xmath350 , this is further bounded from above by @xmath365^d}\sum_{\underset{\gamma(0)=0}{\gamma:\gamma(n)=x}}e^{-f_n(\gamma;\omega)}z(n,2n,0,\omega)\right\}}\\ & \leq\log{(z_n(\omega)z(n,2n,0,\omega))}= \log{z_n(\omega)}+\log{z(n,2n,0,\omega)}. \end{split}\ ] ] from the concentration and this , for any @xmath49 , there exist constant @xmath317 independent of @xmath30 such that , @xmath366\leq{}2q[\log{z_n}]+cn^{\chi}\ ] ] in a similar way to the proof of theorem [ 21 ] , we have for any @xmath105 , @xmath367}{2^kn}\leq{}\frac{q[\log{z_{n}}]}{n}+c'n^{\chi-1},\ ] ] where @xmath368 is a constant independent of @xmath30 . as @xmath358
, we have @xmath369+c'n^{\chi-1 } , .\ ] ] finally , we shall derive the converse estimate by a similar way . indeed
, we replace @xmath350 by @xmath370^d}{z(n,2n , x,\omega)},z_{n}(\omega)\leq{}2\tilde{z}_{n}(\omega)\}$ ] , and on this event , we have @xmath371 which implies @xmath372-c'n^{\chi-1}.\ ] ] this implies . from and
, it is immediate to prove .
thanks to the non - random fluctuation results , it suffices to show the continuity of @xmath373 $ ] and @xmath374 $ ] for fixed @xmath46 . [ finite - vol ] for any @xmath46 fixed , @xmath359 $ ] is continuous function of @xmath54 $ ] .
note that @xmath375 is continuous on the set of locally finite point configurations with respect to the vague topology .
indeed for any @xmath82 , the definition of passage time tells us that only points inside a compact set matter . by local finiteness
we can choose the compact set in such a way that its boundary contains no points of @xmath82 .
now if @xmath376 vaguely as @xmath260 , then the points of @xmath377 inside the compact set converge to those of @xmath82 in the hausdorff metric and then it easily follows that @xmath378 . by using the skorohod representation theorem , we may assume that @xmath20-almost surely , @xmath379 vaguely as @xmath380 . then by continuity , @xmath381 as @xmath380 , @xmath20-almost surely . from this , @xmath222 is continuous in @xmath54 $ ] . since the uniform integrability of @xmath382}$ ] follows from lemma [ greedy ] , @xmath383 convergence follows .
the following lemma can be proved by a similar way to lemma [ finite - vol ] .
[ iii ] @xmath374 $ ] is jointly continuous on @xmath67 . from lemmas
[ finite - vol ] and [ iii ] , we obtain corollaries [ conti1 ] and [ conccc;fe ] .
we gratefully acknowledge useful conversations with ryoki fukushima .
s. boucheron , g. lugosi , and p. massart . .
oxford university press , oxford , 2013 .
f. comets , r. fukushima , s. nakajima , n. yoshida . limiting results for the free energy of directed polymers in random environment with unbounded jumps c. d. howard and c. m. newman .
geodesics and spanning trees for euclidean first - passage percolation . , 29(2):577623 , 2001 .
h. krug and h. spohn .
kinetic roughening of growing surfaces . in : c.godreche ed . , cambridge university press , 1991 q. liu and f. watbled . exponential inequalities for martingales and asymptotic properties of the free energy of directed polymers in a random environment . , 119(10):3101 - 3132 , 2009 .
|
we study the free energy and its relevant quantity for the directed polymer in random environment . the polymer is allowed to make unbounded jumps and the environment is given by the bernoulli variables .
we first establish the concentration of the ground state energy of polymer at zero temperature .
secondly , we also prove the same property of the free energy at finite temperature . in the proof ,
we use the fact that the maximum jump of any polymer nearly minimizing energy is not too large with high probability .
this is an interesting property itself from the first passage percolation viewpoint .
|
dilute magnetic impurities in metal bring the single - impurity kondo effect @xcite .
the anti - ferromagnetic coupling between spins on impurities @xmath1 , such as the rkky interaction , would compete with the kondo effect . to study such competition effect ,
the two impurities in metal have been studied extensively @xcite . if the kondo binding energy is much larger than @xmath1 ( @xmath2 ) , each local spin on the magnetic impurity forms the kondo singlet state with the conduction electrons . on the other hand for @xmath3 ,
the two local spins form the local spin singlet state . from the numerical renormalization group ( nrg )
calculation , jones
_ had pointed out that the transition between the two states occurs as a quantum critical phenomenon @xcite .
however the advanced investigation made clear that the critical transition is an artifice of the model neglecting the parity splitting terms , such as the d - d hopping term between the impurity atoms @xcite . in this paper
we will investigate the effect of the d - d hopping term in double - quantum - dot ( dqd ) systems in detail , and give the new realization to the two - impurity kondo problem
. it might be difficult to observe the two - impurity effect in metal systems as pointed out by previous studies , because the alloy contains many types of impurity pairs , and because the coupling between impurities is fixed in each material .
recently , the kondo effect is observed in the single quantum dot systems @xcite .
the experimental data show good agreement with the results of numerical calculation based on the single - impurity anderson model @xcite .
these works demonstrated that the quantum dot systems are suitable for sensitive experiment of the kondo problem . on the dqd systems
, each dot corresponds to an impurity atom , and the coupling between the dots can be changed freely by applying the split gate voltage between the dots @xcite .
it would be expected that we can investigate the two - impurity effect systematically in the dqd systems .
for the dqd systems , which the two dots are connected to the left lead and the right lead in series as ` lead - dot - dot - lead ' , there are several theoretical works including the kondo effect @xcite .
we have reported the large enhancement of the tunneling conductance through the two dots when the condition @xmath4 holds , by using the nrg calculation @xcite . here
@xmath5 is the anti - ferromagnetic kinetic exchange coupling between the two dots , @xmath6 is the hopping between the two dots , @xmath7 is the coulomb repulsion on the dot , and @xmath8 is the kondo temperature at @xmath9 .
we note that the anti - ferromagnetic coupling is the inevitable effect due to the kinetic process @xmath6 and the coulomb repulsion on the dot @xmath7 .
there are investigations with the slave boson mean field theory ( sbmft ) .
_ had already studied the same model of us , however they could not find the relation @xmath10 on the peak of the conductance pointed out by us because the sbmft can not treat the kinetic exchange process properly @xcite .
_ introduced the anti - ferromagnetic coupling @xmath1 between the two dots by artifice in the model , and discussed the effect related to the critical transition on the conductance by using the sbmft @xcite .
however the introduction of the artificial @xmath1 in the model and the calculation of the conductance within the sbmft framework bring some questions as follows : ( a ) does the effect of anti - ferromagnetic coupling pointed by georges _
actually appear in the dqd systems ? , because the hopping itself breaks the quantum critical transition @xcite .
if it appears , however , ( b ) how the conflicting effects of @xmath6 , the kinetic exchange coupling that would cause the critical transition and the parity splitting that suppress the critical transition , compete ?
and then , ( c ) how they appear in the conductance ?
since the sbmft could not treat the kinetic exchange process properly , this approximation for the two - impurity kondo problem like the dqd systems seems to be unfavorable .
the reliable calculation is necessary for such a sensitive problem . in this paper
, we present the detailed investigation of the kondo effect in the dqd systems .
the numerical calculation is performed by using the nrg method .
this numerical method is known to be a reliable one for the two - impurity kondo problem @xcite .
we calculate the tunneling conductance through the two dots .
we note that some preliminary results were presented at sces98 @xcite , and the one of the central results was presented at lt22 @xcite .
we find that the following states continuously appear when the hopping between the two dots increases from very small value ; ( i ) kondo singlet state ( @xmath11 , @xmath12 ) , ( ii ) singlet state between local spins on the dots ( @xmath11 , @xmath13 ) , and ( iii ) double occupancy in the bonding orbital of the two dots ( @xmath14 ) .
the conductance shows peaks at the transition regions between these states .
the ` main peak ' at the boundary between ( i ) and ( ii ) with the condition @xmath15 has the unitarity limit value of @xmath0 because of coherent connection through the lead - dot - dot - lead . especially for the strongly correlated cases ,
the width of the main peak becomes very narrow and the characteristic temperature of the peak is largely suppressed compared with the kondo temperature of the single dot systems @xmath8 .
these anomalies of the main peak closely relate to the quantum critical phenomenon in the two - impurity kondo problem .
the quantitative calculation in this paper gives the new realization for the two - impurity kondo problem , and suggests the possibility of the systematic study of the anomalous two - impurity kondo effect in the dqd systems .
the formulation is presented in [ sec:2 ] .
the numerical results are presented in [ sec:3 ] .
the summary and discussion are given in
[ sec:4 ] .
we investigate the following model hamiltonian for the dqd systems that the two dots are connected to the left lead and the right lead in series ; @xmath16 @xmath17 gives the electrons in the left and right leads .
@xmath18 gives that in the left and right dots .
@xmath19 gives the tunneling between the left lead and the left dot , and between the right lead and the right dot .
the suffices @xmath20 , @xmath21 mean the left and the right , respectively .
@xmath22 is the annihilation operator of the electron in the left lead , @xmath23 is that in the left dot .
@xmath24 is the number operator of the left dot .
@xmath25 is the energy of the state @xmath26 in the left lead .
@xmath27 is the energy of the orbital in the left dot .
the quantity @xmath6 is the matrix element between the left and right dots , and we call it as the hopping between the dots hereafter .
@xmath28 is the coulomb interaction between the electrons in the left dot .
@xmath29 is the matrix element between the left dot and the left lead .
here we consider only the single orbital in each of the dots .
this situation is justified in the case that the typical energy splitting between the orbitals in the dot is larger than the typical broadening of the energy levels , @xmath30 , and when the temperature is smaller than the typical coulomb repulsion between the electrons in the dots , @xmath31 @xcite .
( the kondo effect is not important in the case of @xmath32 . )
we consider only the on - site coulomb interaction between the electrons .
furthermore , we consider only the nearest neighboring tunneling , between the dot and its adjacent - site lead , between the two dots .
the energies @xmath27 , @xmath33 can be changed by applying the gate voltage on the dots .
@xmath34 ( @xmath35 ) can also be changed by applying the split gate voltage between the left ( right ) dot and the left ( right ) lead .
@xmath6 can be changed by applying the split gate voltage between the left dot and the right dot . in this paper
we consider only the symmetric case with respect to the exchange of the left and the right .
this situation is written with the following relations ; @xmath36 , @xmath37 , and @xmath38 , @xmath39 is the density of states in the leads . here
we consider that there are no @xmath26-dependence in the matrix element and the density of states . )
the model can be mapped into the two - channel anderson hamiltonian by the unitary transform for the operators of the dots and the operators of the leads @xcite .
furthermore we consider the situation that there is one electron in each dot by adjusting the gate voltage on the dot , @xmath40 , @xmath41 .
we solve the hamiltonian by using the nrg method , and calculate the conductance from the current correlation function within the linear response theory @xcite .
( for detailed calculation of the conductance , see appendix of ref .
@xcite . ) at zero temperature , the conductance can be re - written by using the effective parameters of the fixed point non - interacting anderson hamiltonian as follows ; @xmath42 we have used the relation , @xmath43 , @xmath44 , ( @xmath45 ) at @xmath46 . here the suffix @xmath47 denote the even and odd parity orbitals in the two dots .
we note that the even orbital is the bonding orbital , and the odd orbital is the anti - bonding orbital .
we now consider the case of @xmath48 , then @xmath49 , @xmath50 . here
@xmath51 is the effective hopping between the dots , and @xmath52 is the effective hybridization strength between the leads and the dots . at @xmath46 , we calculate the effective parameters from the analysis of the flow chart of the renormalized energy level structure in the nrg calculation , and then calculate the conductance from eq .
( [ eq : g_t=0 ] ) .
in numerical calculation we choose the half of the band width as an energy unit .
the coulomb repulsion is fixed at @xmath53 through this paper .
we calculate the conductance as a function of the hopping @xmath6 for various hybridization strength @xmath54 .
( as noted in previously , @xmath6 and @xmath54 can be changed by applying the split gate voltage between the dots , between the dots and the leads , respectively . ) the gate voltage on the dots is fixed at @xmath55 , then the dqd is in the half - filled case , i. e. each dot contains one electron . in [ sec:3.1 ] and
[ sec:3.2 ] we present the numerical results at zero temperature @xmath46 , and in [ sec:3.3 ]
we present the results at finite temperatures .
first we present the conductance in the strongly correlated case with the hybridization strength satisfying @xmath56 , ( @xmath57 , i.e. @xmath58 ) .
we show the conductance at @xmath46 as a function of the hopping @xmath6 in fig .
[ fig:01 ] .
( the occupation number and the phase shift are also shown in fig . [ fig:01 ] . ) there are two peaks in the conductance , the large peak near @xmath59 , and the small peak near @xmath60 .
( hereafter we call the large peak as the ` main peak ' . )
why these peaks appear ?
in the later paragraph we will analyze various quantities for the parameter cases showing the peaks .
= 8.5 cm = 8.5 cm the density of states on the even orbital of the two dots , @xmath61 , for several @xmath6 cases is shown in fig .
[ fig:02 ] .
( the relation @xmath62 holds for @xmath63 , where @xmath64 is the density of states on the odd orbital of the two dots . ) at @xmath9
, there is the kondo peak on the fermi energy .
( fermi energy corresponds to @xmath65 . ) naturally , this kondo peak is caused by the kondo singlet states between the left lead and the left dot , and between the right lead and the right dot .
as @xmath6 increases to @xmath66 , the conductance has the main peak and the strength of @xmath67 becomes half of that at @xmath9 . in this region
we can consider that the kondo effect with the spins on the orbitals extending the two dots , the even and the odd orbitals , occurs .
as @xmath6 still increases , the conductance decreases rapidly , and the strength of @xmath68 is largely suppressed as shown at @xmath69 .
this suppression means the disappearance of the kondo coupling between the leads and the dots .
@xmath6 becomes still large , the conductance has the small peak at @xmath60 . at @xmath70 , the density of states on the even and odd orbitals
have peaks at @xmath71 , respectively , from fig .
[ fig:02 ] . at the same time
the occupation numbers begin to change as shown in fig .
[ fig:01 ] .
( @xmath72 , @xmath73 at @xmath74 . ) here we note the following two points : first , the condition @xmath15 holds at @xmath75 , where @xmath76 .
( @xmath77 is the kondo temperature at @xmath9 , with the expression @xmath78 $ ] @xcite , then @xmath79 at @xmath80 . )
second , as seen from fig .
[ fig:01 ] , the occupation numbers of the even and the odd orbitals for @xmath81 are almost same with each other , @xmath82 . for @xmath83 , the two electrons occupy the even orbital .
the border between them is at @xmath84 . above analysis
implies the following scenario . in the case of @xmath85 ,
the hopping @xmath6 causes the anti - ferromagnetic kinetic exchange coupling , @xmath86 . for smaller hopping case with @xmath87
, there are the kondo singlet states between the left lead and the left dot , and between the right lead and the right dot each other .
as @xmath6 increases and then @xmath88 , the two local spins on each dot form the local singlet state .
at the transition region between two states we have a main peak with the unitarity limit value of @xmath0 in the conductance .
this will indicate that the leads and the dots are coherently connected by the even and the odd orbital states .
when @xmath6 becomes still large and the condition @xmath89 holds , the local spins do not appear , instead , the two electrons occupy the even orbital .
the small peak of the conductance reflects the transition of the electronic states in the dqd .
= 8.5 cm here we show the effective parameters @xmath90 and @xmath52 as a function of @xmath6 in fig .
[ fig:03 ] .
( we note that the effective parameters have been used already for the calculation of the conductance shown in fig .
[ fig:01 ] . ) in @xmath91 case , the effective parameters behave as @xmath92 , @xmath93 .
then the conductance coincides with the non - interacting one when one substitutes the effective parameters into eq .
( [ eq : g_t=0 ] ) . as @xmath6 increases with the condition near @xmath94 , @xmath95 once decreases , and it has local minimum , and then it increases . at the same time
the slope of @xmath51 once decreases and then increases .
when @xmath51 and @xmath52 coincide with each other , the conductance has a peak at @xmath96 , i. e. @xmath97 .
as @xmath6 increases slightly beyond this point , the conductance sharply decreases because @xmath95 decreases to the minimum even though @xmath90 increases .
here we stress that the relation @xmath97 holds when @xmath98 , and at the same time @xmath52 becomes very small in the transition region .
when @xmath6 increases further , the ratio @xmath99 increases gradually in the region @xmath100 . at @xmath101 ,
the ratio @xmath102 begins to decreases and then increases .
therefore the conductance shows a broad peak near the region @xmath101 . for @xmath103 case ,
the effective parameters behave @xmath104 , @xmath105 .
we note that the conductance has the expression of the non - interacting one itself in @xmath103 region .
finally we compare between the phase shift and the occupation number shown in fig .
[ fig:01 ] .
the phase shift of the odd orbital @xmath106 rapidly changes from @xmath107 to @xmath108 near @xmath96 , even though @xmath109 still remains at @xmath110 .
friedel s sum rule in each channel does not hold , as already pointed out previously @xcite .
it seems that this behavior is enhanced when the anti - ferromagnetic coupling between the two sites competes with the kondo effect . in this subsection
we present the numerical results of the conductance for various @xmath111 cases within @xmath112 .
( @xmath113 .
the hybridization strength is changed in @xmath114 , and the coulomb repulsion is fixed at @xmath53 . )
we confirm the scenario shown in the previous subsection that @xmath4 holds at the main peak of the conductance with @xmath115 .
we also demonstrate how the kinetic exchange process appear in the conductance for arbitrary @xmath116 cases .
the calculated conductance is shown in fig .
[ fig:04 ] .
the horizontal axis is the hopping normalized by the hybridization strength , @xmath117 . from inset of the fig .
[ fig:04 ] , the conductance almost overlaps on the non - interacting curve in the region @xmath118 and @xmath119 .
exactly , these regions should be classified @xmath120 and @xmath121 , respectively , from the analysis in the previous subsection .
the conductance is very small in these regions , however this uniform properties should be useful to arrange the experimental data under uncertain @xmath122 cases .
= 8.5 cm all curves have a main peak with strength @xmath0 . for weakly correlated cases of @xmath123 ,
the conductance almost coincides with the non - interacting one through all @xmath6 region . as @xmath122 increases , the main peak shifts to the small @xmath117 side , and the peak width becomes narrower .
we already found the relation @xmath97 at the main peak position for @xmath124 case in [ sec:3.1 ] .
here we show the ratio @xmath125 at the main peak position for various @xmath116 cases in table [ table:01 ] .
we can see the relation @xmath126 commonly @xcite .
( the relation at the main peak , @xmath97 , would be generalized to @xmath127 to the weakly correlated cases , where @xmath128 is the singlet binding energy between the two dots . ) from the analysis in the previous and present subsection we can conclude the following effect of the hopping term . for the small @xmath6 case with @xmath129 , `` ( i )
the kondo singlet state is formed on the left ( right ) dot with its adjacent - site lead '' . on the other hand for the large @xmath6 case with @xmath130 , `` ( ii ) the local spins on each of the dots couple as the singlet state '' . in the intermediate region ,
the kondo effect of the local spins on the orbitals extending on the two dots ( i.e. even and odd orbitals ) occurs .
the main peak of the conductance appears around the boundary between ( i ) and ( ii ) reflecting the coherent connection of the leads and the dots .
as @xmath122 increases , the kondo temperature @xmath8 exponentially decreases , the condition @xmath4 holds at the smaller @xmath117 , then the main peak shifts to the smaller @xmath117 side . at the same time
, the width of the peak becomes extremely narrow compared with the decreasing of @xmath8 .
this fact has been already shown as the steep minimum of @xmath52 in fig .
[ fig:03 ] .
we note that this narrowing closely relates to the quantum critical transition between the kondo singlet state and the local singlet state in the two - impurity kondo model @xcite .
the shifting and narrowing behaviors shown here are also pointed out with the sbmft with artificial addition of the anti - ferromagnetic coupling between dots to the model @xcite .
however , the sbmft calculation should be checked by the method treating the kinetic exchange term properly . as noted in the introduction ,
the hopping term causes two conflicting effects on the critical transition of the two - impurity systems .
one is the kinetic exchange coupling @xmath131 , which causes the `` critical '' transition through the competition with the kondo effect .
another is the parity splitting , which suppresses the `` critical '' transition .
the calculation in this section is the first reliable quantitative results of the two - impurity kondo problem in the dqd systems .
there is also another small peak ( or shoulder ) structure for the strongly correlated cases of @xmath132 ( @xmath133 ) at larger @xmath6 side of the main peak . in the previous subsection
, we found that the small peak appears around the boundary between ( ii ) and ( iii ) .
however for the weakly correlated cases , the small peak could not be recognized because the condition of the border ( i)-(ii ) and ( ii)-(iii ) could not be distinguished clearly . .
ratios @xmath134 and @xmath135 at the main peak position of the conductance .
[ cols="^,^,^,^,^,^",options="header " , ] 2 finally , we note the accuracy of the conductance calculated from eqs .
( [ eq : g_p])-([eq : p ] ) by using the nrg method .
it is not so accurate at very high temperatures for the small peak . in the case of @xmath136 ,
two dots completely decouple , then the conductance should be zero .
however as shows in fig .
[ fig:10 ] , the calculated conductance has finite value in @xmath137 and it has maximum at @xmath138 .
thus the result at very high temperatures has ambiguities .
this improper finite conductance would be caused by estimation of eq .
( [ eq : g_p ] ) at @xmath139 instead of @xmath140 .
the finite value in the current spectrum at @xmath141 would reflect the largeness of the dynamical charge fluctuation in the dots .
we calculated the tunneling conductance through the two quantum dots that connected to the left lead and the right lead in series .
we investigated the effect of the kinetic exchange coupling between the dots , and also the competition with the kondo effect .
as the hopping between the two dots increased , ( i ) kondo singlet state , ( ii ) local spin singlet state , and ( iii ) molecular orbital like state with double occupancy in even state , continuously appeared . for @xmath11 cases , the kondo binding between the left ( right ) lead and the left ( right ) dot , @xmath8 , and the anti - ferromagnetic kinetic exchange coupling between the two dots , @xmath86 , competed .
the boundary between ( i ) and ( ii ) was characterized as @xmath10 , and the tunneling conductance showed a peak .
this peak had the unitarity limit value of @xmath0 reflecting the coherent connection through the lead - dot - dot - lead . at @xmath101 of the boundary between ( ii ) and ( iii )
, we had a small peak .
the system showed the strongly correlated behaviors for @xmath142 ( @xmath143 ) cases .
the borders of ( i)-(ii ) ( @xmath126 ) and ( ii)-(iii ) ( @xmath101 ) were clearly distinguished , then there were the two peak structures in the conductance .
furthermore the width of the main peak became steeply narrow .
the characteristic temperature of the main peak was strongly reduced compared with the kondo temperature of the single dot systems @xmath144 .
these anomalous behaviors of the main peak related to the quantum critical transition of the two - impurity kondo problem studied in previously .
though the hopping term had conflicting effects on the critical transition of the two - impurity kondo systems , generation of it through the kinetic exchange coupling and suppression of it due to the parity splitting , we found that we see the sign of the anomaly in the tunneling conductance . the quantitative calculation shown in this paper gave the new realization for the two - impurity kondo problem .
this investigation suggested the importance of the systematic study of the dqd systems for the two - impurity kondo problem .
this work was partly supported by a grant - in - aids for scientific research on the priority area `` spin controlled semiconductor nanostructures '' ( no .
11125201 ) from the ministry of education , science , sports and culture .
the numerical computation was partly performed at the supercomputer center of institute for solid state ( university of tokyo ) , the computer center of institute for molecular science ( okazaki national research institute ) , and the computer center of tohoku university .
one of the author ( w. i. ) is supported by jsps fellowship .
o. sakai , w. izumida and s. suzuki : proceedings of the 4-th int .
symposium on advanced physical field ( march , 1999 , tsukuba , japan ) , 143 ; w. izumida and o. sakai : physica b ( in press , proceedings of the international conference on strongly correlated electron systems , ( august , 1999 , nagano , japan ) ) ; w. izumida , o. sakai and s. suzuki : in preparation .
for example of the recent experimental studies of the double dot systems , t. h. oosterkamp , t. fujisawa , w. g. van der wiel , k. ishibashi , r. v. hijman , s. tarucha and l. p.
kouwenhoven : nature * 395 * , 873 ( 1998 ) ; t. fujisawa , t. h. oosterkamp , w. g. van der wiel , b. w. broer , r. aguado , s. tarucha and l. p. kouwenhoven : science * 282 * , 932 ( 1998 ) .
|
tunneling conductance through two quantum dots , which are connected in series to left and right leads , is calculated by using the numerical renormalization group method . as the hopping between the dots increases from very small value , the following states continuously appear ; ( i ) kondo singlet state of each dot with its adjacent - site lead , ( ii ) singlet state between the local spins on the dots , and ( iii ) double occupancy in the bonding orbital of the two dots .
the conductance shows peaks at the transition regions between these states .
especially , the peak at the boundary between ( i ) and ( ii ) has the unitarity limit value of @xmath0 because of coherent connection through the lead - dot - dot - lead . for the strongly correlated cases , the characteristic energy scale of the coherent peak shows anomalous decrease relating to the quantum critical transition known for the two - impurity kondo effect .
the two dots systems give the new realization of the two - impurity kondo problem .
@align#1#2.6ex 2
|
monotone matrices arise in many areas of mathematics , such as stability analysis @xcite , and bounds for eigenvalues and singular values axels90,axels03 . in many applications ,
one is interested in finding the smallest eigenvalue @xmath0 and the associated eigenvector @xmath1 of an irreducible nonsingular monotone matrix @xmath2 .
the smallest eigenvalue @xmath3 of a monotone matrix @xmath4 is defined as @xmath5 , where @xmath6 denotes the set of eigenvalues of @xmath4 . in @xcite , a real matrix @xmath4
is called monotone if and only if @xmath7 is a non - negative matrix .
the irreducible nonsingular @xmath8-matrices are one of the most important classes of matrices for applications such as discretized pdes , markov chains @xcite and electric circuits shi96 , and they have been studied extensively in the literature bpl94 .
it is well known that there exist some monotone matrices that are not @xmath8-matrices , such as matrices that can be written as a product of @xmath8-matrices .
there are some differences between an @xmath8-matrix and a monotone matrix .
for example , an @xmath8-matrix can be expressed in the form @xmath9 with a non - negative matrix @xmath10 and some constant @xmath11 , where @xmath12 denotes the spectral radius , see @xcite .
thus , the smallest eigenvalue @xmath0 of an irreducible nonsingular @xmath8-matrix @xmath4 is equal to @xmath13 .
in contrast , the smallest eigenvalue of a monotone matrix @xmath4 can only be expressed as @xmath14 .
however , the smallest eigenvalue retains the same properties @xcite , that is , the largest eigenvalue of an irreducible non - negative matrix @xmath7 is the perron root , which is simple and equal to the spectral radius of @xmath7 with a positive associated eigenvector . for the computation of the perron vector of a non - negative matrix @xmath10 ,
many methods exist @xcite but the power methods are not structure preserving and can not guarantee the desired positivity of approximations when the perron vector @xmath1 has very small components
. therefore , a central concern is how to preserve strict positivity of approximations to the perron vector . in @xmath15
, noda introduced an inverse iteration method with shifted rayleigh quotient - like approximations @xcite for non - negative matrix eigenvalue problems .
this iteration method is called noda iteration ( ni ) , and it has also been adapted to the computation of the smallest eigenvalue and the eigenvector of an irreducible nonsingular @xmath8-matrix xue96,axy02 .
the major advantages of noda iteration are structure preservation and global convergence .
more precisely , it generates a monotonically decreasing sequence of approximate eigenvalues that is guaranteed to converge to @xmath16 , and maintains the positivity of approximate eigenvectors .
furthermore , the convergence has been proven to be superlinear @xcite and asymptotically quadratic @xcite . in jia14 ,
the authors introduced two inexact strategies for noda iteration , which are called inexact noda iteration ( ini ) to find the perron vector of a non - negative matrix ( or @xmath8-matrix ) .
the proposed ini algorithms are practical , and they always preserve the positivity of approximate eigenvectors . moreover , the convergence of ini with these two strategies is globally linear and superlinear with convergence order @xmath17 , respectively . in this paper , we propose an inexact noda iteration ( ini ) to find the smallest eigenvalue and the associated eigenvector of an irreducible monotone matrix @xmath4 .
the major contribution of this paper is to provide two main relaxation steps for computing the smallest eigenvalue @xmath3 and the associated eigenvector @xmath1 , respectively .
the first step is to use @xmath18 as a stopping criterion for inner iterations , with @xmath19 , where @xmath20 is the current positive approximate eigenvector .
the second step is to update the approximate eigenvalues by using the recurrence relations @xmath21 , where @xmath22 is the next normalized positive approximate eigenvector , so resulting ini algorithms are structure preserving and globally convergent .
the above parameter @xmath23 is called the relaxation factor .
we then establish a rigorous convergence theory of ini with two different relaxation factors @xmath24 , and prove that the convergence of the resulting ini algorithms is globally linear , and superlinear with the relaxation factor @xmath24 as the convergence rate , respectively .
in fact , the inner iterations of ini ( or ni ) require the solution of ill - conditioned linear systems when the sequence of approximate eigenvalues converges to @xmath25 ( or @xmath26 ) . in order to reduce the condition number of the inner linear system
, we propose a modified noda iteration ( mni ) by using rank one update for the inner iterations , and we show that mni and ni are mathematically equivalent . for monotone matrix eigenvalue problems , we also develop an integrated algorithm that combines ini with mni , and we call this modified inexact noda iteration ( mini ) .
this hybrid iterative method can significantly improve the condition number of inner linear systems of ini .
the paper is organized as follows . in section 2
, we introduce the noda iteration and some preliminaries .
section 3 contains the new strategy for inexact noda iteration , and proves some basic properties for it . in section 4 ,
we establish its convergence theory , and derive the asymptotic convergence factor precisely . in section 5
, we present the integrated algorithm that combines ini with mni .
finally , in section 6 we present some numerical examples illustrating the convergence theory and the effectiveness of ini , and we make some concluding remarks in section 7 .
for any real matrix @xmath27 \in \mathbb{r}^{n\times n}$ ] , we write @xmath28 @xmath29 if @xmath30 @xmath29 for all @xmath31 .
we define @xmath32 $ ] . if @xmath28 , we say @xmath10 is a non - negative matrix , and if @xmath33 , we say @xmath10 is a positive matrix . for real matrices @xmath34 and @xmath35 of the same size , if @xmath36 is a non - negative matrix , we write @xmath37 .
a non - negative ( positive ) vector is similarly defined .
a non - negative matrix @xmath10 is said to be reducible if it can be placed into block upper - triangular form by simultaneous row / column permutations ; otherwise it is irreducible . if @xmath38 is not an eigenvalue of @xmath10 , the function @xmath39 is defined as @xmath40 @xmath41 denotes the acute angle of any two nonzero vectors @xmath42 and @xmath43 . throughout the paper
, we use a @xmath44-norm for vectors and matrices , and the superscript @xmath45 denotes its transpose .
we review some fundamental properties of non - negative matrices , monotone matrices and @xmath8-matrices .
a matrix @xmath4 is said to be monotone if @xmath46 implies @xmath47 for any positive vector .
another characterization of monotone matrices is given by the following well known theorem .
[ thm : mono ] @xmath4 is monotone if and only if @xmath4 is non - singular and @xmath48 a monotone matrix @xmath49is an @xmath8-matrix if @xmath50 , @xmath51 for @xmath52 [ thm : m ] let @xmath8 is a nonsingular @xmath8-matrix
. then the following statements are equivalent : 1 .
@xmath53 , @xmath54 for @xmath55 , and @xmath56 ; 2 .
@xmath57 with some @xmath58 and @xmath59 . for a pair of positive vectors @xmath60 and @xmath42 ,
define @xmath61 where @xmath62^t$ ] and @xmath63^t$ ] .
the following lemma gives bounds for the spectral radius of a non - negative matrix @xmath10 .
[ maxmin]let @xmath64be an irreducible non - negative matrix .
if @xmath65 is not an eigenvector of @xmath10 , then @xmath66 the noda iteration @xcite is an inverse iteration shifted by a rayleigh quotient - like approximation of the perron root of an irreducible non - negative matrix @xmath10 .
given an initial vector @xmath67 with @xmath68 , the noda iteration ( ni ) consists of three steps : @xmath69 the main task is to compute a new approximation @xmath70 to @xmath1 by solving the inner linear system ( [ eq : step1 ] ) . from lemma [ maxmin ] , we know that @xmath71 @xmath16 if @xmath72 is not a scalar multiple of the eigenvector @xmath1 .
this result shows that @xmath73 is an irreducible nonsingular @xmath8-matrix , and its inverse is an irreducible non - negative matrix .
therefore , we have @xmath74 and @xmath75 , i.e. , @xmath70 is always a positive vector if @xmath72 is .
after variable transformation , we get @xmath76 from the following relation : @xmath77so @xmath78 is monotonically decreasing .
the inexact noda iteration . based on the noda iteration , in @xcite the authors propose an inexact noda iteration ( ini ) for the computation of the spectral radius of a non - negative irreducible matrix @xmath10 . in this paper , since @xmath4 is a monotone matrix , i.e. , @xmath7 is a non - negative matrix , we replace @xmath10 by @xmath7 in ( [ eq : step1 ] ) , i.e. , @xmath79 when @xmath4 is large and sparse , we see that we must resort to an iterative linear solver to get an approximate solution . in order to reduce the computational cost of ( [ eq : exrelation ] ) , we solve @xmath80 in ( [ eq : exrelation ] ) by inexactly satisfying @xmath81which is equivalent to @xmath82where @xmath83 is the residual vector between @xmath84 and a@xmath72 . here , the residual norm ( inner tolerance ) @xmath85 can be changed at each iterative step @xmath86 .
[ monotone]let @xmath4 be an irreducible monotone matrix and @xmath87 be a fixed constant . for the unit length @xmath88 ,
if @xmath89 and @xmath83 in ( [ eq : inexactsys ] ) satisfies @xmath90then @xmath78 is monotonically decreasing and @xmath91 .
moreover , the convergence of ini is at least globally linear .
based on ( [ eq : inexactsys])-([inexact condi ] ) , we describe ini as algorithm [ alg : iiva ] . 1
. given @xmath92 , @xmath93 with @xmath94 , @xmath95 and @xmath96 .
2 . * for * @xmath97 3 . solve @xmath98 inexactly such that the inner tolerance @xmath99 satisfies condition ( [ inexact condi ] ) 4 .
normalize the vector @xmath100 .
5 . compute @xmath101 .
* until * convergence : resi@xmath102 . using the relation ( [ eq : ine relation ] ) ,
step 5 in algorithm alg : iiva can be rewritten as @xmath103unfortunately , @xmath7 is not explicitly available ; in other words , we need to compute @xmath104exactly for the required approximate eigenvalue @xmath76 .
hence , in the next section , we propose a new strategy to estimate the approximate eigenvalues without increasing the computational cost .
this strategy is practical and preserves the strictly decreasing property of the approximate eigenvalue sequence .
in order to ensure that ini is correctly implemented , we now propose two main relaxation steps to define algorithm [ alg : ini ] : * the residual norm satisfies @xmath105where @xmath106 with a constant upper bound @xmath107 * the update of the approximate eigenvalue satisfies@xmath108 1 . given @xmath109 , @xmath93 with @xmath94 , @xmath95 and @xmath96 .
2 . * for * @xmath97 3 . solve @xmath110 inexactly such that the inner tolerance @xmath99 satisfies condition ( [ eq : strategy ] ) 4 .
normalize the vector @xmath100 .
5 . compute @xmath111 that satisfies condition ( [ eq : updateeig ] ) .
* until * convergence : @xmath112 . in step 3 of algorithm [ alg : ini ]
, it leads to two equivalent inexact relation satisfying @xmath113we remark that @xmath111 in ( [ eq : updateeig ] ) is no longer equal to @xmath114 , and therefore that @xmath111 can not be clearly demonstrated to be greater than its lower bound @xmath115 .
the following lemma ensures that @xmath115 is still the lower bound of @xmath116 .
[ monotone2]let @xmath4 be an irreducible monotone matrix . for the unit length @xmath117 and the relaxation factor
@xmath118 , if @xmath119 , @xmath83 in ( [ eq : inexactsys2 ] ) satisfies condition ( [ eq : strategy ] ) and the approximate eigenvalue satisfies ( [ eq : updateeig ] ) , then the new approximation @xmath75 and the sequence @xmath120 is monotonically decreasing and bounded below by @xmath115 , i.e.,@xmath121 from ( [ eq : updateeig ] ) and @xmath122 , it is easy to know that @xmath123 is monotonically decreasing , i.e. , @xmath124 from ( [ eq : strategy ] ) , @xmath125which implies @xmath126 then @xmath127 consequently , @xmath128 is a nonsingular @xmath8-matrix , and the vector @xmath80 satisfies @xmath129this implies @xmath130 .
we now prove @xmath131 is bounded below by @xmath132 from ( [ eq : normaf ] ) , we have @xmath133and @xmath134hence , we obtain@xmath135then @xmath136from ( [ eq : ine relation2 ] ) , it follows that@xmath137combine ( [ eq : updateeig ] ) , ( [ eq : updatelam ] ) and ( [ eq : gap ] ) , then @xmath138by induction , @xmath123 is bounded below by @xmath139 i.e. , @xmath140 from lemma [ monotone2 ] , since @xmath123 is a monotonically decreasing and bounded sequence , we must have @xmath141 , where @xmath142 or @xmath143 .
we next investigate the case @xmath143 , and present some basic results ; this plays an important role later in proving the convergence of ini .
[ opposite]for algorithm [ alg : ini ] , if @xmath131 is converge to @xmath143 , then ( i ) @xmath144is bounded@xmath145(ii ) @xmath146(iii ) @xmath147 for some constant @xmath148 , where @xmath149 the acute angle of @xmath150 and @xmath1 .
( i ) . from ( [ eq : inqu_gamma ] ) , we get @xmath151(ii ) . from ( [ eq : updateeig ] )
it follows that @xmath152on the other hand , from ( [ eq : bdyk ] ) and ( [ eq : minfk ] ) we have@xmath153thus , it is holds that@xmath154(iii ) suppose there is a subsequence @xmath155 which converges to zero , then @xmath156by the definition of @xmath157 from ( [ eq : lam_relation ] ) and ( [ eq : gap ] ) , we have@xmath158from ( [ eq : minfk ] ) and ( [ eq : boundlambdagap ] ) , @xmath159this is a contradiction .
[ coslower]let @xmath160 be the unit length eigenvector of @xmath4 associated with @xmath161 . for any vector @xmath162 with @xmath163
, it holds that @xmath164 and @xmath165
in sections 4.14.2 , we will prove the global convergence and the convergence rate of ini .
furthermore , we will derive the explicit linear convergence factor and the superlinear convergence order with different @xmath166 .
for an irreducible non - negative matrix @xmath7 , recall that the largest eigenvalue @xmath115 of @xmath7 is simple .
let @xmath1 be the unit length positive eigenvector corresponding to @xmath115 .
then for any orthogonal matrix @xmath167 $ ] it holds ( cf .
@xcite ) that @xmath168
a^{-1}\left [ \begin{array}{cc } \mathbf{x } & v\end{array}\right ] = \left [ \begin{array}{cc } \rho ( a^{-1 } ) & \mathbf{c}^{t } \\ 0 & l\end{array}\right ] \label{eqn : partition}\]]with @xmath169 whose eigenvalues constitute the other eigenvalues of @xmath7 . therefore , we now define @xmath170similar to ( [ eqn : partition ] ) we also have the spectral decomposition @xmath168 a_{k}\left [ \begin{array}{cc } \mathbf{x } & v\end{array}\right ] = \left [ \begin{array}{cc } \varepsilon _ { k } & 0 \\ 0 & l_{k}\end{array}\right ] , \label{eqlpart2}\]]where @xmath171 . for @xmath172 ,
it is easy to verify that @xmath168 a_{k}^{-1}\left [ \begin{array}{cc } \mathbf{x } & v\end{array}\right ] = \left [ \begin{array}{cc } \varepsilon _ { k}^{-1 } & \mathbf{b}_{k}^{t } \\ 0 & l_{k}^{-1}\end{array}\right ] \mbox{\ \ with $ \mathbf{b}_k^t=-\frac{\mathbf{c}^tl_k^{-1}}{\varepsilon_{k}}$ } , \label{eqlpart3}\]]from which we get @xmath173and @xmath174 let @xmath175 be generated by algorithm [ alg : ini ] .
we decompose @xmath72 into the orthogonal direct sum by@xmath176with @xmath177 and @xmath178 @xmath179 @xmath180 being the acute angle between @xmath72 and @xmath1 .
so by definition , we have @xmath181 and @xmath182 evidently , @xmath183 if and only if @xmath184 , i.e. , @xmath185 . since @xmath186 in ini
, it holds that @xmath187 .
therefore , we have @xmath188as @xmath189 , it follows from the above relation that @xmath190using the above relation , we obtain @xmath191 note that if we solve the inner linear system exactly , i.e. , @xmath83 @xmath192 , we recover ni and get @xmath193since l. elsner @xcite proved the quadratic convergence of the proposed noda iteration , for @xmath86 large enough we must have @xmath194it follows that @xmath195 for any given positive constant @xmath196 .
therefore , we have @xmath197 for @xmath198 with @xmath199 large enough . [ main ]
let @xmath4 be an irreducible monotone matrix .
if the sequence @xmath200 is generated by ini with the relaxation strategies ( [ eq : strategy ] ) and ( [ eq : updateeig ] ) , then @xmath200 is monotonically decreasing and @xmath201 . from lemma [ monotone2 ] , the sequence @xmath123 is bounded and monotonically decreasing , and we must have either @xmath202 or @xmath203 .
next we prove by contradiction that , for ini , @xmath202 must hold .
suppose that @xmath204 .
it follows ( iii ) of lemma [ opposite ] show that @xmath205from ( ii ) of lemma [ opposite ] , we have @xmath206this implies the inner tolerance @xmath207 , i.e. , @xmath208 is suitably small . in addition , by lemma coslower , it holds that @xmath209 is uniformly bounded below by @xmath210 , therefore , @xmath211 for @xmath86 large enough . using ( [ eq : tan ] ) , ( [ eq : lowersin ] ) and
( [ eq : tan2 ] ) , we obtain@xmath212define @xmath213note that @xmath214 is a continuous function with respect to @xmath215 for @xmath216 .
then it holds that @xmath217 defined by ( [ niconv ] ) as @xmath218 .
therefore , from ( [ betak ] ) , for @xmath86 large enough we can choose a sufficiently small @xmath219 such that @xmath220for @xmath215 sufficiently small . as a result , we have @xmath221for @xmath198 with @xmath199 large enough and @xmath215 sufficiently small .
this means that @xmath222 , i.e. , @xmath223 . from ( iii ) of lemma [ opposite ] , @xmath224 is uniformly bounded below by a positive constant .
so @xmath225 and @xmath226 , a contradiction .
therefore the initial assumption @xmath203must be false .
theorem [ main ] has proved the global convergence of ini , but the results are only qualitative and do not tell us anything about how fast the ini method converges . in this subsection , we will show the convergence rate of ini with different relaxation factors @xmath24 .
more precisely , we prove that ini converges at least linearly with an asymptotic convergence factor bounded by @xmath227 for @xmath106 and superlinearly for decreasing @xmath228 , respectively .
from ( [ eq : updateeig ] ) we have @xmath229since @xmath230 , from ( [ eq : ratioerr ] ) and ( [ eq : updateeig ] ) , we have @xmath231 [ linearconv]for ini , we have @xmath232 moreover , if @xmath233 then @xmath234 , i.e. , the convergence of ini is at least globally linear . if @xmath235 then @xmath236 that is , the convergence of ini is globally superlinear . from ( [ eq : xaf_k ] ) and ( [ eq : yk+1bound ] ) , we have@xmath237from ( [ bkinv ] ) , we get @xmath238from theorem [ main ] , we know that @xmath239 and @xmath240 , from which it follows that @xmath241 and @xmath242 . on the other hand , since @xmath243 and @xmath244 , from ( [ rela3 ] ) we get @xmath245consequently , we obtain @xmath246leading to @xmath247 it can be seen from ( [ convfactor ] ) that if @xmath24 is small then ini must ultimately converge quickly .
although theorem [ linearconv ] has established the superlinear convergence of ini , it does not reveal the convergence order .
our next concern is to derive the precise convergence order of ini .
this is more informative and instructive because it lets us understand how fast ini converges .
[ quadconvini ] if the inner tolerance @xmath248 in ini satisfies condition ( [ eq : strategy ] ) with the relaxation factors @xmath249then ini converges quadratically ( asymptotically ) in the form of @xmath250for @xmath86 large enough , where the relative error @xmath251 . since @xmath252 we have@xmath253from ( [ eq : ratioerr ] ) , ( [ convfactor ] ) and ( [ eq : gamma ] ) , we have@xmath254dividing both sides of the above inequality by @xmath115 , we get @xmath255
in this section , we propose a modified noda iteration ( mni ) for a non - negative matrix , and show that mni and ni are equivalent .
thus , by combining ini ( algorithm [ alg : ini ] ) with mni we can propose a modified inexact noda iteration for a monotone matrix when @xmath73 tends to a singular matrix , the noda iteration requires us to solve a possibly ill - conditioned linear system ( eq : step1 ) .
hence , we propose a rank one update technique for the ill - conditioned linear system ( [ eq : step1 ] ) , i.e.,@xmath256 , \label{eq : newtonni}\]]where @xmath257 @xmath258 let @xmath259 in general , the linear system ( [ eq : newtonni ] ) is a well - conditioned linear system , unless @xmath10 has the jordan form corresponding to the largest eigenvalue , which contradicts the perron
frobenius theorem . from ( [ eq : newtonni ] ) , @xmath260hence
, we have the following linear system@xmath261or@xmath262 \mathbf{y}_{k+1}=\mathbf{x}_{k},\]]with @xmath263 thus , from ( [ eq : step1 ] ) and ( [ eq : newtonni ] ) , we have the new iterative vector @xmath264this means the noda iteration and the modified noda iteration are mathematically equivalent . based on ( [ eq : newtonni ] ) and ( [ eq : updatexk ] ) , we state our algorithm as follows . 1 .
given @xmath92 , @xmath93 with @xmath94 and @xmath96 .
2 . * for * @xmath97 3 . @xmath265 4 .
solve @xmath266 . 5 .
normalize the vector @xmath100 .
6 . 7 .
solve @xmath267.$ ] 8 .
normalize the vector @xmath268 .
9 . 10 .
. 11 . * until * convergence : @xmath270 .
note that the sequence @xmath271 tends to a singular matrix , meaning that @xmath78 converges to an eigenvalue of @xmath10 , and ( [ eq : step1 ] ) becomes an ill - conditioned linear system .
based on practical experiments , we propose taking @xmath272 and switching from ( [ eq : step1 ] ) to ( [ eq : newtonni ] ) . for a monotone matrix @xmath4 , we replaced @xmath10 by @xmath273 in ( [ eq : newtonni ] ) . the linear system ( [ eq : newtonni ]
) can be rewritten as @xmath274 .
\label{eq : newtonni2}\ ] ] based on mni , by combining ini ( algorithm [ alg : ini ] ) with equation ( [ eq : newtonni2 ] ) , we can propose a modified inexact noda iteration for a monotone matrix , which is described as algorithm [ alg : mini_m ] . 1
. given @xmath92 , @xmath93 with @xmath94 and @xmath96 .
2 . * for * @xmath97 3 . @xmath275 4 .
run ini for monotone matrix @xmath4 ( algorithm [ alg : ini ] ) .
solve @xmath276 $ ] exactly .
normalize the vector @xmath268 .
compute @xmath111 that satisfies condition ( [ eq : updateeig ] ) .
9 . 10 . * until * convergence : @xmath112 .
in this section we present numerical experiments to support our theoretical results for ini , and to illustrate the effectiveness of the proposed mini algorithms .
all numerical tests were performed on an intel ( r ) core ( tm ) i@xmath277 cpu @xmath278@ @xmath279ghz with @xmath280 gb memory using matlab r@xmath281 with the machine precision @xmath282 under the microsoft windows @xmath277 @xmath283-bit .
@xmath284 denotes the number of outer iterations to achieve the convergence , and @xmath285 denotes the total number of inner iterations , which measures the overall efficiency of mni and mini . in view of the above , we have the average number @xmath286 at each outer iteration for our test algorithms . in the tables , positivityillustrates whether the converged perron vector preserves the strict positivity property .
if no , then the percentage in the brace indicates the proportion that the converged perron vector has the positive components .
we also report the cpu time of each algorithm , which measures the overall efficiency too .
we present an example to illustrate the numerical behavior of ni , ini_1 and ini_2 for monotone matrices . the approximate solution @xmath80 of ( [ eq : inexactsys2 ] ) satisfies@xmath287 by requiring the following inner tolerances : * for ni : @xmath288 ; * for ini_1 : @xmath289 with some @xmath290 ; * for ini_2 : @xmath291 for @xmath292 and @xmath293 .
we use the minimal residual method to solve the inner linear systems . for the implementations ,
we use the standard matlab function .
the outer iteration starts with the normalized vector of @xmath294 , and the stopping criterion for outer iterations is @xmath295 where @xmath296 and @xmath297 are the one norm and the infinity norm of a matrix , respectively .
condition ( [ eq : strategy ] ) ensures that the eigenvector in lemma [ monotone2 ] does indeed preserve the strict positivity property .
the formula in ( [ eq : strategy ] ) is not applicable in practice , because it uses @xmath298 , which is unknown at the time it needs to be computed .
therefore , for practical implementations , we suggest a relaxation strategy to replace @xmath298 by @xmath299 .
the quantity @xmath299 is related to the lower bound of the smallest eigenvalue of @xmath4 , i.e. , @xmath300 . for all examples , the stopping criterion for the inner iteration is set at @xmath301and@xmath302 [ exp : fem]we consider the finite - element discretization of the boundary value problem in ( * ? ? ?
* example 4.2.4 ) @xmath303\times \lbrack 0,b ] , \\ a , b & > & 0,\text { \ \ \ \ \ \ \ \ } u = f(x , y)\text { on } \partial \omega , \end{aligned}\]]using piecewise quadratic basis functions on the uniform mesh of @xmath304 isosceles right - angled triangles .
this is a matrix of order @xmath305 with @xmath306 and @xmath307 .
for example 1 , we see that , for this monotone matrix eigenproblem , ini_1 , with two different @xmath308 and @xmath309 exhibits distinct convergence behaviors and uses @xmath310 and @xmath311 outer iterations to achieve the desired accuracy , respectively . as figure [ fig : fem ] indicates , ni and ini_2 typically converge superlinearly , and ini_1 with @xmath312 typically converge linearly .
this confirms our theory and demonstrates that the results of our theorem can be realistic and pronounced .
.the total outer and inner iterations in example exp : fem [ cols="<,>,>,>,>,<",options="header " , ]
we have proposed an inexact noda iteration method for computing the smallest eigenpair of a large irreducible monotone matrix , and have considered the convergence of the modified inexact noda iteration with two relaxation factors .
we have proved that the convergence of ini is globally linear and superlinear , with the asymptotic convergence factor bounded by @xmath313 .
more precisely , the modified inexact noda iteration with inner tolerance @xmath314 converges at least linearly if the relaxation factors meet the condition @xmath315 , and superlinearly if the relaxation factors meet the condition @xmath228 , respectively . the results for ini clearly show how the accuracy of the inner iterations affects the convergence of the outer iterations .
in the experiments , we also compared mini with jacobi davidson type methods ( jdqr , jdsvd ) and the implicitly restarted arnoldi method ( ) .
the contribution of this paper is twofold .
first , mini always preserves the positivity of approximate eigenvectors , while the other three methods often fail to do so .
second , the proposed mini algorithms have been shown to be practical and effective for large monotone matrix eigenvalue problems and @xmath8-matrix singular value problems . height 2pt depth -1.6pt width 23pt , _ on convergence of the inexact rayleigh quotient iteration with the lanczos method used for solving linear systems _ , technical report , september 2011 . ` http://arxiv.org/pdf/0906.2239v4.pdf ` .
a. m. ostrowski , _ on the convergence of the rayleigh quotient iteration for the computation of the characteristic roots and vectors .
v. ( usual rayleigh quotient for non - hermitian matrices and linear elementary divisors ) _ , arch .
rational mech .
, 3 ( 1959 ) , pp .
472481 .
p. shivakumar , j. williams , q. ye and c. marinov , _ on two - sided bounds related to weakly diagonally dominant m - matrices with applications to digital circuit dynamics _ , siam j. matrix anal . appl . , 17 , pp . 298312 ( 1996 )
|
in this paper , we present an inexact noda iteration with inner - outer iterations for finding the smallest eigenvalue and the associated eigenvector of an irreducible monotone matrix .
the proposed inexact noda iteration contains two main relaxation steps for computing the smallest eigenvalue and the associated eigenvector , respectively .
these relaxation steps depend on the relaxation factors , and we analyze how the relaxation factors in the relaxation steps affect the convergence of the outer iterations . by considering two different relaxation factors for solving the inner linear systems involved , we prove that the convergence is globally linear or superlinear , depending on the relaxation factor , and that the relaxation factor also influences the convergence rate .
the proposed inexact noda iterations are structure preserving and maintain the positivity of approximate eigenvectors .
numerical examples are provided to illustrate that the proposed inexact noda iterations are practical , and they always preserve the positivity of approximate eigenvectors .
|
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.